From 39279cc3d2704cfbf9c35dcb5bdd392159ae4625 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 12 Jun 2007 06:35:45 -0400 Subject: Btrfs: split up super.c Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2210 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 2210 insertions(+) create mode 100644 fs/btrfs/inode.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c new file mode 100644 index 00000000000..62a3a778d37 --- /dev/null +++ b/fs/btrfs/inode.c @@ -0,0 +1,2210 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "ctree.h" +#include "disk-io.h" +#include "transaction.h" +#include "btrfs_inode.h" +#include "ioctl.h" +#include "print-tree.h" + +struct btrfs_iget_args { + u64 ino; + struct btrfs_root *root; +}; + +static struct inode_operations btrfs_dir_inode_operations; +static struct inode_operations btrfs_symlink_inode_operations; +static struct inode_operations btrfs_dir_ro_inode_operations; +static struct inode_operations btrfs_file_inode_operations; +static struct address_space_operations btrfs_aops; +static struct address_space_operations btrfs_symlink_aops; +static struct file_operations btrfs_dir_file_operations; + +static struct kmem_cache *btrfs_inode_cachep; +struct kmem_cache *btrfs_trans_handle_cachep; +struct kmem_cache *btrfs_transaction_cachep; +struct kmem_cache *btrfs_bit_radix_cachep; +struct kmem_cache *btrfs_path_cachep; + +#define S_SHIFT 12 +static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { + [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE, + [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR, + [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV, + [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV, + [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO, + [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK, + [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, +}; + +void btrfs_read_locked_inode(struct inode *inode) +{ + struct btrfs_path *path; + struct btrfs_inode_item *inode_item; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_key location; + u64 alloc_group_block; + int ret; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + mutex_lock(&root->fs_info->fs_mutex); + + memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); + ret = btrfs_lookup_inode(NULL, root, path, &location, 0); + if (ret) { + btrfs_free_path(path); + goto make_bad; + } + inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), + path->slots[0], + struct btrfs_inode_item); + + inode->i_mode = btrfs_inode_mode(inode_item); + inode->i_nlink = btrfs_inode_nlink(inode_item); + inode->i_uid = btrfs_inode_uid(inode_item); + inode->i_gid = btrfs_inode_gid(inode_item); + inode->i_size = btrfs_inode_size(inode_item); + inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime); + inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime); + inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime); + inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime); + inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime); + inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime); + inode->i_blocks = btrfs_inode_nblocks(inode_item); + inode->i_generation = btrfs_inode_generation(inode_item); + alloc_group_block = btrfs_inode_block_group(inode_item); + BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, + alloc_group_block); + + btrfs_free_path(path); + inode_item = NULL; + + mutex_unlock(&root->fs_info->fs_mutex); + + switch (inode->i_mode & S_IFMT) { +#if 0 + default: + init_special_inode(inode, inode->i_mode, + btrfs_inode_rdev(inode_item)); + break; +#endif + case S_IFREG: + inode->i_mapping->a_ops = &btrfs_aops; + inode->i_fop = &btrfs_file_operations; + inode->i_op = &btrfs_file_inode_operations; + break; + case S_IFDIR: + inode->i_fop = &btrfs_dir_file_operations; + if (root == root->fs_info->tree_root) + inode->i_op = &btrfs_dir_ro_inode_operations; + else + inode->i_op = &btrfs_dir_inode_operations; + break; + case S_IFLNK: + inode->i_op = &btrfs_symlink_inode_operations; + inode->i_mapping->a_ops = &btrfs_symlink_aops; + break; + } + return; + +make_bad: + btrfs_release_path(root, path); + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + make_bad_inode(inode); +} + +static void fill_inode_item(struct btrfs_inode_item *item, + struct inode *inode) +{ + btrfs_set_inode_uid(item, inode->i_uid); + btrfs_set_inode_gid(item, inode->i_gid); + btrfs_set_inode_size(item, inode->i_size); + btrfs_set_inode_mode(item, inode->i_mode); + btrfs_set_inode_nlink(item, inode->i_nlink); + btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec); + btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec); + btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec); + btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec); + btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec); + btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec); + btrfs_set_inode_nblocks(item, inode->i_blocks); + btrfs_set_inode_generation(item, inode->i_generation); + btrfs_set_inode_block_group(item, + BTRFS_I(inode)->block_group->key.objectid); +} + +static int btrfs_update_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode) +{ + struct btrfs_inode_item *inode_item; + struct btrfs_path *path; + int ret; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + ret = btrfs_lookup_inode(trans, root, path, + &BTRFS_I(inode)->location, 1); + if (ret) { + if (ret > 0) + ret = -ENOENT; + goto failed; + } + + inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), + path->slots[0], + struct btrfs_inode_item); + + fill_inode_item(inode_item, inode); + btrfs_mark_buffer_dirty(path->nodes[0]); + ret = 0; +failed: + btrfs_release_path(root, path); + btrfs_free_path(path); + return ret; +} + + +static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *dir, + struct dentry *dentry) +{ + struct btrfs_path *path; + const char *name = dentry->d_name.name; + int name_len = dentry->d_name.len; + int ret = 0; + u64 objectid; + struct btrfs_dir_item *di; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino, + name, name_len, -1); + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto err; + } + if (!di) { + ret = -ENOENT; + goto err; + } + objectid = btrfs_disk_key_objectid(&di->location); + ret = btrfs_delete_one_dir_name(trans, root, path, di); + BUG_ON(ret); + btrfs_release_path(root, path); + + di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, + objectid, name, name_len, -1); + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto err; + } + if (!di) { + ret = -ENOENT; + goto err; + } + ret = btrfs_delete_one_dir_name(trans, root, path, di); + BUG_ON(ret); + + dentry->d_inode->i_ctime = dir->i_ctime; +err: + btrfs_free_path(path); + if (!ret) { + dir->i_size -= name_len * 2; + btrfs_update_inode(trans, root, dir); + drop_nlink(dentry->d_inode); + btrfs_update_inode(trans, root, dentry->d_inode); + dir->i_sb->s_dirt = 1; + } + return ret; +} + +static int btrfs_unlink(struct inode *dir, struct dentry *dentry) +{ + struct btrfs_root *root; + struct btrfs_trans_handle *trans; + int ret; + + root = BTRFS_I(dir)->root; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + ret = btrfs_unlink_trans(trans, root, dir, dentry); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); + return ret; +} + +static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) +{ + struct inode *inode = dentry->d_inode; + int err; + int ret; + struct btrfs_root *root = BTRFS_I(dir)->root; + struct btrfs_path *path; + struct btrfs_key key; + struct btrfs_trans_handle *trans; + struct btrfs_key found_key; + int found_type; + struct btrfs_leaf *leaf; + char *goodnames = ".."; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + key.objectid = inode->i_ino; + key.offset = (u64)-1; + key.flags = (u32)-1; + while(1) { + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) { + err = ret; + goto out; + } + BUG_ON(ret == 0); + if (path->slots[0] == 0) { + err = -ENOENT; + goto out; + } + path->slots[0]--; + leaf = btrfs_buffer_leaf(path->nodes[0]); + btrfs_disk_key_to_cpu(&found_key, + &leaf->items[path->slots[0]].key); + found_type = btrfs_key_type(&found_key); + if (found_key.objectid != inode->i_ino) { + err = -ENOENT; + goto out; + } + if ((found_type != BTRFS_DIR_ITEM_KEY && + found_type != BTRFS_DIR_INDEX_KEY) || + (!btrfs_match_dir_item_name(root, path, goodnames, 2) && + !btrfs_match_dir_item_name(root, path, goodnames, 1))) { + err = -ENOTEMPTY; + goto out; + } + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + + if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1) + break; + btrfs_release_path(root, path); + } + ret = 0; + btrfs_release_path(root, path); + + /* now the directory is empty */ + err = btrfs_unlink_trans(trans, root, dir, dentry); + if (!err) { + inode->i_size = 0; + } +out: + btrfs_release_path(root, path); + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + ret = btrfs_end_transaction(trans, root); + btrfs_btree_balance_dirty(root); + if (ret && !err) + err = ret; + return err; +} + +static int btrfs_free_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode) +{ + struct btrfs_path *path; + int ret; + + clear_inode(inode); + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + ret = btrfs_lookup_inode(trans, root, path, + &BTRFS_I(inode)->location, -1); + BUG_ON(ret); + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + btrfs_free_path(path); + return ret; +} + +/* + * truncates go from a high offset to a low offset. So, walk + * from hi to lo in the node and issue readas. Stop when you find + * keys from a different objectid + */ +static void reada_truncate(struct btrfs_root *root, struct btrfs_path *path, + u64 objectid) +{ + struct btrfs_node *node; + int i; + int nritems; + u64 item_objectid; + u64 blocknr; + int slot; + int ret; + + if (!path->nodes[1]) + return; + node = btrfs_buffer_node(path->nodes[1]); + slot = path->slots[1]; + if (slot == 0) + return; + nritems = btrfs_header_nritems(&node->header); + for (i = slot - 1; i >= 0; i--) { + item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); + if (item_objectid != objectid) + break; + blocknr = btrfs_node_blockptr(node, i); + ret = readahead_tree_block(root, blocknr); + if (ret) + break; + } +} + +/* + * this can truncate away extent items, csum items and directory items. + * It starts at a high offset and removes keys until it can't find + * any higher than i_size. + * + * csum items that cross the new i_size are truncated to the new size + * as well. + */ +static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode) +{ + int ret; + struct btrfs_path *path; + struct btrfs_key key; + struct btrfs_disk_key *found_key; + u32 found_type; + struct btrfs_leaf *leaf; + struct btrfs_file_extent_item *fi; + u64 extent_start = 0; + u64 extent_num_blocks = 0; + u64 item_end = 0; + int found_extent; + int del_item; + + path = btrfs_alloc_path(); + BUG_ON(!path); + /* FIXME, add redo link to tree so we don't leak on crash */ + key.objectid = inode->i_ino; + key.offset = (u64)-1; + key.flags = (u32)-1; + while(1) { + btrfs_init_path(path); + fi = NULL; + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) { + goto error; + } + if (ret > 0) { + BUG_ON(path->slots[0] == 0); + path->slots[0]--; + } + reada_truncate(root, path, inode->i_ino); + leaf = btrfs_buffer_leaf(path->nodes[0]); + found_key = &leaf->items[path->slots[0]].key; + found_type = btrfs_disk_key_type(found_key); + + if (btrfs_disk_key_objectid(found_key) != inode->i_ino) + break; + if (found_type != BTRFS_CSUM_ITEM_KEY && + found_type != BTRFS_DIR_ITEM_KEY && + found_type != BTRFS_DIR_INDEX_KEY && + found_type != BTRFS_EXTENT_DATA_KEY) + break; + + item_end = btrfs_disk_key_offset(found_key); + if (found_type == BTRFS_EXTENT_DATA_KEY) { + fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), + path->slots[0], + struct btrfs_file_extent_item); + if (btrfs_file_extent_type(fi) != + BTRFS_FILE_EXTENT_INLINE) { + item_end += btrfs_file_extent_num_blocks(fi) << + inode->i_blkbits; + } + } + if (found_type == BTRFS_CSUM_ITEM_KEY) { + ret = btrfs_csum_truncate(trans, root, path, + inode->i_size); + BUG_ON(ret); + } + if (item_end < inode->i_size) { + if (found_type) { + btrfs_set_key_type(&key, found_type - 1); + continue; + } + break; + } + if (btrfs_disk_key_offset(found_key) >= inode->i_size) + del_item = 1; + else + del_item = 0; + found_extent = 0; + + /* FIXME, shrink the extent if the ref count is only 1 */ + if (found_type == BTRFS_EXTENT_DATA_KEY && + btrfs_file_extent_type(fi) != + BTRFS_FILE_EXTENT_INLINE) { + u64 num_dec; + if (!del_item) { + u64 orig_num_blocks = + btrfs_file_extent_num_blocks(fi); + extent_num_blocks = inode->i_size - + btrfs_disk_key_offset(found_key) + + root->blocksize - 1; + extent_num_blocks >>= inode->i_blkbits; + btrfs_set_file_extent_num_blocks(fi, + extent_num_blocks); + inode->i_blocks -= (orig_num_blocks - + extent_num_blocks) << 3; + mark_buffer_dirty(path->nodes[0]); + } else { + extent_start = + btrfs_file_extent_disk_blocknr(fi); + extent_num_blocks = + btrfs_file_extent_disk_num_blocks(fi); + /* FIXME blocksize != 4096 */ + num_dec = btrfs_file_extent_num_blocks(fi) << 3; + if (extent_start != 0) { + found_extent = 1; + inode->i_blocks -= num_dec; + } + } + } + if (del_item) { + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + } else { + break; + } + btrfs_release_path(root, path); + if (found_extent) { + ret = btrfs_free_extent(trans, root, extent_start, + extent_num_blocks, 0); + BUG_ON(ret); + } + } + ret = 0; +error: + btrfs_release_path(root, path); + btrfs_free_path(path); + inode->i_sb->s_dirt = 1; + return ret; +} + +/* + * taken from block_truncate_page, but does cow as it zeros out + * any bytes left in the last page in the file. + */ +static int btrfs_truncate_page(struct address_space *mapping, loff_t from) +{ + struct inode *inode = mapping->host; + unsigned blocksize = 1 << inode->i_blkbits; + pgoff_t index = from >> PAGE_CACHE_SHIFT; + unsigned offset = from & (PAGE_CACHE_SIZE-1); + struct page *page; + char *kaddr; + int ret = 0; + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 alloc_hint = 0; + struct btrfs_key ins; + struct btrfs_trans_handle *trans; + + if ((offset & (blocksize - 1)) == 0) + goto out; + + ret = -ENOMEM; + page = grab_cache_page(mapping, index); + if (!page) + goto out; + + if (!PageUptodate(page)) { + ret = mpage_readpage(page, btrfs_get_block); + lock_page(page); + if (!PageUptodate(page)) { + ret = -EIO; + goto out; + } + } + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + + ret = btrfs_drop_extents(trans, root, inode, + page->index << PAGE_CACHE_SHIFT, + (page->index + 1) << PAGE_CACHE_SHIFT, + &alloc_hint); + BUG_ON(ret); + ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, + alloc_hint, (u64)-1, &ins, 1); + BUG_ON(ret); + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + page->index << PAGE_CACHE_SHIFT, + ins.objectid, 1, 1); + BUG_ON(ret); + SetPageChecked(page); + kaddr = kmap(page); + memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + btrfs_csum_file_block(trans, root, inode->i_ino, + page->index << PAGE_CACHE_SHIFT, + kaddr, PAGE_CACHE_SIZE); + kunmap(page); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + set_page_dirty(page); + unlock_page(page); + page_cache_release(page); +out: + return ret; +} + +static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) +{ + struct inode *inode = dentry->d_inode; + int err; + + err = inode_change_ok(inode, attr); + if (err) + return err; + + if (S_ISREG(inode->i_mode) && + attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 mask = root->blocksize - 1; + u64 pos = (inode->i_size + mask) & ~mask; + u64 hole_size; + + if (attr->ia_size <= pos) + goto out; + + btrfs_truncate_page(inode->i_mapping, inode->i_size); + + hole_size = (attr->ia_size - pos + mask) & ~mask; + hole_size >>= inode->i_blkbits; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + err = btrfs_insert_file_extent(trans, root, inode->i_ino, + pos, 0, 0, hole_size); + BUG_ON(err); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + } +out: + err = inode_setattr(inode, attr); + + return err; +} +void btrfs_delete_inode(struct inode *inode) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret; + + truncate_inode_pages(&inode->i_data, 0); + if (is_bad_inode(inode)) { + goto no_delete; + } + inode->i_size = 0; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + ret = btrfs_truncate_in_trans(trans, root, inode); + BUG_ON(ret); + btrfs_free_inode(trans, root, inode); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); + return; +no_delete: + clear_inode(inode); +} + +/* + * this returns the key found in the dir entry in the location pointer. + * If no dir entries were found, location->objectid is 0. + */ +static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, + struct btrfs_key *location) +{ + const char *name = dentry->d_name.name; + int namelen = dentry->d_name.len; + struct btrfs_dir_item *di; + struct btrfs_path *path; + struct btrfs_root *root = BTRFS_I(dir)->root; + int ret; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, + namelen, 0); + if (!di || IS_ERR(di)) { + location->objectid = 0; + ret = 0; + goto out; + } + btrfs_disk_key_to_cpu(location, &di->location); +out: + btrfs_release_path(root, path); + btrfs_free_path(path); + return ret; +} + +/* + * when we hit a tree root in a directory, the btrfs part of the inode + * needs to be changed to reflect the root directory of the tree root. This + * is kind of like crossing a mount point. + */ +static int fixup_tree_root_location(struct btrfs_root *root, + struct btrfs_key *location, + struct btrfs_root **sub_root) +{ + struct btrfs_path *path; + struct btrfs_root_item *ri; + + if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY) + return 0; + if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) + return 0; + + path = btrfs_alloc_path(); + BUG_ON(!path); + mutex_lock(&root->fs_info->fs_mutex); + + *sub_root = btrfs_read_fs_root(root->fs_info, location); + if (IS_ERR(*sub_root)) + return PTR_ERR(*sub_root); + + ri = &(*sub_root)->root_item; + location->objectid = btrfs_root_dirid(ri); + location->flags = 0; + btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); + location->offset = 0; + + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + return 0; +} + +static int btrfs_init_locked_inode(struct inode *inode, void *p) +{ + struct btrfs_iget_args *args = p; + inode->i_ino = args->ino; + BTRFS_I(inode)->root = args->root; + return 0; +} + +static int btrfs_find_actor(struct inode *inode, void *opaque) +{ + struct btrfs_iget_args *args = opaque; + return (args->ino == inode->i_ino && + args->root == BTRFS_I(inode)->root); +} + +struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, + struct btrfs_root *root) +{ + struct inode *inode; + struct btrfs_iget_args args; + args.ino = objectid; + args.root = root; + + inode = iget5_locked(s, objectid, btrfs_find_actor, + btrfs_init_locked_inode, + (void *)&args); + return inode; +} + +static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, + struct nameidata *nd) +{ + struct inode * inode; + struct btrfs_inode *bi = BTRFS_I(dir); + struct btrfs_root *root = bi->root; + struct btrfs_root *sub_root = root; + struct btrfs_key location; + int ret; + + if (dentry->d_name.len > BTRFS_NAME_LEN) + return ERR_PTR(-ENAMETOOLONG); + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_inode_by_name(dir, dentry, &location); + mutex_unlock(&root->fs_info->fs_mutex); + if (ret < 0) + return ERR_PTR(ret); + inode = NULL; + if (location.objectid) { + ret = fixup_tree_root_location(root, &location, &sub_root); + if (ret < 0) + return ERR_PTR(ret); + if (ret > 0) + return ERR_PTR(-ENOENT); + inode = btrfs_iget_locked(dir->i_sb, location.objectid, + sub_root); + if (!inode) + return ERR_PTR(-EACCES); + if (inode->i_state & I_NEW) { + /* the inode and parent dir are two different roots */ + if (sub_root != root) { + igrab(inode); + sub_root->inode = inode; + } + BTRFS_I(inode)->root = sub_root; + memcpy(&BTRFS_I(inode)->location, &location, + sizeof(location)); + btrfs_read_locked_inode(inode); + unlock_new_inode(inode); + } + } + return d_splice_alias(inode, dentry); +} + +/* + * readahead one full node of leaves as long as their keys include + * the objectid supplied + */ +static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path, + u64 objectid) +{ + struct btrfs_node *node; + int i; + u32 nritems; + u64 item_objectid; + u64 blocknr; + int slot; + int ret; + + if (!path->nodes[1]) + return; + node = btrfs_buffer_node(path->nodes[1]); + slot = path->slots[1]; + nritems = btrfs_header_nritems(&node->header); + for (i = slot + 1; i < nritems; i++) { + item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); + if (item_objectid != objectid) + break; + blocknr = btrfs_node_blockptr(node, i); + ret = readahead_tree_block(root, blocknr); + if (ret) + break; + } +} +static unsigned char btrfs_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK +}; + +static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) +{ + struct inode *inode = filp->f_path.dentry->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_item *item; + struct btrfs_dir_item *di; + struct btrfs_key key; + struct btrfs_path *path; + int ret; + u32 nritems; + struct btrfs_leaf *leaf; + int slot; + int advance; + unsigned char d_type; + int over = 0; + u32 di_cur; + u32 di_total; + u32 di_len; + int key_type = BTRFS_DIR_INDEX_KEY; + + /* FIXME, use a real flag for deciding about the key type */ + if (root->fs_info->tree_root == root) + key_type = BTRFS_DIR_ITEM_KEY; + mutex_lock(&root->fs_info->fs_mutex); + key.objectid = inode->i_ino; + key.flags = 0; + btrfs_set_key_type(&key, key_type); + key.offset = filp->f_pos; + path = btrfs_alloc_path(); + btrfs_init_path(path); + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto err; + advance = 0; + reada_leaves(root, path, inode->i_ino); + while(1) { + leaf = btrfs_buffer_leaf(path->nodes[0]); + nritems = btrfs_header_nritems(&leaf->header); + slot = path->slots[0]; + if (advance || slot >= nritems) { + if (slot >= nritems -1) { + reada_leaves(root, path, inode->i_ino); + ret = btrfs_next_leaf(root, path); + if (ret) + break; + leaf = btrfs_buffer_leaf(path->nodes[0]); + nritems = btrfs_header_nritems(&leaf->header); + slot = path->slots[0]; + } else { + slot++; + path->slots[0]++; + } + } + advance = 1; + item = leaf->items + slot; + if (btrfs_disk_key_objectid(&item->key) != key.objectid) + break; + if (btrfs_disk_key_type(&item->key) != key_type) + break; + if (btrfs_disk_key_offset(&item->key) < filp->f_pos) + continue; + filp->f_pos = btrfs_disk_key_offset(&item->key); + advance = 1; + di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); + di_cur = 0; + di_total = btrfs_item_size(leaf->items + slot); + while(di_cur < di_total) { + d_type = btrfs_filetype_table[btrfs_dir_type(di)]; + over = filldir(dirent, (const char *)(di + 1), + btrfs_dir_name_len(di), + btrfs_disk_key_offset(&item->key), + btrfs_disk_key_objectid(&di->location), + d_type); + if (over) + goto nopos; + di_len = btrfs_dir_name_len(di) + sizeof(*di); + di_cur += di_len; + di = (struct btrfs_dir_item *)((char *)di + di_len); + } + } + filp->f_pos++; +nopos: + ret = 0; +err: + btrfs_release_path(root, path); + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +int btrfs_write_inode(struct inode *inode, int wait) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + if (wait) { + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + ret = btrfs_commit_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + } + return ret; +} + +/* + * This is somewhat expense, updating the tree every time the + * inode changes. But, it is most likely to find the inode in cache. + * FIXME, needs more benchmarking...there are no reasons other than performance + * to keep or drop this code. + */ +void btrfs_dirty_inode(struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + btrfs_update_inode(trans, root, inode); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); +} + +static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 objectid, + struct btrfs_block_group_cache *group, + int mode) +{ + struct inode *inode; + struct btrfs_inode_item inode_item; + struct btrfs_key *location; + int ret; + int owner; + + inode = new_inode(root->fs_info->sb); + if (!inode) + return ERR_PTR(-ENOMEM); + + BTRFS_I(inode)->root = root; + if (mode & S_IFDIR) + owner = 0; + else + owner = 1; + group = btrfs_find_block_group(root, group, 0, 0, owner); + BTRFS_I(inode)->block_group = group; + + inode->i_uid = current->fsuid; + inode->i_gid = current->fsgid; + inode->i_mode = mode; + inode->i_ino = objectid; + inode->i_blocks = 0; + inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; + fill_inode_item(&inode_item, inode); + location = &BTRFS_I(inode)->location; + location->objectid = objectid; + location->flags = 0; + location->offset = 0; + btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); + + ret = btrfs_insert_inode(trans, root, objectid, &inode_item); + BUG_ON(ret); + + insert_inode_hash(inode); + return inode; +} + +static inline u8 btrfs_inode_type(struct inode *inode) +{ + return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT]; +} + +static int btrfs_add_link(struct btrfs_trans_handle *trans, + struct dentry *dentry, struct inode *inode) +{ + int ret; + struct btrfs_key key; + struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root; + key.objectid = inode->i_ino; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); + key.offset = 0; + + ret = btrfs_insert_dir_item(trans, root, + dentry->d_name.name, dentry->d_name.len, + dentry->d_parent->d_inode->i_ino, + &key, btrfs_inode_type(inode)); + if (ret == 0) { + dentry->d_parent->d_inode->i_size += dentry->d_name.len * 2; + ret = btrfs_update_inode(trans, root, + dentry->d_parent->d_inode); + } + return ret; +} + +static int btrfs_add_nondir(struct btrfs_trans_handle *trans, + struct dentry *dentry, struct inode *inode) +{ + int err = btrfs_add_link(trans, dentry, inode); + if (!err) { + d_instantiate(dentry, inode); + return 0; + } + if (err > 0) + err = -EEXIST; + return err; +} + +static int btrfs_create(struct inode *dir, struct dentry *dentry, + int mode, struct nameidata *nd) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(dir)->root; + struct inode *inode; + int err; + int drop_inode = 0; + u64 objectid; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + + err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); + if (err) { + err = -ENOSPC; + goto out_unlock; + } + + inode = btrfs_new_inode(trans, root, objectid, + BTRFS_I(dir)->block_group, mode); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_unlock; + + btrfs_set_trans_block_group(trans, inode); + err = btrfs_add_nondir(trans, dentry, inode); + if (err) + drop_inode = 1; + else { + inode->i_mapping->a_ops = &btrfs_aops; + inode->i_fop = &btrfs_file_operations; + inode->i_op = &btrfs_file_inode_operations; + } + dir->i_sb->s_dirt = 1; + btrfs_update_inode_block_group(trans, inode); + btrfs_update_inode_block_group(trans, dir); +out_unlock: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + if (drop_inode) { + inode_dec_link_count(inode); + iput(inode); + } + btrfs_btree_balance_dirty(root); + return err; +} + +static int btrfs_link(struct dentry *old_dentry, struct inode *dir, + struct dentry *dentry) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(dir)->root; + struct inode *inode = old_dentry->d_inode; + int err; + int drop_inode = 0; + + if (inode->i_nlink == 0) + return -ENOENT; + + inc_nlink(inode); + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + atomic_inc(&inode->i_count); + err = btrfs_add_nondir(trans, dentry, inode); + if (err) + drop_inode = 1; + dir->i_sb->s_dirt = 1; + btrfs_update_inode_block_group(trans, dir); + btrfs_update_inode(trans, root, inode); + + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + if (drop_inode) { + inode_dec_link_count(inode); + iput(inode); + } + btrfs_btree_balance_dirty(root); + return err; +} + +static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 objectid, u64 dirid) +{ + int ret; + char buf[2]; + struct btrfs_key key; + + buf[0] = '.'; + buf[1] = '.'; + + key.objectid = objectid; + key.offset = 0; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); + + ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid, + &key, BTRFS_FT_DIR); + if (ret) + goto error; + key.objectid = dirid; + ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid, + &key, BTRFS_FT_DIR); + if (ret) + goto error; +error: + return ret; +} + +static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) +{ + struct inode *inode; + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(dir)->root; + int err = 0; + int drop_on_err = 0; + u64 objectid; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + goto out_unlock; + } + + err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); + if (err) { + err = -ENOSPC; + goto out_unlock; + } + + inode = btrfs_new_inode(trans, root, objectid, + BTRFS_I(dir)->block_group, S_IFDIR | mode); + if (IS_ERR(inode)) { + err = PTR_ERR(inode); + goto out_fail; + } + drop_on_err = 1; + inode->i_op = &btrfs_dir_inode_operations; + inode->i_fop = &btrfs_dir_file_operations; + btrfs_set_trans_block_group(trans, inode); + + err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino); + if (err) + goto out_fail; + + inode->i_size = 6; + err = btrfs_update_inode(trans, root, inode); + if (err) + goto out_fail; + err = btrfs_add_link(trans, dentry, inode); + if (err) + goto out_fail; + d_instantiate(dentry, inode); + drop_on_err = 0; + dir->i_sb->s_dirt = 1; + btrfs_update_inode_block_group(trans, inode); + btrfs_update_inode_block_group(trans, dir); + +out_fail: + btrfs_end_transaction(trans, root); +out_unlock: + mutex_unlock(&root->fs_info->fs_mutex); + if (drop_on_err) + iput(inode); + btrfs_btree_balance_dirty(root); + return err; +} + +/* + * FIBMAP and others want to pass in a fake buffer head. They need to + * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy + * any packed file data into the fake bh + */ +#define BTRFS_GET_BLOCK_NO_CREATE 0 +#define BTRFS_GET_BLOCK_CREATE 1 +#define BTRFS_GET_BLOCK_NO_DIRECT 2 + +/* + * FIXME create==1 doe not work. + */ +static int btrfs_get_block_lock(struct inode *inode, sector_t iblock, + struct buffer_head *result, int create) +{ + int ret; + int err = 0; + u64 blocknr; + u64 extent_start = 0; + u64 extent_end = 0; + u64 objectid = inode->i_ino; + u32 found_type; + u64 alloc_hint = 0; + struct btrfs_path *path; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_file_extent_item *item; + struct btrfs_leaf *leaf; + struct btrfs_disk_key *found_key; + struct btrfs_trans_handle *trans = NULL; + + path = btrfs_alloc_path(); + BUG_ON(!path); + btrfs_init_path(path); + if (create & BTRFS_GET_BLOCK_CREATE) { + WARN_ON(1); + /* this almost but not quite works */ + trans = btrfs_start_transaction(root, 1); + if (!trans) { + err = -ENOMEM; + goto out; + } + ret = btrfs_drop_extents(trans, root, inode, + iblock << inode->i_blkbits, + (iblock + 1) << inode->i_blkbits, + &alloc_hint); + BUG_ON(ret); + } + + ret = btrfs_lookup_file_extent(NULL, root, path, + inode->i_ino, + iblock << inode->i_blkbits, 0); + if (ret < 0) { + err = ret; + goto out; + } + + if (ret != 0) { + if (path->slots[0] == 0) { + btrfs_release_path(root, path); + goto not_found; + } + path->slots[0]--; + } + + item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], + struct btrfs_file_extent_item); + leaf = btrfs_buffer_leaf(path->nodes[0]); + blocknr = btrfs_file_extent_disk_blocknr(item); + blocknr += btrfs_file_extent_offset(item); + + /* are we inside the extent that was found? */ + found_key = &leaf->items[path->slots[0]].key; + found_type = btrfs_disk_key_type(found_key); + if (btrfs_disk_key_objectid(found_key) != objectid || + found_type != BTRFS_EXTENT_DATA_KEY) { + extent_end = 0; + extent_start = 0; + goto not_found; + } + found_type = btrfs_file_extent_type(item); + extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key); + if (found_type == BTRFS_FILE_EXTENT_REG) { + extent_start = extent_start >> inode->i_blkbits; + extent_end = extent_start + btrfs_file_extent_num_blocks(item); + err = 0; + if (btrfs_file_extent_disk_blocknr(item) == 0) + goto out; + if (iblock >= extent_start && iblock < extent_end) { + btrfs_map_bh_to_logical(root, result, blocknr + + iblock - extent_start); + goto out; + } + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + char *ptr; + char *map; + u32 size; + + if (create & BTRFS_GET_BLOCK_NO_DIRECT) { + err = -EINVAL; + goto out; + } + size = btrfs_file_extent_inline_len(leaf->items + + path->slots[0]); + extent_end = (extent_start + size) >> inode->i_blkbits; + extent_start >>= inode->i_blkbits; + if (iblock < extent_start || iblock > extent_end) { + goto not_found; + } + ptr = btrfs_file_extent_inline_start(item); + map = kmap(result->b_page); + memcpy(map, ptr, size); + memset(map + size, 0, PAGE_CACHE_SIZE - size); + flush_dcache_page(result->b_page); + kunmap(result->b_page); + set_buffer_uptodate(result); + SetPageChecked(result->b_page); + btrfs_map_bh_to_logical(root, result, 0); + } +not_found: + if (create & BTRFS_GET_BLOCK_CREATE) { + struct btrfs_key ins; + ret = btrfs_alloc_extent(trans, root, inode->i_ino, + 1, alloc_hint, (u64)-1, + &ins, 1); + BUG_ON(ret); + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + iblock << inode->i_blkbits, + ins.objectid, ins.offset, + ins.offset); + BUG_ON(ret); + SetPageChecked(result->b_page); + btrfs_map_bh_to_logical(root, result, ins.objectid); + } +out: + if (trans) + err = btrfs_end_transaction(trans, root); + btrfs_free_path(path); + return err; +} + +int btrfs_get_block(struct inode *inode, sector_t iblock, + struct buffer_head *result, int create) +{ + int err; + struct btrfs_root *root = BTRFS_I(inode)->root; + mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_get_block_lock(inode, iblock, result, create); + mutex_unlock(&root->fs_info->fs_mutex); + return err; +} + +static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock, + struct buffer_head *result, int create) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + mutex_lock(&root->fs_info->fs_mutex); + btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT); + mutex_unlock(&root->fs_info->fs_mutex); + return 0; +} + +static sector_t btrfs_bmap(struct address_space *as, sector_t block) +{ + return generic_block_bmap(as, block, btrfs_get_block_bmap); +} + +static int btrfs_prepare_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{ + return block_prepare_write(page, from, to, btrfs_get_block); +} + +static int btrfs_readpage(struct file *file, struct page *page) +{ + return mpage_readpage(page, btrfs_get_block); +} + +/* + * Aside from a tiny bit of packed file data handling, this is the + * same as the generic code. + * + * While block_write_full_page is writing back the dirty buffers under + * the page lock, whoever dirtied the buffers may decide to clean them + * again at any time. We handle that by only looking at the buffer + * state inside lock_buffer(). + * + * If block_write_full_page() is called for regular writeback + * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a + * locked buffer. This only can happen if someone has written the buffer + * directly, with submit_bh(). At the address_space level PageWriteback + * prevents this contention from occurring. + */ +static int __btrfs_write_full_page(struct inode *inode, struct page *page, + struct writeback_control *wbc) +{ + int err; + sector_t block; + sector_t last_block; + struct buffer_head *bh, *head; + const unsigned blocksize = 1 << inode->i_blkbits; + int nr_underway = 0; + + BUG_ON(!PageLocked(page)); + + last_block = (i_size_read(inode) - 1) >> inode->i_blkbits; + + if (!page_has_buffers(page)) { + create_empty_buffers(page, blocksize, + (1 << BH_Dirty)|(1 << BH_Uptodate)); + } + + /* + * Be very careful. We have no exclusion from __set_page_dirty_buffers + * here, and the (potentially unmapped) buffers may become dirty at + * any time. If a buffer becomes dirty here after we've inspected it + * then we just miss that fact, and the page stays dirty. + * + * Buffers outside i_size may be dirtied by __set_page_dirty_buffers; + * handle that here by just cleaning them. + */ + + block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); + head = page_buffers(page); + bh = head; + + /* + * Get all the dirty buffers mapped to disk addresses and + * handle any aliases from the underlying blockdev's mapping. + */ + do { + if (block > last_block) { + /* + * mapped buffers outside i_size will occur, because + * this page can be outside i_size when there is a + * truncate in progress. + */ + /* + * The buffer was zeroed by block_write_full_page() + */ + clear_buffer_dirty(bh); + set_buffer_uptodate(bh); + } else if (!buffer_mapped(bh) && buffer_dirty(bh)) { + WARN_ON(bh->b_size != blocksize); + err = btrfs_get_block(inode, block, bh, 0); + if (err) { + goto recover; + } + if (buffer_new(bh)) { + /* blockdev mappings never come here */ + clear_buffer_new(bh); + } + } + bh = bh->b_this_page; + block++; + } while (bh != head); + + do { + if (!buffer_mapped(bh)) + continue; + /* + * If it's a fully non-blocking write attempt and we cannot + * lock the buffer then redirty the page. Note that this can + * potentially cause a busy-wait loop from pdflush and kswapd + * activity, but those code paths have their own higher-level + * throttling. + */ + if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) { + lock_buffer(bh); + } else if (test_set_buffer_locked(bh)) { + redirty_page_for_writepage(wbc, page); + continue; + } + if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) { + mark_buffer_async_write(bh); + } else { + unlock_buffer(bh); + } + } while ((bh = bh->b_this_page) != head); + + /* + * The page and its buffers are protected by PageWriteback(), so we can + * drop the bh refcounts early. + */ + BUG_ON(PageWriteback(page)); + set_page_writeback(page); + + do { + struct buffer_head *next = bh->b_this_page; + if (buffer_async_write(bh)) { + submit_bh(WRITE, bh); + nr_underway++; + } + bh = next; + } while (bh != head); + unlock_page(page); + + err = 0; +done: + if (nr_underway == 0) { + /* + * The page was marked dirty, but the buffers were + * clean. Someone wrote them back by hand with + * ll_rw_block/submit_bh. A rare case. + */ + int uptodate = 1; + do { + if (!buffer_uptodate(bh)) { + uptodate = 0; + break; + } + bh = bh->b_this_page; + } while (bh != head); + if (uptodate) + SetPageUptodate(page); + end_page_writeback(page); + } + return err; + +recover: + /* + * ENOSPC, or some other error. We may already have added some + * blocks to the file, so we need to write these out to avoid + * exposing stale data. + * The page is currently locked and not marked for writeback + */ + bh = head; + /* Recovery: lock and submit the mapped buffers */ + do { + if (buffer_mapped(bh) && buffer_dirty(bh)) { + lock_buffer(bh); + mark_buffer_async_write(bh); + } else { + /* + * The buffer may have been set dirty during + * attachment to a dirty page. + */ + clear_buffer_dirty(bh); + } + } while ((bh = bh->b_this_page) != head); + SetPageError(page); + BUG_ON(PageWriteback(page)); + set_page_writeback(page); + do { + struct buffer_head *next = bh->b_this_page; + if (buffer_async_write(bh)) { + clear_buffer_dirty(bh); + submit_bh(WRITE, bh); + nr_underway++; + } + bh = next; + } while (bh != head); + unlock_page(page); + goto done; +} + +static int btrfs_writepage(struct page *page, struct writeback_control *wbc) +{ + struct inode * const inode = page->mapping->host; + loff_t i_size = i_size_read(inode); + const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT; + unsigned offset; + void *kaddr; + + /* Is the page fully inside i_size? */ + if (page->index < end_index) + return __btrfs_write_full_page(inode, page, wbc); + + /* Is the page fully outside i_size? (truncate in progress) */ + offset = i_size & (PAGE_CACHE_SIZE-1); + if (page->index >= end_index+1 || !offset) { + /* + * The page may have dirty, unmapped buffers. For example, + * they may have been added in ext3_writepage(). Make them + * freeable here, so the page does not leak. + */ + block_invalidatepage(page, 0); + unlock_page(page); + return 0; /* don't care */ + } + + /* + * The page straddles i_size. It must be zeroed out on each and every + * writepage invokation because it may be mmapped. "A file is mapped + * in multiples of the page size. For a file that is not a multiple of + * the page size, the remaining memory is zeroed when mapped, and + * writes to that region are not written out to the file." + */ + kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + return __btrfs_write_full_page(inode, page, wbc); +} + +static void btrfs_truncate(struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret; + struct btrfs_trans_handle *trans; + + if (!S_ISREG(inode->i_mode)) + return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + + btrfs_truncate_page(inode->i_mapping, inode->i_size); + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + + /* FIXME, add redo link to tree so we don't leak on crash */ + ret = btrfs_truncate_in_trans(trans, root, inode); + BUG_ON(ret); + btrfs_update_inode(trans, root, inode); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); +} + +int btrfs_commit_write(struct file *file, struct page *page, + unsigned from, unsigned to) +{ + struct inode *inode = page->mapping->host; + struct buffer_head *bh; + loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; + + SetPageUptodate(page); + bh = page_buffers(page); + set_buffer_uptodate(bh); + if (buffer_mapped(bh) && bh->b_blocknr != 0) { + set_page_dirty(page); + } + if (pos > inode->i_size) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + return 0; +} + +static int create_subvol(struct btrfs_root *root, char *name, int namelen) +{ + struct btrfs_trans_handle *trans; + struct btrfs_key key; + struct btrfs_root_item root_item; + struct btrfs_inode_item *inode_item; + struct buffer_head *subvol; + struct btrfs_leaf *leaf; + struct btrfs_root *new_root; + struct inode *inode; + struct inode *dir; + int ret; + u64 objectid; + u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + BUG_ON(!trans); + + subvol = btrfs_alloc_free_block(trans, root, 0); + if (subvol == NULL) + return -ENOSPC; + leaf = btrfs_buffer_leaf(subvol); + btrfs_set_header_nritems(&leaf->header, 0); + btrfs_set_header_level(&leaf->header, 0); + btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol)); + btrfs_set_header_generation(&leaf->header, trans->transid); + btrfs_set_header_owner(&leaf->header, root->root_key.objectid); + memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid, + sizeof(leaf->header.fsid)); + mark_buffer_dirty(subvol); + + inode_item = &root_item.inode; + memset(inode_item, 0, sizeof(*inode_item)); + btrfs_set_inode_generation(inode_item, 1); + btrfs_set_inode_size(inode_item, 3); + btrfs_set_inode_nlink(inode_item, 1); + btrfs_set_inode_nblocks(inode_item, 1); + btrfs_set_inode_mode(inode_item, S_IFDIR | 0755); + + btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol)); + btrfs_set_root_refs(&root_item, 1); + brelse(subvol); + subvol = NULL; + + ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, + 0, &objectid); + BUG_ON(ret); + + btrfs_set_root_dirid(&root_item, new_dirid); + + key.objectid = objectid; + key.offset = 1; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, + &root_item); + BUG_ON(ret); + + /* + * insert the directory item + */ + key.offset = (u64)-1; + dir = root->fs_info->sb->s_root->d_inode; + ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, + name, namelen, dir->i_ino, &key, + BTRFS_FT_DIR); + BUG_ON(ret); + + ret = btrfs_commit_transaction(trans, root); + BUG_ON(ret); + + new_root = btrfs_read_fs_root(root->fs_info, &key); + BUG_ON(!new_root); + + trans = btrfs_start_transaction(new_root, 1); + BUG_ON(!trans); + + inode = btrfs_new_inode(trans, new_root, new_dirid, + BTRFS_I(dir)->block_group, S_IFDIR | 0700); + inode->i_op = &btrfs_dir_inode_operations; + inode->i_fop = &btrfs_dir_file_operations; + + ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid); + BUG_ON(ret); + + inode->i_nlink = 1; + inode->i_size = 6; + ret = btrfs_update_inode(trans, new_root, inode); + BUG_ON(ret); + + ret = btrfs_commit_transaction(trans, new_root); + BUG_ON(ret); + + iput(inode); + + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); + return 0; +} + +static int create_snapshot(struct btrfs_root *root, char *name, int namelen) +{ + struct btrfs_trans_handle *trans; + struct btrfs_key key; + struct btrfs_root_item new_root_item; + int ret; + u64 objectid; + + if (!root->ref_cows) + return -EINVAL; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + BUG_ON(!trans); + + ret = btrfs_update_inode(trans, root, root->inode); + BUG_ON(ret); + + ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, + 0, &objectid); + BUG_ON(ret); + + memcpy(&new_root_item, &root->root_item, + sizeof(new_root_item)); + + key.objectid = objectid; + key.offset = 1; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node)); + + ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, + &new_root_item); + BUG_ON(ret); + + /* + * insert the directory item + */ + key.offset = (u64)-1; + ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, + name, namelen, + root->fs_info->sb->s_root->d_inode->i_ino, + &key, BTRFS_FT_DIR); + + BUG_ON(ret); + + ret = btrfs_inc_root_ref(trans, root); + BUG_ON(ret); + + ret = btrfs_commit_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); + return 0; +} + +int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int + cmd, unsigned long arg) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_ioctl_vol_args vol_args; + int ret = 0; + struct btrfs_dir_item *di; + int namelen; + struct btrfs_path *path; + u64 root_dirid; + + switch (cmd) { + case BTRFS_IOC_SNAP_CREATE: + if (copy_from_user(&vol_args, + (struct btrfs_ioctl_vol_args __user *)arg, + sizeof(vol_args))) + return -EFAULT; + namelen = strlen(vol_args.name); + if (namelen > BTRFS_VOL_NAME_MAX) + return -EINVAL; + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, + mutex_lock(&root->fs_info->fs_mutex); + di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, + path, root_dirid, + vol_args.name, namelen, 0); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_free_path(path); + if (di && !IS_ERR(di)) + return -EEXIST; + + if (root == root->fs_info->tree_root) + ret = create_subvol(root, vol_args.name, namelen); + else + ret = create_snapshot(root, vol_args.name, namelen); + WARN_ON(ret); + break; + default: + return -ENOTTY; + } + return ret; +} + +#ifdef CONFIG_COMPAT +long btrfs_compat_ioctl(struct file *file, unsigned int cmd, + unsigned long arg) +{ + struct inode *inode = file->f_path.dentry->d_inode; + int ret; + lock_kernel(); + ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg)); + unlock_kernel(); + return ret; + +} +#endif + +/* + * Called inside transaction, so use GFP_NOFS + */ +struct inode *btrfs_alloc_inode(struct super_block *sb) +{ + struct btrfs_inode *ei; + + ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS); + if (!ei) + return NULL; + return &ei->vfs_inode; +} + +void btrfs_destroy_inode(struct inode *inode) +{ + WARN_ON(!list_empty(&inode->i_dentry)); + WARN_ON(inode->i_data.nrpages); + + kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); +} + +static void init_once(void * foo, struct kmem_cache * cachep, + unsigned long flags) +{ + struct btrfs_inode *ei = (struct btrfs_inode *) foo; + + inode_init_once(&ei->vfs_inode); +} + +void btrfs_destroy_cachep(void) +{ + if (btrfs_inode_cachep) + kmem_cache_destroy(btrfs_inode_cachep); + if (btrfs_trans_handle_cachep) + kmem_cache_destroy(btrfs_trans_handle_cachep); + if (btrfs_transaction_cachep) + kmem_cache_destroy(btrfs_transaction_cachep); + if (btrfs_bit_radix_cachep) + kmem_cache_destroy(btrfs_bit_radix_cachep); + if (btrfs_path_cachep) + kmem_cache_destroy(btrfs_path_cachep); +} + +int btrfs_init_cachep(void) +{ + btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache", + sizeof(struct btrfs_inode), + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD), + init_once, NULL); + if (!btrfs_inode_cachep) + goto fail; + btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache", + sizeof(struct btrfs_trans_handle), + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD), + NULL, NULL); + if (!btrfs_trans_handle_cachep) + goto fail; + btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache", + sizeof(struct btrfs_transaction), + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD), + NULL, NULL); + if (!btrfs_transaction_cachep) + goto fail; + btrfs_path_cachep = kmem_cache_create("btrfs_path_cache", + sizeof(struct btrfs_transaction), + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD), + NULL, NULL); + if (!btrfs_path_cachep) + goto fail; + btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix", + 256, + 0, (SLAB_RECLAIM_ACCOUNT| + SLAB_MEM_SPREAD | + SLAB_DESTROY_BY_RCU), + NULL, NULL); + if (!btrfs_bit_radix_cachep) + goto fail; + return 0; +fail: + btrfs_destroy_cachep(); + return -ENOMEM; +} + +static int btrfs_getattr(struct vfsmount *mnt, + struct dentry *dentry, struct kstat *stat) +{ + struct inode *inode = dentry->d_inode; + generic_fillattr(inode, stat); + stat->blksize = 256 * 1024; + return 0; +} + +static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, + struct inode * new_dir,struct dentry *new_dentry) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(old_dir)->root; + struct inode *new_inode = new_dentry->d_inode; + struct inode *old_inode = old_dentry->d_inode; + struct timespec ctime = CURRENT_TIME; + struct btrfs_path *path; + struct btrfs_dir_item *di; + int ret; + + if (S_ISDIR(old_inode->i_mode) && new_inode && + new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) { + return -ENOTEMPTY; + } + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, new_dir); + path = btrfs_alloc_path(); + if (!path) { + ret = -ENOMEM; + goto out_fail; + } + + old_dentry->d_inode->i_nlink++; + old_dir->i_ctime = old_dir->i_mtime = ctime; + new_dir->i_ctime = new_dir->i_mtime = ctime; + old_inode->i_ctime = ctime; + if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) { + struct btrfs_key *location = &BTRFS_I(new_dir)->location; + u64 old_parent_oid; + di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino, + "..", 2, -1); + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto out_fail; + } + if (!di) { + ret = -ENOENT; + goto out_fail; + } + old_parent_oid = btrfs_disk_key_objectid(&di->location); + ret = btrfs_del_item(trans, root, path); + if (ret) { + ret = -EIO; + goto out_fail; + } + btrfs_release_path(root, path); + + di = btrfs_lookup_dir_index_item(trans, root, path, + old_inode->i_ino, + old_parent_oid, + "..", 2, -1); + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto out_fail; + } + if (!di) { + ret = -ENOENT; + goto out_fail; + } + ret = btrfs_del_item(trans, root, path); + if (ret) { + ret = -EIO; + goto out_fail; + } + btrfs_release_path(root, path); + + ret = btrfs_insert_dir_item(trans, root, "..", 2, + old_inode->i_ino, location, + BTRFS_FT_DIR); + if (ret) + goto out_fail; + } + + + ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry); + if (ret) + goto out_fail; + + if (new_inode) { + new_inode->i_ctime = CURRENT_TIME; + ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry); + if (ret) + goto out_fail; + if (S_ISDIR(new_inode->i_mode)) + clear_nlink(new_inode); + else + drop_nlink(new_inode); + btrfs_update_inode(trans, root, new_inode); + } + ret = btrfs_add_link(trans, new_dentry, old_inode); + if (ret) + goto out_fail; + +out_fail: + btrfs_free_path(path); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +static int btrfs_symlink(struct inode *dir, struct dentry *dentry, + const char *symname) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(dir)->root; + struct btrfs_path *path; + struct btrfs_key key; + struct inode *inode; + int err; + int drop_inode = 0; + u64 objectid; + int name_len; + int datasize; + char *ptr; + struct btrfs_file_extent_item *ei; + + name_len = strlen(symname) + 1; + if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) + return -ENAMETOOLONG; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + + err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); + if (err) { + err = -ENOSPC; + goto out_unlock; + } + + inode = btrfs_new_inode(trans, root, objectid, + BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_unlock; + + btrfs_set_trans_block_group(trans, inode); + err = btrfs_add_nondir(trans, dentry, inode); + if (err) + drop_inode = 1; + else { + inode->i_mapping->a_ops = &btrfs_aops; + inode->i_fop = &btrfs_file_operations; + inode->i_op = &btrfs_file_inode_operations; + } + dir->i_sb->s_dirt = 1; + btrfs_update_inode_block_group(trans, inode); + btrfs_update_inode_block_group(trans, dir); + if (drop_inode) + goto out_unlock; + + path = btrfs_alloc_path(); + BUG_ON(!path); + key.objectid = inode->i_ino; + key.offset = 0; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); + datasize = btrfs_file_extent_calc_inline_size(name_len); + err = btrfs_insert_empty_item(trans, root, path, &key, + datasize); + BUG_ON(err); + ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), + path->slots[0], struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(ei, trans->transid); + btrfs_set_file_extent_type(ei, + BTRFS_FILE_EXTENT_INLINE); + ptr = btrfs_file_extent_inline_start(ei); + btrfs_memcpy(root, path->nodes[0]->b_data, + ptr, symname, name_len); + mark_buffer_dirty(path->nodes[0]); + btrfs_free_path(path); + inode->i_op = &btrfs_symlink_inode_operations; + inode->i_mapping->a_ops = &btrfs_symlink_aops; + inode->i_size = name_len - 1; + btrfs_update_inode(trans, root, inode); + err = 0; + +out_unlock: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + if (drop_inode) { + inode_dec_link_count(inode); + iput(inode); + } + btrfs_btree_balance_dirty(root); + return err; +} + +static struct inode_operations btrfs_dir_inode_operations = { + .lookup = btrfs_lookup, + .create = btrfs_create, + .unlink = btrfs_unlink, + .link = btrfs_link, + .mkdir = btrfs_mkdir, + .rmdir = btrfs_rmdir, + .rename = btrfs_rename, + .symlink = btrfs_symlink, + .setattr = btrfs_setattr, +}; + +static struct inode_operations btrfs_dir_ro_inode_operations = { + .lookup = btrfs_lookup, +}; + +static struct file_operations btrfs_dir_file_operations = { + .llseek = generic_file_llseek, + .read = generic_read_dir, + .readdir = btrfs_readdir, + .ioctl = btrfs_ioctl, +#ifdef CONFIG_COMPAT + .compat_ioctl = btrfs_compat_ioctl, +#endif +}; + +static struct address_space_operations btrfs_aops = { + .readpage = btrfs_readpage, + .writepage = btrfs_writepage, + .sync_page = block_sync_page, + .prepare_write = btrfs_prepare_write, + .commit_write = btrfs_commit_write, + .bmap = btrfs_bmap, +}; + +static struct address_space_operations btrfs_symlink_aops = { + .readpage = btrfs_readpage, + .writepage = btrfs_writepage, +}; + +static struct inode_operations btrfs_file_inode_operations = { + .truncate = btrfs_truncate, + .getattr = btrfs_getattr, + .setattr = btrfs_setattr, +}; + +static struct inode_operations btrfs_symlink_inode_operations = { + .readlink = generic_readlink, + .follow_link = page_follow_link_light, + .put_link = page_put_link, +}; -- cgit v1.2.3-70-g09d2 From 8a712645c3fbbf43293af8640af337b7cf1fef01 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 12 Jun 2007 08:21:35 -0400 Subject: Btrfs: no slashes in subvolume names Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 62a3a778d37..1cabc1951b0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1831,6 +1831,8 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int namelen = strlen(vol_args.name); if (namelen > BTRFS_VOL_NAME_MAX) return -EINVAL; + if (strchr(vol_args.name, '/')) + return -EINVAL; path = btrfs_alloc_path(); if (!path) return -ENOMEM; -- cgit v1.2.3-70-g09d2 From 6cbd55707802b98843f953d1ae6d8f5bcd9a76c0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 12 Jun 2007 09:07:21 -0400 Subject: Btrfs: add GPLv2 Signed-off-by: Chris Mason --- fs/btrfs/COPYING | 356 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/TODO | 1 - fs/btrfs/bit-radix.c | 18 +++ fs/btrfs/bit-radix.h | 18 +++ fs/btrfs/btrfs_inode.h | 18 +++ fs/btrfs/ctree.c | 18 +++ fs/btrfs/ctree.h | 18 +++ fs/btrfs/dir-item.c | 18 +++ fs/btrfs/disk-io.c | 18 +++ fs/btrfs/disk-io.h | 18 +++ fs/btrfs/extent-tree.c | 18 +++ fs/btrfs/file-item.c | 18 +++ fs/btrfs/file.c | 18 +++ fs/btrfs/hash.c | 18 +++ fs/btrfs/hash.h | 18 +++ fs/btrfs/inode-item.c | 18 +++ fs/btrfs/inode-map.c | 18 +++ fs/btrfs/inode.c | 18 +++ fs/btrfs/ioctl.h | 18 +++ fs/btrfs/print-tree.c | 18 +++ fs/btrfs/print-tree.h | 18 +++ fs/btrfs/root-tree.c | 18 +++ fs/btrfs/super.c | 18 +++ fs/btrfs/sysfs.c | 18 +++ fs/btrfs/transaction.c | 18 +++ fs/btrfs/transaction.h | 18 +++ 26 files changed, 788 insertions(+), 1 deletion(-) create mode 100644 fs/btrfs/COPYING (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/COPYING b/fs/btrfs/COPYING new file mode 100644 index 00000000000..ca442d313d8 --- /dev/null +++ b/fs/btrfs/COPYING @@ -0,0 +1,356 @@ + + NOTE! This copyright does *not* cover user programs that use kernel + services by normal system calls - this is merely considered normal use + of the kernel, and does *not* fall under the heading of "derived work". + Also note that the GPL below is copyrighted by the Free Software + Foundation, but the instance of code that it refers to (the Linux + kernel) is copyrighted by me and others who actually wrote it. + + Also note that the only valid version of the GPL as far as the kernel + is concerned is _this_ particular version of the license (ie v2, not + v2.2 or v3.x or whatever), unless explicitly otherwise stated. + + Linus Torvalds + +---------------------------------------- + + GNU GENERAL PUBLIC LICENSE + Version 2, June 1991 + + Copyright (C) 1989, 1991 Free Software Foundation, Inc. + 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + Everyone is permitted to copy and distribute verbatim copies + of this license document, but changing it is not allowed. + + Preamble + + The licenses for most software are designed to take away your +freedom to share and change it. By contrast, the GNU General Public +License is intended to guarantee your freedom to share and change free +software--to make sure the software is free for all its users. This +General Public License applies to most of the Free Software +Foundation's software and to any other program whose authors commit to +using it. (Some other Free Software Foundation software is covered by +the GNU Library General Public License instead.) You can apply it to +your programs, too. + + When we speak of free software, we are referring to freedom, not +price. Our General Public Licenses are designed to make sure that you +have the freedom to distribute copies of free software (and charge for +this service if you wish), that you receive source code or can get it +if you want it, that you can change the software or use pieces of it +in new free programs; and that you know you can do these things. + + To protect your rights, we need to make restrictions that forbid +anyone to deny you these rights or to ask you to surrender the rights. +These restrictions translate to certain responsibilities for you if you +distribute copies of the software, or if you modify it. + + For example, if you distribute copies of such a program, whether +gratis or for a fee, you must give the recipients all the rights that +you have. You must make sure that they, too, receive or can get the +source code. And you must show them these terms so they know their +rights. + + We protect your rights with two steps: (1) copyright the software, and +(2) offer you this license which gives you legal permission to copy, +distribute and/or modify the software. + + Also, for each author's protection and ours, we want to make certain +that everyone understands that there is no warranty for this free +software. If the software is modified by someone else and passed on, we +want its recipients to know that what they have is not the original, so +that any problems introduced by others will not reflect on the original +authors' reputations. + + Finally, any free program is threatened constantly by software +patents. We wish to avoid the danger that redistributors of a free +program will individually obtain patent licenses, in effect making the +program proprietary. To prevent this, we have made it clear that any +patent must be licensed for everyone's free use or not licensed at all. + + The precise terms and conditions for copying, distribution and +modification follow. + + GNU GENERAL PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. This License applies to any program or other work which contains +a notice placed by the copyright holder saying it may be distributed +under the terms of this General Public License. The "Program", below, +refers to any such program or work, and a "work based on the Program" +means either the Program or any derivative work under copyright law: +that is to say, a work containing the Program or a portion of it, +either verbatim or with modifications and/or translated into another +language. (Hereinafter, translation is included without limitation in +the term "modification".) Each licensee is addressed as "you". + +Activities other than copying, distribution and modification are not +covered by this License; they are outside its scope. The act of +running the Program is not restricted, and the output from the Program +is covered only if its contents constitute a work based on the +Program (independent of having been made by running the Program). +Whether that is true depends on what the Program does. + + 1. You may copy and distribute verbatim copies of the Program's +source code as you receive it, in any medium, provided that you +conspicuously and appropriately publish on each copy an appropriate +copyright notice and disclaimer of warranty; keep intact all the +notices that refer to this License and to the absence of any warranty; +and give any other recipients of the Program a copy of this License +along with the Program. + +You may charge a fee for the physical act of transferring a copy, and +you may at your option offer warranty protection in exchange for a fee. + + 2. You may modify your copy or copies of the Program or any portion +of it, thus forming a work based on the Program, and copy and +distribute such modifications or work under the terms of Section 1 +above, provided that you also meet all of these conditions: + + a) You must cause the modified files to carry prominent notices + stating that you changed the files and the date of any change. + + b) You must cause any work that you distribute or publish, that in + whole or in part contains or is derived from the Program or any + part thereof, to be licensed as a whole at no charge to all third + parties under the terms of this License. + + c) If the modified program normally reads commands interactively + when run, you must cause it, when started running for such + interactive use in the most ordinary way, to print or display an + announcement including an appropriate copyright notice and a + notice that there is no warranty (or else, saying that you provide + a warranty) and that users may redistribute the program under + these conditions, and telling the user how to view a copy of this + License. (Exception: if the Program itself is interactive but + does not normally print such an announcement, your work based on + the Program is not required to print an announcement.) + +These requirements apply to the modified work as a whole. If +identifiable sections of that work are not derived from the Program, +and can be reasonably considered independent and separate works in +themselves, then this License, and its terms, do not apply to those +sections when you distribute them as separate works. But when you +distribute the same sections as part of a whole which is a work based +on the Program, the distribution of the whole must be on the terms of +this License, whose permissions for other licensees extend to the +entire whole, and thus to each and every part regardless of who wrote it. + +Thus, it is not the intent of this section to claim rights or contest +your rights to work written entirely by you; rather, the intent is to +exercise the right to control the distribution of derivative or +collective works based on the Program. + +In addition, mere aggregation of another work not based on the Program +with the Program (or with a work based on the Program) on a volume of +a storage or distribution medium does not bring the other work under +the scope of this License. + + 3. You may copy and distribute the Program (or a work based on it, +under Section 2) in object code or executable form under the terms of +Sections 1 and 2 above provided that you also do one of the following: + + a) Accompany it with the complete corresponding machine-readable + source code, which must be distributed under the terms of Sections + 1 and 2 above on a medium customarily used for software interchange; or, + + b) Accompany it with a written offer, valid for at least three + years, to give any third party, for a charge no more than your + cost of physically performing source distribution, a complete + machine-readable copy of the corresponding source code, to be + distributed under the terms of Sections 1 and 2 above on a medium + customarily used for software interchange; or, + + c) Accompany it with the information you received as to the offer + to distribute corresponding source code. (This alternative is + allowed only for noncommercial distribution and only if you + received the program in object code or executable form with such + an offer, in accord with Subsection b above.) + +The source code for a work means the preferred form of the work for +making modifications to it. For an executable work, complete source +code means all the source code for all modules it contains, plus any +associated interface definition files, plus the scripts used to +control compilation and installation of the executable. However, as a +special exception, the source code distributed need not include +anything that is normally distributed (in either source or binary +form) with the major components (compiler, kernel, and so on) of the +operating system on which the executable runs, unless that component +itself accompanies the executable. + +If distribution of executable or object code is made by offering +access to copy from a designated place, then offering equivalent +access to copy the source code from the same place counts as +distribution of the source code, even though third parties are not +compelled to copy the source along with the object code. + + 4. You may not copy, modify, sublicense, or distribute the Program +except as expressly provided under this License. Any attempt +otherwise to copy, modify, sublicense or distribute the Program is +void, and will automatically terminate your rights under this License. +However, parties who have received copies, or rights, from you under +this License will not have their licenses terminated so long as such +parties remain in full compliance. + + 5. You are not required to accept this License, since you have not +signed it. However, nothing else grants you permission to modify or +distribute the Program or its derivative works. These actions are +prohibited by law if you do not accept this License. Therefore, by +modifying or distributing the Program (or any work based on the +Program), you indicate your acceptance of this License to do so, and +all its terms and conditions for copying, distributing or modifying +the Program or works based on it. + + 6. Each time you redistribute the Program (or any work based on the +Program), the recipient automatically receives a license from the +original licensor to copy, distribute or modify the Program subject to +these terms and conditions. You may not impose any further +restrictions on the recipients' exercise of the rights granted herein. +You are not responsible for enforcing compliance by third parties to +this License. + + 7. If, as a consequence of a court judgment or allegation of patent +infringement or for any other reason (not limited to patent issues), +conditions are imposed on you (whether by court order, agreement or +otherwise) that contradict the conditions of this License, they do not +excuse you from the conditions of this License. If you cannot +distribute so as to satisfy simultaneously your obligations under this +License and any other pertinent obligations, then as a consequence you +may not distribute the Program at all. For example, if a patent +license would not permit royalty-free redistribution of the Program by +all those who receive copies directly or indirectly through you, then +the only way you could satisfy both it and this License would be to +refrain entirely from distribution of the Program. + +If any portion of this section is held invalid or unenforceable under +any particular circumstance, the balance of the section is intended to +apply and the section as a whole is intended to apply in other +circumstances. + +It is not the purpose of this section to induce you to infringe any +patents or other property right claims or to contest validity of any +such claims; this section has the sole purpose of protecting the +integrity of the free software distribution system, which is +implemented by public license practices. Many people have made +generous contributions to the wide range of software distributed +through that system in reliance on consistent application of that +system; it is up to the author/donor to decide if he or she is willing +to distribute software through any other system and a licensee cannot +impose that choice. + +This section is intended to make thoroughly clear what is believed to +be a consequence of the rest of this License. + + 8. If the distribution and/or use of the Program is restricted in +certain countries either by patents or by copyrighted interfaces, the +original copyright holder who places the Program under this License +may add an explicit geographical distribution limitation excluding +those countries, so that distribution is permitted only in or among +countries not thus excluded. In such case, this License incorporates +the limitation as if written in the body of this License. + + 9. The Free Software Foundation may publish revised and/or new versions +of the General Public License from time to time. Such new versions will +be similar in spirit to the present version, but may differ in detail to +address new problems or concerns. + +Each version is given a distinguishing version number. If the Program +specifies a version number of this License which applies to it and "any +later version", you have the option of following the terms and conditions +either of that version or of any later version published by the Free +Software Foundation. If the Program does not specify a version number of +this License, you may choose any version ever published by the Free Software +Foundation. + + 10. If you wish to incorporate parts of the Program into other free +programs whose distribution conditions are different, write to the author +to ask for permission. For software which is copyrighted by the Free +Software Foundation, write to the Free Software Foundation; we sometimes +make exceptions for this. Our decision will be guided by the two goals +of preserving the free status of all derivatives of our free software and +of promoting the sharing and reuse of software generally. + + NO WARRANTY + + 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY +FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN +OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES +PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED +OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS +TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE +PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, +REPAIR OR CORRECTION. + + 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING +WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR +REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, +INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING +OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED +TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY +YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER +PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE +POSSIBILITY OF SUCH DAMAGES. + + END OF TERMS AND CONDITIONS + + How to Apply These Terms to Your New Programs + + If you develop a new program, and you want it to be of the greatest +possible use to the public, the best way to achieve this is to make it +free software which everyone can redistribute and change under these terms. + + To do so, attach the following notices to the program. It is safest +to attach them to the start of each source file to most effectively +convey the exclusion of warranty; and each file should have at least +the "copyright" line and a pointer to where the full notice is found. + + + Copyright (C) + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + + +Also add information on how to contact you by electronic and paper mail. + +If the program is interactive, make it output a short notice like this +when it starts in an interactive mode: + + Gnomovision version 69, Copyright (C) year name of author + Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'. + This is free software, and you are welcome to redistribute it + under certain conditions; type `show c' for details. + +The hypothetical commands `show w' and `show c' should show the appropriate +parts of the General Public License. Of course, the commands you use may +be called something other than `show w' and `show c'; they could even be +mouse-clicks or menu items--whatever suits your program. + +You should also get your employer (if you work as a programmer) or your +school, if any, to sign a "copyright disclaimer" for the program, if +necessary. Here is a sample; alter the names: + + Yoyodyne, Inc., hereby disclaims all copyright interest in the program + `Gnomovision' (which makes passes at compilers) written by James Hacker. + + , 1 April 1989 + Ty Coon, President of Vice + +This General Public License does not permit incorporating your program into +proprietary programs. If your program is a subroutine library, you may +consider it more useful to permit linking proprietary applications with the +library. If this is what you want to do, use the GNU Library General +Public License instead of this License. diff --git a/fs/btrfs/TODO b/fs/btrfs/TODO index 4a48a2365d5..d9b6d38c603 100644 --- a/fs/btrfs/TODO +++ b/fs/btrfs/TODO @@ -1,4 +1,3 @@ -* fix printk warnings * cleanup, add more error checking, get rid of BUG_ONs * Fix ENOSPC handling * Make allocator smarter diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c index 8f9cd427723..a4ddf4de979 100644 --- a/fs/btrfs/bit-radix.c +++ b/fs/btrfs/bit-radix.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "bit-radix.h" diff --git a/fs/btrfs/bit-radix.h b/fs/btrfs/bit-radix.h index 4e717e30db4..c100f54d5c3 100644 --- a/fs/btrfs/bit-radix.h +++ b/fs/btrfs/bit-radix.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __BIT_RADIX__ #define __BIT_RADIX__ #include diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 6b50076b02d..783f06b6325 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __BTRFS_I__ #define __BTRFS_I__ diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 20e01f95ce6..4736155af41 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 24410d42e3c..4bd648d68e8 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __BTRFS__ #define __BTRFS__ diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index b408a3d20c7..0d179a44ebf 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 87810117254..8c68a64c0ce 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include #include diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 822ccb8aa4a..7b76ccc4875 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __DISKIO__ #define __DISKIO__ diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index fe02fbfa6d5..975a0eb9241 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index d5a98827e38..37dd3801a10 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 0325dc03859..566e526b60f 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include #include diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c index 32de1ea1b64..e5c76903d72 100644 --- a/fs/btrfs/hash.c +++ b/fs/btrfs/hash.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + /* * Original copy from: * linux/fs/ext3/hash.c diff --git a/fs/btrfs/hash.h b/fs/btrfs/hash.h index e56c4670093..d3be0267058 100644 --- a/fs/btrfs/hash.h +++ b/fs/btrfs/hash.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __HASH__ #define __HASH__ int btrfs_name_hash(const char *name, int len, u64 *hash_result); diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index c3b990b661c..449e9dbdf95 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 318e27a6378..58930fc48be 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1cabc1951b0..46f869d6d0f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include #include diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 497fadd98c3..23bed48c51a 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __IOCTL_ #define __IOCTL_ #include diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 21791f03756..1fa70568300 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h index 396041a05cf..205cd03601a 100644 --- a/fs/btrfs/print-tree.h +++ b/fs/btrfs/print-tree.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __PRINT_TREE_ #define __PRINT_TREE_ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l); diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 13c30284fd8..7a2d63c1ee8 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 94b3535a2a1..ef8ad18b275 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include #include diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index ed8b7e20cfb..ead9dc8051b 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 4f3c8ddcb40..30914521cf0 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #include #include #include "ctree.h" diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index f25b4900db4..5fb1d322aa1 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __TRANSACTION__ #define __TRANSACTION__ #include "btrfs_inode.h" -- cgit v1.2.3-70-g09d2 From 340887809d92a4dd6c71caa97d658cef32c098c3 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 12 Jun 2007 11:36:58 -0400 Subject: Btrfs: i386 fixes from axboe Signed-off-by: Chris Mason --- fs/btrfs/INSTALL | 51 ++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/disk-io.c | 2 +- fs/btrfs/file.c | 7 +++++-- fs/btrfs/inode.c | 3 +-- fs/btrfs/transaction.c | 1 + 5 files changed, 59 insertions(+), 5 deletions(-) create mode 100644 fs/btrfs/INSTALL (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/INSTALL b/fs/btrfs/INSTALL new file mode 100644 index 00000000000..e83ff6e8c8d --- /dev/null +++ b/fs/btrfs/INSTALL @@ -0,0 +1,51 @@ +Install Instructions + +Btrfs puts snapshots and subvolumes into the root directory of the FS. This +directory can only be changed by btrfsctl right now, and normal filesystem +operations do not work on it. The default subvolume is called 'default', +and you can create files and directories in mount_point/default + +Btrfs uses the crypto manager interface in the kernel for file and +metadata checksums. You need to compile the kernel with: + +CONFIG_CRYPTO=y +CONFIG_CRYPTO_MANAGER=m +CONFIG_CRYPTO_CRC32C=m + +cryptomanager and crc32c can be static as well. Once your kernel is +setup, typing make in the btrfs module sources will build against the +running kernel. When the build is complete: + +modprobe crc32c +modprobe cryptomgr +insmod btrfs.ko + +The Btrfs utility programs require libuuid to build. This can be found +in the e2fsprogs sources, and is usually available as libuuid or +e2fsprogs-devel from various distros. + +Building the utilities is just make ; make install. The programs go +into /usr/local/bin. The commands available are: + +mkfs.btrfs: create a filesystem + +btrfsctl: control program to create snapshots and subvolumes: + + mount /dev/sda2 /mnt + btrfsctl -s new_subvol_name /mnt + btrfsctl -s snapshot_of_default /mnt/default + btrfsctl -s snapshot_of_new_subvol /mnt/new_subvol_name + btrfsctl -s snapshot_of_a_snapshot /mnt/snapshot_of_new_subvol + ls /mnt + default snapshot_of_a_snapshot snapshot_of_new_subvol + new_subvol_name snapshot_of_default + + Snapshots and subvolumes cannot be deleted right now, but you can + rm -rf all the files and directories inside them. + +btrfsck: do a limited check of the FS extent trees. + +debug-tree: print all of the FS metadata in text form. Example: + + debug-tree /dev/sda2 >& big_output_file + diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 8c68a64c0ce..b0ebba6f2e4 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -106,7 +106,7 @@ struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root, int err; u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits); - page = grab_cache_page(mapping, index); + page = find_or_create_page(mapping, index, GFP_NOFS); if (!page) return NULL; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 566e526b60f..83836fb3d4e 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -112,7 +112,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, if (buffer_mapped(bh) && bh->b_blocknr == 0) { struct btrfs_key key; struct btrfs_path *path; - char *ptr; + char *ptr, *kaddr; u32 datasize; /* create an inline extent, and copy the data in */ @@ -135,8 +135,11 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, btrfs_set_file_extent_type(ei, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); + kaddr = kmap_atomic(bh->b_page, KM_USER0); btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, bh->b_data, offset + write_bytes); + ptr, kaddr + bh_offset(bh), + offset + write_bytes); + kunmap_atomic(kaddr, KM_USER0); mark_buffer_dirty(path->nodes[0]); btrfs_free_path(path); } else if (buffer_mapped(bh)) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 46f869d6d0f..d3d5a4cdb6a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1754,6 +1754,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) BTRFS_I(dir)->block_group, S_IFDIR | 0700); inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; + new_root->inode = inode; ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid); BUG_ON(ret); @@ -1766,8 +1767,6 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) ret = btrfs_commit_transaction(trans, new_root); BUG_ON(ret); - iput(inode); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root); return 0; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 30914521cf0..cd4f1d14038 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -18,6 +18,7 @@ #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" -- cgit v1.2.3-70-g09d2 From f1ace244c8c1e16eaa5c8b3b5339849651e31ede Mon Sep 17 00:00:00 2001 From: Aneesh Date: Wed, 13 Jun 2007 16:18:26 -0400 Subject: btrfs: Code cleanup Attaching below is some of the code cleanups that i came across while reading the code. a) alloc_path already calls init_path. b) Mention that btrfs_inode is the in memory copy.Ext4 have ext4_inode_info as the in memory copy ext4_inode as the disk copy Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/ctree.c | 1 - fs/btrfs/dir-item.c | 1 - fs/btrfs/extent-tree.c | 4 ---- fs/btrfs/file-item.c | 2 -- fs/btrfs/inode-item.c | 1 - fs/btrfs/inode.c | 10 +--------- fs/btrfs/root-tree.c | 3 --- 8 files changed, 2 insertions(+), 21 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 783f06b6325..6bce46be49f 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -19,6 +19,7 @@ #ifndef __BTRFS_I__ #define __BTRFS_I__ +/* in memory btrfs inode */ struct btrfs_inode { struct btrfs_root *root; struct btrfs_block_group_cache *block_group; diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 4736155af41..498dea0854c 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1647,7 +1647,6 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size); if (!ret) { ptr = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 0d179a44ebf..ff10cf5a842 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -74,7 +74,6 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root ret = btrfs_name_hash(name, name_len, &key.offset); BUG_ON(ret); path = btrfs_alloc_path(); - btrfs_init_path(path); data_size = sizeof(*dir_item) + name_len; dir_item = insert_with_overflow(trans, root, path, &key, data_size, name, name_len); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 975a0eb9241..5f4eaba8b85 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -402,7 +402,6 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, &ins, 0); path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); key.objectid = blocknr; key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); @@ -437,7 +436,6 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_extent_item *item; path = btrfs_alloc_path(); - btrfs_init_path(path); key.objectid = blocknr; key.offset = num_blocks; key.flags = 0; @@ -807,7 +805,6 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root find_free_extent(trans, root, 0, 0, (u64)-1, 0, &ins, 0); path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); if (ret) { @@ -1463,7 +1460,6 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); level = btrfs_header_level(btrfs_buffer_header(snap)); orig_level = level; diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 37dd3801a10..3c140cc1dd0 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -38,7 +38,6 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); file_key.objectid = objectid; file_key.offset = pos; file_key.flags = 0; @@ -274,7 +273,6 @@ int btrfs_csum_verify_file_block(struct btrfs_root *root, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); file_key.objectid = objectid; file_key.offset = offset; file_key.flags = 0; diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index 449e9dbdf95..e8b5a9f7c0c 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -35,7 +35,6 @@ int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_insert_item(trans, root, &key, inode_item, sizeof(*inode_item)); btrfs_release_path(root, path); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d3d5a4cdb6a..e639cc9e089 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -79,7 +79,6 @@ void btrfs_read_locked_inode(struct inode *inode) path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); mutex_lock(&root->fs_info->fs_mutex); memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); @@ -177,7 +176,6 @@ static int btrfs_update_inode(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location, 1); if (ret) { @@ -214,7 +212,6 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino, name, name_len, -1); if (IS_ERR(di)) { @@ -289,7 +286,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -360,7 +356,6 @@ static int btrfs_free_inode(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location, -1); BUG_ON(ret); @@ -687,7 +682,6 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, namelen, 0); if (!di || IS_ERR(di)) { @@ -874,7 +868,6 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; path = btrfs_alloc_path(); - btrfs_init_path(path); ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; @@ -1263,7 +1256,6 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); if (create & BTRFS_GET_BLOCK_CREATE) { WARN_ON(1); /* this almost but not quite works */ @@ -1280,7 +1272,7 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock, } ret = btrfs_lookup_file_extent(NULL, root, path, - inode->i_ino, + objectid, iblock << inode->i_blkbits, 0); if (ret < 0) { err = ret; diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 7a2d63c1ee8..8e1b9046d5e 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -36,7 +36,6 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); if (ret < 0) goto out; @@ -70,7 +69,6 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_search_slot(trans, root, key, path, 0, 1); if (ret < 0) goto out; @@ -106,7 +104,6 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, path = btrfs_alloc_path(); BUG_ON(!path); - btrfs_init_path(path); ret = btrfs_search_slot(trans, root, key, path, -1, 1); if (ret < 0) goto out; -- cgit v1.2.3-70-g09d2 From 9ebefb180bad4914a31c4e1748ba187a30e1e990 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 15 Jun 2007 13:50:00 -0400 Subject: Btrfs: patch queue: page_mkwrite Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 + fs/btrfs/disk-io.c | 4 +- fs/btrfs/file-item.c | 2 +- fs/btrfs/file.c | 176 +++++------------------------- fs/btrfs/inode.c | 303 ++++++++++++++++++++++++++++++++++++++++++++++++++- 5 files changed, 328 insertions(+), 159 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 4bd648d68e8..1e83ad720b1 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1142,6 +1142,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); +int btrfs_readpage(struct file *file, struct page *page); void btrfs_delete_inode(struct inode *inode); void btrfs_read_locked_inode(struct inode *inode); int btrfs_write_inode(struct inode *inode, int wait); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index b0ebba6f2e4..602b63dc76c 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -155,9 +155,9 @@ int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, desc.tfm = tfm; desc.flags = 0; sg_init_one(&sg, data, len); - spin_lock(&root->fs_info->hash_lock); + spin_lock_irq(&root->fs_info->hash_lock); ret = crypto_hash_digest(&desc, &sg, 1, result); - spin_unlock(&root->fs_info->hash_lock); + spin_unlock_irq(&root->fs_info->hash_lock); if (ret) { printk("digest failed\n"); } diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 3c140cc1dd0..1068993ab1c 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -228,6 +228,7 @@ found: path->nodes[0]->b_data, root->fs_info->sb->s_blocksize); ret = btrfs_csum_data(root, data, len, &item->csum); +// printk("file %lu offset %llu csum %X\n", objectid, (unsigned long long)offset, *(int *)(&item->csum)); btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_release_path(root, path); @@ -298,4 +299,3 @@ fail: mutex_unlock(&root->fs_info->fs_mutex); return ret; } - diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 83836fb3d4e..de8d47b44e1 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -103,10 +103,6 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes); /* FIXME, one block at a time */ - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - bh = page_buffers(pages[i]); if (buffer_mapped(bh) && bh->b_blocknr == 0) { @@ -115,6 +111,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, char *ptr, *kaddr; u32 datasize; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + /* create an inline extent, and copy the data in */ path = btrfs_alloc_path(); BUG_ON(!path); @@ -135,24 +135,19 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, btrfs_set_file_extent_type(ei, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); + kaddr = kmap_atomic(bh->b_page, KM_USER0); btrfs_memcpy(root, path->nodes[0]->b_data, ptr, kaddr + bh_offset(bh), offset + write_bytes); kunmap_atomic(kaddr, KM_USER0); + mark_buffer_dirty(path->nodes[0]); btrfs_free_path(path); - } else if (buffer_mapped(bh)) { - /* csum the file data */ - btrfs_csum_file_block(trans, root, inode->i_ino, - pages[i]->index << PAGE_CACHE_SHIFT, - kmap(pages[i]), PAGE_CACHE_SIZE); - kunmap(pages[i]); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); } - SetPageChecked(pages[i]); - ret = btrfs_end_transaction(trans, root); - BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); ret = btrfs_commit_write(file, pages[i], offset, offset + this_write); @@ -503,7 +498,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, if ((pos & (PAGE_CACHE_SIZE - 1))) { pinned[0] = grab_cache_page(inode->i_mapping, first_index); if (!PageUptodate(pinned[0])) { - ret = mpage_readpage(pinned[0], btrfs_get_block); + ret = btrfs_readpage(NULL, pinned[0]); BUG_ON(ret); wait_on_page_locked(pinned[0]); } else { @@ -513,7 +508,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, if ((pos + count) & (PAGE_CACHE_SIZE - 1)) { pinned[1] = grab_cache_page(inode->i_mapping, last_index); if (!PageUptodate(pinned[1])) { - ret = mpage_readpage(pinned[1], btrfs_get_block); + ret = btrfs_readpage(NULL, pinned[1]); BUG_ON(ret); wait_on_page_locked(pinned[1]); } else { @@ -633,138 +628,6 @@ out: return num_written ? num_written : err; } -/* - * FIXME, do this by stuffing the csum we want in the info hanging off - * page->private. For now, verify file csums on read - */ -static int btrfs_read_actor(read_descriptor_t *desc, struct page *page, - unsigned long offset, unsigned long size) -{ - char *kaddr; - unsigned long left, count = desc->count; - struct inode *inode = page->mapping->host; - - if (size > count) - size = count; - - if (!PageChecked(page)) { - /* FIXME, do it per block */ - struct btrfs_root *root = BTRFS_I(inode)->root; - int ret; - struct buffer_head *bh; - - if (page_has_buffers(page)) { - bh = page_buffers(page); - if (!buffer_mapped(bh)) { - SetPageChecked(page); - goto checked; - } - } - - ret = btrfs_csum_verify_file_block(root, - page->mapping->host->i_ino, - page->index << PAGE_CACHE_SHIFT, - kmap(page), PAGE_CACHE_SIZE); - if (ret) { - if (ret != -ENOENT) { - printk("failed to verify ino %lu page %lu ret %d\n", - page->mapping->host->i_ino, - page->index, ret); - memset(page_address(page), 1, PAGE_CACHE_SIZE); - flush_dcache_page(page); - } - } - SetPageChecked(page); - kunmap(page); - } -checked: - /* - * Faults on the destination of a read are common, so do it before - * taking the kmap. - */ - if (!fault_in_pages_writeable(desc->arg.buf, size)) { - kaddr = kmap_atomic(page, KM_USER0); - left = __copy_to_user_inatomic(desc->arg.buf, - kaddr + offset, size); - kunmap_atomic(kaddr, KM_USER0); - if (left == 0) - goto success; - } - - /* Do it the slow way */ - kaddr = kmap(page); - left = __copy_to_user(desc->arg.buf, kaddr + offset, size); - kunmap(page); - - if (left) { - size -= left; - desc->error = -EFAULT; - } -success: - desc->count = count - size; - desc->written += size; - desc->arg.buf += size; - return size; -} - -/** - * btrfs_file_aio_read - filesystem read routine, with a mod to csum verify - * @iocb: kernel I/O control block - * @iov: io vector request - * @nr_segs: number of segments in the iovec - * @pos: current file position - */ -static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov, - unsigned long nr_segs, loff_t pos) -{ - struct file *filp = iocb->ki_filp; - ssize_t retval; - unsigned long seg; - size_t count; - loff_t *ppos = &iocb->ki_pos; - - count = 0; - for (seg = 0; seg < nr_segs; seg++) { - const struct iovec *iv = &iov[seg]; - - /* - * If any segment has a negative length, or the cumulative - * length ever wraps negative then return -EINVAL. - */ - count += iv->iov_len; - if (unlikely((ssize_t)(count|iv->iov_len) < 0)) - return -EINVAL; - if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len)) - continue; - if (seg == 0) - return -EFAULT; - nr_segs = seg; - count -= iv->iov_len; /* This segment is no good */ - break; - } - retval = 0; - if (count) { - for (seg = 0; seg < nr_segs; seg++) { - read_descriptor_t desc; - - desc.written = 0; - desc.arg.buf = iov[seg].iov_base; - desc.count = iov[seg].iov_len; - if (desc.count == 0) - continue; - desc.error = 0; - do_generic_file_read(filp, ppos, &desc, - btrfs_read_actor); - retval += desc.written; - if (desc.error) { - retval = retval ?: desc.error; - break; - } - } - } - return retval; -} - static int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) { @@ -789,12 +652,25 @@ out: return ret > 0 ? EIO : ret; } +static struct vm_operations_struct btrfs_file_vm_ops = { + .nopage = filemap_nopage, + .populate = filemap_populate, + .page_mkwrite = btrfs_page_mkwrite, +}; + +static int btrfs_file_mmap(struct file *filp, struct vm_area_struct *vma) +{ + vma->vm_ops = &btrfs_file_vm_ops; + file_accessed(filp); + return 0; +} + struct file_operations btrfs_file_operations = { .llseek = generic_file_llseek, .read = do_sync_read, - .aio_read = btrfs_file_aio_read, + .aio_read = generic_file_aio_read, .write = btrfs_file_write, - .mmap = generic_file_mmap, + .mmap = btrfs_file_mmap, .open = generic_file_open, .ioctl = btrfs_ioctl, .fsync = btrfs_sync_file, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e639cc9e089..def33ac90d7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -31,6 +31,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -561,7 +562,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) goto out; if (!PageUptodate(page)) { - ret = mpage_readpage(page, btrfs_get_block); + ret = btrfs_readpage(NULL, page); lock_page(page); if (!PageUptodate(page)) { ret = -EIO; @@ -1257,8 +1258,10 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock, path = btrfs_alloc_path(); BUG_ON(!path); if (create & BTRFS_GET_BLOCK_CREATE) { - WARN_ON(1); - /* this almost but not quite works */ + /* + * danger!, this only works if the page is properly up + * to date somehow + */ trans = btrfs_start_transaction(root, 1); if (!trans) { err = -ENOMEM; @@ -1353,7 +1356,6 @@ not_found: ins.objectid, ins.offset, ins.offset); BUG_ON(ret); - SetPageChecked(result->b_page); btrfs_map_bh_to_logical(root, result, ins.objectid); } out: @@ -1374,6 +1376,40 @@ int btrfs_get_block(struct inode *inode, sector_t iblock, return err; } +int btrfs_get_block_csum(struct inode *inode, sector_t iblock, + struct buffer_head *result, int create) +{ + int ret; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct page *page = result->b_page; + u64 offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(result); + struct btrfs_csum_item *item; + struct btrfs_path *path = NULL; + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_get_block_lock(inode, iblock, result, create); + if (ret) + goto out; + + path = btrfs_alloc_path(); + item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, offset, 0); + if (IS_ERR(item)) { + ret = PTR_ERR(item); + /* a csum that isn't present is a preallocated region. */ + if (ret == -ENOENT || ret == -EFBIG) + ret = 0; + result->b_private = 0; + goto out; + } + memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE); +printk("get_block_sum file %lu offset %llu csum %X\n", inode->i_ino, (unsigned long long)offset, *(int *)(&item->csum)); +out: + if (path) + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock, struct buffer_head *result, int create) { @@ -1395,9 +1431,198 @@ static int btrfs_prepare_write(struct file *file, struct page *page, return block_prepare_write(page, from, to, btrfs_get_block); } -static int btrfs_readpage(struct file *file, struct page *page) +static void buffer_io_error(struct buffer_head *bh) +{ + char b[BDEVNAME_SIZE]; + + printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n", + bdevname(bh->b_bdev, b), + (unsigned long long)bh->b_blocknr); +} + +/* + * I/O completion handler for block_read_full_page() - pages + * which come unlocked at the end of I/O. + */ +static void btrfs_end_buffer_async_read(struct buffer_head *bh, int uptodate) +{ + unsigned long flags; + struct buffer_head *first; + struct buffer_head *tmp; + struct page *page; + int page_uptodate = 1; + struct inode *inode; + int ret; + + BUG_ON(!buffer_async_read(bh)); + + page = bh->b_page; + inode = page->mapping->host; + if (uptodate) { + void *kaddr; + struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; + if (bh->b_private) { + char csum[BTRFS_CRC32_SIZE]; + kaddr = kmap_atomic(page, KM_IRQ0); + ret = btrfs_csum_data(root, kaddr + bh_offset(bh), + bh->b_size, csum); + BUG_ON(ret); + if (memcmp(csum, &bh->b_private, BTRFS_CRC32_SIZE)) { + u64 offset; + offset = (page->index << PAGE_CACHE_SHIFT) + + bh_offset(bh); + printk("btrfs csum failed ino %lu off %llu\n", + page->mapping->host->i_ino, + (unsigned long long)offset); + memset(kaddr + bh_offset(bh), 1, bh->b_size); + flush_dcache_page(page); +printk("bad verify file %lu offset %llu bh_private %lX csum %X\n", inode->i_ino, (unsigned long long)offset, (unsigned long)(bh->b_private), *(int *)csum); + } + kunmap_atomic(kaddr, KM_IRQ0); + } + set_buffer_uptodate(bh); + } else { + clear_buffer_uptodate(bh); + if (printk_ratelimit()) + buffer_io_error(bh); + SetPageError(page); + } + + /* + * Be _very_ careful from here on. Bad things can happen if + * two buffer heads end IO at almost the same time and both + * decide that the page is now completely done. + */ + first = page_buffers(page); + local_irq_save(flags); + bit_spin_lock(BH_Uptodate_Lock, &first->b_state); + clear_buffer_async_read(bh); + unlock_buffer(bh); + tmp = bh; + do { + if (!buffer_uptodate(tmp)) + page_uptodate = 0; + if (buffer_async_read(tmp)) { + BUG_ON(!buffer_locked(tmp)); + goto still_busy; + } + tmp = tmp->b_this_page; + } while (tmp != bh); + bit_spin_unlock(BH_Uptodate_Lock, &first->b_state); + local_irq_restore(flags); + + /* + * If none of the buffers had errors and they are all + * uptodate then we can set the page uptodate. + */ + if (page_uptodate && !PageError(page)) + SetPageUptodate(page); + unlock_page(page); + return; + +still_busy: + bit_spin_unlock(BH_Uptodate_Lock, &first->b_state); + local_irq_restore(flags); + return; +} + +/* + * Generic "read page" function for block devices that have the normal + * get_block functionality. This is most of the block device filesystems. + * Reads the page asynchronously --- the unlock_buffer() and + * set/clear_buffer_uptodate() functions propagate buffer state into the + * page struct once IO has completed. + */ +int btrfs_readpage(struct file *file, struct page *page) { - return mpage_readpage(page, btrfs_get_block); + struct inode *inode = page->mapping->host; + sector_t iblock, lblock; + struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE]; + unsigned int blocksize; + int nr, i; + int fully_mapped = 1; + + BUG_ON(!PageLocked(page)); + blocksize = 1 << inode->i_blkbits; + if (!page_has_buffers(page)) + create_empty_buffers(page, blocksize, 0); + head = page_buffers(page); + + iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); + lblock = (i_size_read(inode)+blocksize-1) >> inode->i_blkbits; + bh = head; + nr = 0; + i = 0; + + do { + if (buffer_uptodate(bh)) + continue; + + if (!buffer_mapped(bh)) { + int err = 0; + + fully_mapped = 0; + if (iblock < lblock) { + WARN_ON(bh->b_size != blocksize); + err = btrfs_get_block_csum(inode, iblock, + bh, 0); + if (err) + SetPageError(page); + } + if (!buffer_mapped(bh)) { + void *kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr + i * blocksize, 0, blocksize); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + if (!err) + set_buffer_uptodate(bh); + continue; + } + /* + * get_block() might have updated the buffer + * synchronously + */ + if (buffer_uptodate(bh)) + continue; + } + arr[nr++] = bh; + } while (i++, iblock++, (bh = bh->b_this_page) != head); + + if (fully_mapped) + SetPageMappedToDisk(page); + + if (!nr) { + /* + * All buffers are uptodate - we can set the page uptodate + * as well. But not if get_block() returned an error. + */ + if (!PageError(page)) + SetPageUptodate(page); + unlock_page(page); + return 0; + } + + /* Stage two: lock the buffers */ + for (i = 0; i < nr; i++) { + bh = arr[i]; + lock_buffer(bh); + bh->b_end_io = btrfs_end_buffer_async_read; + set_buffer_async_read(bh); + } + + /* + * Stage 3: start the IO. Check for uptodateness + * inside the buffer lock in case another process reading + * the underlying blockdev brought it uptodate (the sct fix). + */ + for (i = 0; i < nr; i++) { + bh = arr[i]; + if (buffer_uptodate(bh)) + btrfs_end_buffer_async_read(bh, 1); + else + submit_bh(READ, bh); + } + return 0; } /* @@ -1424,6 +1649,7 @@ static int __btrfs_write_full_page(struct inode *inode, struct page *page, struct buffer_head *bh, *head; const unsigned blocksize = 1 << inode->i_blkbits; int nr_underway = 0; + struct btrfs_root *root = BTRFS_I(inode)->root; BUG_ON(!PageLocked(page)); @@ -1496,6 +1722,24 @@ static int __btrfs_write_full_page(struct inode *inode, struct page *page, continue; } if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) { + struct btrfs_trans_handle *trans; + int ret; + u64 off = page->index << PAGE_CACHE_SHIFT; + char *kaddr; + + off += bh_offset(bh); + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + kaddr = kmap(page); + ret = btrfs_csum_file_block(trans, root, inode->i_ino, + off, kaddr + bh_offset(bh), + bh->b_size); + kunmap(page); + BUG_ON(ret); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); mark_buffer_async_write(bh); } else { unlock_buffer(bh); @@ -1617,6 +1861,53 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) return __btrfs_write_full_page(inode, page, wbc); } +/* + * btrfs_page_mkwrite() is not allowed to change the file size as it gets + * called from a page fault handler when a page is first dirtied. Hence we must + * be careful to check for EOF conditions here. We set the page up correctly + * for a written page which means we get ENOSPC checking when writing into + * holes and correct delalloc and unwritten extent mapping on filesystems that + * support these features. + * + * We are not allowed to take the i_mutex here so we have to play games to + * protect against truncate races as the page could now be beyond EOF. Because + * vmtruncate() writes the inode size before removing pages, once we have the + * page lock we can determine safely if the page is beyond EOF. If it is not + * beyond EOF, then the page is guaranteed safe against truncation until we + * unlock the page. + */ +int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) +{ + struct inode *inode = vma->vm_file->f_path.dentry->d_inode; + unsigned long end; + loff_t size; + int ret = -EINVAL; + + lock_page(page); + wait_on_page_writeback(page); +printk("btrfs_page_mkwrite %lu %lu\n", page->mapping->host->i_ino, page->index); + size = i_size_read(inode); + if ((page->mapping != inode->i_mapping) || + ((page->index << PAGE_CACHE_SHIFT) > size)) { + /* page got truncated out from underneath us */ + goto out_unlock; + } + + /* page is wholly or partially inside EOF */ + if (((page->index + 1) << PAGE_CACHE_SHIFT) > size) + end = size & ~PAGE_CACHE_MASK; + else + end = PAGE_CACHE_SIZE; + + ret = btrfs_prepare_write(NULL, page, 0, end); + if (!ret) + ret = btrfs_commit_write(NULL, page, 0, end); + +out_unlock: + unlock_page(page); + return ret; +} + static void btrfs_truncate(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; -- cgit v1.2.3-70-g09d2 From 8c2383c3dd2cb5bb39598ce4fa97154bc591020a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 18 Jun 2007 09:57:58 -0400 Subject: Subject: Rework btrfs_file_write to only allocate while page locks are held Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 2 + fs/btrfs/extent-tree.c | 1 - fs/btrfs/file-item.c | 1 - fs/btrfs/file.c | 244 +++++++++++++++++++++++++++---------------------- fs/btrfs/inode.c | 11 ++- fs/btrfs/print-tree.c | 18 ++++ fs/btrfs/transaction.c | 1 + 7 files changed, 161 insertions(+), 117 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 602b63dc76c..15cc9ec9230 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -541,6 +541,8 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root else ret = submit_bh(WRITE, bh); if (ret == -EOPNOTSUPP) { + get_bh(bh); + lock_buffer(bh); set_buffer_uptodate(bh); root->fs_info->do_barriers = 0; ret = submit_bh(WRITE, bh); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 5f4eaba8b85..e0738c65bc5 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1477,7 +1477,6 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root break; if (wret < 0) ret = wret; - btrfs_btree_balance_dirty(root); } for (i = 0; i <= orig_level; i++) { if (path->nodes[i]) { diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 1068993ab1c..cf894f09f6c 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -228,7 +228,6 @@ found: path->nodes[0]->b_data, root->fs_info->sb->s_blocksize); ret = btrfs_csum_data(root, data, len, &item->csum); -// printk("file %lu offset %llu csum %X\n", objectid, (unsigned long long)offset, *(int *)(&item->csum)); btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_release_path(root, path); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index de8d47b44e1..6b455c2b3f0 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -207,6 +207,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, } path->slots[0]--; } +next_slot: keep = 0; bookend = 0; found_extent = 0; @@ -214,39 +215,48 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, extent = NULL; leaf = btrfs_buffer_leaf(path->nodes[0]); slot = path->slots[0]; + ret = 0; btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key); if (key.offset >= end || key.objectid != inode->i_ino) { - ret = 0; goto out; } - if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) { - ret = 0; + if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) { goto out; } - extent = btrfs_item_ptr(leaf, slot, - struct btrfs_file_extent_item); - found_type = btrfs_file_extent_type(extent); - if (found_type == BTRFS_FILE_EXTENT_REG) { - extent_end = key.offset + - (btrfs_file_extent_num_blocks(extent) << - inode->i_blkbits); - found_extent = 1; - } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - found_inline = 1; - extent_end = key.offset + - btrfs_file_extent_inline_len(leaf->items + slot); + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { + extent = btrfs_item_ptr(leaf, slot, + struct btrfs_file_extent_item); + found_type = btrfs_file_extent_type(extent); + if (found_type == BTRFS_FILE_EXTENT_REG) { + extent_end = key.offset + + (btrfs_file_extent_num_blocks(extent) << + inode->i_blkbits); + found_extent = 1; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + found_inline = 1; + extent_end = key.offset + + btrfs_file_extent_inline_len(leaf->items + + slot); + } + } else { + extent_end = search_start; } /* we found nothing we can drop */ - if (!found_extent && !found_inline) { - ret = 0; - goto out; - } - - /* we found nothing inside the range */ - if (search_start >= extent_end) { - ret = 0; - goto out; + if ((!found_extent && !found_inline) || + search_start >= extent_end) { + int nextret; + u32 nritems; + nritems = btrfs_header_nritems( + btrfs_buffer_header(path->nodes[0])); + if (slot >= nritems - 1) { + nextret = btrfs_next_leaf(root, path); + if (nextret) + goto out; + } else { + path->slots[0]++; + } + goto next_slot; } /* FIXME, there's only one inline extent allowed right now */ @@ -272,7 +282,6 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, WARN_ON(found_inline); bookend = 1; } - /* truncate existing extent */ if (start > key.offset) { u64 new_num; @@ -337,10 +346,14 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, ins.offset = end; ins.flags = 0; btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); - btrfs_release_path(root, path); ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*extent)); + + if (ret) { + btrfs_print_leaf(root, btrfs_buffer_leaf(path->nodes[0])); + printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end); + } BUG_ON(ret); extent = btrfs_item_ptr( btrfs_buffer_leaf(path->nodes[0]), @@ -387,8 +400,7 @@ static int prepare_pages(struct btrfs_root *root, loff_t pos, unsigned long first_index, unsigned long last_index, - size_t write_bytes, - u64 alloc_extent_start) + size_t write_bytes) { int i; unsigned long index = pos >> PAGE_CACHE_SHIFT; @@ -399,6 +411,16 @@ static int prepare_pages(struct btrfs_root *root, struct buffer_head *bh; struct buffer_head *head; loff_t isize = i_size_read(inode); + struct btrfs_trans_handle *trans; + u64 hint_block; + u64 num_blocks; + u64 alloc_extent_start; + u64 start_pos; + struct btrfs_key ins; + + start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1); + num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> + inode->i_blkbits; memset(pages, 0, num_pages * sizeof(struct page *)); @@ -408,6 +430,72 @@ static int prepare_pages(struct btrfs_root *root, err = -ENOMEM; goto failed_release; } + } + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + if (!trans) { + err = -ENOMEM; + mutex_unlock(&root->fs_info->fs_mutex); + goto out_unlock; + } + btrfs_set_trans_block_group(trans, inode); + /* FIXME blocksize != 4096 */ + inode->i_blocks += num_blocks << 3; + hint_block = 0; + + /* FIXME...EIEIO, ENOSPC and more */ + + /* step one, delete the existing extents in this range */ + /* FIXME blocksize != pagesize */ + if (start_pos < inode->i_size) { + err = btrfs_drop_extents(trans, root, inode, + start_pos, (pos + write_bytes + root->blocksize -1) & + ~((u64)root->blocksize - 1), &hint_block); + BUG_ON(err); + } + + /* insert any holes we need to create */ + if (inode->i_size < start_pos) { + u64 last_pos_in_file; + u64 hole_size; + u64 mask = root->blocksize - 1; + last_pos_in_file = (isize + mask) & ~mask; + hole_size = (start_pos - last_pos_in_file + mask) & ~mask; + hole_size >>= inode->i_blkbits; + if (last_pos_in_file < start_pos) { + err = btrfs_insert_file_extent(trans, root, + inode->i_ino, + last_pos_in_file, + 0, 0, hole_size); + } + BUG_ON(err); + } + + /* + * either allocate an extent for the new bytes or setup the key + * to show we are doing inline data in the extent + */ + if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || + pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { + err = btrfs_alloc_extent(trans, root, inode->i_ino, + num_blocks, hint_block, (u64)-1, + &ins, 1); + BUG_ON(err); + err = btrfs_insert_file_extent(trans, root, inode->i_ino, + start_pos, ins.objectid, ins.offset, + ins.offset); + BUG_ON(err); + } else { + ins.offset = 0; + ins.objectid = 0; + } + BUG_ON(err); + alloc_extent_start = ins.objectid; + err = btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + for (i = 0; i < num_pages; i++) { cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); wait_on_page_writeback(pages[i]); offset = pos & (PAGE_CACHE_SIZE -1); @@ -444,6 +532,11 @@ failed_truncate: if (pos > isize) vmtruncate(inode, isize); return err; + +out_unlock: + mutex_unlock(&root->fs_info->fs_mutex); + goto failed_release; + } static ssize_t btrfs_file_write(struct file *file, const char __user *buf, @@ -455,16 +548,14 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, int ret = 0; struct inode *inode = file->f_path.dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; - struct page *pages[8]; + struct page **pages = NULL; + int nrptrs; struct page *pinned[2]; unsigned long first_index; unsigned long last_index; - u64 start_pos; - u64 num_blocks; - u64 alloc_extent_start; - u64 hint_block; - struct btrfs_trans_handle *trans; - struct btrfs_key ins; + + nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, + PAGE_CACHE_SIZE / (sizeof(struct page *))); pinned[0] = NULL; pinned[1] = NULL; if (file->f_flags & O_DIRECT) @@ -482,9 +573,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, goto out; file_update_time(file); - start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1); - num_blocks = (count + pos - start_pos + root->blocksize - 1) >> - inode->i_blkbits; + pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); mutex_lock(&inode->i_mutex); first_index = pos >> PAGE_CACHE_SHIFT; @@ -516,87 +605,20 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, } } - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - if (!trans) { - err = -ENOMEM; - mutex_unlock(&root->fs_info->fs_mutex); - goto out_unlock; - } - btrfs_set_trans_block_group(trans, inode); - /* FIXME blocksize != 4096 */ - inode->i_blocks += num_blocks << 3; - hint_block = 0; - - /* FIXME...EIEIO, ENOSPC and more */ - - /* step one, delete the existing extents in this range */ - if (start_pos < inode->i_size) { - /* FIXME blocksize != pagesize */ - ret = btrfs_drop_extents(trans, root, inode, - start_pos, - (pos + count + root->blocksize -1) & - ~((u64)root->blocksize - 1), - &hint_block); - BUG_ON(ret); - } - - /* insert any holes we need to create */ - if (inode->i_size < start_pos) { - u64 last_pos_in_file; - u64 hole_size; - u64 mask = root->blocksize - 1; - last_pos_in_file = (inode->i_size + mask) & ~mask; - hole_size = (start_pos - last_pos_in_file + mask) & ~mask; - hole_size >>= inode->i_blkbits; - if (last_pos_in_file < start_pos) { - ret = btrfs_insert_file_extent(trans, root, - inode->i_ino, - last_pos_in_file, - 0, 0, hole_size); - } - BUG_ON(ret); - } - - /* - * either allocate an extent for the new bytes or setup the key - * to show we are doing inline data in the extent - */ - if (inode->i_size >= PAGE_CACHE_SIZE || pos + count < inode->i_size || - pos + count - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { - ret = btrfs_alloc_extent(trans, root, inode->i_ino, - num_blocks, hint_block, (u64)-1, - &ins, 1); - BUG_ON(ret); - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - start_pos, ins.objectid, ins.offset, - ins.offset); - BUG_ON(ret); - } else { - ins.offset = 0; - ins.objectid = 0; - } - BUG_ON(ret); - alloc_extent_start = ins.objectid; - ret = btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); - while(count > 0) { size_t offset = pos & (PAGE_CACHE_SIZE - 1); - size_t write_bytes = min(count, - (size_t)PAGE_CACHE_SIZE - offset); + size_t write_bytes = min(count, nrptrs * PAGE_CACHE_SIZE - + offset); size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + WARN_ON(num_pages > nrptrs); memset(pages, 0, sizeof(pages)); ret = prepare_pages(root, file, pages, num_pages, pos, first_index, last_index, - write_bytes, alloc_extent_start); + write_bytes); BUG_ON(ret); - /* FIXME blocks != pagesize */ - if (alloc_extent_start) - alloc_extent_start += num_pages; ret = btrfs_copy_from_user(pos, num_pages, write_bytes, pages, buf); BUG_ON(ret); @@ -611,13 +633,13 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, pos += write_bytes; num_written += write_bytes; - balance_dirty_pages_ratelimited(inode->i_mapping); + balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); btrfs_btree_balance_dirty(root); cond_resched(); } -out_unlock: mutex_unlock(&inode->i_mutex); out: + kfree(pages); if (pinned[0]) page_cache_release(pinned[0]); if (pinned[1]) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index def33ac90d7..94f1c28c25b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -962,7 +962,6 @@ void btrfs_dirty_inode(struct inode *inode) btrfs_update_inode(trans, root, inode); btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); } static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, @@ -1402,7 +1401,6 @@ int btrfs_get_block_csum(struct inode *inode, sector_t iblock, goto out; } memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE); -printk("get_block_sum file %lu offset %llu csum %X\n", inode->i_ino, (unsigned long long)offset, *(int *)(&item->csum)); out: if (path) btrfs_free_path(path); @@ -1476,7 +1474,6 @@ static void btrfs_end_buffer_async_read(struct buffer_head *bh, int uptodate) (unsigned long long)offset); memset(kaddr + bh_offset(bh), 1, bh->b_size); flush_dcache_page(page); -printk("bad verify file %lu offset %llu bh_private %lX csum %X\n", inode->i_ino, (unsigned long long)offset, (unsigned long)(bh->b_private), *(int *)csum); } kunmap_atomic(kaddr, KM_IRQ0); } @@ -1655,6 +1652,13 @@ static int __btrfs_write_full_page(struct inode *inode, struct page *page, last_block = (i_size_read(inode) - 1) >> inode->i_blkbits; + /* no csumming allowed when from PF_MEMALLOC */ + if (current->flags & PF_MEMALLOC) { + redirty_page_for_writepage(wbc, page); + unlock_page(page); + return 0; + } + if (!page_has_buffers(page)) { create_empty_buffers(page, blocksize, (1 << BH_Dirty)|(1 << BH_Uptodate)); @@ -1885,7 +1889,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) lock_page(page); wait_on_page_writeback(page); -printk("btrfs_page_mkwrite %lu %lu\n", page->mapping->host->i_ino, page->index); size = i_size_read(inode); if ((page->mapping != inode->i_mapping) || ((page->index << PAGE_CACHE_SHIFT) > size)) { diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 1fa70568300..e66acd4e274 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -31,6 +31,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l) struct btrfs_dir_item *di; struct btrfs_inode_item *ii; struct btrfs_block_group_item *bi; + struct btrfs_file_extent_item *fi; u32 type; printk("leaf %llu total ptrs %d free space %d\n", @@ -75,6 +76,23 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l) printk("\t\textent data refs %u\n", btrfs_extent_refs(ei)); break; + + case BTRFS_EXTENT_DATA_KEY: + fi = btrfs_item_ptr(l, i, + struct btrfs_file_extent_item); + if (btrfs_file_extent_type(fi) == + BTRFS_FILE_EXTENT_INLINE) { + printk("\t\tinline extent data size %u\n", + btrfs_file_extent_inline_len(l->items + i)); + break; + } + printk("\t\textent data disk block %llu nr %llu\n", + (unsigned long long)btrfs_file_extent_disk_blocknr(fi), + (unsigned long long)btrfs_file_extent_disk_num_blocks(fi)); + printk("\t\textent data offset %llu nr %llu\n", + (unsigned long long)btrfs_file_extent_offset(fi), + (unsigned long long)btrfs_file_extent_num_blocks(fi)); + break; case BTRFS_BLOCK_GROUP_ITEM_KEY: bi = btrfs_item_ptr(l, i, struct btrfs_block_group_item); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index cd4f1d14038..b5b99a85763 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -279,6 +279,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, BUG_ON(ret); kfree(dirty); mutex_unlock(&tree_root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(tree_root); } return 0; } -- cgit v1.2.3-70-g09d2 From 54aa1f4dfdacd60a19c4471220b24e581be6f774 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 22 Jun 2007 14:16:25 -0400 Subject: Btrfs: Audit callers and return codes to make sure -ENOSPC gets up the stack Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 106 ++++++++++++++++++++++++------- fs/btrfs/ctree.h | 3 - fs/btrfs/dir-item.c | 6 +- fs/btrfs/disk-io.c | 2 +- fs/btrfs/extent-tree.c | 141 ++++++++++++++++++++++++++++++++++-------- fs/btrfs/file-item.c | 46 ++------------ fs/btrfs/file.c | 131 ++++++++++++++++++++++++--------------- fs/btrfs/inode.c | 165 ++++++++++++++++++++++++++++++++----------------- fs/btrfs/root-tree.c | 1 - fs/btrfs/super.c | 3 +- fs/btrfs/transaction.c | 21 ++++--- 11 files changed, 410 insertions(+), 215 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 498dea0854c..606a19b5916 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -73,6 +73,7 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root { struct buffer_head *cow; struct btrfs_node *cow_node; + int ret; if (btrfs_header_generation(btrfs_buffer_header(buf)) == trans->transid) { @@ -80,6 +81,8 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root return 0; } cow = btrfs_alloc_free_block(trans, root, buf->b_blocknr); + if (IS_ERR(cow)) + return PTR_ERR(cow); cow_node = btrfs_buffer_node(cow); if (buf->b_size != root->blocksize || cow->b_size != root->blocksize) WARN_ON(1); @@ -87,7 +90,9 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_header_blocknr(&cow_node->header, bh_blocknr(cow)); btrfs_set_header_generation(&cow_node->header, trans->transid); btrfs_set_header_owner(&cow_node->header, root->root_key.objectid); - btrfs_inc_ref(trans, root, buf); + ret = btrfs_inc_ref(trans, root, buf); + if (ret) + return ret; if (buf == root->node) { root->node = cow; get_bh(cow); @@ -320,6 +325,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root int wret; int pslot; int orig_slot = path->slots[level]; + int err_on_enospc = 0; u64 orig_ptr; if (level == 0) @@ -363,29 +369,43 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root BTRFS_NODEPTRS_PER_BLOCK(root) / 4) return 0; + if (btrfs_header_nritems(&mid->header) < 2) + err_on_enospc = 1; + left_buf = read_node_slot(root, parent_buf, pslot - 1); right_buf = read_node_slot(root, parent_buf, pslot + 1); /* first, try to make some room in the middle buffer */ if (left_buf) { - btrfs_cow_block(trans, root, left_buf, parent_buf, pslot - 1, - &left_buf); + wret = btrfs_cow_block(trans, root, left_buf, + parent_buf, pslot - 1, &left_buf); + if (wret) { + ret = wret; + goto enospc; + } left = btrfs_buffer_node(left_buf); orig_slot += btrfs_header_nritems(&left->header); wret = push_node_left(trans, root, left_buf, mid_buf); if (wret < 0) ret = wret; + if (btrfs_header_nritems(&mid->header) < 2) + err_on_enospc = 1; } /* * then try to empty the right most buffer into the middle */ if (right_buf) { - btrfs_cow_block(trans, root, right_buf, parent_buf, pslot + 1, - &right_buf); + wret = btrfs_cow_block(trans, root, right_buf, + parent_buf, pslot + 1, &right_buf); + if (wret) { + ret = wret; + goto enospc; + } + right = btrfs_buffer_node(right_buf); wret = push_node_left(trans, root, mid_buf, right_buf); - if (wret < 0) + if (wret < 0 && wret != -ENOSPC) ret = wret; if (btrfs_header_nritems(&right->header) == 0) { u64 blocknr = bh_blocknr(right_buf); @@ -421,8 +441,10 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root */ BUG_ON(!left_buf); wret = balance_node_right(trans, root, mid_buf, left_buf); - if (wret < 0) + if (wret < 0) { ret = wret; + goto enospc; + } BUG_ON(wret == 1); } if (btrfs_header_nritems(&mid->header) == 0) { @@ -467,7 +489,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_node_blockptr(btrfs_buffer_node(path->nodes[level]), path->slots[level])) BUG(); - +enospc: if (right_buf) btrfs_block_release(root, right_buf); if (left_buf) @@ -519,10 +541,15 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; } else { - btrfs_cow_block(trans, root, left_buf, parent_buf, - pslot - 1, &left_buf); - left = btrfs_buffer_node(left_buf); - wret = push_node_left(trans, root, left_buf, mid_buf); + ret = btrfs_cow_block(trans, root, left_buf, parent_buf, + pslot - 1, &left_buf); + if (ret) + wret = 1; + else { + left = btrfs_buffer_node(left_buf); + wret = push_node_left(trans, root, + left_buf, mid_buf); + } } if (wret < 0) ret = wret; @@ -561,11 +588,16 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; } else { - btrfs_cow_block(trans, root, right_buf, - parent_buf, pslot + 1, &right_buf); - right = btrfs_buffer_node(right_buf); - wret = balance_node_right(trans, root, - right_buf, mid_buf); + ret = btrfs_cow_block(trans, root, right_buf, + parent_buf, pslot + 1, + &right_buf); + if (ret) + wret = 1; + else { + right = btrfs_buffer_node(right_buf); + wret = balance_node_right(trans, root, + right_buf, mid_buf); + } } if (wret < 0) ret = wret; @@ -631,6 +663,10 @@ again: p->nodes[level + 1], p->slots[level + 1], &cow_buf); + if (wret) { + btrfs_block_release(root, cow_buf); + return wret; + } b = cow_buf; c = btrfs_buffer_node(b); } @@ -737,6 +773,7 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root src_nritems = btrfs_header_nritems(&src->header); dst_nritems = btrfs_header_nritems(&dst->header); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; + if (push_items <= 0) { return 1; } @@ -827,6 +864,8 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(path->nodes[level-1] != root->node); t = btrfs_alloc_free_block(trans, root, root->node->b_blocknr); + if (IS_ERR(t)) + return PTR_ERR(t); c = btrfs_buffer_node(t); memset(c, 0, root->blocksize); btrfs_set_header_nritems(&c->header, 1); @@ -929,10 +968,15 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_header_nritems(&c->header) < BTRFS_NODEPTRS_PER_BLOCK(root) - 1) return 0; + if (ret < 0) + return ret; } c_nritems = btrfs_header_nritems(&c->header); split_buffer = btrfs_alloc_free_block(trans, root, t->b_blocknr); + if (IS_ERR(split_buffer)) + return PTR_ERR(split_buffer); + split = btrfs_buffer_node(split_buffer); btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header)); btrfs_set_header_level(&split->header, btrfs_header_level(&c->header)); @@ -1022,6 +1066,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_item *item; u32 left_nritems; u32 right_nritems; + int ret; slot = path->slots[1]; if (!path->nodes[1]) { @@ -1041,7 +1086,12 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root return 1; } /* cow and double check */ - btrfs_cow_block(trans, root, right_buf, upper, slot + 1, &right_buf); + ret = btrfs_cow_block(trans, root, right_buf, upper, + slot + 1, &right_buf); + if (ret) { + btrfs_block_release(root, right_buf); + return 1; + } right = btrfs_buffer_leaf(right_buf); free_space = btrfs_leaf_free_space(root, right); if (free_space < data_size + sizeof(struct btrfs_item)) { @@ -1162,7 +1212,11 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root } /* cow and double check */ - btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t); + ret = btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t); + if (ret) { + /* we hit -ENOSPC, but it isn't fatal here */ + return 1; + } left = btrfs_buffer_leaf(t); free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { @@ -1309,8 +1363,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root slot = path->slots[0]; nritems = btrfs_header_nritems(&l->header); mid = (nritems + 1)/ 2; + right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr); - BUG_ON(!right_buffer); + if (IS_ERR(right_buffer)) + return PTR_ERR(right_buffer); + right = btrfs_buffer_leaf(right_buffer); memset(&right->header, 0, sizeof(right->header)); btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer)); @@ -1407,7 +1464,9 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root if (!double_split) return ret; right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr); - BUG_ON(!right_buffer); + if (IS_ERR(right_buffer)) + return PTR_ERR(right_buffer); + right = btrfs_buffer_leaf(right_buffer); memset(&right->header, 0, sizeof(right->header)); btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer)); @@ -1655,7 +1714,6 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root ptr, data, data_size); btrfs_mark_buffer_dirty(path->nodes[0]); } - btrfs_release_path(root, path); btrfs_free_path(path); return ret; } @@ -1775,12 +1833,12 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, slot = path->slots[1]; get_bh(leaf_buf); wret = push_leaf_left(trans, root, path, 1); - if (wret < 0) + if (wret < 0 && wret != -ENOSPC) ret = wret; if (path->nodes[0] == leaf_buf && btrfs_header_nritems(&leaf->header)) { wret = push_leaf_right(trans, root, path, 1); - if (wret < 0) + if (wret < 0 && wret != -ENOSPC) ret = wret; } if (btrfs_header_nritems(&leaf->header) == 0) { diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 19a1d998fca..77071f27397 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1126,9 +1126,6 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 offset, char *data, size_t len); -int btrfs_csum_verify_file_block(struct btrfs_root *root, - u64 objectid, u64 offset, - char *data, size_t len); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index ff10cf5a842..d5db8bdea3d 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -47,6 +47,8 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle if (ret) return ERR_PTR(ret); } + if (ret < 0) + return ERR_PTR(ret); WARN_ON(ret > 0); leaf = btrfs_buffer_leaf(path->nodes[0]); item = leaf->items + path->slots[0]; @@ -225,14 +227,13 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, struct btrfs_leaf *leaf; u32 sub_item_len; u32 item_len; - int ret; + int ret = 0; leaf = btrfs_buffer_leaf(path->nodes[0]); sub_item_len = sizeof(*di) + btrfs_dir_name_len(di); item_len = btrfs_item_size(leaf->items + path->slots[0]); if (sub_item_len == btrfs_item_size(leaf->items + path->slots[0])) { ret = btrfs_del_item(trans, root, path); - BUG_ON(ret); } else { char *ptr = (char *)di; char *start = btrfs_item_ptr(leaf, path->slots[0], char); @@ -240,7 +241,6 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, item_len - (ptr + sub_item_len - start)); ret = btrfs_truncate_item(trans, root, path, item_len - sub_item_len); - BUG_ON(ret); } return 0; } diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3712e946f9c..865a284aa06 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -580,7 +580,7 @@ int close_ctree(struct btrfs_root *root) btrfs_transaction_flush_work(root); mutex_lock(&fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); - btrfs_commit_transaction(trans, root); + ret = btrfs_commit_transaction(trans, root); /* run commit again to drop the original snapshot */ trans = btrfs_start_transaction(root, 1); btrfs_commit_transaction(trans, root); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index b38c3e92f0c..8025e9f8ef1 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -100,6 +100,8 @@ static int cache_block_group(struct btrfs_root *root, if (slot >= btrfs_header_nritems(&leaf->header)) { reada_extent_leaves(root, path, limit); ret = btrfs_next_leaf(root, path); + if (ret < 0) + goto err; if (ret == 0) { continue; } else { @@ -148,6 +150,7 @@ static int cache_block_group(struct btrfs_root *root, } block_group->cached = 1; +err: btrfs_free_path(path); return 0; } @@ -201,7 +204,9 @@ static u64 find_search_start(struct btrfs_root *root, last = max(last, cache->last_prealloc); } again: - cache_block_group(root, cache); + ret = cache_block_group(root, cache); + if (ret) + goto out; while(1) { ret = find_first_radix_bit(&root->fs_info->extent_map_radix, gang, last, ARRAY_SIZE(gang)); @@ -398,16 +403,23 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_key ins; u32 refs; - find_free_extent(trans, root->fs_info->extent_root, 0, 0, (u64)-1, 0, - &ins, 0); path = btrfs_alloc_path(); - BUG_ON(!path); + if (!path) + return -ENOMEM; + ret = find_free_extent(trans, root->fs_info->extent_root, 0, 0, + (u64)-1, 0, &ins, 0); + if (ret) { + btrfs_free_path(path); + return ret; + } key.objectid = blocknr; key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); key.offset = num_blocks; ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 1); + if (ret < 0) + return ret; if (ret != 0) { BUG(); } @@ -442,12 +454,14 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans, btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 0); + if (ret < 0) + goto out; if (ret != 0) BUG(); l = btrfs_buffer_leaf(path->nodes[0]); item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); *refs = btrfs_extent_refs(item); - btrfs_release_path(root->fs_info->extent_root, path); +out: btrfs_free_path(path); return 0; } @@ -469,6 +483,8 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, int i; int leaf; int ret; + int faili; + int err; if (!root->ref_cows) return 0; @@ -491,14 +507,45 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, continue; ret = btrfs_inc_extent_ref(trans, root, disk_blocknr, btrfs_file_extent_disk_num_blocks(fi)); - BUG_ON(ret); + if (ret) { + faili = i; + goto fail; + } } else { blocknr = btrfs_node_blockptr(buf_node, i); ret = btrfs_inc_extent_ref(trans, root, blocknr, 1); - BUG_ON(ret); + if (ret) { + faili = i; + goto fail; + } } } return 0; +fail: + for (i =0; i < faili; i++) { + if (leaf) { + u64 disk_blocknr; + key = &buf_leaf->items[i].key; + if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY) + continue; + fi = btrfs_item_ptr(buf_leaf, i, + struct btrfs_file_extent_item); + if (btrfs_file_extent_type(fi) == + BTRFS_FILE_EXTENT_INLINE) + continue; + disk_blocknr = btrfs_file_extent_disk_blocknr(fi); + if (disk_blocknr == 0) + continue; + err = btrfs_free_extent(trans, root, disk_blocknr, + btrfs_file_extent_disk_num_blocks(fi), 0); + BUG_ON(err); + } else { + blocknr = btrfs_node_blockptr(buf_node, i); + err = btrfs_free_extent(trans, root, blocknr, 1, 0); + BUG_ON(err); + } + } + return ret; } static int write_one_cache_group(struct btrfs_trans_handle *trans, @@ -512,15 +559,20 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, struct btrfs_block_group_item *bi; struct btrfs_key ins; - find_free_extent(trans, extent_root, 0, 0, (u64)-1, 0, &ins, 0); + ret = find_free_extent(trans, extent_root, 0, 0, (u64)-1, 0, &ins, 0); + /* FIXME, set bit to recalc cache groups on next mount */ + if (ret) + return ret; ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); + if (ret < 0) + goto fail; BUG_ON(ret); bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], struct btrfs_block_group_item); memcpy(bi, &cache->item, sizeof(*bi)); mark_buffer_dirty(path->nodes[0]); btrfs_release_path(extent_root, path); - +fail: finish_current_insert(trans, extent_root); pending_ret = del_pending_extents(trans, extent_root); if (ret) @@ -543,6 +595,7 @@ static int write_dirty_block_radix(struct btrfs_trans_handle *trans, int werr = 0; int i; struct btrfs_path *path; + unsigned long off = 0; path = btrfs_alloc_path(); if (!path) @@ -550,18 +603,28 @@ static int write_dirty_block_radix(struct btrfs_trans_handle *trans, while(1) { ret = radix_tree_gang_lookup_tag(radix, (void **)cache, - 0, ARRAY_SIZE(cache), + off, ARRAY_SIZE(cache), BTRFS_BLOCK_GROUP_DIRTY); if (!ret) break; for (i = 0; i < ret; i++) { - radix_tree_tag_clear(radix, cache[i]->key.objectid + - cache[i]->key.offset - 1, - BTRFS_BLOCK_GROUP_DIRTY); err = write_one_cache_group(trans, root, path, cache[i]); - if (err) + /* + * if we fail to write the cache group, we want + * to keep it marked dirty in hopes that a later + * write will work + */ + if (err) { werr = err; + off = cache[i]->key.objectid + + cache[i]->key.offset; + continue; + } + + radix_tree_tag_clear(radix, cache[i]->key.objectid + + cache[i]->key.offset - 1, + BTRFS_BLOCK_GROUP_DIRTY); } } btrfs_free_path(path); @@ -801,14 +864,20 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); key.offset = num_blocks; - find_free_extent(trans, root, 0, 0, (u64)-1, 0, &ins, 0); path = btrfs_alloc_path(); - BUG_ON(!path); + if (!path) + return -ENOMEM; - ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); + ret = find_free_extent(trans, root, 0, 0, (u64)-1, 0, &ins, 0); if (ret) { - BUG(); + btrfs_free_path(path); + return ret; } + + ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); + if (ret < 0) + return ret; + BUG_ON(ret); ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], struct btrfs_extent_item); BUG_ON(ei->refs == 0); @@ -827,8 +896,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_super_blocks_used(info->disk_super, super_blocks_used - num_blocks); ret = btrfs_del_item(trans, extent_root, path); - if (ret) - BUG(); + if (ret) { + return ret; + } ret = update_block_group(trans, root, blocknr, num_blocks, 0, mark_free, 0); BUG_ON(ret); @@ -1075,7 +1145,6 @@ next: path->slots[0]++; cond_resched(); } - // FIXME -ENOSPC check_pending: /* we have to make sure we didn't find an extent that has already * been allocated by the map tree or the original allocation @@ -1246,7 +1315,15 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, ret = find_free_extent(trans, root, num_blocks, search_start, search_end, hint_block, ins, data); if (ret) { - return ret; + if (search_start == 0) + return ret; + search_end = search_start - 1; + search_start = 0; + hint_block = search_start; + ret = find_free_extent(trans, root, num_blocks, search_start, + search_end, hint_block, ins, data); + if (ret) + return ret; } /* @@ -1271,7 +1348,16 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, search_end, hint_block, &prealloc_key, 0); if (ret) { - return ret; + if (search_start == 0) + return ret; + search_end = search_start - 1; + search_start = 0; + hint_block = search_start; + ret = find_free_extent(trans, root, 0, search_start, + search_end, hint_block, + &prealloc_key, 0); + if (ret) + return ret; } } @@ -1309,11 +1395,14 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, ret = btrfs_alloc_extent(trans, root, root->root_key.objectid, 1, hint, (unsigned long)-1, &ins, 0); if (ret) { - BUG(); - return NULL; + BUG_ON(ret > 0); + return ERR_PTR(ret); } - BUG_ON(ret); buf = btrfs_find_create_tree_block(root, ins.objectid); + if (!buf) { + btrfs_free_extent(trans, root, ins.objectid, 1, 0); + return ERR_PTR(-ENOMEM); + } set_buffer_uptodate(buf); set_buffer_checked(buf); set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index); diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index cf894f09f6c..68859934ae2 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -45,6 +45,8 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, ret = btrfs_insert_empty_item(trans, root, path, &file_key, sizeof(*item)); + if (ret < 0) + goto out; BUG_ON(ret); item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], struct btrfs_file_extent_item); @@ -55,10 +57,9 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, btrfs_set_file_extent_generation(item, trans->transid); btrfs_set_file_extent_type(item, BTRFS_FILE_EXTENT_REG); btrfs_mark_buffer_dirty(path->nodes[0]); - - btrfs_release_path(root, path); +out: btrfs_free_path(path); - return 0; + return ret; } struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, @@ -213,6 +214,8 @@ insert: csum_offset = 0; ret = btrfs_insert_empty_item(trans, root, path, &file_key, BTRFS_CRC32_SIZE); + if (ret < 0) + goto fail; if (ret != 0) { WARN_ON(1); goto fail; @@ -261,40 +264,3 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, return ret; } -int btrfs_csum_verify_file_block(struct btrfs_root *root, - u64 objectid, u64 offset, - char *data, size_t len) -{ - int ret; - struct btrfs_key file_key; - struct btrfs_path *path; - struct btrfs_csum_item *item; - char result[BTRFS_CRC32_SIZE]; - - path = btrfs_alloc_path(); - BUG_ON(!path); - file_key.objectid = objectid; - file_key.offset = offset; - file_key.flags = 0; - btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); - mutex_lock(&root->fs_info->fs_mutex); - - item = btrfs_lookup_csum(NULL, root, path, objectid, offset, 0); - if (IS_ERR(item)) { - ret = PTR_ERR(item); - /* a csum that isn't present is a preallocated region. */ - if (ret == -ENOENT || ret == -EFBIG) - ret = -ENOENT; - goto fail; - } - - ret = btrfs_csum_data(root, data, len, result); - WARN_ON(ret); - if (memcmp(result, &item->csum, BTRFS_CRC32_SIZE)) - ret = 1; -fail: - btrfs_release_path(root, path); - btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); - return ret; -} diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 23f02935968..fef7ba1e707 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -81,6 +81,62 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) } } +static int insert_inline_extent(struct btrfs_root *root, struct inode *inode, + u64 offset, ssize_t size, + struct buffer_head *bh) +{ + struct btrfs_key key; + struct btrfs_path *path; + char *ptr, *kaddr; + struct btrfs_trans_handle *trans; + struct btrfs_file_extent_item *ei; + u32 datasize; + int err = 0; + int ret; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + + key.objectid = inode->i_ino; + key.offset = offset; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); + BUG_ON(size >= PAGE_CACHE_SIZE); + datasize = btrfs_file_extent_calc_inline_size(size); + + ret = btrfs_insert_empty_item(trans, root, path, &key, + datasize); + if (ret) { + err = ret; + goto fail; + } + ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), + path->slots[0], struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(ei, trans->transid); + btrfs_set_file_extent_type(ei, + BTRFS_FILE_EXTENT_INLINE); + ptr = btrfs_file_extent_inline_start(ei); + + kaddr = kmap_atomic(bh->b_page, KM_USER0); + btrfs_memcpy(root, path->nodes[0]->b_data, + ptr, kaddr + bh_offset(bh), + size); + kunmap_atomic(kaddr, KM_USER0); + mark_buffer_dirty(path->nodes[0]); +fail: + btrfs_free_path(path); + ret = btrfs_end_transaction(trans, root); + if (ret && !err) + err = ret; + mutex_unlock(&root->fs_info->fs_mutex); + return err; +} + static int dirty_and_release_pages(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct file *file, @@ -96,57 +152,22 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, int this_write; struct inode *inode = file->f_path.dentry->d_inode; struct buffer_head *bh; - struct btrfs_file_extent_item *ei; for (i = 0; i < num_pages; i++) { offset = pos & (PAGE_CACHE_SIZE -1); this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes); - /* FIXME, one block at a time */ + /* FIXME, one block at a time */ bh = page_buffers(pages[i]); if (buffer_mapped(bh) && bh->b_blocknr == 0) { - struct btrfs_key key; - struct btrfs_path *path; - char *ptr, *kaddr; - u32 datasize; - - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - - /* create an inline extent, and copy the data in */ - path = btrfs_alloc_path(); - BUG_ON(!path); - key.objectid = inode->i_ino; - key.offset = pages[i]->index << PAGE_CACHE_SHIFT; - key.flags = 0; - btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); - BUG_ON(write_bytes >= PAGE_CACHE_SIZE); - datasize = offset + - btrfs_file_extent_calc_inline_size(write_bytes); - - ret = btrfs_insert_empty_item(trans, root, path, &key, - datasize); - BUG_ON(ret); - ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(ei, trans->transid); - btrfs_set_file_extent_type(ei, - BTRFS_FILE_EXTENT_INLINE); - ptr = btrfs_file_extent_inline_start(ei); - - kaddr = kmap_atomic(bh->b_page, KM_USER0); - btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, kaddr + bh_offset(bh), - offset + write_bytes); - kunmap_atomic(kaddr, KM_USER0); - - mark_buffer_dirty(path->nodes[0]); - btrfs_free_path(path); - ret = btrfs_end_transaction(trans, root); - BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); + ret = insert_inline_extent(root, inode, + pages[i]->index << PAGE_CACHE_SHIFT, + offset + this_write, bh); + if (ret) { + err = ret; + goto failed; + } } ret = btrfs_commit_write(file, pages[i], offset, @@ -321,6 +342,7 @@ next_slot: btrfs_file_extent_disk_blocknr(extent); } ret = btrfs_del_item(trans, root, path); + /* TODO update progress marker and return */ BUG_ON(ret); btrfs_release_path(root, path); extent = NULL; @@ -452,7 +474,8 @@ static int prepare_pages(struct btrfs_root *root, err = btrfs_drop_extents(trans, root, inode, start_pos, (pos + write_bytes + root->blocksize -1) & ~((u64)root->blocksize - 1), &hint_block); - BUG_ON(err); + if (err) + goto failed_release; } /* insert any holes we need to create */ @@ -469,7 +492,8 @@ static int prepare_pages(struct btrfs_root *root, last_pos_in_file, 0, 0, hole_size); } - BUG_ON(err); + if (err) + goto failed_release; } /* @@ -481,11 +505,13 @@ static int prepare_pages(struct btrfs_root *root, err = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, hint_block, (u64)-1, &ins, 1); - BUG_ON(err); + if (err) + goto failed_truncate; err = btrfs_insert_file_extent(trans, root, inode->i_ino, start_pos, ins.objectid, ins.offset, ins.offset); - BUG_ON(err); + if (err) + goto failed_truncate; } else { ins.offset = 0; ins.objectid = 0; @@ -618,16 +644,21 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, ret = prepare_pages(root, file, pages, num_pages, pos, first_index, last_index, write_bytes); - BUG_ON(ret); + if (ret) + goto out; ret = btrfs_copy_from_user(pos, num_pages, write_bytes, pages, buf); - BUG_ON(ret); + if (ret) { + btrfs_drop_pages(pages, num_pages); + goto out; + } ret = dirty_and_release_pages(NULL, root, file, pages, num_pages, pos, write_bytes); - BUG_ON(ret); btrfs_drop_pages(pages, num_pages); + if (ret) + goto out; buf += write_bytes; count -= write_bytes; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 94f1c28c25b..6d031daa777 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -212,7 +212,11 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, struct btrfs_dir_item *di; path = btrfs_alloc_path(); - BUG_ON(!path); + if (!path) { + ret = -ENOMEM; + goto err; + } + di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino, name, name_len, -1); if (IS_ERR(di)) { @@ -225,7 +229,8 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, } objectid = btrfs_disk_key_objectid(&di->location); ret = btrfs_delete_one_dir_name(trans, root, path, di); - BUG_ON(ret); + if (ret) + goto err; btrfs_release_path(root, path); di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, @@ -239,7 +244,6 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, goto err; } ret = btrfs_delete_one_dir_name(trans, root, path, di); - BUG_ON(ret); dentry->d_inode->i_ctime = dir->i_ctime; err: @@ -248,7 +252,7 @@ err: dir->i_size -= name_len * 2; btrfs_update_inode(trans, root, dir); drop_nlink(dentry->d_inode); - btrfs_update_inode(trans, root, dentry->d_inode); + ret = btrfs_update_inode(trans, root, dentry->d_inode); dir->i_sb->s_dirt = 1; } return ret; @@ -359,9 +363,10 @@ static int btrfs_free_inode(struct btrfs_trans_handle *trans, BUG_ON(!path); ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location, -1); - BUG_ON(ret); - ret = btrfs_del_item(trans, root, path); - BUG_ON(ret); + if (ret > 0) + ret = -ENOENT; + if (!ret) + ret = btrfs_del_item(trans, root, path); btrfs_free_path(path); return ret; } @@ -516,7 +521,8 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, } if (del_item) { ret = btrfs_del_item(trans, root, path); - BUG_ON(ret); + if (ret) + goto error; } else { break; } @@ -577,19 +583,22 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) page->index << PAGE_CACHE_SHIFT, (page->index + 1) << PAGE_CACHE_SHIFT, &alloc_hint); - BUG_ON(ret); + if (ret) + goto out; ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, alloc_hint, (u64)-1, &ins, 1); - BUG_ON(ret); + if (ret) + goto out; ret = btrfs_insert_file_extent(trans, root, inode->i_ino, page->index << PAGE_CACHE_SHIFT, ins.objectid, 1, 1); - BUG_ON(ret); + if (ret) + goto out; SetPageChecked(page); kaddr = kmap(page); memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); flush_dcache_page(page); - btrfs_csum_file_block(trans, root, inode->i_ino, + ret = btrfs_csum_file_block(trans, root, inode->i_ino, page->index << PAGE_CACHE_SHIFT, kaddr, PAGE_CACHE_SIZE); kunmap(page); @@ -633,9 +642,10 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_set_trans_block_group(trans, inode); err = btrfs_insert_file_extent(trans, root, inode->i_ino, pos, 0, 0, hole_size); - BUG_ON(err); btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); + if (err) + return err; } out: err = inode_setattr(inode, attr); @@ -657,12 +667,20 @@ void btrfs_delete_inode(struct inode *inode) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); ret = btrfs_truncate_in_trans(trans, root, inode); - BUG_ON(ret); - btrfs_free_inode(trans, root, inode); + if (ret) + goto no_delete_lock; + ret = btrfs_free_inode(trans, root, inode); + if (ret) + goto no_delete_lock; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root); return; + +no_delete_lock: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root); no_delete: clear_inode(inode); } @@ -946,7 +964,7 @@ int btrfs_write_inode(struct inode *inode, int wait) } /* - * This is somewhat expense, updating the tree every time the + * This is somewhat expensive, updating the tree every time the * inode changes. But, it is most likely to find the inode in cache. * FIXME, needs more benchmarking...there are no reasons other than performance * to keep or drop this code. @@ -1002,8 +1020,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); ret = btrfs_insert_inode(trans, root, objectid, &inode_item); - BUG_ON(ret); - + if (ret) + return ERR_PTR(ret); insert_inode_hash(inode); return inode; } @@ -1121,7 +1139,9 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, dir); - btrfs_update_inode(trans, root, inode); + err = btrfs_update_inode(trans, root, inode); + if (err) + drop_inode = 1; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); @@ -1349,17 +1369,26 @@ not_found: ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, alloc_hint, (u64)-1, &ins, 1); - BUG_ON(ret); + if (ret) { + err = ret; + goto out; + } ret = btrfs_insert_file_extent(trans, root, inode->i_ino, iblock << inode->i_blkbits, ins.objectid, ins.offset, ins.offset); - BUG_ON(ret); + if (ret) { + err = ret; + goto out; + } btrfs_map_bh_to_logical(root, result, ins.objectid); } out: - if (trans) - err = btrfs_end_transaction(trans, root); + if (trans) { + ret = btrfs_end_transaction(trans, root); + if (!err) + err = ret; + } btrfs_free_path(path); return err; } @@ -1375,8 +1404,8 @@ int btrfs_get_block(struct inode *inode, sector_t iblock, return err; } -int btrfs_get_block_csum(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create) +static int btrfs_get_block_csum(struct inode *inode, sector_t iblock, + struct buffer_head *result, int create) { int ret; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1397,7 +1426,7 @@ int btrfs_get_block_csum(struct inode *inode, sector_t iblock, /* a csum that isn't present is a preallocated region. */ if (ret == -ENOENT || ret == -EFBIG) ret = 0; - result->b_private = 0; + result->b_private = NULL; goto out; } memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE); @@ -1736,11 +1765,10 @@ static int __btrfs_write_full_page(struct inode *inode, struct page *page, trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); kaddr = kmap(page); - ret = btrfs_csum_file_block(trans, root, inode->i_ino, + btrfs_csum_file_block(trans, root, inode->i_ino, off, kaddr + bh_offset(bh), bh->b_size); kunmap(page); - BUG_ON(ret); ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); @@ -1930,7 +1958,6 @@ static void btrfs_truncate(struct inode *inode) /* FIXME, add redo link to tree so we don't leak on crash */ ret = btrfs_truncate_in_trans(trans, root, inode); - BUG_ON(ret); btrfs_update_inode(trans, root, inode); ret = btrfs_end_transaction(trans, root); BUG_ON(ret); @@ -1970,6 +1997,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) struct inode *inode; struct inode *dir; int ret; + int err; u64 objectid; u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; @@ -1978,8 +2006,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) BUG_ON(!trans); subvol = btrfs_alloc_free_block(trans, root, 0); - if (subvol == NULL) - return -ENOSPC; + if (IS_ERR(subvol)) + return PTR_ERR(subvol); leaf = btrfs_buffer_leaf(subvol); btrfs_set_header_nritems(&leaf->header, 0); btrfs_set_header_level(&leaf->header, 0); @@ -2005,7 +2033,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, 0, &objectid); - BUG_ON(ret); + if (ret) + goto fail; btrfs_set_root_dirid(&root_item, new_dirid); @@ -2015,7 +2044,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &root_item); - BUG_ON(ret); + if (ret) + goto fail; /* * insert the directory item @@ -2025,10 +2055,12 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, name, namelen, dir->i_ino, &key, BTRFS_FT_DIR); - BUG_ON(ret); + if (ret) + goto fail; ret = btrfs_commit_transaction(trans, root); - BUG_ON(ret); + if (ret) + goto fail_commit; new_root = btrfs_read_fs_root(root->fs_info, &key); BUG_ON(!new_root); @@ -2038,24 +2070,29 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) inode = btrfs_new_inode(trans, new_root, new_dirid, BTRFS_I(dir)->block_group, S_IFDIR | 0700); + if (IS_ERR(inode)) + goto fail; inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; new_root->inode = inode; ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid); - BUG_ON(ret); + if (ret) + goto fail; inode->i_nlink = 1; inode->i_size = 6; ret = btrfs_update_inode(trans, new_root, inode); - BUG_ON(ret); - - ret = btrfs_commit_transaction(trans, new_root); - BUG_ON(ret); - + if (ret) + goto fail; +fail: + err = btrfs_commit_transaction(trans, root); + if (err && !ret) + ret = err; +fail_commit: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root); - return 0; + return ret; } static int create_snapshot(struct btrfs_root *root, char *name, int namelen) @@ -2064,6 +2101,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) struct btrfs_key key; struct btrfs_root_item new_root_item; int ret; + int err; u64 objectid; if (!root->ref_cows) @@ -2074,11 +2112,13 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) BUG_ON(!trans); ret = btrfs_update_inode(trans, root, root->inode); - BUG_ON(ret); + if (ret) + goto fail; ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, 0, &objectid); - BUG_ON(ret); + if (ret) + goto fail; memcpy(&new_root_item, &root->root_item, sizeof(new_root_item)); @@ -2091,7 +2131,8 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); - BUG_ON(ret); + if (ret) + goto fail; /* * insert the directory item @@ -2102,16 +2143,20 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) root->fs_info->sb->s_root->d_inode->i_ino, &key, BTRFS_FT_DIR); - BUG_ON(ret); + if (ret) + goto fail; ret = btrfs_inc_root_ref(trans, root); - BUG_ON(ret); + if (ret) + goto fail; - ret = btrfs_commit_transaction(trans, root); - BUG_ON(ret); +fail: + err = btrfs_commit_transaction(trans, root); + if (err && !ret) + ret = err; mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root); - return 0; + return ret; } int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int @@ -2148,12 +2193,13 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int btrfs_free_path(path); if (di && !IS_ERR(di)) return -EEXIST; + if (IS_ERR(di)) + return PTR_ERR(di); if (root == root->fs_info->tree_root) ret = create_subvol(root, vol_args.name, namelen); else ret = create_snapshot(root, vol_args.name, namelen); - WARN_ON(ret); break; default: return -ENOTTY; @@ -2316,7 +2362,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, old_parent_oid = btrfs_disk_key_objectid(&di->location); ret = btrfs_del_item(trans, root, path); if (ret) { - ret = -EIO; goto out_fail; } btrfs_release_path(root, path); @@ -2335,7 +2380,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, } ret = btrfs_del_item(trans, root, path); if (ret) { - ret = -EIO; goto out_fail; } btrfs_release_path(root, path); @@ -2361,7 +2405,9 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, clear_nlink(new_inode); else drop_nlink(new_inode); - btrfs_update_inode(trans, root, new_inode); + ret = btrfs_update_inode(trans, root, new_inode); + if (ret) + goto out_fail; } ret = btrfs_add_link(trans, new_dentry, old_inode); if (ret) @@ -2433,7 +2479,10 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, datasize = btrfs_file_extent_calc_inline_size(name_len); err = btrfs_insert_empty_item(trans, root, path, &key, datasize); - BUG_ON(err); + if (err) { + drop_inode = 1; + goto out_unlock; + } ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], struct btrfs_file_extent_item); btrfs_set_file_extent_generation(ei, trans->transid); @@ -2447,13 +2496,13 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; inode->i_size = name_len - 1; - btrfs_update_inode(trans, root, inode); - err = 0; + err = btrfs_update_inode(trans, root, inode); + if (err) + drop_inode = 1; out_unlock: btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - if (drop_inode) { inode_dec_link_count(inode); iput(inode); diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 8e1b9046d5e..ac0fae7780f 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -90,7 +90,6 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root { int ret; ret = btrfs_insert_item(trans, root, key, item, sizeof(*item)); - BUG_ON(ret); return ret; } diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index ef8ad18b275..c11ecf50020 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -125,9 +125,8 @@ static int btrfs_sync_fs(struct super_block *sb, int wait) trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); sb->s_dirt = 0; - BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); - return 0; + return ret; } static void btrfs_write_super(struct super_block *sb) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index b5b99a85763..321f8852755 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -219,7 +219,8 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, struct btrfs_root *root; int i; int ret; - int err; + int err = 0; + while(1) { ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, ARRAY_SIZE(gang), @@ -251,11 +252,12 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, err = btrfs_insert_root(trans, root->fs_info->tree_root, &root->root_key, &root->root_item); - BUG_ON(err); + if (err) + break; list_add(&dirty->list, list); } } - return 0; + return err; } static int drop_dirty_roots(struct btrfs_root *tree_root, @@ -263,7 +265,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, { struct dirty_root *dirty; struct btrfs_trans_handle *trans; - int ret; + int ret = 0; while(!list_empty(list)) { mutex_lock(&tree_root->fs_info->fs_mutex); dirty = list_entry(list->next, struct dirty_root, list); @@ -274,14 +276,15 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, BUG_ON(ret); ret = btrfs_del_root(trans, tree_root, &dirty->snap_key); - BUG_ON(ret); + if (ret) + break; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); kfree(dirty); mutex_unlock(&tree_root->fs_info->fs_mutex); btrfs_btree_balance_dirty(tree_root); } - return 0; + return ret; } int btrfs_commit_transaction(struct btrfs_trans_handle *trans, @@ -321,9 +324,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, } finish_wait(&trans->transaction->writer_wait, &wait); WARN_ON(cur_trans != trans->transaction); - add_dirty_roots(trans, &root->fs_info->fs_roots_radix, &dirty_fs_roots); + ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, + &dirty_fs_roots); + BUG_ON(ret); + ret = btrfs_commit_tree_roots(trans, root); BUG_ON(ret); + cur_trans = root->fs_info->running_transaction; root->fs_info->running_transaction = NULL; if (cur_trans->list.prev != &root->fs_info->trans_list) { -- cgit v1.2.3-70-g09d2 From 5eda7b5e9b0bed864dd18284c7df9b3c8207dad7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 22 Jun 2007 14:16:25 -0400 Subject: Btrfs: Add the ability to find and remove dead roots after a crash. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 17 ++++++++++-- fs/btrfs/disk-io.c | 44 +++++++++++++++++++---------- fs/btrfs/disk-io.h | 3 ++ fs/btrfs/inode.c | 2 ++ fs/btrfs/root-tree.c | 75 ++++++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/transaction.c | 54 +++++++++++++++++++++++++++++++----- fs/btrfs/transaction.h | 1 + 7 files changed, 164 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 77071f27397..fb6fffb71dd 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -122,12 +122,12 @@ struct btrfs_super_block { u8 fsid[16]; /* FS specific uuid */ __le64 blocknr; /* this block number */ __le64 magic; - __le32 blocksize; __le64 generation; __le64 root; __le64 total_blocks; __le64 blocks_used; __le64 root_dir_objectid; + __le32 blocksize; } __attribute__ ((__packed__)); /* @@ -226,10 +226,12 @@ struct btrfs_root_item { struct btrfs_inode_item inode; __le64 root_dirid; __le64 blocknr; - __le32 flags; __le64 block_limit; __le64 blocks_used; + __le32 flags; __le32 refs; + struct btrfs_disk_key drop_progress; + u8 drop_level; } __attribute__ ((__packed__)); #define BTRFS_FILE_EXTENT_REG 0 @@ -800,6 +802,16 @@ static inline void btrfs_set_root_refs(struct btrfs_root_item *item, u32 val) item->refs = cpu_to_le32(val); } +static inline u32 btrfs_root_flags(struct btrfs_root_item *item) +{ + return le32_to_cpu(item->flags); +} + +static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val) +{ + item->flags = cpu_to_le32(val); +} + static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s) { return le64_to_cpu(s->blocknr); @@ -1076,6 +1088,7 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root *item); int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct btrfs_root_item *item, struct btrfs_key *key); +int btrfs_find_dead_roots(struct btrfs_root *root); /* dir-item.c */ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, u64 dir, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 865a284aa06..d7615e1578c 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -326,8 +326,8 @@ static int find_and_setup_root(int blocksize, return 0; } -struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, - struct btrfs_key *location) +struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, + struct btrfs_key *location) { struct btrfs_root *root; struct btrfs_root *tree_root = fs_info->tree_root; @@ -336,11 +336,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, u64 highest_inode; int ret = 0; - root = radix_tree_lookup(&fs_info->fs_roots_radix, - (unsigned long)location->objectid); - if (root) - return root; - root = kmalloc(sizeof(*root), GFP_NOFS); + root = kzalloc(sizeof(*root), GFP_NOFS); if (!root) return ERR_PTR(-ENOMEM); if (location->offset == (u64)-1) { @@ -383,6 +379,28 @@ out: BUG_ON(!root->node); insert: root->ref_cows = 1; + ret = btrfs_find_highest_inode(root, &highest_inode); + if (ret == 0) { + root->highest_inode = highest_inode; + root->last_inode_alloc = highest_inode; + } + return root; +} + +struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, + struct btrfs_key *location) +{ + struct btrfs_root *root; + int ret; + + root = radix_tree_lookup(&fs_info->fs_roots_radix, + (unsigned long)location->objectid); + if (root) + return root; + + root = btrfs_read_fs_root_no_radix(fs_info, location); + if (IS_ERR(root)) + return root; ret = radix_tree_insert(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, root); @@ -391,11 +409,6 @@ insert: kfree(root); return ERR_PTR(ret); } - ret = btrfs_find_highest_inode(root, &highest_inode); - if (ret == 0) { - root->highest_inode = highest_inode; - root->last_inode_alloc = highest_inode; - } return root; } @@ -489,6 +502,9 @@ struct btrfs_root *open_ctree(struct super_block *sb) btrfs_read_block_groups(extent_root); fs_info->generation = btrfs_super_generation(disk_super) + 1; + ret = btrfs_find_dead_roots(tree_root); + if (ret) + goto fail_tree_root; mutex_unlock(&fs_info->fs_mutex); return tree_root; @@ -538,7 +554,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root return 0; } -static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) +int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) { radix_tree_delete(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid); @@ -565,7 +581,7 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info) if (!ret) break; for (i = 0; i < ret; i++) - free_fs_root(fs_info, gang[i]); + btrfs_free_fs_root(fs_info, gang[i]); } return 0; } diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 7b76ccc4875..c4a695ac44f 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -65,6 +65,8 @@ int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, char *result); struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_key *location); +struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, + struct btrfs_key *location); u64 bh_blocknr(struct buffer_head *bh); int btrfs_insert_dev_radix(struct btrfs_root *root, struct block_device *bdev, @@ -75,4 +77,5 @@ int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, u64 logical); int btrfs_releasepage(struct page *page, gfp_t flags); void btrfs_btree_balance_dirty(struct btrfs_root *root); +int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); #endif diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6d031daa777..9d2a0a3f674 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2028,6 +2028,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol)); btrfs_set_root_refs(&root_item, 1); + memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); + root_item.drop_level = 0; brelse(subvol); subvol = NULL; diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index ac0fae7780f..737e5a38d17 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -18,6 +18,7 @@ #include #include "ctree.h" +#include "transaction.h" #include "disk-io.h" #include "print-tree.h" @@ -32,7 +33,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, search_key.objectid = objectid; search_key.flags = (u32)-1; - search_key.offset = (u32)-1; + search_key.offset = (u64)-1; path = btrfs_alloc_path(); BUG_ON(!path); @@ -50,6 +51,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item), sizeof(*item)); btrfs_disk_key_to_cpu(key, &l->items[slot].key); +printk("find last finds key %Lu %u %Lu slot %d search for obj %Lu\n", key->objectid, key->flags, key->offset, slot, objectid); ret = 0; out: btrfs_release_path(root, path); @@ -93,6 +95,67 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root return ret; } +int btrfs_find_dead_roots(struct btrfs_root *root) +{ + struct btrfs_root *dead_root; + struct btrfs_item *item; + struct btrfs_root_item *ri; + struct btrfs_key key; + struct btrfs_path *path; + int ret; + u32 nritems; + struct btrfs_leaf *leaf; + int slot; + + key.objectid = 0; + key.flags = 0; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + key.offset = 0; + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto err; + while(1) { + leaf = btrfs_buffer_leaf(path->nodes[0]); + nritems = btrfs_header_nritems(&leaf->header); + slot = path->slots[0]; + if (slot >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret) + break; + leaf = btrfs_buffer_leaf(path->nodes[0]); + nritems = btrfs_header_nritems(&leaf->header); + slot = path->slots[0]; + } + item = leaf->items + slot; + btrfs_disk_key_to_cpu(&key, &item->key); + if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY) + goto next; + ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item); + if (btrfs_root_refs(ri) != 0) + goto next; + dead_root = btrfs_read_fs_root_no_radix(root->fs_info, &key); + if (IS_ERR(root)) { + ret = PTR_ERR(root); + goto err; + } +printk("found dead root %Lu %u %Lu\n", key.objectid, key.flags, key.offset); + ret = btrfs_add_dead_root(dead_root, + &root->fs_info->dead_roots); + if (ret) + goto err; +next: + slot++; + path->slots[0]++; + } + ret = 0; +err: + btrfs_free_path(path); + return ret; +} + int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key) { @@ -111,14 +174,8 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, path->slots[0], struct btrfs_root_item); refs = btrfs_root_refs(ri); - BUG_ON(refs == 0); - if (refs == 1) { - ret = btrfs_del_item(trans, root, path); - } else { - btrfs_set_root_refs(ri, refs - 1); - WARN_ON(1); - mark_buffer_dirty(path->nodes[0]); - } + BUG_ON(refs != 0); + ret = btrfs_del_item(trans, root, path); out: btrfs_release_path(root, path); btrfs_free_path(path); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 321f8852755..85a2a5e2714 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -85,11 +85,15 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, if (root != root->fs_info->tree_root && root->last_trans < running_trans_id) { - radix_tree_tag_set(&root->fs_info->fs_roots_radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_TRANS_TAG); - root->commit_root = root->node; - get_bh(root->node); + if (root->root_item.refs != 0) { + radix_tree_tag_set(&root->fs_info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_TRANS_TAG); + root->commit_root = root->node; + get_bh(root->node); + } else { + WARN_ON(1); + } } root->last_trans = running_trans_id; h->transid = running_trans_id; @@ -208,8 +212,24 @@ struct dirty_root { struct btrfs_key snap_key; struct buffer_head *commit_root; struct btrfs_root *root; + int free_on_drop; }; +int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list) +{ + struct dirty_root *dirty; + + dirty = kmalloc(sizeof(*dirty), GFP_NOFS); + if (!dirty) + return -ENOMEM; + memcpy(&dirty->snap_key, &root->root_key, sizeof(root->root_key)); + dirty->commit_root = root->node; + dirty->root = root; + dirty->free_on_drop = 1; + list_add(&dirty->list, dead_list); + return 0; +} + static int add_dirty_roots(struct btrfs_trans_handle *trans, struct radix_tree_root *radix, struct list_head *list) @@ -217,9 +237,11 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, struct dirty_root *dirty; struct btrfs_root *gang[8]; struct btrfs_root *root; + struct btrfs_root_item tmp_item; int i; int ret; int err = 0; + u32 refs; while(1) { ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, @@ -246,6 +268,9 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, dirty->commit_root = root->commit_root; root->commit_root = NULL; dirty->root = root; + dirty->free_on_drop = 0; + memcpy(&tmp_item, &root->root_item, sizeof(tmp_item)); + root->root_key.offset = root->fs_info->generation; btrfs_set_root_blocknr(&root->root_item, bh_blocknr(root->node)); @@ -254,7 +279,18 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, &root->root_item); if (err) break; - list_add(&dirty->list, list); + + refs = btrfs_root_refs(&tmp_item); + btrfs_set_root_refs(&tmp_item, refs - 1); + err = btrfs_update_root(trans, root->fs_info->tree_root, + &dirty->snap_key, + &tmp_item); + + BUG_ON(err); + if (refs == 1) + list_add(&dirty->list, list); + else + kfree(dirty); } } return err; @@ -270,16 +306,20 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, mutex_lock(&tree_root->fs_info->fs_mutex); dirty = list_entry(list->next, struct dirty_root, list); list_del_init(&dirty->list); + trans = btrfs_start_transaction(tree_root, 1); +printk("deleting root %Lu %u %Lu\n", dirty->snap_key.objectid, dirty->snap_key.flags, dirty->snap_key.offset); ret = btrfs_drop_snapshot(trans, dirty->root, dirty->commit_root); BUG_ON(ret); - ret = btrfs_del_root(trans, tree_root, &dirty->snap_key); if (ret) break; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); + + if (dirty->free_on_drop) + kfree(dirty->root); kfree(dirty); mutex_unlock(&tree_root->fs_info->fs_mutex); btrfs_btree_balance_dirty(tree_root); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 5fb1d322aa1..ebf44f3e111 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -69,5 +69,6 @@ void btrfs_transaction_flush_work(struct btrfs_root *root); void btrfs_transaction_queue_work(struct btrfs_root *root, int delay); void btrfs_init_transaction_sys(void); void btrfs_exit_transaction_sys(void); +int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list); #endif -- cgit v1.2.3-70-g09d2 From 79c44584ea346024e929ac0f2a3577aac33b7b3f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 25 Jun 2007 10:09:33 -0400 Subject: Btrfs: Fix mtime and ctime updates on parent dirs Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 9d2a0a3f674..eba06e7cf41 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -250,6 +250,7 @@ err: btrfs_free_path(path); if (!ret) { dir->i_size -= name_len * 2; + dir->i_mtime = dir->i_ctime = CURRENT_TIME; btrfs_update_inode(trans, root, dir); drop_nlink(dentry->d_inode); ret = btrfs_update_inode(trans, root, dentry->d_inode); @@ -1037,6 +1038,7 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, int ret; struct btrfs_key key; struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root; + struct inode *parent_inode; key.objectid = inode->i_ino; key.flags = 0; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); @@ -1047,7 +1049,9 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_parent->d_inode->i_ino, &key, btrfs_inode_type(inode)); if (ret == 0) { - dentry->d_parent->d_inode->i_size += dentry->d_name.len * 2; + parent_inode = dentry->d_parent->d_inode; + parent_inode->i_size += dentry->d_name.len * 2; + parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; ret = btrfs_update_inode(trans, root, dentry->d_parent->d_inode); } -- cgit v1.2.3-70-g09d2 From ccd467d60e81b48cdbecae93532b66bcdedca91d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 28 Jun 2007 15:57:36 -0400 Subject: Btrfs: crash recovery fixes Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 13 +++++++++++- fs/btrfs/ctree.h | 12 ++++------- fs/btrfs/disk-io.c | 22 ++++++++++++-------- fs/btrfs/disk-io.h | 1 + fs/btrfs/extent-tree.c | 37 ++++++++++++++++++++++++++++++---- fs/btrfs/file.c | 15 ++++++++++---- fs/btrfs/inode.c | 6 +++--- fs/btrfs/transaction.c | 54 +++++++++++++++++++++++++++++++------------------- 8 files changed, 112 insertions(+), 48 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 606a19b5916..9eb646529ed 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -75,6 +75,17 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_node *cow_node; int ret; + WARN_ON(!buffer_uptodate(buf)); + if (trans->transaction != root->fs_info->running_transaction) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->running_transaction->transid); + WARN_ON(1); + } + if (trans->transid != root->fs_info->generation) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->generation); + WARN_ON(1); + } if (btrfs_header_generation(btrfs_buffer_header(buf)) == trans->transid) { *cow_ret = buf; @@ -107,7 +118,7 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1); } btrfs_block_release(root, buf); - mark_buffer_dirty(cow); + btrfs_mark_buffer_dirty(cow); *cow_ret = cow; return 0; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 1998f86df08..0287bd51d87 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1013,18 +1013,13 @@ static inline void btrfs_memmove(struct btrfs_root *root, memmove(dst, src, nr); } -static inline void btrfs_mark_buffer_dirty(struct buffer_head *bh) -{ - WARN_ON(!atomic_read(&bh->b_count)); - mark_buffer_dirty(bh); -} - /* helper function to cast into the data area of the leaf. */ #define btrfs_item_ptr(leaf, slot, type) \ ((type *)(btrfs_leaf_data(leaf) + \ btrfs_item_offset((leaf)->items + (slot)))) /* extent-tree.c */ +int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy); struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, u64 blocknr); @@ -1044,8 +1039,9 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 blocknr, u64 num_blocks, int pin); -int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct - btrfs_root *root); +int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct radix_tree_root *unpin_radix); int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 blocknr, u64 num_blocks); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 7081729d5b1..d1bf5bc1bc1 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -270,14 +270,6 @@ fail: return NULL; } -int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf) -{ - WARN_ON(atomic_read(&buf->b_count) == 0); - mark_buffer_dirty(buf); - return 0; -} - int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *buf) { @@ -621,6 +613,20 @@ int close_ctree(struct btrfs_root *root) return 0; } +void btrfs_mark_buffer_dirty(struct buffer_head *bh) +{ + struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root; + u64 transid = btrfs_header_generation(btrfs_buffer_header(bh)); + WARN_ON(!atomic_read(&bh->b_count)); + if (transid != root->fs_info->generation) { + printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", + (unsigned long long)bh->b_blocknr, + transid, root->fs_info->generation); + WARN_ON(1); + } + mark_buffer_dirty(bh); +} + void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf) { brelse(buf); diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index c4a695ac44f..9e2c261b41a 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -78,4 +78,5 @@ int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, int btrfs_releasepage(struct page *page, gfp_t flags); void btrfs_btree_balance_dirty(struct btrfs_root *root); int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); +void btrfs_mark_buffer_dirty(struct buffer_head *bh); #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 01dc3057928..14b93268920 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -523,6 +523,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, } return 0; fail: + WARN_ON(1); for (i =0; i < faili; i++) { if (leaf) { u64 disk_blocknr; @@ -572,7 +573,7 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], struct btrfs_block_group_item); memcpy(bi, &cache->item, sizeof(*bi)); - mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_release_path(extent_root, path); fail: finish_current_insert(trans, extent_root); @@ -739,8 +740,30 @@ static int try_remove_page(struct address_space *mapping, unsigned long index) return ret; } -int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct - btrfs_root *root) +int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy) +{ + unsigned long gang[8]; + u64 last = 0; + struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix; + int ret; + int i; + + while(1) { + ret = find_first_radix_bit(pinned_radix, gang, last, + ARRAY_SIZE(gang)); + if (!ret) + break; + for (i = 0 ; i < ret; i++) { + set_radix_bit(copy, gang[i]); + last = gang[i] + 1; + } + } + return 0; +} + +int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct radix_tree_root *unpin_radix) { unsigned long gang[8]; struct inode *btree_inode = root->fs_info->btree_inode; @@ -752,7 +775,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct struct radix_tree_root *extent_radix = &root->fs_info->extent_map_radix; while(1) { - ret = find_first_radix_bit(pinned_radix, gang, 0, + ret = find_first_radix_bit(unpin_radix, gang, 0, ARRAY_SIZE(gang)); if (!ret) break; @@ -760,6 +783,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct first = gang[0]; for (i = 0; i < ret; i++) { clear_radix_bit(pinned_radix, gang[i]); + clear_radix_bit(unpin_radix, gang[i]); block_group = btrfs_lookup_block_group(root->fs_info, gang[i]); if (block_group) { @@ -1309,6 +1333,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, if (data) { ret = find_free_extent(trans, root, 0, 0, search_end, 0, &prealloc_key, 0, 0, 0); + BUG_ON(ret); if (ret) return ret; exclude_nr = info->extent_tree_prealloc_nr; @@ -1319,6 +1344,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, ret = find_free_extent(trans, root, num_blocks, search_start, search_end, hint_block, ins, exclude_start, exclude_nr, data); + BUG_ON(ret); if (ret) return ret; @@ -1334,10 +1360,12 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, if (!data) { exclude_start = ins->objectid; exclude_nr = ins->offset; + hint_block = exclude_start + exclude_nr; ret = find_free_extent(trans, root, 0, search_start, search_end, hint_block, &prealloc_key, exclude_start, exclude_nr, 0); + BUG_ON(ret); if (ret) return ret; } @@ -1348,6 +1376,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, ret = btrfs_insert_item(trans, extent_root, ins, &extent_item, sizeof(extent_item)); + BUG_ON(ret); finish_current_insert(trans, extent_root); pending_ret = del_pending_extents(trans, extent_root); if (ret) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index fef7ba1e707..2456cc3e1cf 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -127,7 +127,7 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode, ptr, kaddr + bh_offset(bh), size); kunmap_atomic(kaddr, KM_USER0); - mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_free_path(path); ret = btrfs_end_transaction(trans, root); @@ -211,11 +211,13 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, int found_type; int found_extent; int found_inline; + int recow; path = btrfs_alloc_path(); if (!path) return -ENOMEM; while(1) { + recow = 0; btrfs_release_path(root, path); ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, search_start, -1); @@ -244,6 +246,10 @@ next_slot: if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) { goto out; } + if (recow) { + search_start = key.offset; + continue; + } if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); @@ -274,6 +280,7 @@ next_slot: nextret = btrfs_next_leaf(root, path); if (nextret) goto out; + recow = 1; } else { path->slots[0]++; } @@ -321,7 +328,7 @@ next_slot: } btrfs_set_file_extent_num_blocks(extent, new_num); - mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(path->nodes[0]); } else { WARN_ON(1); } @@ -452,6 +459,8 @@ static int prepare_pages(struct btrfs_root *root, err = -ENOMEM; goto failed_release; } + cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); + wait_on_page_writeback(pages[i]); } mutex_lock(&root->fs_info->fs_mutex); @@ -522,8 +531,6 @@ static int prepare_pages(struct btrfs_root *root, mutex_unlock(&root->fs_info->fs_mutex); for (i = 0; i < num_pages; i++) { - cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); - wait_on_page_writeback(pages[i]); offset = pos & (PAGE_CACHE_SIZE -1); this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes); if (!page_has_buffers(pages[i])) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index eba06e7cf41..4fc0367d54f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -506,7 +506,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, extent_num_blocks); inode->i_blocks -= (orig_num_blocks - extent_num_blocks) << 3; - mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(path->nodes[0]); } else { extent_start = btrfs_file_extent_disk_blocknr(fi); @@ -2020,7 +2020,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) btrfs_set_header_owner(&leaf->header, root->root_key.objectid); memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid, sizeof(leaf->header.fsid)); - mark_buffer_dirty(subvol); + btrfs_mark_buffer_dirty(subvol); inode_item = &root_item.inode; memset(inode_item, 0, sizeof(*inode_item)); @@ -2497,7 +2497,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, ptr = btrfs_file_extent_inline_start(ei); btrfs_memcpy(root, path->nodes[0]->b_data, ptr, symname, name_len); - mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_free_path(path); inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 3b2face593e..bec38ae8aa1 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -85,6 +85,8 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, if (root != root->fs_info->tree_root && root->last_trans < running_trans_id) { + WARN_ON(root == root->fs_info->extent_root); + WARN_ON(root->ref_cows != 1); if (root->root_item.refs != 0) { radix_tree_tag_set(&root->fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, @@ -113,10 +115,11 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->trans_mutex); cur_trans = root->fs_info->running_transaction; + WARN_ON(cur_trans != trans->transaction); WARN_ON(cur_trans->num_writers < 1); + cur_trans->num_writers--; if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); - cur_trans->num_writers--; put_transaction(cur_trans); mutex_unlock(&root->fs_info->trans_mutex); memset(trans, 0, sizeof(*trans)); @@ -194,6 +197,7 @@ static int wait_for_commit(struct btrfs_root *root, struct btrfs_transaction *commit) { DEFINE_WAIT(wait); + mutex_lock(&root->fs_info->trans_mutex); while(!commit->commit_done) { prepare_to_wait(&commit->commit_wait, &wait, TASK_UNINTERRUPTIBLE); @@ -203,6 +207,7 @@ static int wait_for_commit(struct btrfs_root *root, schedule(); mutex_lock(&root->fs_info->trans_mutex); } + mutex_unlock(&root->fs_info->trans_mutex); finish_wait(&commit->commit_wait, &wait); return 0; } @@ -279,7 +284,6 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, &root->root_item); if (err) break; - refs = btrfs_root_refs(&tmp_item); btrfs_set_root_refs(&tmp_item, refs - 1); err = btrfs_update_root(trans, root->fs_info->tree_root, @@ -333,31 +337,53 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction *cur_trans; struct btrfs_transaction *prev_trans = NULL; struct list_head dirty_fs_roots; + struct radix_tree_root pinned_copy; DEFINE_WAIT(wait); + init_bit_radix(&pinned_copy); INIT_LIST_HEAD(&dirty_fs_roots); mutex_lock(&root->fs_info->trans_mutex); if (trans->transaction->in_commit) { cur_trans = trans->transaction; trans->transaction->use_count++; + mutex_unlock(&root->fs_info->trans_mutex); btrfs_end_transaction(trans, root); + + mutex_unlock(&root->fs_info->fs_mutex); ret = wait_for_commit(root, cur_trans); BUG_ON(ret); put_transaction(cur_trans); - mutex_unlock(&root->fs_info->trans_mutex); + mutex_lock(&root->fs_info->fs_mutex); return 0; } - cur_trans = trans->transaction; trans->transaction->in_commit = 1; + cur_trans = trans->transaction; + if (cur_trans->list.prev != &root->fs_info->trans_list) { + prev_trans = list_entry(cur_trans->list.prev, + struct btrfs_transaction, list); + if (!prev_trans->commit_done) { + prev_trans->use_count++; + mutex_unlock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->trans_mutex); + + wait_for_commit(root, prev_trans); + put_transaction(prev_trans); + + mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->trans_mutex); + } + } while (trans->transaction->num_writers > 1) { WARN_ON(cur_trans != trans->transaction); prepare_to_wait(&trans->transaction->writer_wait, &wait, TASK_UNINTERRUPTIBLE); if (trans->transaction->num_writers <= 1) break; + mutex_unlock(&root->fs_info->fs_mutex); mutex_unlock(&root->fs_info->trans_mutex); schedule(); + mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); finish_wait(&trans->transaction->writer_wait, &wait); } @@ -372,34 +398,22 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, cur_trans = root->fs_info->running_transaction; root->fs_info->running_transaction = NULL; - if (cur_trans->list.prev != &root->fs_info->trans_list) { - prev_trans = list_entry(cur_trans->list.prev, - struct btrfs_transaction, list); - if (prev_trans->commit_done) - prev_trans = NULL; - else - prev_trans->use_count++; - } btrfs_set_super_generation(&root->fs_info->super_copy, cur_trans->transid); btrfs_set_super_root(&root->fs_info->super_copy, bh_blocknr(root->fs_info->tree_root->node)); memcpy(root->fs_info->disk_super, &root->fs_info->super_copy, sizeof(root->fs_info->super_copy)); + + btrfs_copy_pinned(root, &pinned_copy); + mutex_unlock(&root->fs_info->trans_mutex); mutex_unlock(&root->fs_info->fs_mutex); ret = btrfs_write_and_wait_transaction(trans, root); - if (prev_trans) { - mutex_lock(&root->fs_info->trans_mutex); - wait_for_commit(root, prev_trans); - put_transaction(prev_trans); - mutex_unlock(&root->fs_info->trans_mutex); - } BUG_ON(ret); write_ctree_super(trans, root); - mutex_lock(&root->fs_info->fs_mutex); - btrfs_finish_extent_commit(trans, root); + btrfs_finish_extent_commit(trans, root, &pinned_copy); mutex_lock(&root->fs_info->trans_mutex); cur_trans->commit_done = 1; wake_up(&cur_trans->commit_wait); -- cgit v1.2.3-70-g09d2 From ec6b910fb330f29e169c9f502c3ac209515af8d1 Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Wed, 11 Jul 2007 10:00:37 -0400 Subject: Btrfs: trivial include fixups Almost none of the files including module.h need to do so, remove them. Include sched.h in extent-tree.c to silence a warning about cond_resched() being undeclared. Signed-off-by: Zach Brown Signed-off-by: Chris Mason --- fs/btrfs/bit-radix.c | 1 - fs/btrfs/ctree.c | 1 - fs/btrfs/dir-item.c | 1 - fs/btrfs/disk-io.c | 1 - fs/btrfs/extent-tree.c | 2 +- fs/btrfs/file-item.c | 1 - fs/btrfs/file.c | 1 - fs/btrfs/inode-item.c | 1 - fs/btrfs/inode-map.c | 1 - fs/btrfs/inode.c | 1 - fs/btrfs/print-tree.c | 1 - fs/btrfs/root-tree.c | 1 - fs/btrfs/sysfs.c | 1 - fs/btrfs/transaction.c | 1 - 14 files changed, 1 insertion(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c index a4ddf4de979..e8bf876db39 100644 --- a/fs/btrfs/bit-radix.c +++ b/fs/btrfs/bit-radix.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "bit-radix.h" #define BIT_ARRAY_BYTES 256 diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 9eb646529ed..eace2ee7644 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index d5db8bdea3d..49db5fa7ced 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "hash.h" diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index d1bf5bc1bc1..60db85bc4ba 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include #include #include diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 14b93268920..5ace2c33f1a 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -16,7 +16,7 @@ * Boston, MA 021110-1307, USA. */ -#include +#include #include "ctree.h" #include "disk-io.h" #include "print-tree.h" diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 68859934ae2..ea4dd4c5fce 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 2456cc3e1cf..f327525f541 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include #include #include diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index e8b5a9f7c0c..b7834617753 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 58930fc48be..40547086625 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4fc0367d54f..d75ca5cf59b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include #include #include diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index e66acd4e274..bafa78603bd 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "print-tree.h" diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 8b0eca042ea..271dd2b5c32 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "transaction.h" #include "disk-io.h" diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index ead9dc8051b..db8917e5b25 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index bec38ae8aa1..0f494fe365c 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include #include #include "ctree.h" -- cgit v1.2.3-70-g09d2 From 618e21d5957c88c5ea346dfd7d1dbac96410fc5b Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 11 Jul 2007 10:18:17 -0400 Subject: Btrfs: Implement mknod Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 72 +++++++++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 66 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d75ca5cf59b..f95cf5d4a69 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -46,6 +46,7 @@ struct btrfs_iget_args { static struct inode_operations btrfs_dir_inode_operations; static struct inode_operations btrfs_symlink_inode_operations; static struct inode_operations btrfs_dir_ro_inode_operations; +static struct inode_operations btrfs_special_inode_operations; static struct inode_operations btrfs_file_inode_operations; static struct address_space_operations btrfs_aops; static struct address_space_operations btrfs_symlink_aops; @@ -75,6 +76,7 @@ void btrfs_read_locked_inode(struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_key location; u64 alloc_group_block; + u32 rdev; int ret; path = btrfs_alloc_path(); @@ -104,6 +106,8 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime); inode->i_blocks = btrfs_inode_nblocks(inode_item); inode->i_generation = btrfs_inode_generation(inode_item); + inode->i_rdev = 0; + rdev = btrfs_inode_rdev(inode_item); alloc_group_block = btrfs_inode_block_group(inode_item); BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, alloc_group_block); @@ -114,12 +118,6 @@ void btrfs_read_locked_inode(struct inode *inode) mutex_unlock(&root->fs_info->fs_mutex); switch (inode->i_mode & S_IFMT) { -#if 0 - default: - init_special_inode(inode, inode->i_mode, - btrfs_inode_rdev(inode_item)); - break; -#endif case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; inode->i_fop = &btrfs_file_operations; @@ -136,6 +134,9 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; break; + default: + init_special_inode(inode, inode->i_mode, rdev); + break; } return; @@ -162,6 +163,7 @@ static void fill_inode_item(struct btrfs_inode_item *item, btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec); btrfs_set_inode_nblocks(item, inode->i_blocks); btrfs_set_inode_generation(item, inode->i_generation); + btrfs_set_inode_rdev(item, inode->i_rdev); btrfs_set_inode_block_group(item, BTRFS_I(inode)->block_group->key.objectid); } @@ -1070,6 +1072,58 @@ static int btrfs_add_nondir(struct btrfs_trans_handle *trans, return err; } +static int btrfs_mknod(struct inode *dir, struct dentry *dentry, + int mode, dev_t rdev) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(dir)->root; + struct inode *inode; + int err; + int drop_inode = 0; + u64 objectid; + + if (!new_valid_dev(rdev)) + return -EINVAL; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); + + err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); + if (err) { + err = -ENOSPC; + goto out_unlock; + } + + inode = btrfs_new_inode(trans, root, objectid, + BTRFS_I(dir)->block_group, mode); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_unlock; + + btrfs_set_trans_block_group(trans, inode); + err = btrfs_add_nondir(trans, dentry, inode); + if (err) + drop_inode = 1; + else { + inode->i_op = &btrfs_special_inode_operations; + init_special_inode(inode, inode->i_mode, rdev); + } + dir->i_sb->s_dirt = 1; + btrfs_update_inode_block_group(trans, inode); + btrfs_update_inode_block_group(trans, dir); +out_unlock: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + if (drop_inode) { + inode_dec_link_count(inode); + iput(inode); + } + btrfs_btree_balance_dirty(root); + return err; +} + static int btrfs_create(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd) { @@ -2526,6 +2580,7 @@ static struct inode_operations btrfs_dir_inode_operations = { .rename = btrfs_rename, .symlink = btrfs_symlink, .setattr = btrfs_setattr, + .mknod = btrfs_mknod, }; static struct inode_operations btrfs_dir_ro_inode_operations = { @@ -2562,6 +2617,11 @@ static struct inode_operations btrfs_file_inode_operations = { .setattr = btrfs_setattr, }; +static struct inode_operations btrfs_special_inode_operations = { + .getattr = btrfs_getattr, + .setattr = btrfs_setattr, +}; + static struct inode_operations btrfs_symlink_inode_operations = { .readlink = generic_readlink, .follow_link = page_follow_link_light, -- cgit v1.2.3-70-g09d2 From 92fee66d49b812ea57eae53c5baaf8942ac9af06 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jul 2007 12:31:35 -0400 Subject: Btrfs: deal with api changes in 2.6.23-rc1 Signed-off-by: Chris Mason --- fs/btrfs/file.c | 9 +++++++-- fs/btrfs/inode.c | 50 ++++++++++++++++++++++++++------------------------ 2 files changed, 33 insertions(+), 26 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index f327525f541..1fe38fe8415 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -30,6 +30,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -713,8 +714,12 @@ out: } static struct vm_operations_struct btrfs_file_vm_ops = { - .nopage = filemap_nopage, - .populate = filemap_populate, +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) + .nopage = filemap_nopage, + .populate = filemap_populate, +#else + .fault = filemap_fault, +#endif .page_mkwrite = btrfs_page_mkwrite, }; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f95cf5d4a69..fa9c531c86b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -31,6 +31,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -2323,42 +2324,43 @@ void btrfs_destroy_cachep(void) kmem_cache_destroy(btrfs_path_cachep); } +static struct kmem_cache *cache_create(const char *name, size_t size, + unsigned long extra_flags, + void (*ctor)(void *, struct kmem_cache *, + unsigned long)) +{ + return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT | + SLAB_MEM_SPREAD | extra_flags), ctor +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) + ,NULL +#endif + ); +} + int btrfs_init_cachep(void) { - btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache", - sizeof(struct btrfs_inode), - 0, (SLAB_RECLAIM_ACCOUNT| - SLAB_MEM_SPREAD), - init_once, NULL); + btrfs_inode_cachep = cache_create("btrfs_inode_cache", + sizeof(struct btrfs_inode), + 0, init_once); if (!btrfs_inode_cachep) goto fail; - btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache", + btrfs_trans_handle_cachep = cache_create("btrfs_trans_handle_cache", sizeof(struct btrfs_trans_handle), - 0, (SLAB_RECLAIM_ACCOUNT| - SLAB_MEM_SPREAD), - NULL, NULL); + 0, NULL); if (!btrfs_trans_handle_cachep) goto fail; - btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache", + btrfs_transaction_cachep = cache_create("btrfs_transaction_cache", sizeof(struct btrfs_transaction), - 0, (SLAB_RECLAIM_ACCOUNT| - SLAB_MEM_SPREAD), - NULL, NULL); + 0, NULL); if (!btrfs_transaction_cachep) goto fail; - btrfs_path_cachep = kmem_cache_create("btrfs_path_cache", - sizeof(struct btrfs_transaction), - 0, (SLAB_RECLAIM_ACCOUNT| - SLAB_MEM_SPREAD), - NULL, NULL); + btrfs_path_cachep = cache_create("btrfs_path_cache", + sizeof(struct btrfs_transaction), + 0, NULL); if (!btrfs_path_cachep) goto fail; - btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix", - 256, - 0, (SLAB_RECLAIM_ACCOUNT| - SLAB_MEM_SPREAD | - SLAB_DESTROY_BY_RCU), - NULL, NULL); + btrfs_bit_radix_cachep = cache_create("btrfs_radix", 256, + SLAB_DESTROY_BY_RCU, NULL); if (!btrfs_bit_radix_cachep) goto fail; return 0; -- cgit v1.2.3-70-g09d2 From 3c69faecb8d83cb2ef085a98b196a3fecea67725 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 7 Aug 2007 15:52:22 -0400 Subject: Btrfs: Fold some btree readahead routines into something more generic. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/ctree.h | 1 + fs/btrfs/extent-tree.c | 30 +------------------- fs/btrfs/inode.c | 69 ++------------------------------------------ 4 files changed, 81 insertions(+), 96 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index aa824e2c521..7a08491e208 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -639,6 +639,73 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, return 1; } +/* + * readahead one full node of leaves + */ +static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, + int slot) +{ + struct btrfs_node *node; + int i; + u32 nritems; + u64 item_objectid; + u64 blocknr; + u64 search; + u64 cluster_start; + int ret; + int nread = 0; + int direction = path->reada; + struct radix_tree_root found; + unsigned long gang[8]; + struct buffer_head *bh; + + if (!path->nodes[1]) + return; + + node = btrfs_buffer_node(path->nodes[1]); + search = btrfs_node_blockptr(node, slot); + bh = btrfs_find_tree_block(root, search); + if (bh) { + brelse(bh); + return; + } + + init_bit_radix(&found); + nritems = btrfs_header_nritems(&node->header); + for (i = slot; i < nritems; i++) { + item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); + blocknr = btrfs_node_blockptr(node, i); + set_radix_bit(&found, blocknr); + } + if (direction > 0) { + cluster_start = search - 4; + if (cluster_start > search) + cluster_start = 0; + } else + cluster_start = search + 4; + while(1) { + ret = find_first_radix_bit(&found, gang, 0, ARRAY_SIZE(gang)); + if (!ret) + break; + for (i = 0; i < ret; i++) { + blocknr = gang[i]; + clear_radix_bit(&found, blocknr); + if (nread > 64) + continue; + if (direction > 0 && cluster_start <= blocknr && + cluster_start + 8 > blocknr) { + cluster_start = blocknr; + readahead_tree_block(root, blocknr); + nread++; + } else if (direction < 0 && cluster_start >= blocknr && + blocknr + 8 > cluster_start) { + cluster_start = blocknr; + readahead_tree_block(root, blocknr); + nread++; + } + } + } +} /* * look for key in the tree. path is filled in with nodes along the way * if key is found, we return zero and you can find the item in the leaf @@ -660,9 +727,11 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root struct buffer_head *cow_buf; struct btrfs_node *c; struct btrfs_root_item *root_item = &root->root_item; + u64 blocknr; int slot; int ret; int level; + int should_reada = p->reada; u8 lowest_level = 0; if (btrfs_root_refs(root_item) == 0 && root->ref_cows) { @@ -728,7 +797,11 @@ again: /* this is only true while dropping a snapshot */ if (level == lowest_level) break; + blocknr = btrfs_node_blockptr(c, slot); + if (level == 1 && should_reada) + reada_for_search(root, p, slot); b = read_tree_block(root, btrfs_node_blockptr(c, slot)); + } else { struct btrfs_leaf *l = (struct btrfs_leaf *)c; p->slots[level] = slot; @@ -1915,6 +1988,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) blocknr = btrfs_node_blockptr(c_node, slot); if (next) btrfs_block_release(root, next); + if (level == 1 && path->reada) + reada_for_search(root, path, slot); next = read_tree_block(root, blocknr); break; } @@ -1927,6 +2002,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) path->slots[level] = 0; if (!level) break; + if (level == 1 && path->reada) + reada_for_search(root, path, slot); next = read_tree_block(root, btrfs_node_blockptr(btrfs_buffer_node(next), 0)); } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 73c2e75a136..c5a18d5d7f7 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -177,6 +177,7 @@ struct btrfs_node { struct btrfs_path { struct buffer_head *nodes[BTRFS_MAX_LEVEL]; int slots[BTRFS_MAX_LEVEL]; + int reada; }; /* diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 9455974dabe..5d4d5d8db8e 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -32,33 +32,6 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct static int del_pending_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); -static void reada_extent_leaves(struct btrfs_root *root, - struct btrfs_path *path, u64 limit) -{ - struct btrfs_node *node; - int i; - int nritems; - u64 item_objectid; - u64 blocknr; - int slot; - int ret; - - if (!path->nodes[1]) - return; - node = btrfs_buffer_node(path->nodes[1]); - slot = path->slots[1] + 1; - nritems = btrfs_header_nritems(&node->header); - for (i = slot; i < nritems && i < slot + 8; i++) { - item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); - if (item_objectid > limit) - break; - blocknr = btrfs_node_blockptr(node, i); - ret = readahead_tree_block(root, blocknr); - if (ret) - break; - } -} - static int cache_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *block_group) { @@ -84,6 +57,7 @@ static int cache_block_group(struct btrfs_root *root, path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = 1; key.objectid = block_group->key.objectid; key.flags = 0; key.offset = 0; @@ -94,12 +68,10 @@ static int cache_block_group(struct btrfs_root *root, if (ret && path->slots[0] > 0) path->slots[0]--; limit = block_group->key.objectid + block_group->key.offset; - reada_extent_leaves(root, path, limit); while(1) { leaf = btrfs_buffer_leaf(path->nodes[0]); slot = path->slots[0]; if (slot >= btrfs_header_nritems(&leaf->header)) { - reada_extent_leaves(root, path, limit); ret = btrfs_next_leaf(root, path); if (ret < 0) goto err; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index fa9c531c86b..3889032fc44 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -374,40 +374,6 @@ static int btrfs_free_inode(struct btrfs_trans_handle *trans, return ret; } -/* - * truncates go from a high offset to a low offset. So, walk - * from hi to lo in the node and issue readas. Stop when you find - * keys from a different objectid - */ -static void reada_truncate(struct btrfs_root *root, struct btrfs_path *path, - u64 objectid) -{ - struct btrfs_node *node; - int i; - int nritems; - u64 item_objectid; - u64 blocknr; - int slot; - int ret; - - if (!path->nodes[1]) - return; - node = btrfs_buffer_node(path->nodes[1]); - slot = path->slots[1]; - if (slot == 0) - return; - nritems = btrfs_header_nritems(&node->header); - for (i = slot - 1; i >= 0; i--) { - item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); - if (item_objectid != objectid) - break; - blocknr = btrfs_node_blockptr(node, i); - ret = readahead_tree_block(root, blocknr); - if (ret) - break; - } -} - /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find @@ -434,6 +400,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int del_item; path = btrfs_alloc_path(); + path->reada = -1; BUG_ON(!path); /* FIXME, add redo link to tree so we don't leak on crash */ key.objectid = inode->i_ino; @@ -450,7 +417,6 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, BUG_ON(path->slots[0] == 0); path->slots[0]--; } - reada_truncate(root, path, inode->i_ino); leaf = btrfs_buffer_leaf(path->nodes[0]); found_key = &leaf->items[path->slots[0]].key; found_type = btrfs_disk_key_type(found_key); @@ -827,36 +793,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, return d_splice_alias(inode, dentry); } -/* - * readahead one full node of leaves as long as their keys include - * the objectid supplied - */ -static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path, - u64 objectid) -{ - struct btrfs_node *node; - int i; - u32 nritems; - u64 item_objectid; - u64 blocknr; - int slot; - int ret; - - if (!path->nodes[1]) - return; - node = btrfs_buffer_node(path->nodes[1]); - slot = path->slots[1]; - nritems = btrfs_header_nritems(&node->header); - for (i = slot + 1; i < nritems; i++) { - item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); - if (item_objectid != objectid) - break; - blocknr = btrfs_node_blockptr(node, i); - ret = readahead_tree_block(root, blocknr); - if (ret) - break; - } -} static unsigned char btrfs_filetype_table[] = { DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK }; @@ -890,18 +826,17 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; path = btrfs_alloc_path(); + path->reada = 1; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; advance = 0; - reada_leaves(root, path, inode->i_ino); while(1) { leaf = btrfs_buffer_leaf(path->nodes[0]); nritems = btrfs_header_nritems(&leaf->header); slot = path->slots[0]; if (advance || slot >= nritems) { if (slot >= nritems -1) { - reada_leaves(root, path, inode->i_ino); ret = btrfs_next_leaf(root, path); if (ret) break; -- cgit v1.2.3-70-g09d2 From 6702ed490ca0bb44e17131818a5a18b773957c5a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 7 Aug 2007 16:15:09 -0400 Subject: Btrfs: Add run time btree defrag, and an ioctl to force btree defrag This adds two types of btree defrag, a run time form that tries to defrag recently allocated blocks in the btree when they are still in ram, and an ioctl that forces defrag of all btree blocks. File data blocks are not defragged yet, but this can make a huge difference in sequential btree reads. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.c | 195 +++++++++++++++++++++++++++++++++---------- fs/btrfs/ctree.h | 19 ++++- fs/btrfs/disk-io.c | 8 ++ fs/btrfs/extent-tree.c | 53 +++++++----- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 33 +++++++- fs/btrfs/ioctl.h | 2 +- fs/btrfs/transaction.c | 105 +++++++++++++++++++---- fs/btrfs/transaction.h | 1 + fs/btrfs/tree-defrag.c | 222 +++++++++++++++++++++++++++++++++++++++++++++++++ 11 files changed, 553 insertions(+), 89 deletions(-) create mode 100644 fs/btrfs/tree-defrag.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index a4e2df66bdb..9321438c5ee 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -4,7 +4,7 @@ ifneq ($(KERNELRELEASE),) obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ - transaction.o bit-radix.o inode.o file.o + transaction.o bit-radix.o inode.o file.o tree-defrag.o #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \ # root-tree.o dir-item.o hash.o file-item.o inode-item.o \ diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 7a08491e208..c7e47e77723 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -65,44 +65,44 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) memset(p, 0, sizeof(*p)); } -static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root +static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *buf, struct buffer_head *parent, int parent_slot, struct buffer_head - **cow_ret) + **cow_ret, u64 search_start, u64 empty_size) { struct buffer_head *cow; struct btrfs_node *cow_node; - int ret; + int ret = 0; + int different_trans = 0; + WARN_ON(root->ref_cows && trans->transid != root->last_trans); WARN_ON(!buffer_uptodate(buf)); - if (trans->transaction != root->fs_info->running_transaction) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, - root->fs_info->running_transaction->transid); - WARN_ON(1); - } - if (trans->transid != root->fs_info->generation) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, - root->fs_info->generation); - WARN_ON(1); - } - if (btrfs_header_generation(btrfs_buffer_header(buf)) == - trans->transid) { - *cow_ret = buf; - return 0; - } - cow = btrfs_alloc_free_block(trans, root, buf->b_blocknr); + cow = btrfs_alloc_free_block(trans, root, search_start, empty_size); if (IS_ERR(cow)) return PTR_ERR(cow); + cow_node = btrfs_buffer_node(cow); if (buf->b_size != root->blocksize || cow->b_size != root->blocksize) WARN_ON(1); + memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize); btrfs_set_header_blocknr(&cow_node->header, bh_blocknr(cow)); btrfs_set_header_generation(&cow_node->header, trans->transid); btrfs_set_header_owner(&cow_node->header, root->root_key.objectid); - ret = btrfs_inc_ref(trans, root, buf); - if (ret) - return ret; + + WARN_ON(btrfs_header_generation(btrfs_buffer_header(buf)) > + trans->transid); + if (btrfs_header_generation(btrfs_buffer_header(buf)) != + trans->transid) { + different_trans = 1; + ret = btrfs_inc_ref(trans, root, buf); + if (ret) + return ret; + } else { + WARN_ON(!root->ref_cows); + clean_tree_block(trans, root, buf); + } + if (buf == root->node) { root->node = cow; get_bh(cow); @@ -114,6 +114,8 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot, bh_blocknr(cow)); btrfs_mark_buffer_dirty(parent); + WARN_ON(btrfs_header_generation(btrfs_buffer_header(parent)) != + trans->transid); btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1); } btrfs_block_release(root, buf); @@ -122,6 +124,115 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root return 0; } +int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root + *root, struct buffer_head *buf, struct buffer_head + *parent, int parent_slot, struct buffer_head + **cow_ret) +{ + u64 search_start; + if (trans->transaction != root->fs_info->running_transaction) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->running_transaction->transid); + WARN_ON(1); + } + if (trans->transid != root->fs_info->generation) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->generation); + WARN_ON(1); + } + if (btrfs_header_generation(btrfs_buffer_header(buf)) == + trans->transid) { + *cow_ret = buf; + return 0; + } + + search_start = bh_blocknr(buf) & ~((u64)65535); + return __btrfs_cow_block(trans, root, buf, parent, + parent_slot, cow_ret, search_start, 0); +} + +static int close_blocks(u64 blocknr, u64 other) +{ + if (blocknr < other && other - blocknr < 8) + return 1; + if (blocknr > other && blocknr - other < 8) + return 1; + return 0; +} + +int btrfs_realloc_node(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct buffer_head *parent, + int cache_only) +{ + struct btrfs_node *parent_node; + struct buffer_head *cur_bh; + struct buffer_head *tmp_bh; + u64 blocknr; + u64 search_start = 0; + u64 other; + u32 parent_nritems; + int start_slot; + int end_slot; + int i; + int err = 0; + + if (trans->transaction != root->fs_info->running_transaction) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->running_transaction->transid); + WARN_ON(1); + } + if (trans->transid != root->fs_info->generation) { + printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + root->fs_info->generation); + WARN_ON(1); + } + parent_node = btrfs_buffer_node(parent); + parent_nritems = btrfs_header_nritems(&parent_node->header); + + start_slot = 0; + end_slot = parent_nritems; + + if (parent_nritems == 1) + return 0; + + for (i = start_slot; i < end_slot; i++) { + int close = 1; + blocknr = btrfs_node_blockptr(parent_node, i); + if (i > 0) { + other = btrfs_node_blockptr(parent_node, i - 1); + close = close_blocks(blocknr, other); + } + if (close && i < end_slot - 1) { + other = btrfs_node_blockptr(parent_node, i + 1); + close = close_blocks(blocknr, other); + } + if (close) + continue; + + cur_bh = btrfs_find_tree_block(root, blocknr); + if (!cur_bh || !buffer_uptodate(cur_bh) || + buffer_locked(cur_bh)) { + if (cache_only) { + brelse(cur_bh); + continue; + } + brelse(cur_bh); + cur_bh = read_tree_block(root, blocknr); + } + if (search_start == 0) { + search_start = bh_blocknr(cur_bh) & ~((u64)65535); + } + err = __btrfs_cow_block(trans, root, cur_bh, parent, i, + &tmp_bh, search_start, + min(8, end_slot - i)); + if (err) + break; + search_start = bh_blocknr(tmp_bh); + brelse(tmp_bh); + } + return err; +} + /* * The leaf data grows from end-to-front in the node. * this returns the address of the start of the last item, @@ -221,6 +332,7 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, parent_slot = path->slots[level + 1]; parent_key = &parent->ptrs[parent_slot].key; + BUG_ON(memcmp(parent_key, &leaf->items[0].key, sizeof(struct btrfs_disk_key))); BUG_ON(btrfs_node_blockptr(parent, parent_slot) != @@ -643,7 +755,7 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, * readahead one full node of leaves */ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, - int slot) + int level, int slot) { struct btrfs_node *node; int i; @@ -659,10 +771,13 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, unsigned long gang[8]; struct buffer_head *bh; - if (!path->nodes[1]) + if (level == 0) + return; + + if (!path->nodes[level]) return; - node = btrfs_buffer_node(path->nodes[1]); + node = btrfs_buffer_node(path->nodes[level]); search = btrfs_node_blockptr(node, slot); bh = btrfs_find_tree_block(root, search); if (bh) { @@ -690,7 +805,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, for (i = 0; i < ret; i++) { blocknr = gang[i]; clear_radix_bit(&found, blocknr); - if (nread > 64) + if (nread > 32) continue; if (direction > 0 && cluster_start <= blocknr && cluster_start + 8 > blocknr) { @@ -726,7 +841,6 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root struct buffer_head *b; struct buffer_head *cow_buf; struct btrfs_node *c; - struct btrfs_root_item *root_item = &root->root_item; u64 blocknr; int slot; int ret; @@ -734,11 +848,8 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root int should_reada = p->reada; u8 lowest_level = 0; - if (btrfs_root_refs(root_item) == 0 && root->ref_cows) { - lowest_level = root_item->drop_level; - WARN_ON(ins_len || cow); - } - + lowest_level = p->lowest_level; + WARN_ON(lowest_level && ins_len); WARN_ON(p->nodes[0] != NULL); WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex)); again: @@ -798,8 +909,8 @@ again: if (level == lowest_level) break; blocknr = btrfs_node_blockptr(c, slot); - if (level == 1 && should_reada) - reada_for_search(root, p, slot); + if (should_reada) + reada_for_search(root, p, level, slot); b = read_tree_block(root, btrfs_node_blockptr(c, slot)); } else { @@ -960,7 +1071,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(path->nodes[level]); BUG_ON(path->nodes[level-1] != root->node); - t = btrfs_alloc_free_block(trans, root, root->node->b_blocknr); + t = btrfs_alloc_free_block(trans, root, root->node->b_blocknr, 0); if (IS_ERR(t)) return PTR_ERR(t); c = btrfs_buffer_node(t); @@ -1070,7 +1181,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root } c_nritems = btrfs_header_nritems(&c->header); - split_buffer = btrfs_alloc_free_block(trans, root, t->b_blocknr); + split_buffer = btrfs_alloc_free_block(trans, root, t->b_blocknr, 0); if (IS_ERR(split_buffer)) return PTR_ERR(split_buffer); @@ -1461,7 +1572,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root nritems = btrfs_header_nritems(&l->header); mid = (nritems + 1)/ 2; - right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr); + right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0); if (IS_ERR(right_buffer)) return PTR_ERR(right_buffer); @@ -1560,7 +1671,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root if (!double_split) return ret; - right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr); + right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0); if (IS_ERR(right_buffer)) return PTR_ERR(right_buffer); @@ -1988,8 +2099,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) blocknr = btrfs_node_blockptr(c_node, slot); if (next) btrfs_block_release(root, next); - if (level == 1 && path->reada) - reada_for_search(root, path, slot); + if (path->reada) + reada_for_search(root, path, level, slot); next = read_tree_block(root, blocknr); break; } @@ -2002,8 +2113,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) path->slots[level] = 0; if (!level) break; - if (level == 1 && path->reada) - reada_for_search(root, path, slot); + if (path->reada) + reada_for_search(root, path, level, slot); next = read_tree_block(root, btrfs_node_blockptr(btrfs_buffer_node(next), 0)); } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index c5a18d5d7f7..42aa20301bc 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -178,6 +178,7 @@ struct btrfs_path { struct buffer_head *nodes[BTRFS_MAX_LEVEL]; int slots[BTRFS_MAX_LEVEL]; int reada; + int lowest_level; }; /* @@ -338,6 +339,9 @@ struct btrfs_root { u64 highest_inode; u64 last_inode_alloc; int ref_cows; + struct btrfs_key defrag_progress; + int defrag_running; + int defrag_level; }; /* the lower bits in the key flags defines the item type */ @@ -1031,10 +1035,11 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 hint); + struct btrfs_root *root, u64 hint, + u64 empty_size); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 owner, - u64 num_blocks, u64 search_start, + u64 num_blocks, u64 empty_size, u64 search_start, u64 search_end, struct btrfs_key *ins, int data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *buf); @@ -1051,6 +1056,10 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, int btrfs_free_block_groups(struct btrfs_fs_info *info); int btrfs_read_block_groups(struct btrfs_root *root); /* ctree.c */ +int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root + *root, struct buffer_head *buf, struct buffer_head + *parent, int parent_slot, struct buffer_head + **cow_ret); int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u32 data_size); int btrfs_truncate_item(struct btrfs_trans_handle *trans, @@ -1060,6 +1069,9 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_path *p, int ins_len, int cow); +int btrfs_realloc_node(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct buffer_head *parent, + int cache_only); void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); void btrfs_free_path(struct btrfs_path *p); @@ -1171,4 +1183,7 @@ extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 start, u64 end, u64 *hint_block); +/* tree-defrag.c */ +int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int cache_only); #endif diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 60db85bc4ba..c948416fea3 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -273,7 +273,9 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *buf) { WARN_ON(atomic_read(&buf->b_count) == 0); + lock_buffer(buf); clear_buffer_dirty(buf); + unlock_buffer(buf); return 0; } @@ -294,6 +296,9 @@ static int __setup_root(int blocksize, root->last_inode_alloc = 0; memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); + memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); + root->defrag_running = 0; + root->defrag_level = 0; root->root_key.objectid = objectid; return 0; } @@ -585,6 +590,7 @@ int close_ctree(struct btrfs_root *root) fs_info->closing = 1; btrfs_transaction_flush_work(root); mutex_lock(&fs_info->fs_mutex); + btrfs_defrag_dirty_roots(root->fs_info); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); /* run commit again to drop the original snapshot */ @@ -616,7 +622,9 @@ void btrfs_mark_buffer_dirty(struct buffer_head *bh) { struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root; u64 transid = btrfs_header_generation(btrfs_buffer_header(bh)); + WARN_ON(!atomic_read(&bh->b_count)); + if (transid != root->fs_info->generation) { printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", (unsigned long long)bh->b_blocknr, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 5d4d5d8db8e..26b8d340649 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -23,7 +23,8 @@ #include "transaction.h" static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *orig_root, u64 num_blocks, u64 search_start, + *orig_root, u64 num_blocks, u64 empty_size, + u64 search_start, u64 search_end, u64 hint_block, struct btrfs_key *ins, u64 exclude_start, u64 exclude_nr, int data); @@ -379,7 +380,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); if (!path) return -ENOMEM; - ret = find_free_extent(trans, root->fs_info->extent_root, 0, 0, + ret = find_free_extent(trans, root->fs_info->extent_root, 0, 0, 0, (u64)-1, 0, &ins, 0, 0, 0); if (ret) { btrfs_free_path(path); @@ -533,7 +534,7 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, struct btrfs_block_group_item *bi; struct btrfs_key ins; - ret = find_free_extent(trans, extent_root, 0, 0, (u64)-1, 0, &ins, + ret = find_free_extent(trans, extent_root, 0, 0, 0, (u64)-1, 0, &ins, 0, 0, 0); /* FIXME, set bit to recalc cache groups on next mount */ if (ret) @@ -708,6 +709,7 @@ static int update_block_group(struct btrfs_trans_handle *trans, static int try_remove_page(struct address_space *mapping, unsigned long index) { int ret; + return 0; ret = invalidate_mapping_pages(mapping, index, index); return ret; } @@ -866,7 +868,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root if (!path) return -ENOMEM; - ret = find_free_extent(trans, root, 0, 0, (u64)-1, 0, &ins, 0, 0, 0); + ret = find_free_extent(trans, root, 0, 0, 0, (u64)-1, 0, &ins, 0, 0, 0); if (ret) { btrfs_free_path(path); return ret; @@ -983,8 +985,8 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root * Any available blocks before search_start are skipped. */ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *orig_root, u64 num_blocks, u64 search_start, u64 - search_end, u64 hint_block, + *orig_root, u64 num_blocks, u64 empty_size, + u64 search_start, u64 search_end, u64 hint_block, struct btrfs_key *ins, u64 exclude_start, u64 exclude_nr, int data) { @@ -1042,6 +1044,7 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root data, 1); } + total_needed += empty_size; path = btrfs_alloc_path(); check_failed: @@ -1157,9 +1160,11 @@ check_pending: goto error; } search_start = orig_search_start; - if (wrapped) + if (wrapped) { + if (!full_scan) + total_needed -= empty_size; full_scan = 1; - else + } else wrapped = 1; goto new_group; } @@ -1238,9 +1243,11 @@ new_group: ret = -ENOSPC; goto error; } - if (wrapped) + if (wrapped) { + if (!full_scan) + total_needed -= empty_size; full_scan = 1; - else + } else wrapped = 1; } block_group = btrfs_lookup_block_group(info, search_start); @@ -1264,7 +1271,7 @@ error: */ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 owner, - u64 num_blocks, u64 hint_block, + u64 num_blocks, u64 empty_size, u64 hint_block, u64 search_end, struct btrfs_key *ins, int data) { int ret; @@ -1303,7 +1310,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, * in the correct block group. */ if (data) { - ret = find_free_extent(trans, root, 0, 0, + ret = find_free_extent(trans, root, 0, 0, 0, search_end, 0, &prealloc_key, 0, 0, 0); BUG_ON(ret); if (ret) @@ -1313,8 +1320,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, } /* do the real allocation */ - ret = find_free_extent(trans, root, num_blocks, search_start, - search_end, hint_block, ins, + ret = find_free_extent(trans, root, num_blocks, empty_size, + search_start, search_end, hint_block, ins, exclude_start, exclude_nr, data); BUG_ON(ret); if (ret) @@ -1333,7 +1340,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, exclude_start = ins->objectid; exclude_nr = ins->offset; hint_block = exclude_start + exclude_nr; - ret = find_free_extent(trans, root, 0, search_start, + ret = find_free_extent(trans, root, 0, 0, search_start, search_end, hint_block, &prealloc_key, exclude_start, exclude_nr, 0); @@ -1368,14 +1375,16 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, * returns the tree buffer or NULL. */ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 hint) + struct btrfs_root *root, u64 hint, + u64 empty_size) { struct btrfs_key ins; int ret; struct buffer_head *buf; ret = btrfs_alloc_extent(trans, root, root->root_key.objectid, - 1, hint, (unsigned long)-1, &ins, 0); + 1, empty_size, hint, + (unsigned long)-1, &ins, 0); if (ret) { BUG_ON(ret > 0); return ERR_PTR(ret); @@ -1385,6 +1394,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, btrfs_free_extent(trans, root, ins.objectid, 1, 0); return ERR_PTR(-ENOMEM); } + WARN_ON(buffer_dirty(buf)); set_buffer_uptodate(buf); set_buffer_checked(buf); set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index); @@ -1591,13 +1601,15 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_key key; struct btrfs_disk_key *found_key; struct btrfs_node *node; + btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); + level = root_item->drop_level; + path->lowest_level = level; wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); - if (ret < 0) { + if (wret < 0) { ret = wret; goto out; } - level = root_item->drop_level; node = btrfs_buffer_node(path->nodes[level]); found_key = &node->ptrs[path->slots[level]].key; WARN_ON(memcmp(found_key, &root_item->drop_progress, @@ -1617,8 +1629,6 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root ret = wret; num_walks++; if (num_walks > 10) { - struct btrfs_key key; - btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); ret = -EAGAIN; get_bh(root->node); break; @@ -1627,6 +1637,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root for (i = 0; i <= orig_level; i++) { if (path->nodes[i]) { btrfs_block_release(root, path->nodes[i]); + path->nodes[i] = 0; } } out: diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1fe38fe8415..00b118a2db6 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -512,7 +512,7 @@ static int prepare_pages(struct btrfs_root *root, if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { err = btrfs_alloc_extent(trans, root, inode->i_ino, - num_blocks, hint_block, (u64)-1, + num_blocks, 0, hint_block, (u64)-1, &ins, 1); if (err) goto failed_truncate; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3889032fc44..12aa043b9f6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -554,7 +554,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) &alloc_hint); if (ret) goto out; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, + ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) goto out; @@ -1360,7 +1360,7 @@ not_found: if (create & BTRFS_GET_BLOCK_CREATE) { struct btrfs_key ins; ret = btrfs_alloc_extent(trans, root, inode->i_ino, - 1, alloc_hint, (u64)-1, + 1, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { err = ret; @@ -1998,7 +1998,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - subvol = btrfs_alloc_free_block(trans, root, 0); + subvol = btrfs_alloc_free_block(trans, root, 0, 0); if (IS_ERR(subvol)) return PTR_ERR(subvol); leaf = btrfs_buffer_leaf(subvol); @@ -2159,7 +2159,9 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_ioctl_vol_args vol_args; + struct btrfs_trans_handle *trans; int ret = 0; + int err; struct btrfs_dir_item *di; int namelen; struct btrfs_path *path; @@ -2196,6 +2198,31 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int else ret = create_snapshot(root, vol_args.name, namelen); break; + + case BTRFS_IOC_DEFRAG: + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + memset(&root->defrag_progress, 0, + sizeof(root->defrag_progress)); + while (1) { + root->defrag_running = 1; + err = btrfs_defrag_leaves(trans, root, 0); + + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + + btrfs_btree_balance_dirty(root); + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + if (err != -EAGAIN) + break; + } + root->defrag_running = 0; + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + ret = 0; + break; default: return -ENOTTY; } diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 23bed48c51a..8bc47dec286 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -28,6 +28,6 @@ struct btrfs_ioctl_vol_args { #define BTRFS_IOC_SNAP_CREATE _IOW(BTRFS_IOCTL_MAGIC, 1, \ struct btrfs_ioctl_vol_args) -#define BTRFS_IOC_ADD_DISK _IOW(BTRFS_IOCTL_MAGIC, 2, \ +#define BTRFS_IOC_DEFRAG _IOW(BTRFS_IOCTL_MAGIC, 2, \ struct btrfs_ioctl_vol_args) #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 498626470a0..338a7199363 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -29,6 +29,7 @@ extern struct kmem_cache *btrfs_transaction_cachep; static struct workqueue_struct *trans_wq; #define BTRFS_ROOT_TRANS_TAG 0 +#define BTRFS_ROOT_DEFRAG_TAG 1 static void put_transaction(struct btrfs_transaction *transaction) { @@ -69,35 +70,41 @@ static int join_transaction(struct btrfs_root *root) return 0; } +static int record_root_in_trans(struct btrfs_root *root) +{ + u64 running_trans_id = root->fs_info->running_transaction->transid; + if (root->ref_cows && root->last_trans < running_trans_id) { + WARN_ON(root == root->fs_info->extent_root); + if (root->root_item.refs != 0) { + radix_tree_tag_set(&root->fs_info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_TRANS_TAG); + radix_tree_tag_set(&root->fs_info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_DEFRAG_TAG); + root->commit_root = root->node; + get_bh(root->node); + } else { + WARN_ON(1); + } + root->last_trans = running_trans_id; + } + return 0; +} + struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, int num_blocks) { struct btrfs_trans_handle *h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); int ret; - u64 running_trans_id; mutex_lock(&root->fs_info->trans_mutex); ret = join_transaction(root); BUG_ON(ret); - running_trans_id = root->fs_info->running_transaction->transid; - if (root != root->fs_info->tree_root && root->last_trans < - running_trans_id) { - WARN_ON(root == root->fs_info->extent_root); - WARN_ON(root->ref_cows != 1); - if (root->root_item.refs != 0) { - radix_tree_tag_set(&root->fs_info->fs_roots_radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_TRANS_TAG); - root->commit_root = root->node; - get_bh(root->node); - } else { - WARN_ON(1); - } - } - root->last_trans = running_trans_id; - h->transid = running_trans_id; + record_root_in_trans(root); + h->transid = root->fs_info->running_transaction->transid; h->transaction = root->fs_info->running_transaction; h->blocks_reserved = num_blocks; h->blocks_used = 0; @@ -155,6 +162,15 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, gang[i]); if (!page) continue; + if (PageWriteback(page)) { + if (PageDirty(page)) + wait_on_page_writeback(page); + else { + unlock_page(page); + page_cache_release(page); + continue; + } + } err = write_one_page(page, 0); if (err) werr = err; @@ -299,6 +315,58 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, return err; } +int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info) +{ + struct btrfs_root *gang[1]; + struct btrfs_root *root; + struct btrfs_root *tree_root = info->tree_root; + struct btrfs_trans_handle *trans; + int i; + int ret; + int err = 0; + u64 last = 0; + + trans = btrfs_start_transaction(tree_root, 1); + while(1) { + ret = radix_tree_gang_lookup_tag(&info->fs_roots_radix, + (void **)gang, last, + ARRAY_SIZE(gang), + BTRFS_ROOT_DEFRAG_TAG); + if (ret == 0) + break; + for (i = 0; i < ret; i++) { + root = gang[i]; + last = root->root_key.objectid + 1; + radix_tree_tag_clear(&info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_DEFRAG_TAG); + if (root->defrag_running) + continue; + + while (1) { + mutex_lock(&root->fs_info->trans_mutex); + record_root_in_trans(root); + mutex_unlock(&root->fs_info->trans_mutex); + + root->defrag_running = 1; + err = btrfs_defrag_leaves(trans, root, 1); + btrfs_end_transaction(trans, tree_root); + mutex_unlock(&info->fs_mutex); + + btrfs_btree_balance_dirty(root); + + mutex_lock(&info->fs_mutex); + trans = btrfs_start_transaction(tree_root, 1); + if (err != -EAGAIN) + break; + } + root->defrag_running = 0; + } + } + btrfs_end_transaction(trans, tree_root); + return err; +} + static int drop_dirty_roots(struct btrfs_root *tree_root, struct list_head *list) { @@ -475,6 +543,7 @@ void btrfs_transaction_cleaner(struct work_struct *work) goto out; } mutex_unlock(&root->fs_info->trans_mutex); + btrfs_defrag_dirty_roots(root->fs_info); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); out: diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index ebf44f3e111..8b2714e6556 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -70,5 +70,6 @@ void btrfs_transaction_queue_work(struct btrfs_root *root, int delay); void btrfs_init_transaction_sys(void); void btrfs_exit_transaction_sys(void); int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list); +int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info); #endif diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c new file mode 100644 index 00000000000..15d0a486fb5 --- /dev/null +++ b/fs/btrfs/tree-defrag.c @@ -0,0 +1,222 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include "ctree.h" +#include "disk-io.h" +#include "print-tree.h" +#include "transaction.h" + +static void reada_defrag(struct btrfs_root *root, + struct btrfs_node *node) +{ + int i; + u32 nritems; + u64 blocknr; + int ret; + + nritems = btrfs_header_nritems(&node->header); + for (i = 0; i < nritems; i++) { + blocknr = btrfs_node_blockptr(node, i); + ret = readahead_tree_block(root, blocknr); + if (ret) + break; + } +} + +static int defrag_walk_down(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int *level, + int cache_only) +{ + struct buffer_head *next; + struct buffer_head *cur; + u64 blocknr; + int ret = 0; + + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + + while(*level > 0) { + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + cur = path->nodes[*level]; + + if (!cache_only && *level > 1 && path->slots[*level] == 0) + reada_defrag(root, btrfs_buffer_node(cur)); + + if (btrfs_header_level(btrfs_buffer_header(cur)) != *level) + WARN_ON(1); + + if (path->slots[*level] >= + btrfs_header_nritems(btrfs_buffer_header(cur))) + break; + + if (*level == 1) { + ret = btrfs_realloc_node(trans, root, + path->nodes[*level], + cache_only); + break; + } + blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur), + path->slots[*level]); + + if (cache_only) { + next = btrfs_find_tree_block(root, blocknr); + if (!next || !buffer_uptodate(next) || + buffer_locked(next)) { + brelse(next); + path->slots[*level]++; + continue; + } + } else { + next = read_tree_block(root, blocknr); + } + ret = btrfs_cow_block(trans, root, next, path->nodes[*level], + path->slots[*level], &next); + BUG_ON(ret); + ret = btrfs_realloc_node(trans, root, next, cache_only); + BUG_ON(ret); + WARN_ON(*level <= 0); + if (path->nodes[*level-1]) + btrfs_block_release(root, path->nodes[*level-1]); + path->nodes[*level-1] = next; + *level = btrfs_header_level(btrfs_buffer_header(next)); + path->slots[*level] = 0; + } + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + btrfs_block_release(root, path->nodes[*level]); + path->nodes[*level] = NULL; + *level += 1; + WARN_ON(ret); + return 0; +} + +static int defrag_walk_up(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int *level, + int cache_only) +{ + int i; + int slot; + struct btrfs_node *node; + + for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { + slot = path->slots[i]; + if (slot < btrfs_header_nritems( + btrfs_buffer_header(path->nodes[i])) - 1) { + path->slots[i]++; + *level = i; + node = btrfs_buffer_node(path->nodes[i]); + WARN_ON(i == 0); + btrfs_disk_key_to_cpu(&root->defrag_progress, + &node->ptrs[path->slots[i]].key); + root->defrag_level = i; + return 0; + } else { + btrfs_block_release(root, path->nodes[*level]); + path->nodes[*level] = NULL; + *level = i + 1; + } + } + return 1; +} + +int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int cache_only) +{ + struct btrfs_path *path = NULL; + struct buffer_head *tmp; + int ret = 0; + int wret; + int level; + int orig_level; + int i; + int num_runs = 0; + + if (root->ref_cows == 0) { + goto out; + } + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + level = btrfs_header_level(btrfs_buffer_header(root->node)); + orig_level = level; + if (level == 0) { + goto out; + } + if (root->defrag_progress.objectid == 0) { + get_bh(root->node); + ret = btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); + BUG_ON(ret); + ret = btrfs_realloc_node(trans, root, root->node, cache_only); + BUG_ON(ret); + path->nodes[level] = root->node; + path->slots[level] = 0; + } else { + level = root->defrag_level; + path->lowest_level = level; + wret = btrfs_search_slot(trans, root, &root->defrag_progress, + path, 0, 1); + + if (wret < 0) { + ret = wret; + goto out; + } + while(level > 0 && !path->nodes[level]) + level--; + if (!path->nodes[level]) { + ret = 0; + goto out; + } + } + + while(1) { + wret = defrag_walk_down(trans, root, path, &level, cache_only); + if (wret > 0) + break; + if (wret < 0) + ret = wret; + + wret = defrag_walk_up(trans, root, path, &level, cache_only); + if (wret > 0) + break; + if (wret < 0) + ret = wret; + if (num_runs++ > 8) { + ret = -EAGAIN; + break; + } + } + for (i = 0; i <= orig_level; i++) { + if (path->nodes[i]) { + btrfs_block_release(root, path->nodes[i]); + path->nodes[i] = 0; + } + } +out: + if (path) + btrfs_free_path(path); + if (ret != -EAGAIN) { + memset(&root->defrag_progress, 0, + sizeof(root->defrag_progress)); + } + return ret; +} -- cgit v1.2.3-70-g09d2 From e9d0b13b5bbb58c9b840e407a8d181442f799966 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 10 Aug 2007 14:06:19 -0400 Subject: Btrfs: Btree defrag on the extent-mapping tree as well Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 18 ++++++---- fs/btrfs/ctree.h | 4 ++- fs/btrfs/extent-tree.c | 28 +++++++++++++++- fs/btrfs/inode.c | 23 ++----------- fs/btrfs/super.c | 2 ++ fs/btrfs/transaction.c | 91 +++++++++++++++++++++++++++----------------------- fs/btrfs/transaction.h | 2 ++ fs/btrfs/tree-defrag.c | 36 ++++++++++++++++---- 8 files changed, 125 insertions(+), 79 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index c7e47e77723..ee1ae00d282 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -99,7 +99,6 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root if (ret) return ret; } else { - WARN_ON(!root->ref_cows); clean_tree_block(trans, root, buf); } @@ -162,13 +161,14 @@ static int close_blocks(u64 blocknr, u64 other) int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *parent, - int cache_only) + int cache_only, u64 *last_ret) { struct btrfs_node *parent_node; struct buffer_head *cur_bh; struct buffer_head *tmp_bh; u64 blocknr; - u64 search_start = 0; + u64 search_start = *last_ret; + u64 last_block = 0; u64 other; u32 parent_nritems; int start_slot; @@ -198,6 +198,8 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, for (i = start_slot; i < end_slot; i++) { int close = 1; blocknr = btrfs_node_blockptr(parent_node, i); + if (last_block == 0) + last_block = blocknr; if (i > 0) { other = btrfs_node_blockptr(parent_node, i - 1); close = close_blocks(blocknr, other); @@ -206,8 +208,10 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, other = btrfs_node_blockptr(parent_node, i + 1); close = close_blocks(blocknr, other); } - if (close) + if (close) { + last_block = blocknr; continue; + } cur_bh = btrfs_find_tree_block(root, blocknr); if (!cur_bh || !buffer_uptodate(cur_bh) || @@ -219,9 +223,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, brelse(cur_bh); cur_bh = read_tree_block(root, blocknr); } - if (search_start == 0) { - search_start = bh_blocknr(cur_bh) & ~((u64)65535); - } + if (search_start == 0) + search_start = last_block & ~((u64)65535); + err = __btrfs_cow_block(trans, root, cur_bh, parent, i, &tmp_bh, search_start, min(8, end_slot - i)); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 59e09e37ab9..d3cd564b3b3 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1019,6 +1019,8 @@ static inline void btrfs_memmove(struct btrfs_root *root, btrfs_item_offset((leaf)->items + (slot)))) /* extent-tree.c */ +int btrfs_extent_post_op(struct btrfs_trans_handle *trans, + struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy); struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, @@ -1066,7 +1068,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root ins_len, int cow); int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *parent, - int cache_only); + int cache_only, u64 *last_ret); void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); void btrfs_free_path(struct btrfs_path *p); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 120c448f58f..3418bb62b99 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -396,6 +396,14 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, return 0; } +int btrfs_extent_post_op(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + finish_current_insert(trans, root->fs_info->extent_root); + del_pending_extents(trans, root->fs_info->extent_root); + return 0; +} + static int lookup_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 blocknr, u64 num_blocks, u32 *refs) @@ -1374,7 +1382,25 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(ret); continue; } - next = read_tree_block(root, blocknr); + next = btrfs_find_tree_block(root, blocknr); + if (!next || !buffer_uptodate(next)) { + brelse(next); + mutex_unlock(&root->fs_info->fs_mutex); + next = read_tree_block(root, blocknr); + mutex_lock(&root->fs_info->fs_mutex); + + /* we dropped the lock, check one more time */ + ret = lookup_extent_ref(trans, root, blocknr, 1, &refs); + BUG_ON(ret); + if (refs != 1) { + path->slots[*level]++; + brelse(next); + ret = btrfs_free_extent(trans, root, + blocknr, 1, 1); + BUG_ON(ret); + continue; + } + } WARN_ON(*level <= 0); if (path->nodes[*level-1]) btrfs_block_release(root, path->nodes[*level-1]); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 12aa043b9f6..5c05ecbc572 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2159,9 +2159,7 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_ioctl_vol_args vol_args; - struct btrfs_trans_handle *trans; int ret = 0; - int err; struct btrfs_dir_item *di; int namelen; struct btrfs_path *path; @@ -2201,25 +2199,8 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int case BTRFS_IOC_DEFRAG: mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - memset(&root->defrag_progress, 0, - sizeof(root->defrag_progress)); - while (1) { - root->defrag_running = 1; - err = btrfs_defrag_leaves(trans, root, 0); - - btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); - - btrfs_btree_balance_dirty(root); - - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - if (err != -EAGAIN) - break; - } - root->defrag_running = 0; - btrfs_end_transaction(trans, root); + btrfs_defrag_root(root, 0); + btrfs_defrag_root(root->fs_info->extent_root, 0); mutex_unlock(&root->fs_info->fs_mutex); ret = 0; break; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 2e797d5fb28..74f3de47423 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -121,7 +121,9 @@ static int btrfs_sync_fs(struct super_block *sb, int wait) filemap_flush(root->fs_info->btree_inode->i_mapping); return 0; } + btrfs_clean_old_snapshots(root); mutex_lock(&root->fs_info->fs_mutex); + btrfs_defrag_dirty_roots(root->fs_info); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); sb->s_dirt = 0; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 204337c5ca0..c9d52dc83e4 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -317,18 +317,47 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, return err; } +int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) +{ + struct btrfs_fs_info *info = root->fs_info; + int ret; + struct btrfs_trans_handle *trans; + + if (root->defrag_running) + return 0; + + trans = btrfs_start_transaction(root, 1); + while (1) { + root->defrag_running = 1; + ret = btrfs_defrag_leaves(trans, root, cacheonly); + btrfs_end_transaction(trans, root); + mutex_unlock(&info->fs_mutex); + + btrfs_btree_balance_dirty(root); + cond_resched(); + + mutex_lock(&info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + if (ret != -EAGAIN) + break; + } + root->defrag_running = 0; + radix_tree_tag_clear(&info->fs_roots_radix, + (unsigned long)root->root_key.objectid, + BTRFS_ROOT_DEFRAG_TAG); + btrfs_end_transaction(trans, root); + return 0; +} + int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info) { struct btrfs_root *gang[1]; struct btrfs_root *root; - struct btrfs_root *tree_root = info->tree_root; - struct btrfs_trans_handle *trans; int i; int ret; int err = 0; u64 last = 0; - trans = btrfs_start_transaction(tree_root, 1); while(1) { ret = radix_tree_gang_lookup_tag(&info->fs_roots_radix, (void **)gang, last, @@ -339,37 +368,10 @@ int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info) for (i = 0; i < ret; i++) { root = gang[i]; last = root->root_key.objectid + 1; - radix_tree_tag_clear(&info->fs_roots_radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_DEFRAG_TAG); - if (root->defrag_running) - continue; - - while (1) { - mutex_lock(&root->fs_info->trans_mutex); - record_root_in_trans(root); - mutex_unlock(&root->fs_info->trans_mutex); - - root->defrag_running = 1; - err = btrfs_defrag_leaves(trans, root, 1); - btrfs_end_transaction(trans, tree_root); - mutex_unlock(&info->fs_mutex); - - btrfs_btree_balance_dirty(root); - cond_resched(); - - mutex_lock(&info->fs_mutex); - trans = btrfs_start_transaction(tree_root, 1); - if (err != -EAGAIN) - break; - } - root->defrag_running = 0; - radix_tree_tag_clear(&info->fs_roots_radix, - (unsigned long)root->root_key.objectid, - BTRFS_ROOT_DEFRAG_TAG); + btrfs_defrag_root(root, 1); } } - btrfs_end_transaction(trans, tree_root); + btrfs_defrag_root(info->extent_root, 1); return err; } @@ -527,6 +529,20 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, return ret; } +int btrfs_clean_old_snapshots(struct btrfs_root *root) +{ + struct list_head dirty_roots; + INIT_LIST_HEAD(&dirty_roots); + + mutex_lock(&root->fs_info->trans_mutex); + list_splice_init(&root->fs_info->dead_roots, &dirty_roots); + mutex_unlock(&root->fs_info->trans_mutex); + + if (!list_empty(&dirty_roots)) { + drop_dirty_roots(root, &dirty_roots); + } + return 0; +} void btrfs_transaction_cleaner(struct work_struct *work) { struct btrfs_fs_info *fs_info = container_of(work, @@ -536,12 +552,10 @@ void btrfs_transaction_cleaner(struct work_struct *work) struct btrfs_root *root = fs_info->tree_root; struct btrfs_transaction *cur; struct btrfs_trans_handle *trans; - struct list_head dirty_roots; unsigned long now; unsigned long delay = HZ * 30; int ret; - INIT_LIST_HEAD(&dirty_roots); mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); cur = root->fs_info->running_transaction; @@ -561,14 +575,7 @@ void btrfs_transaction_cleaner(struct work_struct *work) ret = btrfs_commit_transaction(trans, root); out: mutex_unlock(&root->fs_info->fs_mutex); - - mutex_lock(&root->fs_info->trans_mutex); - list_splice_init(&root->fs_info->dead_roots, &dirty_roots); - mutex_unlock(&root->fs_info->trans_mutex); - - if (!list_empty(&dirty_roots)) { - drop_dirty_roots(root, &dirty_roots); - } + btrfs_clean_old_snapshots(root); btrfs_transaction_queue_work(root, delay); } diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 65a395eeca9..d5f491d3757 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -73,5 +73,7 @@ void btrfs_init_transaction_sys(void); void btrfs_exit_transaction_sys(void); int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list); int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info); +int btrfs_defrag_root(struct btrfs_root *root, int cacheonly); +int btrfs_clean_old_snapshots(struct btrfs_root *root); #endif diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 7ea66b4aa5c..a09064a9a41 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -42,16 +42,20 @@ static void reada_defrag(struct btrfs_root *root, static int defrag_walk_down(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level, - int cache_only) + int cache_only, u64 *last_ret) { struct buffer_head *next; struct buffer_head *cur; u64 blocknr; int ret = 0; + int is_extent = 0; WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); + if (root->fs_info->extent_root == root) + is_extent = 1; + while(*level > 0) { WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); @@ -70,7 +74,10 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, if (*level == 1) { ret = btrfs_realloc_node(trans, root, path->nodes[*level], - cache_only); + cache_only, last_ret); + if (is_extent) + btrfs_extent_post_op(trans, root); + break; } blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur), @@ -90,8 +97,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, ret = btrfs_cow_block(trans, root, next, path->nodes[*level], path->slots[*level], &next); BUG_ON(ret); - ret = btrfs_realloc_node(trans, root, next, cache_only); + ret = btrfs_realloc_node(trans, root, next, cache_only, + last_ret); BUG_ON(ret); + + if (is_extent) + btrfs_extent_post_op(trans, root); + WARN_ON(*level <= 0); if (path->nodes[*level-1]) btrfs_block_release(root, path->nodes[*level-1]); @@ -148,10 +160,14 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, int level; int orig_level; int i; + int is_extent = 0; + u64 last_ret = 0; + + if (root->fs_info->extent_root == root) + is_extent = 1; - if (root->ref_cows == 0) { + if (root->ref_cows == 0 && !is_extent) goto out; - } path = btrfs_alloc_path(); if (!path) return -ENOMEM; @@ -165,16 +181,21 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, get_bh(root->node); ret = btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); BUG_ON(ret); - ret = btrfs_realloc_node(trans, root, root->node, cache_only); + ret = btrfs_realloc_node(trans, root, root->node, cache_only, + &last_ret); BUG_ON(ret); path->nodes[level] = root->node; path->slots[level] = 0; + if (is_extent) + btrfs_extent_post_op(trans, root); } else { level = root->defrag_level; path->lowest_level = level; wret = btrfs_search_slot(trans, root, &root->defrag_progress, path, 0, 1); + if (is_extent) + btrfs_extent_post_op(trans, root); if (wret < 0) { ret = wret; goto out; @@ -188,7 +209,8 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, } while(1) { - wret = defrag_walk_down(trans, root, path, &level, cache_only); + wret = defrag_walk_down(trans, root, path, &level, cache_only, + &last_ret); if (wret > 0) break; if (wret < 0) -- cgit v1.2.3-70-g09d2 From 15ee9bc7ed171248d1405df5854da5fa91bfdc39 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 10 Aug 2007 16:22:09 -0400 Subject: Btrfs: delay commits during fsync to allow more writers Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 5 +++++ fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 1 + fs/btrfs/file.c | 22 ++++++++++++++++++---- fs/btrfs/inode.c | 2 ++ fs/btrfs/transaction.c | 44 ++++++++++++++++++++++++++++++++------------ fs/btrfs/transaction.h | 7 +++++++ 7 files changed, 66 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 6bce46be49f..f446f161f73 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -25,6 +25,11 @@ struct btrfs_inode { struct btrfs_block_group_cache *block_group; struct btrfs_key location; struct inode vfs_inode; + + /* + * transid of the trans_handle that last modified this inode + */ + u64 last_trans; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) { diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index d3cd564b3b3..c7f5161271d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -301,6 +301,7 @@ struct btrfs_fs_info { struct radix_tree_root extent_map_radix; struct radix_tree_root extent_ins_radix; u64 generation; + u64 last_trans_committed; struct btrfs_transaction *running_transaction; struct btrfs_super_block *disk_super; struct btrfs_super_block super_copy; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 192211274ce..b2f79878d51 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -435,6 +435,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) INIT_LIST_HEAD(&fs_info->dead_roots); sb_set_blocksize(sb, 4096); fs_info->running_transaction = NULL; + fs_info->last_trans_committed = 0; fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; fs_info->sb = sb; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 00b118a2db6..6933ab11a5c 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -694,22 +694,36 @@ static int btrfs_sync_file(struct file *file, { struct inode *inode = dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; - int ret; + int ret = 0; struct btrfs_trans_handle *trans; /* - * FIXME, use inode generation number to check if we can skip the - * commit + * check the transaction that last modified this inode + * and see if its already been committed */ mutex_lock(&root->fs_info->fs_mutex); + if (!BTRFS_I(inode)->last_trans) + goto out; + mutex_lock(&root->fs_info->trans_mutex); + if (BTRFS_I(inode)->last_trans <= + root->fs_info->last_trans_committed) { + BTRFS_I(inode)->last_trans = 0; + mutex_unlock(&root->fs_info->trans_mutex); + goto out; + } + mutex_unlock(&root->fs_info->trans_mutex); + + /* + * ok we haven't committed the transaction yet, lets do a commit + */ trans = btrfs_start_transaction(root, 1); if (!trans) { ret = -ENOMEM; goto out; } ret = btrfs_commit_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); out: + mutex_unlock(&root->fs_info->fs_mutex); return ret > 0 ? EIO : ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5c05ecbc572..398484179d8 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -193,6 +193,7 @@ static int btrfs_update_inode(struct btrfs_trans_handle *trans, fill_inode_item(inode_item, inode); btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_set_inode_last_trans(trans, inode); ret = 0; failed: btrfs_release_path(root, path); @@ -2234,6 +2235,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS); if (!ei) return NULL; + ei->last_trans = 0; return &ei->vfs_inode; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index c9d52dc83e4..18abea80279 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -55,7 +55,8 @@ static int join_transaction(struct btrfs_root *root) BUG_ON(!cur_trans); root->fs_info->generation++; root->fs_info->running_transaction = cur_trans; - cur_trans->num_writers = 0; + cur_trans->num_writers = 1; + cur_trans->num_joined = 0; cur_trans->transid = root->fs_info->generation; init_waitqueue_head(&cur_trans->writer_wait); init_waitqueue_head(&cur_trans->commit_wait); @@ -65,8 +66,11 @@ static int join_transaction(struct btrfs_root *root) cur_trans->start_time = get_seconds(); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); init_bit_radix(&cur_trans->dirty_pages); + } else { + cur_trans->num_writers++; + cur_trans->num_joined++; } - cur_trans->num_writers++; + return 0; } @@ -428,12 +432,14 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - int ret = 0; + unsigned long joined = 0; + unsigned long timeout = 1; struct btrfs_transaction *cur_trans; struct btrfs_transaction *prev_trans = NULL; struct list_head dirty_fs_roots; struct radix_tree_root pinned_copy; DEFINE_WAIT(wait); + int ret; init_bit_radix(&pinned_copy); INIT_LIST_HEAD(&dirty_fs_roots); @@ -448,7 +454,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->fs_mutex); ret = wait_for_commit(root, cur_trans); BUG_ON(ret); + + mutex_lock(&root->fs_info->trans_mutex); put_transaction(cur_trans); + mutex_unlock(&root->fs_info->trans_mutex); + mutex_lock(&root->fs_info->fs_mutex); return 0; } @@ -463,26 +473,35 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->trans_mutex); wait_for_commit(root, prev_trans); - put_transaction(prev_trans); mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); + put_transaction(prev_trans); } } - while (trans->transaction->num_writers > 1) { + + do { + joined = cur_trans->num_joined; WARN_ON(cur_trans != trans->transaction); - prepare_to_wait(&trans->transaction->writer_wait, &wait, + prepare_to_wait(&cur_trans->writer_wait, &wait, TASK_UNINTERRUPTIBLE); - if (trans->transaction->num_writers <= 1) - break; + + if (cur_trans->num_writers > 1) + timeout = MAX_SCHEDULE_TIMEOUT; + else + timeout = 1; + mutex_unlock(&root->fs_info->fs_mutex); mutex_unlock(&root->fs_info->trans_mutex); - schedule(); + + schedule_timeout(timeout); + mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); - finish_wait(&trans->transaction->writer_wait, &wait); - } - finish_wait(&trans->transaction->writer_wait, &wait); + finish_wait(&cur_trans->writer_wait, &wait); + } while (cur_trans->num_writers > 1 || + (cur_trans->num_joined != joined)); + WARN_ON(cur_trans != trans->transaction); ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, &dirty_fs_roots); @@ -511,6 +530,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_finish_extent_commit(trans, root, &pinned_copy); mutex_lock(&root->fs_info->trans_mutex); cur_trans->commit_done = 1; + root->fs_info->last_trans_committed = cur_trans->transid; wake_up(&cur_trans->commit_wait); put_transaction(cur_trans); put_transaction(cur_trans); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index d5f491d3757..e451783a1a4 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -23,6 +23,7 @@ struct btrfs_transaction { u64 transid; unsigned long num_writers; + unsigned long num_joined; int in_commit; int use_count; int commit_done; @@ -57,6 +58,12 @@ static inline void btrfs_update_inode_block_group(struct BTRFS_I(inode)->block_group = trans->block_group; } +static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans, + struct inode *inode) +{ + BTRFS_I(inode)->last_trans = trans->transaction->transid; +} + int btrfs_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, -- cgit v1.2.3-70-g09d2 From 2cc58cf24f69be8632a3b29d653c318bf3bd8c84 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 27 Aug 2007 16:49:44 -0400 Subject: Btrfs: Do more extensive readahead during tree searches Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 62 +++++++++++++++++++++++++++++++++++++++----------- fs/btrfs/extent-tree.c | 4 ++-- fs/btrfs/inode.c | 2 +- 3 files changed, 52 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 51eea5ccb64..c0782a5b04c 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -43,8 +43,10 @@ struct btrfs_path *btrfs_alloc_path(void) { struct btrfs_path *path; path = kmem_cache_alloc(btrfs_path_cachep, GFP_NOFS); - if (path) + if (path) { btrfs_init_path(path); + path->reada = 1; + } return path; } @@ -159,6 +161,34 @@ static int close_blocks(u64 blocknr, u64 other) return 0; } +static int should_defrag_leaf(struct buffer_head *bh) +{ + struct btrfs_leaf *leaf = btrfs_buffer_leaf(bh); + struct btrfs_disk_key *key; + u32 nritems; + + if (buffer_defrag(bh)) + return 1; + + nritems = btrfs_header_nritems(&leaf->header); + if (nritems == 0) + return 0; + + key = &leaf->items[0].key; + if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY) + return 1; + + key = &leaf->items[nritems-1].key; + if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY) + return 1; + if (nritems > 4) { + key = &leaf->items[nritems/2].key; + if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY) + return 1; + } + return 0; +} + int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct buffer_head *parent, int cache_only, u64 *last_ret) @@ -217,7 +247,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, cur_bh = btrfs_find_tree_block(root, blocknr); if (!cur_bh || !buffer_uptodate(cur_bh) || - buffer_locked(cur_bh) || !buffer_defrag(cur_bh)) { + buffer_locked(cur_bh) || + (parent_level != 1 && !buffer_defrag(cur_bh)) || + (parent_level == 1 && !should_defrag_leaf(cur_bh))) { if (cache_only) { brelse(cur_bh); continue; @@ -297,6 +329,7 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path, parent = btrfs_buffer_node(path->nodes[level + 1]); slot = path->slots[level]; + BUG_ON(!buffer_uptodate(path->nodes[level])); BUG_ON(nritems == 0); if (parent) { struct btrfs_disk_key *parent_key; @@ -511,9 +544,6 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root err_on_enospc = 1; left_buf = read_node_slot(root, parent_buf, pslot - 1); - right_buf = read_node_slot(root, parent_buf, pslot + 1); - - /* first, try to make some room in the middle buffer */ if (left_buf) { wret = btrfs_cow_block(trans, root, left_buf, parent_buf, pslot - 1, &left_buf); @@ -521,6 +551,19 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root ret = wret; goto enospc; } + } + right_buf = read_node_slot(root, parent_buf, pslot + 1); + if (right_buf) { + wret = btrfs_cow_block(trans, root, right_buf, + parent_buf, pslot + 1, &right_buf); + if (wret) { + ret = wret; + goto enospc; + } + } + + /* first, try to make some room in the middle buffer */ + if (left_buf) { left = btrfs_buffer_node(left_buf); orig_slot += btrfs_header_nritems(&left->header); wret = push_node_left(trans, root, left_buf, mid_buf); @@ -534,13 +577,6 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root * then try to empty the right most buffer into the middle */ if (right_buf) { - wret = btrfs_cow_block(trans, root, right_buf, - parent_buf, pslot + 1, &right_buf); - if (wret) { - ret = wret; - goto enospc; - } - right = btrfs_buffer_node(right_buf); wret = push_node_left(trans, root, mid_buf, right_buf); if (wret < 0 && wret != -ENOSPC) @@ -817,7 +853,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, for (i = 0; i < ret; i++) { blocknr = gang[i]; clear_radix_bit(&found, blocknr); - if (nread > 32) + if (path->reada == 1 && nread > 16) continue; if (close_blocks(cluster_start, blocknr)) { readahead_tree_block(root, blocknr); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index b3641234473..4049aadbeda 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -52,7 +52,7 @@ static int cache_block_group(struct btrfs_root *root, path = btrfs_alloc_path(); if (!path) return -ENOMEM; - path->reada = 1; + path->reada = 2; key.objectid = block_group->key.objectid; key.flags = 0; key.offset = 0; @@ -1015,7 +1015,7 @@ check_failed: ins->objectid = search_start; ins->offset = 0; start_found = 0; - path->reada = 1; + path->reada = 2; ret = btrfs_search_slot(trans, root, ins, path, 0, 0); if (ret < 0) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 398484179d8..7e4cf62ada5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -827,7 +827,7 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; path = btrfs_alloc_path(); - path->reada = 1; + path->reada = 2; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; -- cgit v1.2.3-70-g09d2 From 83df7c1d8b885e216c68c3f0469bec9b93666b74 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 27 Aug 2007 16:49:44 -0400 Subject: Btrfs: Make sure to cow the root during a snapshot Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7e4cf62ada5..504d37dfa6c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2096,6 +2096,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) struct btrfs_trans_handle *trans; struct btrfs_key key; struct btrfs_root_item new_root_item; + struct buffer_head *tmp; int ret; int err; u64 objectid; @@ -2123,6 +2124,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) key.offset = 1; key.flags = 0; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, -- cgit v1.2.3-70-g09d2 From a52d9a8033c454cd9b4697cfafb467509fc1693f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 27 Aug 2007 16:49:44 -0400 Subject: Btrfs: Extent based page cache code. This uses an rbtree of extents and tests instead of buffer heads. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 3 +- fs/btrfs/btrfs_inode.h | 3 + fs/btrfs/ctree.h | 7 + fs/btrfs/disk-io.h | 1 - fs/btrfs/extent_map.c | 1656 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/extent_map.h | 89 +++ fs/btrfs/file.c | 295 ++++----- fs/btrfs/inode.c | 816 +++++++++--------------- fs/btrfs/super.c | 2 + 9 files changed, 2203 insertions(+), 669 deletions(-) create mode 100644 fs/btrfs/extent_map.c create mode 100644 fs/btrfs/extent_map.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 9321438c5ee..ea7a2287035 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -4,7 +4,8 @@ ifneq ($(KERNELRELEASE),) obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ - transaction.o bit-radix.o inode.o file.o tree-defrag.o + transaction.o bit-radix.o inode.o file.o tree-defrag.o \ + extent_map.o #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \ # root-tree.o dir-item.o hash.o file-item.o inode-item.o \ diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index f446f161f73..d1d5af471c3 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -19,11 +19,14 @@ #ifndef __BTRFS_I__ #define __BTRFS_I__ +#include "extent_map.h" + /* in memory btrfs inode */ struct btrfs_inode { struct btrfs_root *root; struct btrfs_block_group_cache *block_group; struct btrfs_key location; + struct extent_map_tree extent_tree; struct inode vfs_inode; /* diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index c7f5161271d..947f061ed11 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1176,7 +1176,14 @@ int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to); int btrfs_get_block(struct inode *inode, sector_t iblock, struct buffer_head *result, int create); +struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, + size_t page_offset, u64 start, u64 end, + int create); +int btrfs_update_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode); /* file.c */ +int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 81fd18cbd82..5261733b873 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -77,7 +77,6 @@ int btrfs_insert_dev_radix(struct btrfs_root *root, u64 num_blocks); int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, u64 logical); -int btrfs_releasepage(struct page *page, gfp_t flags); void btrfs_btree_balance_dirty(struct btrfs_root *root); int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); void btrfs_mark_buffer_dirty(struct buffer_head *bh); diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c new file mode 100644 index 00000000000..d378edf0964 --- /dev/null +++ b/fs/btrfs/extent_map.c @@ -0,0 +1,1656 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "extent_map.h" + +static struct kmem_cache *extent_map_cache; +static struct kmem_cache *extent_state_cache; + +struct tree_entry { + u64 start; + u64 end; + int in_tree; + struct rb_node rb_node; +}; + +/* bits for the extent state */ +#define EXTENT_DIRTY 1 +#define EXTENT_WRITEBACK (1 << 1) +#define EXTENT_UPTODATE (1 << 2) +#define EXTENT_LOCKED (1 << 3) +#define EXTENT_NEW (1 << 4) +#define EXTENT_DELALLOC (1 << 5) + +#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) + +static LIST_HEAD(all_states); +spinlock_t state_lock = SPIN_LOCK_UNLOCKED; + +void __init extent_map_init(void) +{ + extent_map_cache = kmem_cache_create("extent_map", + sizeof(struct extent_map), 0, + SLAB_RECLAIM_ACCOUNT | + SLAB_DESTROY_BY_RCU, + NULL); + extent_state_cache = kmem_cache_create("extent_state", + sizeof(struct extent_state), 0, + SLAB_RECLAIM_ACCOUNT | + SLAB_DESTROY_BY_RCU, + NULL); +} + +void __exit extent_map_exit(void) +{ + while(!list_empty(&all_states)) { + struct extent_state *state; + struct list_head *cur = all_states.next; + state = list_entry(cur, struct extent_state, list); + printk("found leaked state %Lu %Lu state %d in_tree %d\n", + state->start, state->end, state->state, state->in_tree); + list_del(&state->list); + kfree(state); + } + if (extent_map_cache) + kmem_cache_destroy(extent_map_cache); + if (extent_state_cache) + kmem_cache_destroy(extent_state_cache); +} + +void extent_map_tree_init(struct extent_map_tree *tree, + struct address_space *mapping, gfp_t mask) +{ + tree->map.rb_node = NULL; + tree->state.rb_node = NULL; + rwlock_init(&tree->lock); + tree->mapping = mapping; +} +EXPORT_SYMBOL(extent_map_tree_init); + +struct extent_map *alloc_extent_map(gfp_t mask) +{ + struct extent_map *em; + em = kmem_cache_alloc(extent_map_cache, mask); + if (!em || IS_ERR(em)) + return em; + em->in_tree = 0; + atomic_set(&em->refs, 1); + return em; +} +EXPORT_SYMBOL(alloc_extent_map); + +void free_extent_map(struct extent_map *em) +{ + if (atomic_dec_and_test(&em->refs)) { + WARN_ON(em->in_tree); + kmem_cache_free(extent_map_cache, em); + } +} +EXPORT_SYMBOL(free_extent_map); + + +struct extent_state *alloc_extent_state(gfp_t mask) +{ + struct extent_state *state; + state = kmem_cache_alloc(extent_state_cache, mask); + if (!state || IS_ERR(state)) + return state; + state->state = 0; + state->in_tree = 0; + atomic_set(&state->refs, 1); + init_waitqueue_head(&state->wq); + spin_lock_irq(&state_lock); + list_add(&state->list, &all_states); + spin_unlock_irq(&state_lock); + return state; +} +EXPORT_SYMBOL(alloc_extent_state); + +void free_extent_state(struct extent_state *state) +{ + if (atomic_dec_and_test(&state->refs)) { + WARN_ON(state->in_tree); + spin_lock_irq(&state_lock); + list_del_init(&state->list); + spin_unlock_irq(&state_lock); + kmem_cache_free(extent_state_cache, state); + } +} +EXPORT_SYMBOL(free_extent_state); + +static struct rb_node *tree_insert(struct rb_root *root, u64 offset, + struct rb_node *node) +{ + struct rb_node ** p = &root->rb_node; + struct rb_node * parent = NULL; + struct tree_entry *entry; + + while(*p) { + parent = *p; + entry = rb_entry(parent, struct tree_entry, rb_node); + + if (offset < entry->start) + p = &(*p)->rb_left; + else if (offset > entry->end) + p = &(*p)->rb_right; + else + return parent; + } + + entry = rb_entry(node, struct tree_entry, rb_node); + entry->in_tree = 1; + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct rb_node *__tree_search(struct rb_root *root, u64 offset, + struct rb_node **prev_ret) +{ + struct rb_node * n = root->rb_node; + struct rb_node *prev = NULL; + struct tree_entry *entry; + struct tree_entry *prev_entry = NULL; + + while(n) { + entry = rb_entry(n, struct tree_entry, rb_node); + prev = n; + prev_entry = entry; + + if (offset < entry->start) + n = n->rb_left; + else if (offset > entry->end) + n = n->rb_right; + else + return n; + } + if (!prev_ret) + return NULL; + while(prev && offset > prev_entry->end) { + prev = rb_next(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *prev_ret = prev; + return NULL; +} + +static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) +{ + struct rb_node *prev; + struct rb_node *ret; + ret = __tree_search(root, offset, &prev); + if (!ret) + return prev; + return ret; +} + +static int tree_delete(struct rb_root *root, u64 offset) +{ + struct rb_node *node; + struct tree_entry *entry; + + node = __tree_search(root, offset, NULL); + if (!node) + return -ENOENT; + entry = rb_entry(node, struct tree_entry, rb_node); + entry->in_tree = 0; + rb_erase(node, root); + return 0; +} + +/* + * add_extent_mapping tries a simple backward merge with existing + * mappings. The extent_map struct passed in will be inserted into + * the tree directly (no copies made, just a reference taken). + */ +int add_extent_mapping(struct extent_map_tree *tree, + struct extent_map *em) +{ + int ret = 0; + struct extent_map *prev = NULL; + struct rb_node *rb; + + write_lock_irq(&tree->lock); + rb = tree_insert(&tree->map, em->end, &em->rb_node); + if (rb) { + prev = rb_entry(rb, struct extent_map, rb_node); + printk("found extent map %Lu %Lu on insert of %Lu %Lu\n", prev->start, prev->end, em->start, em->end); + ret = -EEXIST; + goto out; + } + atomic_inc(&em->refs); + if (em->start != 0) { + rb = rb_prev(&em->rb_node); + if (rb) + prev = rb_entry(rb, struct extent_map, rb_node); + if (prev && prev->end + 1 == em->start && + ((em->block_start == 0 && prev->block_start == 0) || + (em->block_start == prev->block_end + 1))) { + em->start = prev->start; + em->block_start = prev->block_start; + rb_erase(&prev->rb_node, &tree->map); + prev->in_tree = 0; + free_extent_map(prev); + } + } +out: + write_unlock_irq(&tree->lock); + return ret; +} +EXPORT_SYMBOL(add_extent_mapping); + +/* + * lookup_extent_mapping returns the first extent_map struct in the + * tree that intersects the [start, end] (inclusive) range. There may + * be additional objects in the tree that intersect, so check the object + * returned carefully to make sure you don't need additional lookups. + */ +struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, + u64 start, u64 end) +{ + struct extent_map *em; + struct rb_node *rb_node; + + read_lock_irq(&tree->lock); + rb_node = tree_search(&tree->map, start); + if (!rb_node) { + em = NULL; + goto out; + } + if (IS_ERR(rb_node)) { + em = ERR_PTR(PTR_ERR(rb_node)); + goto out; + } + em = rb_entry(rb_node, struct extent_map, rb_node); + if (em->end < start || em->start > end) { + em = NULL; + goto out; + } + atomic_inc(&em->refs); +out: + read_unlock_irq(&tree->lock); + return em; +} +EXPORT_SYMBOL(lookup_extent_mapping); + +/* + * removes an extent_map struct from the tree. No reference counts are + * dropped, and no checks are done to see if the range is in use + */ +int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) +{ + int ret; + + write_lock_irq(&tree->lock); + ret = tree_delete(&tree->map, em->end); + write_unlock_irq(&tree->lock); + return ret; +} +EXPORT_SYMBOL(remove_extent_mapping); + +/* + * utility function to look for merge candidates inside a given range. + * Any extents with matching state are merged together into a single + * extent in the tree. Extents with EXTENT_IO in their state field + * are not merged because the end_io handlers need to be able to do + * operations on them without sleeping (or doing allocations/splits). + * + * This should be called with the tree lock held. + */ +static int merge_state(struct extent_map_tree *tree, + struct extent_state *state) +{ + struct extent_state *other; + struct rb_node *other_node; + + if (state->state & EXTENT_IOBITS) + return 0; + + other_node = rb_prev(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->end == state->start - 1 && + other->state == state->state) { + state->start = other->start; + other->in_tree = 0; + rb_erase(&other->rb_node, &tree->state); + free_extent_state(other); + } + } + other_node = rb_next(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->start == state->end + 1 && + other->state == state->state) { + other->start = state->start; + state->in_tree = 0; + rb_erase(&state->rb_node, &tree->state); + free_extent_state(state); + } + } + return 0; +} + +/* + * insert an extent_state struct into the tree. 'bits' are set on the + * struct before it is inserted. + * + * This may return -EEXIST if the extent is already there, in which case the + * state struct is freed. + * + * The tree lock is not taken internally. This is a utility function and + * probably isn't what you want to call (see set/clear_extent_bit). + */ +static int insert_state(struct extent_map_tree *tree, + struct extent_state *state, u64 start, u64 end, + int bits) +{ + struct rb_node *node; + + if (end < start) { + printk("end < start %Lu %Lu\n", end, start); + WARN_ON(1); + } + state->state |= bits; + state->start = start; + state->end = end; + if ((end & 4095) == 0) { + printk("insert state %Lu %Lu strange end\n", start, end); + WARN_ON(1); + } + node = tree_insert(&tree->state, end, &state->rb_node); + if (node) { + struct extent_state *found; + found = rb_entry(node, struct extent_state, rb_node); +printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); + free_extent_state(state); + return -EEXIST; + } + merge_state(tree, state); + return 0; +} + +/* + * split a given extent state struct in two, inserting the preallocated + * struct 'prealloc' as the newly created second half. 'split' indicates an + * offset inside 'orig' where it should be split. + * + * Before calling, + * the tree has 'orig' at [orig->start, orig->end]. After calling, there + * are two extent state structs in the tree: + * prealloc: [orig->start, split - 1] + * orig: [ split, orig->end ] + * + * The tree locks are not taken by this function. They need to be held + * by the caller. + */ +static int split_state(struct extent_map_tree *tree, struct extent_state *orig, + struct extent_state *prealloc, u64 split) +{ + struct rb_node *node; + prealloc->start = orig->start; + prealloc->end = split - 1; + prealloc->state = orig->state; + orig->start = split; + if ((prealloc->end & 4095) == 0) { + printk("insert state %Lu %Lu strange end\n", prealloc->start, + prealloc->end); + WARN_ON(1); + } + node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); + if (node) { + struct extent_state *found; + found = rb_entry(node, struct extent_state, rb_node); +printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); + free_extent_state(prealloc); + return -EEXIST; + } + return 0; +} + +/* + * utility function to clear some bits in an extent state struct. + * it will optionally wake up any one waiting on this state (wake == 1), or + * forcibly remove the state from the tree (delete == 1). + * + * If no bits are set on the state struct after clearing things, the + * struct is freed and removed from the tree + */ +static int clear_state_bit(struct extent_map_tree *tree, + struct extent_state *state, int bits, int wake, + int delete) +{ + int ret = state->state & bits; + state->state &= ~bits; + if (wake) + wake_up(&state->wq); + if (delete || state->state == 0) { + if (state->in_tree) { + rb_erase(&state->rb_node, &tree->state); + state->in_tree = 0; + free_extent_state(state); + } else { + WARN_ON(1); + } + } else { + merge_state(tree, state); + } + return ret; +} + +/* + * clear some bits on a range in the tree. This may require splitting + * or inserting elements in the tree, so the gfp mask is used to + * indicate which allocations or sleeping are allowed. + * + * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove + * the given range from the tree regardless of state (ie for truncate). + * + * the range [start, end] is inclusive. + * + * This takes the tree lock, and returns < 0 on error, > 0 if any of the + * bits were already set, or zero if none of the bits were already set. + */ +int clear_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, + int bits, int wake, int delete, gfp_t mask) +{ + struct extent_state *state; + struct extent_state *prealloc = NULL; + struct rb_node *node; + int err; + int set = 0; + +again: + if (!prealloc && (mask & __GFP_WAIT)) { + prealloc = alloc_extent_state(mask); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irq(&tree->lock); + /* + * this search will find the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + goto out; + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > end) + goto out; + WARN_ON(state->end < start); + + /* + * | ---- desired range ---- | + * | state | or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip + * bits on second half. + * + * If the extent we found extends past our range, we + * just split and search again. It'll get split again + * the next time though. + * + * If the extent we found is inside our range, we clear + * the desired bit on it. + */ + + if (state->start < start) { + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + if (err) + goto out; + if (state->end <= end) { + start = state->end + 1; + set |= clear_state_bit(tree, state, bits, + wake, delete); + } else { + start = state->start; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and clear the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + err = split_state(tree, state, prealloc, end + 1); + BUG_ON(err == -EEXIST); + + if (wake) + wake_up(&state->wq); + set |= clear_state_bit(tree, prealloc, bits, + wake, delete); + prealloc = NULL; + goto out; + } + + start = state->end + 1; + set |= clear_state_bit(tree, state, bits, wake, delete); + goto search_again; + +out: + write_unlock_irq(&tree->lock); + if (prealloc) + free_extent_state(prealloc); + + return set; + +search_again: + if (start >= end) + goto out; + write_unlock_irq(&tree->lock); + if (mask & __GFP_WAIT) + cond_resched(); + goto again; +} +EXPORT_SYMBOL(clear_extent_bit); + +static int wait_on_state(struct extent_map_tree *tree, + struct extent_state *state) +{ + DEFINE_WAIT(wait); + prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); + read_unlock_irq(&tree->lock); + schedule(); + read_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + return 0; +} + +/* + * waits for one or more bits to clear on a range in the state tree. + * The range [start, end] is inclusive. + * The tree lock is taken by this function + */ +int wait_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits) +{ + struct extent_state *state; + struct rb_node *node; + + read_lock_irq(&tree->lock); +again: + while (1) { + /* + * this search will find all the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + break; + + state = rb_entry(node, struct extent_state, rb_node); + + if (state->start > end) + goto out; + + if (state->state & bits) { + start = state->start; + atomic_inc(&state->refs); + wait_on_state(tree, state); + free_extent_state(state); + goto again; + } + start = state->end + 1; + + if (start > end) + break; + + if (need_resched()) { + read_unlock_irq(&tree->lock); + cond_resched(); + read_lock_irq(&tree->lock); + } + } +out: + read_unlock_irq(&tree->lock); + return 0; +} +EXPORT_SYMBOL(wait_extent_bit); + +/* + * set some bits on a range in the tree. This may require allocations + * or sleeping, so the gfp mask is used to indicate what is allowed. + * + * If 'exclusive' == 1, this will fail with -EEXIST if some part of the + * range already has the desired bits set. The start of the existing + * range is returned in failed_start in this case. + * + * [start, end] is inclusive + * This takes the tree lock. + */ +int set_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits, + int exclusive, u64 *failed_start, gfp_t mask) +{ + struct extent_state *state; + struct extent_state *prealloc = NULL; + struct rb_node *node; + int err = 0; + int set; + u64 last_start; + u64 last_end; +again: + if (!prealloc && (mask & __GFP_WAIT)) { + prealloc = alloc_extent_state(mask); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node) { + err = insert_state(tree, prealloc, start, end, bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + goto out; + } + + state = rb_entry(node, struct extent_state, rb_node); + last_start = state->start; + last_end = state->end; + + /* + * | ---- desired range ---- | + * | state | + * + * Just lock what we found and keep going + */ + if (state->start == start && state->end <= end) { + set = state->state & bits; + if (set && exclusive) { + *failed_start = state->start; + err = -EEXIST; + goto out; + } + state->state |= bits; + start = state->end + 1; + merge_state(tree, state); + goto search_again; + } + + /* + * | ---- desired range ---- | + * | state | + * or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip bits on + * second half. + * + * If the extent we found extends past our + * range, we just split and search again. It'll get split + * again the next time though. + * + * If the extent we found is inside our range, we set the + * desired bit on it. + */ + if (state->start < start) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + if (err) + goto out; + if (state->end <= end) { + state->state |= bits; + start = state->end + 1; + merge_state(tree, state); + } else { + start = state->start; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and set the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, end + 1); + BUG_ON(err == -EEXIST); + + prealloc->state |= bits; + merge_state(tree, prealloc); + prealloc = NULL; + goto out; + } + + /* + * | ---- desired range ---- | + * | state | or | state | + * + * There's a hole, we need to insert something in it and + * ignore the extent we found. + */ + if (state->start > start) { + u64 this_end; + if (end < last_start) + this_end = end; + else + this_end = last_start -1; + err = insert_state(tree, prealloc, start, this_end, + bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + if (err) + goto out; + start = this_end + 1; + goto search_again; + } + goto search_again; + +out: + write_unlock_irq(&tree->lock); + if (prealloc) + free_extent_state(prealloc); + + return err; + +search_again: + if (start > end) + goto out; + write_unlock_irq(&tree->lock); + if (mask & __GFP_WAIT) + cond_resched(); + goto again; +} +EXPORT_SYMBOL(set_extent_bit); + +/* wrappers around set/clear extent bit */ +int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_dirty); + +int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_DIRTY, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_dirty); + +int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_new); + +int clear_extent_new(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_new); + +int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_uptodate); + +int clear_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_uptodate); + +int set_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_WRITEBACK, + 0, NULL, mask); +} +EXPORT_SYMBOL(set_extent_writeback); + +int clear_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); +} +EXPORT_SYMBOL(clear_extent_writeback); + +int wait_on_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end) +{ + return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK); +} +EXPORT_SYMBOL(wait_on_extent_writeback); + +/* + * locks a range in ascending order, waiting for any locked regions + * it hits on the way. [start,end] are inclusive, and this will sleep. + */ +int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask) +{ + int err; + u64 failed_start; + while (1) { + err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, + &failed_start, mask); + if (err == -EEXIST && (mask & __GFP_WAIT)) { + wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); + start = failed_start; + } else { + break; + } + WARN_ON(start > end); + } + return err; +} +EXPORT_SYMBOL(lock_extent); + +int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask); +} +EXPORT_SYMBOL(unlock_extent); + +/* + * helper function to set pages and extents in the tree dirty + */ +int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + BUG_ON(!page); + __set_page_dirty_nobuffers(page); + page_cache_release(page); + index++; + } + set_extent_dirty(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(set_range_dirty); + +/* + * helper function to set both pages and extents in the tree writeback + */ +int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + BUG_ON(!page); + set_page_writeback(page); + page_cache_release(page); + index++; + } + set_extent_writeback(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(set_range_writeback); + +/* + * helper function to lock both pages and extents in the tree. + * pages must be locked first. + */ +int lock_range(struct extent_map_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + int err; + + while (index <= end_index) { + page = grab_cache_page(tree->mapping, index); + if (!page) { + err = -ENOMEM; + goto failed; + } + if (IS_ERR(page)) { + err = PTR_ERR(page); + goto failed; + } + index++; + } + lock_extent(tree, start, end, GFP_NOFS); + return 0; + +failed: + /* + * we failed above in getting the page at 'index', so we undo here + * up to but not including the page at 'index' + */ + end_index = index; + index = start >> PAGE_CACHE_SHIFT; + while (index < end_index) { + page = find_get_page(tree->mapping, index); + unlock_page(page); + page_cache_release(page); + index++; + } + return err; +} +EXPORT_SYMBOL(lock_range); + +/* + * helper function to unlock both pages and extents in the tree. + */ +int unlock_range(struct extent_map_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + unlock_page(page); + page_cache_release(page); + index++; + } + unlock_extent(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(unlock_range); + +/* + * searches a range in the state tree for a given mask. + * If 'filled' == 1, this returns 1 only if ever extent in the tree + * has the bits set. Otherwise, 1 is returned if any bit in the + * range is found set. + */ +static int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end, + int bits, int filled) +{ + struct extent_state *state = NULL; + struct rb_node *node; + int bitset = 0; + + read_lock_irq(&tree->lock); + node = tree_search(&tree->state, start); + while (node && start <= end) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > end) + break; + + if (filled && state->start > start) { + bitset = 0; + break; + } + if (state->state & bits) { + bitset = 1; + if (!filled) + break; + } else if (filled) { + bitset = 0; + break; + } + start = state->end + 1; + if (start > end) + break; + node = rb_next(node); + } + read_unlock_irq(&tree->lock); + return bitset; +} + +/* + * helper function to set a given page up to date if all the + * extents in the tree for that page are up to date + */ +static int check_page_uptodate(struct extent_map_tree *tree, + struct page *page) +{ + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1)) + SetPageUptodate(page); + return 0; +} + +/* + * helper function to unlock a page if all the extents in the tree + * for that page are unlocked + */ +static int check_page_locked(struct extent_map_tree *tree, + struct page *page) +{ + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0)) + unlock_page(page); + return 0; +} + +/* + * helper function to end page writeback if all the extents + * in the tree for that page are done with writeback + */ +static int check_page_writeback(struct extent_map_tree *tree, + struct page *page) +{ + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0)) + end_page_writeback(page); + return 0; +} + +/* lots and lots of room for performance fixes in the end_bio funcs */ + +/* + * after a writepage IO is done, we need to: + * clear the uptodate bits on error + * clear the writeback bits in the extent tree for this IO + * end_page_writeback if the page has no more pending IO + * + * Scheduling is not allowed, so the extent state tree is expected + * to have one and only one object corresponding to this IO. + */ +static int end_bio_extent_writepage(struct bio *bio, + unsigned int bytes_done, int err) +{ + const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_map_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + + if (bio->bi_size) + return 1; + + do { + struct page *page = bvec->bv_page; + start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (!uptodate) { + clear_extent_uptodate(tree, start, end, GFP_ATOMIC); + ClearPageUptodate(page); + SetPageError(page); + } + clear_extent_writeback(tree, start, end, GFP_ATOMIC); + + if (whole_page) + end_page_writeback(page); + else + check_page_writeback(tree, page); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); + return 0; +} + +/* + * after a readpage IO is done, we need to: + * clear the uptodate bits on error + * set the uptodate bits if things worked + * set the page up to date if all extents in the tree are uptodate + * clear the lock bit in the extent tree + * unlock the page if there are no other extents locked for it + * + * Scheduling is not allowed, so the extent state tree is expected + * to have one and only one object corresponding to this IO. + */ +static int end_bio_extent_readpage(struct bio *bio, + unsigned int bytes_done, int err) +{ + const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_map_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + + if (bio->bi_size) + return 1; + + do { + struct page *page = bvec->bv_page; + start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + if (whole_page) + SetPageUptodate(page); + else + check_page_uptodate(tree, page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + if (whole_page) + unlock_page(page); + else + check_page_locked(tree, page); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); + return 0; +} + +/* + * IO done from prepare_write is pretty simple, we just unlock + * the structs in the extent tree when done, and set the uptodate bits + * as appropriate. + */ +static int end_bio_extent_preparewrite(struct bio *bio, + unsigned int bytes_done, int err) +{ + const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_map_tree *tree = bio->bi_private; + u64 start; + u64 end; + + if (bio->bi_size) + return 1; + + do { + struct page *page = bvec->bv_page; + start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); + return 0; +} + +static int submit_extent_page(int rw, struct extent_map_tree *tree, + struct page *page, sector_t sector, + size_t size, unsigned long offset, + struct block_device *bdev, + bio_end_io_t end_io_func) +{ + struct bio *bio; + int ret = 0; + + bio = bio_alloc(GFP_NOIO, 1); + + bio->bi_sector = sector; + bio->bi_bdev = bdev; + bio->bi_io_vec[0].bv_page = page; + bio->bi_io_vec[0].bv_len = size; + bio->bi_io_vec[0].bv_offset = offset; + + bio->bi_vcnt = 1; + bio->bi_idx = 0; + bio->bi_size = size; + + bio->bi_end_io = end_io_func; + bio->bi_private = tree; + + bio_get(bio); + submit_bio(rw, bio); + + if (bio_flagged(bio, BIO_EOPNOTSUPP)) + ret = -EOPNOTSUPP; + + bio_put(bio); + return ret; +} + +/* + * basic readpage implementation. Locked extent state structs are inserted + * into the tree that are removed when the IO is done (by the end_io + * handlers) + */ +int extent_read_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent) +{ + struct inode *inode = page->mapping->host; + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + u64 cur_end; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_offset = 0; + size_t iosize; + size_t blocksize = inode->i_sb->s_blocksize; + + if (!PagePrivate(page)) { + SetPagePrivate(page); + set_page_private(page, 1); + page_cache_get(page); + } + + end = page_end; + lock_extent(tree, start, end, GFP_NOFS); + + while (cur <= end) { + if (cur >= last_byte) { + iosize = PAGE_CACHE_SIZE - page_offset; + zero_user_page(page, page_offset, iosize, KM_USER0); + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + break; + } + em = get_extent(inode, page, page_offset, cur, end, 0); + if (IS_ERR(em) || !em) { + SetPageError(page); + unlock_extent(tree, cur, end, GFP_NOFS); + break; + } + + extent_offset = cur - em->start; + BUG_ON(em->end < cur); + BUG_ON(end < cur); + + iosize = min(em->end - cur, end - cur) + 1; + cur_end = min(em->end, end); + iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + sector = (em->block_start + extent_offset) >> 9; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + /* we've found a hole, just zero and go on */ + if (block_start == 0) { + zero_user_page(page, page_offset, iosize, KM_USER0); + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + /* the get_extent function already copied into the page */ + if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, bdev, + end_bio_extent_readpage); + if (ret) + SetPageError(page); + cur = cur + iosize; + page_offset += iosize; + nr++; + } + if (!nr) { + if (!PageError(page)) + SetPageUptodate(page); + unlock_page(page); + } + return 0; +} +EXPORT_SYMBOL(extent_read_full_page); + +/* + * the writepage semantics are similar to regular writepage. extent + * records are inserted to lock ranges in the tree, and as dirty areas + * are found, they are marked writeback. Then the lock bits are removed + * and the end_io handler clears the writeback ranges + */ +int extent_write_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + struct inode *inode = page->mapping->host; + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_offset = 0; + size_t iosize; + size_t blocksize; + loff_t i_size = i_size_read(inode); + unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; + + if (page->index > end_index) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; + } + + if (page->index == end_index) { + size_t offset = i_size & (PAGE_CACHE_SIZE - 1); + zero_user_page(page, offset, + PAGE_CACHE_SIZE - offset, KM_USER0); + } + + if (!PagePrivate(page)) { + SetPagePrivate(page); + set_page_private(page, 1); + page_cache_get(page); + } + + end = page_end; + lock_extent(tree, start, page_end, GFP_NOFS); + + if (last_byte <= start) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + goto done; + } + + set_extent_uptodate(tree, start, page_end, GFP_NOFS); + blocksize = inode->i_sb->s_blocksize; + + while (cur <= end) { + if (cur >= last_byte) { + clear_extent_dirty(tree, cur, page_end, GFP_NOFS); + break; + } + em = get_extent(inode, page, page_offset, cur, end, 1); + if (IS_ERR(em) || !em) { + SetPageError(page); + break; + } + + extent_offset = cur - em->start; + BUG_ON(em->end < cur); + BUG_ON(end < cur); + iosize = min(em->end - cur, end - cur) + 1; + iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + sector = (em->block_start + extent_offset) >> 9; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + if (block_start == 0 || block_start == EXTENT_MAP_INLINE) { + clear_extent_dirty(tree, cur, + cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + /* leave this out until we have a page_mkwrite call */ + if (0 && !test_range_bit(tree, cur, cur + iosize - 1, + EXTENT_DIRTY, 0)) { + cur = cur + iosize; + page_offset += iosize; + continue; + } + clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); + set_range_writeback(tree, cur, cur + iosize - 1); + ret = submit_extent_page(WRITE, tree, page, + sector, iosize, page_offset, bdev, + end_bio_extent_writepage); + if (ret) + SetPageError(page); + cur = cur + iosize; + page_offset += iosize; + nr++; + } +done: + WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0)); + unlock_extent(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; +} +EXPORT_SYMBOL(extent_write_full_page); + +/* + * basic invalidatepage code, this waits on any locked or writeback + * ranges corresponding to the page, and then deletes any extent state + * records from the tree + */ +int extent_invalidatepage(struct extent_map_tree *tree, + struct page *page, unsigned long offset) +{ + u64 start = (page->index << PAGE_CACHE_SHIFT); + u64 end = start + PAGE_CACHE_SIZE - 1; + size_t blocksize = page->mapping->host->i_sb->s_blocksize; + + start += (offset + blocksize -1) & ~(blocksize - 1); + if (start > end) + return 0; + + lock_extent(tree, start, end, GFP_NOFS); + wait_on_extent_writeback(tree, start, end); + clear_extent_bit(tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY, + 1, 1, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(extent_invalidatepage); + +/* + * simple commit_write call, set_range_dirty is used to mark both + * the pages and the extent records as dirty + */ +int extent_commit_write(struct extent_map_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to) +{ + loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; + + if (!PagePrivate(page)) { + SetPagePrivate(page); + set_page_private(page, 1); + page_cache_get(page); + } + + set_page_dirty(page); + + if (pos > inode->i_size) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + return 0; +} +EXPORT_SYMBOL(extent_commit_write); + +int extent_prepare_write(struct extent_map_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to, get_extent_t *get_extent) +{ + u64 page_start = page->index << PAGE_CACHE_SHIFT; + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + u64 block_start; + u64 orig_block_start; + u64 block_end; + u64 cur_end; + struct extent_map *em; + unsigned blocksize = 1 << inode->i_blkbits; + size_t page_offset = 0; + size_t block_off_start; + size_t block_off_end; + int err = 0; + int iocount = 0; + int ret = 0; + int isnew; + + if (!PagePrivate(page)) { + SetPagePrivate(page); + set_page_private(page, 1); + page_cache_get(page); + } + block_start = (page_start + from) & ~((u64)blocksize - 1); + block_end = (page_start + to - 1) | (blocksize - 1); + orig_block_start = block_start; + + lock_extent(tree, page_start, page_end, GFP_NOFS); + while(block_start <= block_end) { + em = get_extent(inode, page, page_offset, block_start, + block_end, 1); + if (IS_ERR(em) || !em) { + goto err; + } + cur_end = min(block_end, em->end); + block_off_start = block_start & (PAGE_CACHE_SIZE - 1); + block_off_end = block_off_start + blocksize; + isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS); + + if (!PageUptodate(page) && isnew && + (block_off_end > to || block_off_start < from)) { + void *kaddr; + + kaddr = kmap_atomic(page, KM_USER0); + if (block_off_end > to) + memset(kaddr + to, 0, block_off_end - to); + if (block_off_start < from) + memset(kaddr + block_off_start, 0, + from - block_off_start); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + } + if (!isnew && !PageUptodate(page) && + (block_off_end > to || block_off_start < from) && + !test_range_bit(tree, block_start, cur_end, + EXTENT_UPTODATE, 1)) { + u64 sector; + u64 extent_offset = block_start - em->start; + size_t iosize; + sector = (em->block_start + extent_offset) >> 9; + iosize = (cur_end - block_start + blocksize - 1) & + ~((u64)blocksize - 1); + /* + * we've already got the extent locked, but we + * need to split the state such that our end_bio + * handler can clear the lock. + */ + set_extent_bit(tree, block_start, + block_start + iosize - 1, + EXTENT_LOCKED, 0, NULL, GFP_NOFS); + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, em->bdev, + end_bio_extent_preparewrite); + iocount++; + block_start = block_start + iosize; + } else { + set_extent_uptodate(tree, block_start, cur_end, + GFP_NOFS); + unlock_extent(tree, block_start, cur_end, GFP_NOFS); + block_start = cur_end + 1; + } + page_offset = block_start & (PAGE_CACHE_SIZE - 1); + free_extent_map(em); + } + if (iocount) { + wait_extent_bit(tree, orig_block_start, + block_end, EXTENT_LOCKED); + } + check_page_uptodate(tree, page); +err: + /* FIXME, zero out newly allocated blocks on error */ + return err; +} +EXPORT_SYMBOL(extent_prepare_write); + +/* + * a helper for releasepage. As long as there are no locked extents + * in the range corresponding to the page, both state records and extent + * map records are removed + */ +int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page) +{ + struct extent_map *em; + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + u64 orig_start = start; + + while (start <= end) { + em = lookup_extent_mapping(tree, start, end); + if (!em || IS_ERR(em)) + break; + if (test_range_bit(tree, em->start, em->end, + EXTENT_LOCKED, 0)) { + free_extent_map(em); + start = em->end + 1; +printk("range still locked %Lu %Lu\n", em->start, em->end); + break; + } + remove_extent_mapping(tree, em); + start = em->end + 1; + /* once for the rb tree */ + free_extent_map(em); + /* once for us */ + free_extent_map(em); + } + WARN_ON(test_range_bit(tree, orig_start, end, EXTENT_WRITEBACK, 0)); + clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, + 1, 1, GFP_NOFS); + return 1; +} +EXPORT_SYMBOL(try_release_extent_mapping); + diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h new file mode 100644 index 00000000000..108944aab4b --- /dev/null +++ b/fs/btrfs/extent_map.h @@ -0,0 +1,89 @@ +#ifndef __EXTENTMAP__ +#define __EXTENTMAP__ + +#include + +#define EXTENT_MAP_INLINE (u64)-2 +#define EXTENT_MAP_DELALLOC (u64)-1 + +struct extent_map_tree { + struct rb_root map; + struct rb_root state; + struct address_space *mapping; + rwlock_t lock; +}; + +/* note, this must start with the same fields as fs/extent_map.c:tree_entry */ +struct extent_map { + u64 start; + u64 end; /* inclusive */ + int in_tree; + struct rb_node rb_node; + /* block_start and block_end are in bytes */ + u64 block_start; + u64 block_end; /* inclusive */ + struct block_device *bdev; + atomic_t refs; +}; + +/* note, this must start with the same fields as fs/extent_map.c:tree_entry */ +struct extent_state { + u64 start; + u64 end; /* inclusive */ + int in_tree; + struct rb_node rb_node; + wait_queue_head_t wq; + atomic_t refs; + unsigned long state; + struct list_head list; +}; + +struct extent_buffer { + u64 start; + u64 end; /* inclusive */ + char *addr; + struct page *pages[]; +}; + +typedef struct extent_map *(get_extent_t)(struct inode *inode, + struct page *page, + size_t page_offset, + u64 start, u64 end, + int create); + +void extent_map_tree_init(struct extent_map_tree *tree, + struct address_space *mapping, gfp_t mask); +struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, + u64 start, u64 end); +int add_extent_mapping(struct extent_map_tree *tree, + struct extent_map *em); +int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em); +int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page); +int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); +int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); +struct extent_map *alloc_extent_map(gfp_t mask); +void free_extent_map(struct extent_map *em); +int extent_read_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent); +void __init extent_map_init(void); +void __exit extent_map_exit(void); +int extent_clean_all_trees(struct extent_map_tree *tree); +int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask); +int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask); +int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask); +int extent_invalidatepage(struct extent_map_tree *tree, + struct page *page, unsigned long offset); +int extent_write_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc); +int extent_prepare_write(struct extent_map_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to, get_extent_t *get_extent); +int extent_commit_write(struct extent_map_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to); +int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); +#endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 6933ab11a5c..71a481894ab 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -81,14 +81,14 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) } } -static int insert_inline_extent(struct btrfs_root *root, struct inode *inode, +static int insert_inline_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, u64 offset, ssize_t size, - struct buffer_head *bh) + struct page *page, size_t page_offset) { struct btrfs_key key; struct btrfs_path *path; char *ptr, *kaddr; - struct btrfs_trans_handle *trans; struct btrfs_file_extent_item *ei; u32 datasize; int err = 0; @@ -98,8 +98,6 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode, if (!path) return -ENOMEM; - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); key.objectid = inode->i_ino; @@ -122,18 +120,13 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); - kaddr = kmap_atomic(bh->b_page, KM_USER0); + kaddr = kmap_atomic(page, KM_USER0); btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, kaddr + bh_offset(bh), - size); + ptr, kaddr + page_offset, size); kunmap_atomic(kaddr, KM_USER0); btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_free_path(path); - ret = btrfs_end_transaction(trans, root); - if (ret && !err) - err = ret; - mutex_unlock(&root->fs_info->fs_mutex); return err; } @@ -145,45 +138,143 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, loff_t pos, size_t write_bytes) { - int i; - int offset; int err = 0; - int ret; - int this_write; + int i; struct inode *inode = file->f_path.dentry->d_inode; - struct buffer_head *bh; + struct extent_map *em; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct btrfs_key ins; + u64 hint_block; + u64 num_blocks; + u64 start_pos; + u64 end_of_last_block; + u64 end_pos = pos + write_bytes; + loff_t isize = i_size_read(inode); - for (i = 0; i < num_pages; i++) { - offset = pos & (PAGE_CACHE_SIZE -1); - this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes); + em = alloc_extent_map(GFP_NOFS); + if (!em) + return -ENOMEM; - /* FIXME, one block at a time */ - bh = page_buffers(pages[i]); + em->bdev = inode->i_sb->s_bdev; - if (buffer_mapped(bh) && bh->b_blocknr == 0) { - ret = insert_inline_extent(root, inode, - pages[i]->index << PAGE_CACHE_SHIFT, - offset + this_write, bh); - if (ret) { - err = ret; - goto failed; - } - } + start_pos = pos & ~((u64)root->blocksize - 1); + num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> + inode->i_blkbits; - ret = btrfs_commit_write(file, pages[i], offset, - offset + this_write); - pos += this_write; - if (ret) { - err = ret; + end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + if (!trans) { + err = -ENOMEM; + goto out_unlock; + } + btrfs_set_trans_block_group(trans, inode); + inode->i_blocks += num_blocks << 3; + hint_block = 0; + + if ((end_of_last_block & 4095) == 0) { + printk("strange end of last %Lu %lu %Lu\n", start_pos, write_bytes, end_of_last_block); + } + set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS); + + /* FIXME...EIEIO, ENOSPC and more */ + + /* step one, delete the existing extents in this range */ + /* FIXME blocksize != pagesize */ + if (start_pos < inode->i_size) { + err = btrfs_drop_extents(trans, root, inode, + start_pos, (pos + write_bytes + root->blocksize -1) & + ~((u64)root->blocksize - 1), &hint_block); + if (err) + goto failed; + } + + /* insert any holes we need to create */ + if (inode->i_size < start_pos) { + u64 last_pos_in_file; + u64 hole_size; + u64 mask = root->blocksize - 1; + last_pos_in_file = (isize + mask) & ~mask; + hole_size = (start_pos - last_pos_in_file + mask) & ~mask; + hole_size >>= inode->i_blkbits; + if (last_pos_in_file < start_pos) { + err = btrfs_insert_file_extent(trans, root, + inode->i_ino, + last_pos_in_file, + 0, 0, hole_size); + } + if (err) goto failed; + } + + /* + * either allocate an extent for the new bytes or setup the key + * to show we are doing inline data in the extent + */ + if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || + pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { + err = btrfs_alloc_extent(trans, root, inode->i_ino, + num_blocks, 0, hint_block, (u64)-1, + &ins, 1); + BUG_ON(err); + err = btrfs_insert_file_extent(trans, root, inode->i_ino, + start_pos, ins.objectid, ins.offset, + ins.offset); + BUG_ON(err); + em->start = start_pos; + em->end = end_of_last_block; + em->block_start = ins.objectid << inode->i_blkbits; + em->block_end = em->block_start + + (ins.offset << inode->i_blkbits) - 1; + set_extent_dirty(em_tree, start_pos, end_of_last_block, + GFP_NOFS); + err = add_extent_mapping(em_tree, em); + for (i = 0; i < num_pages; i++) { + struct page *p = pages[i]; + SetPageUptodate(p); + __set_page_dirty_nobuffers(p); } - WARN_ON(this_write > write_bytes); - write_bytes -= this_write; + } else { + struct page *p = pages[0]; + err = insert_inline_extent(trans, root, inode, start_pos, + end_pos - start_pos, p, 0); + BUG_ON(err); + em->start = start_pos; + em->end = end_pos; + em->block_start = EXTENT_MAP_INLINE; + em->block_end = EXTENT_MAP_INLINE; + add_extent_mapping(em_tree, em); + } + if (end_pos > isize) { + i_size_write(inode, end_pos); + btrfs_update_inode(trans, root, inode); } failed: + err = btrfs_end_transaction(trans, root); +out_unlock: + mutex_unlock(&root->fs_info->fs_mutex); + free_extent_map(em); return err; } +int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) +{ + struct extent_map *em; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + + while(1) { + em = lookup_extent_mapping(em_tree, start, end); + if (!em) + break; + remove_extent_mapping(em_tree, em); + /* once for us */ + free_extent_map(em); + /* once for the tree*/ + free_extent_map(em); + } + return 0; +} + /* * this is very complex, but the basic idea is to drop all extents * in the range start - end. hint_block is filled in with a block number @@ -213,6 +304,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, int found_inline; int recow; + btrfs_drop_extent_cache(inode, start, end - 1); + path = btrfs_alloc_path(); if (!path) return -ENOMEM; @@ -434,18 +527,9 @@ static int prepare_pages(struct btrfs_root *root, int i; unsigned long index = pos >> PAGE_CACHE_SHIFT; struct inode *inode = file->f_path.dentry->d_inode; - int offset; int err = 0; - int this_write; - struct buffer_head *bh; - struct buffer_head *head; - loff_t isize = i_size_read(inode); - struct btrfs_trans_handle *trans; - u64 hint_block; u64 num_blocks; - u64 alloc_extent_start; u64 start_pos; - struct btrfs_key ins; start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1); num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> @@ -457,119 +541,17 @@ static int prepare_pages(struct btrfs_root *root, pages[i] = grab_cache_page(inode->i_mapping, index + i); if (!pages[i]) { err = -ENOMEM; - goto failed_release; + BUG_ON(1); } cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); wait_on_page_writeback(pages[i]); - } - - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - if (!trans) { - err = -ENOMEM; - mutex_unlock(&root->fs_info->fs_mutex); - goto out_unlock; - } - btrfs_set_trans_block_group(trans, inode); - /* FIXME blocksize != 4096 */ - inode->i_blocks += num_blocks << 3; - hint_block = 0; - - /* FIXME...EIEIO, ENOSPC and more */ - - /* step one, delete the existing extents in this range */ - /* FIXME blocksize != pagesize */ - if (start_pos < inode->i_size) { - err = btrfs_drop_extents(trans, root, inode, - start_pos, (pos + write_bytes + root->blocksize -1) & - ~((u64)root->blocksize - 1), &hint_block); - if (err) - goto failed_release; - } - - /* insert any holes we need to create */ - if (inode->i_size < start_pos) { - u64 last_pos_in_file; - u64 hole_size; - u64 mask = root->blocksize - 1; - last_pos_in_file = (isize + mask) & ~mask; - hole_size = (start_pos - last_pos_in_file + mask) & ~mask; - hole_size >>= inode->i_blkbits; - if (last_pos_in_file < start_pos) { - err = btrfs_insert_file_extent(trans, root, - inode->i_ino, - last_pos_in_file, - 0, 0, hole_size); - } - if (err) - goto failed_release; - } - - /* - * either allocate an extent for the new bytes or setup the key - * to show we are doing inline data in the extent - */ - if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || - pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { - err = btrfs_alloc_extent(trans, root, inode->i_ino, - num_blocks, 0, hint_block, (u64)-1, - &ins, 1); - if (err) - goto failed_truncate; - err = btrfs_insert_file_extent(trans, root, inode->i_ino, - start_pos, ins.objectid, ins.offset, - ins.offset); - if (err) - goto failed_truncate; - } else { - ins.offset = 0; - ins.objectid = 0; - } - BUG_ON(err); - alloc_extent_start = ins.objectid; - err = btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); - - for (i = 0; i < num_pages; i++) { - offset = pos & (PAGE_CACHE_SIZE -1); - this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes); - if (!page_has_buffers(pages[i])) { - create_empty_buffers(pages[i], - root->fs_info->sb->s_blocksize, - (1 << BH_Uptodate)); + if (!PagePrivate(pages[i])) { + SetPagePrivate(pages[i]); + set_page_private(pages[i], 1); + page_cache_get(pages[i]); } - head = page_buffers(pages[i]); - bh = head; - do { - err = btrfs_map_bh_to_logical(root, bh, - alloc_extent_start); - BUG_ON(err); - if (err) - goto failed_truncate; - bh = bh->b_this_page; - if (alloc_extent_start) - alloc_extent_start++; - } while (bh != head); - pos += this_write; - WARN_ON(this_write > write_bytes); - write_bytes -= this_write; } return 0; - -failed_release: - btrfs_drop_pages(pages, num_pages); - return err; - -failed_truncate: - btrfs_drop_pages(pages, num_pages); - if (pos > isize) - vmtruncate(inode, isize); - return err; - -out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); - goto failed_release; - } static ssize_t btrfs_file_write(struct file *file, const char __user *buf, @@ -685,7 +667,6 @@ out: page_cache_release(pinned[1]); *ppos = pos; current->backing_dev_info = NULL; - mark_inode_dirty(inode); return num_written ? num_written : err; } @@ -714,8 +695,8 @@ static int btrfs_sync_file(struct file *file, mutex_unlock(&root->fs_info->trans_mutex); /* - * ok we haven't committed the transaction yet, lets do a commit - */ + * ok we haven't committed the transaction yet, lets do a commit + */ trans = btrfs_start_transaction(root, 1); if (!trans) { ret = -ENOMEM; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 504d37dfa6c..fd4cb65fcf1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -121,6 +121,8 @@ void btrfs_read_locked_inode(struct inode *inode) switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + inode->i_mapping, GFP_NOFS); inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; break; @@ -169,7 +171,7 @@ static void fill_inode_item(struct btrfs_inode_item *item, BTRFS_I(inode)->block_group->key.objectid); } -static int btrfs_update_inode(struct btrfs_trans_handle *trans, +int btrfs_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) { @@ -400,6 +402,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int found_extent; int del_item; + btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1); path = btrfs_alloc_path(); path->reada = -1; BUG_ON(!path); @@ -511,6 +514,50 @@ error: return ret; } +static int btrfs_cow_one_page(struct btrfs_trans_handle *trans, + struct inode *inode, struct page *page, + size_t zero_start) +{ + char *kaddr; + int ret = 0; + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 alloc_hint = 0; + u64 page_start = page->index << PAGE_CACHE_SHIFT; + struct btrfs_key ins; + + if (!PagePrivate(page)) { + SetPagePrivate(page); + set_page_private(page, 1); + page_cache_get(page); + } + + btrfs_set_trans_block_group(trans, inode); + + ret = btrfs_drop_extents(trans, root, inode, + page_start, page_start + PAGE_CACHE_SIZE, + &alloc_hint); + if (ret) + goto out; + ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0, + alloc_hint, (u64)-1, &ins, 1); + if (ret) + goto out; + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + page_start, ins.objectid, 1, 1); + if (ret) + goto out; + SetPageChecked(page); + kaddr = kmap(page); + if (zero_start != PAGE_CACHE_SIZE) { + memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); + flush_dcache_page(page); + } + kunmap(page); + +out: + return ret; +} + /* * taken from block_truncate_page, but does cow as it zeros out * any bytes left in the last page in the file. @@ -518,16 +565,14 @@ error: static int btrfs_truncate_page(struct address_space *mapping, loff_t from) { struct inode *inode = mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; unsigned blocksize = 1 << inode->i_blkbits; pgoff_t index = from >> PAGE_CACHE_SHIFT; unsigned offset = from & (PAGE_CACHE_SIZE-1); struct page *page; - char *kaddr; int ret = 0; - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 alloc_hint = 0; - struct btrfs_key ins; - struct btrfs_trans_handle *trans; + u64 page_start; if ((offset & (blocksize - 1)) == 0) goto out; @@ -536,7 +581,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) page = grab_cache_page(mapping, index); if (!page) goto out; - if (!PageUptodate(page)) { ret = btrfs_readpage(NULL, page); lock_page(page); @@ -545,37 +589,24 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) goto out; } } + page_start = page->index << PAGE_CACHE_SHIFT; + mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - - ret = btrfs_drop_extents(trans, root, inode, - page->index << PAGE_CACHE_SHIFT, - (page->index + 1) << PAGE_CACHE_SHIFT, - &alloc_hint); - if (ret) - goto out; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0, - alloc_hint, (u64)-1, &ins, 1); - if (ret) - goto out; - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - page->index << PAGE_CACHE_SHIFT, - ins.objectid, 1, 1); - if (ret) - goto out; - SetPageChecked(page); - kaddr = kmap(page); - memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); - flush_dcache_page(page); - ret = btrfs_csum_file_block(trans, root, inode->i_ino, - page->index << PAGE_CACHE_SHIFT, - kaddr, PAGE_CACHE_SIZE); - kunmap(page); + ret = btrfs_cow_one_page(trans, inode, page, offset); + if (!ret) { + char *kaddr = kmap(page); + ret = btrfs_csum_file_block(trans, root, inode->i_ino, + page_start, kaddr, PAGE_CACHE_SIZE); + kunmap(page); + } + set_extent_dirty(&BTRFS_I(inode)->extent_tree, + page_start, page_start + PAGE_CACHE_SIZE - 1, + GFP_NOFS); + set_page_dirty(page); btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - set_page_dirty(page); unlock_page(page); page_cache_release(page); out: @@ -1095,6 +1126,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping->a_ops = &btrfs_aops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + inode->i_mapping, GFP_NOFS); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -1238,6 +1271,182 @@ out_unlock: return err; } +struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, + size_t page_offset, u64 start, u64 end, + int create) +{ + int ret; + int err = 0; + u64 blocknr; + u64 extent_start = 0; + u64 extent_end = 0; + u64 objectid = inode->i_ino; + u32 found_type; + int failed_insert = 0; + struct btrfs_path *path; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_file_extent_item *item; + struct btrfs_leaf *leaf; + struct btrfs_disk_key *found_key; + struct extent_map *em = NULL; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct btrfs_trans_handle *trans = NULL; + + path = btrfs_alloc_path(); + BUG_ON(!path); + mutex_lock(&root->fs_info->fs_mutex); + +again: + em = lookup_extent_mapping(em_tree, start, end); + if (em) { + goto out; + } + if (!em) { + em = alloc_extent_map(GFP_NOFS); + if (!em) { + err = -ENOMEM; + goto out; + } + em->start = 0; + em->end = 0; + } + em->bdev = inode->i_sb->s_bdev; + ret = btrfs_lookup_file_extent(NULL, root, path, + objectid, start, 0); + if (ret < 0) { + err = ret; + goto out; + } + + if (ret != 0) { + if (path->slots[0] == 0) + goto not_found; + path->slots[0]--; + } + + item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], + struct btrfs_file_extent_item); + leaf = btrfs_buffer_leaf(path->nodes[0]); + blocknr = btrfs_file_extent_disk_blocknr(item); + blocknr += btrfs_file_extent_offset(item); + + /* are we inside the extent that was found? */ + found_key = &leaf->items[path->slots[0]].key; + found_type = btrfs_disk_key_type(found_key); + if (btrfs_disk_key_objectid(found_key) != objectid || + found_type != BTRFS_EXTENT_DATA_KEY) { + goto not_found; + } + + found_type = btrfs_file_extent_type(item); + extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key); + if (found_type == BTRFS_FILE_EXTENT_REG) { + extent_end = extent_start + + (btrfs_file_extent_num_blocks(item) << inode->i_blkbits); + err = 0; + if (start < extent_start || start > extent_end) { + em->start = start; + if (start < extent_start) { + em->end = extent_end - 1; + } else { + em->end = end; + } + goto not_found_em; + } + if (btrfs_file_extent_disk_blocknr(item) == 0) { + em->start = extent_start; + em->end = extent_end - 1; + em->block_start = 0; + em->block_end = 0; + goto insert; + } + em->block_start = blocknr << inode->i_blkbits; + em->block_end = em->block_start + + (btrfs_file_extent_num_blocks(item) << + inode->i_blkbits) - 1; + em->start = extent_start; + em->end = extent_end - 1; + goto insert; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + char *ptr; + char *map; + u32 size; + + size = btrfs_file_extent_inline_len(leaf->items + + path->slots[0]); + extent_end = extent_start + size; + if (start < extent_start || start > extent_end) { + em->start = start; + if (start < extent_start) { + em->end = extent_end - 1; + } else { + em->end = end; + } + goto not_found_em; + } + em->block_start = EXTENT_MAP_INLINE; + em->block_end = EXTENT_MAP_INLINE; + em->start = extent_start; + em->end = extent_end - 1; + if (!page) { + goto insert; + } + ptr = btrfs_file_extent_inline_start(item); + map = kmap(page); + memcpy(map + page_offset, ptr, size); + flush_dcache_page(result->b_page); + kunmap(page); + set_extent_uptodate(em_tree, extent_start, + extent_end, GFP_NOFS); + goto insert; + } else { + printk("unkknown found_type %d\n", found_type); + WARN_ON(1); + } +not_found: + em->start = start; + em->end = end; +not_found_em: + em->block_start = 0; + em->block_end = 0; +insert: + btrfs_release_path(root, path); + if (em->start > start || em->end < start) { + printk("bad extent! %Lu %Lu start %Lu end %Lu\n", em->start, em->end, start, end); + WARN_ON(1); + err = -EIO; + goto out; + } + ret = add_extent_mapping(em_tree, em); + if (ret == -EEXIST) { + free_extent_map(em); + failed_insert++; + if (failed_insert > 5) { + printk("failing to insert %Lu %Lu\n", start, end); + err = -EIO; + goto out; + } + em = NULL; + goto again; + } + err = 0; +out: + btrfs_free_path(path); + if (trans) { + ret = btrfs_end_transaction(trans, root); + if (!err) + err = ret; + } + mutex_unlock(&root->fs_info->fs_mutex); + if (err) { + free_extent_map(em); + WARN_ON(1); + return ERR_PTR(err); + } + return em; +} + + /* * FIBMAP and others want to pass in a fake buffer head. They need to * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy @@ -1398,46 +1607,22 @@ int btrfs_get_block(struct inode *inode, sector_t iblock, return err; } -static int btrfs_get_block_csum(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create) -{ - int ret; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct page *page = result->b_page; - u64 offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(result); - struct btrfs_csum_item *item; - struct btrfs_path *path = NULL; - - mutex_lock(&root->fs_info->fs_mutex); - ret = btrfs_get_block_lock(inode, iblock, result, create); - if (ret) - goto out; - - path = btrfs_alloc_path(); - item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, offset, 0); - if (IS_ERR(item)) { - ret = PTR_ERR(item); - /* a csum that isn't present is a preallocated region. */ - if (ret == -ENOENT || ret == -EFBIG) - ret = 0; - result->b_private = NULL; - goto out; - } - memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE); -out: - if (path) - btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); - return ret; -} - static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock, struct buffer_head *result, int create) { struct btrfs_root *root = BTRFS_I(inode)->root; - mutex_lock(&root->fs_info->fs_mutex); - btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT); - mutex_unlock(&root->fs_info->fs_mutex); + u64 start = iblock << inode->i_blkbits; + u64 end = start + root->blocksize -1; + struct extent_map *em; + + em = btrfs_get_extent(inode, NULL, 0, start, end, 0); + if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE && + em->block_start != 0) { + u64 offset; + offset = start - em->start; + start = (em->block_start + offset) >> inode->i_blkbits; + btrfs_map_bh_to_logical(root, result, start); + } return 0; } @@ -1449,442 +1634,50 @@ static sector_t btrfs_bmap(struct address_space *as, sector_t block) static int btrfs_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to) { - return block_prepare_write(page, from, to, btrfs_get_block); + return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree, + page->mapping->host, page, from, to, + btrfs_get_extent); } -static void buffer_io_error(struct buffer_head *bh) +int btrfs_readpage(struct file *file, struct page *page) { - char b[BDEVNAME_SIZE]; - - printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n", - bdevname(bh->b_bdev, b), - (unsigned long long)bh->b_blocknr); + struct extent_map_tree *tree; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + return extent_read_full_page(tree, page, btrfs_get_extent); } - -/* - * I/O completion handler for block_read_full_page() - pages - * which come unlocked at the end of I/O. - */ -static void btrfs_end_buffer_async_read(struct buffer_head *bh, int uptodate) +static int btrfs_writepage(struct page *page, struct writeback_control *wbc) { - unsigned long flags; - struct buffer_head *first; - struct buffer_head *tmp; - struct page *page; - int page_uptodate = 1; - struct inode *inode; - int ret; - - BUG_ON(!buffer_async_read(bh)); - - page = bh->b_page; - inode = page->mapping->host; - if (uptodate) { - void *kaddr; - struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; - if (bh->b_private) { - char csum[BTRFS_CRC32_SIZE]; - kaddr = kmap_atomic(page, KM_IRQ0); - ret = btrfs_csum_data(root, kaddr + bh_offset(bh), - bh->b_size, csum); - BUG_ON(ret); - if (memcmp(csum, &bh->b_private, BTRFS_CRC32_SIZE)) { - u64 offset; - offset = (page->index << PAGE_CACHE_SHIFT) + - bh_offset(bh); - printk("btrfs csum failed ino %lu off %llu\n", - page->mapping->host->i_ino, - (unsigned long long)offset); - memset(kaddr + bh_offset(bh), 1, bh->b_size); - flush_dcache_page(page); - } - kunmap_atomic(kaddr, KM_IRQ0); - } - set_buffer_uptodate(bh); - } else { - clear_buffer_uptodate(bh); - if (printk_ratelimit()) - buffer_io_error(bh); - SetPageError(page); - } - - /* - * Be _very_ careful from here on. Bad things can happen if - * two buffer heads end IO at almost the same time and both - * decide that the page is now completely done. - */ - first = page_buffers(page); - local_irq_save(flags); - bit_spin_lock(BH_Uptodate_Lock, &first->b_state); - clear_buffer_async_read(bh); - unlock_buffer(bh); - tmp = bh; - do { - if (!buffer_uptodate(tmp)) - page_uptodate = 0; - if (buffer_async_read(tmp)) { - BUG_ON(!buffer_locked(tmp)); - goto still_busy; - } - tmp = tmp->b_this_page; - } while (tmp != bh); - bit_spin_unlock(BH_Uptodate_Lock, &first->b_state); - local_irq_restore(flags); - - /* - * If none of the buffers had errors and they are all - * uptodate then we can set the page uptodate. - */ - if (page_uptodate && !PageError(page)) - SetPageUptodate(page); - unlock_page(page); - return; - -still_busy: - bit_spin_unlock(BH_Uptodate_Lock, &first->b_state); - local_irq_restore(flags); - return; + struct extent_map_tree *tree; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } -/* - * Generic "read page" function for block devices that have the normal - * get_block functionality. This is most of the block device filesystems. - * Reads the page asynchronously --- the unlock_buffer() and - * set/clear_buffer_uptodate() functions propagate buffer state into the - * page struct once IO has completed. - */ -int btrfs_readpage(struct file *file, struct page *page) +static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) { - struct inode *inode = page->mapping->host; - sector_t iblock, lblock; - struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE]; - unsigned int blocksize; - int nr, i; - int fully_mapped = 1; - - BUG_ON(!PageLocked(page)); - blocksize = 1 << inode->i_blkbits; - if (!page_has_buffers(page)) - create_empty_buffers(page, blocksize, 0); - head = page_buffers(page); - - iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); - lblock = (i_size_read(inode)+blocksize-1) >> inode->i_blkbits; - bh = head; - nr = 0; - i = 0; - - do { - if (buffer_uptodate(bh)) - continue; - - if (!buffer_mapped(bh)) { - int err = 0; - - fully_mapped = 0; - if (iblock < lblock) { - WARN_ON(bh->b_size != blocksize); - err = btrfs_get_block_csum(inode, iblock, - bh, 0); - if (err) - SetPageError(page); - } - if (!buffer_mapped(bh)) { - void *kaddr = kmap_atomic(page, KM_USER0); - memset(kaddr + i * blocksize, 0, blocksize); - flush_dcache_page(page); - kunmap_atomic(kaddr, KM_USER0); - if (!err) - set_buffer_uptodate(bh); - continue; - } - /* - * get_block() might have updated the buffer - * synchronously - */ - if (buffer_uptodate(bh)) - continue; - } - arr[nr++] = bh; - } while (i++, iblock++, (bh = bh->b_this_page) != head); - - if (fully_mapped) - SetPageMappedToDisk(page); - - if (!nr) { - /* - * All buffers are uptodate - we can set the page uptodate - * as well. But not if get_block() returned an error. - */ - if (!PageError(page)) - SetPageUptodate(page); - unlock_page(page); - return 0; - } - - /* Stage two: lock the buffers */ - for (i = 0; i < nr; i++) { - bh = arr[i]; - lock_buffer(bh); - bh->b_end_io = btrfs_end_buffer_async_read; - set_buffer_async_read(bh); - } - - /* - * Stage 3: start the IO. Check for uptodateness - * inside the buffer lock in case another process reading - * the underlying blockdev brought it uptodate (the sct fix). - */ - for (i = 0; i < nr; i++) { - bh = arr[i]; - if (buffer_uptodate(bh)) - btrfs_end_buffer_async_read(bh, 1); - else - submit_bh(READ, bh); - } - return 0; -} - -/* - * Aside from a tiny bit of packed file data handling, this is the - * same as the generic code. - * - * While block_write_full_page is writing back the dirty buffers under - * the page lock, whoever dirtied the buffers may decide to clean them - * again at any time. We handle that by only looking at the buffer - * state inside lock_buffer(). - * - * If block_write_full_page() is called for regular writeback - * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a - * locked buffer. This only can happen if someone has written the buffer - * directly, with submit_bh(). At the address_space level PageWriteback - * prevents this contention from occurring. - */ -static int __btrfs_write_full_page(struct inode *inode, struct page *page, - struct writeback_control *wbc) -{ - int err; - sector_t block; - sector_t last_block; - struct buffer_head *bh, *head; - const unsigned blocksize = 1 << inode->i_blkbits; - int nr_underway = 0; - struct btrfs_root *root = BTRFS_I(inode)->root; - - BUG_ON(!PageLocked(page)); - - last_block = (i_size_read(inode) - 1) >> inode->i_blkbits; - - /* no csumming allowed when from PF_MEMALLOC */ - if (current->flags & PF_MEMALLOC) { - redirty_page_for_writepage(wbc, page); - unlock_page(page); - return 0; - } + struct extent_map_tree *tree; + int ret; - if (!page_has_buffers(page)) { - create_empty_buffers(page, blocksize, - (1 << BH_Dirty)|(1 << BH_Uptodate)); + if (page->private != 1) { + WARN_ON(1); + return try_to_free_buffers(page); } - - /* - * Be very careful. We have no exclusion from __set_page_dirty_buffers - * here, and the (potentially unmapped) buffers may become dirty at - * any time. If a buffer becomes dirty here after we've inspected it - * then we just miss that fact, and the page stays dirty. - * - * Buffers outside i_size may be dirtied by __set_page_dirty_buffers; - * handle that here by just cleaning them. - */ - - block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits); - head = page_buffers(page); - bh = head; - - /* - * Get all the dirty buffers mapped to disk addresses and - * handle any aliases from the underlying blockdev's mapping. - */ - do { - if (block > last_block) { - /* - * mapped buffers outside i_size will occur, because - * this page can be outside i_size when there is a - * truncate in progress. - */ - /* - * The buffer was zeroed by block_write_full_page() - */ - clear_buffer_dirty(bh); - set_buffer_uptodate(bh); - } else if (!buffer_mapped(bh) && buffer_dirty(bh)) { - WARN_ON(bh->b_size != blocksize); - err = btrfs_get_block(inode, block, bh, 0); - if (err) { - goto recover; - } - if (buffer_new(bh)) { - /* blockdev mappings never come here */ - clear_buffer_new(bh); - } - } - bh = bh->b_this_page; - block++; - } while (bh != head); - - do { - if (!buffer_mapped(bh)) - continue; - /* - * If it's a fully non-blocking write attempt and we cannot - * lock the buffer then redirty the page. Note that this can - * potentially cause a busy-wait loop from pdflush and kswapd - * activity, but those code paths have their own higher-level - * throttling. - */ - if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) { - lock_buffer(bh); - } else if (test_set_buffer_locked(bh)) { - redirty_page_for_writepage(wbc, page); - continue; - } - if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) { - struct btrfs_trans_handle *trans; - int ret; - u64 off = page->index << PAGE_CACHE_SHIFT; - char *kaddr; - - off += bh_offset(bh); - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - kaddr = kmap(page); - btrfs_csum_file_block(trans, root, inode->i_ino, - off, kaddr + bh_offset(bh), - bh->b_size); - kunmap(page); - ret = btrfs_end_transaction(trans, root); - BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); - mark_buffer_async_write(bh); - } else { - unlock_buffer(bh); - } - } while ((bh = bh->b_this_page) != head); - - /* - * The page and its buffers are protected by PageWriteback(), so we can - * drop the bh refcounts early. - */ - BUG_ON(PageWriteback(page)); - set_page_writeback(page); - - do { - struct buffer_head *next = bh->b_this_page; - if (buffer_async_write(bh)) { - submit_bh(WRITE, bh); - nr_underway++; - } - bh = next; - } while (bh != head); - unlock_page(page); - - err = 0; -done: - if (nr_underway == 0) { - /* - * The page was marked dirty, but the buffers were - * clean. Someone wrote them back by hand with - * ll_rw_block/submit_bh. A rare case. - */ - int uptodate = 1; - do { - if (!buffer_uptodate(bh)) { - uptodate = 0; - break; - } - bh = bh->b_this_page; - } while (bh != head); - if (uptodate) - SetPageUptodate(page); - end_page_writeback(page); + tree = &BTRFS_I(page->mapping->host)->extent_tree; + ret = try_release_extent_mapping(tree, page); + if (ret == 1) { + ClearPagePrivate(page); + set_page_private(page, 0); + page_cache_release(page); } - return err; - -recover: - /* - * ENOSPC, or some other error. We may already have added some - * blocks to the file, so we need to write these out to avoid - * exposing stale data. - * The page is currently locked and not marked for writeback - */ - bh = head; - /* Recovery: lock and submit the mapped buffers */ - do { - if (buffer_mapped(bh) && buffer_dirty(bh)) { - lock_buffer(bh); - mark_buffer_async_write(bh); - } else { - /* - * The buffer may have been set dirty during - * attachment to a dirty page. - */ - clear_buffer_dirty(bh); - } - } while ((bh = bh->b_this_page) != head); - SetPageError(page); - BUG_ON(PageWriteback(page)); - set_page_writeback(page); - do { - struct buffer_head *next = bh->b_this_page; - if (buffer_async_write(bh)) { - clear_buffer_dirty(bh); - submit_bh(WRITE, bh); - nr_underway++; - } - bh = next; - } while (bh != head); - unlock_page(page); - goto done; + return ret; } -static int btrfs_writepage(struct page *page, struct writeback_control *wbc) +static void btrfs_invalidatepage(struct page *page, unsigned long offset) { - struct inode * const inode = page->mapping->host; - loff_t i_size = i_size_read(inode); - const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT; - unsigned offset; - void *kaddr; - - /* Is the page fully inside i_size? */ - if (page->index < end_index) - return __btrfs_write_full_page(inode, page, wbc); - - /* Is the page fully outside i_size? (truncate in progress) */ - offset = i_size & (PAGE_CACHE_SIZE-1); - if (page->index >= end_index+1 || !offset) { - /* - * The page may have dirty, unmapped buffers. For example, - * they may have been added in ext3_writepage(). Make them - * freeable here, so the page does not leak. - */ - block_invalidatepage(page, 0); - unlock_page(page); - return 0; /* don't care */ - } + struct extent_map_tree *tree; - /* - * The page straddles i_size. It must be zeroed out on each and every - * writepage invokation because it may be mmapped. "A file is mapped - * in multiples of the page size. For a file that is not a multiple of - * the page size, the remaining memory is zeroed when mapped, and - * writes to that region are not written out to the file." - */ - kaddr = kmap_atomic(page, KM_USER0); - memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); - flush_dcache_page(page); - kunmap_atomic(kaddr, KM_USER0); - return __btrfs_write_full_page(inode, page, wbc); + tree = &BTRFS_I(page->mapping->host)->extent_tree; + extent_invalidatepage(tree, page, offset); + btrfs_releasepage(page, GFP_NOFS); } /* @@ -1905,28 +1698,39 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) { struct inode *inode = vma->vm_file->f_path.dentry->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; unsigned long end; loff_t size; int ret = -EINVAL; + u64 page_start; lock_page(page); wait_on_page_writeback(page); size = i_size_read(inode); + page_start = page->index << PAGE_CACHE_SHIFT; + if ((page->mapping != inode->i_mapping) || - ((page->index << PAGE_CACHE_SHIFT) > size)) { + (page_start > size)) { /* page got truncated out from underneath us */ goto out_unlock; } /* page is wholly or partially inside EOF */ - if (((page->index + 1) << PAGE_CACHE_SHIFT) > size) + if (page_start + PAGE_CACHE_SIZE > size) end = size & ~PAGE_CACHE_MASK; else end = PAGE_CACHE_SIZE; - ret = btrfs_prepare_write(NULL, page, 0, end); - if (!ret) - ret = btrfs_commit_write(NULL, page, 0, end); + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + ret = btrfs_cow_one_page(trans, inode, page, end); + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + set_extent_dirty(&BTRFS_I(inode)->extent_tree, + page_start, page_start + PAGE_CACHE_SIZE - 1, + GFP_NOFS); + set_page_dirty(page); out_unlock: unlock_page(page); @@ -1962,21 +1766,8 @@ static void btrfs_truncate(struct inode *inode) int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to) { - struct inode *inode = page->mapping->host; - struct buffer_head *bh; - loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; - - SetPageUptodate(page); - bh = page_buffers(page); - set_buffer_uptodate(bh); - if (buffer_mapped(bh) && bh->b_blocknr != 0) { - set_page_dirty(page); - } - if (pos > inode->i_size) { - i_size_write(inode, pos); - mark_inode_dirty(inode); - } - return 0; + return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree, + page->mapping->host, page, from, to); } static int create_subvol(struct btrfs_root *root, char *name, int namelen) @@ -2471,6 +2262,8 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping->a_ops = &btrfs_aops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + inode->i_mapping, GFP_NOFS); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2553,6 +2346,9 @@ static struct address_space_operations btrfs_aops = { .prepare_write = btrfs_prepare_write, .commit_write = btrfs_commit_write, .bmap = btrfs_bmap, + .invalidatepage = btrfs_invalidatepage, + .releasepage = btrfs_releasepage, + .set_page_dirty = __set_page_dirty_nobuffers, }; static struct address_space_operations btrfs_symlink_aops = { diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 74f3de47423..7b7c32c73a2 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -185,6 +185,7 @@ static int __init init_btrfs_fs(void) err = btrfs_init_cachep(); if (err) return err; + extent_map_init(); return register_filesystem(&btrfs_fs_type); } @@ -192,6 +193,7 @@ static void __exit exit_btrfs_fs(void) { btrfs_exit_transaction_sys(); btrfs_destroy_cachep(); + extent_map_exit(); unregister_filesystem(&btrfs_fs_type); } -- cgit v1.2.3-70-g09d2 From b888db2bd7b67f190b32934e6a86181f262ac3ec Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 27 Aug 2007 16:49:44 -0400 Subject: Btrfs: Add delayed allocation to the extent based page tree code Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 9 --- fs/btrfs/extent_map.c | 147 +++++++++++++++++++++++++++++++++++++------------ fs/btrfs/extent_map.h | 3 + fs/btrfs/file.c | 50 +++++++---------- fs/btrfs/inode.c | 141 ++++++++++++++++++++++++++--------------------- fs/btrfs/super.c | 1 + 6 files changed, 216 insertions(+), 135 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 4049aadbeda..c31e84d4265 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -39,7 +39,6 @@ static int cache_block_group(struct btrfs_root *root, u64 i; u64 last = 0; u64 hole_size; - u64 limit; int found = 0; root = root->fs_info->extent_root; @@ -62,7 +61,6 @@ static int cache_block_group(struct btrfs_root *root, return ret; if (ret && path->slots[0] > 0) path->slots[0]--; - limit = block_group->key.objectid + block_group->key.offset; while(1) { leaf = btrfs_buffer_leaf(path->nodes[0]); slot = path->slots[0]; @@ -982,7 +980,6 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_block_group_cache *block_group; int full_scan = 0; int wrapped = 0; - u64 limit; WARN_ON(num_blocks < 1); ins->flags = 0; @@ -1049,12 +1046,6 @@ check_failed: l = btrfs_buffer_leaf(path->nodes[0]); slot = path->slots[0]; if (slot >= btrfs_header_nritems(&l->header)) { - if (start_found) - limit = last_block + - (block_group->key.offset >> 1); - else - limit = search_start + - (block_group->key.offset >> 1); ret = btrfs_next_leaf(root, path); if (ret == 0) continue; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index d378edf0964..a9c7419615b 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -30,9 +30,6 @@ struct tree_entry { #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) -static LIST_HEAD(all_states); -spinlock_t state_lock = SPIN_LOCK_UNLOCKED; - void __init extent_map_init(void) { extent_map_cache = kmem_cache_create("extent_map", @@ -49,15 +46,6 @@ void __init extent_map_init(void) void __exit extent_map_exit(void) { - while(!list_empty(&all_states)) { - struct extent_state *state; - struct list_head *cur = all_states.next; - state = list_entry(cur, struct extent_state, list); - printk("found leaked state %Lu %Lu state %d in_tree %d\n", - state->start, state->end, state->state, state->in_tree); - list_del(&state->list); - kfree(state); - } if (extent_map_cache) kmem_cache_destroy(extent_map_cache); if (extent_state_cache) @@ -69,6 +57,7 @@ void extent_map_tree_init(struct extent_map_tree *tree, { tree->map.rb_node = NULL; tree->state.rb_node = NULL; + tree->fill_delalloc = NULL; rwlock_init(&tree->lock); tree->mapping = mapping; } @@ -106,9 +95,6 @@ struct extent_state *alloc_extent_state(gfp_t mask) state->in_tree = 0; atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); - spin_lock_irq(&state_lock); - list_add(&state->list, &all_states); - spin_unlock_irq(&state_lock); return state; } EXPORT_SYMBOL(alloc_extent_state); @@ -117,9 +103,6 @@ void free_extent_state(struct extent_state *state) { if (atomic_dec_and_test(&state->refs)) { WARN_ON(state->in_tree); - spin_lock_irq(&state_lock); - list_del_init(&state->list); - spin_unlock_irq(&state_lock); kmem_cache_free(extent_state_cache, state); } } @@ -369,7 +352,7 @@ static int insert_state(struct extent_map_tree *tree, if (node) { struct extent_state *found; found = rb_entry(node, struct extent_state, rb_node); -printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); + printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); free_extent_state(state); return -EEXIST; } @@ -408,7 +391,7 @@ static int split_state(struct extent_map_tree *tree, struct extent_state *orig, if (node) { struct extent_state *found; found = rb_entry(node, struct extent_state, rb_node); -printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); + printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); free_extent_state(prealloc); return -EEXIST; } @@ -792,10 +775,20 @@ int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_dirty); +int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, + EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_delalloc); + int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask) { - return clear_extent_bit(tree, start, end, EXTENT_DIRTY, 0, 0, mask); + return clear_extent_bit(tree, start, end, + EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask); } EXPORT_SYMBOL(clear_extent_dirty); @@ -922,6 +915,62 @@ int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(set_range_writeback); +u64 find_lock_delalloc_range(struct extent_map_tree *tree, + u64 start, u64 lock_start, u64 *end, u64 max_bytes) +{ + struct rb_node *node; + struct extent_state *state; + u64 cur_start = start; + u64 found = 0; + u64 total_bytes = 0; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ +search_again: + node = tree_search(&tree->state, cur_start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != cur_start) { + goto out; + } + if (!(state->state & EXTENT_DELALLOC)) { + goto out; + } + if (state->start >= lock_start) { + if (state->state & EXTENT_LOCKED) { + DEFINE_WAIT(wait); + atomic_inc(&state->refs); + write_unlock_irq(&tree->lock); + schedule(); + write_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + free_extent_state(state); + goto search_again; + } + state->state |= EXTENT_LOCKED; + } + found++; + *end = state->end; + cur_start = state->end + 1; + node = rb_next(node); + if (!node) + break; + total_bytes = state->end - state->start + 1; + if (total_bytes >= max_bytes) + break; + } +out: + write_unlock_irq(&tree->lock); + return found; +} + /* * helper function to lock both pages and extents in the tree. * pages must be locked first. @@ -1285,6 +1334,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, if (!PagePrivate(page)) { SetPagePrivate(page); set_page_private(page, 1); + WARN_ON(!page->mapping->a_ops->invalidatepage); page_cache_get(page); } @@ -1384,7 +1434,10 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, size_t blocksize; loff_t i_size = i_size_read(inode); unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; + u64 nr_delalloc; + u64 delalloc_end; + WARN_ON(!PageLocked(page)); if (page->index > end_index) { clear_extent_dirty(tree, start, page_end, GFP_NOFS); unlock_page(page); @@ -1400,11 +1453,34 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, if (!PagePrivate(page)) { SetPagePrivate(page); set_page_private(page, 1); + WARN_ON(!page->mapping->a_ops->invalidatepage); page_cache_get(page); } - end = page_end; lock_extent(tree, start, page_end, GFP_NOFS); + nr_delalloc = find_lock_delalloc_range(tree, start, page_end + 1, + &delalloc_end, + 128 * 1024 * 1024); + if (nr_delalloc) { + tree->fill_delalloc(inode, start, delalloc_end); + if (delalloc_end >= page_end + 1) { + clear_extent_bit(tree, page_end + 1, delalloc_end, + EXTENT_LOCKED | EXTENT_DELALLOC, + 1, 0, GFP_NOFS); + } + clear_extent_bit(tree, start, page_end, EXTENT_DELALLOC, + 0, 0, GFP_NOFS); + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { + printk("found delalloc bits after clear extent_bit\n"); + } + } else if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { + printk("found delalloc bits after find_delalloc_range returns 0\n"); + } + + end = page_end; + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { + printk("found delalloc bits after lock_extent\n"); + } if (last_byte <= start) { clear_extent_dirty(tree, start, page_end, GFP_NOFS); @@ -1419,7 +1495,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, clear_extent_dirty(tree, cur, page_end, GFP_NOFS); break; } - em = get_extent(inode, page, page_offset, cur, end, 1); + em = get_extent(inode, page, page_offset, cur, end, 0); if (IS_ERR(em) || !em) { SetPageError(page); break; @@ -1507,6 +1583,7 @@ int extent_commit_write(struct extent_map_tree *tree, if (!PagePrivate(page)) { SetPagePrivate(page); set_page_private(page, 1); + WARN_ON(!page->mapping->a_ops->invalidatepage); page_cache_get(page); } @@ -1543,6 +1620,7 @@ int extent_prepare_write(struct extent_map_tree *tree, if (!PagePrivate(page)) { SetPagePrivate(page); set_page_private(page, 1); + WARN_ON(!page->mapping->a_ops->invalidatepage); page_cache_get(page); } block_start = (page_start + from) & ~((u64)blocksize - 1); @@ -1628,29 +1706,28 @@ int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page) u64 start = page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; u64 orig_start = start; + int ret = 1; while (start <= end) { em = lookup_extent_mapping(tree, start, end); if (!em || IS_ERR(em)) break; - if (test_range_bit(tree, em->start, em->end, - EXTENT_LOCKED, 0)) { + if (!test_range_bit(tree, em->start, em->end, + EXTENT_LOCKED, 0)) { + remove_extent_mapping(tree, em); + /* once for the rb tree */ free_extent_map(em); - start = em->end + 1; -printk("range still locked %Lu %Lu\n", em->start, em->end); - break; } - remove_extent_mapping(tree, em); start = em->end + 1; - /* once for the rb tree */ - free_extent_map(em); /* once for us */ free_extent_map(em); } - WARN_ON(test_range_bit(tree, orig_start, end, EXTENT_WRITEBACK, 0)); - clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, - 1, 1, GFP_NOFS); - return 1; + if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) + ret = 0; + else + clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, + 1, 1, GFP_NOFS); + return ret; } EXPORT_SYMBOL(try_release_extent_mapping); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 108944aab4b..e91a2e9619e 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -11,6 +11,7 @@ struct extent_map_tree { struct rb_root state; struct address_space *mapping; rwlock_t lock; + int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); }; /* note, this must start with the same fields as fs/extent_map.c:tree_entry */ @@ -74,6 +75,8 @@ int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); +int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask); int extent_invalidatepage(struct extent_map_tree *tree, struct page *page, unsigned long offset); int extent_write_full_page(struct extent_map_tree *tree, struct page *page, diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 71a481894ab..d3d39e4a279 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -143,7 +143,6 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, struct inode *inode = file->f_path.dentry->d_inode; struct extent_map *em; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - struct btrfs_key ins; u64 hint_block; u64 num_blocks; u64 start_pos; @@ -162,6 +161,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, inode->i_blkbits; end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1; + lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); if (!trans) { @@ -179,16 +179,6 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, /* FIXME...EIEIO, ENOSPC and more */ - /* step one, delete the existing extents in this range */ - /* FIXME blocksize != pagesize */ - if (start_pos < inode->i_size) { - err = btrfs_drop_extents(trans, root, inode, - start_pos, (pos + write_bytes + root->blocksize -1) & - ~((u64)root->blocksize - 1), &hint_block); - if (err) - goto failed; - } - /* insert any holes we need to create */ if (inode->i_size < start_pos) { u64 last_pos_in_file; @@ -213,29 +203,28 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, */ if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { - err = btrfs_alloc_extent(trans, root, inode->i_ino, - num_blocks, 0, hint_block, (u64)-1, - &ins, 1); - BUG_ON(err); - err = btrfs_insert_file_extent(trans, root, inode->i_ino, - start_pos, ins.objectid, ins.offset, - ins.offset); - BUG_ON(err); - em->start = start_pos; - em->end = end_of_last_block; - em->block_start = ins.objectid << inode->i_blkbits; - em->block_end = em->block_start + - (ins.offset << inode->i_blkbits) - 1; - set_extent_dirty(em_tree, start_pos, end_of_last_block, - GFP_NOFS); - err = add_extent_mapping(em_tree, em); + u64 last_end; for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; SetPageUptodate(p); - __set_page_dirty_nobuffers(p); + set_page_dirty(p); } + last_end = pages[num_pages -1]->index << PAGE_CACHE_SHIFT; + last_end += PAGE_CACHE_SIZE - 1; + set_extent_delalloc(em_tree, start_pos, end_of_last_block, + GFP_NOFS); } else { struct page *p = pages[0]; + /* step one, delete the existing extents in this range */ + /* FIXME blocksize != pagesize */ + if (start_pos < inode->i_size) { + err = btrfs_drop_extents(trans, root, inode, start_pos, + (pos + write_bytes + root->blocksize -1) & + ~((u64)root->blocksize - 1), &hint_block); + if (err) + goto failed; + } + err = insert_inline_extent(trans, root, inode, start_pos, end_pos - start_pos, p, 0); BUG_ON(err); @@ -253,6 +242,7 @@ failed: err = btrfs_end_transaction(trans, root); out_unlock: mutex_unlock(&root->fs_info->fs_mutex); + unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); free_extent_map(em); return err; } @@ -531,7 +521,7 @@ static int prepare_pages(struct btrfs_root *root, u64 num_blocks; u64 start_pos; - start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1); + start_pos = pos & ~((u64)root->blocksize - 1); num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> inode->i_blkbits; @@ -548,8 +538,10 @@ static int prepare_pages(struct btrfs_root *root, if (!PagePrivate(pages[i])) { SetPagePrivate(pages[i]); set_page_private(pages[i], 1); + WARN_ON(!pages[i]->mapping->a_ops->invalidatepage); page_cache_get(pages[i]); } + WARN_ON(!PageLocked(pages[i])); } return 0; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index fd4cb65fcf1..dc181089aa7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -70,6 +70,40 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, }; +static int run_delalloc_range(struct inode *inode, u64 start, u64 end) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + struct btrfs_key ins; + u64 alloc_hint = 0; + u64 num_blocks; + int ret; + u64 blocksize = 1 << inode->i_blkbits; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + BUG_ON(!trans); + num_blocks = (end - start + blocksize) & ~(blocksize - 1); + ret = btrfs_drop_extents(trans, root, inode, + start, start + num_blocks, &alloc_hint); + num_blocks = num_blocks >> inode->i_blkbits; + ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0, + alloc_hint, (u64)-1, &ins, 1); + if (ret) { + WARN_ON(1); + goto out; + } + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + start, ins.objectid, ins.offset, + ins.offset); +out: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + + void btrfs_read_locked_inode(struct inode *inode) { struct btrfs_path *path; @@ -121,8 +155,7 @@ void btrfs_read_locked_inode(struct inode *inode) switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, - inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; break; @@ -450,11 +483,17 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, BUG_ON(ret); } if (item_end < inode->i_size) { - if (found_type) { - btrfs_set_key_type(&key, found_type - 1); - continue; + if (found_type == BTRFS_DIR_ITEM_KEY) { + found_type = BTRFS_INODE_ITEM_KEY; + } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { + found_type = BTRFS_CSUM_ITEM_KEY; + } else if (found_type) { + found_type--; + } else { + break; } - break; + btrfs_set_key_type(&key, found_type - 1); + continue; } if (btrfs_disk_key_offset(found_key) >= inode->i_size) del_item = 1; @@ -514,47 +553,34 @@ error: return ret; } -static int btrfs_cow_one_page(struct btrfs_trans_handle *trans, - struct inode *inode, struct page *page, +static int btrfs_cow_one_page(struct inode *inode, struct page *page, size_t zero_start) { char *kaddr; int ret = 0; - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 alloc_hint = 0; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; u64 page_start = page->index << PAGE_CACHE_SHIFT; - struct btrfs_key ins; + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; if (!PagePrivate(page)) { SetPagePrivate(page); set_page_private(page, 1); + WARN_ON(!page->mapping->a_ops->invalidatepage); page_cache_get(page); } - btrfs_set_trans_block_group(trans, inode); - - ret = btrfs_drop_extents(trans, root, inode, - page_start, page_start + PAGE_CACHE_SIZE, - &alloc_hint); - if (ret) - goto out; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0, - alloc_hint, (u64)-1, &ins, 1); - if (ret) - goto out; - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - page_start, ins.objectid, 1, 1); - if (ret) - goto out; - SetPageChecked(page); - kaddr = kmap(page); + lock_extent(em_tree, page_start, page_end, GFP_NOFS); + set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start, + page_end, GFP_NOFS); if (zero_start != PAGE_CACHE_SIZE) { + kaddr = kmap(page); memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); flush_dcache_page(page); + kunmap(page); } - kunmap(page); + set_page_dirty(page); + unlock_extent(em_tree, page_start, page_end, GFP_NOFS); -out: return ret; } @@ -565,8 +591,6 @@ out: static int btrfs_truncate_page(struct address_space *mapping, loff_t from) { struct inode *inode = mapping->host; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; unsigned blocksize = 1 << inode->i_blkbits; pgoff_t index = from >> PAGE_CACHE_SHIFT; unsigned offset = from & (PAGE_CACHE_SIZE-1); @@ -591,21 +615,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) } page_start = page->index << PAGE_CACHE_SHIFT; - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - ret = btrfs_cow_one_page(trans, inode, page, offset); - if (!ret) { - char *kaddr = kmap(page); - ret = btrfs_csum_file_block(trans, root, inode->i_ino, - page_start, kaddr, PAGE_CACHE_SIZE); - kunmap(page); - } - set_extent_dirty(&BTRFS_I(inode)->extent_tree, - page_start, page_start + PAGE_CACHE_SIZE - 1, - GFP_NOFS); - set_page_dirty(page); - btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); + ret = btrfs_cow_one_page(inode, page, offset); unlock_page(page); page_cache_release(page); @@ -757,6 +767,8 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) struct btrfs_iget_args *args = p; inode->i_ino = args->ino; BTRFS_I(inode)->root = args->root; + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + inode->i_mapping, GFP_NOFS); return 0; } @@ -968,7 +980,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, if (!inode) return ERR_PTR(-ENOMEM); + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->root = root; + if (mode & S_IFDIR) owner = 0; else @@ -1128,6 +1143,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -1344,9 +1360,11 @@ again: extent_end = extent_start + (btrfs_file_extent_num_blocks(item) << inode->i_blkbits); err = 0; - if (start < extent_start || start > extent_end) { + if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { + if (end < extent_start) + goto not_found; em->end = extent_end - 1; } else { em->end = end; @@ -1375,9 +1393,11 @@ again: size = btrfs_file_extent_inline_len(leaf->items + path->slots[0]); extent_end = extent_start + size; - if (start < extent_start || start > extent_end) { + if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { + if (end < extent_start) + goto not_found; em->end = extent_end - 1; } else { em->end = end; @@ -1412,8 +1432,7 @@ not_found_em: insert: btrfs_release_path(root, path); if (em->start > start || em->end < start) { - printk("bad extent! %Lu %Lu start %Lu end %Lu\n", em->start, em->end, start, end); - WARN_ON(1); + printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end); err = -EIO; goto out; } @@ -1648,6 +1667,13 @@ int btrfs_readpage(struct file *file, struct page *page) static int btrfs_writepage(struct page *page, struct writeback_control *wbc) { struct extent_map_tree *tree; + + + if (current->flags & PF_MEMALLOC) { + redirty_page_for_writepage(wbc, page); + unlock_page(page); + return 0; + } tree = &BTRFS_I(page->mapping->host)->extent_tree; return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } @@ -1698,8 +1724,6 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) { struct inode *inode = vma->vm_file->f_path.dentry->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; unsigned long end; loff_t size; int ret = -EINVAL; @@ -1722,15 +1746,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) else end = PAGE_CACHE_SIZE; - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - ret = btrfs_cow_one_page(trans, inode, page, end); - btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); - set_extent_dirty(&BTRFS_I(inode)->extent_tree, - page_start, page_start + PAGE_CACHE_SIZE - 1, - GFP_NOFS); - set_page_dirty(page); + ret = btrfs_cow_one_page(inode, page, end); out_unlock: unlock_page(page); @@ -2264,6 +2280,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 7b7c32c73a2..7da95538745 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -83,6 +83,7 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) bi->location.offset = 0; bi->location.flags = 0; bi->root = tree_root; + btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY); if (!inode) { -- cgit v1.2.3-70-g09d2 From 58176a9604c5db1784d2c979aea472b3be40b6f0 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 29 Aug 2007 15:47:34 -0400 Subject: Btrfs: Add per-root block accounting and sysfs entries Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.h | 54 +++++++++++ fs/btrfs/disk-io.c | 31 ++++++- fs/btrfs/disk-io.h | 3 +- fs/btrfs/extent-tree.c | 17 +++- fs/btrfs/inode.c | 13 ++- fs/btrfs/super.c | 14 +++ fs/btrfs/sysfs.c | 236 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/transaction.c | 34 ++++++- 9 files changed, 393 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index ea7a2287035..ee505c87ecf 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -5,7 +5,7 @@ obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ - extent_map.o + extent_map.o sysfs.o #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \ # root-tree.o dir-item.o hash.o file-item.o inode-item.o \ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 947f061ed11..ffeee546a0d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -22,6 +22,7 @@ #include #include #include +#include #include "bit-radix.h" struct btrfs_trans_handle; @@ -313,6 +314,8 @@ struct btrfs_fs_info { struct list_head trans_list; struct list_head dead_roots; struct delayed_work trans_work; + struct kobject super_kobj; + struct completion kobj_unregister; int do_barriers; int closing; }; @@ -328,6 +331,8 @@ struct btrfs_root { struct btrfs_key root_key; struct btrfs_fs_info *fs_info; struct inode *inode; + struct kobject root_kobj; + struct completion kobj_unregister; u64 objectid; u64 last_trans; u32 blocksize; @@ -338,6 +343,7 @@ struct btrfs_root { struct btrfs_key defrag_progress; int defrag_running; int defrag_level; + char *name; }; /* the lower bits in the key flags defines the item type */ @@ -814,6 +820,28 @@ static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val) item->flags = cpu_to_le32(val); } +static inline void btrfs_set_root_blocks_used(struct btrfs_root_item *item, + u64 val) +{ + item->blocks_used = cpu_to_le64(val); +} + +static inline u64 btrfs_root_blocks_used(struct btrfs_root_item *item) +{ + return le64_to_cpu(item->blocks_used); +} + +static inline void btrfs_set_root_block_limit(struct btrfs_root_item *item, + u64 val) +{ + item->block_limit = cpu_to_le64(val); +} + +static inline u64 btrfs_root_block_limit(struct btrfs_root_item *item) +{ + return le64_to_cpu(item->block_limit); +} + static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s) { return le64_to_cpu(s->blocknr); @@ -1014,6 +1042,23 @@ static inline void btrfs_memmove(struct btrfs_root *root, memmove(dst, src, nr); } +static inline int btrfs_set_root_name(struct btrfs_root *root, + const char *name, int len) +{ + /* if we already have a name just free it */ + if (root->name) + kfree(root->name); + + root->name = kmalloc(len+1, GFP_KERNEL); + if (!root->name) + return -ENOMEM; + + memcpy(root->name, name, len); + root->name[len] ='\0'; + + return 0; +} + /* helper function to cast into the data area of the leaf. */ #define btrfs_item_ptr(leaf, slot, type) \ ((type *)(btrfs_leaf_data(leaf) + \ @@ -1191,4 +1236,13 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, /* tree-defrag.c */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only); + +/* sysfs.c */ +int btrfs_init_sysfs(void); +void btrfs_exit_sysfs(void); +int btrfs_sysfs_add_super(struct btrfs_fs_info *fs); +int btrfs_sysfs_add_root(struct btrfs_root *root); +void btrfs_sysfs_del_root(struct btrfs_root *root); +void btrfs_sysfs_del_super(struct btrfs_fs_info *root); + #endif diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index b2f79878d51..c25ef0a68f1 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -294,9 +294,12 @@ static int __setup_root(int blocksize, root->last_trans = 0; root->highest_inode = 0; root->last_inode_alloc = 0; + root->name = NULL; memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); + memset(&root->root_kobj, 0, sizeof(root->root_kobj)); + init_completion(&root->kobj_unregister); root->defrag_running = 0; root->defrag_level = 0; root->root_key.objectid = objectid; @@ -384,7 +387,8 @@ insert: } struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, - struct btrfs_key *location) + struct btrfs_key *location, + const char *name, int namelen) { struct btrfs_root *root; int ret; @@ -405,6 +409,22 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, kfree(root); return ERR_PTR(ret); } + + ret = btrfs_set_root_name(root, name, namelen); + if (ret) { + brelse(root->node); + kfree(root); + return ERR_PTR(ret); + } + + ret = btrfs_sysfs_add_root(root); + if (ret) { + brelse(root->node); + kfree(root->name); + kfree(root); + return ERR_PTR(ret); + } + return root; } @@ -433,6 +453,8 @@ struct btrfs_root *open_ctree(struct super_block *sb) INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL); INIT_LIST_HEAD(&fs_info->trans_list); INIT_LIST_HEAD(&fs_info->dead_roots); + memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); + init_completion(&fs_info->kobj_unregister); sb_set_blocksize(sb, 4096); fs_info->running_transaction = NULL; fs_info->last_trans_committed = 0; @@ -500,8 +522,10 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->generation = btrfs_super_generation(disk_super) + 1; ret = btrfs_find_dead_roots(tree_root); - if (ret) + if (ret) { + mutex_unlock(&fs_info->fs_mutex); goto fail_tree_root; + } mutex_unlock(&fs_info->fs_mutex); return tree_root; @@ -553,12 +577,15 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) { radix_tree_delete(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid); + btrfs_sysfs_del_root(root); if (root->inode) iput(root->inode); if (root->node) brelse(root->node); if (root->commit_root) brelse(root->commit_root); + if (root->name) + kfree(root->name); kfree(root); return 0; } diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 5261733b873..da6bb72750f 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -66,7 +66,8 @@ struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr); int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, char *result); struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, - struct btrfs_key *location); + struct btrfs_key *location, + const char *name, int namelen); struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, struct btrfs_key *location); u64 bh_blocknr(struct buffer_head *bh); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c31e84d4265..ff3f7c2be60 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -858,16 +858,23 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_extent_refs(ei, refs); btrfs_mark_buffer_dirty(path->nodes[0]); if (refs == 0) { - u64 super_blocks_used; + u64 super_blocks_used, root_blocks_used; if (pin) { ret = pin_down_block(root, blocknr, 0); BUG_ON(ret); } + /* block accounting for super block */ super_blocks_used = btrfs_super_blocks_used(&info->super_copy); btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used - num_blocks); + + /* block accounting for root item */ + root_blocks_used = btrfs_root_blocks_used(&root->root_item); + btrfs_set_root_blocks_used(&root->root_item, + root_blocks_used - num_blocks); + ret = btrfs_del_item(trans, extent_root, path); if (ret) { return ret; @@ -1175,7 +1182,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, { int ret; int pending_ret; - u64 super_blocks_used; + u64 super_blocks_used, root_blocks_used; u64 search_start = 0; struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; @@ -1193,10 +1200,16 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, if (ret) return ret; + /* block accounting for super block */ super_blocks_used = btrfs_super_blocks_used(&info->super_copy); btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used + num_blocks); + /* block accounting for root item */ + root_blocks_used = btrfs_root_blocks_used(&root->root_item); + btrfs_set_root_blocks_used(&root->root_item, root_blocks_used + + num_blocks); + if (root == extent_root) { BUG_ON(num_blocks != 1); set_radix_bit(&root->fs_info->extent_ins_radix, ins->objectid); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index dc181089aa7..2e3918e6049 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -733,7 +733,8 @@ out: */ static int fixup_tree_root_location(struct btrfs_root *root, struct btrfs_key *location, - struct btrfs_root **sub_root) + struct btrfs_root **sub_root, + struct dentry *dentry) { struct btrfs_path *path; struct btrfs_root_item *ri; @@ -747,7 +748,9 @@ static int fixup_tree_root_location(struct btrfs_root *root, BUG_ON(!path); mutex_lock(&root->fs_info->fs_mutex); - *sub_root = btrfs_read_fs_root(root->fs_info, location); + *sub_root = btrfs_read_fs_root(root->fs_info, location, + dentry->d_name.name, + dentry->d_name.len); if (IS_ERR(*sub_root)) return PTR_ERR(*sub_root); @@ -812,7 +815,8 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, return ERR_PTR(ret); inode = NULL; if (location.objectid) { - ret = fixup_tree_root_location(root, &location, &sub_root); + ret = fixup_tree_root_location(root, &location, &sub_root, + dentry); if (ret < 0) return ERR_PTR(ret); if (ret > 0) @@ -1829,6 +1833,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol)); btrfs_set_root_refs(&root_item, 1); + btrfs_set_root_blocks_used(&root_item, 0); memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); root_item.drop_level = 0; brelse(subvol); @@ -1865,7 +1870,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail_commit; - new_root = btrfs_read_fs_root(root->fs_info, &key); + new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); BUG_ON(!new_root); trans = btrfs_start_transaction(new_root, 1); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 7da95538745..a68101ae1a3 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -45,12 +45,14 @@ static struct super_operations btrfs_super_ops; static void btrfs_put_super (struct super_block * sb) { struct btrfs_root *root = btrfs_sb(sb); + struct btrfs_fs_info *fs = root->fs_info; int ret; ret = close_ctree(root); if (ret) { printk("close ctree returns %d\n", ret); } + btrfs_sysfs_del_super(fs); sb->s_fs_info = NULL; } @@ -101,6 +103,12 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) err = -ENOMEM; goto fail_close; } + + /* this does the super kobj at the same time */ + err = btrfs_sysfs_add_super(tree_root->fs_info); + if (err) + goto fail_close; + sb->s_root = root_dentry; btrfs_transaction_queue_work(tree_root, HZ * 30); return 0; @@ -182,6 +190,11 @@ static struct super_operations btrfs_super_ops = { static int __init init_btrfs_fs(void) { int err; + + err = btrfs_init_sysfs(); + if (err) + return err; + btrfs_init_transaction_sys(); err = btrfs_init_cachep(); if (err) @@ -196,6 +209,7 @@ static void __exit exit_btrfs_fs(void) btrfs_destroy_cachep(); extent_map_exit(); unregister_filesystem(&btrfs_fs_type); + btrfs_exit_sysfs(); } module_init(init_btrfs_fs) diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index db8917e5b25..2058783373e 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -16,6 +16,242 @@ * Boston, MA 021110-1307, USA. */ +#include +#include +#include +#include +#include +#include +#include + #include "ctree.h" #include "disk-io.h" #include "transaction.h" + +static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%llu\n", + (unsigned long long)btrfs_root_blocks_used(&root->root_item)); +} + +static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%llu\n", + (unsigned long long)btrfs_root_block_limit(&root->root_item)); +} + +static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%llu\n", + (unsigned long long)btrfs_super_blocks_used(fs->disk_super)); +} + +static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%llu\n", + (unsigned long long)btrfs_super_total_blocks(fs->disk_super)); +} + +static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf) +{ + return snprintf(buf, PAGE_SIZE, "%llu\n", + (unsigned long long)btrfs_super_blocksize(fs->disk_super)); +} + +/* this is for root attrs (subvols/snapshots) */ +struct btrfs_root_attr { + struct attribute attr; + ssize_t (*show)(struct btrfs_root *, char *); + ssize_t (*store)(struct btrfs_root *, const char *, size_t); +}; + +#define ROOT_ATTR(name, mode, show, store) \ +static struct btrfs_root_attr btrfs_root_attr_##name = __ATTR(name, mode, show, store) + +ROOT_ATTR(blocks_used, 0444, root_blocks_used_show, NULL); +ROOT_ATTR(block_limit, 0644, root_block_limit_show, NULL); + +static struct attribute *btrfs_root_attrs[] = { + &btrfs_root_attr_blocks_used.attr, + &btrfs_root_attr_block_limit.attr, + NULL, +}; + +/* this is for super attrs (actual full fs) */ +struct btrfs_super_attr { + struct attribute attr; + ssize_t (*show)(struct btrfs_fs_info *, char *); + ssize_t (*store)(struct btrfs_fs_info *, const char *, size_t); +}; + +#define SUPER_ATTR(name, mode, show, store) \ +static struct btrfs_super_attr btrfs_super_attr_##name = __ATTR(name, mode, show, store) + +SUPER_ATTR(blocks_used, 0444, super_blocks_used_show, NULL); +SUPER_ATTR(total_blocks, 0444, super_total_blocks_show, NULL); +SUPER_ATTR(blocksize, 0444, super_blocksize_show, NULL); + +static struct attribute *btrfs_super_attrs[] = { + &btrfs_super_attr_blocks_used.attr, + &btrfs_super_attr_total_blocks.attr, + &btrfs_super_attr_blocksize.attr, + NULL, +}; + +static ssize_t btrfs_super_attr_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info, + super_kobj); + struct btrfs_super_attr *a = container_of(attr, + struct btrfs_super_attr, + attr); + + return a->show ? a->show(fs, buf) : 0; +} + +static ssize_t btrfs_super_attr_store(struct kobject *kobj, + struct attribute *attr, + const char *buf, size_t len) +{ + struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info, + super_kobj); + struct btrfs_super_attr *a = container_of(attr, + struct btrfs_super_attr, + attr); + + return a->store ? a->store(fs, buf, len) : 0; +} + +static ssize_t btrfs_root_attr_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct btrfs_root *root = container_of(kobj, struct btrfs_root, + root_kobj); + struct btrfs_root_attr *a = container_of(attr, + struct btrfs_root_attr, + attr); + + return a->show ? a->show(root, buf) : 0; +} + +static ssize_t btrfs_root_attr_store(struct kobject *kobj, + struct attribute *attr, + const char *buf, size_t len) +{ + struct btrfs_root *root = container_of(kobj, struct btrfs_root, + root_kobj); + struct btrfs_root_attr *a = container_of(attr, + struct btrfs_root_attr, + attr); + return a->store ? a->store(root, buf, len) : 0; +} + +static void btrfs_super_release(struct kobject *kobj) +{ + struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info, + super_kobj); + complete(&fs->kobj_unregister); +} + +static void btrfs_root_release(struct kobject *kobj) +{ + struct btrfs_root *root = container_of(kobj, struct btrfs_root, + root_kobj); + complete(&root->kobj_unregister); +} + +static struct sysfs_ops btrfs_super_attr_ops = { + .show = btrfs_super_attr_show, + .store = btrfs_super_attr_store, +}; + +static struct sysfs_ops btrfs_root_attr_ops = { + .show = btrfs_root_attr_show, + .store = btrfs_root_attr_store, +}; + +static struct kobj_type btrfs_root_ktype = { + .default_attrs = btrfs_root_attrs, + .sysfs_ops = &btrfs_root_attr_ops, + .release = btrfs_root_release, +}; + +static struct kobj_type btrfs_super_ktype = { + .default_attrs = btrfs_super_attrs, + .sysfs_ops = &btrfs_super_attr_ops, + .release = btrfs_super_release, +}; + +static struct kset btrfs_kset = { + .kobj = {.name = "btrfs"}, +}; + +int btrfs_sysfs_add_super(struct btrfs_fs_info *fs) +{ + int error; + + fs->super_kobj.kset = &btrfs_kset; + fs->super_kobj.ktype = &btrfs_super_ktype; + + error = kobject_set_name(&fs->super_kobj, "%s", + fs->sb->s_id); + if (error) + goto fail; + + error = kobject_register(&fs->super_kobj); + if (error) + goto fail; + + return 0; + +fail: + printk(KERN_ERR "btrfs: sysfs creation for super failed\n"); + return error; +} + +int btrfs_sysfs_add_root(struct btrfs_root *root) +{ + int error; + + root->root_kobj.ktype = &btrfs_root_ktype; + root->root_kobj.parent = &root->fs_info->super_kobj; + + error = kobject_set_name(&root->root_kobj, "%s", root->name); + if (error) { + goto fail; + } + + error = kobject_register(&root->root_kobj); + if (error) + goto fail; + + return 0; + +fail: + printk(KERN_ERR "btrfs: sysfs creation for root failed\n"); + return error; +} + +void btrfs_sysfs_del_root(struct btrfs_root *root) +{ + kobject_unregister(&root->root_kobj); + wait_for_completion(&root->kobj_unregister); +} + +void btrfs_sysfs_del_super(struct btrfs_fs_info *fs) +{ + kobject_unregister(&fs->super_kobj); + wait_for_completion(&fs->kobj_unregister); +} + +int btrfs_init_sysfs() +{ + kobj_set_kset_s(&btrfs_kset, fs_subsys); + return kset_register(&btrfs_kset); +} + +void btrfs_exit_sysfs() +{ + kset_unregister(&btrfs_kset); +} diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 18abea80279..2b15daa3a9f 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -236,6 +236,7 @@ static int wait_for_commit(struct btrfs_root *root, struct dirty_root { struct list_head list; struct btrfs_root *root; + struct btrfs_root *latest_root; }; int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list) @@ -278,6 +279,15 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, btrfs_root_blocknr(&root->root_item)); brelse(root->commit_root); root->commit_root = NULL; + + /* make sure to update the root on disk + * so we get any updates to the block used + * counts + */ + err = btrfs_update_root(trans, + root->fs_info->tree_root, + &root->root_key, + &root->root_item); continue; } dirty = kmalloc(sizeof(*dirty), GFP_NOFS); @@ -291,6 +301,7 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, memcpy(dirty->root, root, sizeof(*root)); dirty->root->node = root->commit_root; + dirty->latest_root = root; root->commit_root = NULL; root->root_key.offset = root->fs_info->generation; @@ -384,20 +395,29 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, { struct dirty_root *dirty; struct btrfs_trans_handle *trans; + u64 num_blocks; + u64 blocks_used; int ret = 0; int err; while(!list_empty(list)) { + struct btrfs_root *root; + mutex_lock(&tree_root->fs_info->fs_mutex); dirty = list_entry(list->next, struct dirty_root, list); list_del_init(&dirty->list); + num_blocks = btrfs_root_blocks_used(&dirty->root->root_item); + root = dirty->latest_root; + while(1) { trans = btrfs_start_transaction(tree_root, 1); + ret = btrfs_drop_snapshot(trans, dirty->root); if (ret != -EAGAIN) { break; } + err = btrfs_update_root(trans, tree_root, &dirty->root->root_key, @@ -414,9 +434,19 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, mutex_lock(&tree_root->fs_info->fs_mutex); } BUG_ON(ret); + + num_blocks -= btrfs_root_blocks_used(&dirty->root->root_item); + blocks_used = btrfs_root_blocks_used(&root->root_item); + if (num_blocks) { + record_root_in_trans(root); + btrfs_set_root_blocks_used(&root->root_item, + blocks_used - num_blocks); + } ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); - if (ret) + if (ret) { + BUG(); break; + } ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); @@ -534,10 +564,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, wake_up(&cur_trans->commit_wait); put_transaction(cur_trans); put_transaction(cur_trans); + if (root->fs_info->closing) list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); else list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); + mutex_unlock(&root->fs_info->trans_mutex); kmem_cache_free(btrfs_trans_handle_cachep, trans); -- cgit v1.2.3-70-g09d2 From 1b4ab1bb4bbd3045cd041259c30b826654599e96 Mon Sep 17 00:00:00 2001 From: Yan Date: Wed, 29 Aug 2007 09:11:44 -0400 Subject: Btrfs: Fix mknod to properly send rdev info back to disk Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2e3918e6049..483eea69b0a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1095,6 +1095,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, else { inode->i_op = &btrfs_special_inode_operations; init_special_inode(inode, inode->i_mode, rdev); + btrfs_update_inode(trans, root, inode); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); -- cgit v1.2.3-70-g09d2 From 07157aacb1ecd394a5494910b5f80d7e7d407d37 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 30 Aug 2007 08:50:51 -0400 Subject: Btrfs: Add file data csums back in via hooks in the extent map code Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 94 ++++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/extent_map.h | 15 +++++++- fs/btrfs/inode.c | 98 +++++++++++++++++++++++++++++++++++++++++++++++++-- 3 files changed, 193 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index a9c7419615b..6053f9db073 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -57,7 +57,7 @@ void extent_map_tree_init(struct extent_map_tree *tree, { tree->map.rb_node = NULL; tree->state.rb_node = NULL; - tree->fill_delalloc = NULL; + tree->ops = NULL; rwlock_init(&tree->lock); tree->mapping = mapping; } @@ -93,6 +93,7 @@ struct extent_state *alloc_extent_state(gfp_t mask) return state; state->state = 0; state->in_tree = 0; + state->private = 0; atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); return state; @@ -1034,6 +1035,61 @@ int unlock_range(struct extent_map_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(unlock_range); +int set_state_private(struct extent_map_tree *tree, u64 start, u64 private) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 0; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + state->private = private; +out: + write_unlock_irq(&tree->lock); + return ret; + +} + +int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 0; + + read_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + *private = state->private; +out: + read_unlock_irq(&tree->lock); + return ret; +} + /* * searches a range in the state tree for a given mask. * If 'filled' == 1, this returns 1 only if ever extent in the tree @@ -1185,12 +1241,13 @@ static int end_bio_extent_writepage(struct bio *bio, static int end_bio_extent_readpage(struct bio *bio, unsigned int bytes_done, int err) { - const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct extent_map_tree *tree = bio->bi_private; u64 start; u64 end; int whole_page; + int ret; if (bio->bi_size) return 1; @@ -1208,6 +1265,11 @@ static int end_bio_extent_readpage(struct bio *bio, if (--bvec >= bio->bi_io_vec) prefetchw(&bvec->bv_page->flags); + if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { + ret = tree->ops->readpage_end_io_hook(page, start, end); + if (ret) + uptodate = 0; + } if (uptodate) { set_extent_uptodate(tree, start, end, GFP_ATOMIC); if (whole_page) @@ -1388,9 +1450,16 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, continue; } - ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, bdev, - end_bio_extent_readpage); + ret = 0; + if (tree->ops && tree->ops->readpage_io_hook) { + ret = tree->ops->readpage_io_hook(page, cur, + cur + iosize - 1); + } + if (!ret) { + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, + bdev, end_bio_extent_readpage); + } if (ret) SetPageError(page); cur = cur + iosize; @@ -1462,7 +1531,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, &delalloc_end, 128 * 1024 * 1024); if (nr_delalloc) { - tree->fill_delalloc(inode, start, delalloc_end); + tree->ops->fill_delalloc(inode, start, delalloc_end); if (delalloc_end >= page_end + 1) { clear_extent_bit(tree, page_end + 1, delalloc_end, EXTENT_LOCKED | EXTENT_DELALLOC, @@ -1528,12 +1597,17 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, continue; } clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); - set_range_writeback(tree, cur, cur + iosize - 1); - ret = submit_extent_page(WRITE, tree, page, - sector, iosize, page_offset, bdev, - end_bio_extent_writepage); + ret = tree->ops->writepage_io_hook(page, cur, cur + iosize - 1); if (ret) SetPageError(page); + else { + set_range_writeback(tree, cur, cur + iosize - 1); + ret = submit_extent_page(WRITE, tree, page, sector, + iosize, page_offset, bdev, + end_bio_extent_writepage); + if (ret) + SetPageError(page); + } cur = cur + iosize; page_offset += iosize; nr++; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index e91a2e9619e..0eae6c420ac 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -6,12 +6,19 @@ #define EXTENT_MAP_INLINE (u64)-2 #define EXTENT_MAP_DELALLOC (u64)-1 +struct extent_map_ops { + int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); + int (*writepage_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end); +}; + struct extent_map_tree { struct rb_root map; struct rb_root state; struct address_space *mapping; rwlock_t lock; - int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); + struct extent_map_ops *ops; }; /* note, this must start with the same fields as fs/extent_map.c:tree_entry */ @@ -36,6 +43,10 @@ struct extent_state { wait_queue_head_t wq; atomic_t refs; unsigned long state; + + /* for use by the FS */ + u64 private; + struct list_head list; }; @@ -89,4 +100,6 @@ int extent_commit_write(struct extent_map_tree *tree, struct inode *inode, struct page *page, unsigned from, unsigned to); int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); +int set_state_private(struct extent_map_tree *tree, u64 start, u64 private); +int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private); #endif diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 483eea69b0a..3ee6b2fadf5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -52,6 +52,7 @@ static struct inode_operations btrfs_file_inode_operations; static struct address_space_operations btrfs_aops; static struct address_space_operations btrfs_symlink_aops; static struct file_operations btrfs_dir_file_operations; +static struct extent_map_ops btrfs_extent_map_ops; static struct kmem_cache *btrfs_inode_cachep; struct kmem_cache *btrfs_trans_handle_cachep; @@ -103,6 +104,90 @@ out: return ret; } +int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) +{ + struct inode *inode = page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + char *kaddr; + int ret; + u64 page_start = page->index << PAGE_CACHE_SHIFT; + size_t offset = start - page_start; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + kaddr = kmap(page); + btrfs_csum_file_block(trans, root, inode->i_ino, + start, kaddr + offset, end - start + 1); + kunmap(page); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) +{ + int ret = 0; + struct inode *inode = page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct btrfs_csum_item *item; + struct btrfs_path *path = NULL; + u64 private; + + mutex_lock(&root->fs_info->fs_mutex); + path = btrfs_alloc_path(); + item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); + if (IS_ERR(item)) { + ret = PTR_ERR(item); + /* a csum that isn't present is a preallocated region. */ + if (ret == -ENOENT || ret == -EFBIG) + ret = 0; + private = 0; + goto out; + } + memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE); + set_state_private(em_tree, start, private); +out: + if (path) + btrfs_free_path(path); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) +{ + char csum[BTRFS_CRC32_SIZE]; + size_t offset = start - (page->index << PAGE_CACHE_SHIFT); + struct inode *inode = page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + char *kaddr; + u64 private; + int ret; + + ret = get_state_private(em_tree, start, &private); + kaddr = kmap_atomic(page, KM_IRQ0); + if (ret) { + goto zeroit; + } + ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum); + BUG_ON(ret); + if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) { + goto zeroit; + } + kunmap_atomic(kaddr, KM_IRQ0); + return 0; + +zeroit: + printk("btrfs csum failed ino %lu off %llu\n", + page->mapping->host->i_ino, (unsigned long long)start); + memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); + kunmap_atomic(kaddr, KM_IRQ0); + return 0; +} void btrfs_read_locked_inode(struct inode *inode) { @@ -155,7 +240,7 @@ void btrfs_read_locked_inode(struct inode *inode) switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; - BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; + BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; break; @@ -1148,7 +1233,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, inode->i_mapping, GFP_NOFS); - BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; + BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2286,7 +2371,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, inode->i_mapping, GFP_NOFS); - BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range; + BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2362,6 +2447,13 @@ static struct file_operations btrfs_dir_file_operations = { #endif }; +static struct extent_map_ops btrfs_extent_map_ops = { + .fill_delalloc = run_delalloc_range, + .writepage_io_hook = btrfs_writepage_io_hook, + .readpage_io_hook = btrfs_readpage_io_hook, + .readpage_end_io_hook = btrfs_readpage_end_io_hook, +}; + static struct address_space_operations btrfs_aops = { .readpage = btrfs_readpage, .writepage = btrfs_writepage, -- cgit v1.2.3-70-g09d2 From 2bf5a725a3b82efeaf7b292c085e69a9388a89ea Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 30 Aug 2007 11:54:02 -0400 Subject: Btrfs: fsx delalloc fixes Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 9 ++- fs/btrfs/file.c | 26 +++++--- fs/btrfs/inode.c | 177 ++++---------------------------------------------- 3 files changed, 38 insertions(+), 174 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 6053f9db073..8ad6f8efc5a 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -77,6 +77,8 @@ EXPORT_SYMBOL(alloc_extent_map); void free_extent_map(struct extent_map *em) { + if (!em) + return; if (atomic_dec_and_test(&em->refs)) { WARN_ON(em->in_tree); kmem_cache_free(extent_map_cache, em); @@ -102,6 +104,8 @@ EXPORT_SYMBOL(alloc_extent_state); void free_extent_state(struct extent_state *state) { + if (!state) + return; if (atomic_dec_and_test(&state->refs)) { WARN_ON(state->in_tree); kmem_cache_free(extent_state_cache, state); @@ -1395,8 +1399,8 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, if (!PagePrivate(page)) { SetPagePrivate(page); - set_page_private(page, 1); WARN_ON(!page->mapping->a_ops->invalidatepage); + set_page_private(page, 1); page_cache_get(page); } @@ -1638,7 +1642,8 @@ int extent_invalidatepage(struct extent_map_tree *tree, lock_extent(tree, start, end, GFP_NOFS); wait_on_extent_writeback(tree, start, end); - clear_extent_bit(tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY, + clear_extent_bit(tree, start, end, + EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, 1, 1, GFP_NOFS); return 0; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index d3d39e4a279..07b121d4bd9 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -186,8 +186,16 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 mask = root->blocksize - 1; last_pos_in_file = (isize + mask) & ~mask; hole_size = (start_pos - last_pos_in_file + mask) & ~mask; - hole_size >>= inode->i_blkbits; + if (last_pos_in_file < start_pos) { + err = btrfs_drop_extents(trans, root, inode, + last_pos_in_file, + last_pos_in_file + hole_size, + &hint_block); + if (err) + goto failed; + + hole_size >>= inode->i_blkbits; err = btrfs_insert_file_extent(trans, root, inode->i_ino, last_pos_in_file, @@ -217,13 +225,11 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, struct page *p = pages[0]; /* step one, delete the existing extents in this range */ /* FIXME blocksize != pagesize */ - if (start_pos < inode->i_size) { - err = btrfs_drop_extents(trans, root, inode, start_pos, - (pos + write_bytes + root->blocksize -1) & - ~((u64)root->blocksize - 1), &hint_block); - if (err) - goto failed; - } + err = btrfs_drop_extents(trans, root, inode, start_pos, + (pos + write_bytes + root->blocksize -1) & + ~((u64)root->blocksize - 1), &hint_block); + if (err) + goto failed; err = insert_inline_extent(trans, root, inode, start_pos, end_pos - start_pos, p, 0); @@ -400,6 +406,8 @@ next_slot: keep = 1; WARN_ON(start & (root->blocksize - 1)); if (found_extent) { + btrfs_drop_extent_cache(inode, key.offset, + start - 1 ); new_num = (start - key.offset) >> inode->i_blkbits; old_num = btrfs_file_extent_num_blocks(extent); @@ -464,7 +472,7 @@ next_slot: if (ret) { btrfs_print_leaf(root, btrfs_buffer_leaf(path->nodes[0])); - printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end); + printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end, keep); } BUG_ON(ret); extent = btrfs_item_ptr( diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3ee6b2fadf5..64710fa77d0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -721,25 +721,35 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + u64 mask = root->blocksize - 1; u64 pos = (inode->i_size + mask) & ~mask; + u64 block_end = attr->ia_size | mask; u64 hole_size; + u64 alloc_hint; if (attr->ia_size <= pos) goto out; btrfs_truncate_page(inode->i_mapping, inode->i_size); + lock_extent(em_tree, pos, block_end, GFP_NOFS); hole_size = (attr->ia_size - pos + mask) & ~mask; - hole_size >>= inode->i_blkbits; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); + err = btrfs_drop_extents(trans, root, inode, + pos, pos + hole_size, &alloc_hint); + + hole_size >>= inode->i_blkbits; + err = btrfs_insert_file_extent(trans, root, inode->i_ino, pos, 0, 0, hole_size); btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); + unlock_extent(em_tree, pos, block_end, GFP_NOFS); if (err) return err; } @@ -1529,13 +1539,13 @@ insert: ret = add_extent_mapping(em_tree, em); if (ret == -EEXIST) { free_extent_map(em); + em = NULL; failed_insert++; if (failed_insert > 5) { printk("failing to insert %Lu %Lu\n", start, end); err = -EIO; goto out; } - em = NULL; goto again; } err = 0; @@ -1555,167 +1565,6 @@ out: return em; } - -/* - * FIBMAP and others want to pass in a fake buffer head. They need to - * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy - * any packed file data into the fake bh - */ -#define BTRFS_GET_BLOCK_NO_CREATE 0 -#define BTRFS_GET_BLOCK_CREATE 1 -#define BTRFS_GET_BLOCK_NO_DIRECT 2 - -/* - * FIXME create==1 doe not work. - */ -static int btrfs_get_block_lock(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create) -{ - int ret; - int err = 0; - u64 blocknr; - u64 extent_start = 0; - u64 extent_end = 0; - u64 objectid = inode->i_ino; - u32 found_type; - u64 alloc_hint = 0; - struct btrfs_path *path; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_file_extent_item *item; - struct btrfs_leaf *leaf; - struct btrfs_disk_key *found_key; - struct btrfs_trans_handle *trans = NULL; - - path = btrfs_alloc_path(); - BUG_ON(!path); - if (create & BTRFS_GET_BLOCK_CREATE) { - /* - * danger!, this only works if the page is properly up - * to date somehow - */ - trans = btrfs_start_transaction(root, 1); - if (!trans) { - err = -ENOMEM; - goto out; - } - ret = btrfs_drop_extents(trans, root, inode, - iblock << inode->i_blkbits, - (iblock + 1) << inode->i_blkbits, - &alloc_hint); - BUG_ON(ret); - } - - ret = btrfs_lookup_file_extent(NULL, root, path, - objectid, - iblock << inode->i_blkbits, 0); - if (ret < 0) { - err = ret; - goto out; - } - - if (ret != 0) { - if (path->slots[0] == 0) { - btrfs_release_path(root, path); - goto not_found; - } - path->slots[0]--; - } - - item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], - struct btrfs_file_extent_item); - leaf = btrfs_buffer_leaf(path->nodes[0]); - blocknr = btrfs_file_extent_disk_blocknr(item); - blocknr += btrfs_file_extent_offset(item); - - /* are we inside the extent that was found? */ - found_key = &leaf->items[path->slots[0]].key; - found_type = btrfs_disk_key_type(found_key); - if (btrfs_disk_key_objectid(found_key) != objectid || - found_type != BTRFS_EXTENT_DATA_KEY) { - extent_end = 0; - extent_start = 0; - goto not_found; - } - found_type = btrfs_file_extent_type(item); - extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key); - if (found_type == BTRFS_FILE_EXTENT_REG) { - extent_start = extent_start >> inode->i_blkbits; - extent_end = extent_start + btrfs_file_extent_num_blocks(item); - err = 0; - if (btrfs_file_extent_disk_blocknr(item) == 0) - goto out; - if (iblock >= extent_start && iblock < extent_end) { - btrfs_map_bh_to_logical(root, result, blocknr + - iblock - extent_start); - goto out; - } - } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - char *ptr; - char *map; - u32 size; - - if (create & BTRFS_GET_BLOCK_NO_DIRECT) { - err = -EINVAL; - goto out; - } - size = btrfs_file_extent_inline_len(leaf->items + - path->slots[0]); - extent_end = (extent_start + size) >> inode->i_blkbits; - extent_start >>= inode->i_blkbits; - if (iblock < extent_start || iblock > extent_end) { - goto not_found; - } - ptr = btrfs_file_extent_inline_start(item); - map = kmap(result->b_page); - memcpy(map, ptr, size); - memset(map + size, 0, PAGE_CACHE_SIZE - size); - flush_dcache_page(result->b_page); - kunmap(result->b_page); - set_buffer_uptodate(result); - SetPageChecked(result->b_page); - btrfs_map_bh_to_logical(root, result, 0); - } -not_found: - if (create & BTRFS_GET_BLOCK_CREATE) { - struct btrfs_key ins; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, - 1, 0, alloc_hint, (u64)-1, - &ins, 1); - if (ret) { - err = ret; - goto out; - } - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - iblock << inode->i_blkbits, - ins.objectid, ins.offset, - ins.offset); - if (ret) { - err = ret; - goto out; - } - btrfs_map_bh_to_logical(root, result, ins.objectid); - } -out: - if (trans) { - ret = btrfs_end_transaction(trans, root); - if (!err) - err = ret; - } - btrfs_free_path(path); - return err; -} - -int btrfs_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create) -{ - int err; - struct btrfs_root *root = BTRFS_I(inode)->root; - mutex_lock(&root->fs_info->fs_mutex); - err = btrfs_get_block_lock(inode, iblock, result, create); - mutex_unlock(&root->fs_info->fs_mutex); - return err; -} - static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock, struct buffer_head *result, int create) { @@ -2469,6 +2318,8 @@ static struct address_space_operations btrfs_aops = { static struct address_space_operations btrfs_symlink_aops = { .readpage = btrfs_readpage, .writepage = btrfs_writepage, + .invalidatepage = btrfs_invalidatepage, + .releasepage = btrfs_releasepage, }; static struct inode_operations btrfs_file_inode_operations = { -- cgit v1.2.3-70-g09d2 From 86479a04eef8f304a13aeb8b64bcc8e506a68268 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Sep 2007 19:58:16 -0400 Subject: Add support for defragging files via btrfsctl -d. Avoid OOM on extent tree defrag. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 4 +++ fs/btrfs/disk-io.h | 2 ++ fs/btrfs/extent_map.c | 16 +++++---- fs/btrfs/inode.c | 93 +++++++++++++++++++++++++++++++++++++++++++------- fs/btrfs/tree-defrag.c | 3 ++ 5 files changed, 100 insertions(+), 18 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 89e282c35ce..b41f48ade41 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -217,6 +217,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, root->fs_info->generation); WARN_ON(1); } + if (buffer_defrag_done(parent)) + return 0; + parent_node = btrfs_buffer_node(parent); parent_nritems = btrfs_header_nritems(&parent_node->header); parent_level = btrfs_header_level(&parent_node->header); @@ -274,6 +277,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, *last_ret = search_start; if (parent_level == 1) clear_buffer_defrag(tmp_bh); + set_buffer_defrag_done(tmp_bh); brelse(tmp_bh); } return err; diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index da6bb72750f..714fa877438 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -26,9 +26,11 @@ enum btrfs_bh_state_bits { BH_Checked = BH_PrivateStart, BH_Defrag, + BH_DefragDone, }; BUFFER_FNS(Checked, checked); BUFFER_FNS(Defrag, defrag); +BUFFER_FNS(DefragDone, defrag_done); static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh) { diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 8ad6f8efc5a..33f7a18dddf 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -10,6 +10,12 @@ #include #include "extent_map.h" +/* temporary define until extent_map moves out of btrfs */ +struct kmem_cache *btrfs_cache_create(const char *name, size_t size, + unsigned long extra_flags, + void (*ctor)(void *, struct kmem_cache *, + unsigned long)); + static struct kmem_cache *extent_map_cache; static struct kmem_cache *extent_state_cache; @@ -32,14 +38,12 @@ struct tree_entry { void __init extent_map_init(void) { - extent_map_cache = kmem_cache_create("extent_map", - sizeof(struct extent_map), 0, - SLAB_RECLAIM_ACCOUNT | + extent_map_cache = btrfs_cache_create("extent_map", + sizeof(struct extent_map), SLAB_DESTROY_BY_RCU, NULL); - extent_state_cache = kmem_cache_create("extent_state", - sizeof(struct extent_state), 0, - SLAB_RECLAIM_ACCOUNT | + extent_state_cache = btrfs_cache_create("extent_state", + sizeof(struct extent_state), SLAB_DESTROY_BY_RCU, NULL); } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 64710fa77d0..6b3e4404dc6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1904,6 +1904,70 @@ fail: return ret; } +static unsigned long force_ra(struct address_space *mapping, + struct file_ra_state *ra, struct file *file, + pgoff_t offset, pgoff_t last_index) +{ + pgoff_t req_size; + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) + req_size = last_index - offset + 1; + offset = page_cache_readahead(mapping, ra, file, offset, req_size); + return offset; +#else + req_size = min(last_index - offset + 1, (pgoff_t)128); + page_cache_sync_readahead(mapping, ra, file, offset, req_size); + return offset + req_size; +#endif +} + +int btrfs_defrag_file(struct file *file) { + struct inode *inode = file->f_path.dentry->d_inode; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct page *page; + unsigned long last_index; + unsigned long ra_index = 0; + u64 page_start; + u64 page_end; + unsigned long i; + + mutex_lock(&inode->i_mutex); + last_index = inode->i_size >> PAGE_CACHE_SHIFT; + for (i = 0; i <= last_index; i++) { + if (i == ra_index) { + ra_index = force_ra(inode->i_mapping, &file->f_ra, + file, ra_index, last_index); + } + page = grab_cache_page(inode->i_mapping, i); + if (!page) + goto out_unlock; + if (!PageUptodate(page)) { + btrfs_readpage(NULL, page); + lock_page(page); + if (!PageUptodate(page)) { + unlock_page(page); + page_cache_release(page); + goto out_unlock; + } + } + page_start = page->index << PAGE_CACHE_SHIFT; + page_end = page_start + PAGE_CACHE_SIZE - 1; + + lock_extent(em_tree, page_start, page_end, GFP_NOFS); + set_extent_delalloc(em_tree, page_start, + page_end, GFP_NOFS); + unlock_extent(em_tree, page_start, page_end, GFP_NOFS); + set_page_dirty(page); + unlock_page(page); + page_cache_release(page); + balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); + } + +out_unlock: + mutex_unlock(&inode->i_mutex); + return 0; +} + int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) { @@ -1948,10 +2012,14 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int break; case BTRFS_IOC_DEFRAG: - mutex_lock(&root->fs_info->fs_mutex); - btrfs_defrag_root(root, 0); - btrfs_defrag_root(root->fs_info->extent_root, 0); - mutex_unlock(&root->fs_info->fs_mutex); + if (S_ISDIR(inode->i_mode)) { + mutex_lock(&root->fs_info->fs_mutex); + btrfs_defrag_root(root, 0); + btrfs_defrag_root(root->fs_info->extent_root, 0); + mutex_unlock(&root->fs_info->fs_mutex); + } else if (S_ISREG(inode->i_mode)) { + btrfs_defrag_file(filp); + } ret = 0; break; default: @@ -2018,7 +2086,7 @@ void btrfs_destroy_cachep(void) kmem_cache_destroy(btrfs_path_cachep); } -static struct kmem_cache *cache_create(const char *name, size_t size, +struct kmem_cache *btrfs_cache_create(const char *name, size_t size, unsigned long extra_flags, void (*ctor)(void *, struct kmem_cache *, unsigned long)) @@ -2033,27 +2101,28 @@ static struct kmem_cache *cache_create(const char *name, size_t size, int btrfs_init_cachep(void) { - btrfs_inode_cachep = cache_create("btrfs_inode_cache", + btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache", sizeof(struct btrfs_inode), 0, init_once); if (!btrfs_inode_cachep) goto fail; - btrfs_trans_handle_cachep = cache_create("btrfs_trans_handle_cache", - sizeof(struct btrfs_trans_handle), - 0, NULL); + btrfs_trans_handle_cachep = + btrfs_cache_create("btrfs_trans_handle_cache", + sizeof(struct btrfs_trans_handle), + 0, NULL); if (!btrfs_trans_handle_cachep) goto fail; - btrfs_transaction_cachep = cache_create("btrfs_transaction_cache", + btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache", sizeof(struct btrfs_transaction), 0, NULL); if (!btrfs_transaction_cachep) goto fail; - btrfs_path_cachep = cache_create("btrfs_path_cache", + btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache", sizeof(struct btrfs_transaction), 0, NULL); if (!btrfs_path_cachep) goto fail; - btrfs_bit_radix_cachep = cache_create("btrfs_radix", 256, + btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256, SLAB_DESTROY_BY_RCU, NULL); if (!btrfs_bit_radix_cachep) goto fail; diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 35fd20d2464..420597127ed 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -113,6 +113,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, } WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); + clear_buffer_defrag(path->nodes[*level]); + clear_buffer_defrag_done(path->nodes[*level]); btrfs_block_release(root, path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; @@ -143,6 +145,7 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans, return 0; } else { clear_buffer_defrag(path->nodes[*level]); + clear_buffer_defrag_done(path->nodes[*level]); btrfs_block_release(root, path->nodes[*level]); path->nodes[*level] = NULL; *level = i + 1; -- cgit v1.2.3-70-g09d2 From 011410bd859a481a335d7db1fb559542c5663fd5 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Sep 2007 19:58:36 -0400 Subject: Btrfs: Add more synchronization before creating a snapshot File data checksums are only done during writepage, so we have to make sure all pages are written when the snapshot is taken. This also adds some locking so that new writes don't race in and add new dirty pages. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 1 + fs/btrfs/file.c | 2 ++ fs/btrfs/inode.c | 10 +++++++++- 4 files changed, 13 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index cd75c906048..03e1c4ad341 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -333,6 +333,7 @@ struct btrfs_root { struct inode *inode; struct kobject root_kobj; struct completion kobj_unregister; + struct rw_semaphore snap_sem; u64 objectid; u64 last_trans; u32 blocksize; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index c25ef0a68f1..6c953a0e0aa 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -300,6 +300,7 @@ static int __setup_root(int blocksize, memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); memset(&root->root_kobj, 0, sizeof(root->root_kobj)); init_completion(&root->kobj_unregister); + init_rwsem(&root->snap_sem); root->defrag_running = 0; root->defrag_level = 0; root->root_key.objectid = objectid; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index acef8cb7fb7..77f1950b0aa 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -160,6 +160,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> inode->i_blkbits; + down_read(&BTRFS_I(inode)->root->snap_sem); end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1; lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); mutex_lock(&root->fs_info->fs_mutex); @@ -250,6 +251,7 @@ out_unlock: mutex_unlock(&root->fs_info->fs_mutex); unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); free_extent_map(em); + up_read(&BTRFS_I(inode)->root->snap_sem); return err; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6b3e4404dc6..1ace6d11e09 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -686,6 +686,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) if ((offset & (blocksize - 1)) == 0) goto out; + down_read(&BTRFS_I(inode)->root->snap_sem); ret = -ENOMEM; page = grab_cache_page(mapping, index); if (!page) @@ -704,6 +705,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) unlock_page(page); page_cache_release(page); + up_read(&BTRFS_I(inode)->root->snap_sem); out: return ret; } @@ -1668,6 +1670,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) int ret = -EINVAL; u64 page_start; + down_read(&BTRFS_I(inode)->root->snap_sem); lock_page(page); wait_on_page_writeback(page); size = i_size_read(inode); @@ -1688,6 +1691,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) ret = btrfs_cow_one_page(inode, page, end); out_unlock: + up_read(&BTRFS_I(inode)->root->snap_sem); unlock_page(page); return ret; } @@ -1851,6 +1855,10 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (!root->ref_cows) return -EINVAL; + down_write(&root->snap_sem); + freeze_bdev(root->fs_info->sb->s_bdev); + thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb); + mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); @@ -1894,12 +1902,12 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) ret = btrfs_inc_root_ref(trans, root); if (ret) goto fail; - fail: err = btrfs_commit_transaction(trans, root); if (err && !ret) ret = err; mutex_unlock(&root->fs_info->fs_mutex); + up_write(&root->snap_sem); btrfs_btree_balance_dirty(root); return ret; } -- cgit v1.2.3-70-g09d2 From d396c6f554d182b7bce4e2bb2acb47f80430c05f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 10 Sep 2007 20:02:30 -0400 Subject: Btrfs: [PATCH] extent_map: provide generic bmap generic_bmap is completely trivial, while the extent to bh mapping in btrfs is rather complex. So provide a extent_bmap instead that takes a get_extent callback and can be used by filesystem using the extent_map code. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 19 +++++++++++++++++++ fs/btrfs/extent_map.h | 2 ++ fs/btrfs/inode.c | 23 ++--------------------- 3 files changed, 23 insertions(+), 21 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 3021564ac62..ff2b1dd25bf 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1816,3 +1816,22 @@ int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page) } EXPORT_SYMBOL(try_release_extent_mapping); +sector_t extent_bmap(struct address_space *mapping, sector_t iblock, + get_extent_t *get_extent) +{ + struct inode *inode = mapping->host; + u64 start = iblock << inode->i_blkbits; + u64 end = start + (1 << inode->i_blkbits) - 1; + struct extent_map *em; + + em = get_extent(inode, NULL, 0, start, end, 0); + if (!em || IS_ERR(em)) + return 0; + + // XXX(hch): block 0 is valid in some cases, e.g. XFS RT device + if (em->block_start == EXTENT_MAP_INLINE || + em->block_start == 0) + return 0; + + return (em->block_start + start - em->start) >> inode->i_blkbits; +} diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 0eae6c420ac..da41aca74e0 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -99,6 +99,8 @@ int extent_prepare_write(struct extent_map_tree *tree, int extent_commit_write(struct extent_map_tree *tree, struct inode *inode, struct page *page, unsigned from, unsigned to); +sector_t extent_bmap(struct address_space *mapping, sector_t iblock, + get_extent_t *get_extent); int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); int set_state_private(struct extent_map_tree *tree, u64 start, u64 private); int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1ace6d11e09..39c1fe178f4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1567,28 +1567,9 @@ out: return em; } -static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create) +static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) { - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 start = iblock << inode->i_blkbits; - u64 end = start + root->blocksize -1; - struct extent_map *em; - - em = btrfs_get_extent(inode, NULL, 0, start, end, 0); - if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE && - em->block_start != 0) { - u64 offset; - offset = start - em->start; - start = (em->block_start + offset) >> inode->i_blkbits; - btrfs_map_bh_to_logical(root, result, start); - } - return 0; -} - -static sector_t btrfs_bmap(struct address_space *as, sector_t block) -{ - return generic_block_bmap(as, block, btrfs_get_block_bmap); + return extent_bmap(mapping, iblock, btrfs_get_extent); } static int btrfs_prepare_write(struct file *file, struct page *page, -- cgit v1.2.3-70-g09d2 From 93a6925ec10c2d5792ffad35af8366179c9db57c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 14 Sep 2007 09:42:31 -0400 Subject: Btrfs: Fix extra link count dec in rename Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 39c1fe178f4..cde8ddde609 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2214,13 +2214,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry); if (ret) goto out_fail; - if (S_ISDIR(new_inode->i_mode)) - clear_nlink(new_inode); - else - drop_nlink(new_inode); - ret = btrfs_update_inode(trans, root, new_inode); - if (ret) - goto out_fail; } ret = btrfs_add_link(trans, new_dentry, old_inode); if (ret) -- cgit v1.2.3-70-g09d2 From 34287aa3605d52ef5a4b6e5e2cca98667166508b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 14 Sep 2007 10:22:47 -0400 Subject: Btrfs: use unlocked_ioctl No reason to grab the BKL before calling into the btrfs ioctl code. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 5 +---- fs/btrfs/file.c | 4 ++-- fs/btrfs/inode.c | 23 +++++------------------ 3 files changed, 8 insertions(+), 24 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 6d6e94b6380..800a3499cc3 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1219,10 +1219,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb); void btrfs_destroy_inode(struct inode *inode); int btrfs_init_cachep(void); void btrfs_destroy_cachep(void); -int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg); -long btrfs_compat_ioctl(struct file *file, unsigned int cmd, - unsigned long arg); +long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); int btrfs_commit_write(struct file *file, struct page *page, diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index c26ca54ed4d..ca7e5d4474e 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -732,10 +732,10 @@ struct file_operations btrfs_file_operations = { .write = btrfs_file_write, .mmap = btrfs_file_mmap, .open = generic_file_open, - .ioctl = btrfs_ioctl, .fsync = btrfs_sync_file, + .unlocked_ioctl = btrfs_ioctl, #ifdef CONFIG_COMPAT - .compat_ioctl = btrfs_compat_ioctl, + .compat_ioctl = btrfs_ioctl, #endif }; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index cde8ddde609..3901a7fa094 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1957,9 +1957,10 @@ out_unlock: return 0; } -int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int +long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { + struct inode *inode = file->f_path.dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_ioctl_vol_args vol_args; int ret = 0; @@ -2007,7 +2008,7 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int btrfs_defrag_root(root->fs_info->extent_root, 0); mutex_unlock(&root->fs_info->fs_mutex); } else if (S_ISREG(inode->i_mode)) { - btrfs_defrag_file(filp); + btrfs_defrag_file(file); } ret = 0; break; @@ -2017,20 +2018,6 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int return ret; } -#ifdef CONFIG_COMPAT -long btrfs_compat_ioctl(struct file *file, unsigned int cmd, - unsigned long arg) -{ - struct inode *inode = file->f_path.dentry->d_inode; - int ret; - lock_kernel(); - ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg)); - unlock_kernel(); - return ret; - -} -#endif - /* * Called inside transaction, so use GFP_NOFS */ @@ -2341,9 +2328,9 @@ static struct file_operations btrfs_dir_file_operations = { .llseek = generic_file_llseek, .read = generic_read_dir, .readdir = btrfs_readdir, - .ioctl = btrfs_ioctl, + .unlocked_ioctl = btrfs_ioctl, #ifdef CONFIG_COMPAT - .compat_ioctl = btrfs_compat_ioctl, + .compat_ioctl = btrfs_ioctl, #endif }; -- cgit v1.2.3-70-g09d2 From d03581f4345e30db078ea8f81189e8177280bbb8 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 14 Sep 2007 10:22:57 -0400 Subject: split up btrfs_ioctl Add a helper per ioctl function to make the code more readable. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 106 +++++++++++++++++++++++++++++++------------------------ 1 file changed, 59 insertions(+), 47 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3901a7fa094..e24b875327d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1957,65 +1957,77 @@ out_unlock: return 0; } -long btrfs_ioctl(struct file *file, unsigned int - cmd, unsigned long arg) +static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) { - struct inode *inode = file->f_path.dentry->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_ioctl_vol_args vol_args; - int ret = 0; struct btrfs_dir_item *di; - int namelen; struct btrfs_path *path; + int namelen; u64 root_dirid; - switch (cmd) { - case BTRFS_IOC_SNAP_CREATE: - if (copy_from_user(&vol_args, - (struct btrfs_ioctl_vol_args __user *)arg, - sizeof(vol_args))) - return -EFAULT; - namelen = strlen(vol_args.name); - if (namelen > BTRFS_VOL_NAME_MAX) - return -EINVAL; - if (strchr(vol_args.name, '/')) - return -EINVAL; - path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; - root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, + if (copy_from_user(&vol_args, arg, sizeof(vol_args))) + return -EFAULT; + + namelen = strlen(vol_args.name); + if (namelen > BTRFS_VOL_NAME_MAX) + return -EINVAL; + if (strchr(vol_args.name, '/')) + return -EINVAL; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, + mutex_lock(&root->fs_info->fs_mutex); + di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, + path, root_dirid, + vol_args.name, namelen, 0); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_free_path(path); + if (di && !IS_ERR(di)) + return -EEXIST; + if (IS_ERR(di)) + return PTR_ERR(di); + + if (root == root->fs_info->tree_root) + return create_subvol(root, vol_args.name, namelen); + return create_snapshot(root, vol_args.name, namelen); +} + +static int btrfs_ioctl_defrag(struct file *file) +{ + struct inode *inode = file->f_path.dentry->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + + switch (inode->i_mode & S_IFMT) { + case S_IFDIR: mutex_lock(&root->fs_info->fs_mutex); - di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, - path, root_dirid, - vol_args.name, namelen, 0); + btrfs_defrag_root(root, 0); + btrfs_defrag_root(root->fs_info->extent_root, 0); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_free_path(path); - if (di && !IS_ERR(di)) - return -EEXIST; - if (IS_ERR(di)) - return PTR_ERR(di); - - if (root == root->fs_info->tree_root) - ret = create_subvol(root, vol_args.name, namelen); - else - ret = create_snapshot(root, vol_args.name, namelen); break; + case S_IFREG: + btrfs_defrag_file(file); + break; + } + + return 0; +} +long btrfs_ioctl(struct file *file, unsigned int + cmd, unsigned long arg) +{ + struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root; + + switch (cmd) { + case BTRFS_IOC_SNAP_CREATE: + return btrfs_ioctl_snap_create(root, (void __user *)arg); case BTRFS_IOC_DEFRAG: - if (S_ISDIR(inode->i_mode)) { - mutex_lock(&root->fs_info->fs_mutex); - btrfs_defrag_root(root, 0); - btrfs_defrag_root(root->fs_info->extent_root, 0); - mutex_unlock(&root->fs_info->fs_mutex); - } else if (S_ISREG(inode->i_mode)) { - btrfs_defrag_file(file); - } - ret = 0; - break; - default: - return -ENOTTY; + return btrfs_ioctl_defrag(file); } - return ret; + + return -ENOTTY; } /* -- cgit v1.2.3-70-g09d2 From d3c2fdcf7b79079f60ac64e61d886964d4647910 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Sep 2007 10:58:06 -0400 Subject: Btrfs: Use balance_dirty_pages_nr on btree blocks btrfs_btree_balance_dirty is changed to pass the number of pages dirtied for more accurate dirty throttling. This lets the VM make better decisions about when to force some writeback. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 5 +++-- fs/btrfs/disk-io.h | 2 +- fs/btrfs/extent-tree.c | 1 + fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 47 +++++++++++++++++++++++++++++++++++------------ fs/btrfs/transaction.c | 14 ++++++++++---- 6 files changed, 51 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 4296839eea1..ef6d76f2ec2 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -666,7 +666,8 @@ void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf) brelse(buf); } -void btrfs_btree_balance_dirty(struct btrfs_root *root) +void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) { - balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping); + balance_dirty_pages_ratelimited_nr( + root->fs_info->btree_inode->i_mapping, nr); } diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 714fa877438..fc6b520c6e0 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -80,7 +80,7 @@ int btrfs_insert_dev_radix(struct btrfs_root *root, u64 num_blocks); int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, u64 logical); -void btrfs_btree_balance_dirty(struct btrfs_root *root); +void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr); int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); void btrfs_mark_buffer_dirty(struct buffer_head *bh); #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 9151850266b..fe772f9b06c 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1275,6 +1275,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, set_buffer_checked(buf); set_buffer_defrag(buf); set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index); + trans->blocks_used++; return buf; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 26826a54335..698eaea612f 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -655,7 +655,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, num_written += write_bytes; balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, 1); cond_resched(); } mutex_unlock(&inode->i_mutex); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e24b875327d..bd00df33fb3 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -387,15 +387,17 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) struct btrfs_root *root; struct btrfs_trans_handle *trans; int ret; + unsigned long nr; root = BTRFS_I(dir)->root; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); ret = btrfs_unlink_trans(trans, root, dir, dentry); + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return ret; } @@ -412,6 +414,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) int found_type; struct btrfs_leaf *leaf; char *goodnames = ".."; + unsigned long nr; path = btrfs_alloc_path(); BUG_ON(!path); @@ -467,8 +470,9 @@ out: btrfs_release_path(root, path); btrfs_free_path(path); mutex_unlock(&root->fs_info->fs_mutex); + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); if (ret && !err) err = ret; return err; @@ -764,6 +768,7 @@ void btrfs_delete_inode(struct inode *inode) { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(inode)->root; + unsigned long nr; int ret; truncate_inode_pages(&inode->i_data, 0); @@ -780,15 +785,17 @@ void btrfs_delete_inode(struct inode *inode) ret = btrfs_free_inode(trans, root, inode); if (ret) goto no_delete_lock; + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return; no_delete_lock: + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); no_delete: clear_inode(inode); } @@ -1165,6 +1172,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, int err; int drop_inode = 0; u64 objectid; + unsigned long nr; if (!new_valid_dev(rdev)) return -EINVAL; @@ -1198,6 +1206,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); out_unlock: + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); @@ -1205,7 +1214,7 @@ out_unlock: inode_dec_link_count(inode); iput(inode); } - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return err; } @@ -1217,6 +1226,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, struct inode *inode; int err; int drop_inode = 0; + unsigned long nr; u64 objectid; mutex_lock(&root->fs_info->fs_mutex); @@ -1251,6 +1261,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); out_unlock: + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); @@ -1258,7 +1269,7 @@ out_unlock: inode_dec_link_count(inode); iput(inode); } - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return err; } @@ -1268,6 +1279,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; struct inode *inode = old_dentry->d_inode; + unsigned long nr; int err; int drop_inode = 0; @@ -1288,6 +1300,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, if (err) drop_inode = 1; + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); @@ -1295,7 +1308,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, inode_dec_link_count(inode); iput(inode); } - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return err; } @@ -1336,6 +1349,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) int err = 0; int drop_on_err = 0; u64 objectid; + unsigned long nr = 1; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); @@ -1380,12 +1394,13 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) btrfs_update_inode_block_group(trans, dir); out_fail: + nr = trans->blocks_used; btrfs_end_transaction(trans, root); out_unlock: mutex_unlock(&root->fs_info->fs_mutex); if (drop_on_err) iput(inode); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return err; } @@ -1682,6 +1697,7 @@ static void btrfs_truncate(struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; int ret; struct btrfs_trans_handle *trans; + unsigned long nr; if (!S_ISREG(inode->i_mode)) return; @@ -1697,10 +1713,11 @@ static void btrfs_truncate(struct inode *inode) /* FIXME, add redo link to tree so we don't leak on crash */ ret = btrfs_truncate_in_trans(trans, root, inode); btrfs_update_inode(trans, root, inode); + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); } int btrfs_commit_write(struct file *file, struct page *page, @@ -1725,6 +1742,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) int err; u64 objectid; u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; + unsigned long nr = 1; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); @@ -1814,12 +1832,13 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; fail: + nr = trans->blocks_used; err = btrfs_commit_transaction(trans, root); if (err && !ret) ret = err; fail_commit: mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return ret; } @@ -1832,6 +1851,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) int ret; int err; u64 objectid; + unsigned long nr; if (!root->ref_cows) return -EINVAL; @@ -1884,12 +1904,13 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; fail: + nr = trans->blocks_used; err = btrfs_commit_transaction(trans, root); if (err && !ret) ret = err; mutex_unlock(&root->fs_info->fs_mutex); up_write(&root->snap_sem); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return ret; } @@ -2240,6 +2261,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, int datasize; char *ptr; struct btrfs_file_extent_item *ei; + unsigned long nr; name_len = strlen(symname) + 1; if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) @@ -2309,13 +2331,14 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, drop_inode = 1; out_unlock: + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { inode_dec_link_count(inode); iput(inode); } - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(root, nr); return err; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 29755593de6..087074db0bd 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -18,6 +18,7 @@ #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -340,6 +341,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) struct btrfs_fs_info *info = root->fs_info; int ret; struct btrfs_trans_handle *trans; + unsigned long nr; if (root->defrag_running) return 0; @@ -348,10 +350,11 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) while (1) { root->defrag_running = 1; ret = btrfs_defrag_leaves(trans, root, cacheonly); + nr = trans->blocks_used; btrfs_end_transaction(trans, root); mutex_unlock(&info->fs_mutex); - btrfs_btree_balance_dirty(root); + btrfs_btree_balance_dirty(info->tree_root, nr); cond_resched(); mutex_lock(&info->fs_mutex); @@ -398,6 +401,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, { struct dirty_root *dirty; struct btrfs_trans_handle *trans; + unsigned long nr; u64 num_blocks; u64 blocks_used; int ret = 0; @@ -426,11 +430,11 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, &dirty->root->root_item); if (err) ret = err; + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); mutex_unlock(&tree_root->fs_info->fs_mutex); - - btrfs_btree_balance_dirty(tree_root); + btrfs_btree_balance_dirty(tree_root, nr); schedule(); mutex_lock(&tree_root->fs_info->fs_mutex); @@ -449,13 +453,15 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, BUG(); break; } + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); kfree(dirty->root); kfree(dirty); mutex_unlock(&tree_root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(tree_root); + + btrfs_btree_balance_dirty(tree_root, nr); schedule(); } return ret; -- cgit v1.2.3-70-g09d2 From a61721d5b771d1735415a6b4c75409ab1fc9c845 Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 17 Sep 2007 11:08:38 -0400 Subject: fix found_type decrement in btrfs_truncate_in_trans found_type has already been decreased by codes above the change, I think decrease it by one again doesn't make sense. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bd00df33fb3..211ca32c65c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -581,7 +581,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, } else { break; } - btrfs_set_key_type(&key, found_type - 1); + btrfs_set_key_type(&key, found_type); continue; } if (btrfs_disk_key_offset(found_key) >= inode->i_size) -- cgit v1.2.3-70-g09d2 From 23223584e44cb1b0b69ba3a9ad74956f292a416b Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 17 Sep 2007 11:08:52 -0400 Subject: create btrfs_path slab with the correct size Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 211ca32c65c..192fdda0d56 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2127,7 +2127,7 @@ int btrfs_init_cachep(void) if (!btrfs_transaction_cachep) goto fail; btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache", - sizeof(struct btrfs_transaction), + sizeof(struct btrfs_path), 0, NULL); if (!btrfs_path_cachep) goto fail; -- cgit v1.2.3-70-g09d2 From bab9fb035faa5119991b64cfee1b14c1d6f6a31b Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 17 Sep 2007 11:13:11 -0400 Subject: Btrfs: truncate: don't update inode->i_blocks when extent is a hole I think check whether extent is a hole before update 'inode->i_blocks' is unconditional required. (original codes check it only when del_item isn't equal to 0) Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 192fdda0d56..43bf4337859 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -595,6 +595,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, btrfs_file_extent_type(fi) != BTRFS_FILE_EXTENT_INLINE) { u64 num_dec; + extent_start = btrfs_file_extent_disk_blocknr(fi); if (!del_item) { u64 orig_num_blocks = btrfs_file_extent_num_blocks(fi); @@ -604,12 +605,13 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, extent_num_blocks >>= inode->i_blkbits; btrfs_set_file_extent_num_blocks(fi, extent_num_blocks); - inode->i_blocks -= (orig_num_blocks - - extent_num_blocks) << 3; + num_dec = (orig_num_blocks - + extent_num_blocks) << 3; + if (extent_start != 0) { + inode->i_blocks -= num_dec; + } btrfs_mark_buffer_dirty(path->nodes[0]); } else { - extent_start = - btrfs_file_extent_disk_blocknr(fi); extent_num_blocks = btrfs_file_extent_disk_num_blocks(fi); /* FIXME blocksize != 4096 */ -- cgit v1.2.3-70-g09d2 From 8e1cd766643f040fd88b9c4221353a3f6b77db1e Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 17 Sep 2007 11:13:12 -0400 Subject: Btrfs: Fix double free and off by one in inode.c The first change removes potential double free, the second fix a off by one error. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 43bf4337859..88b3ef20be6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -206,7 +206,6 @@ void btrfs_read_locked_inode(struct inode *inode) memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); ret = btrfs_lookup_inode(NULL, root, path, &location, 0); if (ret) { - btrfs_free_path(path); goto make_bad; } inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), @@ -1536,7 +1535,7 @@ again: flush_dcache_page(result->b_page); kunmap(page); set_extent_uptodate(em_tree, extent_start, - extent_end, GFP_NOFS); + extent_end - 1, GFP_NOFS); goto insert; } else { printk("unkknown found_type %d\n", found_type); -- cgit v1.2.3-70-g09d2 From b3cfa35a493866cffd00952438b6c77364d42d3c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Sep 2007 11:25:58 -0400 Subject: Btrfs: factor page private preparations into a helper Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 40 +++++++++++++++------------------------- fs/btrfs/extent_map.h | 1 + fs/btrfs/file.c | 7 +------ fs/btrfs/inode.c | 7 +------ 4 files changed, 18 insertions(+), 37 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index ab5bde31b68..0ab368e091f 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1379,6 +1379,16 @@ static int submit_extent_page(int rw, struct extent_map_tree *tree, return ret; } +void set_page_extent_mapped(struct page *page) +{ + if (!PagePrivate(page)) { + SetPagePrivate(page); + WARN_ON(!page->mapping->a_ops->invalidatepage); + set_page_private(page, 1); + page_cache_get(page); + } +} + /* * basic readpage implementation. Locked extent state structs are inserted * into the tree that are removed when the IO is done (by the end_io @@ -1405,12 +1415,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, size_t iosize; size_t blocksize = inode->i_sb->s_blocksize; - if (!PagePrivate(page)) { - SetPagePrivate(page); - WARN_ON(!page->mapping->a_ops->invalidatepage); - set_page_private(page, 1); - page_cache_get(page); - } + set_page_extent_mapped(page); end = page_end; lock_extent(tree, start, end, GFP_NOFS); @@ -1531,12 +1536,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, PAGE_CACHE_SIZE - offset, KM_USER0); } - if (!PagePrivate(page)) { - SetPagePrivate(page); - set_page_private(page, 1); - WARN_ON(!page->mapping->a_ops->invalidatepage); - page_cache_get(page); - } + set_page_extent_mapped(page); lock_extent(tree, start, page_end, GFP_NOFS); nr_delalloc = find_lock_delalloc_range(tree, start, page_end + 1, @@ -1672,13 +1672,7 @@ int extent_commit_write(struct extent_map_tree *tree, { loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; - if (!PagePrivate(page)) { - SetPagePrivate(page); - set_page_private(page, 1); - WARN_ON(!page->mapping->a_ops->invalidatepage); - page_cache_get(page); - } - + set_page_extent_mapped(page); set_page_dirty(page); if (pos > inode->i_size) { @@ -1709,12 +1703,8 @@ int extent_prepare_write(struct extent_map_tree *tree, int ret = 0; int isnew; - if (!PagePrivate(page)) { - SetPagePrivate(page); - set_page_private(page, 1); - WARN_ON(!page->mapping->a_ops->invalidatepage); - page_cache_get(page); - } + set_page_extent_mapped(page); + block_start = (page_start + from) & ~((u64)blocksize - 1); block_end = (page_start + to - 1) | (blocksize - 1); orig_block_start = block_start; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 40b53ee274f..d913ce01248 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -105,4 +105,5 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock, int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); int set_state_private(struct extent_map_tree *tree, u64 start, u64 private); int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private); +void set_page_extent_mapped(struct page *page); #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 698eaea612f..4cc459c943e 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -543,12 +543,7 @@ static int prepare_pages(struct btrfs_root *root, } cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); wait_on_page_writeback(pages[i]); - if (!PagePrivate(pages[i])) { - SetPagePrivate(pages[i]); - set_page_private(pages[i], 1); - WARN_ON(!pages[i]->mapping->a_ops->invalidatepage); - page_cache_get(pages[i]); - } + set_page_extent_mapped(pages[i]); WARN_ON(!PageLocked(pages[i])); } return 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 88b3ef20be6..0df325f89d6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -652,12 +652,7 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, u64 page_start = page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; - if (!PagePrivate(page)) { - SetPagePrivate(page); - set_page_private(page, 1); - WARN_ON(!page->mapping->a_ops->invalidatepage); - page_cache_get(page); - } + set_page_extent_mapped(page); lock_extent(em_tree, page_start, page_end, GFP_NOFS); set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start, -- cgit v1.2.3-70-g09d2 From 50b78c24d594e6985f36a9a8d8e2fb74e1664e90 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 20 Sep 2007 14:14:42 -0400 Subject: btrfs_get_extent should treat inline extents as though they hold a whole block Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0df325f89d6..b03d40a907c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1505,13 +1505,13 @@ again: size = btrfs_file_extent_inline_len(leaf->items + path->slots[0]); - extent_end = extent_start + size; + extent_end = extent_start | ((u64)root->blocksize - 1); if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { if (end < extent_start) goto not_found; - em->end = extent_end - 1; + em->end = extent_end; } else { em->end = end; } @@ -1520,17 +1520,19 @@ again: em->block_start = EXTENT_MAP_INLINE; em->block_end = EXTENT_MAP_INLINE; em->start = extent_start; - em->end = extent_end - 1; + em->end = extent_end; if (!page) { goto insert; } ptr = btrfs_file_extent_inline_start(item); map = kmap(page); memcpy(map + page_offset, ptr, size); - flush_dcache_page(result->b_page); + memset(map + page_offset + size, 0, + root->blocksize - (page_offset + size)); + flush_dcache_page(page); kunmap(page); set_extent_uptodate(em_tree, extent_start, - extent_end - 1, GFP_NOFS); + extent_end, GFP_NOFS); goto insert; } else { printk("unkknown found_type %d\n", found_type); -- cgit v1.2.3-70-g09d2 From 5f39d397dfbe140a14edecd4e73c34ce23c4f9ee Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Oct 2007 16:14:19 -0400 Subject: Btrfs: Create extent_buffer interface for large blocksizes Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 + fs/btrfs/ctree.c | 1484 ++++++++++++++++++++++++++---------------------- fs/btrfs/ctree.h | 849 ++++++++++----------------- fs/btrfs/dir-item.c | 102 ++-- fs/btrfs/disk-io.c | 391 ++++++------- fs/btrfs/disk-io.h | 55 +- fs/btrfs/extent-tree.c | 294 +++++----- fs/btrfs/extent_map.c | 672 +++++++++++++++++++++- fs/btrfs/extent_map.h | 75 ++- fs/btrfs/file-item.c | 63 +- fs/btrfs/file.c | 129 +++-- fs/btrfs/inode-item.c | 22 +- fs/btrfs/inode-map.c | 17 +- fs/btrfs/inode.c | 435 ++++++++------ fs/btrfs/print-tree.c | 96 ++-- fs/btrfs/print-tree.h | 4 +- fs/btrfs/root-tree.c | 51 +- fs/btrfs/super.c | 7 +- fs/btrfs/sysfs.c | 10 +- fs/btrfs/transaction.c | 61 +- fs/btrfs/transaction.h | 5 +- fs/btrfs/tree-defrag.c | 58 +- 22 files changed, 2794 insertions(+), 2088 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 6341cf7a9f9..a3b51085d7f 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -23,4 +23,6 @@ modules_install: clean: $(MAKE) -C $(KERNELDIR) M=`pwd` clean +tester: + $(MAKE) -C $(KERNELDIR) M=`pwd` tree-defrag.o transaction.o sysfs.o super.o root-tree.o inode-map.o inode-item.o inode.o file-item.o file.o extent_map.o disk-io.o ctree.o dir-item.o extent-tree.o endif diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index b41f48ade41..f60920e8a0e 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -16,21 +16,24 @@ * Boston, MA 021110-1307, USA. */ +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" +#include "print-tree.h" static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level); static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *ins_key, struct btrfs_path *path, int data_size); -static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct buffer_head *dst, struct buffer_head - *src); -static int balance_node_right(struct btrfs_trans_handle *trans, struct - btrfs_root *root, struct buffer_head *dst_buf, - struct buffer_head *src_buf); +static int push_node_left(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *dst, + struct extent_buffer *src); +static int balance_node_right(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *dst_buf, + struct extent_buffer *src_buf); static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level, int slot); @@ -62,40 +65,38 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) for (i = 0; i < BTRFS_MAX_LEVEL; i++) { if (!p->nodes[i]) break; - btrfs_block_release(root, p->nodes[i]); + free_extent_buffer(p->nodes[i]); } memset(p, 0, sizeof(*p)); } -static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct buffer_head *buf, struct buffer_head - *parent, int parent_slot, struct buffer_head - **cow_ret, u64 search_start, u64 empty_size) +static int __btrfs_cow_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *buf, + struct extent_buffer *parent, int parent_slot, + struct extent_buffer **cow_ret, + u64 search_start, u64 empty_size) { - struct buffer_head *cow; - struct btrfs_node *cow_node; + struct extent_buffer *cow; int ret = 0; int different_trans = 0; WARN_ON(root->ref_cows && trans->transid != root->last_trans); - WARN_ON(!buffer_uptodate(buf)); + cow = btrfs_alloc_free_block(trans, root, search_start, empty_size); if (IS_ERR(cow)) return PTR_ERR(cow); - cow_node = btrfs_buffer_node(cow); - if (buf->b_size != root->blocksize || cow->b_size != root->blocksize) + if (buf->len != root->sectorsize || cow->len != root->sectorsize) WARN_ON(1); - memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize); - btrfs_set_header_blocknr(&cow_node->header, bh_blocknr(cow)); - btrfs_set_header_generation(&cow_node->header, trans->transid); - btrfs_set_header_owner(&cow_node->header, root->root_key.objectid); + copy_extent_buffer(cow, buf, 0, 0, cow->len); + btrfs_set_header_blocknr(cow, extent_buffer_blocknr(cow)); + btrfs_set_header_generation(cow, trans->transid); + btrfs_set_header_owner(cow, root->root_key.objectid); - WARN_ON(btrfs_header_generation(btrfs_buffer_header(buf)) > - trans->transid); - if (btrfs_header_generation(btrfs_buffer_header(buf)) != - trans->transid) { + WARN_ON(btrfs_header_generation(buf) > trans->transid); + if (btrfs_header_generation(buf) != trans->transid) { different_trans = 1; ret = btrfs_inc_ref(trans, root, buf); if (ret) @@ -106,29 +107,29 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root if (buf == root->node) { root->node = cow; - get_bh(cow); + extent_buffer_get(cow); if (buf != root->commit_root) { - btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1); + btrfs_free_extent(trans, root, + extent_buffer_blocknr(buf), 1, 1); } - btrfs_block_release(root, buf); + free_extent_buffer(buf); } else { - btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot, - bh_blocknr(cow)); + btrfs_set_node_blockptr(parent, parent_slot, + extent_buffer_blocknr(cow)); btrfs_mark_buffer_dirty(parent); - WARN_ON(btrfs_header_generation(btrfs_buffer_header(parent)) != - trans->transid); - btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1); + WARN_ON(btrfs_header_generation(parent) != trans->transid); + btrfs_free_extent(trans, root, extent_buffer_blocknr(buf),1,1); } - btrfs_block_release(root, buf); + free_extent_buffer(buf); btrfs_mark_buffer_dirty(cow); *cow_ret = cow; return 0; } -int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct buffer_head *buf, struct buffer_head - *parent, int parent_slot, struct buffer_head - **cow_ret) +int btrfs_cow_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *buf, + struct extent_buffer *parent, int parent_slot, + struct extent_buffer **cow_ret) { u64 search_start; if (trans->transaction != root->fs_info->running_transaction) { @@ -141,13 +142,12 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root root->fs_info->generation); WARN_ON(1); } - if (btrfs_header_generation(btrfs_buffer_header(buf)) == - trans->transid) { + if (btrfs_header_generation(buf) == trans->transid) { *cow_ret = buf; return 0; } - search_start = bh_blocknr(buf) & ~((u64)65535); + search_start = extent_buffer_blocknr(buf) & ~((u64)65535); return __btrfs_cow_block(trans, root, buf, parent, parent_slot, cow_ret, search_start, 0); } @@ -161,9 +161,11 @@ static int close_blocks(u64 blocknr, u64 other) return 0; } -static int should_defrag_leaf(struct buffer_head *bh) +#if 0 +static int should_defrag_leaf(struct extent_buffer *eb) { - struct btrfs_leaf *leaf = btrfs_buffer_leaf(bh); + return 0; + struct btrfs_leaf *leaf = btrfs_buffer_leaf(eb); struct btrfs_disk_key *key; u32 nritems; @@ -188,14 +190,17 @@ static int should_defrag_leaf(struct buffer_head *bh) } return 0; } +#endif int btrfs_realloc_node(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct buffer_head *parent, + struct btrfs_root *root, struct extent_buffer *parent, int cache_only, u64 *last_ret) { + return 0; +#if 0 struct btrfs_node *parent_node; - struct buffer_head *cur_bh; - struct buffer_head *tmp_bh; + struct extent_buffer *cur_eb; + struct extent_buffer *tmp_eb; u64 blocknr; u64 search_start = *last_ret; u64 last_block = 0; @@ -281,6 +286,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, brelse(tmp_bh); } return err; +#endif } /* @@ -289,12 +295,12 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, * which is the stop of the leaf data stack */ static inline unsigned int leaf_data_end(struct btrfs_root *root, - struct btrfs_leaf *leaf) + struct extent_buffer *leaf) { - u32 nr = btrfs_header_nritems(&leaf->header); + u32 nr = btrfs_header_nritems(leaf); if (nr == 0) return BTRFS_LEAF_DATA_SIZE(root); - return btrfs_item_offset(leaf->items + nr - 1); + return btrfs_item_offset_nr(leaf, nr - 1); } /* @@ -310,9 +316,9 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2) return 1; if (k1.objectid < k2->objectid) return -1; - if (k1.flags > k2->flags) + if (k1.type > k2->type) return 1; - if (k1.flags < k2->flags) + if (k1.type < k2->type) return -1; if (k1.offset > k2->offset) return 1; @@ -324,37 +330,39 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2) static int check_node(struct btrfs_root *root, struct btrfs_path *path, int level) { - struct btrfs_node *parent = NULL; - struct btrfs_node *node = btrfs_buffer_node(path->nodes[level]); + struct extent_buffer *parent = NULL; + struct extent_buffer *node = path->nodes[level]; + struct btrfs_disk_key parent_key; + struct btrfs_disk_key node_key; int parent_slot; int slot; struct btrfs_key cpukey; - u32 nritems = btrfs_header_nritems(&node->header); + u32 nritems = btrfs_header_nritems(node); if (path->nodes[level + 1]) - parent = btrfs_buffer_node(path->nodes[level + 1]); + parent = path->nodes[level + 1]; slot = path->slots[level]; - BUG_ON(!buffer_uptodate(path->nodes[level])); BUG_ON(nritems == 0); if (parent) { - struct btrfs_disk_key *parent_key; - parent_slot = path->slots[level + 1]; - parent_key = &parent->ptrs[parent_slot].key; - BUG_ON(memcmp(parent_key, &node->ptrs[0].key, + btrfs_node_key(parent, &parent_key, parent_slot); + btrfs_node_key(node, &node_key, 0); + BUG_ON(memcmp(&parent_key, &node_key, sizeof(struct btrfs_disk_key))); BUG_ON(btrfs_node_blockptr(parent, parent_slot) != - btrfs_header_blocknr(&node->header)); + btrfs_header_blocknr(node)); } BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root)); if (slot != 0) { - btrfs_disk_key_to_cpu(&cpukey, &node->ptrs[slot - 1].key); - BUG_ON(comp_keys(&node->ptrs[slot].key, &cpukey) <= 0); + btrfs_node_key_to_cpu(node, &cpukey, slot - 1); + btrfs_node_key(node, &node_key, slot); + BUG_ON(comp_keys(&node_key, &cpukey) <= 0); } if (slot < nritems - 1) { - btrfs_disk_key_to_cpu(&cpukey, &node->ptrs[slot + 1].key); - BUG_ON(comp_keys(&node->ptrs[slot].key, &cpukey) >= 0); + btrfs_node_key_to_cpu(node, &cpukey, slot + 1); + btrfs_node_key(node, &node_key, slot); + BUG_ON(comp_keys(&node_key, &cpukey) >= 0); } return 0; } @@ -362,83 +370,172 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path, static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, int level) { - struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[level]); - struct btrfs_node *parent = NULL; + struct extent_buffer *leaf = path->nodes[level]; + struct extent_buffer *parent = NULL; int parent_slot; - int slot = path->slots[0]; struct btrfs_key cpukey; + struct btrfs_disk_key parent_key; + struct btrfs_disk_key leaf_key; + int slot = path->slots[0]; - u32 nritems = btrfs_header_nritems(&leaf->header); + u32 nritems = btrfs_header_nritems(leaf); if (path->nodes[level + 1]) - parent = btrfs_buffer_node(path->nodes[level + 1]); - - BUG_ON(btrfs_leaf_free_space(root, leaf) < 0); + parent = path->nodes[level + 1]; if (nritems == 0) return 0; if (parent) { - struct btrfs_disk_key *parent_key; - parent_slot = path->slots[level + 1]; - parent_key = &parent->ptrs[parent_slot].key; + btrfs_node_key(parent, &parent_key, parent_slot); + btrfs_item_key(leaf, &leaf_key, 0); - BUG_ON(memcmp(parent_key, &leaf->items[0].key, + BUG_ON(memcmp(&parent_key, &leaf_key, sizeof(struct btrfs_disk_key))); BUG_ON(btrfs_node_blockptr(parent, parent_slot) != - btrfs_header_blocknr(&leaf->header)); + btrfs_header_blocknr(leaf)); + } +#if 0 + for (i = 0; nritems > 1 && i < nritems - 2; i++) { + btrfs_item_key_to_cpu(leaf, &cpukey, i + 1); + btrfs_item_key(leaf, &leaf_key, i); + if (comp_keys(&leaf_key, &cpukey) >= 0) { + btrfs_print_leaf(root, leaf); + printk("slot %d offset bad key\n", i); + BUG_ON(1); + } + if (btrfs_item_offset_nr(leaf, i) != + btrfs_item_end_nr(leaf, i + 1)) { + btrfs_print_leaf(root, leaf); + printk("slot %d offset bad\n", i); + BUG_ON(1); + } + if (i == 0) { + if (btrfs_item_offset_nr(leaf, i) + + btrfs_item_size_nr(leaf, i) != + BTRFS_LEAF_DATA_SIZE(root)) { + btrfs_print_leaf(root, leaf); + printk("slot %d first offset bad\n", i); + BUG_ON(1); + } + } } - if (slot != 0) { - btrfs_disk_key_to_cpu(&cpukey, &leaf->items[slot - 1].key); - BUG_ON(comp_keys(&leaf->items[slot].key, &cpukey) <= 0); - BUG_ON(btrfs_item_offset(leaf->items + slot - 1) != - btrfs_item_end(leaf->items + slot)); + if (nritems > 0) { + if (btrfs_item_size_nr(leaf, nritems - 1) > 4096) { + btrfs_print_leaf(root, leaf); + printk("slot %d bad size \n", nritems - 1); + BUG_ON(1); + } + } +#endif + if (slot != 0 && slot < nritems - 1) { + btrfs_item_key(leaf, &leaf_key, slot); + btrfs_item_key_to_cpu(leaf, &cpukey, slot - 1); + if (comp_keys(&leaf_key, &cpukey) <= 0) { + btrfs_print_leaf(root, leaf); + printk("slot %d offset bad key\n", slot); + BUG_ON(1); + } + if (btrfs_item_offset_nr(leaf, slot - 1) != + btrfs_item_end_nr(leaf, slot)) { + btrfs_print_leaf(root, leaf); + printk("slot %d offset bad\n", slot); + BUG_ON(1); + } } if (slot < nritems - 1) { - btrfs_disk_key_to_cpu(&cpukey, &leaf->items[slot + 1].key); - BUG_ON(comp_keys(&leaf->items[slot].key, &cpukey) >= 0); - BUG_ON(btrfs_item_offset(leaf->items + slot) != - btrfs_item_end(leaf->items + slot + 1)); + btrfs_item_key(leaf, &leaf_key, slot); + btrfs_item_key_to_cpu(leaf, &cpukey, slot + 1); + BUG_ON(comp_keys(&leaf_key, &cpukey) >= 0); + if (btrfs_item_offset_nr(leaf, slot) != + btrfs_item_end_nr(leaf, slot + 1)) { + btrfs_print_leaf(root, leaf); + printk("slot %d offset bad\n", slot); + BUG_ON(1); + } } - BUG_ON(btrfs_item_offset(leaf->items) + - btrfs_item_size(leaf->items) != BTRFS_LEAF_DATA_SIZE(root)); + BUG_ON(btrfs_item_offset_nr(leaf, 0) + + btrfs_item_size_nr(leaf, 0) != BTRFS_LEAF_DATA_SIZE(root)); return 0; } static int check_block(struct btrfs_root *root, struct btrfs_path *path, int level) { - struct btrfs_node *node = btrfs_buffer_node(path->nodes[level]); - if (memcmp(node->header.fsid, root->fs_info->disk_super->fsid, - sizeof(node->header.fsid))) - BUG(); + struct extent_buffer *buf = path->nodes[level]; + char fsid[BTRFS_FSID_SIZE]; + + read_extent_buffer(buf, fsid, (unsigned long)btrfs_header_fsid(buf), + BTRFS_FSID_SIZE); + + if (memcmp(fsid, root->fs_info->fsid, BTRFS_FSID_SIZE)) { + int i = 0; + printk("warning bad block %Lu\n", buf->start); + if (!btrfs_buffer_uptodate(buf)) { + WARN_ON(1); + } + for (i = 0; i < BTRFS_FSID_SIZE; i++) { + printk("%x:%x ", root->fs_info->fsid[i], fsid[i]); + } + printk("\n"); + // BUG(); + } if (level == 0) return check_leaf(root, path, level); return check_node(root, path, level); } /* - * search for key in the array p. items p are item_size apart - * and there are 'max' items in p + * search for key in the extent_buffer. The items start at offset p, + * and they are item_size apart. There are 'max' items in p. + * * the slot in the array is returned via slot, and it points to * the place where you would insert key if it is not found in * the array. * * slot may point to max if the key is bigger than all of the keys */ -static int generic_bin_search(char *p, int item_size, struct btrfs_key *key, - int max, int *slot) +static int generic_bin_search(struct extent_buffer *eb, unsigned long p, + int item_size, struct btrfs_key *key, + int max, int *slot) { int low = 0; int high = max; int mid; int ret; struct btrfs_disk_key *tmp; + struct btrfs_disk_key unaligned; + unsigned long offset; + char *map_token = NULL; + char *kaddr = NULL; + unsigned long map_start = 0; + unsigned long map_len = 0; while(low < high) { mid = (low + high) / 2; - tmp = (struct btrfs_disk_key *)(p + mid * item_size); + offset = p + mid * item_size; + + if (!map_token || offset < map_start || + (offset + sizeof(struct btrfs_disk_key)) > + map_start + map_len) { + if (map_token) + unmap_extent_buffer(eb, map_token, KM_USER0); + map_extent_buffer(eb, offset, &map_token, &kaddr, + &map_start, &map_len, KM_USER0); + + } + if (offset + sizeof(struct btrfs_disk_key) > + map_start + map_len) { + unmap_extent_buffer(eb, map_token, KM_USER0); + read_extent_buffer(eb, &unaligned, + offset, sizeof(unaligned)); + map_token = NULL; + tmp = &unaligned; + } else { + tmp = (struct btrfs_disk_key *)(kaddr + offset - + map_start); + } ret = comp_keys(tmp, key); if (ret < 0) @@ -447,10 +544,13 @@ static int generic_bin_search(char *p, int item_size, struct btrfs_key *key, high = mid; else { *slot = mid; + unmap_extent_buffer(eb, map_token, KM_USER0); return 0; } } *slot = low; + if (map_token) + unmap_extent_buffer(eb, map_token, KM_USER0); return 1; } @@ -458,46 +558,42 @@ static int generic_bin_search(char *p, int item_size, struct btrfs_key *key, * simple bin_search frontend that does the right thing for * leaves vs nodes */ -static int bin_search(struct btrfs_node *c, struct btrfs_key *key, int *slot) +static int bin_search(struct extent_buffer *eb, struct btrfs_key *key, + int level, int *slot) { - if (btrfs_is_leaf(c)) { - struct btrfs_leaf *l = (struct btrfs_leaf *)c; - return generic_bin_search((void *)l->items, + if (level == 0) { + return generic_bin_search(eb, + offsetof(struct btrfs_leaf, items), sizeof(struct btrfs_item), - key, btrfs_header_nritems(&c->header), + key, btrfs_header_nritems(eb), slot); } else { - return generic_bin_search((void *)c->ptrs, + return generic_bin_search(eb, + offsetof(struct btrfs_node, ptrs), sizeof(struct btrfs_key_ptr), - key, btrfs_header_nritems(&c->header), + key, btrfs_header_nritems(eb), slot); } return -1; } -static struct buffer_head *read_node_slot(struct btrfs_root *root, - struct buffer_head *parent_buf, - int slot) +static struct extent_buffer *read_node_slot(struct btrfs_root *root, + struct extent_buffer *parent, int slot) { - struct btrfs_node *node = btrfs_buffer_node(parent_buf); if (slot < 0) return NULL; - if (slot >= btrfs_header_nritems(&node->header)) + if (slot >= btrfs_header_nritems(parent)) return NULL; - return read_tree_block(root, btrfs_node_blockptr(node, slot)); + return read_tree_block(root, btrfs_node_blockptr(parent, slot)); } static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { - struct buffer_head *right_buf; - struct buffer_head *mid_buf; - struct buffer_head *left_buf; - struct buffer_head *parent_buf = NULL; - struct btrfs_node *right = NULL; - struct btrfs_node *mid; - struct btrfs_node *left = NULL; - struct btrfs_node *parent = NULL; + struct extent_buffer *right = NULL; + struct extent_buffer *mid; + struct extent_buffer *left = NULL; + struct extent_buffer *parent = NULL; int ret = 0; int wret; int pslot; @@ -508,60 +604,57 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root if (level == 0) return 0; - mid_buf = path->nodes[level]; - mid = btrfs_buffer_node(mid_buf); + mid = path->nodes[level]; orig_ptr = btrfs_node_blockptr(mid, orig_slot); if (level < BTRFS_MAX_LEVEL - 1) - parent_buf = path->nodes[level + 1]; + parent = path->nodes[level + 1]; pslot = path->slots[level + 1]; /* * deal with the case where there is only one pointer in the root * by promoting the node below to a root */ - if (!parent_buf) { - struct buffer_head *child; - u64 blocknr = bh_blocknr(mid_buf); + if (!parent) { + struct extent_buffer *child; + u64 blocknr = extent_buffer_blocknr(mid); - if (btrfs_header_nritems(&mid->header) != 1) + if (btrfs_header_nritems(mid) != 1) return 0; /* promote the child to a root */ - child = read_node_slot(root, mid_buf, 0); + child = read_node_slot(root, mid, 0); BUG_ON(!child); root->node = child; path->nodes[level] = NULL; - clean_tree_block(trans, root, mid_buf); - wait_on_buffer(mid_buf); + clean_tree_block(trans, root, mid); + wait_on_tree_block_writeback(root, mid); /* once for the path */ - btrfs_block_release(root, mid_buf); + free_extent_buffer(mid); /* once for the root ptr */ - btrfs_block_release(root, mid_buf); + free_extent_buffer(mid); return btrfs_free_extent(trans, root, blocknr, 1, 1); } - parent = btrfs_buffer_node(parent_buf); - - if (btrfs_header_nritems(&mid->header) > + if (btrfs_header_nritems(mid) > BTRFS_NODEPTRS_PER_BLOCK(root) / 4) return 0; - if (btrfs_header_nritems(&mid->header) < 2) + if (btrfs_header_nritems(mid) < 2) err_on_enospc = 1; - left_buf = read_node_slot(root, parent_buf, pslot - 1); - if (left_buf) { - wret = btrfs_cow_block(trans, root, left_buf, - parent_buf, pslot - 1, &left_buf); + left = read_node_slot(root, parent, pslot - 1); + if (left) { + wret = btrfs_cow_block(trans, root, left, + parent, pslot - 1, &left); if (wret) { ret = wret; goto enospc; } } - right_buf = read_node_slot(root, parent_buf, pslot + 1); - if (right_buf) { - wret = btrfs_cow_block(trans, root, right_buf, - parent_buf, pslot + 1, &right_buf); + right = read_node_slot(root, parent, pslot + 1); + if (right) { + wret = btrfs_cow_block(trans, root, right, + parent, pslot + 1, &right); if (wret) { ret = wret; goto enospc; @@ -569,30 +662,27 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root } /* first, try to make some room in the middle buffer */ - if (left_buf) { - left = btrfs_buffer_node(left_buf); - orig_slot += btrfs_header_nritems(&left->header); - wret = push_node_left(trans, root, left_buf, mid_buf); + if (left) { + orig_slot += btrfs_header_nritems(left); + wret = push_node_left(trans, root, left, mid); if (wret < 0) ret = wret; - if (btrfs_header_nritems(&mid->header) < 2) + if (btrfs_header_nritems(mid) < 2) err_on_enospc = 1; } /* * then try to empty the right most buffer into the middle */ - if (right_buf) { - right = btrfs_buffer_node(right_buf); - wret = push_node_left(trans, root, mid_buf, right_buf); + if (right) { + wret = push_node_left(trans, root, mid, right); if (wret < 0 && wret != -ENOSPC) ret = wret; - if (btrfs_header_nritems(&right->header) == 0) { - u64 blocknr = bh_blocknr(right_buf); - clean_tree_block(trans, root, right_buf); - wait_on_buffer(right_buf); - btrfs_block_release(root, right_buf); - right_buf = NULL; + if (btrfs_header_nritems(right) == 0) { + u64 blocknr = extent_buffer_blocknr(right); + clean_tree_block(trans, root, right); + wait_on_tree_block_writeback(root, right); + free_extent_buffer(right); right = NULL; wret = del_ptr(trans, root, path, level + 1, pslot + 1); @@ -602,14 +692,13 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root if (wret) ret = wret; } else { - btrfs_memcpy(root, parent, - &parent->ptrs[pslot + 1].key, - &right->ptrs[0].key, - sizeof(struct btrfs_disk_key)); - btrfs_mark_buffer_dirty(parent_buf); + struct btrfs_disk_key right_key; + btrfs_node_key(right, &right_key, 0); + btrfs_set_node_key(parent, &right_key, pslot + 1); + btrfs_mark_buffer_dirty(parent); } } - if (btrfs_header_nritems(&mid->header) == 1) { + if (btrfs_header_nritems(mid) == 1) { /* * we're not allowed to leave a node with one item in the * tree during a delete. A deletion from lower in the tree @@ -619,21 +708,20 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root * otherwise we would have pulled some pointers from the * right */ - BUG_ON(!left_buf); - wret = balance_node_right(trans, root, mid_buf, left_buf); + BUG_ON(!left); + wret = balance_node_right(trans, root, mid, left); if (wret < 0) { ret = wret; goto enospc; } BUG_ON(wret == 1); } - if (btrfs_header_nritems(&mid->header) == 0) { + if (btrfs_header_nritems(mid) == 0) { /* we've managed to empty the middle node, drop it */ - u64 blocknr = bh_blocknr(mid_buf); - clean_tree_block(trans, root, mid_buf); - wait_on_buffer(mid_buf); - btrfs_block_release(root, mid_buf); - mid_buf = NULL; + u64 blocknr = extent_buffer_blocknr(mid); + clean_tree_block(trans, root, mid); + wait_on_tree_block_writeback(root, mid); + free_extent_buffer(mid); mid = NULL; wret = del_ptr(trans, root, path, level + 1, pslot); if (wret) @@ -643,37 +731,36 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root ret = wret; } else { /* update the parent key to reflect our changes */ - btrfs_memcpy(root, parent, - &parent->ptrs[pslot].key, &mid->ptrs[0].key, - sizeof(struct btrfs_disk_key)); - btrfs_mark_buffer_dirty(parent_buf); + struct btrfs_disk_key mid_key; + btrfs_node_key(mid, &mid_key, 0); + btrfs_set_node_key(parent, &mid_key, pslot); + btrfs_mark_buffer_dirty(parent); } /* update the path */ - if (left_buf) { - if (btrfs_header_nritems(&left->header) > orig_slot) { - get_bh(left_buf); - path->nodes[level] = left_buf; + if (left) { + if (btrfs_header_nritems(left) > orig_slot) { + extent_buffer_get(left); + path->nodes[level] = left; path->slots[level + 1] -= 1; path->slots[level] = orig_slot; - if (mid_buf) - btrfs_block_release(root, mid_buf); + if (mid) + free_extent_buffer(mid); } else { - orig_slot -= btrfs_header_nritems(&left->header); + orig_slot -= btrfs_header_nritems(left); path->slots[level] = orig_slot; } } /* double check we haven't messed things up */ check_block(root, path, level); if (orig_ptr != - btrfs_node_blockptr(btrfs_buffer_node(path->nodes[level]), - path->slots[level])) + btrfs_node_blockptr(path->nodes[level], path->slots[level])) BUG(); enospc: - if (right_buf) - btrfs_block_release(root, right_buf); - if (left_buf) - btrfs_block_release(root, left_buf); + if (right) + free_extent_buffer(right); + if (left) + free_extent_buffer(left); return ret; } @@ -682,14 +769,10 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { - struct buffer_head *right_buf; - struct buffer_head *mid_buf; - struct buffer_head *left_buf; - struct buffer_head *parent_buf = NULL; - struct btrfs_node *right = NULL; - struct btrfs_node *mid; - struct btrfs_node *left = NULL; - struct btrfs_node *parent = NULL; + struct extent_buffer *right = NULL; + struct extent_buffer *mid; + struct extent_buffer *left = NULL; + struct extent_buffer *parent = NULL; int ret = 0; int wret; int pslot; @@ -699,107 +782,101 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, if (level == 0) return 1; - mid_buf = path->nodes[level]; - mid = btrfs_buffer_node(mid_buf); + mid = path->nodes[level]; orig_ptr = btrfs_node_blockptr(mid, orig_slot); if (level < BTRFS_MAX_LEVEL - 1) - parent_buf = path->nodes[level + 1]; + parent = path->nodes[level + 1]; pslot = path->slots[level + 1]; - if (!parent_buf) + if (!parent) return 1; - parent = btrfs_buffer_node(parent_buf); - left_buf = read_node_slot(root, parent_buf, pslot - 1); + left = read_node_slot(root, parent, pslot - 1); /* first, try to make some room in the middle buffer */ - if (left_buf) { + if (left) { u32 left_nr; - left = btrfs_buffer_node(left_buf); - left_nr = btrfs_header_nritems(&left->header); + left_nr = btrfs_header_nritems(left); if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; } else { - ret = btrfs_cow_block(trans, root, left_buf, parent_buf, - pslot - 1, &left_buf); + ret = btrfs_cow_block(trans, root, left, parent, + pslot - 1, &left); if (ret) wret = 1; else { - left = btrfs_buffer_node(left_buf); wret = push_node_left(trans, root, - left_buf, mid_buf); + left, mid); } } if (wret < 0) ret = wret; if (wret == 0) { + struct btrfs_disk_key disk_key; orig_slot += left_nr; - btrfs_memcpy(root, parent, - &parent->ptrs[pslot].key, - &mid->ptrs[0].key, - sizeof(struct btrfs_disk_key)); - btrfs_mark_buffer_dirty(parent_buf); - if (btrfs_header_nritems(&left->header) > orig_slot) { - path->nodes[level] = left_buf; + btrfs_node_key(mid, &disk_key, 0); + btrfs_set_node_key(parent, &disk_key, pslot); + btrfs_mark_buffer_dirty(parent); + if (btrfs_header_nritems(left) > orig_slot) { + path->nodes[level] = left; path->slots[level + 1] -= 1; path->slots[level] = orig_slot; - btrfs_block_release(root, mid_buf); + free_extent_buffer(mid); } else { orig_slot -= - btrfs_header_nritems(&left->header); + btrfs_header_nritems(left); path->slots[level] = orig_slot; - btrfs_block_release(root, left_buf); + free_extent_buffer(left); } check_node(root, path, level); return 0; } - btrfs_block_release(root, left_buf); + free_extent_buffer(left); } - right_buf = read_node_slot(root, parent_buf, pslot + 1); + right= read_node_slot(root, parent, pslot + 1); /* * then try to empty the right most buffer into the middle */ - if (right_buf) { + if (right) { u32 right_nr; - right = btrfs_buffer_node(right_buf); - right_nr = btrfs_header_nritems(&right->header); + right_nr = btrfs_header_nritems(right); if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; } else { - ret = btrfs_cow_block(trans, root, right_buf, - parent_buf, pslot + 1, - &right_buf); + ret = btrfs_cow_block(trans, root, right, + parent, pslot + 1, + &right); if (ret) wret = 1; else { - right = btrfs_buffer_node(right_buf); wret = balance_node_right(trans, root, - right_buf, mid_buf); + right, mid); } } if (wret < 0) ret = wret; if (wret == 0) { - btrfs_memcpy(root, parent, - &parent->ptrs[pslot + 1].key, - &right->ptrs[0].key, - sizeof(struct btrfs_disk_key)); - btrfs_mark_buffer_dirty(parent_buf); - if (btrfs_header_nritems(&mid->header) <= orig_slot) { - path->nodes[level] = right_buf; + struct btrfs_disk_key disk_key; + + btrfs_node_key(right, &disk_key, 0); + btrfs_set_node_key(parent, &disk_key, pslot + 1); + btrfs_mark_buffer_dirty(parent); + + if (btrfs_header_nritems(mid) <= orig_slot) { + path->nodes[level] = right; path->slots[level + 1] += 1; path->slots[level] = orig_slot - - btrfs_header_nritems(&mid->header); - btrfs_block_release(root, mid_buf); + btrfs_header_nritems(mid); + free_extent_buffer(mid); } else { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); } check_node(root, path, level); return 0; } - btrfs_block_release(root, right_buf); + free_extent_buffer(right); } check_node(root, path, level); return 1; @@ -811,10 +888,9 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, int level, int slot) { - struct btrfs_node *node; + struct extent_buffer *node; int i; u32 nritems; - u64 item_objectid; u64 blocknr; u64 search; u64 cluster_start; @@ -823,7 +899,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, int direction = path->reada; struct radix_tree_root found; unsigned long gang[8]; - struct buffer_head *bh; + struct extent_buffer *eb; if (level == 0) return; @@ -831,18 +907,17 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, if (!path->nodes[level]) return; - node = btrfs_buffer_node(path->nodes[level]); + node = path->nodes[level]; search = btrfs_node_blockptr(node, slot); - bh = btrfs_find_tree_block(root, search); - if (bh) { - brelse(bh); + eb = btrfs_find_tree_block(root, search); + if (eb) { + free_extent_buffer(eb); return; } init_bit_radix(&found); - nritems = btrfs_header_nritems(&node->header); + nritems = btrfs_header_nritems(node); for (i = slot; i < nritems; i++) { - item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key); blocknr = btrfs_node_blockptr(node, i); set_radix_bit(&found, blocknr); } @@ -886,8 +961,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_path *p, int ins_len, int cow) { - struct buffer_head *b; - struct btrfs_node *c; + struct extent_buffer *b; u64 blocknr; int slot; int ret; @@ -901,10 +975,9 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex)); again: b = root->node; - get_bh(b); + extent_buffer_get(b); while (b) { - c = btrfs_buffer_node(b); - level = btrfs_header_level(&c->header); + level = btrfs_header_level(b); if (cow) { int wret; wret = btrfs_cow_block(trans, root, b, @@ -912,32 +985,30 @@ again: p->slots[level + 1], &b); if (wret) { - btrfs_block_release(root, b); + free_extent_buffer(b); return wret; } - c = btrfs_buffer_node(b); } BUG_ON(!cow && ins_len); - if (level != btrfs_header_level(&c->header)) + if (level != btrfs_header_level(b)) WARN_ON(1); - level = btrfs_header_level(&c->header); + level = btrfs_header_level(b); p->nodes[level] = b; ret = check_block(root, p, level); if (ret) return -1; - ret = bin_search(c, key, &slot); - if (!btrfs_is_leaf(c)) { + ret = bin_search(b, key, level, &slot); + if (level != 0) { if (ret && slot > 0) slot -= 1; p->slots[level] = slot; - if (ins_len > 0 && btrfs_header_nritems(&c->header) >= + if (ins_len > 0 && btrfs_header_nritems(b) >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { int sret = split_node(trans, root, p, level); BUG_ON(sret > 0); if (sret) return sret; b = p->nodes[level]; - c = btrfs_buffer_node(b); slot = p->slots[level]; } else if (ins_len < 0) { int sret = balance_level(trans, root, p, @@ -947,22 +1018,19 @@ again: b = p->nodes[level]; if (!b) goto again; - c = btrfs_buffer_node(b); slot = p->slots[level]; - BUG_ON(btrfs_header_nritems(&c->header) == 1); + BUG_ON(btrfs_header_nritems(b) == 1); } /* this is only true while dropping a snapshot */ if (level == lowest_level) break; - blocknr = btrfs_node_blockptr(c, slot); + blocknr = btrfs_node_blockptr(b, slot); if (should_reada) reada_for_search(root, p, level, slot); - b = read_tree_block(root, btrfs_node_blockptr(c, slot)); - + b = read_tree_block(root, btrfs_node_blockptr(b, slot)); } else { - struct btrfs_leaf *l = (struct btrfs_leaf *)c; p->slots[level] = slot; - if (ins_len > 0 && btrfs_leaf_free_space(root, l) < + if (ins_len > 0 && btrfs_leaf_free_space(root, b) < sizeof(struct btrfs_item) + ins_len) { int sret = split_leaf(trans, root, key, p, ins_len); @@ -986,19 +1054,20 @@ again: * If this fails to write a tree block, it returns -1, but continues * fixing up the blocks in ram so the tree is consistent. */ -static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, struct btrfs_disk_key - *key, int level) +static int fixup_low_keys(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_path *path, + struct btrfs_disk_key *key, int level) { int i; int ret = 0; + struct extent_buffer *t; + for (i = level; i < BTRFS_MAX_LEVEL; i++) { - struct btrfs_node *t; int tslot = path->slots[i]; if (!path->nodes[i]) break; - t = btrfs_buffer_node(path->nodes[i]); - btrfs_memcpy(root, t, &t->ptrs[tslot].key, key, sizeof(*key)); + t = path->nodes[i]; + btrfs_set_node_key(t, key, tslot); btrfs_mark_buffer_dirty(path->nodes[i]); if (tslot != 0) break; @@ -1014,18 +1083,16 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root * error, and > 0 if there was no room in the left hand block. */ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct buffer_head *dst_buf, struct - buffer_head *src_buf) + *root, struct extent_buffer *dst, + struct extent_buffer *src) { - struct btrfs_node *src = btrfs_buffer_node(src_buf); - struct btrfs_node *dst = btrfs_buffer_node(dst_buf); int push_items = 0; int src_nritems; int dst_nritems; int ret = 0; - src_nritems = btrfs_header_nritems(&src->header); - dst_nritems = btrfs_header_nritems(&dst->header); + src_nritems = btrfs_header_nritems(src); + dst_nritems = btrfs_header_nritems(dst); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; if (push_items <= 0) { @@ -1035,17 +1102,21 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root if (src_nritems < push_items) push_items = src_nritems; - btrfs_memcpy(root, dst, dst->ptrs + dst_nritems, src->ptrs, - push_items * sizeof(struct btrfs_key_ptr)); + copy_extent_buffer(dst, src, + btrfs_node_key_ptr_offset(dst_nritems), + btrfs_node_key_ptr_offset(0), + push_items * sizeof(struct btrfs_key_ptr)); + if (push_items < src_nritems) { - btrfs_memmove(root, src, src->ptrs, src->ptrs + push_items, - (src_nritems - push_items) * - sizeof(struct btrfs_key_ptr)); - } - btrfs_set_header_nritems(&src->header, src_nritems - push_items); - btrfs_set_header_nritems(&dst->header, dst_nritems + push_items); - btrfs_mark_buffer_dirty(src_buf); - btrfs_mark_buffer_dirty(dst_buf); + memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0), + btrfs_node_key_ptr_offset(push_items), + (src_nritems - push_items) * + sizeof(struct btrfs_key_ptr)); + } + btrfs_set_header_nritems(src, src_nritems - push_items); + btrfs_set_header_nritems(dst, dst_nritems + push_items); + btrfs_mark_buffer_dirty(src); + btrfs_mark_buffer_dirty(dst); return ret; } @@ -1058,24 +1129,22 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root * * this will only push up to 1/2 the contents of the left node over */ -static int balance_node_right(struct btrfs_trans_handle *trans, struct - btrfs_root *root, struct buffer_head *dst_buf, - struct buffer_head *src_buf) +static int balance_node_right(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *dst, + struct extent_buffer *src) { - struct btrfs_node *src = btrfs_buffer_node(src_buf); - struct btrfs_node *dst = btrfs_buffer_node(dst_buf); int push_items = 0; int max_push; int src_nritems; int dst_nritems; int ret = 0; - src_nritems = btrfs_header_nritems(&src->header); - dst_nritems = btrfs_header_nritems(&dst->header); + src_nritems = btrfs_header_nritems(src); + dst_nritems = btrfs_header_nritems(dst); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; - if (push_items <= 0) { + if (push_items <= 0) return 1; - } max_push = src_nritems / 2 + 1; /* don't try to empty the node */ @@ -1085,18 +1154,21 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct if (max_push < push_items) push_items = max_push; - btrfs_memmove(root, dst, dst->ptrs + push_items, dst->ptrs, - dst_nritems * sizeof(struct btrfs_key_ptr)); + memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items), + btrfs_node_key_ptr_offset(0), + (dst_nritems) * + sizeof(struct btrfs_key_ptr)); - btrfs_memcpy(root, dst, dst->ptrs, - src->ptrs + src_nritems - push_items, - push_items * sizeof(struct btrfs_key_ptr)); + copy_extent_buffer(dst, src, + btrfs_node_key_ptr_offset(0), + btrfs_node_key_ptr_offset(src_nritems - push_items), + push_items * sizeof(struct btrfs_key_ptr)); - btrfs_set_header_nritems(&src->header, src_nritems - push_items); - btrfs_set_header_nritems(&dst->header, dst_nritems + push_items); + btrfs_set_header_nritems(src, src_nritems - push_items); + btrfs_set_header_nritems(dst, dst_nritems + push_items); - btrfs_mark_buffer_dirty(src_buf); - btrfs_mark_buffer_dirty(dst_buf); + btrfs_mark_buffer_dirty(src); + btrfs_mark_buffer_dirty(dst); return ret; } @@ -1107,45 +1179,46 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct * * returns zero on success or < 0 on failure. */ -static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, int level) +static int insert_new_root(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int level) { - struct buffer_head *t; - struct btrfs_node *lower; - struct btrfs_node *c; - struct btrfs_disk_key *lower_key; + struct extent_buffer *lower; + struct extent_buffer *c; + struct btrfs_disk_key lower_key; BUG_ON(path->nodes[level]); BUG_ON(path->nodes[level-1] != root->node); - t = btrfs_alloc_free_block(trans, root, root->node->b_blocknr, 0); - if (IS_ERR(t)) - return PTR_ERR(t); - c = btrfs_buffer_node(t); - memset(c, 0, root->blocksize); - btrfs_set_header_nritems(&c->header, 1); - btrfs_set_header_level(&c->header, level); - btrfs_set_header_blocknr(&c->header, bh_blocknr(t)); - btrfs_set_header_generation(&c->header, trans->transid); - btrfs_set_header_owner(&c->header, root->root_key.objectid); - lower = btrfs_buffer_node(path->nodes[level-1]); - memcpy(c->header.fsid, root->fs_info->disk_super->fsid, - sizeof(c->header.fsid)); - if (btrfs_is_leaf(lower)) - lower_key = &((struct btrfs_leaf *)lower)->items[0].key; + c = btrfs_alloc_free_block(trans, root, + extent_buffer_blocknr(root->node), 0); + if (IS_ERR(c)) + return PTR_ERR(c); + memset_extent_buffer(c, 0, 0, root->nodesize); + btrfs_set_header_nritems(c, 1); + btrfs_set_header_level(c, level); + btrfs_set_header_blocknr(c, extent_buffer_blocknr(c)); + btrfs_set_header_generation(c, trans->transid); + btrfs_set_header_owner(c, root->root_key.objectid); + lower = path->nodes[level-1]; + + write_extent_buffer(c, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(c), + BTRFS_FSID_SIZE); + if (level == 1) + btrfs_item_key(lower, &lower_key, 0); else - lower_key = &lower->ptrs[0].key; - btrfs_memcpy(root, c, &c->ptrs[0].key, lower_key, - sizeof(struct btrfs_disk_key)); - btrfs_set_node_blockptr(c, 0, bh_blocknr(path->nodes[level - 1])); + btrfs_node_key(lower, &lower_key, 0); + btrfs_set_node_key(c, &lower_key, 0); + btrfs_set_node_blockptr(c, 0, extent_buffer_blocknr(lower)); - btrfs_mark_buffer_dirty(t); + btrfs_mark_buffer_dirty(c); /* the super has an extra ref to root->node */ - btrfs_block_release(root, root->node); - root->node = t; - get_bh(t); - path->nodes[level] = t; + free_extent_buffer(root->node); + root->node = c; + extent_buffer_get(c); + path->nodes[level] = c; path->slots[level] = 0; return 0; } @@ -1163,26 +1236,26 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_disk_key *key, u64 blocknr, int slot, int level) { - struct btrfs_node *lower; + struct extent_buffer *lower; int nritems; BUG_ON(!path->nodes[level]); - lower = btrfs_buffer_node(path->nodes[level]); - nritems = btrfs_header_nritems(&lower->header); + lower = path->nodes[level]; + nritems = btrfs_header_nritems(lower); if (slot > nritems) BUG(); if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root)) BUG(); if (slot != nritems) { - btrfs_memmove(root, lower, lower->ptrs + slot + 1, - lower->ptrs + slot, + memmove_extent_buffer(lower, + btrfs_node_key_ptr_offset(slot + 1), + btrfs_node_key_ptr_offset(slot), (nritems - slot) * sizeof(struct btrfs_key_ptr)); } - btrfs_memcpy(root, lower, &lower->ptrs[slot].key, - key, sizeof(struct btrfs_disk_key)); + btrfs_set_node_key(lower, key, slot); btrfs_set_node_blockptr(lower, slot, blocknr); - btrfs_set_header_nritems(&lower->header, nritems + 1); - btrfs_mark_buffer_dirty(path->nodes[level]); + btrfs_set_header_nritems(lower, nritems + 1); + btrfs_mark_buffer_dirty(lower); check_node(root, path, level); return 0; } @@ -1199,69 +1272,73 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { - struct buffer_head *t; - struct btrfs_node *c; - struct buffer_head *split_buffer; - struct btrfs_node *split; + struct extent_buffer *c; + struct extent_buffer *split; + struct btrfs_disk_key disk_key; int mid; int ret; int wret; u32 c_nritems; - t = path->nodes[level]; - c = btrfs_buffer_node(t); - if (t == root->node) { + c = path->nodes[level]; + if (c == root->node) { /* trying to split the root, lets make a new one */ ret = insert_new_root(trans, root, path, level + 1); if (ret) return ret; } else { ret = push_nodes_for_insert(trans, root, path, level); - t = path->nodes[level]; - c = btrfs_buffer_node(t); - if (!ret && - btrfs_header_nritems(&c->header) < + c = path->nodes[level]; + if (!ret && btrfs_header_nritems(c) < BTRFS_NODEPTRS_PER_BLOCK(root) - 1) return 0; if (ret < 0) return ret; } - c_nritems = btrfs_header_nritems(&c->header); - split_buffer = btrfs_alloc_free_block(trans, root, t->b_blocknr, 0); - if (IS_ERR(split_buffer)) - return PTR_ERR(split_buffer); + c_nritems = btrfs_header_nritems(c); + split = btrfs_alloc_free_block(trans, root, + extent_buffer_blocknr(c), 0); + if (IS_ERR(split)) + return PTR_ERR(split); + + btrfs_set_header_flags(split, btrfs_header_flags(c)); + btrfs_set_header_level(split, btrfs_header_level(c)); + btrfs_set_header_blocknr(split, extent_buffer_blocknr(split)); + btrfs_set_header_generation(split, trans->transid); + btrfs_set_header_owner(split, root->root_key.objectid); + write_extent_buffer(split, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(split), + BTRFS_FSID_SIZE); - split = btrfs_buffer_node(split_buffer); - btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header)); - btrfs_set_header_level(&split->header, btrfs_header_level(&c->header)); - btrfs_set_header_blocknr(&split->header, bh_blocknr(split_buffer)); - btrfs_set_header_generation(&split->header, trans->transid); - btrfs_set_header_owner(&split->header, root->root_key.objectid); - memcpy(split->header.fsid, root->fs_info->disk_super->fsid, - sizeof(split->header.fsid)); mid = (c_nritems + 1) / 2; - btrfs_memcpy(root, split, split->ptrs, c->ptrs + mid, - (c_nritems - mid) * sizeof(struct btrfs_key_ptr)); - btrfs_set_header_nritems(&split->header, c_nritems - mid); - btrfs_set_header_nritems(&c->header, mid); + + copy_extent_buffer(split, c, + btrfs_node_key_ptr_offset(0), + btrfs_node_key_ptr_offset(mid), + (c_nritems - mid) * sizeof(struct btrfs_key_ptr)); + btrfs_set_header_nritems(split, c_nritems - mid); + btrfs_set_header_nritems(c, mid); ret = 0; - btrfs_mark_buffer_dirty(t); - btrfs_mark_buffer_dirty(split_buffer); - wret = insert_ptr(trans, root, path, &split->ptrs[0].key, - bh_blocknr(split_buffer), path->slots[level + 1] + 1, + btrfs_mark_buffer_dirty(c); + btrfs_mark_buffer_dirty(split); + + btrfs_node_key(split, &disk_key, 0); + wret = insert_ptr(trans, root, path, &disk_key, + extent_buffer_blocknr(split), + path->slots[level + 1] + 1, level + 1); if (wret) ret = wret; if (path->slots[level] >= mid) { path->slots[level] -= mid; - btrfs_block_release(root, t); - path->nodes[level] = split_buffer; + free_extent_buffer(c); + path->nodes[level] = split; path->slots[level + 1] += 1; } else { - btrfs_block_release(root, split_buffer); + free_extent_buffer(split); } return ret; } @@ -1271,16 +1348,16 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root * and nr indicate which items in the leaf to check. This totals up the * space used both by the item structs and the item data */ -static int leaf_space_used(struct btrfs_leaf *l, int start, int nr) +static int leaf_space_used(struct extent_buffer *l, int start, int nr) { int data_len; - int nritems = btrfs_header_nritems(&l->header); + int nritems = btrfs_header_nritems(l); int end = min(nritems, start + nr) - 1; if (!nr) return 0; - data_len = btrfs_item_end(l->items + start); - data_len = data_len - btrfs_item_offset(l->items + end); + data_len = btrfs_item_end_nr(l, start); + data_len = data_len - btrfs_item_offset_nr(l, end); data_len += sizeof(struct btrfs_item) * nr; WARN_ON(data_len < 0); return data_len; @@ -1291,10 +1368,17 @@ static int leaf_space_used(struct btrfs_leaf *l, int start, int nr) * the start of the leaf data. IOW, how much room * the leaf has left for both items and data */ -int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf) +int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf) { - int nritems = btrfs_header_nritems(&leaf->header); - return BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems); + int nritems = btrfs_header_nritems(leaf); + int ret; + ret = BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems); + if (ret < 0) { + printk("leaf free space ret %d, leaf data size %lu, used %d nritems %d\n", + ret, BTRFS_LEAF_DATA_SIZE(root), + leaf_space_used(leaf, 0, nritems), nritems); + } + return ret; } /* @@ -1307,12 +1391,10 @@ int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf) static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int data_size) { - struct buffer_head *left_buf = path->nodes[0]; - struct btrfs_leaf *left = btrfs_buffer_leaf(left_buf); - struct btrfs_leaf *right; - struct buffer_head *right_buf; - struct buffer_head *upper; - struct btrfs_node *upper_node; + struct extent_buffer *left = path->nodes[0]; + struct extent_buffer *right; + struct extent_buffer *upper; + struct btrfs_disk_key disk_key; int slot; int i; int free_space; @@ -1321,6 +1403,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_item *item; u32 left_nritems; u32 right_nritems; + u32 data_end; int ret; slot = path->slots[1]; @@ -1328,102 +1411,109 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root return 1; } upper = path->nodes[1]; - upper_node = btrfs_buffer_node(upper); - if (slot >= btrfs_header_nritems(&upper_node->header) - 1) { + if (slot >= btrfs_header_nritems(upper) - 1) return 1; - } - right_buf = read_tree_block(root, - btrfs_node_blockptr(btrfs_buffer_node(upper), slot + 1)); - right = btrfs_buffer_leaf(right_buf); + + right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1)); free_space = btrfs_leaf_free_space(root, right); if (free_space < data_size + sizeof(struct btrfs_item)) { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); return 1; } + /* cow and double check */ - ret = btrfs_cow_block(trans, root, right_buf, upper, - slot + 1, &right_buf); + ret = btrfs_cow_block(trans, root, right, upper, + slot + 1, &right); if (ret) { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); return 1; } - right = btrfs_buffer_leaf(right_buf); free_space = btrfs_leaf_free_space(root, right); if (free_space < data_size + sizeof(struct btrfs_item)) { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); return 1; } - left_nritems = btrfs_header_nritems(&left->header); + left_nritems = btrfs_header_nritems(left); if (left_nritems == 0) { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); return 1; } + for (i = left_nritems - 1; i >= 1; i--) { - item = left->items + i; + item = btrfs_item_nr(left, i); if (path->slots[0] == i) push_space += data_size + sizeof(*item); - if (btrfs_item_size(item) + sizeof(*item) + push_space > + if (btrfs_item_size(left, item) + sizeof(*item) + push_space > free_space) break; push_items++; - push_space += btrfs_item_size(item) + sizeof(*item); + push_space += btrfs_item_size(left, item) + sizeof(*item); } + if (push_items == 0) { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); return 1; } + if (push_items == left_nritems) WARN_ON(1); - right_nritems = btrfs_header_nritems(&right->header); + /* push left to right */ - push_space = btrfs_item_end(left->items + left_nritems - push_items); + right_nritems = btrfs_header_nritems(right); + push_space = btrfs_item_end_nr(left, left_nritems - push_items); push_space -= leaf_data_end(root, left); + /* make room in the right data area */ - btrfs_memmove(root, right, btrfs_leaf_data(right) + - leaf_data_end(root, right) - push_space, - btrfs_leaf_data(right) + - leaf_data_end(root, right), BTRFS_LEAF_DATA_SIZE(root) - - leaf_data_end(root, right)); + data_end = leaf_data_end(root, right); + memmove_extent_buffer(right, + btrfs_leaf_data(right) + data_end - push_space, + btrfs_leaf_data(right) + data_end, + BTRFS_LEAF_DATA_SIZE(root) - data_end); + /* copy from the left data area */ - btrfs_memcpy(root, right, btrfs_leaf_data(right) + + copy_extent_buffer(right, left, btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) - push_space, btrfs_leaf_data(left) + leaf_data_end(root, left), push_space); - btrfs_memmove(root, right, right->items + push_items, right->items, - right_nritems * sizeof(struct btrfs_item)); + + memmove_extent_buffer(right, btrfs_item_nr_offset(push_items), + btrfs_item_nr_offset(0), + right_nritems * sizeof(struct btrfs_item)); + /* copy the items from left to right */ - btrfs_memcpy(root, right, right->items, left->items + - left_nritems - push_items, - push_items * sizeof(struct btrfs_item)); + copy_extent_buffer(right, left, btrfs_item_nr_offset(0), + btrfs_item_nr_offset(left_nritems - push_items), + push_items * sizeof(struct btrfs_item)); /* update the item pointers */ right_nritems += push_items; - btrfs_set_header_nritems(&right->header, right_nritems); + btrfs_set_header_nritems(right, right_nritems); push_space = BTRFS_LEAF_DATA_SIZE(root); for (i = 0; i < right_nritems; i++) { - btrfs_set_item_offset(right->items + i, push_space - - btrfs_item_size(right->items + i)); - push_space = btrfs_item_offset(right->items + i); + item = btrfs_item_nr(right, i); + btrfs_set_item_offset(right, item, push_space - + btrfs_item_size(right, item)); + push_space = btrfs_item_offset(right, item); } left_nritems -= push_items; - btrfs_set_header_nritems(&left->header, left_nritems); + btrfs_set_header_nritems(left, left_nritems); - btrfs_mark_buffer_dirty(left_buf); - btrfs_mark_buffer_dirty(right_buf); + btrfs_mark_buffer_dirty(left); + btrfs_mark_buffer_dirty(right); - btrfs_memcpy(root, upper_node, &upper_node->ptrs[slot + 1].key, - &right->items[0].key, sizeof(struct btrfs_disk_key)); + btrfs_item_key(right, &disk_key, 0); + btrfs_set_node_key(upper, &disk_key, slot + 1); btrfs_mark_buffer_dirty(upper); /* then fixup the leaf pointer in the path */ if (path->slots[0] >= left_nritems) { path->slots[0] -= left_nritems; - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = right_buf; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = right; path->slots[1] += 1; } else { - btrfs_block_release(root, right_buf); + free_extent_buffer(right); } if (path->nodes[1]) check_node(root, path, 1); @@ -1436,10 +1526,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int data_size) { - struct buffer_head *right_buf = path->nodes[0]; - struct btrfs_leaf *right = btrfs_buffer_leaf(right_buf); - struct buffer_head *t; - struct btrfs_leaf *left; + struct btrfs_disk_key disk_key; + struct extent_buffer *right = path->nodes[0]; + struct extent_buffer *left; int slot; int i; int free_space; @@ -1447,119 +1536,128 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root int push_items = 0; struct btrfs_item *item; u32 old_left_nritems; + u32 right_nritems; int ret = 0; int wret; slot = path->slots[1]; - if (slot == 0) { + if (slot == 0) return 1; - } - if (!path->nodes[1]) { + if (!path->nodes[1]) return 1; - } - t = read_tree_block(root, - btrfs_node_blockptr(btrfs_buffer_node(path->nodes[1]), slot - 1)); - left = btrfs_buffer_leaf(t); + + left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1], + slot - 1)); free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { - btrfs_block_release(root, t); + free_extent_buffer(left); return 1; } /* cow and double check */ - ret = btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t); + ret = btrfs_cow_block(trans, root, left, + path->nodes[1], slot - 1, &left); if (ret) { /* we hit -ENOSPC, but it isn't fatal here */ - btrfs_block_release(root, t); + free_extent_buffer(left); return 1; } - left = btrfs_buffer_leaf(t); free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { - btrfs_block_release(root, t); + free_extent_buffer(left); return 1; } - if (btrfs_header_nritems(&right->header) == 0) { - btrfs_block_release(root, t); + right_nritems = btrfs_header_nritems(right); + if (right_nritems == 0) { + free_extent_buffer(left); return 1; } - for (i = 0; i < btrfs_header_nritems(&right->header) - 1; i++) { - item = right->items + i; + for (i = 0; i < right_nritems - 1; i++) { + item = btrfs_item_nr(right, i); if (path->slots[0] == i) push_space += data_size + sizeof(*item); - if (btrfs_item_size(item) + sizeof(*item) + push_space > + if (btrfs_item_size(right, item) + sizeof(*item) + push_space > free_space) break; push_items++; - push_space += btrfs_item_size(item) + sizeof(*item); + push_space += btrfs_item_size(right, item) + sizeof(*item); } if (push_items == 0) { - btrfs_block_release(root, t); + free_extent_buffer(left); return 1; } - if (push_items == btrfs_header_nritems(&right->header)) + if (push_items == btrfs_header_nritems(right)) WARN_ON(1); + /* push data from right to left */ - btrfs_memcpy(root, left, left->items + - btrfs_header_nritems(&left->header), - right->items, push_items * sizeof(struct btrfs_item)); + copy_extent_buffer(left, right, + btrfs_item_nr_offset(btrfs_header_nritems(left)), + btrfs_item_nr_offset(0), + push_items * sizeof(struct btrfs_item)); + push_space = BTRFS_LEAF_DATA_SIZE(root) - - btrfs_item_offset(right->items + push_items -1); - btrfs_memcpy(root, left, btrfs_leaf_data(left) + + btrfs_item_offset_nr(right, push_items -1); + + copy_extent_buffer(left, right, btrfs_leaf_data(left) + leaf_data_end(root, left) - push_space, btrfs_leaf_data(right) + - btrfs_item_offset(right->items + push_items - 1), + btrfs_item_offset_nr(right, push_items - 1), push_space); - old_left_nritems = btrfs_header_nritems(&left->header); + old_left_nritems = btrfs_header_nritems(left); BUG_ON(old_left_nritems < 0); for (i = old_left_nritems; i < old_left_nritems + push_items; i++) { - u32 ioff = btrfs_item_offset(left->items + i); - btrfs_set_item_offset(left->items + i, ioff - - (BTRFS_LEAF_DATA_SIZE(root) - - btrfs_item_offset(left->items + - old_left_nritems - 1))); + u32 ioff; + item = btrfs_item_nr(left, i); + ioff = btrfs_item_offset(left, item); + btrfs_set_item_offset(left, item, + ioff - (BTRFS_LEAF_DATA_SIZE(root) - + btrfs_item_offset_nr(left, old_left_nritems - 1))); } - btrfs_set_header_nritems(&left->header, old_left_nritems + push_items); + btrfs_set_header_nritems(left, old_left_nritems + push_items); /* fixup right node */ - push_space = btrfs_item_offset(right->items + push_items - 1) - - leaf_data_end(root, right); - btrfs_memmove(root, right, btrfs_leaf_data(right) + - BTRFS_LEAF_DATA_SIZE(root) - push_space, - btrfs_leaf_data(right) + - leaf_data_end(root, right), push_space); - btrfs_memmove(root, right, right->items, right->items + push_items, - (btrfs_header_nritems(&right->header) - push_items) * - sizeof(struct btrfs_item)); - btrfs_set_header_nritems(&right->header, - btrfs_header_nritems(&right->header) - - push_items); + push_space = btrfs_item_offset_nr(right, push_items - 1) - + leaf_data_end(root, right); + memmove_extent_buffer(right, btrfs_leaf_data(right) + + BTRFS_LEAF_DATA_SIZE(root) - push_space, + btrfs_leaf_data(right) + + leaf_data_end(root, right), push_space); + + memmove_extent_buffer(right, btrfs_item_nr_offset(0), + btrfs_item_nr_offset(push_items), + (btrfs_header_nritems(right) - push_items) * + sizeof(struct btrfs_item)); + + right_nritems = btrfs_header_nritems(right) - push_items; + btrfs_set_header_nritems(right, right_nritems); push_space = BTRFS_LEAF_DATA_SIZE(root); - for (i = 0; i < btrfs_header_nritems(&right->header); i++) { - btrfs_set_item_offset(right->items + i, push_space - - btrfs_item_size(right->items + i)); - push_space = btrfs_item_offset(right->items + i); + for (i = 0; i < right_nritems; i++) { + item = btrfs_item_nr(right, i); + btrfs_set_item_offset(right, item, push_space - + btrfs_item_size(right, item)); + push_space = btrfs_item_offset(right, item); } - btrfs_mark_buffer_dirty(t); - btrfs_mark_buffer_dirty(right_buf); + btrfs_mark_buffer_dirty(left); + btrfs_mark_buffer_dirty(right); - wret = fixup_low_keys(trans, root, path, &right->items[0].key, 1); + btrfs_item_key(right, &disk_key, 0); + wret = fixup_low_keys(trans, root, path, &disk_key, 1); if (wret) ret = wret; /* then fixup the leaf pointer in the path */ if (path->slots[0] < push_items) { path->slots[0] += old_left_nritems; - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = t; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = left; path->slots[1] -= 1; } else { - btrfs_block_release(root, t); + free_extent_buffer(left); path->slots[0] -= push_items; } BUG_ON(path->slots[0] < 0); @@ -1578,13 +1676,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *ins_key, struct btrfs_path *path, int data_size) { - struct buffer_head *l_buf; - struct btrfs_leaf *l; + struct extent_buffer *l; u32 nritems; int mid; int slot; - struct btrfs_leaf *right; - struct buffer_head *right_buffer; + struct extent_buffer *right; int space_needed = data_size + sizeof(struct btrfs_item); int data_copy_size; int rt_data_off; @@ -1603,8 +1699,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root if (wret < 0) return wret; } - l_buf = path->nodes[0]; - l = btrfs_buffer_leaf(l_buf); + l = path->nodes[0]; /* did the pushes work? */ if (btrfs_leaf_free_space(root, l) >= @@ -1617,36 +1712,38 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root return ret; } slot = path->slots[0]; - nritems = btrfs_header_nritems(&l->header); + nritems = btrfs_header_nritems(l); mid = (nritems + 1)/ 2; - right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0); - if (IS_ERR(right_buffer)) - return PTR_ERR(right_buffer); - - right = btrfs_buffer_leaf(right_buffer); - memset(&right->header, 0, sizeof(right->header)); - btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer)); - btrfs_set_header_generation(&right->header, trans->transid); - btrfs_set_header_owner(&right->header, root->root_key.objectid); - btrfs_set_header_level(&right->header, 0); - memcpy(right->header.fsid, root->fs_info->disk_super->fsid, - sizeof(right->header.fsid)); + right = btrfs_alloc_free_block(trans, root, + extent_buffer_blocknr(l), 0); + if (IS_ERR(right)) + return PTR_ERR(right); + + memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header)); + btrfs_set_header_blocknr(right, extent_buffer_blocknr(right)); + btrfs_set_header_generation(right, trans->transid); + btrfs_set_header_owner(right, root->root_key.objectid); + btrfs_set_header_level(right, 0); + write_extent_buffer(right, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(right), + BTRFS_FSID_SIZE); + if (mid <= slot) { if (nritems == 1 || leaf_space_used(l, mid, nritems - mid) + space_needed > BTRFS_LEAF_DATA_SIZE(root)) { if (slot >= nritems) { btrfs_cpu_key_to_disk(&disk_key, ins_key); - btrfs_set_header_nritems(&right->header, 0); + btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, &disk_key, - bh_blocknr(right_buffer), + extent_buffer_blocknr(right), path->slots[1] + 1, 1); if (wret) ret = wret; - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = right_buffer; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = right; path->slots[0] = 0; path->slots[1] += 1; return ret; @@ -1659,15 +1756,15 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root BTRFS_LEAF_DATA_SIZE(root)) { if (slot == 0) { btrfs_cpu_key_to_disk(&disk_key, ins_key); - btrfs_set_header_nritems(&right->header, 0); + btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, &disk_key, - bh_blocknr(right_buffer), + extent_buffer_blocknr(right), path->slots[1], 1); if (wret) ret = wret; - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = right_buffer; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = right; path->slots[0] = 0; if (path->slots[1] == 0) { wret = fixup_low_keys(trans, root, @@ -1681,61 +1778,74 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root double_split = 1; } } - btrfs_set_header_nritems(&right->header, nritems - mid); - data_copy_size = btrfs_item_end(l->items + mid) - - leaf_data_end(root, l); - btrfs_memcpy(root, right, right->items, l->items + mid, - (nritems - mid) * sizeof(struct btrfs_item)); - btrfs_memcpy(root, right, + nritems = nritems - mid; + btrfs_set_header_nritems(right, nritems); + data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(root, l); + + copy_extent_buffer(right, l, btrfs_item_nr_offset(0), + btrfs_item_nr_offset(mid), + nritems * sizeof(struct btrfs_item)); + + copy_extent_buffer(right, l, btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) - data_copy_size, btrfs_leaf_data(l) + leaf_data_end(root, l), data_copy_size); + rt_data_off = BTRFS_LEAF_DATA_SIZE(root) - - btrfs_item_end(l->items + mid); + btrfs_item_end_nr(l, mid); - for (i = 0; i < btrfs_header_nritems(&right->header); i++) { - u32 ioff = btrfs_item_offset(right->items + i); - btrfs_set_item_offset(right->items + i, ioff + rt_data_off); + for (i = 0; i < nritems; i++) { + struct btrfs_item *item = btrfs_item_nr(right, i); + u32 ioff = btrfs_item_offset(right, item); + btrfs_set_item_offset(right, item, ioff + rt_data_off); } - btrfs_set_header_nritems(&l->header, mid); + btrfs_set_header_nritems(l, mid); ret = 0; - wret = insert_ptr(trans, root, path, &right->items[0].key, - bh_blocknr(right_buffer), path->slots[1] + 1, 1); + btrfs_item_key(right, &disk_key, 0); + wret = insert_ptr(trans, root, path, &disk_key, + extent_buffer_blocknr(right), path->slots[1] + 1, 1); if (wret) ret = wret; - btrfs_mark_buffer_dirty(right_buffer); - btrfs_mark_buffer_dirty(l_buf); + + btrfs_mark_buffer_dirty(right); + btrfs_mark_buffer_dirty(l); BUG_ON(path->slots[0] != slot); + if (mid <= slot) { - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = right_buffer; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = right; path->slots[0] -= mid; path->slots[1] += 1; } else - btrfs_block_release(root, right_buffer); + free_extent_buffer(right); + BUG_ON(path->slots[0] < 0); check_node(root, path, 1); + check_leaf(root, path, 0); if (!double_split) return ret; - right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0); - if (IS_ERR(right_buffer)) - return PTR_ERR(right_buffer); - - right = btrfs_buffer_leaf(right_buffer); - memset(&right->header, 0, sizeof(right->header)); - btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer)); - btrfs_set_header_generation(&right->header, trans->transid); - btrfs_set_header_owner(&right->header, root->root_key.objectid); - btrfs_set_header_level(&right->header, 0); - memcpy(right->header.fsid, root->fs_info->disk_super->fsid, - sizeof(right->header.fsid)); + + right = btrfs_alloc_free_block(trans, root, + extent_buffer_blocknr(l), 0); + if (IS_ERR(right)) + return PTR_ERR(right); + + memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header)); + btrfs_set_header_blocknr(right, extent_buffer_blocknr(right)); + btrfs_set_header_generation(right, trans->transid); + btrfs_set_header_owner(right, root->root_key.objectid); + btrfs_set_header_level(right, 0); + write_extent_buffer(right, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(right), + BTRFS_FSID_SIZE); + btrfs_cpu_key_to_disk(&disk_key, ins_key); - btrfs_set_header_nritems(&right->header, 0); + btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, &disk_key, - bh_blocknr(right_buffer), + extent_buffer_blocknr(right), path->slots[1], 1); if (wret) ret = wret; @@ -1744,8 +1854,8 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root if (wret) ret = wret; } - btrfs_block_release(root, path->nodes[0]); - path->nodes[0] = right_buffer; + free_extent_buffer(path->nodes[0]); + path->nodes[0] = right; path->slots[0] = 0; check_node(root, path, 1); check_leaf(root, path, 0); @@ -1760,8 +1870,8 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, int ret = 0; int slot; int slot_orig; - struct btrfs_leaf *leaf; - struct buffer_head *leaf_buf; + struct extent_buffer *leaf; + struct btrfs_item *item; u32 nritems; unsigned int data_end; unsigned int old_data_start; @@ -1770,15 +1880,14 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, int i; slot_orig = path->slots[0]; - leaf_buf = path->nodes[0]; - leaf = btrfs_buffer_leaf(leaf_buf); + leaf = path->nodes[0]; - nritems = btrfs_header_nritems(&leaf->header); + nritems = btrfs_header_nritems(leaf); data_end = leaf_data_end(root, leaf); slot = path->slots[0]; - old_data_start = btrfs_item_offset(leaf->items + slot); - old_size = btrfs_item_size(leaf->items + slot); + old_data_start = btrfs_item_offset_nr(leaf, slot); + old_size = btrfs_item_size_nr(leaf, slot); BUG_ON(old_size <= new_size); size_diff = old_size - new_size; @@ -1790,32 +1899,38 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, */ /* first correct the data pointers */ for (i = slot; i < nritems; i++) { - u32 ioff = btrfs_item_offset(leaf->items + i); - btrfs_set_item_offset(leaf->items + i, - ioff + size_diff); + u32 ioff; + item = btrfs_item_nr(leaf, i); + ioff = btrfs_item_offset(leaf, item); + btrfs_set_item_offset(leaf, item, ioff + size_diff); } /* shift the data */ - btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) + + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end + size_diff, btrfs_leaf_data(leaf) + data_end, old_data_start + new_size - data_end); - btrfs_set_item_size(leaf->items + slot, new_size); - btrfs_mark_buffer_dirty(leaf_buf); + + item = btrfs_item_nr(leaf, slot); + btrfs_set_item_size(leaf, item, new_size); + btrfs_mark_buffer_dirty(leaf); ret = 0; - if (btrfs_leaf_free_space(root, leaf) < 0) + if (btrfs_leaf_free_space(root, leaf) < 0) { + btrfs_print_leaf(root, leaf); BUG(); + } check_leaf(root, path, 0); return ret; } -int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, u32 data_size) +int btrfs_extend_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_path *path, + u32 data_size) { int ret = 0; int slot; int slot_orig; - struct btrfs_leaf *leaf; - struct buffer_head *leaf_buf; + struct extent_buffer *leaf; + struct btrfs_item *item; u32 nritems; unsigned int data_end; unsigned int old_data; @@ -1823,16 +1938,17 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root int i; slot_orig = path->slots[0]; - leaf_buf = path->nodes[0]; - leaf = btrfs_buffer_leaf(leaf_buf); + leaf = path->nodes[0]; - nritems = btrfs_header_nritems(&leaf->header); + nritems = btrfs_header_nritems(leaf); data_end = leaf_data_end(root, leaf); - if (btrfs_leaf_free_space(root, leaf) < data_size) + if (btrfs_leaf_free_space(root, leaf) < data_size) { + btrfs_print_leaf(root, leaf); BUG(); + } slot = path->slots[0]; - old_data = btrfs_item_end(leaf->items + slot); + old_data = btrfs_item_end_nr(leaf, slot); BUG_ON(slot < 0); BUG_ON(slot >= nritems); @@ -1842,22 +1958,28 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root */ /* first correct the data pointers */ for (i = slot; i < nritems; i++) { - u32 ioff = btrfs_item_offset(leaf->items + i); - btrfs_set_item_offset(leaf->items + i, - ioff - data_size); + u32 ioff; + item = btrfs_item_nr(leaf, i); + ioff = btrfs_item_offset(leaf, item); + btrfs_set_item_offset(leaf, item, ioff - data_size); } + /* shift the data */ - btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) + + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end - data_size, btrfs_leaf_data(leaf) + data_end, old_data - data_end); + data_end = old_data; - old_size = btrfs_item_size(leaf->items + slot); - btrfs_set_item_size(leaf->items + slot, old_size + data_size); - btrfs_mark_buffer_dirty(leaf_buf); + old_size = btrfs_item_size_nr(leaf, slot); + item = btrfs_item_nr(leaf, slot); + btrfs_set_item_size(leaf, item, old_size + data_size); + btrfs_mark_buffer_dirty(leaf); ret = 0; - if (btrfs_leaf_free_space(root, leaf) < 0) + if (btrfs_leaf_free_space(root, leaf) < 0) { + btrfs_print_leaf(root, leaf); BUG(); + } check_leaf(root, path, 0); return ret; } @@ -1866,15 +1988,16 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root * Given a key and some data, insert an item into the tree. * This does all the path init required, making room in the tree if needed. */ -int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, struct btrfs_key - *cpu_key, u32 data_size) +int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct btrfs_key *cpu_key, u32 data_size) { + struct extent_buffer *leaf; + struct btrfs_item *item; int ret = 0; int slot; int slot_orig; - struct btrfs_leaf *leaf; - struct buffer_head *leaf_buf; u32 nritems; unsigned int data_end; struct btrfs_disk_key disk_key; @@ -1884,6 +2007,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root /* create a root if there isn't one */ if (!root->node) BUG(); + ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1); if (ret == 0) { return -EEXIST; @@ -1892,57 +2016,68 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root goto out; slot_orig = path->slots[0]; - leaf_buf = path->nodes[0]; - leaf = btrfs_buffer_leaf(leaf_buf); + leaf = path->nodes[0]; - nritems = btrfs_header_nritems(&leaf->header); + nritems = btrfs_header_nritems(leaf); data_end = leaf_data_end(root, leaf); if (btrfs_leaf_free_space(root, leaf) < sizeof(struct btrfs_item) + data_size) { BUG(); } + slot = path->slots[0]; BUG_ON(slot < 0); + if (slot != nritems) { int i; - unsigned int old_data = btrfs_item_end(leaf->items + slot); + unsigned int old_data = btrfs_item_end_nr(leaf, slot); + if (old_data < data_end) { + btrfs_print_leaf(root, leaf); + printk("slot %d old_data %d data_end %d\n", + slot, old_data, data_end); + BUG_ON(1); + } /* * item0..itemN ... dataN.offset..dataN.size .. data0.size */ /* first correct the data pointers */ for (i = slot; i < nritems; i++) { - u32 ioff = btrfs_item_offset(leaf->items + i); - btrfs_set_item_offset(leaf->items + i, - ioff - data_size); + u32 ioff; + item = btrfs_item_nr(leaf, i); + ioff = btrfs_item_offset(leaf, item); + btrfs_set_item_offset(leaf, item, ioff - data_size); } /* shift the items */ - btrfs_memmove(root, leaf, leaf->items + slot + 1, - leaf->items + slot, + memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1), + btrfs_item_nr_offset(slot), (nritems - slot) * sizeof(struct btrfs_item)); /* shift the data */ - btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) + + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end - data_size, btrfs_leaf_data(leaf) + data_end, old_data - data_end); data_end = old_data; } + /* setup the item for the new data */ - btrfs_memcpy(root, leaf, &leaf->items[slot].key, &disk_key, - sizeof(struct btrfs_disk_key)); - btrfs_set_item_offset(leaf->items + slot, data_end - data_size); - btrfs_set_item_size(leaf->items + slot, data_size); - btrfs_set_header_nritems(&leaf->header, nritems + 1); - btrfs_mark_buffer_dirty(leaf_buf); + btrfs_set_item_key(leaf, &disk_key, slot); + item = btrfs_item_nr(leaf, slot); + btrfs_set_item_offset(leaf, item, data_end - data_size); + btrfs_set_item_size(leaf, item, data_size); + btrfs_set_header_nritems(leaf, nritems + 1); + btrfs_mark_buffer_dirty(leaf); ret = 0; if (slot == 0) ret = fixup_low_keys(trans, root, path, &disk_key, 1); - if (btrfs_leaf_free_space(root, leaf) < 0) + if (btrfs_leaf_free_space(root, leaf) < 0) { + btrfs_print_leaf(root, leaf); BUG(); + } check_leaf(root, path, 0); out: return ret; @@ -1958,17 +2093,17 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root { int ret = 0; struct btrfs_path *path; - u8 *ptr; + struct extent_buffer *leaf; + unsigned long ptr; path = btrfs_alloc_path(); BUG_ON(!path); ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size); if (!ret) { - ptr = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], u8); - btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, data, data_size); - btrfs_mark_buffer_dirty(path->nodes[0]); + leaf = path->nodes[0]; + ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); + write_extent_buffer(leaf, data, ptr, data_size); + btrfs_mark_buffer_dirty(leaf); } btrfs_free_path(path); return ret; @@ -1984,30 +2119,30 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level, int slot) { - struct btrfs_node *node; - struct buffer_head *parent = path->nodes[level]; + struct extent_buffer *parent = path->nodes[level]; u32 nritems; int ret = 0; int wret; - node = btrfs_buffer_node(parent); - nritems = btrfs_header_nritems(&node->header); + nritems = btrfs_header_nritems(parent); if (slot != nritems -1) { - btrfs_memmove(root, node, node->ptrs + slot, - node->ptrs + slot + 1, + memmove_extent_buffer(parent, + btrfs_node_key_ptr_offset(slot), + btrfs_node_key_ptr_offset(slot + 1), sizeof(struct btrfs_key_ptr) * (nritems - slot - 1)); } nritems--; - btrfs_set_header_nritems(&node->header, nritems); + btrfs_set_header_nritems(parent, nritems); if (nritems == 0 && parent == root->node) { - struct btrfs_header *header = btrfs_buffer_header(root->node); - BUG_ON(btrfs_header_level(header) != 1); + BUG_ON(btrfs_header_level(root->node) != 1); /* just turn the root into a leaf and break */ - btrfs_set_header_level(header, 0); + btrfs_set_header_level(root->node, 0); } else if (slot == 0) { - wret = fixup_low_keys(trans, root, path, &node->ptrs[0].key, - level + 1); + struct btrfs_disk_key disk_key; + + btrfs_node_key(parent, &disk_key, 0); + wret = fixup_low_keys(trans, root, path, &disk_key, level + 1); if (wret) ret = wret; } @@ -2023,59 +2158,67 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path) { int slot; - struct btrfs_leaf *leaf; - struct buffer_head *leaf_buf; + struct extent_buffer *leaf; + struct btrfs_item *item; int doff; int dsize; int ret = 0; int wret; u32 nritems; - leaf_buf = path->nodes[0]; - leaf = btrfs_buffer_leaf(leaf_buf); + leaf = path->nodes[0]; slot = path->slots[0]; - doff = btrfs_item_offset(leaf->items + slot); - dsize = btrfs_item_size(leaf->items + slot); - nritems = btrfs_header_nritems(&leaf->header); + doff = btrfs_item_offset_nr(leaf, slot); + dsize = btrfs_item_size_nr(leaf, slot); + nritems = btrfs_header_nritems(leaf); if (slot != nritems - 1) { int i; int data_end = leaf_data_end(root, leaf); - btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) + + + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end + dsize, btrfs_leaf_data(leaf) + data_end, doff - data_end); + for (i = slot + 1; i < nritems; i++) { - u32 ioff = btrfs_item_offset(leaf->items + i); - btrfs_set_item_offset(leaf->items + i, ioff + dsize); + u32 ioff; + item = btrfs_item_nr(leaf, i); + ioff = btrfs_item_offset(leaf, item); + btrfs_set_item_offset(leaf, item, ioff + dsize); } - btrfs_memmove(root, leaf, leaf->items + slot, - leaf->items + slot + 1, + memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot), + btrfs_item_nr_offset(slot + 1), sizeof(struct btrfs_item) * (nritems - slot - 1)); } - btrfs_set_header_nritems(&leaf->header, nritems - 1); + btrfs_set_header_nritems(leaf, nritems - 1); nritems--; + /* delete the leaf if we've emptied it */ if (nritems == 0) { - if (leaf_buf == root->node) { - btrfs_set_header_level(&leaf->header, 0); + if (leaf == root->node) { + btrfs_set_header_level(leaf, 0); } else { - clean_tree_block(trans, root, leaf_buf); - wait_on_buffer(leaf_buf); + clean_tree_block(trans, root, leaf); + wait_on_tree_block_writeback(root, leaf); wret = del_ptr(trans, root, path, 1, path->slots[1]); if (wret) ret = wret; wret = btrfs_free_extent(trans, root, - bh_blocknr(leaf_buf), 1, 1); + extent_buffer_blocknr(leaf), + 1, 1); if (wret) ret = wret; } } else { int used = leaf_space_used(leaf, 0, nritems); if (slot == 0) { + struct btrfs_disk_key disk_key; + + btrfs_item_key(leaf, &disk_key, 0); wret = fixup_low_keys(trans, root, path, - &leaf->items[0].key, 1); + &disk_key, 1); if (wret) ret = wret; } @@ -2087,34 +2230,40 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, * for possible call to del_ptr below */ slot = path->slots[1]; - get_bh(leaf_buf); + extent_buffer_get(leaf); + wret = push_leaf_left(trans, root, path, 1); if (wret < 0 && wret != -ENOSPC) ret = wret; - if (path->nodes[0] == leaf_buf && - btrfs_header_nritems(&leaf->header)) { + + if (path->nodes[0] == leaf && + btrfs_header_nritems(leaf)) { wret = push_leaf_right(trans, root, path, 1); if (wret < 0 && wret != -ENOSPC) ret = wret; } - if (btrfs_header_nritems(&leaf->header) == 0) { - u64 blocknr = bh_blocknr(leaf_buf); - clean_tree_block(trans, root, leaf_buf); - wait_on_buffer(leaf_buf); + + if (btrfs_header_nritems(leaf) == 0) { + u64 blocknr = extent_buffer_blocknr(leaf); + + clean_tree_block(trans, root, leaf); + wait_on_tree_block_writeback(root, leaf); + wret = del_ptr(trans, root, path, 1, slot); if (wret) ret = wret; - btrfs_block_release(root, leaf_buf); + + free_extent_buffer(leaf); wret = btrfs_free_extent(trans, root, blocknr, 1, 1); if (wret) ret = wret; } else { - btrfs_mark_buffer_dirty(leaf_buf); - btrfs_block_release(root, leaf_buf); + btrfs_mark_buffer_dirty(leaf); + free_extent_buffer(leaf); } } else { - btrfs_mark_buffer_dirty(leaf_buf); + btrfs_mark_buffer_dirty(leaf); } } return ret; @@ -2130,25 +2279,27 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) int slot; int level = 1; u64 blocknr; - struct buffer_head *c; - struct btrfs_node *c_node; - struct buffer_head *next = NULL; + struct extent_buffer *c; + struct extent_buffer *next = NULL; while(level < BTRFS_MAX_LEVEL) { if (!path->nodes[level]) return 1; + slot = path->slots[level] + 1; c = path->nodes[level]; - c_node = btrfs_buffer_node(c); - if (slot >= btrfs_header_nritems(&c_node->header)) { + if (slot >= btrfs_header_nritems(c)) { level++; continue; } - blocknr = btrfs_node_blockptr(c_node, slot); + + blocknr = btrfs_node_blockptr(c, slot); if (next) - btrfs_block_release(root, next); + free_extent_buffer(next); + if (path->reada) reada_for_search(root, path, level, slot); + next = read_tree_block(root, blocknr); break; } @@ -2156,15 +2307,14 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) while(1) { level--; c = path->nodes[level]; - btrfs_block_release(root, c); + free_extent_buffer(c); path->nodes[level] = next; path->slots[level] = 0; if (!level) break; if (path->reada) reada_for_search(root, path, level, 0); - next = read_tree_block(root, - btrfs_node_blockptr(btrfs_buffer_node(next), 0)); + next = read_tree_block(root, btrfs_node_blockptr(next, 0)); } return 0; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 800a3499cc3..c4b82980685 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -20,10 +20,10 @@ #define __BTRFS__ #include -#include #include #include #include "bit-radix.h" +#include "extent_map.h" struct btrfs_trans_handle; struct btrfs_transaction; @@ -32,7 +32,7 @@ extern struct kmem_cache *btrfs_transaction_cachep; extern struct kmem_cache *btrfs_bit_radix_cachep; extern struct kmem_cache *btrfs_path_cachep; -#define BTRFS_MAGIC "_BtRfS_M" +#define BTRFS_MAGIC "_B2RfS_M" #define BTRFS_ROOT_TREE_OBJECTID 1ULL #define BTRFS_EXTENT_TREE_OBJECTID 2ULL @@ -78,41 +78,41 @@ extern struct kmem_cache *btrfs_path_cachep; */ struct btrfs_disk_key { __le64 objectid; - __le32 flags; + u8 type; __le64 offset; } __attribute__ ((__packed__)); struct btrfs_key { u64 objectid; - u32 flags; + u8 type; u64 offset; } __attribute__ ((__packed__)); +#define BTRFS_FSID_SIZE 16 /* * every tree block (leaf or node) starts with this header. */ struct btrfs_header { u8 csum[BTRFS_CSUM_SIZE]; - u8 fsid[16]; /* FS specific uuid */ + u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ __le64 blocknr; /* which block this node is supposed to live in */ __le64 generation; __le64 owner; - __le16 nritems; + __le32 nritems; __le16 flags; u8 level; } __attribute__ ((__packed__)); #define BTRFS_MAX_LEVEL 8 -#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->blocksize - \ +#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->nodesize - \ sizeof(struct btrfs_header)) / \ (sizeof(struct btrfs_disk_key) + sizeof(u64))) #define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header)) -#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->blocksize)) +#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->leafsize)) #define BTRFS_MAX_INLINE_DATA_SIZE(r) (BTRFS_LEAF_DATA_SIZE(r) - \ sizeof(struct btrfs_item) - \ sizeof(struct btrfs_file_extent_item)) -struct buffer_head; /* * the super block basically lists the main trees of the FS * it currently lacks any block count etc etc @@ -128,7 +128,9 @@ struct btrfs_super_block { __le64 total_blocks; __le64 blocks_used; __le64 root_dir_objectid; - __le32 blocksize; + __le32 sectorsize; + __le32 nodesize; + __le32 leafsize; } __attribute__ ((__packed__)); /* @@ -138,7 +140,7 @@ struct btrfs_super_block { struct btrfs_item { struct btrfs_disk_key key; __le32 offset; - __le16 size; + __le32 size; } __attribute__ ((__packed__)); /* @@ -176,7 +178,7 @@ struct btrfs_node { * used while walking the tree. */ struct btrfs_path { - struct buffer_head *nodes[BTRFS_MAX_LEVEL]; + struct extent_buffer *nodes[BTRFS_MAX_LEVEL]; int slots[BTRFS_MAX_LEVEL]; int reada; int lowest_level; @@ -292,6 +294,7 @@ struct btrfs_block_group_cache { }; struct btrfs_fs_info { + u8 fsid[BTRFS_FSID_SIZE]; struct btrfs_root *extent_root; struct btrfs_root *tree_root; struct radix_tree_root fs_roots_radix; @@ -304,9 +307,8 @@ struct btrfs_fs_info { u64 generation; u64 last_trans_committed; struct btrfs_transaction *running_transaction; - struct btrfs_super_block *disk_super; struct btrfs_super_block super_copy; - struct buffer_head *sb_buffer; + struct extent_buffer *sb_buffer; struct super_block *sb; struct inode *btree_inode; struct mutex trans_mutex; @@ -325,8 +327,8 @@ struct btrfs_fs_info { * and for the extent tree extent_root root. */ struct btrfs_root { - struct buffer_head *node; - struct buffer_head *commit_root; + struct extent_buffer *node; + struct extent_buffer *commit_root; struct btrfs_root_item root_item; struct btrfs_key root_key; struct btrfs_fs_info *fs_info; @@ -336,7 +338,16 @@ struct btrfs_root { struct rw_semaphore snap_sem; u64 objectid; u64 last_trans; - u32 blocksize; + + /* data allocations are done in sectorsize units */ + u32 sectorsize; + + /* node allocations are done in nodesize units */ + u32 nodesize; + + /* leaf allocations are done in leafsize units */ + u32 leafsize; + u32 type; u64 highest_inode; u64 last_inode_alloc; @@ -347,12 +358,6 @@ struct btrfs_root { char *name; }; -/* the lower bits in the key flags defines the item type */ -#define BTRFS_KEY_TYPE_MAX 256 -#define BTRFS_KEY_TYPE_SHIFT 24 -#define BTRFS_KEY_TYPE_MASK (((u32)BTRFS_KEY_TYPE_MAX - 1) << \ - BTRFS_KEY_TYPE_SHIFT) - /* * inode items have the data typically returned from stat and store other * info about object characteristics. There is one for every file and dir in @@ -402,246 +407,253 @@ struct btrfs_root { */ #define BTRFS_STRING_ITEM_KEY 253 +/* some macros to generate set/get funcs for the struct fields. This + * assumes there is a lefoo_to_cpu for every type, so lets make a simple + * one for u8: + */ +#define le8_to_cpu(v) (v) +#define cpu_to_le8(v) (v) +#define __le8 u8 + +#define read_eb_member(eb, ptr, type, member, result) ( \ + read_extent_buffer(eb, (char *)(result), \ + ((unsigned long)(ptr)) + \ + offsetof(type, member), \ + sizeof(((type *)0)->member))) + +#define write_eb_member(eb, ptr, type, member, result) ( \ + write_extent_buffer(eb, (char *)(result), \ + ((unsigned long)(ptr)) + \ + offsetof(type, member), \ + sizeof(((type *)0)->member))) + +#define BTRFS_SETGET_FUNCS(name, type, member, bits) \ +static inline u##bits btrfs_##name(struct extent_buffer *eb, \ + type *s) \ +{ \ + __le##bits res; \ + read_eb_member(eb, s, type, member, &res); \ + return le##bits##_to_cpu(res); \ +} \ +static inline void btrfs_set_##name(struct extent_buffer *eb, \ + type *s, u##bits val) \ +{ \ + val = cpu_to_le##bits(val); \ + write_eb_member(eb, s, type, member, &val); \ +} + +#define BTRFS_SETGET_HEADER_FUNCS(name, type, member, bits) \ +static inline u##bits btrfs_##name(struct extent_buffer *eb) \ +{ \ + __le##bits res; \ + read_eb_member(eb, NULL, type, member, &res); \ + return le##bits##_to_cpu(res); \ +} \ +static inline void btrfs_set_##name(struct extent_buffer *eb, \ + u##bits val) \ +{ \ + val = cpu_to_le##bits(val); \ + write_eb_member(eb, NULL, type, member, &val); \ +} -static inline u64 btrfs_block_group_used(struct btrfs_block_group_item *bi) -{ - return le64_to_cpu(bi->used); -} - -static inline void btrfs_set_block_group_used(struct - btrfs_block_group_item *bi, - u64 val) -{ - bi->used = cpu_to_le64(val); -} - -static inline u64 btrfs_inode_generation(struct btrfs_inode_item *i) -{ - return le64_to_cpu(i->generation); -} - -static inline void btrfs_set_inode_generation(struct btrfs_inode_item *i, - u64 val) -{ - i->generation = cpu_to_le64(val); -} - -static inline u64 btrfs_inode_size(struct btrfs_inode_item *i) -{ - return le64_to_cpu(i->size); -} - -static inline void btrfs_set_inode_size(struct btrfs_inode_item *i, u64 val) -{ - i->size = cpu_to_le64(val); -} - -static inline u64 btrfs_inode_nblocks(struct btrfs_inode_item *i) -{ - return le64_to_cpu(i->nblocks); -} - -static inline void btrfs_set_inode_nblocks(struct btrfs_inode_item *i, u64 val) -{ - i->nblocks = cpu_to_le64(val); -} - -static inline u64 btrfs_inode_block_group(struct btrfs_inode_item *i) -{ - return le64_to_cpu(i->block_group); -} - -static inline void btrfs_set_inode_block_group(struct btrfs_inode_item *i, - u64 val) -{ - i->block_group = cpu_to_le64(val); -} - -static inline u32 btrfs_inode_nlink(struct btrfs_inode_item *i) -{ - return le32_to_cpu(i->nlink); -} - -static inline void btrfs_set_inode_nlink(struct btrfs_inode_item *i, u32 val) -{ - i->nlink = cpu_to_le32(val); -} - -static inline u32 btrfs_inode_uid(struct btrfs_inode_item *i) -{ - return le32_to_cpu(i->uid); -} - -static inline void btrfs_set_inode_uid(struct btrfs_inode_item *i, u32 val) -{ - i->uid = cpu_to_le32(val); -} - -static inline u32 btrfs_inode_gid(struct btrfs_inode_item *i) -{ - return le32_to_cpu(i->gid); -} - -static inline void btrfs_set_inode_gid(struct btrfs_inode_item *i, u32 val) -{ - i->gid = cpu_to_le32(val); -} - -static inline u32 btrfs_inode_mode(struct btrfs_inode_item *i) -{ - return le32_to_cpu(i->mode); -} - -static inline void btrfs_set_inode_mode(struct btrfs_inode_item *i, u32 val) -{ - i->mode = cpu_to_le32(val); +#define BTRFS_SETGET_STACK_FUNCS(name, type, member, bits) \ +static inline u##bits btrfs_##name(type *s) \ +{ \ + return le##bits##_to_cpu(s->member); \ +} \ +static inline void btrfs_set_##name(type *s, u##bits val) \ +{ \ + s->member = cpu_to_le##bits(val); \ } -static inline u32 btrfs_inode_rdev(struct btrfs_inode_item *i) -{ - return le32_to_cpu(i->rdev); -} +/* struct btrfs_block_group_item */ +BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item, + used, 64); +BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item, + used, 64); -static inline void btrfs_set_inode_rdev(struct btrfs_inode_item *i, u32 val) -{ - i->rdev = cpu_to_le32(val); -} +/* struct btrfs_inode_item */ +BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); +BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64); +BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64); +BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64); +BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32); +BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32); +BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32); +BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32); +BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 32); +BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 16); +BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item, + compat_flags, 16); -static inline u16 btrfs_inode_flags(struct btrfs_inode_item *i) +static inline struct btrfs_inode_timespec * +btrfs_inode_atime(struct btrfs_inode_item *inode_item) { - return le16_to_cpu(i->flags); + unsigned long ptr = (unsigned long)inode_item; + ptr += offsetof(struct btrfs_inode_item, atime); + return (struct btrfs_inode_timespec *)ptr; } -static inline void btrfs_set_inode_flags(struct btrfs_inode_item *i, u16 val) +static inline struct btrfs_inode_timespec * +btrfs_inode_mtime(struct btrfs_inode_item *inode_item) { - i->flags = cpu_to_le16(val); + unsigned long ptr = (unsigned long)inode_item; + ptr += offsetof(struct btrfs_inode_item, mtime); + return (struct btrfs_inode_timespec *)ptr; } -static inline u16 btrfs_inode_compat_flags(struct btrfs_inode_item *i) +static inline struct btrfs_inode_timespec * +btrfs_inode_ctime(struct btrfs_inode_item *inode_item) { - return le16_to_cpu(i->compat_flags); + unsigned long ptr = (unsigned long)inode_item; + ptr += offsetof(struct btrfs_inode_item, ctime); + return (struct btrfs_inode_timespec *)ptr; } -static inline void btrfs_set_inode_compat_flags(struct btrfs_inode_item *i, - u16 val) +static inline struct btrfs_inode_timespec * +btrfs_inode_otime(struct btrfs_inode_item *inode_item) { - i->compat_flags = cpu_to_le16(val); + unsigned long ptr = (unsigned long)inode_item; + ptr += offsetof(struct btrfs_inode_item, otime); + return (struct btrfs_inode_timespec *)ptr; } -static inline u64 btrfs_timespec_sec(struct btrfs_inode_timespec *ts) -{ - return le64_to_cpu(ts->sec); -} +BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_inode_timespec, sec, 64); +BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_inode_timespec, nsec, 32); -static inline void btrfs_set_timespec_sec(struct btrfs_inode_timespec *ts, - u64 val) -{ - ts->sec = cpu_to_le64(val); -} +/* struct btrfs_extent_item */ +BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32); +BTRFS_SETGET_FUNCS(extent_owner, struct btrfs_extent_item, owner, 32); -static inline u32 btrfs_timespec_nsec(struct btrfs_inode_timespec *ts) -{ - return le32_to_cpu(ts->nsec); -} +BTRFS_SETGET_STACK_FUNCS(stack_extent_refs, struct btrfs_extent_item, + refs, 32); +BTRFS_SETGET_STACK_FUNCS(stack_extent_owner, struct btrfs_extent_item, + owner, 32); -static inline void btrfs_set_timespec_nsec(struct btrfs_inode_timespec *ts, - u32 val) -{ - ts->nsec = cpu_to_le32(val); -} +/* struct btrfs_node */ +BTRFS_SETGET_FUNCS(key_blockptr, struct btrfs_key_ptr, blockptr, 64); -static inline u32 btrfs_extent_refs(struct btrfs_extent_item *ei) +static inline u64 btrfs_node_blockptr(struct extent_buffer *eb, int nr) { - return le32_to_cpu(ei->refs); + unsigned long ptr; + ptr = offsetof(struct btrfs_node, ptrs) + + sizeof(struct btrfs_key_ptr) * nr; + return btrfs_key_blockptr(eb, (struct btrfs_key_ptr *)ptr); } -static inline void btrfs_set_extent_refs(struct btrfs_extent_item *ei, u32 val) +static inline void btrfs_set_node_blockptr(struct extent_buffer *eb, + int nr, u64 val) { - ei->refs = cpu_to_le32(val); + unsigned long ptr; + ptr = offsetof(struct btrfs_node, ptrs) + + sizeof(struct btrfs_key_ptr) * nr; + btrfs_set_key_blockptr(eb, (struct btrfs_key_ptr *)ptr, val); } -static inline u64 btrfs_extent_owner(struct btrfs_extent_item *ei) +static unsigned long btrfs_node_key_ptr_offset(int nr) { - return le64_to_cpu(ei->owner); + return offsetof(struct btrfs_node, ptrs) + + sizeof(struct btrfs_key_ptr) * nr; } -static inline void btrfs_set_extent_owner(struct btrfs_extent_item *ei, u64 val) +static void btrfs_node_key(struct extent_buffer *eb, + struct btrfs_disk_key *disk_key, int nr) { - ei->owner = cpu_to_le64(val); + unsigned long ptr; + ptr = btrfs_node_key_ptr_offset(nr); + read_eb_member(eb, (struct btrfs_key_ptr *)ptr, + struct btrfs_key_ptr, key, disk_key); } - -static inline u64 btrfs_node_blockptr(struct btrfs_node *n, int nr) +static inline void btrfs_set_node_key(struct extent_buffer *eb, + struct btrfs_disk_key *disk_key, int nr) { - return le64_to_cpu(n->ptrs[nr].blockptr); + unsigned long ptr; + ptr = btrfs_node_key_ptr_offset(nr); + write_eb_member(eb, (struct btrfs_key_ptr *)ptr, + struct btrfs_key_ptr, key, disk_key); } +/* struct btrfs_item */ +BTRFS_SETGET_FUNCS(item_offset, struct btrfs_item, offset, 32); +BTRFS_SETGET_FUNCS(item_size, struct btrfs_item, size, 32); -static inline void btrfs_set_node_blockptr(struct btrfs_node *n, int nr, - u64 val) +static inline unsigned long btrfs_item_nr_offset(int nr) { - n->ptrs[nr].blockptr = cpu_to_le64(val); + return offsetof(struct btrfs_leaf, items) + + sizeof(struct btrfs_item) * nr; } -static inline u32 btrfs_item_offset(struct btrfs_item *item) +static inline struct btrfs_item *btrfs_item_nr(struct extent_buffer *eb, + int nr) { - return le32_to_cpu(item->offset); + return (struct btrfs_item *)btrfs_item_nr_offset(nr); } -static inline void btrfs_set_item_offset(struct btrfs_item *item, u32 val) +static inline u32 btrfs_item_end(struct extent_buffer *eb, + struct btrfs_item *item) { - item->offset = cpu_to_le32(val); + return btrfs_item_offset(eb, item) + btrfs_item_size(eb, item); } -static inline u32 btrfs_item_end(struct btrfs_item *item) +static inline u32 btrfs_item_end_nr(struct extent_buffer *eb, int nr) { - return le32_to_cpu(item->offset) + le16_to_cpu(item->size); + return btrfs_item_end(eb, btrfs_item_nr(eb, nr)); } -static inline u16 btrfs_item_size(struct btrfs_item *item) +static inline u32 btrfs_item_offset_nr(struct extent_buffer *eb, int nr) { - return le16_to_cpu(item->size); + return btrfs_item_offset(eb, btrfs_item_nr(eb, nr)); } -static inline void btrfs_set_item_size(struct btrfs_item *item, u16 val) +static inline u32 btrfs_item_size_nr(struct extent_buffer *eb, int nr) { - item->size = cpu_to_le16(val); + return btrfs_item_size(eb, btrfs_item_nr(eb, nr)); } -static inline u16 btrfs_dir_flags(struct btrfs_dir_item *d) +static inline void btrfs_item_key(struct extent_buffer *eb, + struct btrfs_disk_key *disk_key, int nr) { - return le16_to_cpu(d->flags); + struct btrfs_item *item = btrfs_item_nr(eb, nr); + read_eb_member(eb, item, struct btrfs_item, key, disk_key); } -static inline void btrfs_set_dir_flags(struct btrfs_dir_item *d, u16 val) +static inline void btrfs_set_item_key(struct extent_buffer *eb, + struct btrfs_disk_key *disk_key, int nr) { - d->flags = cpu_to_le16(val); + struct btrfs_item *item = btrfs_item_nr(eb, nr); + write_eb_member(eb, item, struct btrfs_item, key, disk_key); } -static inline u8 btrfs_dir_type(struct btrfs_dir_item *d) -{ - return d->type; -} +/* struct btrfs_dir_item */ +BTRFS_SETGET_FUNCS(dir_flags, struct btrfs_dir_item, flags, 16); +BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8); +BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16); -static inline void btrfs_set_dir_type(struct btrfs_dir_item *d, u8 val) +static inline void btrfs_dir_item_key(struct extent_buffer *eb, + struct btrfs_dir_item *item, + struct btrfs_disk_key *key) { - d->type = val; + read_eb_member(eb, item, struct btrfs_dir_item, location, key); } -static inline u16 btrfs_dir_name_len(struct btrfs_dir_item *d) +static inline void btrfs_set_dir_item_key(struct extent_buffer *eb, + struct btrfs_dir_item *item, + struct btrfs_disk_key *key) { - return le16_to_cpu(d->name_len); + write_eb_member(eb, item, struct btrfs_dir_item, location, key); } -static inline void btrfs_set_dir_name_len(struct btrfs_dir_item *d, u16 val) -{ - d->name_len = cpu_to_le16(val); -} +/* struct btrfs_disk_key */ +BTRFS_SETGET_STACK_FUNCS(disk_key_objectid, struct btrfs_disk_key, + objectid, 64); +BTRFS_SETGET_STACK_FUNCS(disk_key_offset, struct btrfs_disk_key, offset, 64); +BTRFS_SETGET_STACK_FUNCS(disk_key_type, struct btrfs_disk_key, type, 8); static inline void btrfs_disk_key_to_cpu(struct btrfs_key *cpu, struct btrfs_disk_key *disk) { cpu->offset = le64_to_cpu(disk->offset); - cpu->flags = le32_to_cpu(disk->flags); + cpu->type = disk->type; cpu->objectid = le64_to_cpu(disk->objectid); } @@ -649,400 +661,167 @@ static inline void btrfs_cpu_key_to_disk(struct btrfs_disk_key *disk, struct btrfs_key *cpu) { disk->offset = cpu_to_le64(cpu->offset); - disk->flags = cpu_to_le32(cpu->flags); + disk->type = cpu->type; disk->objectid = cpu_to_le64(cpu->objectid); } -static inline u64 btrfs_disk_key_objectid(struct btrfs_disk_key *disk) -{ - return le64_to_cpu(disk->objectid); -} - -static inline void btrfs_set_disk_key_objectid(struct btrfs_disk_key *disk, - u64 val) -{ - disk->objectid = cpu_to_le64(val); -} - -static inline u64 btrfs_disk_key_offset(struct btrfs_disk_key *disk) -{ - return le64_to_cpu(disk->offset); -} - -static inline void btrfs_set_disk_key_offset(struct btrfs_disk_key *disk, - u64 val) -{ - disk->offset = cpu_to_le64(val); -} - -static inline u32 btrfs_disk_key_flags(struct btrfs_disk_key *disk) -{ - return le32_to_cpu(disk->flags); -} - -static inline void btrfs_set_disk_key_flags(struct btrfs_disk_key *disk, - u32 val) -{ - disk->flags = cpu_to_le32(val); -} - -static inline u32 btrfs_disk_key_type(struct btrfs_disk_key *key) -{ - return le32_to_cpu(key->flags) >> BTRFS_KEY_TYPE_SHIFT; -} - -static inline void btrfs_set_disk_key_type(struct btrfs_disk_key *key, - u32 val) -{ - u32 flags = btrfs_disk_key_flags(key); - BUG_ON(val >= BTRFS_KEY_TYPE_MAX); - val = val << BTRFS_KEY_TYPE_SHIFT; - flags = (flags & ~BTRFS_KEY_TYPE_MASK) | val; - btrfs_set_disk_key_flags(key, flags); -} - -static inline u32 btrfs_key_type(struct btrfs_key *key) -{ - return key->flags >> BTRFS_KEY_TYPE_SHIFT; -} - -static inline void btrfs_set_key_type(struct btrfs_key *key, u32 val) -{ - BUG_ON(val >= BTRFS_KEY_TYPE_MAX); - val = val << BTRFS_KEY_TYPE_SHIFT; - key->flags = (key->flags & ~(BTRFS_KEY_TYPE_MASK)) | val; -} - -static inline u64 btrfs_header_blocknr(struct btrfs_header *h) -{ - return le64_to_cpu(h->blocknr); -} - -static inline void btrfs_set_header_blocknr(struct btrfs_header *h, u64 blocknr) -{ - h->blocknr = cpu_to_le64(blocknr); -} - -static inline u64 btrfs_header_generation(struct btrfs_header *h) +static inline void btrfs_node_key_to_cpu(struct extent_buffer *eb, + struct btrfs_key *key, int nr) { - return le64_to_cpu(h->generation); + struct btrfs_disk_key disk_key; + btrfs_node_key(eb, &disk_key, nr); + btrfs_disk_key_to_cpu(key, &disk_key); } -static inline void btrfs_set_header_generation(struct btrfs_header *h, - u64 val) +static inline void btrfs_item_key_to_cpu(struct extent_buffer *eb, + struct btrfs_key *key, int nr) { - h->generation = cpu_to_le64(val); + struct btrfs_disk_key disk_key; + btrfs_item_key(eb, &disk_key, nr); + btrfs_disk_key_to_cpu(key, &disk_key); } -static inline u64 btrfs_header_owner(struct btrfs_header *h) +static inline void btrfs_dir_item_key_to_cpu(struct extent_buffer *eb, + struct btrfs_dir_item *item, + struct btrfs_key *key) { - return le64_to_cpu(h->owner); + struct btrfs_disk_key disk_key; + btrfs_dir_item_key(eb, item, &disk_key); + btrfs_disk_key_to_cpu(key, &disk_key); } -static inline void btrfs_set_header_owner(struct btrfs_header *h, - u64 val) -{ - h->owner = cpu_to_le64(val); -} - -static inline u16 btrfs_header_nritems(struct btrfs_header *h) -{ - return le16_to_cpu(h->nritems); -} - -static inline void btrfs_set_header_nritems(struct btrfs_header *h, u16 val) -{ - h->nritems = cpu_to_le16(val); -} - -static inline u16 btrfs_header_flags(struct btrfs_header *h) -{ - return le16_to_cpu(h->flags); -} - -static inline void btrfs_set_header_flags(struct btrfs_header *h, u16 val) -{ - h->flags = cpu_to_le16(val); -} - -static inline int btrfs_header_level(struct btrfs_header *h) -{ - return h->level; -} - -static inline void btrfs_set_header_level(struct btrfs_header *h, int level) -{ - BUG_ON(level > BTRFS_MAX_LEVEL); - h->level = level; -} - -static inline int btrfs_is_leaf(struct btrfs_node *n) -{ - return (btrfs_header_level(&n->header) == 0); -} - -static inline u64 btrfs_root_blocknr(struct btrfs_root_item *item) -{ - return le64_to_cpu(item->blocknr); -} - -static inline void btrfs_set_root_blocknr(struct btrfs_root_item *item, u64 val) -{ - item->blocknr = cpu_to_le64(val); -} - -static inline u64 btrfs_root_dirid(struct btrfs_root_item *item) -{ - return le64_to_cpu(item->root_dirid); -} - -static inline void btrfs_set_root_dirid(struct btrfs_root_item *item, u64 val) -{ - item->root_dirid = cpu_to_le64(val); -} - -static inline u32 btrfs_root_refs(struct btrfs_root_item *item) -{ - return le32_to_cpu(item->refs); -} - -static inline void btrfs_set_root_refs(struct btrfs_root_item *item, u32 val) -{ - item->refs = cpu_to_le32(val); -} - -static inline u32 btrfs_root_flags(struct btrfs_root_item *item) -{ - return le32_to_cpu(item->flags); -} - -static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val) -{ - item->flags = cpu_to_le32(val); -} - -static inline void btrfs_set_root_blocks_used(struct btrfs_root_item *item, - u64 val) -{ - item->blocks_used = cpu_to_le64(val); -} - -static inline u64 btrfs_root_blocks_used(struct btrfs_root_item *item) -{ - return le64_to_cpu(item->blocks_used); -} - -static inline void btrfs_set_root_block_limit(struct btrfs_root_item *item, - u64 val) -{ - item->block_limit = cpu_to_le64(val); -} - -static inline u64 btrfs_root_block_limit(struct btrfs_root_item *item) -{ - return le64_to_cpu(item->block_limit); -} -static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s) +static inline u8 btrfs_key_type(struct btrfs_key *key) { - return le64_to_cpu(s->blocknr); + return key->type; } -static inline void btrfs_set_super_blocknr(struct btrfs_super_block *s, u64 val) +static inline void btrfs_set_key_type(struct btrfs_key *key, u8 val) { - s->blocknr = cpu_to_le64(val); + key->type = val; } -static inline u64 btrfs_super_generation(struct btrfs_super_block *s) -{ - return le64_to_cpu(s->generation); -} +/* struct btrfs_header */ +BTRFS_SETGET_HEADER_FUNCS(header_blocknr, struct btrfs_header, blocknr, 64); +BTRFS_SETGET_HEADER_FUNCS(header_generation, struct btrfs_header, + generation, 64); +BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64); +BTRFS_SETGET_HEADER_FUNCS(header_nritems, struct btrfs_header, nritems, 32); +BTRFS_SETGET_HEADER_FUNCS(header_flags, struct btrfs_header, flags, 16); +BTRFS_SETGET_HEADER_FUNCS(header_level, struct btrfs_header, level, 8); -static inline void btrfs_set_super_generation(struct btrfs_super_block *s, - u64 val) +static inline u8 *btrfs_header_fsid(struct extent_buffer *eb) { - s->generation = cpu_to_le64(val); + unsigned long ptr = offsetof(struct btrfs_header, fsid); + return (u8 *)ptr; } -static inline u64 btrfs_super_root(struct btrfs_super_block *s) +static inline u8 *btrfs_super_fsid(struct extent_buffer *eb) { - return le64_to_cpu(s->root); + unsigned long ptr = offsetof(struct btrfs_super_block, fsid); + return (u8 *)ptr; } -static inline void btrfs_set_super_root(struct btrfs_super_block *s, u64 val) +static inline u8 *btrfs_header_csum(struct extent_buffer *eb) { - s->root = cpu_to_le64(val); + unsigned long ptr = offsetof(struct btrfs_header, csum); + return (u8 *)ptr; } -static inline u64 btrfs_super_total_blocks(struct btrfs_super_block *s) +static inline struct btrfs_node *btrfs_buffer_node(struct extent_buffer *eb) { - return le64_to_cpu(s->total_blocks); + return NULL; } -static inline void btrfs_set_super_total_blocks(struct btrfs_super_block *s, - u64 val) +static inline struct btrfs_leaf *btrfs_buffer_leaf(struct extent_buffer *eb) { - s->total_blocks = cpu_to_le64(val); + return NULL; } -static inline u64 btrfs_super_blocks_used(struct btrfs_super_block *s) +static inline struct btrfs_header *btrfs_buffer_header(struct extent_buffer *eb) { - return le64_to_cpu(s->blocks_used); + return NULL; } -static inline void btrfs_set_super_blocks_used(struct btrfs_super_block *s, - u64 val) +static inline int btrfs_is_leaf(struct extent_buffer *eb) { - s->blocks_used = cpu_to_le64(val); + return (btrfs_header_level(eb) == 0); } -static inline u32 btrfs_super_blocksize(struct btrfs_super_block *s) -{ - return le32_to_cpu(s->blocksize); -} +/* struct btrfs_root_item */ +BTRFS_SETGET_FUNCS(disk_root_refs, struct btrfs_root_item, refs, 32); +BTRFS_SETGET_FUNCS(disk_root_blocknr, struct btrfs_root_item, blocknr, 64); -static inline void btrfs_set_super_blocksize(struct btrfs_super_block *s, - u32 val) -{ - s->blocksize = cpu_to_le32(val); -} +BTRFS_SETGET_STACK_FUNCS(root_blocknr, struct btrfs_root_item, blocknr, 64); +BTRFS_SETGET_STACK_FUNCS(root_dirid, struct btrfs_root_item, root_dirid, 64); +BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32); +BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32); +BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, blocks_used, 64); +BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, block_limit, 64); -static inline u64 btrfs_super_root_dir(struct btrfs_super_block *s) -{ - return le64_to_cpu(s->root_dir_objectid); -} +/* struct btrfs_super_block */ +BTRFS_SETGET_STACK_FUNCS(super_blocknr, struct btrfs_super_block, blocknr, 64); +BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block, + generation, 64); +BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64); +BTRFS_SETGET_STACK_FUNCS(super_total_blocks, struct btrfs_super_block, + total_blocks, 64); +BTRFS_SETGET_STACK_FUNCS(super_blocks_used, struct btrfs_super_block, + blocks_used, 64); +BTRFS_SETGET_STACK_FUNCS(super_sectorsize, struct btrfs_super_block, + sectorsize, 32); +BTRFS_SETGET_STACK_FUNCS(super_nodesize, struct btrfs_super_block, + nodesize, 32); +BTRFS_SETGET_STACK_FUNCS(super_leafsize, struct btrfs_super_block, + leafsize, 32); +BTRFS_SETGET_STACK_FUNCS(super_root_dir, struct btrfs_super_block, + root_dir_objectid, 64); -static inline void btrfs_set_super_root_dir(struct btrfs_super_block *s, u64 - val) +static inline unsigned long btrfs_leaf_data(struct extent_buffer *l) { - s->root_dir_objectid = cpu_to_le64(val); + return offsetof(struct btrfs_leaf, items); } -static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l) -{ - return (u8 *)l->items; -} +/* struct btrfs_file_extent_item */ +BTRFS_SETGET_FUNCS(file_extent_type, struct btrfs_file_extent_item, type, 8); -static inline int btrfs_file_extent_type(struct btrfs_file_extent_item *e) -{ - return e->type; -} -static inline void btrfs_set_file_extent_type(struct btrfs_file_extent_item *e, - u8 val) -{ - e->type = val; -} - -static inline char *btrfs_file_extent_inline_start(struct +static inline unsigned long btrfs_file_extent_inline_start(struct btrfs_file_extent_item *e) { - return (char *)(&e->disk_blocknr); + unsigned long offset = (unsigned long)e; + offset += offsetof(struct btrfs_file_extent_item, disk_blocknr); + return offset; } static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize) { - return (unsigned long)(&((struct - btrfs_file_extent_item *)NULL)->disk_blocknr) + datasize; -} - -static inline u32 btrfs_file_extent_inline_len(struct btrfs_item *e) -{ - struct btrfs_file_extent_item *fe = NULL; - return btrfs_item_size(e) - (unsigned long)(&fe->disk_blocknr); -} - -static inline u64 btrfs_file_extent_disk_blocknr(struct btrfs_file_extent_item - *e) -{ - return le64_to_cpu(e->disk_blocknr); + return offsetof(struct btrfs_file_extent_item, disk_blocknr) + datasize; } -static inline void btrfs_set_file_extent_disk_blocknr(struct - btrfs_file_extent_item - *e, u64 val) +static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb, + struct btrfs_item *e) { - e->disk_blocknr = cpu_to_le64(val); + unsigned long offset; + offset = offsetof(struct btrfs_file_extent_item, disk_blocknr); + return btrfs_item_size(eb, e) - offset; } -static inline u64 btrfs_file_extent_generation(struct btrfs_file_extent_item *e) -{ - return le64_to_cpu(e->generation); -} - -static inline void btrfs_set_file_extent_generation(struct - btrfs_file_extent_item *e, - u64 val) -{ - e->generation = cpu_to_le64(val); -} - -static inline u64 btrfs_file_extent_disk_num_blocks(struct - btrfs_file_extent_item *e) -{ - return le64_to_cpu(e->disk_num_blocks); -} - -static inline void btrfs_set_file_extent_disk_num_blocks(struct - btrfs_file_extent_item - *e, u64 val) -{ - e->disk_num_blocks = cpu_to_le64(val); -} - -static inline u64 btrfs_file_extent_offset(struct btrfs_file_extent_item *e) -{ - return le64_to_cpu(e->offset); -} - -static inline void btrfs_set_file_extent_offset(struct btrfs_file_extent_item - *e, u64 val) -{ - e->offset = cpu_to_le64(val); -} - -static inline u64 btrfs_file_extent_num_blocks(struct btrfs_file_extent_item - *e) -{ - return le64_to_cpu(e->num_blocks); -} - -static inline void btrfs_set_file_extent_num_blocks(struct - btrfs_file_extent_item *e, - u64 val) -{ - e->num_blocks = cpu_to_le64(val); -} +BTRFS_SETGET_FUNCS(file_extent_disk_blocknr, struct btrfs_file_extent_item, + disk_blocknr, 64); +BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item, + generation, 64); +BTRFS_SETGET_FUNCS(file_extent_disk_num_blocks, struct btrfs_file_extent_item, + disk_num_blocks, 64); +BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item, + offset, 64); +BTRFS_SETGET_FUNCS(file_extent_num_blocks, struct btrfs_file_extent_item, + num_blocks, 64); static inline struct btrfs_root *btrfs_sb(struct super_block *sb) { return sb->s_fs_info; } -static inline void btrfs_check_bounds(void *vptr, size_t len, - void *vcontainer, size_t container_len) -{ - char *ptr = vptr; - char *container = vcontainer; - WARN_ON(ptr < container); - WARN_ON(ptr + len > container + container_len); -} - -static inline void btrfs_memcpy(struct btrfs_root *root, - void *dst_block, - void *dst, const void *src, size_t nr) -{ - btrfs_check_bounds(dst, nr, dst_block, root->fs_info->sb->s_blocksize); - memcpy(dst, src, nr); -} - -static inline void btrfs_memmove(struct btrfs_root *root, - void *dst_block, - void *dst, void *src, size_t nr) -{ - btrfs_check_bounds(dst, nr, dst_block, root->fs_info->sb->s_blocksize); - memmove(dst, src, nr); -} - static inline int btrfs_set_root_name(struct btrfs_root *root, const char *name, int len) { @@ -1063,7 +842,11 @@ static inline int btrfs_set_root_name(struct btrfs_root *root, /* helper function to cast into the data area of the leaf. */ #define btrfs_item_ptr(leaf, slot, type) \ ((type *)(btrfs_leaf_data(leaf) + \ - btrfs_item_offset((leaf)->items + (slot)))) + btrfs_item_offset_nr(leaf, slot))) + +#define btrfs_item_ptr_offset(leaf, slot) \ + ((unsigned long)(btrfs_leaf_data(leaf) + \ + btrfs_item_offset_nr(leaf, slot))) /* mount option defines and helpers */ #define BTRFS_MOUNT_SUBVOL 0x000001 @@ -1084,7 +867,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int data, int owner); int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root); -struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, +struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 hint, u64 empty_size); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, @@ -1092,7 +875,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 num_blocks, u64 empty_size, u64 search_start, u64 search_end, struct btrfs_key *ins, int data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf); + struct extent_buffer *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 blocknr, u64 num_blocks, int pin); int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, @@ -1106,10 +889,10 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, int btrfs_free_block_groups(struct btrfs_fs_info *info); int btrfs_read_block_groups(struct btrfs_root *root); /* ctree.c */ -int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct buffer_head *buf, struct buffer_head - *parent, int parent_slot, struct buffer_head - **cow_ret); +int btrfs_cow_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *buf, + struct extent_buffer *parent, int parent_slot, + struct extent_buffer **cow_ret); int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u32 data_size); int btrfs_truncate_item(struct btrfs_trans_handle *trans, @@ -1120,7 +903,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_path *p, int ins_len, int cow); int btrfs_realloc_node(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct buffer_head *parent, + struct btrfs_root *root, struct extent_buffer *parent, int cache_only, u64 *last_ret); void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); @@ -1134,7 +917,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *cpu_key, u32 data_size); int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path); -int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf); +int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf); int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root *root); /* root-item.c */ @@ -1179,9 +962,9 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, int btrfs_find_highest_inode(struct btrfs_root *fs_root, u64 *objectid); /* inode-item.c */ -int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 objectid, struct btrfs_inode_item - *inode_item); +int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 objectid); int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *location, int mod); @@ -1224,8 +1007,6 @@ struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to); -int btrfs_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *result, int create); struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, size_t page_offset, u64 start, u64 end, int create); diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 49db5fa7ced..6f19de41b87 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -33,7 +33,7 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle int ret; char *ptr; struct btrfs_item *item; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size); if (ret == -EEXIST) { @@ -49,11 +49,11 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle if (ret < 0) return ERR_PTR(ret); WARN_ON(ret > 0); - leaf = btrfs_buffer_leaf(path->nodes[0]); - item = leaf->items + path->slots[0]; + leaf = path->nodes[0]; + item = btrfs_item_nr(leaf, path->slots[0]); ptr = btrfs_item_ptr(leaf, path->slots[0], char); - BUG_ON(data_size > btrfs_item_size(item)); - ptr += btrfs_item_size(item) - data_size; + BUG_ON(data_size > btrfs_item_size(leaf, item)); + ptr += btrfs_item_size(leaf, item) - data_size; return (struct btrfs_dir_item *)ptr; } @@ -65,12 +65,13 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root int ret2 = 0; struct btrfs_path *path; struct btrfs_dir_item *dir_item; - char *name_ptr; + struct extent_buffer *leaf; + unsigned long name_ptr; struct btrfs_key key; + struct btrfs_disk_key disk_key; u32 data_size; key.objectid = dir; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY); ret = btrfs_name_hash(name, name_len, &key.offset); BUG_ON(ret); @@ -85,14 +86,16 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root goto out; } - btrfs_cpu_key_to_disk(&dir_item->location, location); - btrfs_set_dir_type(dir_item, type); - btrfs_set_dir_flags(dir_item, 0); - btrfs_set_dir_name_len(dir_item, name_len); - name_ptr = (char *)(dir_item + 1); + leaf = path->nodes[0]; + btrfs_cpu_key_to_disk(&disk_key, location); + btrfs_set_dir_item_key(leaf, dir_item, &disk_key); + btrfs_set_dir_type(leaf, dir_item, type); + btrfs_set_dir_flags(leaf, dir_item, 0); + btrfs_set_dir_name_len(leaf, dir_item, name_len); + name_ptr = (unsigned long)(dir_item + 1); - btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len); - btrfs_mark_buffer_dirty(path->nodes[0]); + write_extent_buffer(leaf, name, name_ptr, name_len); + btrfs_mark_buffer_dirty(leaf); second_insert: /* FIXME, use some real flag for selecting the extra index */ @@ -110,13 +113,15 @@ second_insert: ret2 = PTR_ERR(dir_item); goto out; } - btrfs_cpu_key_to_disk(&dir_item->location, location); - btrfs_set_dir_type(dir_item, type); - btrfs_set_dir_flags(dir_item, 0); - btrfs_set_dir_name_len(dir_item, name_len); - name_ptr = (char *)(dir_item + 1); - btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len); - btrfs_mark_buffer_dirty(path->nodes[0]); + leaf = path->nodes[0]; + btrfs_cpu_key_to_disk(&disk_key, location); + btrfs_set_dir_item_key(leaf, dir_item, &disk_key); + btrfs_set_dir_type(leaf, dir_item, type); + btrfs_set_dir_flags(leaf, dir_item, 0); + btrfs_set_dir_name_len(leaf, dir_item, name_len); + name_ptr = (unsigned long)(dir_item + 1); + write_extent_buffer(leaf, name, name_ptr, name_len); + btrfs_mark_buffer_dirty(leaf); out: btrfs_free_path(path); if (ret) @@ -136,14 +141,15 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_key key; int ins_len = mod < 0 ? -1 : 0; int cow = mod != 0; - struct btrfs_disk_key *found_key; - struct btrfs_leaf *leaf; + struct btrfs_key found_key; + struct extent_buffer *leaf; key.objectid = dir; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY); + ret = btrfs_name_hash(name, name_len, &key.offset); BUG_ON(ret); + ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow); if (ret < 0) return ERR_PTR(ret); @@ -152,12 +158,13 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, return NULL; path->slots[0]--; } - leaf = btrfs_buffer_leaf(path->nodes[0]); - found_key = &leaf->items[path->slots[0]].key; - if (btrfs_disk_key_objectid(found_key) != dir || - btrfs_disk_key_type(found_key) != BTRFS_DIR_ITEM_KEY || - btrfs_disk_key_offset(found_key) != key.offset) + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + if (found_key.objectid != dir || + btrfs_key_type(&found_key) != BTRFS_DIR_ITEM_KEY || + found_key.offset != key.offset) return NULL; return btrfs_match_dir_item_name(root, path, name, name_len); @@ -176,7 +183,6 @@ btrfs_lookup_dir_index_item(struct btrfs_trans_handle *trans, int cow = mod != 0; key.objectid = dir; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY); key.offset = objectid; @@ -193,21 +199,22 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, const char *name, int name_len) { struct btrfs_dir_item *dir_item; - char *name_ptr; + unsigned long name_ptr; u32 total_len; u32 cur = 0; u32 this_len; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; - leaf = btrfs_buffer_leaf(path->nodes[0]); + leaf = path->nodes[0]; dir_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dir_item); - total_len = btrfs_item_size(leaf->items + path->slots[0]); + total_len = btrfs_item_size_nr(leaf, path->slots[0]); while(cur < total_len) { - this_len = sizeof(*dir_item) + btrfs_dir_name_len(dir_item); - name_ptr = (char *)(dir_item + 1); + this_len = sizeof(*dir_item) + + btrfs_dir_name_len(leaf, dir_item); + name_ptr = (unsigned long)(dir_item + 1); - if (btrfs_dir_name_len(dir_item) == name_len && - memcmp(name_ptr, name, name_len) == 0) + if (btrfs_dir_name_len(leaf, dir_item) == name_len && + memcmp_extent_buffer(leaf, name, name_ptr, name_len) == 0) return dir_item; cur += this_len; @@ -223,20 +230,23 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, struct btrfs_dir_item *di) { - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; u32 sub_item_len; u32 item_len; int ret = 0; - leaf = btrfs_buffer_leaf(path->nodes[0]); - sub_item_len = sizeof(*di) + btrfs_dir_name_len(di); - item_len = btrfs_item_size(leaf->items + path->slots[0]); - if (sub_item_len == btrfs_item_size(leaf->items + path->slots[0])) { + leaf = path->nodes[0]; + sub_item_len = sizeof(*di) + btrfs_dir_name_len(leaf, di); + item_len = btrfs_item_size_nr(leaf, path->slots[0]); + if (sub_item_len == item_len) { ret = btrfs_del_item(trans, root, path); } else { - char *ptr = (char *)di; - char *start = btrfs_item_ptr(leaf, path->slots[0], char); - btrfs_memmove(root, leaf, ptr, ptr + sub_item_len, + /* MARKER */ + unsigned long ptr = (unsigned long)di; + unsigned long start; + + start = btrfs_item_ptr_offset(leaf, path->slots[0]); + memmove_extent_buffer(leaf, ptr, ptr + sub_item_len, item_len - (ptr + sub_item_len - start)); ret = btrfs_truncate_item(trans, root, path, item_len - sub_item_len); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index ef6d76f2ec2..0c1f90cbedb 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -23,138 +23,132 @@ #include #include #include +#include // for block_sync_page #include "ctree.h" #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" -u64 bh_blocknr(struct buffer_head *bh) +#if 0 +static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) { - return bh->b_blocknr; -} - -static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf) -{ - struct btrfs_node *node = btrfs_buffer_node(buf); - if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) { - printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n", - (unsigned long long)bh_blocknr(buf), - (unsigned long long)btrfs_header_blocknr(&node->header)); + if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { + printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", + (unsigned long long)extent_buffer_blocknr(buf), + (unsigned long long)btrfs_header_blocknr(buf)); return 1; } return 0; } +#endif -struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr) +struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, + u64 blocknr) { - struct address_space *mapping = root->fs_info->btree_inode->i_mapping; - int blockbits = root->fs_info->sb->s_blocksize_bits; - unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits); - struct page *page; - struct buffer_head *bh; - struct buffer_head *head; - struct buffer_head *ret = NULL; - - - page = find_lock_page(mapping, index); - if (!page) - return NULL; - - if (!page_has_buffers(page)) - goto out_unlock; + struct inode *btree_inode = root->fs_info->btree_inode; + return find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, + blocknr * root->sectorsize, + root->sectorsize, GFP_NOFS); +} - head = page_buffers(page); - bh = head; - do { - if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) { - ret = bh; - get_bh(bh); - goto out_unlock; - } - bh = bh->b_this_page; - } while (bh != head); -out_unlock: - unlock_page(page); - page_cache_release(page); - return ret; +struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, + u64 blocknr) +{ + struct inode *btree_inode = root->fs_info->btree_inode; + return alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, + blocknr * root->sectorsize, + root->sectorsize, GFP_NOFS); } -int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, - u64 logical) +struct extent_map *btree_get_extent(struct inode *inode, struct page *page, + size_t page_offset, u64 start, u64 end, + int create) { - if (logical == 0) { - bh->b_bdev = NULL; - bh->b_blocknr = 0; - set_buffer_mapped(bh); - } else { - map_bh(bh, root->fs_info->sb, logical); + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_map *em; + int ret; + +again: + em = lookup_extent_mapping(em_tree, start, end); + if (em) { + goto out; } - return 0; + em = alloc_extent_map(GFP_NOFS); + if (!em) { + em = ERR_PTR(-ENOMEM); + goto out; + } + em->start = 0; + em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1; + em->block_start = 0; + em->block_end = em->end; + em->bdev = inode->i_sb->s_bdev; + ret = add_extent_mapping(em_tree, em); + if (ret == -EEXIST) { + free_extent_map(em); + em = NULL; + goto again; + } else if (ret) { + em = ERR_PTR(ret); + } +out: + return em; } -struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root, - u64 blocknr) +static int btree_writepage(struct page *page, struct writeback_control *wbc) { - struct address_space *mapping = root->fs_info->btree_inode->i_mapping; - int blockbits = root->fs_info->sb->s_blocksize_bits; - unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits); - struct page *page; - struct buffer_head *bh; - struct buffer_head *head; - struct buffer_head *ret = NULL; - int err; - u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits); + struct extent_map_tree *tree; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + return extent_write_full_page(tree, page, btree_get_extent, wbc); +} +int btree_readpage(struct file *file, struct page *page) +{ + struct extent_map_tree *tree; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + return extent_read_full_page(tree, page, btree_get_extent); +} - page = find_or_create_page(mapping, index, GFP_NOFS); - if (!page) - return NULL; +static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) +{ + struct extent_map_tree *tree; + int ret; - if (!page_has_buffers(page)) - create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0); - head = page_buffers(page); - bh = head; - do { - if (!buffer_mapped(bh)) { - err = btrfs_map_bh_to_logical(root, bh, first_block); - BUG_ON(err); - } - if (bh_blocknr(bh) == blocknr) { - ret = bh; - get_bh(bh); - goto out_unlock; - } - bh = bh->b_this_page; - first_block++; - } while (bh != head); -out_unlock: - unlock_page(page); - if (ret) - touch_buffer(ret); - page_cache_release(page); + BUG_ON(page->private != 1); + tree = &BTRFS_I(page->mapping->host)->extent_tree; + ret = try_release_extent_mapping(tree, page); + if (ret == 1) { + ClearPagePrivate(page); + set_page_private(page, 0); + page_cache_release(page); + } return ret; } -static int btree_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *bh, int create) +static void btree_invalidatepage(struct page *page, unsigned long offset) { - int err; - struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root; - err = btrfs_map_bh_to_logical(root, bh, iblock); - return err; + struct extent_map_tree *tree; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + extent_invalidatepage(tree, page, offset); + btree_releasepage(page, GFP_NOFS); } int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, char *result) { + return 0; +#if 0 u32 crc; crc = crc32c(0, data, len); memcpy(result, &crc, BTRFS_CRC32_SIZE); return 0; +#endif } -static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh, +#if 0 +static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, int verify) { + return 0; char result[BTRFS_CRC32_SIZE]; int ret; struct btrfs_node *node; @@ -176,7 +170,9 @@ static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh, } return 0; } +#endif +#if 0 static int btree_writepage(struct page *page, struct writeback_control *wbc) { struct buffer_head *bh; @@ -195,87 +191,65 @@ static int btree_writepage(struct page *page, struct writeback_control *wbc) } while (bh != head); return block_write_full_page(page, btree_get_block, wbc); } - -static int btree_readpage(struct file * file, struct page * page) -{ - return block_read_full_page(page, btree_get_block); -} +#endif static struct address_space_operations btree_aops = { .readpage = btree_readpage, .writepage = btree_writepage, + .releasepage = btree_releasepage, + .invalidatepage = btree_invalidatepage, .sync_page = block_sync_page, }; int readahead_tree_block(struct btrfs_root *root, u64 blocknr) { - struct buffer_head *bh = NULL; + struct extent_buffer *buf = NULL; + struct inode *btree_inode = root->fs_info->btree_inode; int ret = 0; - bh = btrfs_find_create_tree_block(root, blocknr); - if (!bh) + buf = btrfs_find_create_tree_block(root, blocknr); + if (!buf) return 0; - if (buffer_uptodate(bh)) { - ret = 1; - goto done; - } - if (test_set_buffer_locked(bh)) { - ret = 1; - goto done; - } - if (!buffer_uptodate(bh)) { - get_bh(bh); - bh->b_end_io = end_buffer_read_sync; - submit_bh(READ, bh); - } else { - unlock_buffer(bh); - ret = 1; - } -done: - brelse(bh); + read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, + buf, 0); + free_extent_buffer(buf); return ret; } -struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr) +struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr) { - struct buffer_head *bh = NULL; - - bh = btrfs_find_create_tree_block(root, blocknr); - if (!bh) - return bh; - if (buffer_uptodate(bh)) - goto uptodate; - lock_buffer(bh); - if (!buffer_uptodate(bh)) { - get_bh(bh); - bh->b_end_io = end_buffer_read_sync; - submit_bh(READ, bh); - wait_on_buffer(bh); - if (!buffer_uptodate(bh)) - goto fail; - } else { - unlock_buffer(bh); - } -uptodate: - if (!buffer_checked(bh)) { - csum_tree_block(root, bh, 1); - set_buffer_checked(bh); - } - if (check_tree_block(root, bh)) - goto fail; - return bh; -fail: - brelse(bh); - return NULL; + struct extent_buffer *buf = NULL; + struct inode *btree_inode = root->fs_info->btree_inode; + + buf = btrfs_find_create_tree_block(root, blocknr); + if (!buf) + return NULL; + read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, + buf, 1); + return buf; } int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf) + struct extent_buffer *buf) { - WARN_ON(atomic_read(&buf->b_count) == 0); - lock_buffer(buf); - clear_buffer_dirty(buf); - unlock_buffer(buf); + struct inode *btree_inode = root->fs_info->btree_inode; + clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); + return 0; +} + +int wait_on_tree_block_writeback(struct btrfs_root *root, + struct extent_buffer *buf) +{ + struct inode *btree_inode = root->fs_info->btree_inode; + wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree, + buf); + return 0; +} + +int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf) +{ + struct inode *btree_inode = root->fs_info->btree_inode; + set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); return 0; } @@ -287,7 +261,9 @@ static int __setup_root(int blocksize, root->node = NULL; root->inode = NULL; root->commit_root = NULL; - root->blocksize = blocksize; + root->sectorsize = blocksize; + root->nodesize = blocksize; + root->leafsize = blocksize; root->ref_cows = 0; root->fs_info = fs_info; root->objectid = objectid; @@ -332,7 +308,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, struct btrfs_root *root; struct btrfs_root *tree_root = fs_info->tree_root; struct btrfs_path *path; - struct btrfs_leaf *l; + struct extent_buffer *l; u64 highest_inode; int ret = 0; @@ -361,11 +337,10 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, ret = -ENOENT; goto out; } - l = btrfs_buffer_leaf(path->nodes[0]); - memcpy(&root->root_item, - btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item), + l = path->nodes[0]; + read_extent_buffer(l, &root->root_item, + btrfs_item_ptr_offset(l, path->slots[0]), sizeof(root->root_item)); - memcpy(&root->root_key, location, sizeof(*location)); ret = 0; out: btrfs_release_path(root, path); @@ -406,21 +381,21 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, (unsigned long)root->root_key.objectid, root); if (ret) { - brelse(root->node); + free_extent_buffer(root->node); kfree(root); return ERR_PTR(ret); } ret = btrfs_set_root_name(root, name, namelen); if (ret) { - brelse(root->node); + free_extent_buffer(root->node); kfree(root); return ERR_PTR(ret); } ret = btrfs_sysfs_add_root(root); if (ret) { - brelse(root->node); + free_extent_buffer(root->node); kfree(root->name); kfree(root); return ERR_PTR(ret); @@ -471,6 +446,9 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->btree_inode->i_nlink = 1; fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size; fs_info->btree_inode->i_mapping->a_ops = &btree_aops; + extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, + fs_info->btree_inode->i_mapping, + GFP_NOFS); fs_info->do_barriers = 1; fs_info->closing = 0; @@ -493,10 +471,14 @@ struct btrfs_root *open_ctree(struct super_block *sb) if (!fs_info->sb_buffer) goto fail_iput; - disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data; - fs_info->disk_super = disk_super; - memcpy(&fs_info->super_copy, disk_super, sizeof(fs_info->super_copy)); + read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0, + sizeof(fs_info->super_copy)); + + read_extent_buffer(fs_info->sb_buffer, fs_info->fsid, + (unsigned long)btrfs_super_fsid(fs_info->sb_buffer), + BTRFS_FSID_SIZE); + disk_super = &fs_info->super_copy; if (!btrfs_super_root(disk_super)) goto fail_sb_buffer; @@ -530,9 +512,9 @@ struct btrfs_root *open_ctree(struct super_block *sb) return tree_root; fail_tree_root: - btrfs_block_release(tree_root, tree_root->node); + free_extent_buffer(tree_root->node); fail_sb_buffer: - btrfs_block_release(tree_root, fs_info->sb_buffer); + free_extent_buffer(fs_info->sb_buffer); fail_iput: iput(fs_info->btree_inode); fail: @@ -546,31 +528,13 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root) { int ret; - struct buffer_head *bh = root->fs_info->sb_buffer; - - lock_buffer(bh); - WARN_ON(atomic_read(&bh->b_count) < 1); - clear_buffer_dirty(bh); - csum_tree_block(root, bh, 0); - bh->b_end_io = end_buffer_write_sync; - get_bh(bh); - if (root->fs_info->do_barriers) - ret = submit_bh(WRITE_BARRIER, bh); - else - ret = submit_bh(WRITE, bh); - if (ret == -EOPNOTSUPP) { - get_bh(bh); - lock_buffer(bh); - set_buffer_uptodate(bh); - root->fs_info->do_barriers = 0; - ret = submit_bh(WRITE, bh); - } - wait_on_buffer(bh); - if (!buffer_uptodate(bh)) { - WARN_ON(1); - return -EIO; - } - return 0; + struct extent_buffer *super = root->fs_info->sb_buffer; + struct inode *btree_inode = root->fs_info->btree_inode; + + set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super); + ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping, + super->start, super->len); + return ret; } int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) @@ -581,9 +545,9 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) if (root->inode) iput(root->inode); if (root->node) - brelse(root->node); + free_extent_buffer(root->node); if (root->commit_root) - brelse(root->commit_root); + free_extent_buffer(root->commit_root); if (root->name) kfree(root->name); kfree(root); @@ -629,12 +593,10 @@ int close_ctree(struct btrfs_root *root) mutex_unlock(&fs_info->fs_mutex); if (fs_info->extent_root->node) - btrfs_block_release(fs_info->extent_root, - fs_info->extent_root->node); + free_extent_buffer(fs_info->extent_root->node); if (fs_info->tree_root->node) - btrfs_block_release(fs_info->tree_root, - fs_info->tree_root->node); - btrfs_block_release(root, fs_info->sb_buffer); + free_extent_buffer(fs_info->tree_root->node); + free_extent_buffer(fs_info->sb_buffer); truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); iput(fs_info->btree_inode); @@ -645,25 +607,32 @@ int close_ctree(struct btrfs_root *root) return 0; } -void btrfs_mark_buffer_dirty(struct buffer_head *bh) +int btrfs_buffer_uptodate(struct extent_buffer *buf) +{ + struct inode *btree_inode = buf->pages[0]->mapping->host; + return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf); +} + +int btrfs_set_buffer_uptodate(struct extent_buffer *buf) { - struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root; - u64 transid = btrfs_header_generation(btrfs_buffer_header(bh)); + struct inode *btree_inode = buf->pages[0]->mapping->host; + return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, + buf); +} - WARN_ON(!atomic_read(&bh->b_count)); +void btrfs_mark_buffer_dirty(struct extent_buffer *buf) +{ + struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; + u64 transid = btrfs_header_generation(buf); + struct inode *btree_inode = root->fs_info->btree_inode; if (transid != root->fs_info->generation) { printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", - (unsigned long long)bh->b_blocknr, + (unsigned long long)extent_buffer_blocknr(buf), transid, root->fs_info->generation); WARN_ON(1); } - mark_buffer_dirty(bh); -} - -void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf) -{ - brelse(buf); + set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); } void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index fc6b520c6e0..70d9413c599 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -19,68 +19,35 @@ #ifndef __DISKIO__ #define __DISKIO__ -#include - #define BTRFS_SUPER_INFO_OFFSET (16 * 1024) -enum btrfs_bh_state_bits { - BH_Checked = BH_PrivateStart, - BH_Defrag, - BH_DefragDone, -}; -BUFFER_FNS(Checked, checked); -BUFFER_FNS(Defrag, defrag); -BUFFER_FNS(DefragDone, defrag_done); - -static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh) -{ - return (struct btrfs_node *)bh->b_data; -} - -static inline struct btrfs_leaf *btrfs_buffer_leaf(struct buffer_head *bh) -{ - return (struct btrfs_leaf *)bh->b_data; -} - -static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh) -{ - return &((struct btrfs_node *)bh->b_data)->header; -} - -struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr); +struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr); int readahead_tree_block(struct btrfs_root *root, u64 blocknr); -struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root, - u64 blocknr); -int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf); -int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf); +struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, + u64 blocknr); int clean_tree_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct buffer_head *buf); -int btrfs_commit_transaction(struct btrfs_trans_handle *trans, - struct btrfs_root *root); + struct btrfs_root *root, struct extent_buffer *buf); struct btrfs_root *open_ctree(struct super_block *sb); int close_ctree(struct btrfs_root *root); -void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf); int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root); -struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr); -int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, - char *result); +struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, + u64 blocknr); struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_key *location, const char *name, int namelen); struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, struct btrfs_key *location); -u64 bh_blocknr(struct buffer_head *bh); int btrfs_insert_dev_radix(struct btrfs_root *root, struct block_device *bdev, u64 device_id, u64 block_start, u64 num_blocks); -int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh, - u64 logical); void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr); int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); -void btrfs_mark_buffer_dirty(struct buffer_head *bh); +void btrfs_mark_buffer_dirty(struct extent_buffer *buf); +int btrfs_buffer_uptodate(struct extent_buffer *buf); +int btrfs_set_buffer_uptodate(struct extent_buffer *buf); +int wait_on_tree_block_writeback(struct btrfs_root *root, + struct extent_buffer *buf); #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index f261a8326cd..089c41cbca7 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -33,7 +33,7 @@ static int cache_block_group(struct btrfs_root *root, struct btrfs_path *path; int ret; struct btrfs_key key; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; struct radix_tree_root *extent_radix; int slot; u64 i; @@ -56,7 +56,6 @@ static int cache_block_group(struct btrfs_root *root, path->reada = 2; first_free = block_group->key.objectid; key.objectid = block_group->key.objectid; - key.flags = 0; key.offset = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); @@ -69,9 +68,9 @@ static int cache_block_group(struct btrfs_root *root, path->slots[0]--; while(1) { - leaf = btrfs_buffer_leaf(path->nodes[0]); + leaf = path->nodes[0]; slot = path->slots[0]; - if (slot >= btrfs_header_nritems(&leaf->header)) { + if (slot >= btrfs_header_nritems(leaf)) { ret = btrfs_next_leaf(root, path); if (ret < 0) goto err; @@ -82,7 +81,7 @@ static int cache_block_group(struct btrfs_root *root, } } - btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key); + btrfs_item_key_to_cpu(leaf, &key, slot); if (key.objectid < block_group->key.objectid) { if (key.objectid + key.offset > first_free) first_free = key.objectid + key.offset; @@ -116,8 +115,7 @@ next: hole_size = block_group->key.objectid + block_group->key.offset - last; for (i = 0; i < hole_size; i++) { - set_radix_bit(extent_radix, - last + i); + set_radix_bit(extent_radix, last + i); } } block_group->cached = 1; @@ -366,7 +364,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_path *path; int ret; struct btrfs_key key; - struct btrfs_leaf *l; + struct extent_buffer *l; struct btrfs_extent_item *item; u32 refs; @@ -375,7 +373,6 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, return -ENOMEM; key.objectid = blocknr; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); key.offset = num_blocks; ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, @@ -386,10 +383,10 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, BUG(); } BUG_ON(ret != 0); - l = btrfs_buffer_leaf(path->nodes[0]); + l = path->nodes[0]; item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); - refs = btrfs_extent_refs(item); - btrfs_set_extent_refs(item, refs + 1); + refs = btrfs_extent_refs(l, item); + btrfs_set_extent_refs(l, item, refs + 1); btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_release_path(root->fs_info->extent_root, path); @@ -414,23 +411,25 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_path *path; int ret; struct btrfs_key key; - struct btrfs_leaf *l; + struct extent_buffer *l; struct btrfs_extent_item *item; path = btrfs_alloc_path(); key.objectid = blocknr; key.offset = num_blocks; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 0); if (ret < 0) goto out; - if (ret != 0) + if (ret != 0) { + btrfs_print_leaf(root, path->nodes[0]); + printk("failed to find block number %Lu\n", blocknr); BUG(); - l = btrfs_buffer_leaf(path->nodes[0]); + } + l = path->nodes[0]; item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); - *refs = btrfs_extent_refs(item); + *refs = btrfs_extent_refs(l, item); out: btrfs_free_path(path); return 0; @@ -439,16 +438,16 @@ out: int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - return btrfs_inc_extent_ref(trans, root, bh_blocknr(root->node), 1); + return btrfs_inc_extent_ref(trans, root, + extent_buffer_blocknr(root->node), 1); } int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct buffer_head *buf) + struct extent_buffer *buf) { u64 blocknr; - struct btrfs_node *buf_node; - struct btrfs_leaf *buf_leaf; - struct btrfs_disk_key *key; + u32 nritems; + struct btrfs_key key; struct btrfs_file_extent_item *fi; int i; int leaf; @@ -458,31 +457,31 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (!root->ref_cows) return 0; - buf_node = btrfs_buffer_node(buf); - leaf = btrfs_is_leaf(buf_node); - buf_leaf = btrfs_buffer_leaf(buf); - for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) { + + leaf = btrfs_is_leaf(buf); + nritems = btrfs_header_nritems(buf); + for (i = 0; i < nritems; i++) { if (leaf) { u64 disk_blocknr; - key = &buf_leaf->items[i].key; - if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY) + btrfs_item_key_to_cpu(buf, &key, i); + if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; - fi = btrfs_item_ptr(buf_leaf, i, + fi = btrfs_item_ptr(buf, i, struct btrfs_file_extent_item); - if (btrfs_file_extent_type(fi) == + if (btrfs_file_extent_type(buf, fi) == BTRFS_FILE_EXTENT_INLINE) continue; - disk_blocknr = btrfs_file_extent_disk_blocknr(fi); + disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi); if (disk_blocknr == 0) continue; ret = btrfs_inc_extent_ref(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(fi)); + btrfs_file_extent_disk_num_blocks(buf, fi)); if (ret) { faili = i; goto fail; } } else { - blocknr = btrfs_node_blockptr(buf_node, i); + blocknr = btrfs_node_blockptr(buf, i); ret = btrfs_inc_extent_ref(trans, root, blocknr, 1); if (ret) { faili = i; @@ -496,22 +495,23 @@ fail: for (i =0; i < faili; i++) { if (leaf) { u64 disk_blocknr; - key = &buf_leaf->items[i].key; - if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY) + btrfs_item_key_to_cpu(buf, &key, i); + if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; - fi = btrfs_item_ptr(buf_leaf, i, + fi = btrfs_item_ptr(buf, i, struct btrfs_file_extent_item); - if (btrfs_file_extent_type(fi) == + if (btrfs_file_extent_type(buf, fi) == BTRFS_FILE_EXTENT_INLINE) continue; - disk_blocknr = btrfs_file_extent_disk_blocknr(fi); + disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi); if (disk_blocknr == 0) continue; err = btrfs_free_extent(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(fi), 0); + btrfs_file_extent_disk_num_blocks(buf, + fi), 0); BUG_ON(err); } else { - blocknr = btrfs_node_blockptr(buf_node, i); + blocknr = btrfs_node_blockptr(buf, i); err = btrfs_free_extent(trans, root, blocknr, 1, 0); BUG_ON(err); } @@ -527,16 +527,18 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans, int ret; int pending_ret; struct btrfs_root *extent_root = root->fs_info->extent_root; - struct btrfs_block_group_item *bi; + unsigned long bi; + struct extent_buffer *leaf; ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); if (ret < 0) goto fail; BUG_ON(ret); - bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], - struct btrfs_block_group_item); - memcpy(bi, &cache->item, sizeof(*bi)); - btrfs_mark_buffer_dirty(path->nodes[0]); + + leaf = path->nodes[0]; + bi = btrfs_item_ptr_offset(leaf, path->slots[0]); + write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); + btrfs_mark_buffer_dirty(leaf); btrfs_release_path(extent_root, path); fail: finish_current_insert(trans, extent_root); @@ -768,11 +770,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct unsigned long gang[8]; struct btrfs_fs_info *info = extent_root->fs_info; - btrfs_set_extent_refs(&extent_item, 1); + btrfs_set_stack_extent_refs(&extent_item, 1); ins.offset = 1; - ins.flags = 0; btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); - btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid); + btrfs_set_stack_extent_owner(&extent_item, + extent_root->root_key.objectid); while(1) { ret = find_first_radix_bit(&info->extent_ins_radix, gang, 0, @@ -795,23 +797,20 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending) { int err; - struct btrfs_header *header; - struct buffer_head *bh; + struct extent_buffer *buf; if (!pending) { - bh = btrfs_find_tree_block(root, blocknr); - if (bh) { - if (buffer_uptodate(bh)) { + buf = btrfs_find_tree_block(root, blocknr); + if (buf) { + if (btrfs_buffer_uptodate(buf)) { u64 transid = root->fs_info->running_transaction->transid; - header = btrfs_buffer_header(bh); - if (btrfs_header_generation(header) == - transid) { - btrfs_block_release(root, bh); + if (btrfs_header_generation(buf) == transid) { + free_extent_buffer(buf); return 0; } } - btrfs_block_release(root, bh); + free_extent_buffer(buf); } err = set_radix_bit(&root->fs_info->pinned_radix, blocknr); if (!err) { @@ -839,12 +838,12 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_key key; struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; + struct extent_buffer *leaf; int ret; struct btrfs_extent_item *ei; u32 refs; key.objectid = blocknr; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); key.offset = num_blocks; @@ -856,12 +855,16 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root if (ret < 0) return ret; BUG_ON(ret); - ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], + + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); - BUG_ON(ei->refs == 0); - refs = btrfs_extent_refs(ei) - 1; - btrfs_set_extent_refs(ei, refs); - btrfs_mark_buffer_dirty(path->nodes[0]); + refs = btrfs_extent_refs(leaf, ei); + BUG_ON(refs == 0); + refs -= 1; + btrfs_set_extent_refs(leaf, ei, refs); + btrfs_mark_buffer_dirty(leaf); + if (refs == 0) { u64 super_blocks_used, root_blocks_used; @@ -876,8 +879,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root super_blocks_used - num_blocks); /* block accounting for root item */ - root_blocks_used = btrfs_root_blocks_used(&root->root_item); - btrfs_set_root_blocks_used(&root->root_item, + root_blocks_used = btrfs_root_used(&root->root_item); + btrfs_set_root_used(&root->root_item, root_blocks_used - num_blocks); ret = btrfs_del_item(trans, extent_root, path); @@ -984,7 +987,7 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root u64 test_block; u64 orig_search_start = search_start; int start_found; - struct btrfs_leaf *l; + struct extent_buffer *l; struct btrfs_root * root = orig_root->fs_info->extent_root; struct btrfs_fs_info *info = root->fs_info; int total_needed = num_blocks; @@ -994,10 +997,10 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root int wrapped = 0; WARN_ON(num_blocks < 1); - ins->flags = 0; btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); - level = btrfs_header_level(btrfs_buffer_header(root->node)); + level = btrfs_header_level(root->node); + if (search_end == (u64)-1) search_end = btrfs_super_total_blocks(&info->super_copy); if (hint_block) { @@ -1034,8 +1037,9 @@ check_failed: path->slots[0]--; } - l = btrfs_buffer_leaf(path->nodes[0]); - btrfs_disk_key_to_cpu(&key, &l->items[path->slots[0]].key); + l = path->nodes[0]; + btrfs_item_key_to_cpu(l, &key, path->slots[0]); + /* * a rare case, go back one key if we hit a block group item * instead of an extent item @@ -1055,9 +1059,9 @@ check_failed: } while (1) { - l = btrfs_buffer_leaf(path->nodes[0]); + l = path->nodes[0]; slot = path->slots[0]; - if (slot >= btrfs_header_nritems(&l->header)) { + if (slot >= btrfs_header_nritems(l)) { ret = btrfs_next_leaf(root, path); if (ret == 0) continue; @@ -1075,7 +1079,7 @@ check_failed: goto check_pending; } - btrfs_disk_key_to_cpu(&key, &l->items[slot].key); + btrfs_item_key_to_cpu(l, &key, slot); if (key.objectid >= search_start && key.objectid > last_block && start_found) { if (last_block < search_start) @@ -1183,8 +1187,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root = info->extent_root; struct btrfs_extent_item extent_item; - btrfs_set_extent_refs(&extent_item, 1); - btrfs_set_extent_owner(&extent_item, owner); + btrfs_set_stack_extent_refs(&extent_item, 1); + btrfs_set_stack_extent_owner(&extent_item, owner); WARN_ON(num_blocks < 1); ret = find_free_extent(trans, root, num_blocks, empty_size, @@ -1201,8 +1205,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, num_blocks); /* block accounting for root item */ - root_blocks_used = btrfs_root_blocks_used(&root->root_item); - btrfs_set_root_blocks_used(&root->root_item, root_blocks_used + + root_blocks_used = btrfs_root_used(&root->root_item); + btrfs_set_root_used(&root->root_item, root_blocks_used + num_blocks); if (root == extent_root) { @@ -1241,13 +1245,13 @@ update_block: * helper function to allocate a block for a given tree * returns the tree buffer or NULL. */ -struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 hint, - u64 empty_size) +struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 hint, + u64 empty_size) { struct btrfs_key ins; int ret; - struct buffer_head *buf; + struct extent_buffer *buf; ret = btrfs_alloc_extent(trans, root, root->root_key.objectid, 1, empty_size, hint, (u64)-1, &ins, 0); @@ -1260,53 +1264,57 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, btrfs_free_extent(trans, root, ins.objectid, 1, 0); return ERR_PTR(-ENOMEM); } - WARN_ON(buffer_dirty(buf)); - set_buffer_uptodate(buf); + btrfs_set_buffer_uptodate(buf); + set_extent_dirty(&trans->transaction->dirty_pages, buf->start, + buf->start + buf->len - 1, GFP_NOFS); + /* set_buffer_checked(buf); set_buffer_defrag(buf); - set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index); + */ + /* FIXME!!!!!!!!!!!!!!!! + set_radix_bit(&trans->transaction->dirty_pages, buf->pages[0]->index); + */ trans->blocks_used++; return buf; } static int drop_leaf_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct buffer_head *cur) + struct btrfs_root *root, struct extent_buffer *leaf) { - struct btrfs_disk_key *key; - struct btrfs_leaf *leaf; + struct btrfs_key key; struct btrfs_file_extent_item *fi; int i; int nritems; int ret; - BUG_ON(!btrfs_is_leaf(btrfs_buffer_node(cur))); - leaf = btrfs_buffer_leaf(cur); - nritems = btrfs_header_nritems(&leaf->header); + BUG_ON(!btrfs_is_leaf(leaf)); + nritems = btrfs_header_nritems(leaf); for (i = 0; i < nritems; i++) { u64 disk_blocknr; - key = &leaf->items[i].key; - if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY) + + btrfs_item_key_to_cpu(leaf, &key, i); + if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); - if (btrfs_file_extent_type(fi) == BTRFS_FILE_EXTENT_INLINE) + if (btrfs_file_extent_type(leaf, fi) == + BTRFS_FILE_EXTENT_INLINE) continue; /* * FIXME make sure to insert a trans record that * repeats the snapshot del on crash */ - disk_blocknr = btrfs_file_extent_disk_blocknr(fi); + disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi); if (disk_blocknr == 0) continue; ret = btrfs_free_extent(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(fi), - 0); + btrfs_file_extent_disk_num_blocks(leaf, fi), 0); BUG_ON(ret); } return 0; } static void reada_walk_down(struct btrfs_root *root, - struct btrfs_node *node) + struct extent_buffer *node) { int i; u32 nritems; @@ -1314,7 +1322,7 @@ static void reada_walk_down(struct btrfs_root *root, int ret; u32 refs; - nritems = btrfs_header_nritems(&node->header); + nritems = btrfs_header_nritems(node); for (i = 0; i < nritems; i++) { blocknr = btrfs_node_blockptr(node, i); ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs); @@ -1337,16 +1345,17 @@ static void reada_walk_down(struct btrfs_root *root, static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level) { - struct buffer_head *next; - struct buffer_head *cur; + struct extent_buffer *next; + struct extent_buffer *cur; u64 blocknr; int ret; u32 refs; WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); - ret = lookup_extent_ref(trans, root, bh_blocknr(path->nodes[*level]), - 1, &refs); + ret = lookup_extent_ref(trans, root, + extent_buffer_blocknr(path->nodes[*level]), + 1, &refs); BUG_ON(ret); if (refs > 1) goto out; @@ -1360,21 +1369,20 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root cur = path->nodes[*level]; if (*level > 0 && path->slots[*level] == 0) - reada_walk_down(root, btrfs_buffer_node(cur)); + reada_walk_down(root, cur); - if (btrfs_header_level(btrfs_buffer_header(cur)) != *level) + if (btrfs_header_level(cur) != *level) WARN_ON(1); if (path->slots[*level] >= - btrfs_header_nritems(btrfs_buffer_header(cur))) + btrfs_header_nritems(cur)) break; if (*level == 0) { ret = drop_leaf_ref(trans, root, cur); BUG_ON(ret); break; } - blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur), - path->slots[*level]); + blocknr = btrfs_node_blockptr(cur, path->slots[*level]); ret = lookup_extent_ref(trans, root, blocknr, 1, &refs); BUG_ON(ret); if (refs != 1) { @@ -1384,8 +1392,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root continue; } next = btrfs_find_tree_block(root, blocknr); - if (!next || !buffer_uptodate(next)) { - brelse(next); + if (!next || !btrfs_buffer_uptodate(next)) { + free_extent_buffer(next); mutex_unlock(&root->fs_info->fs_mutex); next = read_tree_block(root, blocknr); mutex_lock(&root->fs_info->fs_mutex); @@ -1395,7 +1403,7 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(ret); if (refs != 1) { path->slots[*level]++; - brelse(next); + free_extent_buffer(next); ret = btrfs_free_extent(trans, root, blocknr, 1, 1); BUG_ON(ret); @@ -1404,17 +1412,17 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root } WARN_ON(*level <= 0); if (path->nodes[*level-1]) - btrfs_block_release(root, path->nodes[*level-1]); + free_extent_buffer(path->nodes[*level-1]); path->nodes[*level-1] = next; - *level = btrfs_header_level(btrfs_buffer_header(next)); + *level = btrfs_header_level(next); path->slots[*level] = 0; } out: WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); ret = btrfs_free_extent(trans, root, - bh_blocknr(path->nodes[*level]), 1, 1); - btrfs_block_release(root, path->nodes[*level]); + extent_buffer_blocknr(path->nodes[*level]), 1, 1); + free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; BUG_ON(ret); @@ -1436,24 +1444,24 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { slot = path->slots[i]; - if (slot < btrfs_header_nritems( - btrfs_buffer_header(path->nodes[i])) - 1) { - struct btrfs_node *node; - node = btrfs_buffer_node(path->nodes[i]); + if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { + struct extent_buffer *node; + struct btrfs_disk_key disk_key; + node = path->nodes[i]; path->slots[i]++; *level = i; WARN_ON(*level == 0); + btrfs_node_key(node, &disk_key, path->slots[i]); memcpy(&root_item->drop_progress, - &node->ptrs[path->slots[i]].key, - sizeof(root_item->drop_progress)); + &disk_key, sizeof(disk_key)); root_item->drop_level = i; return 0; } else { ret = btrfs_free_extent(trans, root, - bh_blocknr(path->nodes[*level]), - 1, 1); + extent_buffer_blocknr(path->nodes[*level]), + 1, 1); BUG_ON(ret); - btrfs_block_release(root, path->nodes[*level]); + free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level = i + 1; } @@ -1480,15 +1488,15 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root path = btrfs_alloc_path(); BUG_ON(!path); - level = btrfs_header_level(btrfs_buffer_header(root->node)); + level = btrfs_header_level(root->node); orig_level = level; if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { path->nodes[level] = root->node; path->slots[level] = 0; } else { struct btrfs_key key; - struct btrfs_disk_key *found_key; - struct btrfs_node *node; + struct btrfs_disk_key found_key; + struct extent_buffer *node; btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); level = root_item->drop_level; @@ -1498,10 +1506,10 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root ret = wret; goto out; } - node = btrfs_buffer_node(path->nodes[level]); - found_key = &node->ptrs[path->slots[level]].key; - WARN_ON(memcmp(found_key, &root_item->drop_progress, - sizeof(*found_key))); + node = path->nodes[level]; + btrfs_node_key(node, &found_key, path->slots[level]); + WARN_ON(memcmp(&found_key, &root_item->drop_progress, + sizeof(found_key))); } while(1) { wret = walk_down_tree(trans, root, path, &level); @@ -1516,12 +1524,12 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root if (wret < 0) ret = wret; ret = -EAGAIN; - get_bh(root->node); + extent_buffer_get(root->node); break; } for (i = 0; i <= orig_level; i++) { if (path->nodes[i]) { - btrfs_block_release(root, path->nodes[i]); + free_extent_buffer(path->nodes[i]); path->nodes[i] = 0; } } @@ -1581,13 +1589,12 @@ int btrfs_read_block_groups(struct btrfs_root *root) struct btrfs_path *path; int ret; int err = 0; - struct btrfs_block_group_item *bi; struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; struct radix_tree_root *radix; struct btrfs_key key; struct btrfs_key found_key; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; u64 group_size_blocks; u64 used; @@ -1596,7 +1603,6 @@ int btrfs_read_block_groups(struct btrfs_root *root) root = info->extent_root; key.objectid = 0; key.offset = group_size_blocks; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); path = btrfs_alloc_path(); @@ -1610,18 +1616,18 @@ int btrfs_read_block_groups(struct btrfs_root *root) err = ret; break; } - leaf = btrfs_buffer_leaf(path->nodes[0]); - btrfs_disk_key_to_cpu(&found_key, - &leaf->items[path->slots[0]].key); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); cache = kmalloc(sizeof(*cache), GFP_NOFS); if (!cache) { err = -1; break; } - bi = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_block_group_item); - if (bi->flags & BTRFS_BLOCK_GROUP_DATA) { + read_extent_buffer(leaf, &cache->item, + btrfs_item_ptr_offset(leaf, path->slots[0]), + sizeof(cache->item)); + if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { radix = &info->block_group_data_radix; cache->data = 1; } else { @@ -1629,7 +1635,6 @@ int btrfs_read_block_groups(struct btrfs_root *root) cache->data = 0; } - memcpy(&cache->item, bi, sizeof(*bi)); memcpy(&cache->key, &found_key, sizeof(found_key)); cache->last_alloc = cache->key.objectid; cache->first_free = cache->key.objectid; @@ -1640,11 +1645,12 @@ int btrfs_read_block_groups(struct btrfs_root *root) key.objectid = found_key.objectid + found_key.offset; btrfs_release_path(root, path); + ret = radix_tree_insert(radix, found_key.objectid + found_key.offset - 1, (void *)cache); BUG_ON(ret); - used = btrfs_block_group_used(bi); + used = btrfs_block_group_used(&cache->item); if (used < div_factor(key.offset, 8)) { radix_tree_tag_set(radix, found_key.objectid + found_key.offset - 1, diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 0ab368e091f..55fefdfab84 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -26,16 +26,6 @@ struct tree_entry { struct rb_node rb_node; }; -/* bits for the extent state */ -#define EXTENT_DIRTY 1 -#define EXTENT_WRITEBACK (1 << 1) -#define EXTENT_UPTODATE (1 << 2) -#define EXTENT_LOCKED (1 << 3) -#define EXTENT_NEW (1 << 4) -#define EXTENT_DELALLOC (1 << 5) - -#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) - void __init extent_map_init(void) { extent_map_cache = btrfs_cache_create("extent_map", @@ -223,7 +213,8 @@ int add_extent_mapping(struct extent_map_tree *tree, if (rb) prev = rb_entry(rb, struct extent_map, rb_node); if (prev && prev->end + 1 == em->start && - ((em->block_start == 0 && prev->block_start == 0) || + ((em->block_start == EXTENT_MAP_HOLE && + prev->block_start == EXTENT_MAP_HOLE) || (em->block_start == prev->block_end + 1))) { em->start = prev->start; em->block_start = prev->block_start; @@ -926,6 +917,40 @@ int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(set_range_writeback); +int find_first_extent_bit(struct extent_map_tree *tree, u64 start, + u64 *start_ret, u64 *end_ret, int bits) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 1; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->state & bits) { + *start_ret = state->start; + *end_ret = state->end; + ret = 0; + } + node = rb_next(node); + if (!node) + break; + } +out: + write_unlock_irq(&tree->lock); + return ret; +} +EXPORT_SYMBOL(find_first_extent_bit); + u64 find_lock_delalloc_range(struct extent_map_tree *tree, u64 start, u64 lock_start, u64 *end, u64 max_bytes) { @@ -1450,7 +1475,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, em = NULL; /* we've found a hole, just zero and go on */ - if (block_start == 0) { + if (block_start == EXTENT_MAP_HOLE) { zero_user_page(page, page_offset, iosize, KM_USER0); set_extent_uptodate(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -1593,7 +1618,8 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, free_extent_map(em); em = NULL; - if (block_start == 0 || block_start == EXTENT_MAP_INLINE) { + if (block_start == EXTENT_MAP_HOLE || + block_start == EXTENT_MAP_INLINE) { clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); cur = cur + iosize; @@ -1630,7 +1656,6 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, nr++; } done: - WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0)); unlock_extent(tree, start, page_end, GFP_NOFS); unlock_page(page); return 0; @@ -1827,8 +1852,623 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock, // XXX(hch): block 0 is valid in some cases, e.g. XFS RT device if (em->block_start == EXTENT_MAP_INLINE || - em->block_start == 0) - return 0; + em->block_start == EXTENT_MAP_HOLE) + return 0; return (em->block_start + start - em->start) >> inode->i_blkbits; } + +struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, + u64 start, unsigned long len, + gfp_t mask) +{ + unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT) + 1; + unsigned long i; + unsigned long index = start >> PAGE_CACHE_SHIFT; + struct extent_buffer *eb; + struct page *p; + struct address_space *mapping = tree->mapping; + int uptodate = 0; + + eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask); + if (!eb || IS_ERR(eb)) + return NULL; + + eb->start = start; + eb->len = len; + atomic_set(&eb->refs, 1); + + for (i = 0; i < num_pages; i++, index++) { + p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); + if (!p) + goto fail; + eb->pages[i] = p; + if (!PageUptodate(p)) + uptodate = 0; + unlock_page(p); + } + if (uptodate) + eb->flags |= EXTENT_UPTODATE; + return eb; +fail: + free_extent_buffer(eb); + return NULL; +} +EXPORT_SYMBOL(alloc_extent_buffer); + +struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, + u64 start, unsigned long len, + gfp_t mask) +{ + unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT) + 1; + unsigned long i; + unsigned long index = start >> PAGE_CACHE_SHIFT; + struct extent_buffer *eb; + struct page *p; + struct address_space *mapping = tree->mapping; + + eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask); + if (!eb || IS_ERR(eb)) + return NULL; + + eb->start = start; + eb->len = len; + atomic_set(&eb->refs, 1); + + for (i = 0; i < num_pages; i++, index++) { + p = find_get_page(mapping, index); + if (!p) + goto fail; + eb->pages[i] = p; + } + return eb; +fail: + free_extent_buffer(eb); + return NULL; +} +EXPORT_SYMBOL(find_extent_buffer); + +void free_extent_buffer(struct extent_buffer *eb) +{ + unsigned long i; + unsigned long num_pages; + + if (!eb) + return; + + if (!atomic_dec_and_test(&eb->refs)) + return; + + num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT) + 1; + + for (i = 0; i < num_pages; i++) { + if (eb->pages[i]) + page_cache_release(eb->pages[i]); + } + kfree(eb); +} +EXPORT_SYMBOL(free_extent_buffer); + +int clear_extent_buffer_dirty(struct extent_map_tree *tree, + struct extent_buffer *eb) +{ + int set; + unsigned long i; + unsigned long num_pages; + struct page *page; + + u64 start = eb->start; + u64 end = start + eb->len - 1; + + set = clear_extent_dirty(tree, start, end, GFP_NOFS); + num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT) + 1; + + for (i = 0; i < num_pages; i++) { + page = eb->pages[i]; + lock_page(page); + /* + * if we're on the last page or the first page and the + * block isn't aligned on a page boundary, do extra checks + * to make sure we don't clean page that is partially dirty + */ + if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || + ((i == num_pages - 1) && + ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) { + start = page->index << PAGE_CACHE_SHIFT; + end = start + PAGE_CACHE_SIZE - 1; + if (test_range_bit(tree, start, end, + EXTENT_DIRTY, 0)) { + unlock_page(page); + continue; + } + } + clear_page_dirty_for_io(page); + unlock_page(page); + } + return 0; +} +EXPORT_SYMBOL(clear_extent_buffer_dirty); + +int wait_on_extent_buffer_writeback(struct extent_map_tree *tree, + struct extent_buffer *eb) +{ + return wait_on_extent_writeback(tree, eb->start, + eb->start + eb->len - 1); +} +EXPORT_SYMBOL(wait_on_extent_buffer_writeback); + +int set_extent_buffer_dirty(struct extent_map_tree *tree, + struct extent_buffer *eb) +{ + return set_range_dirty(tree, eb->start, eb->start + eb->len - 1); +} +EXPORT_SYMBOL(set_extent_buffer_dirty); + +int set_extent_buffer_uptodate(struct extent_map_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + struct page *page; + unsigned long num_pages; + + num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT) + 1; + + set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, + GFP_NOFS); + for (i = 0; i < num_pages; i++) { + page = eb->pages[i]; + if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || + ((i == num_pages - 1) && + ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) { + check_page_uptodate(tree, page); + continue; + } + SetPageUptodate(page); + } + return 0; +} +EXPORT_SYMBOL(set_extent_buffer_uptodate); + +int extent_buffer_uptodate(struct extent_map_tree *tree, + struct extent_buffer *eb) +{ + if (eb->flags & EXTENT_UPTODATE) + return 1; + return test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1); +} +EXPORT_SYMBOL(extent_buffer_uptodate); + +int read_extent_buffer_pages(struct extent_map_tree *tree, + struct extent_buffer *eb, int wait) +{ + unsigned long i; + struct page *page; + int err; + int ret = 0; + unsigned long num_pages; + + if (eb->flags & EXTENT_UPTODATE) + return 0; + + if (test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1)) { + return 0; + } + + num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT) + 1; + for (i = 0; i < num_pages; i++) { + page = eb->pages[i]; + if (PageUptodate(page)) { + continue; + } + if (!wait) { + if (TestSetPageLocked(page)) { + continue; + } + } else { + lock_page(page); + } + if (!PageUptodate(page)) { + err = page->mapping->a_ops->readpage(NULL, page); + if (err) { + ret = err; + } + } else { + unlock_page(page); + } + } + + if (ret || !wait) { + return ret; + } + + for (i = 0; i < num_pages; i++) { + page = eb->pages[i]; + wait_on_page_locked(page); + if (!PageUptodate(page)) { + ret = -EIO; + } + } + eb->flags |= EXTENT_UPTODATE; + return ret; +} +EXPORT_SYMBOL(read_extent_buffer_pages); + +void read_extent_buffer(struct extent_buffer *eb, void *dstv, + unsigned long start, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *dst = (char *)dstv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + page = eb->pages[i]; + offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1); + if (i == 0) + offset += start_offset; + + while(len > 0) { + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + // kaddr = kmap_atomic(page, KM_USER0); + kaddr = page_address(page); + memcpy(dst, kaddr + offset, cur); + // kunmap_atomic(kaddr, KM_USER0); + + dst += cur; + len -= cur; + offset = 0; + i++; + page = eb->pages[i]; + } +} +EXPORT_SYMBOL(read_extent_buffer); + +int map_extent_buffer(struct extent_buffer *eb, unsigned long start, + char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + size_t offset; + char *kaddr; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + + if (i == 0) { + offset = start_offset; + *map_start = 0; + } else { + offset = 0; + *map_start = (i << PAGE_CACHE_SHIFT) - offset; + } + + // kaddr = kmap_atomic(eb->pages[i], km); + kaddr = page_address(eb->pages[i]); + *token = kaddr; + *map = kaddr + offset; + *map_len = PAGE_CACHE_SIZE - offset; + return 0; +} +EXPORT_SYMBOL(map_extent_buffer); + +void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) +{ + // kunmap_atomic(token, km); +} +EXPORT_SYMBOL(unmap_extent_buffer); + +int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, + unsigned long start, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *ptr = (char *)ptrv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + int ret = 0; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + page = eb->pages[i]; + offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1); + if (i == 0) + offset += start_offset; + + while(len > 0) { + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + + // kaddr = kmap_atomic(page, KM_USER0); + kaddr = page_address(page); + ret = memcmp(ptr, kaddr + offset, cur); + // kunmap_atomic(kaddr, KM_USER0); + if (ret) + break; + + ptr += cur; + len -= cur; + offset = 0; + i++; + page = eb->pages[i]; + } + return ret; +} +EXPORT_SYMBOL(memcmp_extent_buffer); + +void write_extent_buffer(struct extent_buffer *eb, const void *srcv, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *src = (char *)srcv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + page = eb->pages[i]; + offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1); + if (i == 0) + offset += start_offset; + + while(len > 0) { + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + // kaddr = kmap_atomic(page, KM_USER0); + kaddr = page_address(page); + memcpy(kaddr + offset, src, cur); + // kunmap_atomic(kaddr, KM_USER0); + + src += cur; + len -= cur; + offset = 0; + i++; + page = eb->pages[i]; + } +} +EXPORT_SYMBOL(write_extent_buffer); + +void memset_extent_buffer(struct extent_buffer *eb, char c, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + page = eb->pages[i]; + offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1); + if (i == 0) + offset += start_offset; + + while(len > 0) { + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + // kaddr = kmap_atomic(page, KM_USER0); + kaddr = page_address(page); + memset(kaddr + offset, c, cur); + // kunmap_atomic(kaddr, KM_USER0); + + len -= cur; + offset = 0; + i++; + page = eb->pages[i]; + } +} +EXPORT_SYMBOL(memset_extent_buffer); + +void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, + unsigned long dst_offset, unsigned long src_offset, + unsigned long len) +{ + u64 dst_len = dst->len; + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + + WARN_ON(src->len != dst_len); + + offset = dst_offset & ((unsigned long)PAGE_CACHE_SIZE - 1); + if (i == 0) + offset += start_offset; + + while(len > 0) { + page = dst->pages[i]; + WARN_ON(!PageUptodate(page)); + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset)); + + // kaddr = kmap_atomic(page, KM_USER1); + kaddr = page_address(page); + read_extent_buffer(src, kaddr + offset, src_offset, cur); + // kunmap_atomic(kaddr, KM_USER1); + + src_offset += cur; + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(copy_extent_buffer); + +static void move_pages(struct page *dst_page, struct page *src_page, + unsigned long dst_off, unsigned long src_off, + unsigned long len) +{ + // char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); + char *dst_kaddr = page_address(dst_page); + if (dst_page == src_page) { + memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len); + } else { + // char *src_kaddr = kmap_atomic(src_page, KM_USER1); + char *src_kaddr = page_address(src_page); + char *p = dst_kaddr + dst_off + len; + char *s = src_kaddr + src_off + len; + + while (len--) + *--p = *--s; + + // kunmap_atomic(src_kaddr, KM_USER1); + } + // kunmap_atomic(dst_kaddr, KM_USER0); +} + +static void copy_pages(struct page *dst_page, struct page *src_page, + unsigned long dst_off, unsigned long src_off, + unsigned long len) +{ + //kmap_atomic(dst_page, KM_USER0); + char *dst_kaddr = page_address(dst_page); + char *src_kaddr; + + if (dst_page != src_page) + src_kaddr = page_address(src_page); // kmap_atomic(src_page, KM_USER1); + else + src_kaddr = dst_kaddr; + + memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len); + /* + kunmap_atomic(dst_kaddr, KM_USER0); + if (dst_page != src_page) + kunmap_atomic(src_kaddr, KM_USER1); + */ +} + +void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + unsigned long src_i; + + if (src_offset + len > dst->len) { + printk("memmove bogus src_offset %lu move len %lu len %lu\n", + src_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset + len > dst->len) { + printk("memmove bogus dst_offset %lu move len %lu len %lu\n", + dst_offset, len, dst->len); + BUG_ON(1); + } + + while(len > 0) { + dst_off_in_page = dst_offset & + ((unsigned long)PAGE_CACHE_SIZE - 1); + src_off_in_page = src_offset & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT; + + if (src_i == 0) + src_off_in_page += start_offset; + if (dst_i == 0) + dst_off_in_page += start_offset; + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - + src_off_in_page)); + cur = min(cur, (unsigned long)(PAGE_CACHE_SIZE - + dst_off_in_page)); + + copy_pages(dst->pages[dst_i], dst->pages[src_i], + dst_off_in_page, src_off_in_page, cur); + + src_offset += cur; + dst_offset += cur; + len -= cur; + } +} +EXPORT_SYMBOL(memcpy_extent_buffer); + +void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + unsigned long dst_end = dst_offset + len - 1; + unsigned long src_end = src_offset + len - 1; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + unsigned long src_i; + + if (src_offset + len > dst->len) { + printk("memmove bogus src_offset %lu move len %lu len %lu\n", + src_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset + len > dst->len) { + printk("memmove bogus dst_offset %lu move len %lu len %lu\n", + dst_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset < src_offset) { + memcpy_extent_buffer(dst, dst_offset, src_offset, len); + return; + } + while(len > 0) { + dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; + + dst_off_in_page = dst_end & + ((unsigned long)PAGE_CACHE_SIZE - 1); + src_off_in_page = src_end & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + if (src_i == 0) + src_off_in_page += start_offset; + if (dst_i == 0) + dst_off_in_page += start_offset; + + cur = min(len, src_off_in_page + 1); + cur = min(cur, dst_off_in_page + 1); +// printk("move pages orig dst %lu src %lu len %lu, this %lu %lu %lu\n", dst_offset, src_offset, len, dst_off_in_page - cur + 1, src_off_in_page - cur + 1, cur); + move_pages(dst->pages[dst_i], dst->pages[src_i], + dst_off_in_page - cur + 1, + src_off_in_page - cur + 1, cur); + + dst_end -= cur - 1; + src_end -= cur - 1; + len -= cur; + } +} +EXPORT_SYMBOL(memmove_extent_buffer); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index d913ce01248..430b997a70f 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -3,9 +3,20 @@ #include +#define EXTENT_MAP_HOLE (u64)-3 #define EXTENT_MAP_INLINE (u64)-2 #define EXTENT_MAP_DELALLOC (u64)-1 +/* bits for the extent state */ +#define EXTENT_DIRTY 1 +#define EXTENT_WRITEBACK (1 << 1) +#define EXTENT_UPTODATE (1 << 2) +#define EXTENT_LOCKED (1 << 3) +#define EXTENT_NEW (1 << 4) +#define EXTENT_DELALLOC (1 << 5) +#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) + + struct extent_map_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); @@ -51,10 +62,13 @@ struct extent_state { struct list_head list; }; +#define EXTENT_BUFFER_SIZE(nr) (sizeof(struct extent_buffer) + \ + (nr - 1) * sizeof(struct page *)) struct extent_buffer { u64 start; - u64 end; /* inclusive */ - char *addr; + unsigned long len; + atomic_t refs; + int flags; struct page *pages[]; }; @@ -87,8 +101,12 @@ int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); +int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, + gfp_t mask); int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); +int find_first_extent_bit(struct extent_map_tree *tree, u64 start, + u64 *start_ret, u64 *end_ret, int bits); int extent_invalidatepage(struct extent_map_tree *tree, struct page *page, unsigned long offset); int extent_write_full_page(struct extent_map_tree *tree, struct page *page, @@ -106,4 +124,57 @@ int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); int set_state_private(struct extent_map_tree *tree, u64 start, u64 private); int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private); void set_page_extent_mapped(struct page *page); + +struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, + u64 start, unsigned long len, + gfp_t mask); +struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, + u64 start, unsigned long len, + gfp_t mask); +void free_extent_buffer(struct extent_buffer *eb); +int read_extent_buffer_pages(struct extent_map_tree *tree, + struct extent_buffer *eb, int wait); + +static inline void extent_buffer_get(struct extent_buffer *eb) +{ + atomic_inc(&eb->refs); +} + +static inline u64 extent_buffer_blocknr(struct extent_buffer *eb) +{ + return eb->start / 4096; +} + +int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, + unsigned long start, + unsigned long len); +void read_extent_buffer(struct extent_buffer *eb, void *dst, + unsigned long start, + unsigned long len); +void write_extent_buffer(struct extent_buffer *eb, const void *src, + unsigned long start, unsigned long len); +void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, + unsigned long dst_offset, unsigned long src_offset, + unsigned long len); +void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len); +void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len); +void memset_extent_buffer(struct extent_buffer *eb, char c, + unsigned long start, unsigned long len); +int wait_on_extent_buffer_writeback(struct extent_map_tree *tree, + struct extent_buffer *eb); +int clear_extent_buffer_dirty(struct extent_map_tree *tree, + struct extent_buffer *eb); +int set_extent_buffer_dirty(struct extent_map_tree *tree, + struct extent_buffer *eb); +int set_extent_buffer_uptodate(struct extent_map_tree *tree, + struct extent_buffer *eb); +int extent_buffer_uptodate(struct extent_map_tree *tree, + struct extent_buffer *eb); +int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, + char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km); +void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); #endif diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index ea4dd4c5fce..226f6d028c3 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -34,12 +34,12 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_file_extent_item *item; struct btrfs_key file_key; struct btrfs_path *path; + struct extent_buffer *leaf; path = btrfs_alloc_path(); BUG_ON(!path); file_key.objectid = objectid; file_key.offset = pos; - file_key.flags = 0; btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY); ret = btrfs_insert_empty_item(trans, root, path, &file_key, @@ -47,15 +47,16 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, if (ret < 0) goto out; BUG_ON(ret); - item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], + leaf = path->nodes[0]; + item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_disk_blocknr(item, offset); - btrfs_set_file_extent_disk_num_blocks(item, disk_num_blocks); - btrfs_set_file_extent_offset(item, 0); - btrfs_set_file_extent_num_blocks(item, num_blocks); - btrfs_set_file_extent_generation(item, trans->transid); - btrfs_set_file_extent_type(item, BTRFS_FILE_EXTENT_REG); - btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_set_file_extent_disk_blocknr(leaf, item, offset); + btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks); + btrfs_set_file_extent_offset(leaf, item, 0); + btrfs_set_file_extent_num_blocks(leaf, item, num_blocks); + btrfs_set_file_extent_generation(leaf, item, trans->transid); + btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG); + btrfs_mark_buffer_dirty(leaf); out: btrfs_free_path(path); return ret; @@ -71,32 +72,30 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_key file_key; struct btrfs_key found_key; struct btrfs_csum_item *item; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; u64 csum_offset = 0; int csums_in_item; file_key.objectid = objectid; file_key.offset = offset; - file_key.flags = 0; btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow); if (ret < 0) goto fail; - leaf = btrfs_buffer_leaf(path->nodes[0]); + leaf = path->nodes[0]; if (ret > 0) { ret = 1; if (path->slots[0] == 0) goto fail; path->slots[0]--; - btrfs_disk_key_to_cpu(&found_key, - &leaf->items[path->slots[0]].key); + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || found_key.objectid != objectid) { goto fail; } csum_offset = (offset - found_key.offset) >> root->fs_info->sb->s_blocksize_bits; - csums_in_item = btrfs_item_size(leaf->items + path->slots[0]); + csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]); csums_in_item /= BTRFS_CRC32_SIZE; if (csum_offset >= csums_in_item) { @@ -127,7 +126,6 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, file_key.objectid = objectid; file_key.offset = offset; - file_key.flags = 0; btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY); ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow); return ret; @@ -138,12 +136,14 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, u64 objectid, u64 offset, char *data, size_t len) { + return 0; +#if 0 int ret; struct btrfs_key file_key; struct btrfs_key found_key; struct btrfs_path *path; struct btrfs_csum_item *item; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; u64 csum_offset; path = btrfs_alloc_path(); @@ -161,8 +161,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, if (ret == -EFBIG) { u32 item_size; /* we found one, but it isn't big enough yet */ - leaf = btrfs_buffer_leaf(path->nodes[0]); - item_size = btrfs_item_size(leaf->items + path->slots[0]); + leaf = path->nodes[0]; + item_size = btrfs_item_size_nr(leaf, path->slots[0]); if ((item_size / BTRFS_CRC32_SIZE) >= MAX_CSUM_ITEMS(root)) { /* already at max size, make a new one */ goto insert; @@ -188,8 +188,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, goto insert; } path->slots[0]--; - leaf = btrfs_buffer_leaf(path->nodes[0]); - btrfs_disk_key_to_cpu(&found_key, &leaf->items[path->slots[0]].key); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); csum_offset = (offset - found_key.offset) >> root->fs_info->sb->s_blocksize_bits; if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || @@ -197,10 +197,10 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, csum_offset >= MAX_CSUM_ITEMS(root)) { goto insert; } - if (csum_offset >= btrfs_item_size(leaf->items + path->slots[0]) / + if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) / BTRFS_CRC32_SIZE) { u32 diff = (csum_offset + 1) * BTRFS_CRC32_SIZE; - diff = diff - btrfs_item_size(leaf->items + path->slots[0]); + diff = diff - btrfs_item_size_nr(leaf, path->slots[0]); if (diff != BTRFS_CRC32_SIZE) goto insert; ret = btrfs_extend_item(trans, root, path, diff); @@ -220,21 +220,20 @@ insert: goto fail; } csum: - item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], - struct btrfs_csum_item); + leaf = path->nodes[0]; + item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item); ret = 0; item = (struct btrfs_csum_item *)((unsigned char *)item + csum_offset * BTRFS_CRC32_SIZE); found: - btrfs_check_bounds(&item->csum, BTRFS_CRC32_SIZE, - path->nodes[0]->b_data, - root->fs_info->sb->s_blocksize); + /* FIXME!!!!!!!!!!!! */ ret = btrfs_csum_data(root, data, len, &item->csum); btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_release_path(root, path); btrfs_free_path(path); return ret; +#endif } int btrfs_csum_truncate(struct btrfs_trans_handle *trans, @@ -242,21 +241,21 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, u64 isize) { struct btrfs_key key; - struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[0]); + struct extent_buffer *leaf = path->nodes[0]; int slot = path->slots[0]; int ret; u32 new_item_size; u64 new_item_span; u64 blocks; - btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key); + btrfs_item_key_to_cpu(leaf, &key, slot); if (isize <= key.offset) return 0; new_item_span = isize - key.offset; - blocks = (new_item_span + root->blocksize - 1) >> + blocks = (new_item_span + root->sectorsize - 1) >> root->fs_info->sb->s_blocksize_bits; new_item_size = blocks * BTRFS_CRC32_SIZE; - if (new_item_size >= btrfs_item_size(leaf->items + slot)) + if (new_item_size >= btrfs_item_size_nr(leaf, slot)) return 0; ret = btrfs_truncate_item(trans, root, path, new_item_size); BUG_ON(ret); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 4cc459c943e..1734ca69555 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -16,7 +16,6 @@ * Boston, MA 021110-1307, USA. */ -#include #include #include #include @@ -88,7 +87,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, { struct btrfs_key key; struct btrfs_path *path; - char *ptr, *kaddr; + struct extent_buffer *leaf; + char *kaddr; + unsigned long ptr; struct btrfs_file_extent_item *ei; u32 datasize; int err = 0; @@ -102,7 +103,6 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, key.objectid = inode->i_ino; key.offset = offset; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); BUG_ON(size >= PAGE_CACHE_SIZE); datasize = btrfs_file_extent_calc_inline_size(size); @@ -113,18 +113,17 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, err = ret; goto fail; } - ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(ei, trans->transid); - btrfs_set_file_extent_type(ei, - BTRFS_FILE_EXTENT_INLINE); + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, ei, trans->transid); + btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); kaddr = kmap_atomic(page, KM_USER0); - btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, kaddr + page_offset, size); + write_extent_buffer(leaf, kaddr + page_offset, ptr, size); kunmap_atomic(kaddr, KM_USER0); - btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(leaf); fail: btrfs_free_path(path); return err; @@ -156,8 +155,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, em->bdev = inode->i_sb->s_bdev; - start_pos = pos & ~((u64)root->blocksize - 1); - num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> + start_pos = pos & ~((u64)root->sectorsize - 1); + num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >> inode->i_blkbits; down_read(&BTRFS_I(inode)->root->snap_sem); @@ -184,7 +183,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, if (inode->i_size < start_pos) { u64 last_pos_in_file; u64 hole_size; - u64 mask = root->blocksize - 1; + u64 mask = root->sectorsize - 1; last_pos_in_file = (isize + mask) & ~mask; hole_size = (start_pos - last_pos_in_file + mask) & ~mask; @@ -227,8 +226,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, /* step one, delete the existing extents in this range */ /* FIXME blocksize != pagesize */ err = btrfs_drop_extents(trans, root, inode, start_pos, - (pos + write_bytes + root->blocksize -1) & - ~((u64)root->blocksize - 1), &hint_block); + (pos + write_bytes + root->sectorsize -1) & + ~((u64)root->sectorsize - 1), &hint_block); if (err) goto failed; @@ -288,7 +287,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, { int ret; struct btrfs_key key; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; int slot; struct btrfs_file_extent_item *extent; u64 extent_end = 0; @@ -327,10 +326,10 @@ next_slot: found_extent = 0; found_inline = 0; extent = NULL; - leaf = btrfs_buffer_leaf(path->nodes[0]); + leaf = path->nodes[0]; slot = path->slots[0]; ret = 0; - btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key); + btrfs_item_key_to_cpu(leaf, &key, slot); if (key.offset >= end || key.objectid != inode->i_ino) { goto out; } @@ -344,17 +343,18 @@ next_slot: if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); - found_type = btrfs_file_extent_type(extent); + found_type = btrfs_file_extent_type(leaf, extent); if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = key.offset + - (btrfs_file_extent_num_blocks(extent) << + (btrfs_file_extent_num_blocks(leaf, extent) << inode->i_blkbits); found_extent = 1; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + struct btrfs_item *item; + item = btrfs_item_nr(leaf, slot); found_inline = 1; extent_end = key.offset + - btrfs_file_extent_inline_len(leaf->items + - slot); + btrfs_file_extent_inline_len(leaf, item); } } else { extent_end = search_start; @@ -365,8 +365,7 @@ next_slot: search_start >= extent_end) { int nextret; u32 nritems; - nritems = btrfs_header_nritems( - btrfs_buffer_header(path->nodes[0])); + nritems = btrfs_header_nritems(leaf); if (slot >= nritems - 1) { nextret = btrfs_next_leaf(root, path); if (nextret) @@ -380,7 +379,7 @@ next_slot: /* FIXME, there's only one inline extent allowed right now */ if (found_inline) { - u64 mask = root->blocksize - 1; + u64 mask = root->sectorsize - 1; search_start = (extent_end + mask) & ~mask; } else search_start = extent_end; @@ -388,10 +387,13 @@ next_slot: if (end < extent_end && end >= key.offset) { if (found_extent) { u64 disk_blocknr = - btrfs_file_extent_disk_blocknr(extent); + btrfs_file_extent_disk_blocknr(leaf,extent); u64 disk_num_blocks = - btrfs_file_extent_disk_num_blocks(extent); - memcpy(&old, extent, sizeof(old)); + btrfs_file_extent_disk_num_blocks(leaf, + extent); + read_extent_buffer(leaf, &old, + (unsigned long)extent, + sizeof(old)); if (disk_blocknr != 0) { ret = btrfs_inc_extent_ref(trans, root, disk_blocknr, disk_num_blocks); @@ -406,20 +408,24 @@ next_slot: u64 new_num; u64 old_num; keep = 1; - WARN_ON(start & (root->blocksize - 1)); + WARN_ON(start & (root->sectorsize - 1)); if (found_extent) { new_num = (start - key.offset) >> inode->i_blkbits; - old_num = btrfs_file_extent_num_blocks(extent); + old_num = btrfs_file_extent_num_blocks(leaf, + extent); *hint_block = - btrfs_file_extent_disk_blocknr(extent); - if (btrfs_file_extent_disk_blocknr(extent)) { + btrfs_file_extent_disk_blocknr(leaf, + extent); + if (btrfs_file_extent_disk_blocknr(leaf, + extent)) { inode->i_blocks -= (old_num - new_num) << 3; } - btrfs_set_file_extent_num_blocks(extent, + btrfs_set_file_extent_num_blocks(leaf, + extent, new_num); - btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(leaf); } else { WARN_ON(1); } @@ -431,13 +437,17 @@ next_slot: u64 extent_num_blocks = 0; if (found_extent) { disk_blocknr = - btrfs_file_extent_disk_blocknr(extent); + btrfs_file_extent_disk_blocknr(leaf, + extent); disk_num_blocks = - btrfs_file_extent_disk_num_blocks(extent); + btrfs_file_extent_disk_num_blocks(leaf, + extent); extent_num_blocks = - btrfs_file_extent_num_blocks(extent); + btrfs_file_extent_num_blocks(leaf, + extent); *hint_block = - btrfs_file_extent_disk_blocknr(extent); + btrfs_file_extent_disk_blocknr(leaf, + extent); } ret = btrfs_del_item(trans, root, path); /* TODO update progress marker and return */ @@ -464,42 +474,37 @@ next_slot: struct btrfs_key ins; ins.objectid = inode->i_ino; ins.offset = end; - ins.flags = 0; btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); btrfs_release_path(root, path); ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*extent)); + leaf = path->nodes[0]; if (ret) { - btrfs_print_leaf(root, btrfs_buffer_leaf(path->nodes[0])); - printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end, keep); + btrfs_print_leaf(root, leaf); + printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.type, ins.offset, start, end, key.offset, extent_end, keep); } BUG_ON(ret); - extent = btrfs_item_ptr( - btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], - struct btrfs_file_extent_item); - btrfs_set_file_extent_disk_blocknr(extent, - btrfs_file_extent_disk_blocknr(&old)); - btrfs_set_file_extent_disk_num_blocks(extent, - btrfs_file_extent_disk_num_blocks(&old)); - - btrfs_set_file_extent_offset(extent, - btrfs_file_extent_offset(&old) + + extent = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + write_extent_buffer(leaf, &old, + (unsigned long)extent, sizeof(old)); + + btrfs_set_file_extent_offset(leaf, extent, + le64_to_cpu(old.offset) + ((end - key.offset) >> inode->i_blkbits)); - WARN_ON(btrfs_file_extent_num_blocks(&old) < + WARN_ON(le64_to_cpu(old.num_blocks) < (extent_end - end) >> inode->i_blkbits); - btrfs_set_file_extent_num_blocks(extent, + btrfs_set_file_extent_num_blocks(leaf, extent, (extent_end - end) >> inode->i_blkbits); - btrfs_set_file_extent_type(extent, + btrfs_set_file_extent_type(leaf, extent, BTRFS_FILE_EXTENT_REG); - btrfs_set_file_extent_generation(extent, - btrfs_file_extent_generation(&old)); btrfs_mark_buffer_dirty(path->nodes[0]); - if (btrfs_file_extent_disk_blocknr(&old) != 0) { + if (le64_to_cpu(old.disk_blocknr) != 0) { inode->i_blocks += - btrfs_file_extent_num_blocks(extent) << 3; + btrfs_file_extent_num_blocks(leaf, + extent) << 3; } ret = 0; goto out; @@ -529,8 +534,8 @@ static int prepare_pages(struct btrfs_root *root, u64 num_blocks; u64 start_pos; - start_pos = pos & ~((u64)root->blocksize - 1); - num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >> + start_pos = pos & ~((u64)root->sectorsize - 1); + num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >> inode->i_blkbits; memset(pages, 0, num_pages * sizeof(struct page *)); diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index b7834617753..35d2608f891 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -20,24 +20,18 @@ #include "disk-io.h" #include "transaction.h" -int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 objectid, struct btrfs_inode_item - *inode_item) +int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 objectid) { - struct btrfs_path *path; struct btrfs_key key; int ret; key.objectid = objectid; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); key.offset = 0; - path = btrfs_alloc_path(); - BUG_ON(!path); - ret = btrfs_insert_item(trans, root, &key, inode_item, - sizeof(*inode_item)); - btrfs_release_path(root, path); - btrfs_free_path(path); + ret = btrfs_insert_empty_item(trans, root, path, &key, + sizeof(struct btrfs_inode_item)); if (ret == 0 && objectid > root->highest_inode) root->highest_inode = objectid; return ret; @@ -51,15 +45,15 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root int cow = mod != 0; int ret; int slot; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; struct btrfs_key found_key; ret = btrfs_search_slot(trans, root, location, path, ins_len, cow); if (ret > 0 && btrfs_key_type(location) == BTRFS_ROOT_ITEM_KEY && location->offset == (u64)-1 && path->slots[0] != 0) { slot = path->slots[0] - 1; - leaf = btrfs_buffer_leaf(path->nodes[0]); - btrfs_disk_key_to_cpu(&found_key, &leaf->items[slot].key); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, slot); if (found_key.objectid == location->objectid && btrfs_key_type(&found_key) == btrfs_key_type(location)) { path->slots[0]--; diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 40547086625..ab74977adf5 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -24,8 +24,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid) { struct btrfs_path *path; int ret; - struct btrfs_leaf *l; + struct extent_buffer *l; struct btrfs_key search_key; + struct btrfs_key found_key; int slot; path = btrfs_alloc_path(); @@ -39,8 +40,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid) BUG_ON(ret == 0); if (path->slots[0] > 0) { slot = path->slots[0] - 1; - l = btrfs_buffer_leaf(path->nodes[0]); - *objectid = btrfs_disk_key_objectid(&l->items[slot].key); + l = path->nodes[0]; + btrfs_item_key_to_cpu(l, &found_key, slot); + *objectid = found_key.objectid; } else { *objectid = BTRFS_FIRST_FREE_OBJECTID; } @@ -64,13 +66,12 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, int slot = 0; u64 last_ino = 0; int start_found; - struct btrfs_leaf *l; + struct extent_buffer *l; struct btrfs_key search_key; u64 search_start = dirid; path = btrfs_alloc_path(); BUG_ON(!path); - search_key.flags = 0; search_start = root->last_inode_alloc; search_start = max(search_start, BTRFS_FIRST_FREE_OBJECTID); search_key.objectid = search_start; @@ -86,9 +87,9 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, path->slots[0]--; while (1) { - l = btrfs_buffer_leaf(path->nodes[0]); + l = path->nodes[0]; slot = path->slots[0]; - if (slot >= btrfs_header_nritems(&l->header)) { + if (slot >= btrfs_header_nritems(l)) { ret = btrfs_next_leaf(root, path); if (ret == 0) continue; @@ -103,7 +104,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, last_ino : search_start; goto found; } - btrfs_disk_key_to_cpu(&key, &l->items[slot].key); + btrfs_item_key_to_cpu(l, &key, slot); if (key.objectid >= search_start) { if (start_found) { if (last_ino < search_start) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b03d40a907c..fbe2836364e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -159,10 +159,8 @@ out: int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) { - char csum[BTRFS_CRC32_SIZE]; size_t offset = start - (page->index << PAGE_CACHE_SHIFT); struct inode *inode = page->mapping->host; - struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; char *kaddr; u64 private; @@ -173,11 +171,15 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) if (ret) { goto zeroit; } + /* + struct btrfs_root *root = BTRFS_I(inode)->root; + char csum[BTRFS_CRC32_SIZE]; ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum); BUG_ON(ret); if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) { goto zeroit; } + */ kunmap_atomic(kaddr, KM_IRQ0); return 0; @@ -192,7 +194,9 @@ zeroit: void btrfs_read_locked_inode(struct inode *inode) { struct btrfs_path *path; + struct extent_buffer *leaf; struct btrfs_inode_item *inode_item; + struct btrfs_inode_timespec *tspec; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_key location; u64 alloc_group_block; @@ -205,29 +209,37 @@ void btrfs_read_locked_inode(struct inode *inode) memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); ret = btrfs_lookup_inode(NULL, root, path, &location, 0); - if (ret) { + if (ret) goto make_bad; - } - inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], - struct btrfs_inode_item); - inode->i_mode = btrfs_inode_mode(inode_item); - inode->i_nlink = btrfs_inode_nlink(inode_item); - inode->i_uid = btrfs_inode_uid(inode_item); - inode->i_gid = btrfs_inode_gid(inode_item); - inode->i_size = btrfs_inode_size(inode_item); - inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime); - inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime); - inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime); - inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime); - inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime); - inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime); - inode->i_blocks = btrfs_inode_nblocks(inode_item); - inode->i_generation = btrfs_inode_generation(inode_item); + leaf = path->nodes[0]; + inode_item = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_inode_item); + + inode->i_mode = btrfs_inode_mode(leaf, inode_item); + inode->i_nlink = btrfs_inode_nlink(leaf, inode_item); + inode->i_uid = btrfs_inode_uid(leaf, inode_item); + inode->i_gid = btrfs_inode_gid(leaf, inode_item); + inode->i_size = btrfs_inode_size(leaf, inode_item); + + tspec = btrfs_inode_atime(inode_item); + inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec); + inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); + + tspec = btrfs_inode_mtime(inode_item); + inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec); + inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); + + tspec = btrfs_inode_ctime(inode_item); + inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec); + inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); + + inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item); + inode->i_generation = btrfs_inode_generation(leaf, inode_item); inode->i_rdev = 0; - rdev = btrfs_inode_rdev(inode_item); - alloc_group_block = btrfs_inode_block_group(inode_item); + rdev = btrfs_inode_rdev(leaf, inode_item); + + alloc_group_block = btrfs_inode_block_group(leaf, inode_item); BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, alloc_group_block); @@ -267,24 +279,35 @@ make_bad: make_bad_inode(inode); } -static void fill_inode_item(struct btrfs_inode_item *item, +static void fill_inode_item(struct extent_buffer *leaf, + struct btrfs_inode_item *item, struct inode *inode) { - btrfs_set_inode_uid(item, inode->i_uid); - btrfs_set_inode_gid(item, inode->i_gid); - btrfs_set_inode_size(item, inode->i_size); - btrfs_set_inode_mode(item, inode->i_mode); - btrfs_set_inode_nlink(item, inode->i_nlink); - btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec); - btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec); - btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec); - btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec); - btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec); - btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec); - btrfs_set_inode_nblocks(item, inode->i_blocks); - btrfs_set_inode_generation(item, inode->i_generation); - btrfs_set_inode_rdev(item, inode->i_rdev); - btrfs_set_inode_block_group(item, + btrfs_set_inode_uid(leaf, item, inode->i_uid); + btrfs_set_inode_gid(leaf, item, inode->i_gid); + btrfs_set_inode_size(leaf, item, inode->i_size); + btrfs_set_inode_mode(leaf, item, inode->i_mode); + btrfs_set_inode_nlink(leaf, item, inode->i_nlink); + + btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item), + inode->i_atime.tv_sec); + btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item), + inode->i_atime.tv_nsec); + + btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item), + inode->i_mtime.tv_sec); + btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item), + inode->i_mtime.tv_nsec); + + btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item), + inode->i_ctime.tv_sec); + btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item), + inode->i_ctime.tv_nsec); + + btrfs_set_inode_nblocks(leaf, item, inode->i_blocks); + btrfs_set_inode_generation(leaf, item, inode->i_generation); + btrfs_set_inode_rdev(leaf, item, inode->i_rdev); + btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group->key.objectid); } @@ -294,6 +317,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, { struct btrfs_inode_item *inode_item; struct btrfs_path *path; + struct extent_buffer *leaf; int ret; path = btrfs_alloc_path(); @@ -306,12 +330,12 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, goto failed; } - inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], + leaf = path->nodes[0]; + inode_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item); - fill_inode_item(inode_item, inode); - btrfs_mark_buffer_dirty(path->nodes[0]); + fill_inode_item(leaf, inode_item, inode); + btrfs_mark_buffer_dirty(leaf); btrfs_set_inode_last_trans(trans, inode); ret = 0; failed: @@ -330,8 +354,9 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, const char *name = dentry->d_name.name; int name_len = dentry->d_name.len; int ret = 0; - u64 objectid; + struct extent_buffer *leaf; struct btrfs_dir_item *di; + struct btrfs_key key; path = btrfs_alloc_path(); if (!path) { @@ -349,14 +374,15 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, ret = -ENOENT; goto err; } - objectid = btrfs_disk_key_objectid(&di->location); + leaf = path->nodes[0]; + btrfs_dir_item_key_to_cpu(leaf, di, &key); ret = btrfs_delete_one_dir_name(trans, root, path, di); if (ret) goto err; btrfs_release_path(root, path); di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, - objectid, name, name_len, -1); + key.objectid, name, name_len, -1); if (IS_ERR(di)) { ret = PTR_ERR(di); goto err; @@ -391,12 +417,15 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) root = BTRFS_I(dir)->root; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); ret = btrfs_unlink_trans(trans, root, dir, dentry); nr = trans->blocks_used; + btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + return ret; } @@ -411,7 +440,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) struct btrfs_trans_handle *trans; struct btrfs_key found_key; int found_type; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; char *goodnames = ".."; unsigned long nr; @@ -419,10 +448,11 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) BUG_ON(!path); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); key.objectid = inode->i_ino; key.offset = (u64)-1; - key.flags = (u32)-1; + key.type = (u8)-1; while(1) { ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) { @@ -435,9 +465,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) goto out; } path->slots[0]--; - leaf = btrfs_buffer_leaf(path->nodes[0]); - btrfs_disk_key_to_cpu(&found_key, - &leaf->items[path->slots[0]].key); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); found_type = btrfs_key_type(&found_key); if (found_key.objectid != inode->i_ino) { err = -ENOENT; @@ -513,9 +542,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int ret; struct btrfs_path *path; struct btrfs_key key; - struct btrfs_disk_key *found_key; + struct btrfs_key found_key; u32 found_type; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; struct btrfs_file_extent_item *fi; u64 extent_start = 0; u64 extent_num_blocks = 0; @@ -527,10 +556,12 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); path->reada = -1; BUG_ON(!path); + /* FIXME, add redo link to tree so we don't leak on crash */ key.objectid = inode->i_ino; key.offset = (u64)-1; - key.flags = (u32)-1; + key.type = (u8)-1; + while(1) { btrfs_init_path(path); fi = NULL; @@ -542,27 +573,28 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, BUG_ON(path->slots[0] == 0); path->slots[0]--; } - leaf = btrfs_buffer_leaf(path->nodes[0]); - found_key = &leaf->items[path->slots[0]].key; - found_type = btrfs_disk_key_type(found_key); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + found_type = btrfs_key_type(&found_key); - if (btrfs_disk_key_objectid(found_key) != inode->i_ino) + if (found_key.objectid != inode->i_ino) break; + if (found_type != BTRFS_CSUM_ITEM_KEY && found_type != BTRFS_DIR_ITEM_KEY && found_type != BTRFS_DIR_INDEX_KEY && found_type != BTRFS_EXTENT_DATA_KEY) break; - item_end = btrfs_disk_key_offset(found_key); + item_end = found_key.offset; if (found_type == BTRFS_EXTENT_DATA_KEY) { - fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], + fi = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - if (btrfs_file_extent_type(fi) != + if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_INLINE) { - item_end += btrfs_file_extent_num_blocks(fi) << - inode->i_blkbits; + item_end += + btrfs_file_extent_num_blocks(leaf, fi) << + inode->i_blkbits; } } if (found_type == BTRFS_CSUM_ITEM_KEY) { @@ -583,7 +615,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, btrfs_set_key_type(&key, found_type); continue; } - if (btrfs_disk_key_offset(found_key) >= inode->i_size) + if (found_key.offset >= inode->i_size) del_item = 1; else del_item = 0; @@ -591,30 +623,31 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, /* FIXME, shrink the extent if the ref count is only 1 */ if (found_type == BTRFS_EXTENT_DATA_KEY && - btrfs_file_extent_type(fi) != + btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_INLINE) { u64 num_dec; - extent_start = btrfs_file_extent_disk_blocknr(fi); + extent_start = btrfs_file_extent_disk_blocknr(leaf, fi); if (!del_item) { u64 orig_num_blocks = - btrfs_file_extent_num_blocks(fi); + btrfs_file_extent_num_blocks(leaf, fi); extent_num_blocks = inode->i_size - - btrfs_disk_key_offset(found_key) + - root->blocksize - 1; + found_key.offset + root->sectorsize - 1; extent_num_blocks >>= inode->i_blkbits; - btrfs_set_file_extent_num_blocks(fi, + btrfs_set_file_extent_num_blocks(leaf, fi, extent_num_blocks); num_dec = (orig_num_blocks - extent_num_blocks) << 3; if (extent_start != 0) { inode->i_blocks -= num_dec; } - btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_mark_buffer_dirty(leaf); } else { extent_num_blocks = - btrfs_file_extent_disk_num_blocks(fi); + btrfs_file_extent_disk_num_blocks(leaf, + fi); /* FIXME blocksize != 4096 */ - num_dec = btrfs_file_extent_num_blocks(fi) << 3; + num_dec = btrfs_file_extent_num_blocks(leaf, + fi) << 3; if (extent_start != 0) { found_extent = 1; inode->i_blocks -= num_dec; @@ -725,7 +758,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - u64 mask = root->blocksize - 1; + u64 mask = root->sectorsize - 1; u64 pos = (inode->i_size + mask) & ~mask; u64 block_end = attr->ia_size | mask; u64 hole_size; @@ -771,9 +804,11 @@ void btrfs_delete_inode(struct inode *inode) if (is_bad_inode(inode)) { goto no_delete; } + inode->i_size = 0; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); ret = btrfs_truncate_in_trans(trans, root, inode); if (ret) @@ -782,6 +817,7 @@ void btrfs_delete_inode(struct inode *inode) if (ret) goto no_delete_lock; nr = trans->blocks_used; + btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); @@ -819,7 +855,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, ret = 0; goto out; } - btrfs_disk_key_to_cpu(location, &di->location); + btrfs_dir_item_key_to_cpu(path->nodes[0], di, location); out: btrfs_release_path(root, path); btrfs_free_path(path); @@ -856,7 +892,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, ri = &(*sub_root)->root_item; location->objectid = btrfs_root_dirid(ri); - location->flags = 0; btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); location->offset = 0; @@ -908,11 +943,14 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, if (dentry->d_name.len > BTRFS_NAME_LEN) return ERR_PTR(-ENAMETOOLONG); + mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_inode_by_name(dir, dentry, &location); mutex_unlock(&root->fs_info->fs_mutex); + if (ret < 0) return ERR_PTR(ret); + inode = NULL; if (location.objectid) { ret = fixup_tree_root_location(root, &location, &sub_root, @@ -952,10 +990,11 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) struct btrfs_item *item; struct btrfs_dir_item *di; struct btrfs_key key; + struct btrfs_key found_key; struct btrfs_path *path; int ret; u32 nritems; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; int slot; int advance; unsigned char d_type; @@ -964,15 +1003,19 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) u32 di_total; u32 di_len; int key_type = BTRFS_DIR_INDEX_KEY; + char tmp_name[32]; + char *name_ptr; + int name_len; /* FIXME, use a real flag for deciding about the key type */ if (root->fs_info->tree_root == root) key_type = BTRFS_DIR_ITEM_KEY; + mutex_lock(&root->fs_info->fs_mutex); key.objectid = inode->i_ino; - key.flags = 0; btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; + path = btrfs_alloc_path(); path->reada = 2; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); @@ -980,16 +1023,16 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) goto err; advance = 0; while(1) { - leaf = btrfs_buffer_leaf(path->nodes[0]); - nritems = btrfs_header_nritems(&leaf->header); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; if (advance || slot >= nritems) { if (slot >= nritems -1) { ret = btrfs_next_leaf(root, path); if (ret) break; - leaf = btrfs_buffer_leaf(path->nodes[0]); - nritems = btrfs_header_nritems(&leaf->header); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; } else { slot++; @@ -997,28 +1040,48 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) } } advance = 1; - item = leaf->items + slot; - if (btrfs_disk_key_objectid(&item->key) != key.objectid) + item = btrfs_item_nr(leaf, slot); + btrfs_item_key_to_cpu(leaf, &found_key, slot); + + if (found_key.objectid != key.objectid) break; - if (btrfs_disk_key_type(&item->key) != key_type) + if (btrfs_key_type(&found_key) != key_type) break; - if (btrfs_disk_key_offset(&item->key) < filp->f_pos) + if (found_key.offset < filp->f_pos) continue; - filp->f_pos = btrfs_disk_key_offset(&item->key); + + filp->f_pos = found_key.offset; advance = 1; di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); di_cur = 0; - di_total = btrfs_item_size(leaf->items + slot); + di_total = btrfs_item_size(leaf, item); while(di_cur < di_total) { - d_type = btrfs_filetype_table[btrfs_dir_type(di)]; - over = filldir(dirent, (const char *)(di + 1), - btrfs_dir_name_len(di), - btrfs_disk_key_offset(&item->key), - btrfs_disk_key_objectid(&di->location), + struct btrfs_key location; + + name_len = btrfs_dir_name_len(leaf, di); + if (name_len < 32) { + name_ptr = tmp_name; + } else { + name_ptr = kmalloc(name_len, GFP_NOFS); + BUG_ON(!name_ptr); + } + read_extent_buffer(leaf, name_ptr, + (unsigned long)(di + 1), name_len); + + d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)]; + btrfs_dir_item_key_to_cpu(leaf, di, &location); + + over = filldir(dirent, name_ptr, name_len, + found_key.offset, + location.objectid, d_type); + + if (name_ptr != tmp_name) + kfree(name_ptr); + if (over) goto nopos; - di_len = btrfs_dir_name_len(di) + sizeof(*di); + di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di); di_cur += di_len; di = (struct btrfs_dir_item *)((char *)di + di_len); } @@ -1075,11 +1138,15 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, int mode) { struct inode *inode; - struct btrfs_inode_item inode_item; + struct btrfs_inode_item *inode_item; struct btrfs_key *location; + struct btrfs_path *path; int ret; int owner; + path = btrfs_alloc_path(); + BUG_ON(!path); + inode = new_inode(root->fs_info->sb); if (!inode) return ERR_PTR(-ENOMEM); @@ -1095,24 +1162,32 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, group = btrfs_find_block_group(root, group, 0, 0, owner); BTRFS_I(inode)->block_group = group; + ret = btrfs_insert_empty_inode(trans, root, path, objectid); + if (ret) + goto fail; + inode->i_uid = current->fsuid; inode->i_gid = current->fsgid; inode->i_mode = mode; inode->i_ino = objectid; inode->i_blocks = 0; inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; - fill_inode_item(&inode_item, inode); + inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_inode_item); + fill_inode_item(path->nodes[0], inode_item, inode); + btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_free_path(path); + location = &BTRFS_I(inode)->location; location->objectid = objectid; - location->flags = 0; location->offset = 0; btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); - ret = btrfs_insert_inode(trans, root, objectid, &inode_item); - if (ret) - return ERR_PTR(ret); insert_inode_hash(inode); return inode; +fail: + btrfs_free_path(path); + return ERR_PTR(ret); } static inline u8 btrfs_inode_type(struct inode *inode) @@ -1127,8 +1202,8 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, struct btrfs_key key; struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root; struct inode *parent_inode; + key.objectid = inode->i_ino; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); key.offset = 0; @@ -1285,14 +1360,18 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, inc_nlink(inode); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, dir); atomic_inc(&inode->i_count); err = btrfs_add_nondir(trans, dentry, inode); + if (err) drop_inode = 1; + dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, dir); err = btrfs_update_inode(trans, root, inode); + if (err) drop_inode = 1; @@ -1321,13 +1400,13 @@ static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans, key.objectid = objectid; key.offset = 0; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid, &key, BTRFS_FT_DIR); if (ret) goto error; + key.objectid = dirid; ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid, &key, BTRFS_FT_DIR); @@ -1350,6 +1429,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); + if (IS_ERR(trans)) { err = PTR_ERR(trans); goto out_unlock; @@ -1367,6 +1447,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) err = PTR_ERR(inode); goto out_fail; } + drop_on_err = 1; inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; @@ -1380,9 +1461,11 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) err = btrfs_update_inode(trans, root, inode); if (err) goto out_fail; + err = btrfs_add_link(trans, dentry, inode); if (err) goto out_fail; + d_instantiate(dentry, inode); drop_on_err = 0; dir->i_sb->s_dirt = 1; @@ -1392,6 +1475,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) out_fail: nr = trans->blocks_used; btrfs_end_transaction(trans, root); + out_unlock: mutex_unlock(&root->fs_info->fs_mutex); if (drop_on_err) @@ -1415,8 +1499,8 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, struct btrfs_path *path; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_file_extent_item *item; - struct btrfs_leaf *leaf; - struct btrfs_disk_key *found_key; + struct extent_buffer *leaf; + struct btrfs_key found_key; struct extent_map *em = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct btrfs_trans_handle *trans = NULL; @@ -1436,8 +1520,8 @@ again: err = -ENOMEM; goto out; } - em->start = 0; - em->end = 0; + em->start = EXTENT_MAP_HOLE; + em->end = EXTENT_MAP_HOLE; } em->bdev = inode->i_sb->s_bdev; ret = btrfs_lookup_file_extent(NULL, root, path, @@ -1453,25 +1537,27 @@ again: path->slots[0]--; } - item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0], + leaf = path->nodes[0]; + item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - leaf = btrfs_buffer_leaf(path->nodes[0]); - blocknr = btrfs_file_extent_disk_blocknr(item); - blocknr += btrfs_file_extent_offset(item); + + blocknr = btrfs_file_extent_disk_blocknr(leaf, item); + blocknr += btrfs_file_extent_offset(leaf, item); /* are we inside the extent that was found? */ - found_key = &leaf->items[path->slots[0]].key; - found_type = btrfs_disk_key_type(found_key); - if (btrfs_disk_key_objectid(found_key) != objectid || + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + found_type = btrfs_key_type(&found_key); + if (found_key.objectid != objectid || found_type != BTRFS_EXTENT_DATA_KEY) { goto not_found; } - found_type = btrfs_file_extent_type(item); - extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key); + found_type = btrfs_file_extent_type(leaf, item); + extent_start = found_key.offset; if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = extent_start + - (btrfs_file_extent_num_blocks(item) << inode->i_blkbits); + (btrfs_file_extent_num_blocks(leaf, item) << + inode->i_blkbits); err = 0; if (start < extent_start || start >= extent_end) { em->start = start; @@ -1484,28 +1570,29 @@ again: } goto not_found_em; } - if (btrfs_file_extent_disk_blocknr(item) == 0) { + if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) { em->start = extent_start; em->end = extent_end - 1; - em->block_start = 0; - em->block_end = 0; + em->block_start = EXTENT_MAP_HOLE; + em->block_end = EXTENT_MAP_HOLE; goto insert; } em->block_start = blocknr << inode->i_blkbits; em->block_end = em->block_start + - (btrfs_file_extent_num_blocks(item) << + (btrfs_file_extent_num_blocks(leaf, item) << inode->i_blkbits) - 1; em->start = extent_start; em->end = extent_end - 1; goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - char *ptr; + unsigned long ptr; char *map; u32 size; - size = btrfs_file_extent_inline_len(leaf->items + - path->slots[0]); - extent_end = extent_start | ((u64)root->blocksize - 1); + size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, + path->slots[0])); + + extent_end = extent_start | ((u64)root->sectorsize - 1); if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { @@ -1517,18 +1604,21 @@ again: } goto not_found_em; } + em->block_start = EXTENT_MAP_INLINE; em->block_end = EXTENT_MAP_INLINE; em->start = extent_start; em->end = extent_end; + if (!page) { goto insert; } + ptr = btrfs_file_extent_inline_start(item); map = kmap(page); - memcpy(map + page_offset, ptr, size); + read_extent_buffer(leaf, map + page_offset, ptr, size); memset(map + page_offset + size, 0, - root->blocksize - (page_offset + size)); + root->sectorsize - (page_offset + size)); flush_dcache_page(page); kunmap(page); set_extent_uptodate(em_tree, extent_start, @@ -1542,8 +1632,8 @@ not_found: em->start = start; em->end = end; not_found_em: - em->block_start = 0; - em->block_end = 0; + em->block_start = EXTENT_MAP_HOLE; + em->block_end = EXTENT_MAP_HOLE; insert: btrfs_release_path(root, path); if (em->start > start || em->end < start) { @@ -1712,6 +1802,7 @@ static void btrfs_truncate(struct inode *inode) ret = btrfs_truncate_in_trans(trans, root, inode); btrfs_update_inode(trans, root, inode); nr = trans->blocks_used; + ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); @@ -1731,8 +1822,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) struct btrfs_key key; struct btrfs_root_item root_item; struct btrfs_inode_item *inode_item; - struct buffer_head *subvol; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; struct btrfs_root *new_root; struct inode *inode; struct inode *dir; @@ -1746,34 +1836,37 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - subvol = btrfs_alloc_free_block(trans, root, 0, 0); - if (IS_ERR(subvol)) - return PTR_ERR(subvol); - leaf = btrfs_buffer_leaf(subvol); - btrfs_set_header_nritems(&leaf->header, 0); - btrfs_set_header_level(&leaf->header, 0); - btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol)); - btrfs_set_header_generation(&leaf->header, trans->transid); - btrfs_set_header_owner(&leaf->header, root->root_key.objectid); - memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid, - sizeof(leaf->header.fsid)); - btrfs_mark_buffer_dirty(subvol); + leaf = btrfs_alloc_free_block(trans, root, 0, 0); + if (IS_ERR(leaf)) + return PTR_ERR(leaf); + + btrfs_set_header_nritems(leaf, 0); + btrfs_set_header_level(leaf, 0); + btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf)); + btrfs_set_header_generation(leaf, trans->transid); + btrfs_set_header_owner(leaf, root->root_key.objectid); + write_extent_buffer(leaf, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(leaf), + BTRFS_FSID_SIZE); + btrfs_mark_buffer_dirty(leaf); inode_item = &root_item.inode; memset(inode_item, 0, sizeof(*inode_item)); - btrfs_set_inode_generation(inode_item, 1); - btrfs_set_inode_size(inode_item, 3); - btrfs_set_inode_nlink(inode_item, 1); - btrfs_set_inode_nblocks(inode_item, 1); - btrfs_set_inode_mode(inode_item, S_IFDIR | 0755); + inode_item->generation = cpu_to_le64(1); + inode_item->size = cpu_to_le64(3); + inode_item->nlink = cpu_to_le32(1); + inode_item->nblocks = cpu_to_le64(1); + inode_item->mode = cpu_to_le32(S_IFDIR | 0755); - btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol)); + btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf)); btrfs_set_root_refs(&root_item, 1); - btrfs_set_root_blocks_used(&root_item, 0); + btrfs_set_root_used(&root_item, 0); + memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); root_item.drop_level = 0; - brelse(subvol); - subvol = NULL; + + free_extent_buffer(leaf); + leaf = NULL; ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, 0, &objectid); @@ -1784,7 +1877,6 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) key.objectid = objectid; key.offset = 1; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &root_item); @@ -1845,7 +1937,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) struct btrfs_trans_handle *trans; struct btrfs_key key; struct btrfs_root_item new_root_item; - struct buffer_head *tmp; + struct extent_buffer *tmp; int ret; int err; u64 objectid; @@ -1876,10 +1968,11 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) key.objectid = objectid; key.offset = 1; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); - btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node)); + btrfs_set_root_blocknr(&new_root_item, + extent_buffer_blocknr(root->node)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); @@ -1904,8 +1997,10 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) fail: nr = trans->blocks_used; err = btrfs_commit_transaction(trans, root); + if (err && !ret) ret = err; + mutex_unlock(&root->fs_info->fs_mutex); up_write(&root->snap_sem); btrfs_btree_balance_dirty(root, nr); @@ -1986,7 +2081,7 @@ static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) if (copy_from_user(&vol_args, arg, sizeof(vol_args))) return -EFAULT; - + namelen = strlen(vol_args.name); if (namelen > BTRFS_VOL_NAME_MAX) return -EINVAL; @@ -2164,8 +2259,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) { return -ENOTEMPTY; } + mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, new_dir); path = btrfs_alloc_path(); if (!path) { @@ -2177,9 +2274,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, old_dir->i_ctime = old_dir->i_mtime = ctime; new_dir->i_ctime = new_dir->i_mtime = ctime; old_inode->i_ctime = ctime; + if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) { struct btrfs_key *location = &BTRFS_I(new_dir)->location; - u64 old_parent_oid; + struct btrfs_key old_parent_key; di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino, "..", 2, -1); if (IS_ERR(di)) { @@ -2190,7 +2288,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, ret = -ENOENT; goto out_fail; } - old_parent_oid = btrfs_disk_key_objectid(&di->location); + btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key); ret = btrfs_del_item(trans, root, path); if (ret) { goto out_fail; @@ -2199,7 +2297,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, di = btrfs_lookup_dir_index_item(trans, root, path, old_inode->i_ino, - old_parent_oid, + old_parent_key.objectid, "..", 2, -1); if (IS_ERR(di)) { ret = PTR_ERR(di); @@ -2257,8 +2355,9 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, u64 objectid; int name_len; int datasize; - char *ptr; + unsigned long ptr; struct btrfs_file_extent_item *ei; + struct extent_buffer *leaf; unsigned long nr; name_len = strlen(symname) + 1; @@ -2302,7 +2401,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, BUG_ON(!path); key.objectid = inode->i_ino; key.offset = 0; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); datasize = btrfs_file_extent_calc_inline_size(name_len); err = btrfs_insert_empty_item(trans, root, path, &key, @@ -2311,16 +2409,17 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, drop_inode = 1; goto out_unlock; } - ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(ei, trans->transid); - btrfs_set_file_extent_type(ei, + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, ei, trans->transid); + btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); - btrfs_memcpy(root, path->nodes[0]->b_data, - ptr, symname, name_len); - btrfs_mark_buffer_dirty(path->nodes[0]); + write_extent_buffer(leaf, symname, ptr, name_len); + btrfs_mark_buffer_dirty(leaf); btrfs_free_path(path); + inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; inode->i_size = name_len - 1; diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index bafa78603bd..a825ce078a5 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -20,10 +20,10 @@ #include "disk-io.h" #include "print-tree.h" -void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l) +void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) { int i; - u32 nr = btrfs_header_nritems(&l->header); + u32 nr = btrfs_header_nritems(l); struct btrfs_item *item; struct btrfs_extent_item *ei; struct btrfs_root_item *ri; @@ -31,119 +31,113 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l) struct btrfs_inode_item *ii; struct btrfs_block_group_item *bi; struct btrfs_file_extent_item *fi; + struct btrfs_key key; + struct btrfs_key found_key; u32 type; printk("leaf %llu total ptrs %d free space %d\n", - (unsigned long long)btrfs_header_blocknr(&l->header), nr, + (unsigned long long)btrfs_header_blocknr(l), nr, btrfs_leaf_free_space(root, l)); for (i = 0 ; i < nr ; i++) { - item = l->items + i; - type = btrfs_disk_key_type(&item->key); + item = btrfs_item_nr(l, i); + btrfs_item_key_to_cpu(l, &key, i); + type = btrfs_key_type(&key); printk("\titem %d key (%llu %x %llu) itemoff %d itemsize %d\n", i, - (unsigned long long)btrfs_disk_key_objectid(&item->key), - btrfs_disk_key_flags(&item->key), - (unsigned long long)btrfs_disk_key_offset(&item->key), - btrfs_item_offset(item), - btrfs_item_size(item)); + (unsigned long long)key.objectid, type, + (unsigned long long)key.offset, + btrfs_item_offset(l, item), btrfs_item_size(l, item)); switch (type) { case BTRFS_INODE_ITEM_KEY: ii = btrfs_item_ptr(l, i, struct btrfs_inode_item); printk("\t\tinode generation %llu size %llu mode %o\n", - (unsigned long long)btrfs_inode_generation(ii), - (unsigned long long)btrfs_inode_size(ii), - btrfs_inode_mode(ii)); + (unsigned long long)btrfs_inode_generation(l, ii), + (unsigned long long)btrfs_inode_size(l, ii), + btrfs_inode_mode(l, ii)); break; case BTRFS_DIR_ITEM_KEY: di = btrfs_item_ptr(l, i, struct btrfs_dir_item); + btrfs_dir_item_key_to_cpu(l, di, &found_key); printk("\t\tdir oid %llu flags %u type %u\n", - (unsigned long long)btrfs_disk_key_objectid( - &di->location), - btrfs_dir_flags(di), - btrfs_dir_type(di)); - printk("\t\tname %.*s\n", - btrfs_dir_name_len(di),(char *)(di + 1)); + (unsigned long long)found_key.objectid, + btrfs_dir_flags(l, di), + btrfs_dir_type(l, di)); break; case BTRFS_ROOT_ITEM_KEY: ri = btrfs_item_ptr(l, i, struct btrfs_root_item); printk("\t\troot data blocknr %llu refs %u\n", - (unsigned long long)btrfs_root_blocknr(ri), - btrfs_root_refs(ri)); + (unsigned long long)btrfs_disk_root_blocknr(l, ri), + btrfs_disk_root_refs(l, ri)); break; case BTRFS_EXTENT_ITEM_KEY: ei = btrfs_item_ptr(l, i, struct btrfs_extent_item); printk("\t\textent data refs %u\n", - btrfs_extent_refs(ei)); + btrfs_extent_refs(l, ei)); break; case BTRFS_EXTENT_DATA_KEY: fi = btrfs_item_ptr(l, i, struct btrfs_file_extent_item); - if (btrfs_file_extent_type(fi) == + if (btrfs_file_extent_type(l, fi) == BTRFS_FILE_EXTENT_INLINE) { printk("\t\tinline extent data size %u\n", - btrfs_file_extent_inline_len(l->items + i)); + btrfs_file_extent_inline_len(l, item)); break; } printk("\t\textent data disk block %llu nr %llu\n", - (unsigned long long)btrfs_file_extent_disk_blocknr(fi), - (unsigned long long)btrfs_file_extent_disk_num_blocks(fi)); + (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi), + (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi)); printk("\t\textent data offset %llu nr %llu\n", - (unsigned long long)btrfs_file_extent_offset(fi), - (unsigned long long)btrfs_file_extent_num_blocks(fi)); + (unsigned long long)btrfs_file_extent_offset(l, fi), + (unsigned long long)btrfs_file_extent_num_blocks(l, fi)); break; case BTRFS_BLOCK_GROUP_ITEM_KEY: bi = btrfs_item_ptr(l, i, struct btrfs_block_group_item); printk("\t\tblock group used %llu\n", - (unsigned long long)btrfs_block_group_used(bi)); - break; - case BTRFS_STRING_ITEM_KEY: - printk("\t\titem data %.*s\n", btrfs_item_size(item), - btrfs_leaf_data(l) + btrfs_item_offset(item)); + (unsigned long long)btrfs_disk_block_group_used(l, bi)); break; }; } } -void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t) +void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c) { int i; u32 nr; - struct btrfs_node *c; + struct btrfs_key key; - if (!t) + if (!c) return; - c = btrfs_buffer_node(t); - nr = btrfs_header_nritems(&c->header); + nr = btrfs_header_nritems(c); if (btrfs_is_leaf(c)) { - btrfs_print_leaf(root, (struct btrfs_leaf *)c); + btrfs_print_leaf(root, c); return; } printk("node %llu level %d total ptrs %d free spc %u\n", - (unsigned long long)btrfs_header_blocknr(&c->header), - btrfs_header_level(&c->header), nr, + (unsigned long long)btrfs_header_blocknr(c), + btrfs_header_level(c), nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr); for (i = 0; i < nr; i++) { + btrfs_node_key_to_cpu(c, &key, i); printk("\tkey %d (%llu %u %llu) block %llu\n", i, - (unsigned long long)c->ptrs[i].key.objectid, - c->ptrs[i].key.flags, - (unsigned long long)c->ptrs[i].key.offset, + (unsigned long long)key.objectid, + key.type, + (unsigned long long)key.offset, (unsigned long long)btrfs_node_blockptr(c, i)); } for (i = 0; i < nr; i++) { - struct buffer_head *next_buf = read_tree_block(root, + struct extent_buffer *next = read_tree_block(root, btrfs_node_blockptr(c, i)); - struct btrfs_node *next = btrfs_buffer_node(next_buf); if (btrfs_is_leaf(next) && - btrfs_header_level(&c->header) != 1) + btrfs_header_level(c) != 1) BUG(); - if (btrfs_header_level(&next->header) != - btrfs_header_level(&c->header) - 1) + if (btrfs_header_level(next) != + btrfs_header_level(c) - 1) BUG(); - btrfs_print_tree(root, next_buf); - btrfs_block_release(root, next_buf); + btrfs_print_tree(root, next); + free_extent_buffer(next); } } diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h index 205cd03601a..da75efe534d 100644 --- a/fs/btrfs/print-tree.h +++ b/fs/btrfs/print-tree.h @@ -18,6 +18,6 @@ #ifndef __PRINT_TREE_ #define __PRINT_TREE_ -void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l); -void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t); +void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l); +void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *t); #endif diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 3b5926dfbeb..88bcdd33f56 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -26,12 +26,13 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, { struct btrfs_path *path; struct btrfs_key search_key; - struct btrfs_leaf *l; + struct btrfs_key found_key; + struct extent_buffer *l; int ret; int slot; search_key.objectid = objectid; - search_key.flags = (u32)-1; + search_key.type = (u8)-1; search_key.offset = (u64)-1; path = btrfs_alloc_path(); @@ -39,17 +40,19 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); if (ret < 0) goto out; + BUG_ON(ret == 0); - l = btrfs_buffer_leaf(path->nodes[0]); + l = path->nodes[0]; BUG_ON(path->slots[0] == 0); slot = path->slots[0] - 1; - if (btrfs_disk_key_objectid(&l->items[slot].key) != objectid) { + btrfs_item_key_to_cpu(l, &found_key, slot); + if (found_key.objectid != objectid) { ret = 1; goto out; } - memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item), - sizeof(*item)); - btrfs_disk_key_to_cpu(key, &l->items[slot].key); + read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot), + sizeof(*item)); + memcpy(key, &found_key, sizeof(found_key)); ret = 0; out: btrfs_release_path(root, path); @@ -62,10 +65,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root *item) { struct btrfs_path *path; - struct btrfs_leaf *l; + struct extent_buffer *l; int ret; int slot; - struct btrfs_root_item *update_item; + unsigned long ptr; path = btrfs_alloc_path(); BUG_ON(!path); @@ -73,10 +76,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root if (ret < 0) goto out; BUG_ON(ret != 0); - l = btrfs_buffer_leaf(path->nodes[0]); + l = path->nodes[0]; slot = path->slots[0]; - update_item = btrfs_item_ptr(l, slot, struct btrfs_root_item); - btrfs_memcpy(root, l, update_item, item, sizeof(*item)); + ptr = btrfs_item_ptr_offset(l, slot); + write_extent_buffer(l, item, ptr, sizeof(*item)); btrfs_mark_buffer_dirty(path->nodes[0]); out: btrfs_release_path(root, path); @@ -103,11 +106,10 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, struct btrfs_path *path; int ret; u32 nritems; - struct btrfs_leaf *leaf; + struct extent_buffer *leaf; int slot; key.objectid = objectid; - key.flags = 0; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); key.offset = 0; path = btrfs_alloc_path(); @@ -117,19 +119,19 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, if (ret < 0) goto err; while(1) { - leaf = btrfs_buffer_leaf(path->nodes[0]); - nritems = btrfs_header_nritems(&leaf->header); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; if (slot >= nritems) { ret = btrfs_next_leaf(root, path); if (ret) break; - leaf = btrfs_buffer_leaf(path->nodes[0]); - nritems = btrfs_header_nritems(&leaf->header); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; } - item = leaf->items + slot; - btrfs_disk_key_to_cpu(&key, &item->key); + item = btrfs_item_nr(leaf, slot); + btrfs_item_key_to_cpu(leaf, &key, slot); if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY) goto next; @@ -140,7 +142,7 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, break; ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item); - if (btrfs_root_refs(ri) != 0) + if (btrfs_disk_root_refs(leaf, ri) != 0) goto next; dead_root = btrfs_read_fs_root_no_radix(root->fs_info, &key); @@ -170,6 +172,7 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, int ret; u32 refs; struct btrfs_root_item *ri; + struct extent_buffer *leaf; path = btrfs_alloc_path(); BUG_ON(!path); @@ -177,10 +180,10 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (ret < 0) goto out; BUG_ON(ret != 0); - ri = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), - path->slots[0], struct btrfs_root_item); + leaf = path->nodes[0]; + ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item); - refs = btrfs_root_refs(ri); + refs = btrfs_disk_root_refs(leaf, ri); BUG_ON(refs != 0); ret = btrfs_del_item(trans, root, path); out: diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 66a01cbbbea..39a1435c68f 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -41,7 +41,7 @@ #include "ioctl.h" #include "print-tree.h" -#define BTRFS_SUPER_MAGIC 0x9123682E +#define BTRFS_SUPER_MAGIC 0x9123683E static struct super_operations btrfs_super_ops; @@ -115,13 +115,12 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) return -EIO; } sb->s_fs_info = tree_root; - disk_super = tree_root->fs_info->disk_super; + disk_super = &tree_root->fs_info->super_copy; inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super), tree_root); bi = BTRFS_I(inode); bi->location.objectid = inode->i_ino; bi->location.offset = 0; - bi->location.flags = 0; bi->root = tree_root; btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY); @@ -281,6 +280,7 @@ error_s: error_bdev: close_bdev_excl(bdev); error: +printk("get_sb failed\n"); return error; } /* end copy & paste */ @@ -295,6 +295,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, btrfs_fill_super, mnt, subvol_name ? subvol_name : "default"); +printk("btrfs_get_sb returns %d\n", ret); return ret; } diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index 2058783373e..9654e90eec8 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -31,31 +31,31 @@ static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_root_blocks_used(&root->root_item)); + (unsigned long long)btrfs_root_used(&root->root_item)); } static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_root_block_limit(&root->root_item)); + (unsigned long long)btrfs_root_limit(&root->root_item)); } static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_super_blocks_used(fs->disk_super)); + (unsigned long long)btrfs_super_blocks_used(&fs->super_copy)); } static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_super_total_blocks(fs->disk_super)); + (unsigned long long)btrfs_super_total_blocks(&fs->super_copy)); } static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_super_blocksize(fs->disk_super)); + (unsigned long long)btrfs_super_sectorsize(&fs->super_copy)); } /* this is for root attrs (subvols/snapshots) */ diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 087074db0bd..750f35a37aa 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -19,6 +19,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -66,7 +67,9 @@ static int join_transaction(struct btrfs_root *root) cur_trans->commit_done = 0; cur_trans->start_time = get_seconds(); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); - init_bit_radix(&cur_trans->dirty_pages); + extent_map_tree_init(&cur_trans->dirty_pages, + root->fs_info->btree_inode->i_mapping, + GFP_NOFS); } else { cur_trans->num_writers++; cur_trans->num_joined++; @@ -88,7 +91,7 @@ static int record_root_in_trans(struct btrfs_root *root) (unsigned long)root->root_key.objectid, BTRFS_ROOT_DEFRAG_TAG); root->commit_root = root->node; - get_bh(root->node); + extent_buffer_get(root->node); } else { WARN_ON(1); } @@ -144,29 +147,30 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans, int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - unsigned long gang[16]; int ret; - int i; int err; int werr = 0; + struct extent_map_tree *dirty_pages; struct page *page; - struct radix_tree_root *dirty_pages; struct inode *btree_inode = root->fs_info->btree_inode; + u64 start; + u64 end; + unsigned long index; if (!trans || !trans->transaction) { return filemap_write_and_wait(btree_inode->i_mapping); } dirty_pages = &trans->transaction->dirty_pages; while(1) { - ret = find_first_radix_bit(dirty_pages, gang, - 0, ARRAY_SIZE(gang)); - if (!ret) + ret = find_first_extent_bit(dirty_pages, 0, &start, &end, + EXTENT_DIRTY); + if (ret) break; - for (i = 0; i < ret; i++) { - /* FIXME EIO */ - clear_radix_bit(dirty_pages, gang[i]); - page = find_lock_page(btree_inode->i_mapping, - gang[i]); + clear_extent_dirty(dirty_pages, start, end, GFP_NOFS); + while(start <= end) { + index = start >> PAGE_CACHE_SHIFT; + start = (index + 1) << PAGE_CACHE_SHIFT; + page = find_lock_page(btree_inode->i_mapping, index); if (!page) continue; if (PageWriteback(page)) { @@ -202,10 +206,11 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, btrfs_write_dirty_block_groups(trans, extent_root); while(1) { old_extent_block = btrfs_root_blocknr(&extent_root->root_item); - if (old_extent_block == bh_blocknr(extent_root->node)) + if (old_extent_block == + extent_buffer_blocknr(extent_root->node)) break; btrfs_set_root_blocknr(&extent_root->root_item, - bh_blocknr(extent_root->node)); + extent_buffer_blocknr(extent_root->node)); ret = btrfs_update_root(trans, tree_root, &extent_root->root_key, &extent_root->root_item); @@ -279,9 +284,9 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, (unsigned long)root->root_key.objectid, BTRFS_ROOT_TRANS_TAG); if (root->commit_root == root->node) { - WARN_ON(bh_blocknr(root->node) != + WARN_ON(extent_buffer_blocknr(root->node) != btrfs_root_blocknr(&root->root_item)); - brelse(root->commit_root); + free_extent_buffer(root->commit_root); root->commit_root = NULL; /* make sure to update the root on disk @@ -310,7 +315,7 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, root->root_key.offset = root->fs_info->generation; btrfs_set_root_blocknr(&root->root_item, - bh_blocknr(root->node)); + extent_buffer_blocknr(root->node)); err = btrfs_insert_root(trans, root->fs_info->tree_root, &root->root_key, &root->root_item); @@ -389,10 +394,10 @@ int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info) for (i = 0; i < ret; i++) { root = gang[i]; last = root->root_key.objectid + 1; - btrfs_defrag_root(root, 1); + // btrfs_defrag_root(root, 1); } } - btrfs_defrag_root(info->extent_root, 1); + // btrfs_defrag_root(info->extent_root, 1); return err; } @@ -414,7 +419,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, dirty = list_entry(list->next, struct dirty_root, list); list_del_init(&dirty->list); - num_blocks = btrfs_root_blocks_used(&dirty->root->root_item); + num_blocks = btrfs_root_used(&dirty->root->root_item); root = dirty->latest_root; while(1) { @@ -441,11 +446,11 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, } BUG_ON(ret); - num_blocks -= btrfs_root_blocks_used(&dirty->root->root_item); - blocks_used = btrfs_root_blocks_used(&root->root_item); + num_blocks -= btrfs_root_used(&dirty->root->root_item); + blocks_used = btrfs_root_used(&root->root_item); if (num_blocks) { record_root_in_trans(root); - btrfs_set_root_blocks_used(&root->root_item, + btrfs_set_root_used(&root->root_item, blocks_used - num_blocks); } ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); @@ -553,9 +558,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_set_super_generation(&root->fs_info->super_copy, cur_trans->transid); btrfs_set_super_root(&root->fs_info->super_copy, - bh_blocknr(root->fs_info->tree_root->node)); - memcpy(root->fs_info->disk_super, &root->fs_info->super_copy, - sizeof(root->fs_info->super_copy)); + extent_buffer_blocknr(root->fs_info->tree_root->node)); + + write_extent_buffer(root->fs_info->sb_buffer, + &root->fs_info->super_copy, 0, + sizeof(root->fs_info->super_copy)); btrfs_copy_pinned(root, &pinned_copy); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 4bc328cbb24..ae39fcfc169 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -28,7 +28,7 @@ struct btrfs_transaction { int use_count; int commit_done; struct list_head list; - struct radix_tree_root dirty_pages; + struct extent_map_tree dirty_pages; unsigned long start_time; wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; @@ -83,5 +83,6 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest, int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info); int btrfs_defrag_root(struct btrfs_root *root, int cacheonly); int btrfs_clean_old_snapshots(struct btrfs_root *root); - +int btrfs_commit_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root); #endif diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 420597127ed..daf019afa0a 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -23,14 +23,14 @@ #include "transaction.h" static void reada_defrag(struct btrfs_root *root, - struct btrfs_node *node) + struct extent_buffer *node) { int i; u32 nritems; u64 blocknr; int ret; - nritems = btrfs_header_nritems(&node->header); + nritems = btrfs_header_nritems(node); for (i = 0; i < nritems; i++) { blocknr = btrfs_node_blockptr(node, i); ret = readahead_tree_block(root, blocknr); @@ -44,8 +44,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, struct btrfs_path *path, int *level, int cache_only, u64 *last_ret) { - struct buffer_head *next; - struct buffer_head *cur; + struct extent_buffer *next; + struct extent_buffer *cur; u64 blocknr; int ret = 0; int is_extent = 0; @@ -62,13 +62,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, cur = path->nodes[*level]; if (!cache_only && *level > 1 && path->slots[*level] == 0) - reada_defrag(root, btrfs_buffer_node(cur)); + reada_defrag(root, cur); - if (btrfs_header_level(btrfs_buffer_header(cur)) != *level) + if (btrfs_header_level(cur) != *level) WARN_ON(1); if (path->slots[*level] >= - btrfs_header_nritems(btrfs_buffer_header(cur))) + btrfs_header_nritems(cur)) break; if (*level == 1) { @@ -80,14 +80,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, break; } - blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur), - path->slots[*level]); + blocknr = btrfs_node_blockptr(cur, path->slots[*level]); if (cache_only) { next = btrfs_find_tree_block(root, blocknr); - if (!next || !buffer_uptodate(next) || - buffer_locked(next) || !buffer_defrag(next)) { - brelse(next); + /* FIXME, test for defrag */ + if (!next || !btrfs_buffer_uptodate(next)) { + free_extent_buffer(next); path->slots[*level]++; continue; } @@ -106,16 +105,18 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, WARN_ON(*level <= 0); if (path->nodes[*level-1]) - btrfs_block_release(root, path->nodes[*level-1]); + free_extent_buffer(path->nodes[*level-1]); path->nodes[*level-1] = next; - *level = btrfs_header_level(btrfs_buffer_header(next)); + *level = btrfs_header_level(next); path->slots[*level] = 0; } WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); +#if 0 clear_buffer_defrag(path->nodes[*level]); clear_buffer_defrag_done(path->nodes[*level]); - btrfs_block_release(root, path->nodes[*level]); +#endif + free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; WARN_ON(ret); @@ -129,24 +130,25 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans, { int i; int slot; - struct btrfs_node *node; + struct extent_buffer *node; for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { slot = path->slots[i]; - if (slot < btrfs_header_nritems( - btrfs_buffer_header(path->nodes[i])) - 1) { + if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { path->slots[i]++; *level = i; - node = btrfs_buffer_node(path->nodes[i]); + node = path->nodes[i]; WARN_ON(i == 0); - btrfs_disk_key_to_cpu(&root->defrag_progress, - &node->ptrs[path->slots[i]].key); + btrfs_node_key_to_cpu(node, &root->defrag_progress, + path->slots[i]); root->defrag_level = i; return 0; } else { + /* clear_buffer_defrag(path->nodes[*level]); clear_buffer_defrag_done(path->nodes[*level]); - btrfs_block_release(root, path->nodes[*level]); + */ + free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level = i + 1; } @@ -158,7 +160,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only) { struct btrfs_path *path = NULL; - struct buffer_head *tmp; + struct extent_buffer *tmp; int ret = 0; int wret; int level; @@ -172,17 +174,18 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, if (root->ref_cows == 0 && !is_extent) goto out; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; - level = btrfs_header_level(btrfs_buffer_header(root->node)); + level = btrfs_header_level(root->node); orig_level = level; if (level == 0) { goto out; } if (root->defrag_progress.objectid == 0) { - get_bh(root->node); + extent_buffer_get(root->node); ret = btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); BUG_ON(ret); ret = btrfs_realloc_node(trans, root, root->node, cache_only, @@ -200,12 +203,15 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, if (is_extent) btrfs_extent_post_op(trans, root); + if (wret < 0) { ret = wret; goto out; } + while(level > 0 && !path->nodes[level]) level--; + if (!path->nodes[level]) { ret = 0; goto out; @@ -230,7 +236,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, } for (i = 0; i <= orig_level; i++) { if (path->nodes[i]) { - btrfs_block_release(root, path->nodes[i]); + free_extent_buffer(path->nodes[i]); path->nodes[i] = 0; } } -- cgit v1.2.3-70-g09d2 From db94535db75e67fab12ccbb7f5ee548e33fed891 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Oct 2007 16:15:53 -0400 Subject: Btrfs: Allow tree blocks larger than the page size Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 319 +++++++++++++++++++++++++++++++++++-------------- fs/btrfs/ctree.h | 112 ++++++++++------- fs/btrfs/disk-io.c | 92 ++++++++------ fs/btrfs/disk-io.h | 9 +- fs/btrfs/extent-tree.c | 285 ++++++++++++++++++++++--------------------- fs/btrfs/extent_map.c | 91 ++++++++++---- fs/btrfs/extent_map.h | 10 +- fs/btrfs/file-item.c | 10 +- fs/btrfs/file.c | 126 ++++++++++--------- fs/btrfs/inode.c | 90 +++++++------- fs/btrfs/print-tree.c | 23 ++-- fs/btrfs/super.c | 6 +- fs/btrfs/sysfs.c | 5 +- fs/btrfs/transaction.c | 39 +++--- fs/btrfs/tree-defrag.c | 18 +-- 15 files changed, 738 insertions(+), 497 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 50ef351ef6b..34de83630ae 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -83,16 +83,15 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, WARN_ON(root->ref_cows && trans->transid != root->last_trans); - cow = btrfs_alloc_free_block(trans, root, search_start, empty_size); + cow = btrfs_alloc_free_block(trans, root, buf->len, + search_start, empty_size); if (IS_ERR(cow)) return PTR_ERR(cow); cow->alloc_addr = (unsigned long)__builtin_return_address(0); - if (buf->len != root->sectorsize || cow->len != root->sectorsize) - WARN_ON(1); copy_extent_buffer(cow, buf, 0, 0, cow->len); - btrfs_set_header_blocknr(cow, extent_buffer_blocknr(cow)); + btrfs_set_header_bytenr(cow, cow->start); btrfs_set_header_generation(cow, trans->transid); btrfs_set_header_owner(cow, root->root_key.objectid); @@ -110,16 +109,16 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, root->node = cow; extent_buffer_get(cow); if (buf != root->commit_root) { - btrfs_free_extent(trans, root, - extent_buffer_blocknr(buf), 1, 1); + btrfs_free_extent(trans, root, buf->start, + buf->len, 1); } free_extent_buffer(buf); } else { btrfs_set_node_blockptr(parent, parent_slot, - extent_buffer_blocknr(cow)); + cow->start); btrfs_mark_buffer_dirty(parent); WARN_ON(btrfs_header_generation(parent) != trans->transid); - btrfs_free_extent(trans, root, extent_buffer_blocknr(buf),1,1); + btrfs_free_extent(trans, root, buf->start, buf->len, 1); } free_extent_buffer(buf); btrfs_mark_buffer_dirty(cow); @@ -149,13 +148,14 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, return 0; } - search_start = extent_buffer_blocknr(buf) & ~((u64)65535); + search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1); ret = __btrfs_cow_block(trans, root, buf, parent, parent_slot, cow_ret, search_start, 0); (*cow_ret)->alloc_addr = (unsigned long)__builtin_return_address(0); return ret; } +#if 0 static int close_blocks(u64 blocknr, u64 other) { if (blocknr < other && other - blocknr < 8) @@ -165,7 +165,6 @@ static int close_blocks(u64 blocknr, u64 other) return 0; } -#if 0 static int should_defrag_leaf(struct extent_buffer *eb) { return 0; @@ -355,7 +354,7 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path, BUG_ON(memcmp(&parent_key, &node_key, sizeof(struct btrfs_disk_key))); BUG_ON(btrfs_node_blockptr(parent, parent_slot) != - btrfs_header_blocknr(node)); + btrfs_header_bytenr(node)); } BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root)); if (slot != 0) { @@ -398,7 +397,7 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, BUG_ON(memcmp(&parent_key, &leaf_key, sizeof(struct btrfs_disk_key))); BUG_ON(btrfs_node_blockptr(parent, parent_slot) != - btrfs_header_blocknr(leaf)); + btrfs_header_bytenr(leaf)); } #if 0 for (i = 0; nritems > 1 && i < nritems - 2; i++) { @@ -467,14 +466,16 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, static int check_block(struct btrfs_root *root, struct btrfs_path *path, int level) { +#if 0 struct extent_buffer *buf = path->nodes[level]; if (memcmp_extent_buffer(buf, root->fs_info->fsid, (unsigned long)btrfs_header_fsid(buf), BTRFS_FSID_SIZE)) { printk("warning bad block %Lu\n", buf->start); - BUG(); + return 1; } +#endif if (level == 0) return check_leaf(root, path, level); return check_node(root, path, level); @@ -585,7 +586,8 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root, return NULL; if (slot >= btrfs_header_nritems(parent)) return NULL; - return read_tree_block(root, btrfs_node_blockptr(parent, slot)); + return read_tree_block(root, btrfs_node_blockptr(parent, slot), + btrfs_level_size(root, btrfs_header_level(parent) - 1)); } static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root @@ -618,7 +620,6 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root */ if (!parent) { struct extent_buffer *child; - u64 blocknr = extent_buffer_blocknr(mid); if (btrfs_header_nritems(mid) != 1) return 0; @@ -632,9 +633,10 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root wait_on_tree_block_writeback(root, mid); /* once for the path */ free_extent_buffer(mid); + ret = btrfs_free_extent(trans, root, mid->start, mid->len, 1); /* once for the root ptr */ free_extent_buffer(mid); - return btrfs_free_extent(trans, root, blocknr, 1, 1); + return ret; } if (btrfs_header_nritems(mid) > BTRFS_NODEPTRS_PER_BLOCK(root) / 4) @@ -680,7 +682,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root if (wret < 0 && wret != -ENOSPC) ret = wret; if (btrfs_header_nritems(right) == 0) { - u64 blocknr = extent_buffer_blocknr(right); + u64 bytenr = right->start; + u32 blocksize = right->len; + clean_tree_block(trans, root, right); wait_on_tree_block_writeback(root, right); free_extent_buffer(right); @@ -689,7 +693,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root 1); if (wret) ret = wret; - wret = btrfs_free_extent(trans, root, blocknr, 1, 1); + wret = btrfs_free_extent(trans, root, bytenr, + blocksize, 1); if (wret) ret = wret; } else { @@ -719,7 +724,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root } if (btrfs_header_nritems(mid) == 0) { /* we've managed to empty the middle node, drop it */ - u64 blocknr = extent_buffer_blocknr(mid); + u64 bytenr = mid->start; + u32 blocksize = mid->len; clean_tree_block(trans, root, mid); wait_on_tree_block_writeback(root, mid); free_extent_buffer(mid); @@ -727,7 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root wret = del_ptr(trans, root, path, level + 1, pslot); if (wret) ret = wret; - wret = btrfs_free_extent(trans, root, blocknr, 1, 1); + wret = btrfs_free_extent(trans, root, bytenr, blocksize, 1); if (wret) ret = wret; } else { @@ -830,7 +836,6 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, path->slots[level] = orig_slot; free_extent_buffer(left); } - check_node(root, path, level); return 0; } free_extent_buffer(left); @@ -874,12 +879,10 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, } else { free_extent_buffer(right); } - check_node(root, path, level); return 0; } free_extent_buffer(right); } - check_node(root, path, level); return 1; } @@ -889,19 +892,23 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, int level, int slot) { + return; +#if 0 struct extent_buffer *node; int i; u32 nritems; - u64 blocknr; + u64 bytenr; u64 search; u64 cluster_start; int ret; int nread = 0; int direction = path->reada; + int level; struct radix_tree_root found; unsigned long gang[8]; struct extent_buffer *eb; + if (level == 0) return; @@ -918,8 +925,9 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, init_bit_radix(&found); nritems = btrfs_header_nritems(node); + level = btrfs_header_level(node) - 1; for (i = slot; i < nritems; i++) { - blocknr = btrfs_node_blockptr(node, i); + bytenr = btrfs_node_blockptr(node, i); set_radix_bit(&found, blocknr); } if (direction > 0) { @@ -944,6 +952,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, } } } +#endif } /* * look for key in the tree. path is filled in with nodes along the way @@ -963,7 +972,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root ins_len, int cow) { struct extent_buffer *b; - u64 blocknr; + u64 bytenr; int slot; int ret; int level; @@ -1027,10 +1036,11 @@ again: /* this is only true while dropping a snapshot */ if (level == lowest_level) break; - blocknr = btrfs_node_blockptr(b, slot); + bytenr = btrfs_node_blockptr(b, slot); if (should_reada) reada_for_search(root, p, level, slot); - b = read_tree_block(root, btrfs_node_blockptr(b, slot)); + b = read_tree_block(root, bytenr, + btrfs_level_size(root, level - 1)); } else { p->slots[level] = slot; if (ins_len > 0 && btrfs_leaf_free_space(root, b) < @@ -1193,14 +1203,14 @@ static int insert_new_root(struct btrfs_trans_handle *trans, BUG_ON(path->nodes[level]); BUG_ON(path->nodes[level-1] != root->node); - c = btrfs_alloc_free_block(trans, root, - extent_buffer_blocknr(root->node), 0); + c = btrfs_alloc_free_block(trans, root, root->nodesize, + root->node->start, 0); if (IS_ERR(c)) return PTR_ERR(c); memset_extent_buffer(c, 0, 0, root->nodesize); btrfs_set_header_nritems(c, 1); btrfs_set_header_level(c, level); - btrfs_set_header_blocknr(c, extent_buffer_blocknr(c)); + btrfs_set_header_bytenr(c, c->start); btrfs_set_header_generation(c, trans->transid); btrfs_set_header_owner(c, root->root_key.objectid); lower = path->nodes[level-1]; @@ -1213,7 +1223,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans, else btrfs_node_key(lower, &lower_key, 0); btrfs_set_node_key(c, &lower_key, 0); - btrfs_set_node_blockptr(c, 0, extent_buffer_blocknr(lower)); + btrfs_set_node_blockptr(c, 0, lower->start); btrfs_mark_buffer_dirty(c); @@ -1237,7 +1247,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans, */ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_disk_key - *key, u64 blocknr, int slot, int level) + *key, u64 bytenr, int slot, int level) { struct extent_buffer *lower; int nritems; @@ -1256,10 +1266,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root (nritems - slot) * sizeof(struct btrfs_key_ptr)); } btrfs_set_node_key(lower, key, slot); - btrfs_set_node_blockptr(lower, slot, blocknr); + btrfs_set_node_blockptr(lower, slot, bytenr); btrfs_set_header_nritems(lower, nritems + 1); btrfs_mark_buffer_dirty(lower); - check_node(root, path, level); return 0; } @@ -1300,14 +1309,14 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root } c_nritems = btrfs_header_nritems(c); - split = btrfs_alloc_free_block(trans, root, - extent_buffer_blocknr(c), 0); + split = btrfs_alloc_free_block(trans, root, root->nodesize, + c->start, 0); if (IS_ERR(split)) return PTR_ERR(split); btrfs_set_header_flags(split, btrfs_header_flags(c)); btrfs_set_header_level(split, btrfs_header_level(c)); - btrfs_set_header_blocknr(split, extent_buffer_blocknr(split)); + btrfs_set_header_bytenr(split, split->start); btrfs_set_header_generation(split, trans->transid); btrfs_set_header_owner(split, root->root_key.objectid); write_extent_buffer(split, root->fs_info->fsid, @@ -1328,8 +1337,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_mark_buffer_dirty(split); btrfs_node_key(split, &disk_key, 0); - wret = insert_ptr(trans, root, path, &disk_key, - extent_buffer_blocknr(split), + wret = insert_ptr(trans, root, path, &disk_key, split->start, path->slots[level + 1] + 1, level + 1); if (wret) @@ -1407,6 +1415,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root u32 left_nritems; u32 right_nritems; u32 data_end; + u32 this_item_size; int ret; slot = path->slots[1]; @@ -1417,7 +1426,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root if (slot >= btrfs_header_nritems(upper) - 1) return 1; - right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1)); + right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1), + root->leafsize); free_space = btrfs_leaf_free_space(root, right); if (free_space < data_size + sizeof(struct btrfs_item)) { free_extent_buffer(right); @@ -1445,13 +1455,27 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root for (i = left_nritems - 1; i >= 1; i--) { item = btrfs_item_nr(left, i); + if (path->slots[0] == i) push_space += data_size + sizeof(*item); - if (btrfs_item_size(left, item) + sizeof(*item) + push_space > - free_space) + + if (!left->map_token) { + map_extent_buffer(left, (unsigned long)item, + sizeof(struct btrfs_item), + &left->map_token, &left->kaddr, + &left->map_start, &left->map_len, + KM_USER1); + } + + this_item_size = btrfs_item_size(left, item); + if (this_item_size + sizeof(*item) + push_space > free_space) break; push_items++; - push_space += btrfs_item_size(left, item) + sizeof(*item); + push_space += this_item_size + sizeof(*item); + } + if (left->map_token) { + unmap_extent_buffer(left, left->map_token, KM_USER1); + left->map_token = NULL; } if (push_items == 0) { @@ -1493,11 +1517,23 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root right_nritems += push_items; btrfs_set_header_nritems(right, right_nritems); push_space = BTRFS_LEAF_DATA_SIZE(root); + for (i = 0; i < right_nritems; i++) { item = btrfs_item_nr(right, i); - btrfs_set_item_offset(right, item, push_space - - btrfs_item_size(right, item)); - push_space = btrfs_item_offset(right, item); + if (!right->map_token) { + map_extent_buffer(right, (unsigned long)item, + sizeof(struct btrfs_item), + &right->map_token, &right->kaddr, + &right->map_start, &right->map_len, + KM_USER1); + } + push_space -= btrfs_item_size(right, item); + btrfs_set_item_offset(right, item, push_space); + } + + if (right->map_token) { + unmap_extent_buffer(right, right->map_token, KM_USER1); + right->map_token = NULL; } left_nritems -= push_items; btrfs_set_header_nritems(left, left_nritems); @@ -1518,8 +1554,6 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root } else { free_extent_buffer(right); } - if (path->nodes[1]) - check_node(root, path, 1); return 0; } /* @@ -1542,6 +1576,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root u32 right_nritems; int ret = 0; int wret; + u32 this_item_size; + u32 old_left_item_size; slot = path->slots[1]; if (slot == 0) @@ -1550,7 +1586,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root return 1; left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1], - slot - 1)); + slot - 1), root->leafsize); free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { free_extent_buffer(left); @@ -1579,14 +1615,30 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root for (i = 0; i < right_nritems - 1; i++) { item = btrfs_item_nr(right, i); + if (!right->map_token) { + map_extent_buffer(right, (unsigned long)item, + sizeof(struct btrfs_item), + &right->map_token, &right->kaddr, + &right->map_start, &right->map_len, + KM_USER1); + } + if (path->slots[0] == i) push_space += data_size + sizeof(*item); - if (btrfs_item_size(right, item) + sizeof(*item) + push_space > - free_space) + + this_item_size = btrfs_item_size(right, item); + if (this_item_size + sizeof(*item) + push_space > free_space) break; + push_items++; - push_space += btrfs_item_size(right, item) + sizeof(*item); + push_space += this_item_size + sizeof(*item); + } + + if (right->map_token) { + unmap_extent_buffer(right, right->map_token, KM_USER1); + right->map_token = NULL; } + if (push_items == 0) { free_extent_buffer(left); return 1; @@ -1611,15 +1663,28 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root old_left_nritems = btrfs_header_nritems(left); BUG_ON(old_left_nritems < 0); + old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1); for (i = old_left_nritems; i < old_left_nritems + push_items; i++) { u32 ioff; + item = btrfs_item_nr(left, i); + if (!left->map_token) { + map_extent_buffer(left, (unsigned long)item, + sizeof(struct btrfs_item), + &left->map_token, &left->kaddr, + &left->map_start, &left->map_len, + KM_USER1); + } + ioff = btrfs_item_offset(left, item); btrfs_set_item_offset(left, item, - ioff - (BTRFS_LEAF_DATA_SIZE(root) - - btrfs_item_offset_nr(left, old_left_nritems - 1))); + ioff - (BTRFS_LEAF_DATA_SIZE(root) - old_left_item_size)); } btrfs_set_header_nritems(left, old_left_nritems + push_items); + if (left->map_token) { + unmap_extent_buffer(left, left->map_token, KM_USER1); + left->map_token = NULL; + } /* fixup right node */ push_space = btrfs_item_offset_nr(right, push_items - 1) - @@ -1640,9 +1705,21 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root for (i = 0; i < right_nritems; i++) { item = btrfs_item_nr(right, i); - btrfs_set_item_offset(right, item, push_space - - btrfs_item_size(right, item)); - push_space = btrfs_item_offset(right, item); + + if (!right->map_token) { + map_extent_buffer(right, (unsigned long)item, + sizeof(struct btrfs_item), + &right->map_token, &right->kaddr, + &right->map_start, &right->map_len, + KM_USER1); + } + + push_space = push_space - btrfs_item_size(right, item); + btrfs_set_item_offset(right, item, push_space); + } + if (right->map_token) { + unmap_extent_buffer(right, right->map_token, KM_USER1); + right->map_token = NULL; } btrfs_mark_buffer_dirty(left); @@ -1664,8 +1741,6 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root path->slots[0] -= push_items; } BUG_ON(path->slots[0] < 0); - if (path->nodes[1]) - check_node(root, path, 1); return ret; } @@ -1718,13 +1793,13 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root nritems = btrfs_header_nritems(l); mid = (nritems + 1)/ 2; - right = btrfs_alloc_free_block(trans, root, - extent_buffer_blocknr(l), 0); + right = btrfs_alloc_free_block(trans, root, root->leafsize, + l->start, 0); if (IS_ERR(right)) return PTR_ERR(right); memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header)); - btrfs_set_header_blocknr(right, extent_buffer_blocknr(right)); + btrfs_set_header_bytenr(right, right->start); btrfs_set_header_generation(right, trans->transid); btrfs_set_header_owner(right, root->root_key.objectid); btrfs_set_header_level(right, 0); @@ -1740,8 +1815,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_cpu_key_to_disk(&disk_key, ins_key); btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, - &disk_key, - extent_buffer_blocknr(right), + &disk_key, right->start, path->slots[1] + 1, 1); if (wret) ret = wret; @@ -1762,7 +1836,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, &disk_key, - extent_buffer_blocknr(right), + right->start, path->slots[1], 1); if (wret) ret = wret; @@ -1799,15 +1873,30 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root for (i = 0; i < nritems; i++) { struct btrfs_item *item = btrfs_item_nr(right, i); - u32 ioff = btrfs_item_offset(right, item); + u32 ioff; + + if (!right->map_token) { + map_extent_buffer(right, (unsigned long)item, + sizeof(struct btrfs_item), + &right->map_token, &right->kaddr, + &right->map_start, &right->map_len, + KM_USER1); + } + + ioff = btrfs_item_offset(right, item); btrfs_set_item_offset(right, item, ioff + rt_data_off); } + if (right->map_token) { + unmap_extent_buffer(right, right->map_token, KM_USER1); + right->map_token = NULL; + } + btrfs_set_header_nritems(l, mid); ret = 0; btrfs_item_key(right, &disk_key, 0); - wret = insert_ptr(trans, root, path, &disk_key, - extent_buffer_blocknr(right), path->slots[1] + 1, 1); + wret = insert_ptr(trans, root, path, &disk_key, right->start, + path->slots[1] + 1, 1); if (wret) ret = wret; @@ -1824,19 +1913,17 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root free_extent_buffer(right); BUG_ON(path->slots[0] < 0); - check_node(root, path, 1); - check_leaf(root, path, 0); if (!double_split) return ret; - right = btrfs_alloc_free_block(trans, root, - extent_buffer_blocknr(l), 0); + right = btrfs_alloc_free_block(trans, root, root->leafsize, + l->start, 0); if (IS_ERR(right)) return PTR_ERR(right); memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header)); - btrfs_set_header_blocknr(right, extent_buffer_blocknr(right)); + btrfs_set_header_bytenr(right, right->start); btrfs_set_header_generation(right, trans->transid); btrfs_set_header_owner(right, root->root_key.objectid); btrfs_set_header_level(right, 0); @@ -1847,8 +1934,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_cpu_key_to_disk(&disk_key, ins_key); btrfs_set_header_nritems(right, 0); wret = insert_ptr(trans, root, path, - &disk_key, - extent_buffer_blocknr(right), + &disk_key, right->start, path->slots[1], 1); if (wret) ret = wret; @@ -1860,8 +1946,6 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root free_extent_buffer(path->nodes[0]); path->nodes[0] = right; path->slots[0] = 0; - check_node(root, path, 1); - check_leaf(root, path, 0); return ret; } @@ -1904,9 +1988,24 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, for (i = slot; i < nritems; i++) { u32 ioff; item = btrfs_item_nr(leaf, i); + + if (!leaf->map_token) { + map_extent_buffer(leaf, (unsigned long)item, + sizeof(struct btrfs_item), + &leaf->map_token, &leaf->kaddr, + &leaf->map_start, &leaf->map_len, + KM_USER1); + } + ioff = btrfs_item_offset(leaf, item); btrfs_set_item_offset(leaf, item, ioff + size_diff); } + + if (leaf->map_token) { + unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); + leaf->map_token = NULL; + } + /* shift the data */ memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end + size_diff, btrfs_leaf_data(leaf) + @@ -1921,7 +2020,6 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, btrfs_print_leaf(root, leaf); BUG(); } - check_leaf(root, path, 0); return ret; } @@ -1963,10 +2061,23 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, for (i = slot; i < nritems; i++) { u32 ioff; item = btrfs_item_nr(leaf, i); + + if (!leaf->map_token) { + map_extent_buffer(leaf, (unsigned long)item, + sizeof(struct btrfs_item), + &leaf->map_token, &leaf->kaddr, + &leaf->map_start, &leaf->map_len, + KM_USER1); + } ioff = btrfs_item_offset(leaf, item); btrfs_set_item_offset(leaf, item, ioff - data_size); } + if (leaf->map_token) { + unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); + leaf->map_token = NULL; + } + /* shift the data */ memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end - data_size, btrfs_leaf_data(leaf) + @@ -1983,7 +2094,6 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, btrfs_print_leaf(root, leaf); BUG(); } - check_leaf(root, path, 0); return ret; } @@ -2046,12 +2156,26 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, * item0..itemN ... dataN.offset..dataN.size .. data0.size */ /* first correct the data pointers */ + WARN_ON(leaf->map_token); for (i = slot; i < nritems; i++) { u32 ioff; + item = btrfs_item_nr(leaf, i); + if (!leaf->map_token) { + map_extent_buffer(leaf, (unsigned long)item, + sizeof(struct btrfs_item), + &leaf->map_token, &leaf->kaddr, + &leaf->map_start, &leaf->map_len, + KM_USER1); + } + ioff = btrfs_item_offset(leaf, item); btrfs_set_item_offset(leaf, item, ioff - data_size); } + if (leaf->map_token) { + unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); + leaf->map_token = NULL; + } /* shift the items */ memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1), @@ -2081,7 +2205,6 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, btrfs_print_leaf(root, leaf); BUG(); } - check_leaf(root, path, 0); out: return ret; } @@ -2186,10 +2309,24 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, for (i = slot + 1; i < nritems; i++) { u32 ioff; + item = btrfs_item_nr(leaf, i); + if (!leaf->map_token) { + map_extent_buffer(leaf, (unsigned long)item, + sizeof(struct btrfs_item), + &leaf->map_token, &leaf->kaddr, + &leaf->map_start, &leaf->map_len, + KM_USER1); + } ioff = btrfs_item_offset(leaf, item); btrfs_set_item_offset(leaf, item, ioff + dsize); } + + if (leaf->map_token) { + unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); + leaf->map_token = NULL; + } + memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot), btrfs_item_nr_offset(slot + 1), sizeof(struct btrfs_item) * @@ -2209,8 +2346,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (wret) ret = wret; wret = btrfs_free_extent(trans, root, - extent_buffer_blocknr(leaf), - 1, 1); + leaf->start, leaf->len, 1); if (wret) ret = wret; } @@ -2247,7 +2383,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, } if (btrfs_header_nritems(leaf) == 0) { - u64 blocknr = extent_buffer_blocknr(leaf); + u64 bytenr = leaf->start; + u32 blocksize = leaf->len; clean_tree_block(trans, root, leaf); wait_on_tree_block_writeback(root, leaf); @@ -2257,8 +2394,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, ret = wret; free_extent_buffer(leaf); - wret = btrfs_free_extent(trans, root, blocknr, - 1, 1); + wret = btrfs_free_extent(trans, root, bytenr, + blocksize, 1); if (wret) ret = wret; } else { @@ -2281,7 +2418,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) { int slot; int level = 1; - u64 blocknr; + u64 bytenr; struct extent_buffer *c; struct extent_buffer *next = NULL; @@ -2296,14 +2433,15 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) continue; } - blocknr = btrfs_node_blockptr(c, slot); + bytenr = btrfs_node_blockptr(c, slot); if (next) free_extent_buffer(next); if (path->reada) reada_for_search(root, path, level, slot); - next = read_tree_block(root, blocknr); + next = read_tree_block(root, bytenr, + btrfs_level_size(root, level -1)); break; } path->slots[level] = slot; @@ -2317,7 +2455,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) break; if (path->reada) reada_for_search(root, path, level, 0); - next = read_tree_block(root, btrfs_node_blockptr(next, 0)); + next = read_tree_block(root, btrfs_node_blockptr(next, 0), + btrfs_level_size(root, level - 1)); } return 0; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 256689551eb..4d05456ec32 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -96,7 +96,7 @@ struct btrfs_key { struct btrfs_header { u8 csum[BTRFS_CSUM_SIZE]; u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */ - __le64 blocknr; /* which block this node is supposed to live in */ + __le64 bytenr; /* which block this node is supposed to live in */ __le64 generation; __le64 owner; __le32 nritems; @@ -122,16 +122,17 @@ struct btrfs_super_block { u8 csum[BTRFS_CSUM_SIZE]; /* the first 3 fields must match struct btrfs_header */ u8 fsid[16]; /* FS specific uuid */ - __le64 blocknr; /* this block number */ + __le64 bytenr; /* this block number */ __le64 magic; __le64 generation; __le64 root; - __le64 total_blocks; - __le64 blocks_used; + __le64 total_bytes; + __le64 bytes_used; __le64 root_dir_objectid; __le32 sectorsize; __le32 nodesize; __le32 leafsize; + u8 root_level; } __attribute__ ((__packed__)); /* @@ -231,13 +232,14 @@ struct btrfs_dir_item { struct btrfs_root_item { struct btrfs_inode_item inode; __le64 root_dirid; - __le64 blocknr; - __le64 block_limit; - __le64 blocks_used; + __le64 bytenr; + __le64 byte_limit; + __le64 bytes_used; __le32 flags; __le32 refs; struct btrfs_disk_key drop_progress; u8 drop_level; + u8 level; } __attribute__ ((__packed__)); #define BTRFS_FILE_EXTENT_REG 0 @@ -250,8 +252,8 @@ struct btrfs_file_extent_item { * disk space consumed by the extent, checksum blocks are included * in these numbers */ - __le64 disk_blocknr; - __le64 disk_num_blocks; + __le64 disk_bytenr; + __le64 disk_num_bytes; /* * the logical offset in file blocks (no csums) * this extent record is for. This allows a file extent to point @@ -263,7 +265,7 @@ struct btrfs_file_extent_item { /* * the logical number of file blocks (no csums included) */ - __le64 num_blocks; + __le64 num_bytes; } __attribute__ ((__packed__)); struct btrfs_csum_item { @@ -429,6 +431,7 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb, \ int err; \ char *map_token; \ char *kaddr; \ + int unmap_on_exit = (eb->map_token == NULL); \ unsigned long map_start; \ unsigned long map_len; \ unsigned long offset = (unsigned long)s + \ @@ -436,12 +439,13 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb, \ err = map_extent_buffer(eb, offset, \ sizeof(((type *)0)->member), \ &map_token, &kaddr, \ - &map_start, &map_len, KM_USER0); \ + &map_start, &map_len, KM_USER1); \ if (!err) { \ __le##bits *tmp = (__le##bits *)(kaddr + offset - \ map_start); \ u##bits res = le##bits##_to_cpu(*tmp); \ - unmap_extent_buffer(eb, map_token, KM_USER0); \ + if (unmap_on_exit) \ + unmap_extent_buffer(eb, map_token, KM_USER1); \ return res; \ } else { \ __le##bits res; \ @@ -457,17 +461,19 @@ static inline void btrfs_set_##name(struct extent_buffer *eb, \ char *kaddr; \ unsigned long map_start; \ unsigned long map_len; \ + int unmap_on_exit = (eb->map_token == NULL); \ unsigned long offset = (unsigned long)s + \ offsetof(type, member); \ err = map_extent_buffer(eb, offset, \ sizeof(((type *)0)->member), \ &map_token, &kaddr, \ - &map_start, &map_len, KM_USER0); \ + &map_start, &map_len, KM_USER1); \ if (!err) { \ __le##bits *tmp = (__le##bits *)(kaddr + offset - \ map_start); \ *tmp = cpu_to_le##bits(val); \ - unmap_extent_buffer(eb, map_token, KM_USER0); \ + if (unmap_on_exit) \ + unmap_extent_buffer(eb, map_token, KM_USER1); \ } else { \ val = cpu_to_le##bits(val); \ write_eb_member(eb, s, type, member, &val); \ @@ -483,15 +489,17 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb) \ unsigned long map_start; \ unsigned long map_len; \ unsigned long offset = offsetof(type, member); \ + int unmap_on_exit = (eb->map_token == NULL); \ err = map_extent_buffer(eb, offset, \ sizeof(((type *)0)->member), \ &map_token, &kaddr, \ - &map_start, &map_len, KM_USER0); \ + &map_start, &map_len, KM_USER1); \ if (!err) { \ __le##bits *tmp = (__le##bits *)(kaddr + offset - \ map_start); \ u##bits res = le##bits##_to_cpu(*tmp); \ - unmap_extent_buffer(eb, map_token, KM_USER0); \ + if (unmap_on_exit) \ + unmap_extent_buffer(eb, map_token, KM_USER1); \ return res; \ } else { \ __le##bits res; \ @@ -508,15 +516,17 @@ static inline void btrfs_set_##name(struct extent_buffer *eb, \ unsigned long map_start; \ unsigned long map_len; \ unsigned long offset = offsetof(type, member); \ + int unmap_on_exit = (eb->map_token == NULL); \ err = map_extent_buffer(eb, offset, \ sizeof(((type *)0)->member), \ &map_token, &kaddr, \ - &map_start, &map_len, KM_USER0); \ + &map_start, &map_len, KM_USER1); \ if (!err) { \ __le##bits *tmp = (__le##bits *)(kaddr + offset - \ map_start); \ *tmp = cpu_to_le##bits(val); \ - unmap_extent_buffer(eb, map_token, KM_USER0); \ + if (unmap_on_exit) \ + unmap_extent_buffer(eb, map_token, KM_USER1); \ } else { \ val = cpu_to_le##bits(val); \ write_eb_member(eb, NULL, type, member, &val); \ @@ -769,7 +779,7 @@ static inline void btrfs_set_key_type(struct btrfs_key *key, u8 val) } /* struct btrfs_header */ -BTRFS_SETGET_HEADER_FUNCS(header_blocknr, struct btrfs_header, blocknr, 64); +BTRFS_SETGET_HEADER_FUNCS(header_bytenr, struct btrfs_header, bytenr, 64); BTRFS_SETGET_HEADER_FUNCS(header_generation, struct btrfs_header, generation, 64); BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64); @@ -817,24 +827,28 @@ static inline int btrfs_is_leaf(struct extent_buffer *eb) /* struct btrfs_root_item */ BTRFS_SETGET_FUNCS(disk_root_refs, struct btrfs_root_item, refs, 32); -BTRFS_SETGET_FUNCS(disk_root_blocknr, struct btrfs_root_item, blocknr, 64); +BTRFS_SETGET_FUNCS(disk_root_bytenr, struct btrfs_root_item, bytenr, 64); +BTRFS_SETGET_FUNCS(disk_root_level, struct btrfs_root_item, level, 8); -BTRFS_SETGET_STACK_FUNCS(root_blocknr, struct btrfs_root_item, blocknr, 64); +BTRFS_SETGET_STACK_FUNCS(root_bytenr, struct btrfs_root_item, bytenr, 64); +BTRFS_SETGET_STACK_FUNCS(root_level, struct btrfs_root_item, level, 8); BTRFS_SETGET_STACK_FUNCS(root_dirid, struct btrfs_root_item, root_dirid, 64); BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32); BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32); -BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, blocks_used, 64); -BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, block_limit, 64); +BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, bytes_used, 64); +BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64); /* struct btrfs_super_block */ -BTRFS_SETGET_STACK_FUNCS(super_blocknr, struct btrfs_super_block, blocknr, 64); +BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64); BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block, generation, 64); BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64); -BTRFS_SETGET_STACK_FUNCS(super_total_blocks, struct btrfs_super_block, - total_blocks, 64); -BTRFS_SETGET_STACK_FUNCS(super_blocks_used, struct btrfs_super_block, - blocks_used, 64); +BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block, + root_level, 8); +BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block, + total_bytes, 64); +BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block, + bytes_used, 64); BTRFS_SETGET_STACK_FUNCS(super_sectorsize, struct btrfs_super_block, sectorsize, 32); BTRFS_SETGET_STACK_FUNCS(super_nodesize, struct btrfs_super_block, @@ -856,33 +870,33 @@ static inline unsigned long btrfs_file_extent_inline_start(struct btrfs_file_extent_item *e) { unsigned long offset = (unsigned long)e; - offset += offsetof(struct btrfs_file_extent_item, disk_blocknr); + offset += offsetof(struct btrfs_file_extent_item, disk_bytenr); return offset; } static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize) { - return offsetof(struct btrfs_file_extent_item, disk_blocknr) + datasize; + return offsetof(struct btrfs_file_extent_item, disk_bytenr) + datasize; } static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb, struct btrfs_item *e) { unsigned long offset; - offset = offsetof(struct btrfs_file_extent_item, disk_blocknr); + offset = offsetof(struct btrfs_file_extent_item, disk_bytenr); return btrfs_item_size(eb, e) - offset; } -BTRFS_SETGET_FUNCS(file_extent_disk_blocknr, struct btrfs_file_extent_item, - disk_blocknr, 64); +BTRFS_SETGET_FUNCS(file_extent_disk_bytenr, struct btrfs_file_extent_item, + disk_bytenr, 64); BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item, generation, 64); -BTRFS_SETGET_FUNCS(file_extent_disk_num_blocks, struct btrfs_file_extent_item, - disk_num_blocks, 64); +BTRFS_SETGET_FUNCS(file_extent_disk_num_bytes, struct btrfs_file_extent_item, + disk_num_bytes, 64); BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item, offset, 64); -BTRFS_SETGET_FUNCS(file_extent_num_blocks, struct btrfs_file_extent_item, - num_blocks, 64); +BTRFS_SETGET_FUNCS(file_extent_num_bytes, struct btrfs_file_extent_item, + num_bytes, 64); static inline struct btrfs_root *btrfs_sb(struct super_block *sb) { @@ -906,6 +920,12 @@ static inline int btrfs_set_root_name(struct btrfs_root *root, return 0; } +static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { + if (level == 0) + return root->leafsize; + return root->nodesize; +} + /* helper function to cast into the data area of the leaf. */ #define btrfs_item_ptr(leaf, slot, type) \ ((type *)(btrfs_leaf_data(leaf) + \ @@ -927,7 +947,7 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans, int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy); struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, - u64 blocknr); + u64 bytenr); struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *hint, u64 search_start, @@ -935,22 +955,22 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 hint, - u64 empty_size); + struct btrfs_root *root, u32 size, + u64 hint, u64 empty_size); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 owner, - u64 num_blocks, u64 empty_size, u64 search_start, + u64 num_bytes, u64 empty_size, u64 search_start, u64 search_end, struct btrfs_key *ins, int data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 blocknr, u64 num_blocks, int pin); + *root, u64 bytenr, u64 num_bytes, int pin); int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_map_tree *unpin); int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 blocknr, u64 num_blocks); + u64 bytenr, u64 num_bytes); int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_free_block_groups(struct btrfs_fs_info *info); @@ -1040,12 +1060,12 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, u64 offset, - u64 disk_num_blocks, - u64 num_blocks); + u64 disk_num_bytes, + u64 num_bytes); int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid, - u64 blocknr, int mod); + u64 bytenr, int mod); int btrfs_csum_file_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 offset, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 2b86a1d779b..fad9298c696 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -28,6 +28,7 @@ #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" +#include "print-tree.h" #if 0 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) @@ -43,26 +44,25 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) #endif struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, - u64 blocknr) + u64 bytenr, u32 blocksize) { struct inode *btree_inode = root->fs_info->btree_inode; struct extent_buffer *eb; eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, - blocknr * root->sectorsize, - root->sectorsize, GFP_NOFS); + bytenr, blocksize, GFP_NOFS); if (eb) eb->alloc_addr = (unsigned long)__builtin_return_address(0); return eb; } struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, - u64 blocknr) + u64 bytenr, u32 blocksize) { struct inode *btree_inode = root->fs_info->btree_inode; struct extent_buffer *eb; + eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, - blocknr * root->sectorsize, - root->sectorsize, GFP_NOFS); + bytenr, blocksize, GFP_NOFS); eb->alloc_addr = (unsigned long)__builtin_return_address(0); return eb; } @@ -208,13 +208,13 @@ static struct address_space_operations btree_aops = { .sync_page = block_sync_page, }; -int readahead_tree_block(struct btrfs_root *root, u64 blocknr) +int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) { struct extent_buffer *buf = NULL; struct inode *btree_inode = root->fs_info->btree_inode; int ret = 0; - buf = btrfs_find_create_tree_block(root, blocknr); + buf = btrfs_find_create_tree_block(root, bytenr, blocksize); if (!buf) return 0; read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, @@ -223,12 +223,13 @@ int readahead_tree_block(struct btrfs_root *root, u64 blocknr) return ret; } -struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr) +struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, + u32 blocksize) { struct extent_buffer *buf = NULL; struct inode *btree_inode = root->fs_info->btree_inode; - buf = btrfs_find_create_tree_block(root, blocknr); + buf = btrfs_find_create_tree_block(root, bytenr, blocksize); if (!buf) return NULL; read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, @@ -261,7 +262,7 @@ int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf) return 0; } -static int __setup_root(int blocksize, +static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, struct btrfs_root *root, struct btrfs_fs_info *fs_info, u64 objectid) @@ -269,9 +270,9 @@ static int __setup_root(int blocksize, root->node = NULL; root->inode = NULL; root->commit_root = NULL; - root->sectorsize = blocksize; - root->nodesize = blocksize; - root->leafsize = blocksize; + root->sectorsize = sectorsize; + root->nodesize = nodesize; + root->leafsize = leafsize; root->ref_cows = 0; root->fs_info = fs_info; root->objectid = objectid; @@ -291,21 +292,23 @@ static int __setup_root(int blocksize, return 0; } -static int find_and_setup_root(int blocksize, - struct btrfs_root *tree_root, +static int find_and_setup_root(struct btrfs_root *tree_root, struct btrfs_fs_info *fs_info, u64 objectid, struct btrfs_root *root) { int ret; + u32 blocksize; - __setup_root(blocksize, root, fs_info, objectid); + __setup_root(tree_root->nodesize, tree_root->leafsize, + tree_root->sectorsize, root, fs_info, objectid); ret = btrfs_find_last_root(tree_root, objectid, &root->root_item, &root->root_key); BUG_ON(ret); - root->node = read_tree_block(root, - btrfs_root_blocknr(&root->root_item)); + blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); + root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), + blocksize); BUG_ON(!root->node); return 0; } @@ -318,14 +321,14 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, struct btrfs_path *path; struct extent_buffer *l; u64 highest_inode; + u32 blocksize; int ret = 0; root = kzalloc(sizeof(*root), GFP_NOFS); if (!root) return ERR_PTR(-ENOMEM); if (location->offset == (u64)-1) { - ret = find_and_setup_root(fs_info->sb->s_blocksize, - fs_info->tree_root, fs_info, + ret = find_and_setup_root(tree_root, fs_info, location->objectid, root); if (ret) { kfree(root); @@ -334,7 +337,8 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, goto insert; } - __setup_root(fs_info->sb->s_blocksize, root, fs_info, + __setup_root(tree_root->nodesize, tree_root->leafsize, + tree_root->sectorsize, root, fs_info, location->objectid); path = btrfs_alloc_path(); @@ -357,8 +361,9 @@ out: kfree(root); return ERR_PTR(ret); } - root->node = read_tree_block(root, - btrfs_root_blocknr(&root->root_item)); + blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); + root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), + blocksize); BUG_ON(!root->node); insert: root->ref_cows = 1; @@ -418,6 +423,10 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *open_ctree(struct super_block *sb) { + u32 sectorsize; + u32 nodesize; + u32 leafsize; + u32 blocksize; struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), GFP_NOFS); struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), @@ -474,12 +483,12 @@ struct btrfs_root *open_ctree(struct super_block *sb) mutex_init(&fs_info->trans_mutex); mutex_init(&fs_info->fs_mutex); - __setup_root(sb->s_blocksize, tree_root, + __setup_root(512, 512, 512, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); fs_info->sb_buffer = read_tree_block(tree_root, - BTRFS_SUPER_INFO_OFFSET / - sb->s_blocksize); + BTRFS_SUPER_INFO_OFFSET, + 512); if (!fs_info->sb_buffer) goto fail_iput; @@ -494,9 +503,15 @@ struct btrfs_root *open_ctree(struct super_block *sb) if (!btrfs_super_root(disk_super)) goto fail_sb_buffer; + nodesize = btrfs_super_nodesize(disk_super); + leafsize = btrfs_super_leafsize(disk_super); + sectorsize = btrfs_super_sectorsize(disk_super); + tree_root->nodesize = nodesize; + tree_root->leafsize = leafsize; + tree_root->sectorsize = sectorsize; + i_size_write(fs_info->btree_inode, - btrfs_super_total_blocks(disk_super) << - fs_info->btree_inode->i_blkbits); + btrfs_super_total_bytes(disk_super)); if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, @@ -504,13 +519,22 @@ struct btrfs_root *open_ctree(struct super_block *sb) printk("btrfs: valid FS not found on %s\n", sb->s_id); goto fail_sb_buffer; } + blocksize = btrfs_level_size(tree_root, + btrfs_super_root_level(disk_super)); tree_root->node = read_tree_block(tree_root, - btrfs_super_root(disk_super)); + btrfs_super_root(disk_super), + blocksize); if (!tree_root->node) goto fail_sb_buffer; +#if 0 + btrfs_print_leaf(tree_root, tree_root->node); + err = -EIO; + goto fail_tree_root; +#endif mutex_lock(&fs_info->fs_mutex); - ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info, + + ret = find_and_setup_root(tree_root, fs_info, BTRFS_EXTENT_TREE_OBJECTID, extent_root); if (ret) { mutex_unlock(&fs_info->fs_mutex); @@ -611,11 +635,11 @@ int close_ctree(struct btrfs_root *root) free_extent_buffer(fs_info->tree_root->node); free_extent_buffer(fs_info->sb_buffer); - truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); - iput(fs_info->btree_inode); btrfs_free_block_groups(root->fs_info); del_fs_roots(fs_info); + truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); + iput(fs_info->btree_inode); kfree(fs_info->extent_root); kfree(fs_info->tree_root); return 0; @@ -642,7 +666,7 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) if (transid != root->fs_info->generation) { printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", - (unsigned long long)extent_buffer_blocknr(buf), + (unsigned long long)buf->start, transid, root->fs_info->generation); WARN_ON(1); } diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 70d9413c599..fd4db5f810c 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -21,10 +21,11 @@ #define BTRFS_SUPER_INFO_OFFSET (16 * 1024) -struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr); -int readahead_tree_block(struct btrfs_root *root, u64 blocknr); +struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, + u32 blocksize); +int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize); struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, - u64 blocknr); + u64 bytenr, u32 blocksize); int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf); struct btrfs_root *open_ctree(struct super_block *sb); @@ -32,7 +33,7 @@ int close_ctree(struct btrfs_root *root); int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, - u64 blocknr); + u64 bytenr, u32 blocksize); struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_key *location, const char *name, int namelen); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 477466d167a..1be8f9f04a1 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -129,7 +129,7 @@ err: struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, - u64 blocknr) + u64 bytenr) { struct extent_map_tree *block_group_cache; struct btrfs_block_group_cache *block_group = NULL; @@ -140,7 +140,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct block_group_cache = &info->block_group_cache; ret = find_first_extent_bit(block_group_cache, - blocknr, &start, &end, + bytenr, &start, &end, BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA); if (ret) { return NULL; @@ -152,7 +152,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct block_group = (struct btrfs_block_group_cache *)ptr; - if (block_group->key.objectid <= blocknr && blocknr <= + if (block_group->key.objectid <= bytenr && bytenr <= block_group->key.objectid + block_group->key.offset) return block_group; @@ -315,7 +315,7 @@ found: int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 blocknr, u64 num_blocks) + u64 bytenr, u64 num_bytes) { struct btrfs_path *path; int ret; @@ -324,13 +324,14 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_extent_item *item; u32 refs; + WARN_ON(num_bytes < root->sectorsize); path = btrfs_alloc_path(); if (!path) return -ENOMEM; - key.objectid = blocknr; + key.objectid = bytenr; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); - key.offset = num_blocks; + key.offset = num_bytes; ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 1); if (ret < 0) @@ -361,8 +362,8 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans, } static int lookup_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 blocknr, - u64 num_blocks, u32 *refs) + struct btrfs_root *root, u64 bytenr, + u64 num_bytes, u32 *refs) { struct btrfs_path *path; int ret; @@ -370,9 +371,10 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans, struct extent_buffer *l; struct btrfs_extent_item *item; + WARN_ON(num_bytes < root->sectorsize); path = btrfs_alloc_path(); - key.objectid = blocknr; - key.offset = num_blocks; + key.objectid = bytenr; + key.offset = num_bytes; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 0); @@ -380,7 +382,7 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans, goto out; if (ret != 0) { btrfs_print_leaf(root, path->nodes[0]); - printk("failed to find block number %Lu\n", blocknr); + printk("failed to find block number %Lu\n", bytenr); BUG(); } l = path->nodes[0]; @@ -394,19 +396,19 @@ out: int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - return btrfs_inc_extent_ref(trans, root, - extent_buffer_blocknr(root->node), 1); + return btrfs_inc_extent_ref(trans, root, root->node->start, + root->node->len); } int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf) { - u64 blocknr; + u64 bytenr; u32 nritems; struct btrfs_key key; struct btrfs_file_extent_item *fi; int i; - int leaf; + int level; int ret; int faili; int err; @@ -414,11 +416,11 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (!root->ref_cows) return 0; - leaf = btrfs_is_leaf(buf); + level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); for (i = 0; i < nritems; i++) { - if (leaf) { - u64 disk_blocknr; + if (level == 0) { + u64 disk_bytenr; btrfs_item_key_to_cpu(buf, &key, i); if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; @@ -427,18 +429,19 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (btrfs_file_extent_type(buf, fi) == BTRFS_FILE_EXTENT_INLINE) continue; - disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi); - if (disk_blocknr == 0) + disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); + if (disk_bytenr == 0) continue; - ret = btrfs_inc_extent_ref(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(buf, fi)); + ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, + btrfs_file_extent_disk_num_bytes(buf, fi)); if (ret) { faili = i; goto fail; } } else { - blocknr = btrfs_node_blockptr(buf, i); - ret = btrfs_inc_extent_ref(trans, root, blocknr, 1); + bytenr = btrfs_node_blockptr(buf, i); + ret = btrfs_inc_extent_ref(trans, root, bytenr, + btrfs_level_size(root, level - 1)); if (ret) { faili = i; goto fail; @@ -449,8 +452,8 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, fail: WARN_ON(1); for (i =0; i < faili; i++) { - if (leaf) { - u64 disk_blocknr; + if (level == 0) { + u64 disk_bytenr; btrfs_item_key_to_cpu(buf, &key, i); if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; @@ -459,16 +462,17 @@ fail: if (btrfs_file_extent_type(buf, fi) == BTRFS_FILE_EXTENT_INLINE) continue; - disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi); - if (disk_blocknr == 0) + disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); + if (disk_bytenr == 0) continue; - err = btrfs_free_extent(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(buf, + err = btrfs_free_extent(trans, root, disk_bytenr, + btrfs_file_extent_disk_num_bytes(buf, fi), 0); BUG_ON(err); } else { - blocknr = btrfs_node_blockptr(buf, i); - err = btrfs_free_extent(trans, root, blocknr, 1, 0); + bytenr = btrfs_node_blockptr(buf, i); + err = btrfs_free_extent(trans, root, bytenr, + btrfs_level_size(root, level - 1), 0); BUG_ON(err); } } @@ -558,31 +562,31 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, static int update_block_group(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 blocknr, u64 num, int alloc, int mark_free, - int data) + u64 bytenr, u64 num_bytes, int alloc, + int mark_free, int data) { struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; - u64 total = num; + u64 total = num_bytes; u64 old_val; - u64 block_in_group; + u64 byte_in_group; u64 start; u64 end; while(total) { - cache = btrfs_lookup_block_group(info, blocknr); + cache = btrfs_lookup_block_group(info, bytenr); if (!cache) { return -1; } - block_in_group = blocknr - cache->key.objectid; - WARN_ON(block_in_group > cache->key.offset); + byte_in_group = bytenr - cache->key.objectid; + WARN_ON(byte_in_group > cache->key.offset); start = cache->key.objectid; end = start + cache->key.offset - 1; set_extent_bits(&info->block_group_cache, start, end, BLOCK_GROUP_DIRTY, GFP_NOFS); old_val = btrfs_block_group_used(&cache->item); - num = min(total, cache->key.offset - block_in_group); + num_bytes = min(total, cache->key.offset - byte_in_group); if (alloc) { if (cache->data != data && old_val < (cache->key.offset >> 1)) { @@ -608,18 +612,18 @@ static int update_block_group(struct btrfs_trans_handle *trans, start, end, bit_to_set, GFP_NOFS); } - old_val += num; + old_val += num_bytes; } else { - old_val -= num; + old_val -= num_bytes; if (mark_free) { set_extent_dirty(&info->free_space_cache, - blocknr, blocknr + num - 1, + bytenr, bytenr + num_bytes - 1, GFP_NOFS); } } btrfs_set_block_group_used(&cache->item, old_val); - total -= num; - blocknr += num; + total -= num_bytes; + bytenr += num_bytes; } return 0; } @@ -701,13 +705,14 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct return 0; } -static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending) +static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, + int pending) { int err = 0; struct extent_buffer *buf; if (!pending) { - buf = btrfs_find_tree_block(root, blocknr); + buf = btrfs_find_tree_block(root, bytenr, num_bytes); if (buf) { if (btrfs_buffer_uptodate(buf)) { u64 transid = @@ -720,10 +725,11 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending) free_extent_buffer(buf); } set_extent_dirty(&root->fs_info->pinned_extents, - blocknr, blocknr, GFP_NOFS); + bytenr, bytenr + num_bytes - 1, GFP_NOFS); } else { set_extent_bits(&root->fs_info->pending_del, - blocknr, blocknr, EXTENT_LOCKED, GFP_NOFS); + bytenr, bytenr + num_bytes - 1, + EXTENT_LOCKED, GFP_NOFS); } BUG_ON(err < 0); return 0; @@ -733,7 +739,7 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending) * remove an extent from the root, returns 0 on success */ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 blocknr, u64 num_blocks, int pin, + *root, u64 bytenr, u64 num_bytes, int pin, int mark_free) { struct btrfs_path *path; @@ -745,9 +751,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_extent_item *ei; u32 refs; - key.objectid = blocknr; + key.objectid = bytenr; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); - key.offset = num_blocks; + key.offset = num_bytes; path = btrfs_alloc_path(); if (!path) @@ -768,28 +774,29 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_mark_buffer_dirty(leaf); if (refs == 0) { - u64 super_blocks_used, root_blocks_used; + u64 super_used; + u64 root_used; if (pin) { - ret = pin_down_block(root, blocknr, 0); + ret = pin_down_bytes(root, bytenr, num_bytes, 0); BUG_ON(ret); } /* block accounting for super block */ - super_blocks_used = btrfs_super_blocks_used(&info->super_copy); - btrfs_set_super_blocks_used(&info->super_copy, - super_blocks_used - num_blocks); + super_used = btrfs_super_bytes_used(&info->super_copy); + btrfs_set_super_bytes_used(&info->super_copy, + super_used - num_bytes); /* block accounting for root item */ - root_blocks_used = btrfs_root_used(&root->root_item); + root_used = btrfs_root_used(&root->root_item); btrfs_set_root_used(&root->root_item, - root_blocks_used - num_blocks); + root_used - num_bytes); ret = btrfs_del_item(trans, extent_root, path); if (ret) { return ret; } - ret = update_block_group(trans, root, blocknr, num_blocks, 0, + ret = update_block_group(trans, root, bytenr, num_bytes, 0, mark_free, 0); BUG_ON(ret); } @@ -836,17 +843,18 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct * remove an extent from the root, returns 0 on success */ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 blocknr, u64 num_blocks, int pin) + *root, u64 bytenr, u64 num_bytes, int pin) { struct btrfs_root *extent_root = root->fs_info->extent_root; int pending_ret; int ret; + WARN_ON(num_bytes < root->sectorsize); if (root == extent_root) { - pin_down_block(root, blocknr, 1); + pin_down_bytes(root, bytenr, num_bytes, 1); return 0; } - ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0); + ret = __free_extent(trans, root, bytenr, num_bytes, pin, pin == 0); pending_ret = del_pending_extents(trans, root->fs_info->extent_root); return ret ? ret : pending_ret; } @@ -860,8 +868,8 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root * Any available blocks before search_start are skipped. */ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *orig_root, u64 num_blocks, u64 empty_size, - u64 search_start, u64 search_end, u64 hint_block, + *orig_root, u64 num_bytes, u64 empty_size, + u64 search_start, u64 search_end, u64 hint_byte, struct btrfs_key *ins, u64 exclude_start, u64 exclude_nr, int data) { @@ -870,30 +878,29 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root int ret; u64 hole_size = 0; int slot = 0; - u64 last_block = 0; + u64 last_byte = 0; u64 orig_search_start = search_start; int start_found; struct extent_buffer *l; struct btrfs_root * root = orig_root->fs_info->extent_root; struct btrfs_fs_info *info = root->fs_info; - int total_needed = num_blocks; + u64 total_needed = num_bytes; int level; struct btrfs_block_group_cache *block_group; int full_scan = 0; int wrapped = 0; - u64 cached_search_start = 0; - WARN_ON(num_blocks < 1); + WARN_ON(num_bytes < root->sectorsize); btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); level = btrfs_header_level(root->node); if (search_end == (u64)-1) - search_end = btrfs_super_total_blocks(&info->super_copy); - if (hint_block) { - block_group = btrfs_lookup_block_group(info, hint_block); + search_end = btrfs_super_total_bytes(&info->super_copy); + if (hint_byte) { + block_group = btrfs_lookup_block_group(info, hint_byte); block_group = btrfs_find_block_group(root, block_group, - hint_block, data, 1); + hint_byte, data, 1); } else { block_group = btrfs_find_block_group(root, trans->block_group, 0, @@ -906,7 +913,6 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root check_failed: search_start = find_search_start(root, &block_group, search_start, total_needed, data); - cached_search_start = search_start; btrfs_init_path(path); ins->objectid = search_start; @@ -958,27 +964,27 @@ check_failed: start_found = 1; goto check_pending; } - ins->objectid = last_block > search_start ? - last_block : search_start; + ins->objectid = last_byte > search_start ? + last_byte : search_start; ins->offset = search_end - ins->objectid; goto check_pending; } btrfs_item_key_to_cpu(l, &key, slot); - if (key.objectid >= search_start && key.objectid > last_block && + if (key.objectid >= search_start && key.objectid > last_byte && start_found) { - if (last_block < search_start) - last_block = search_start; - hole_size = key.objectid - last_block; - if (hole_size >= num_blocks) { - ins->objectid = last_block; + if (last_byte < search_start) + last_byte = search_start; + hole_size = key.objectid - last_byte; + if (hole_size >= num_bytes) { + ins->objectid = last_byte; ins->offset = hole_size; goto check_pending; } } if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) { if (!start_found) { - last_block = key.objectid; + last_byte = key.objectid; start_found = 1; } goto next; @@ -986,9 +992,9 @@ check_failed: start_found = 1; - last_block = key.objectid + key.offset; + last_byte = key.objectid + key.offset; - if (!full_scan && last_block >= block_group->key.objectid + + if (!full_scan && last_byte >= block_group->key.objectid + block_group->key.offset) { btrfs_release_path(root, path); search_start = block_group->key.objectid + @@ -1006,20 +1012,20 @@ check_pending: btrfs_release_path(root, path); BUG_ON(ins->objectid < search_start); - if (ins->objectid + num_blocks >= search_end) + if (ins->objectid + num_bytes >= search_end) goto enospc; if (test_range_bit(&info->extent_ins, ins->objectid, - ins->objectid + num_blocks -1, EXTENT_LOCKED, 0)) { - search_start = ins->objectid + num_blocks; + ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { + search_start = ins->objectid + num_bytes; goto new_group; } if (test_range_bit(&info->pinned_extents, ins->objectid, - ins->objectid + num_blocks -1, EXTENT_DIRTY, 0)) { - search_start = ins->objectid + num_blocks; + ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { + search_start = ins->objectid + num_bytes; goto new_group; } - if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start && + if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && ins->objectid < exclude_start + exclude_nr)) { search_start = exclude_start + exclude_nr; goto new_group; @@ -1029,12 +1035,12 @@ check_pending: if (block_group) trans->block_group = block_group; } - ins->offset = num_blocks; + ins->offset = num_bytes; btrfs_free_path(path); return 0; new_group: - if (search_start + num_blocks >= search_end) { + if (search_start + num_bytes >= search_end) { enospc: search_start = orig_search_start; if (full_scan) { @@ -1069,12 +1075,12 @@ error: */ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 owner, - u64 num_blocks, u64 empty_size, u64 hint_block, + u64 num_bytes, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, int data) { int ret; int pending_ret; - u64 super_blocks_used, root_blocks_used; + u64 super_used, root_used; u64 search_start = 0; struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; @@ -1083,9 +1089,9 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, btrfs_set_stack_extent_refs(&extent_item, 1); btrfs_set_stack_extent_owner(&extent_item, owner); - WARN_ON(num_blocks < 1); - ret = find_free_extent(trans, root, num_blocks, empty_size, - search_start, search_end, hint_block, ins, + WARN_ON(num_bytes < root->sectorsize); + ret = find_free_extent(trans, root, num_bytes, empty_size, + search_start, search_end, hint_byte, ins, trans->alloc_exclude_start, trans->alloc_exclude_nr, data); BUG_ON(ret); @@ -1093,21 +1099,18 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, return ret; /* block accounting for super block */ - super_blocks_used = btrfs_super_blocks_used(&info->super_copy); - btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used + - num_blocks); + super_used = btrfs_super_bytes_used(&info->super_copy); + btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); /* block accounting for root item */ - root_blocks_used = btrfs_root_used(&root->root_item); - btrfs_set_root_used(&root->root_item, root_blocks_used + - num_blocks); + root_used = btrfs_root_used(&root->root_item); + btrfs_set_root_used(&root->root_item, root_used + num_bytes); clear_extent_dirty(&root->fs_info->free_space_cache, ins->objectid, ins->objectid + ins->offset - 1, GFP_NOFS); if (root == extent_root) { - BUG_ON(num_blocks != 1); set_extent_bits(&root->fs_info->extent_ins, ins->objectid, ins->objectid + ins->offset - 1, EXTENT_LOCKED, GFP_NOFS); @@ -1146,7 +1149,8 @@ update_block: * returns the tree buffer or NULL. */ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 hint, + struct btrfs_root *root, + u32 blocksize, u64 hint, u64 empty_size) { struct btrfs_key ins; @@ -1154,14 +1158,15 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct extent_buffer *buf; ret = btrfs_alloc_extent(trans, root, root->root_key.objectid, - 1, empty_size, hint, (u64)-1, &ins, 0); + blocksize, empty_size, hint, + (u64)-1, &ins, 0); if (ret) { BUG_ON(ret > 0); return ERR_PTR(ret); } - buf = btrfs_find_create_tree_block(root, ins.objectid); + buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); if (!buf) { - btrfs_free_extent(trans, root, ins.objectid, 1, 0); + btrfs_free_extent(trans, root, ins.objectid, blocksize, 0); return ERR_PTR(-ENOMEM); } btrfs_set_buffer_uptodate(buf); @@ -1191,7 +1196,7 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans, BUG_ON(!btrfs_is_leaf(leaf)); nritems = btrfs_header_nritems(leaf); for (i = 0; i < nritems; i++) { - u64 disk_blocknr; + u64 disk_bytenr; btrfs_item_key_to_cpu(leaf, &key, i); if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) @@ -1204,11 +1209,11 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans, * FIXME make sure to insert a trans record that * repeats the snapshot del on crash */ - disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi); - if (disk_blocknr == 0) + disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); + if (disk_bytenr == 0) continue; - ret = btrfs_free_extent(trans, root, disk_blocknr, - btrfs_file_extent_disk_num_blocks(leaf, fi), 0); + ret = btrfs_free_extent(trans, root, disk_bytenr, + btrfs_file_extent_disk_num_bytes(leaf, fi), 0); BUG_ON(ret); } return 0; @@ -1219,19 +1224,23 @@ static void reada_walk_down(struct btrfs_root *root, { int i; u32 nritems; - u64 blocknr; + u64 bytenr; int ret; u32 refs; + int level; + u32 blocksize; nritems = btrfs_header_nritems(node); + level = btrfs_header_level(node); for (i = 0; i < nritems; i++) { - blocknr = btrfs_node_blockptr(node, i); - ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs); + bytenr = btrfs_node_blockptr(node, i); + blocksize = btrfs_level_size(root, level - 1); + ret = lookup_extent_ref(NULL, root, bytenr, blocksize, &refs); BUG_ON(ret); if (refs != 1) continue; mutex_unlock(&root->fs_info->fs_mutex); - ret = readahead_tree_block(root, blocknr); + ret = readahead_tree_block(root, bytenr, blocksize); cond_resched(); mutex_lock(&root->fs_info->fs_mutex); if (ret) @@ -1248,15 +1257,16 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root { struct extent_buffer *next; struct extent_buffer *cur; - u64 blocknr; + u64 bytenr; + u32 blocksize; int ret; u32 refs; WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); ret = lookup_extent_ref(trans, root, - extent_buffer_blocknr(path->nodes[*level]), - 1, &refs); + path->nodes[*level]->start, + path->nodes[*level]->len, &refs); BUG_ON(ret); if (refs > 1) goto out; @@ -1283,30 +1293,33 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(ret); break; } - blocknr = btrfs_node_blockptr(cur, path->slots[*level]); - ret = lookup_extent_ref(trans, root, blocknr, 1, &refs); + bytenr = btrfs_node_blockptr(cur, path->slots[*level]); + blocksize = btrfs_level_size(root, *level - 1); + ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); BUG_ON(ret); if (refs != 1) { path->slots[*level]++; - ret = btrfs_free_extent(trans, root, blocknr, 1, 1); + ret = btrfs_free_extent(trans, root, bytenr, + blocksize, 1); BUG_ON(ret); continue; } - next = btrfs_find_tree_block(root, blocknr); + next = btrfs_find_tree_block(root, bytenr, blocksize); if (!next || !btrfs_buffer_uptodate(next)) { free_extent_buffer(next); mutex_unlock(&root->fs_info->fs_mutex); - next = read_tree_block(root, blocknr); + next = read_tree_block(root, bytenr, blocksize); mutex_lock(&root->fs_info->fs_mutex); /* we dropped the lock, check one more time */ - ret = lookup_extent_ref(trans, root, blocknr, 1, &refs); + ret = lookup_extent_ref(trans, root, bytenr, + blocksize, &refs); BUG_ON(ret); if (refs != 1) { path->slots[*level]++; free_extent_buffer(next); ret = btrfs_free_extent(trans, root, - blocknr, 1, 1); + bytenr, blocksize, 1); BUG_ON(ret); continue; } @@ -1321,8 +1334,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root out: WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); - ret = btrfs_free_extent(trans, root, - extent_buffer_blocknr(path->nodes[*level]), 1, 1); + ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, + path->nodes[*level]->len, 1); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; @@ -1359,8 +1372,8 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root return 0; } else { ret = btrfs_free_extent(trans, root, - extent_buffer_blocknr(path->nodes[*level]), - 1, 1); + path->nodes[*level]->start, + path->nodes[*level]->len, 1); BUG_ON(ret); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; @@ -1476,16 +1489,12 @@ int btrfs_read_block_groups(struct btrfs_root *root) struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; - u64 group_size_blocks; block_group_cache = &info->block_group_cache; - group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >> - info->sb->s_blocksize_bits; - root = info->extent_root; key.objectid = 0; - key.offset = group_size_blocks; + key.offset = BTRFS_BLOCK_GROUP_SIZE; btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); path = btrfs_alloc_path(); @@ -1532,7 +1541,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) (u64)cache); if (key.objectid >= - btrfs_super_total_blocks(&info->super_copy)) + btrfs_super_total_bytes(&info->super_copy)) break; } diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index e081558d52f..f658703c42e 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1963,18 +1963,27 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, int i) struct page *p; if (i == 0) return eb->first_page; + i += eb->start >> PAGE_CACHE_SHIFT; + if (eb->last_page && eb->last_page->index == i) + return eb->last_page; + p = find_get_page(eb->first_page->mapping, i); page_cache_release(p); + eb->last_page = p; return p; } +static inline unsigned long num_extent_pages(u64 start, u64 len) +{ + return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT); +} struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, u64 start, unsigned long len, gfp_t mask) { - unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) - - (start >> PAGE_CACHE_SHIFT) + 1; + unsigned long num_pages = num_extent_pages(start, len); unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; @@ -1986,7 +1995,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, if (!eb || IS_ERR(eb)) return NULL; - eb->alloc_addr = __builtin_return_address(0); + eb->alloc_addr = (unsigned long)__builtin_return_address(0); eb->start = start; eb->len = len; atomic_set(&eb->refs, 1); @@ -1994,6 +2003,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, for (i = 0; i < num_pages; i++, index++) { p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); if (!p) { + WARN_ON(1); /* make sure the free only frees the pages we've * grabbed a reference on */ @@ -2021,8 +2031,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, u64 start, unsigned long len, gfp_t mask) { - unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) - - (start >> PAGE_CACHE_SHIFT) + 1; + unsigned long num_pages = num_extent_pages(start, len); unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; @@ -2033,7 +2042,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, if (!eb || IS_ERR(eb)) return NULL; - eb->alloc_addr = __builtin_return_address(0); + eb->alloc_addr = (unsigned long)__builtin_return_address(0); eb->start = start; eb->len = len; atomic_set(&eb->refs, 1); @@ -2070,8 +2079,7 @@ void free_extent_buffer(struct extent_buffer *eb) if (!atomic_dec_and_test(&eb->refs)) return; - num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - - (eb->start >> PAGE_CACHE_SHIFT) + 1; + num_pages = num_extent_pages(eb->start, eb->len); if (eb->first_page) page_cache_release(eb->first_page); @@ -2094,8 +2102,7 @@ int clear_extent_buffer_dirty(struct extent_map_tree *tree, u64 end = start + eb->len - 1; set = clear_extent_dirty(tree, start, end, GFP_NOFS); - num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - - (eb->start >> PAGE_CACHE_SHIFT) + 1; + num_pages = num_extent_pages(eb->start, eb->len); for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); @@ -2145,8 +2152,7 @@ int set_extent_buffer_uptodate(struct extent_map_tree *tree, struct page *page; unsigned long num_pages; - num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - - (eb->start >> PAGE_CACHE_SHIFT) + 1; + num_pages = num_extent_pages(eb->start, eb->len); set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); @@ -2191,8 +2197,7 @@ int read_extent_buffer_pages(struct extent_map_tree *tree, return 0; } - num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) - - (eb->start >> PAGE_CACHE_SHIFT) + 1; + num_pages = num_extent_pages(eb->start, eb->len); for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); if (PageUptodate(page)) { @@ -2267,14 +2272,14 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, } EXPORT_SYMBOL(read_extent_buffer); -int map_extent_buffer(struct extent_buffer *eb, unsigned long start, - unsigned long min_len, - char **token, char **map, - unsigned long *map_start, - unsigned long *map_len, int km) +static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start, + unsigned long min_len, char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) { size_t offset = start & (PAGE_CACHE_SIZE - 1); char *kaddr; + struct page *p; size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; unsigned long end_i = (start_offset + start + min_len) >> @@ -2283,21 +2288,59 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start, if (i != end_i) return -EINVAL; - WARN_ON(start > eb->len); + if (start >= eb->len) { + printk("bad start in map eb start %Lu len %lu caller start %lu min %lu\n", eb->start, eb->len, start, min_len); + WARN_ON(1); + } if (i == 0) { offset = start_offset; *map_start = 0; } else { + offset = 0; *map_start = (i << PAGE_CACHE_SHIFT) - start_offset; } - kaddr = kmap_atomic(extent_buffer_page(eb, i), km); + p = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(p)); + kaddr = kmap_atomic(p, km); *token = kaddr; *map = kaddr + offset; *map_len = PAGE_CACHE_SIZE - offset; return 0; } + +int map_extent_buffer(struct extent_buffer *eb, unsigned long start, + unsigned long min_len, + char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + int err; + int save = 0; + if (eb->map_token) { + if (start >= eb->map_start && + start + min_len <= eb->map_start + eb->map_len) { + *token = eb->map_token; + *map = eb->kaddr; + *map_start = eb->map_start; + *map_len = eb->map_len; + return 0; + } + unmap_extent_buffer(eb, eb->map_token, km); + eb->map_token = NULL; + save = 1; + } + err = __map_extent_buffer(eb, start, min_len, token, map, + map_start, map_len, km); + if (!err && save) { + eb->map_token = *token; + eb->kaddr = *map; + eb->map_start = *map_start; + eb->map_len = *map_len; + } + return err; +} EXPORT_SYMBOL(map_extent_buffer); void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) @@ -2574,7 +2617,6 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, ((unsigned long)PAGE_CACHE_SIZE - 1); src_off_in_page = src_end & ((unsigned long)PAGE_CACHE_SIZE - 1); - if (src_i == 0) src_off_in_page += start_offset; if (dst_i == 0) @@ -2582,14 +2624,13 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, cur = min(len, src_off_in_page + 1); cur = min(cur, dst_off_in_page + 1); - move_pages(extent_buffer_page(dst, dst_i), extent_buffer_page(dst, src_i), dst_off_in_page - cur + 1, src_off_in_page - cur + 1, cur); - dst_end -= cur - 1; - src_end -= cur - 1; + dst_end -= cur; + src_end -= cur; len -= cur; } } diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 75dc600dc0f..4ef8bdd6834 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -70,7 +70,12 @@ struct extent_buffer { struct list_head list; struct list_head leak_list; struct page *first_page; + struct page *last_page; unsigned long alloc_addr; + char *map_token; + char *kaddr; + unsigned long map_start; + unsigned long map_len; }; typedef struct extent_map *(get_extent_t)(struct inode *inode, @@ -147,11 +152,6 @@ static inline void extent_buffer_get(struct extent_buffer *eb) atomic_inc(&eb->refs); } -static inline u64 extent_buffer_blocknr(struct extent_buffer *eb) -{ - return eb->start / 4096; -} - int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, unsigned long start, unsigned long len); diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 226f6d028c3..9a65e97a4e2 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -27,8 +27,8 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, - u64 offset, u64 disk_num_blocks, - u64 num_blocks) + u64 offset, u64 disk_num_bytes, + u64 num_bytes) { int ret = 0; struct btrfs_file_extent_item *item; @@ -50,10 +50,10 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, leaf = path->nodes[0]; item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_disk_blocknr(leaf, item, offset); - btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks); + btrfs_set_file_extent_disk_bytenr(leaf, item, offset); + btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes); btrfs_set_file_extent_offset(leaf, item, 0); - btrfs_set_file_extent_num_blocks(leaf, item, num_blocks); + btrfs_set_file_extent_num_bytes(leaf, item, num_bytes); btrfs_set_file_extent_generation(leaf, item, trans->transid); btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG); btrfs_mark_buffer_dirty(leaf); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1734ca69555..844d8807e44 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -120,9 +120,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); ptr = btrfs_file_extent_inline_start(ei); - kaddr = kmap_atomic(page, KM_USER0); + kaddr = kmap_atomic(page, KM_USER1); write_extent_buffer(leaf, kaddr + page_offset, ptr, size); - kunmap_atomic(kaddr, KM_USER0); + kunmap_atomic(kaddr, KM_USER1); btrfs_mark_buffer_dirty(leaf); fail: btrfs_free_path(path); @@ -142,11 +142,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, struct inode *inode = file->f_path.dentry->d_inode; struct extent_map *em; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - u64 hint_block; - u64 num_blocks; + u64 hint_byte; + u64 num_bytes; u64 start_pos; u64 end_of_last_block; u64 end_pos = pos + write_bytes; + u32 inline_size; loff_t isize = i_size_read(inode); em = alloc_extent_map(GFP_NOFS); @@ -156,11 +157,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, em->bdev = inode->i_sb->s_bdev; start_pos = pos & ~((u64)root->sectorsize - 1); - num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >> - inode->i_blkbits; + num_bytes = (write_bytes + pos - start_pos + + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); down_read(&BTRFS_I(inode)->root->snap_sem); - end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1; + end_of_last_block = start_pos + num_bytes - 1; + lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); @@ -169,8 +171,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, goto out_unlock; } btrfs_set_trans_block_group(trans, inode); - inode->i_blocks += num_blocks << 3; - hint_block = 0; + inode->i_blocks += num_bytes >> 9; + hint_byte = 0; if ((end_of_last_block & 4095) == 0) { printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); @@ -191,11 +193,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, err = btrfs_drop_extents(trans, root, inode, last_pos_in_file, last_pos_in_file + hole_size, - &hint_block); + &hint_byte); if (err) goto failed; - hole_size >>= inode->i_blkbits; err = btrfs_insert_file_extent(trans, root, inode->i_ino, last_pos_in_file, @@ -209,8 +210,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, * either allocate an extent for the new bytes or setup the key * to show we are doing inline data in the extent */ + inline_size = end_pos - start_pos; if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || - pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) { + inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root) || + inline_size >= PAGE_CACHE_SIZE) { u64 last_end; for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; @@ -224,10 +227,9 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, } else { struct page *p = pages[0]; /* step one, delete the existing extents in this range */ - /* FIXME blocksize != pagesize */ err = btrfs_drop_extents(trans, root, inode, start_pos, (pos + write_bytes + root->sectorsize -1) & - ~((u64)root->sectorsize - 1), &hint_block); + ~((u64)root->sectorsize - 1), &hint_byte); if (err) goto failed; @@ -283,7 +285,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) */ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - u64 start, u64 end, u64 *hint_block) + u64 start, u64 end, u64 *hint_byte) { int ret; struct btrfs_key key; @@ -346,8 +348,7 @@ next_slot: found_type = btrfs_file_extent_type(leaf, extent); if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = key.offset + - (btrfs_file_extent_num_blocks(leaf, extent) << - inode->i_blkbits); + btrfs_file_extent_num_bytes(leaf, extent); found_extent = 1; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { struct btrfs_item *item; @@ -386,17 +387,17 @@ next_slot: if (end < extent_end && end >= key.offset) { if (found_extent) { - u64 disk_blocknr = - btrfs_file_extent_disk_blocknr(leaf,extent); - u64 disk_num_blocks = - btrfs_file_extent_disk_num_blocks(leaf, + u64 disk_bytenr = + btrfs_file_extent_disk_bytenr(leaf, extent); + u64 disk_num_bytes = + btrfs_file_extent_disk_num_bytes(leaf, extent); read_extent_buffer(leaf, &old, (unsigned long)extent, sizeof(old)); - if (disk_blocknr != 0) { + if (disk_bytenr != 0) { ret = btrfs_inc_extent_ref(trans, root, - disk_blocknr, disk_num_blocks); + disk_bytenr, disk_num_bytes); BUG_ON(ret); } } @@ -410,21 +411,19 @@ next_slot: keep = 1; WARN_ON(start & (root->sectorsize - 1)); if (found_extent) { - new_num = (start - key.offset) >> - inode->i_blkbits; - old_num = btrfs_file_extent_num_blocks(leaf, - extent); - *hint_block = - btrfs_file_extent_disk_blocknr(leaf, - extent); - if (btrfs_file_extent_disk_blocknr(leaf, - extent)) { + new_num = start - key.offset; + old_num = btrfs_file_extent_num_bytes(leaf, + extent); + *hint_byte = + btrfs_file_extent_disk_bytenr(leaf, + extent); + if (btrfs_file_extent_disk_bytenr(leaf, + extent)) { inode->i_blocks -= - (old_num - new_num) << 3; + (old_num - new_num) >> 9; } - btrfs_set_file_extent_num_blocks(leaf, - extent, - new_num); + btrfs_set_file_extent_num_bytes(leaf, extent, + new_num); btrfs_mark_buffer_dirty(leaf); } else { WARN_ON(1); @@ -432,33 +431,32 @@ next_slot: } /* delete the entire extent */ if (!keep) { - u64 disk_blocknr = 0; - u64 disk_num_blocks = 0; - u64 extent_num_blocks = 0; + u64 disk_bytenr = 0; + u64 disk_num_bytes = 0; + u64 extent_num_bytes = 0; if (found_extent) { - disk_blocknr = - btrfs_file_extent_disk_blocknr(leaf, + disk_bytenr = + btrfs_file_extent_disk_bytenr(leaf, extent); - disk_num_blocks = - btrfs_file_extent_disk_num_blocks(leaf, - extent); - extent_num_blocks = - btrfs_file_extent_num_blocks(leaf, - extent); - *hint_block = - btrfs_file_extent_disk_blocknr(leaf, + disk_num_bytes = + btrfs_file_extent_disk_num_bytes(leaf, extent); + extent_num_bytes = + btrfs_file_extent_num_bytes(leaf, extent); + *hint_byte = + btrfs_file_extent_disk_bytenr(leaf, + extent); } ret = btrfs_del_item(trans, root, path); /* TODO update progress marker and return */ BUG_ON(ret); btrfs_release_path(root, path); extent = NULL; - if (found_extent && disk_blocknr != 0) { - inode->i_blocks -= extent_num_blocks << 3; + if (found_extent && disk_bytenr != 0) { + inode->i_blocks -= extent_num_bytes >> 9; ret = btrfs_free_extent(trans, root, - disk_blocknr, - disk_num_blocks, 0); + disk_bytenr, + disk_num_bytes, 0); } BUG_ON(ret); @@ -491,20 +489,19 @@ next_slot: (unsigned long)extent, sizeof(old)); btrfs_set_file_extent_offset(leaf, extent, - le64_to_cpu(old.offset) + - ((end - key.offset) >> inode->i_blkbits)); - WARN_ON(le64_to_cpu(old.num_blocks) < - (extent_end - end) >> inode->i_blkbits); - btrfs_set_file_extent_num_blocks(leaf, extent, - (extent_end - end) >> inode->i_blkbits); - + le64_to_cpu(old.offset) + end - key.offset); + WARN_ON(le64_to_cpu(old.num_bytes) < + (extent_end - end)); + btrfs_set_file_extent_num_bytes(leaf, extent, + extent_end - end); btrfs_set_file_extent_type(leaf, extent, BTRFS_FILE_EXTENT_REG); + btrfs_mark_buffer_dirty(path->nodes[0]); - if (le64_to_cpu(old.disk_blocknr) != 0) { + if (le64_to_cpu(old.disk_bytenr) != 0) { inode->i_blocks += - btrfs_file_extent_num_blocks(leaf, - extent) << 3; + btrfs_file_extent_num_bytes(leaf, + extent) >> 9; } ret = 0; goto out; @@ -531,12 +528,9 @@ static int prepare_pages(struct btrfs_root *root, unsigned long index = pos >> PAGE_CACHE_SHIFT; struct inode *inode = file->f_path.dentry->d_inode; int err = 0; - u64 num_blocks; u64 start_pos; start_pos = pos & ~((u64)root->sectorsize - 1); - num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >> - inode->i_blkbits; memset(pages, 0, num_pages * sizeof(struct page *)); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index fbe2836364e..d6b3a55ed8e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -77,19 +77,19 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; struct btrfs_key ins; u64 alloc_hint = 0; - u64 num_blocks; + u64 num_bytes; int ret; - u64 blocksize = 1 << inode->i_blkbits; + u64 blocksize = root->sectorsize; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); BUG_ON(!trans); - num_blocks = (end - start + blocksize) & ~(blocksize - 1); + num_bytes = (end - start + blocksize) & ~(blocksize - 1); ret = btrfs_drop_extents(trans, root, inode, - start, start + num_blocks, &alloc_hint); - num_blocks = num_blocks >> inode->i_blkbits; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0, + start, start + num_bytes, &alloc_hint); + + ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { WARN_ON(1); @@ -186,7 +186,8 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) zeroit: printk("btrfs csum failed ino %lu off %llu\n", page->mapping->host->i_ino, (unsigned long long)start); - memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); + memset(kaddr + offset, 1, end - start + 1); + flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); return 0; } @@ -547,7 +548,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, struct extent_buffer *leaf; struct btrfs_file_extent_item *fi; u64 extent_start = 0; - u64 extent_num_blocks = 0; + u64 extent_num_bytes = 0; u64 item_end = 0; int found_extent; int del_item; @@ -593,8 +594,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_INLINE) { item_end += - btrfs_file_extent_num_blocks(leaf, fi) << - inode->i_blkbits; + btrfs_file_extent_num_bytes(leaf, fi); } } if (found_type == BTRFS_CSUM_ITEM_KEY) { @@ -626,28 +626,27 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_INLINE) { u64 num_dec; - extent_start = btrfs_file_extent_disk_blocknr(leaf, fi); + extent_start = btrfs_file_extent_disk_bytenr(leaf, fi); if (!del_item) { - u64 orig_num_blocks = - btrfs_file_extent_num_blocks(leaf, fi); - extent_num_blocks = inode->i_size - + u64 orig_num_bytes = + btrfs_file_extent_num_bytes(leaf, fi); + extent_num_bytes = inode->i_size - found_key.offset + root->sectorsize - 1; - extent_num_blocks >>= inode->i_blkbits; - btrfs_set_file_extent_num_blocks(leaf, fi, - extent_num_blocks); - num_dec = (orig_num_blocks - - extent_num_blocks) << 3; + btrfs_set_file_extent_num_bytes(leaf, fi, + extent_num_bytes); + num_dec = (orig_num_bytes - + extent_num_bytes) >> 9; if (extent_start != 0) { inode->i_blocks -= num_dec; } btrfs_mark_buffer_dirty(leaf); } else { - extent_num_blocks = - btrfs_file_extent_disk_num_blocks(leaf, - fi); + extent_num_bytes = + btrfs_file_extent_disk_num_bytes(leaf, + fi); /* FIXME blocksize != 4096 */ - num_dec = btrfs_file_extent_num_blocks(leaf, - fi) << 3; + num_dec = btrfs_file_extent_num_bytes(leaf, + fi) >> 9; if (extent_start != 0) { found_extent = 1; inode->i_blocks -= num_dec; @@ -664,7 +663,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, btrfs_release_path(root, path); if (found_extent) { ret = btrfs_free_extent(trans, root, extent_start, - extent_num_blocks, 0); + extent_num_bytes, 0); BUG_ON(ret); } } @@ -709,7 +708,8 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, static int btrfs_truncate_page(struct address_space *mapping, loff_t from) { struct inode *inode = mapping->host; - unsigned blocksize = 1 << inode->i_blkbits; + struct btrfs_root *root = BTRFS_I(inode)->root; + u32 blocksize = root->sectorsize; pgoff_t index = from >> PAGE_CACHE_SHIFT; unsigned offset = from & (PAGE_CACHE_SIZE-1); struct page *page; @@ -719,7 +719,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) if ((offset & (blocksize - 1)) == 0) goto out; - down_read(&BTRFS_I(inode)->root->snap_sem); + down_read(&root->snap_sem); ret = -ENOMEM; page = grab_cache_page(mapping, index); if (!page) @@ -778,8 +778,6 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) err = btrfs_drop_extents(trans, root, inode, pos, pos + hole_size, &alloc_hint); - hole_size >>= inode->i_blkbits; - err = btrfs_insert_file_extent(trans, root, inode->i_ino, pos, 0, 0, hole_size); btrfs_end_transaction(trans, root); @@ -1490,7 +1488,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, { int ret; int err = 0; - u64 blocknr; + u64 bytenr; u64 extent_start = 0; u64 extent_end = 0; u64 objectid = inode->i_ino; @@ -1540,10 +1538,6 @@ again: leaf = path->nodes[0]; item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - - blocknr = btrfs_file_extent_disk_blocknr(leaf, item); - blocknr += btrfs_file_extent_offset(leaf, item); - /* are we inside the extent that was found? */ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); found_type = btrfs_key_type(&found_key); @@ -1556,8 +1550,7 @@ again: extent_start = found_key.offset; if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = extent_start + - (btrfs_file_extent_num_blocks(leaf, item) << - inode->i_blkbits); + btrfs_file_extent_num_bytes(leaf, item); err = 0; if (start < extent_start || start >= extent_end) { em->start = start; @@ -1570,17 +1563,18 @@ again: } goto not_found_em; } - if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) { + bytenr = btrfs_file_extent_disk_bytenr(leaf, item); + if (bytenr == 0) { em->start = extent_start; em->end = extent_end - 1; em->block_start = EXTENT_MAP_HOLE; em->block_end = EXTENT_MAP_HOLE; goto insert; } - em->block_start = blocknr << inode->i_blkbits; + bytenr += btrfs_file_extent_offset(leaf, item); + em->block_start = bytenr; em->block_end = em->block_start + - (btrfs_file_extent_num_blocks(leaf, item) << - inode->i_blkbits) - 1; + btrfs_file_extent_num_bytes(leaf, item) - 1; em->start = extent_start; em->end = extent_end - 1; goto insert; @@ -1592,7 +1586,8 @@ again: size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, path->slots[0])); - extent_end = extent_start | ((u64)root->sectorsize - 1); + extent_end = (extent_start + size) | + ((u64)root->sectorsize - 1); if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { @@ -1617,8 +1612,10 @@ again: ptr = btrfs_file_extent_inline_start(item); map = kmap(page); read_extent_buffer(leaf, map + page_offset, ptr, size); + /* memset(map + page_offset + size, 0, root->sectorsize - (page_offset + size)); + */ flush_dcache_page(page); kunmap(page); set_extent_uptodate(em_tree, extent_start, @@ -1836,13 +1833,13 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - leaf = btrfs_alloc_free_block(trans, root, 0, 0); + leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0); if (IS_ERR(leaf)) return PTR_ERR(leaf); btrfs_set_header_nritems(leaf, 0); btrfs_set_header_level(leaf, 0); - btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf)); + btrfs_set_header_bytenr(leaf, leaf->start); btrfs_set_header_generation(leaf, trans->transid); btrfs_set_header_owner(leaf, root->root_key.objectid); write_extent_buffer(leaf, root->fs_info->fsid, @@ -1858,7 +1855,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) inode_item->nblocks = cpu_to_le64(1); inode_item->mode = cpu_to_le32(S_IFDIR | 0755); - btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf)); + btrfs_set_root_bytenr(&root_item, leaf->start); + btrfs_set_root_level(&root_item, 0); btrfs_set_root_refs(&root_item, 1); btrfs_set_root_used(&root_item, 0); @@ -1971,8 +1969,8 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); - btrfs_set_root_blocknr(&new_root_item, - extent_buffer_blocknr(root->node)); + btrfs_set_root_bytenr(&new_root_item, root->node->start); + btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index a825ce078a5..9f8696c8a8e 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -36,7 +36,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) u32 type; printk("leaf %llu total ptrs %d free space %d\n", - (unsigned long long)btrfs_header_blocknr(l), nr, + (unsigned long long)btrfs_header_bytenr(l), nr, btrfs_leaf_free_space(root, l)); for (i = 0 ; i < nr ; i++) { item = btrfs_item_nr(l, i); @@ -65,8 +65,8 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) break; case BTRFS_ROOT_ITEM_KEY: ri = btrfs_item_ptr(l, i, struct btrfs_root_item); - printk("\t\troot data blocknr %llu refs %u\n", - (unsigned long long)btrfs_disk_root_blocknr(l, ri), + printk("\t\troot data bytenr %llu refs %u\n", + (unsigned long long)btrfs_disk_root_bytenr(l, ri), btrfs_disk_root_refs(l, ri)); break; case BTRFS_EXTENT_ITEM_KEY: @@ -84,12 +84,12 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) btrfs_file_extent_inline_len(l, item)); break; } - printk("\t\textent data disk block %llu nr %llu\n", - (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi), - (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi)); + printk("\t\textent data disk bytenr %llu nr %llu\n", + (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi), + (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi)); printk("\t\textent data offset %llu nr %llu\n", (unsigned long long)btrfs_file_extent_offset(l, fi), - (unsigned long long)btrfs_file_extent_num_blocks(l, fi)); + (unsigned long long)btrfs_file_extent_num_bytes(l, fi)); break; case BTRFS_BLOCK_GROUP_ITEM_KEY: bi = btrfs_item_ptr(l, i, @@ -106,16 +106,18 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c) int i; u32 nr; struct btrfs_key key; + int level; if (!c) return; nr = btrfs_header_nritems(c); - if (btrfs_is_leaf(c)) { + level = btrfs_header_level(c); + if (level == 0) { btrfs_print_leaf(root, c); return; } printk("node %llu level %d total ptrs %d free spc %u\n", - (unsigned long long)btrfs_header_blocknr(c), + (unsigned long long)btrfs_header_bytenr(c), btrfs_header_level(c), nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr); for (i = 0; i < nr; i++) { @@ -129,7 +131,8 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c) } for (i = 0; i < nr; i++) { struct extent_buffer *next = read_tree_block(root, - btrfs_node_blockptr(c, i)); + btrfs_node_blockptr(c, i), + btrfs_level_size(root, level - 1)); if (btrfs_is_leaf(next) && btrfs_header_level(c) != 1) BUG(); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 39a1435c68f..5c4370f3a5b 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -303,10 +303,12 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) { struct btrfs_root *root = btrfs_sb(dentry->d_sb); struct btrfs_super_block *disk_super = &root->fs_info->super_copy; + int bits = dentry->d_sb->s_blocksize_bits; buf->f_namelen = BTRFS_NAME_LEN; - buf->f_blocks = btrfs_super_total_blocks(disk_super); - buf->f_bfree = buf->f_blocks - btrfs_super_blocks_used(disk_super); + buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; + buf->f_bfree = buf->f_blocks - + (btrfs_super_bytes_used(disk_super) >> bits); buf->f_bavail = buf->f_bfree; buf->f_bsize = dentry->d_sb->s_blocksize; buf->f_type = BTRFS_SUPER_MAGIC; diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index 9654e90eec8..0bd1fd3d29d 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -42,14 +42,15 @@ static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf) static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf) { + return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_super_blocks_used(&fs->super_copy)); + (unsigned long long)btrfs_super_bytes_used(&fs->super_copy)); } static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", - (unsigned long long)btrfs_super_total_blocks(&fs->super_copy)); + (unsigned long long)btrfs_super_total_bytes(&fs->super_copy)); } static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 55289b71056..60f61345a8d 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -205,12 +205,13 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, btrfs_write_dirty_block_groups(trans, extent_root); while(1) { - old_extent_block = btrfs_root_blocknr(&extent_root->root_item); - if (old_extent_block == - extent_buffer_blocknr(extent_root->node)) + old_extent_block = btrfs_root_bytenr(&extent_root->root_item); + if (old_extent_block == extent_root->node->start) break; - btrfs_set_root_blocknr(&extent_root->root_item, - extent_buffer_blocknr(extent_root->node)); + btrfs_set_root_bytenr(&extent_root->root_item, + extent_root->node->start); + btrfs_set_root_level(&extent_root->root_item, + btrfs_header_level(extent_root->node)); ret = btrfs_update_root(trans, tree_root, &extent_root->root_key, &extent_root->root_item); @@ -284,8 +285,8 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, (unsigned long)root->root_key.objectid, BTRFS_ROOT_TRANS_TAG); if (root->commit_root == root->node) { - WARN_ON(extent_buffer_blocknr(root->node) != - btrfs_root_blocknr(&root->root_item)); + WARN_ON(root->node->start != + btrfs_root_bytenr(&root->root_item)); free_extent_buffer(root->commit_root); root->commit_root = NULL; @@ -314,8 +315,10 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans, root->commit_root = NULL; root->root_key.offset = root->fs_info->generation; - btrfs_set_root_blocknr(&root->root_item, - extent_buffer_blocknr(root->node)); + btrfs_set_root_bytenr(&root->root_item, + root->node->start); + btrfs_set_root_level(&root->root_item, + btrfs_header_level(root->node)); err = btrfs_insert_root(trans, root->fs_info->tree_root, &root->root_key, &root->root_item); @@ -407,8 +410,8 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, struct dirty_root *dirty; struct btrfs_trans_handle *trans; unsigned long nr; - u64 num_blocks; - u64 blocks_used; + u64 num_bytes; + u64 bytes_used; int ret = 0; int err; @@ -419,7 +422,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, dirty = list_entry(list->next, struct dirty_root, list); list_del_init(&dirty->list); - num_blocks = btrfs_root_used(&dirty->root->root_item); + num_bytes = btrfs_root_used(&dirty->root->root_item); root = dirty->latest_root; while(1) { @@ -446,12 +449,12 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, } BUG_ON(ret); - num_blocks -= btrfs_root_used(&dirty->root->root_item); - blocks_used = btrfs_root_used(&root->root_item); - if (num_blocks) { + num_bytes -= btrfs_root_used(&dirty->root->root_item); + bytes_used = btrfs_root_used(&root->root_item); + if (num_bytes) { record_root_in_trans(root); btrfs_set_root_used(&root->root_item, - blocks_used - num_blocks); + bytes_used - num_bytes); } ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); if (ret) { @@ -560,7 +563,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_set_super_generation(&root->fs_info->super_copy, cur_trans->transid); btrfs_set_super_root(&root->fs_info->super_copy, - extent_buffer_blocknr(root->fs_info->tree_root->node)); + root->fs_info->tree_root->node->start); + btrfs_set_super_root_level(&root->fs_info->super_copy, + btrfs_header_level(root->fs_info->tree_root->node)); write_extent_buffer(root->fs_info->sb_buffer, &root->fs_info->super_copy, 0, diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index daf019afa0a..3feac2f2862 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -27,13 +27,15 @@ static void reada_defrag(struct btrfs_root *root, { int i; u32 nritems; - u64 blocknr; + u64 bytenr; + u32 blocksize; int ret; + blocksize = btrfs_level_size(root, btrfs_header_level(node) - 1); nritems = btrfs_header_nritems(node); for (i = 0; i < nritems; i++) { - blocknr = btrfs_node_blockptr(node, i); - ret = readahead_tree_block(root, blocknr); + bytenr = btrfs_node_blockptr(node, i); + ret = readahead_tree_block(root, bytenr, blocksize); if (ret) break; } @@ -46,7 +48,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, { struct extent_buffer *next; struct extent_buffer *cur; - u64 blocknr; + u64 bytenr; int ret = 0; int is_extent = 0; @@ -80,10 +82,11 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, break; } - blocknr = btrfs_node_blockptr(cur, path->slots[*level]); + bytenr = btrfs_node_blockptr(cur, path->slots[*level]); if (cache_only) { - next = btrfs_find_tree_block(root, blocknr); + next = btrfs_find_tree_block(root, bytenr, + btrfs_level_size(root, *level - 1)); /* FIXME, test for defrag */ if (!next || !btrfs_buffer_uptodate(next)) { free_extent_buffer(next); @@ -91,7 +94,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, continue; } } else { - next = read_tree_block(root, blocknr); + next = read_tree_block(root, bytenr, + btrfs_level_size(root, *level - 1)); } ret = btrfs_cow_block(trans, root, next, path->nodes[*level], path->slots[*level], &next); -- cgit v1.2.3-70-g09d2 From 3326d1b07c0cb6a2ff5b835b7a2cffa54124d074 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Oct 2007 16:18:25 -0400 Subject: Btrfs: Allow tails larger than one page Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 24 +++++++-- fs/btrfs/ctree.h | 2 +- fs/btrfs/file.c | 152 ++++++++++++++++++++++++++++++++++++++++++------------- fs/btrfs/inode.c | 32 +++++++----- fs/btrfs/super.c | 2 - 5 files changed, 156 insertions(+), 56 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index bef61ee883e..f691542dfa1 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1766,8 +1766,9 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root /* first try to make some room by pushing left and right */ wret = push_leaf_left(trans, root, path, data_size); - if (wret < 0) + if (wret < 0) { return wret; + } if (wret) { wret = push_leaf_right(trans, root, path, data_size); if (wret < 0) @@ -1777,8 +1778,9 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root /* did the pushes work? */ if (btrfs_leaf_free_space(root, l) >= - sizeof(struct btrfs_item) + data_size) + sizeof(struct btrfs_item) + data_size) { return 0; + } if (!path->nodes[1]) { ret = insert_new_root(trans, root, path, 1); @@ -1822,7 +1824,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root return ret; } mid = slot; - double_split = 1; + if (mid != nritems && + leaf_space_used(l, mid, nritems - mid) + + space_needed > BTRFS_LEAF_DATA_SIZE(root)) { + double_split = 1; + } } } else { if (leaf_space_used(l, 0, mid + 1) + space_needed > @@ -1910,8 +1916,9 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root BUG_ON(path->slots[0] < 0); - if (!double_split) + if (!double_split) { return ret; + } right = btrfs_alloc_free_block(trans, root, root->leafsize, l->start, 0); @@ -2048,7 +2055,11 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, old_data = btrfs_item_end_nr(leaf, slot); BUG_ON(slot < 0); - BUG_ON(slot >= nritems); + if (slot >= nritems) { + btrfs_print_leaf(root, leaf); + printk("slot %d too large, nritems %d\n", slot, nritems); + BUG_ON(1); + } /* * item0..itemN ... dataN.offset..dataN.size .. data0.size @@ -2132,6 +2143,9 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, if (btrfs_leaf_free_space(root, leaf) < sizeof(struct btrfs_item) + data_size) { + btrfs_print_leaf(root, leaf); + printk("not enough freespace need %u have %d\n", + data_size, btrfs_leaf_free_space(root, leaf)); BUG(); } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 54c754dd9a1..18994c53106 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1137,7 +1137,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - u64 start, u64 end, u64 *hint_block); + u64 start, u64 end, u64 inline_end, u64 *hint_block); /* tree-defrag.c */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 844d8807e44..1af2b6534da 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -82,8 +82,9 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) static int insert_inline_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - u64 offset, ssize_t size, - struct page *page, size_t page_offset) + u64 offset, size_t size, + struct page **pages, size_t page_offset, + int num_pages) { struct btrfs_key key; struct btrfs_path *path; @@ -91,9 +92,12 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, char *kaddr; unsigned long ptr; struct btrfs_file_extent_item *ei; + struct page *page; u32 datasize; int err = 0; int ret; + int i; + ssize_t cur_size; path = btrfs_alloc_path(); if (!path) @@ -104,25 +108,97 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, key.objectid = inode->i_ino; key.offset = offset; btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); - BUG_ON(size >= PAGE_CACHE_SIZE); - datasize = btrfs_file_extent_calc_inline_size(size); + datasize = btrfs_file_extent_calc_inline_size(offset + size); - ret = btrfs_insert_empty_item(trans, root, path, &key, - datasize); - if (ret) { + ret = btrfs_search_slot(trans, root, &key, path, 0, 1); + if (ret < 0) { err = ret; goto fail; } - leaf = path->nodes[0]; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(leaf, ei, trans->transid); - btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); - ptr = btrfs_file_extent_inline_start(ei); - - kaddr = kmap_atomic(page, KM_USER1); - write_extent_buffer(leaf, kaddr + page_offset, ptr, size); - kunmap_atomic(kaddr, KM_USER1); + if (ret == 1) { + path->slots[0]--; + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + + if (btrfs_file_extent_type(leaf, ei) != + BTRFS_FILE_EXTENT_INLINE) { + goto insert; + } + btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); + ret = 0; + } + if (ret == 0) { + u32 found_size; + u64 found_start; + + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + + if (btrfs_file_extent_type(leaf, ei) != + BTRFS_FILE_EXTENT_INLINE) { + err = ret; + btrfs_print_leaf(root, leaf); + printk("found wasn't inline offset %Lu inode %lu\n", + offset, inode->i_ino); + goto fail; + } + found_start = key.offset; + found_size = btrfs_file_extent_inline_len(leaf, + btrfs_item_nr(leaf, path->slots[0])); + + if (found_size < offset + size) { + btrfs_release_path(root, path); + ret = btrfs_search_slot(trans, root, &key, path, + offset + size - found_size - + found_start, 1); + BUG_ON(ret != 0); + ret = btrfs_extend_item(trans, root, path, + offset + size - found_size - + found_start); + if (ret) { + err = ret; + goto fail; + } + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + } + } else { +insert: + btrfs_release_path(root, path); + ret = btrfs_insert_empty_item(trans, root, path, &key, + datasize); + if (ret) { + err = ret; + printk("got bad ret %d\n", ret); + goto fail; + } + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, ei, trans->transid); + btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); + } + ptr = btrfs_file_extent_inline_start(ei) + offset; + + cur_size = size; + i = 0; + while (size > 0) { + page = pages[i]; + kaddr = kmap_atomic(page, KM_USER0); + cur_size = min(PAGE_CACHE_SIZE - page_offset, size); + write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); + kunmap_atomic(kaddr, KM_USER0); + page_offset = 0; + ptr += cur_size; + size -= cur_size; + if (i >= num_pages) { + printk("i %d num_pages %d\n", i, num_pages); + } + i++; + } btrfs_mark_buffer_dirty(leaf); fail: btrfs_free_path(path); @@ -193,6 +269,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, err = btrfs_drop_extents(trans, root, inode, last_pos_in_file, last_pos_in_file + hole_size, + last_pos_in_file, &hint_byte); if (err) goto failed; @@ -210,11 +287,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, * either allocate an extent for the new bytes or setup the key * to show we are doing inline data in the extent */ - inline_size = end_pos - start_pos; - if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size || - inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root) || - inline_size >= PAGE_CACHE_SIZE) { + inline_size = end_pos; + if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || + inline_size > 16384 || + inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { u64 last_end; + for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; SetPageUptodate(p); @@ -225,22 +303,18 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, set_extent_delalloc(em_tree, start_pos, end_of_last_block, GFP_NOFS); } else { - struct page *p = pages[0]; + u64 aligned_end; /* step one, delete the existing extents in this range */ + aligned_end = (pos + write_bytes + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); err = btrfs_drop_extents(trans, root, inode, start_pos, - (pos + write_bytes + root->sectorsize -1) & - ~((u64)root->sectorsize - 1), &hint_byte); + aligned_end, end_pos, &hint_byte); if (err) goto failed; - err = insert_inline_extent(trans, root, inode, start_pos, - end_pos - start_pos, p, 0); + end_pos - start_pos, pages, 0, + num_pages); BUG_ON(err); - em->start = start_pos; - em->end = end_pos - 1; - em->block_start = EXTENT_MAP_INLINE; - em->block_end = EXTENT_MAP_INLINE; - add_extent_mapping(em_tree, em); } if (end_pos > isize) { i_size_write(inode, end_pos); @@ -285,7 +359,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) */ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - u64 start, u64 end, u64 *hint_byte) + u64 start, u64 end, u64 inline_end, u64 *hint_byte) { int ret; struct btrfs_key key; @@ -401,8 +475,8 @@ next_slot: BUG_ON(ret); } } - WARN_ON(found_inline); - bookend = 1; + if (!found_inline) + bookend = 1; } /* truncate existing extent */ if (start > key.offset) { @@ -425,8 +499,14 @@ next_slot: btrfs_set_file_extent_num_bytes(leaf, extent, new_num); btrfs_mark_buffer_dirty(leaf); - } else { - WARN_ON(1); + } else if (end > extent_end && + key.offset < inline_end && + inline_end < extent_end) { + u32 new_size; + new_size = btrfs_file_extent_calc_inline_size( + inline_end - key.offset); + btrfs_truncate_item(trans, root, path, + new_size); } } /* delete the entire extent */ diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d6b3a55ed8e..84f496c838c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -87,7 +87,7 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) BUG_ON(!trans); num_bytes = (end - start + blocksize) & ~(blocksize - 1); ret = btrfs_drop_extents(trans, root, inode, - start, start + num_bytes, &alloc_hint); + start, start + num_bytes, start, &alloc_hint); ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0, alloc_hint, (u64)-1, &ins, 1); @@ -776,7 +776,8 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); err = btrfs_drop_extents(trans, root, inode, - pos, pos + hole_size, &alloc_hint); + pos, pos + hole_size, pos, + &alloc_hint); err = btrfs_insert_file_extent(trans, root, inode->i_ino, pos, 0, 0, hole_size); @@ -1581,7 +1582,9 @@ again: } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { unsigned long ptr; char *map; - u32 size; + size_t size; + size_t extent_offset; + size_t copy_size; size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, path->slots[0])); @@ -1600,26 +1603,31 @@ again: goto not_found_em; } + extent_offset = (page->index << PAGE_CACHE_SHIFT) - + extent_start; + ptr = btrfs_file_extent_inline_start(item) + extent_offset; + map = kmap(page); + copy_size = min(PAGE_CACHE_SIZE - page_offset, + size - extent_offset); + em->block_start = EXTENT_MAP_INLINE; em->block_end = EXTENT_MAP_INLINE; - em->start = extent_start; - em->end = extent_end; + em->start = extent_start + extent_offset; + em->end = (em->start + copy_size -1) | + ((u64)root->sectorsize -1); if (!page) { goto insert; } - ptr = btrfs_file_extent_inline_start(item); - map = kmap(page); - read_extent_buffer(leaf, map + page_offset, ptr, size); + read_extent_buffer(leaf, map + page_offset, ptr, copy_size); /* - memset(map + page_offset + size, 0, - root->sectorsize - (page_offset + size)); + memset(map + page_offset + copy_size, 0, + PAGE_CACHE_SIZE - copy_size - page_offset); */ flush_dcache_page(page); kunmap(page); - set_extent_uptodate(em_tree, extent_start, - extent_end, GFP_NOFS); + set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS); goto insert; } else { printk("unkknown found_type %d\n", found_type); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 5c4370f3a5b..f94aa1f97a0 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -280,7 +280,6 @@ error_s: error_bdev: close_bdev_excl(bdev); error: -printk("get_sb failed\n"); return error; } /* end copy & paste */ @@ -295,7 +294,6 @@ static int btrfs_get_sb(struct file_system_type *fs_type, ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, btrfs_fill_super, mnt, subvol_name ? subvol_name : "default"); -printk("btrfs_get_sb returns %d\n", ret); return ret; } -- cgit v1.2.3-70-g09d2 From 19c00ddcc31ad4bdfb86b57085e06d6135b9b1d7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Oct 2007 16:19:22 -0400 Subject: Btrfs: Add back metadata checksumming Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 + fs/btrfs/disk-io.c | 215 +++++++++++++++++++++++++++++++++++-------------- fs/btrfs/extent-tree.c | 4 + fs/btrfs/extent_map.c | 83 +++++++++++++++---- fs/btrfs/extent_map.h | 18 ++++- fs/btrfs/inode.c | 4 - 6 files changed, 245 insertions(+), 81 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index d1c6f023a30..1af0a966f85 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -310,9 +310,11 @@ struct btrfs_fs_info { struct extent_buffer *sb_buffer; struct super_block *sb; struct inode *btree_inode; + spinlock_t hash_lock; struct mutex trans_mutex; struct mutex fs_mutex; struct list_head trans_list; + struct list_head hashers; struct list_head dead_roots; struct delayed_work trans_work; struct kobject super_kobj; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 16f0260fca6..1176e5420c5 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -60,7 +60,7 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, struct extent_buffer *eb; eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, - bytenr, blocksize, GFP_NOFS); + bytenr, blocksize, NULL, GFP_NOFS); return eb; } @@ -99,10 +99,102 @@ out: return em; } +u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) +{ + return crc32c(seed, data, len); +} + +void btrfs_csum_final(u32 crc, char *result) +{ + *(__le32 *)result = ~cpu_to_le32(crc); +} + +static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, + int verify) +{ + char result[BTRFS_CRC32_SIZE]; + unsigned long len; + unsigned long cur_len; + unsigned long offset = BTRFS_CSUM_SIZE; + char *map_token = NULL; + char *kaddr; + unsigned long map_start; + unsigned long map_len; + int err; + u32 crc = ~(u32)0; + + len = buf->len - offset; + while(len > 0) { + err = map_private_extent_buffer(buf, offset, 32, + &map_token, &kaddr, + &map_start, &map_len, KM_USER0); + if (err) { + printk("failed to map extent buffer! %lu\n", + offset); + return 1; + } + cur_len = min(len, map_len - (offset - map_start)); + crc = btrfs_csum_data(root, kaddr + offset - map_start, + crc, cur_len); + len -= cur_len; + offset += cur_len; + unmap_extent_buffer(buf, map_token, KM_USER0); + } + btrfs_csum_final(crc, result); + + if (verify) { + if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { + printk("btrfs: %s checksum verify failed on %llu\n", + root->fs_info->sb->s_id, + buf->start); + return 1; + } + } else { + write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); + } + return 0; +} + + +int csum_dirty_buffer(struct btrfs_root *root, struct page *page) +{ + struct extent_map_tree *tree; + u64 start = page->index << PAGE_CACHE_SHIFT; + u64 found_start; + int found_level; + unsigned long len; + struct extent_buffer *eb; + tree = &BTRFS_I(page->mapping->host)->extent_tree; + + if (page->private == EXTENT_PAGE_PRIVATE) + goto out; + if (!page->private) + goto out; + len = page->private >> 2; + if (len == 0) { + WARN_ON(1); + } + eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); + read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1); + found_start = btrfs_header_bytenr(eb); + if (found_start != start) { + printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", + start, found_start, len); + } + found_level = btrfs_header_level(eb); + csum_tree_block(root, eb, 0); + free_extent_buffer(eb); +out: + return 0; +} + static int btree_writepage(struct page *page, struct writeback_control *wbc) { struct extent_map_tree *tree; + struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; tree = &BTRFS_I(page->mapping->host)->extent_tree; + + csum_dirty_buffer(root, page); return extent_write_full_page(tree, page, btree_get_extent, wbc); } int btree_readpage(struct file *file, struct page *page) @@ -117,7 +209,6 @@ static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) struct extent_map_tree *tree; int ret; - BUG_ON(page->private != 1); tree = &BTRFS_I(page->mapping->host)->extent_tree; ret = try_release_extent_mapping(tree, page); if (ret == 1) { @@ -136,46 +227,6 @@ static void btree_invalidatepage(struct page *page, unsigned long offset) btree_releasepage(page, GFP_NOFS); } -int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len, - char *result) -{ - return 0; -#if 0 - u32 crc; - crc = crc32c(0, data, len); - memcpy(result, &crc, BTRFS_CRC32_SIZE); - return 0; -#endif -} - -#if 0 -static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, - int verify) -{ - return 0; - char result[BTRFS_CRC32_SIZE]; - int ret; - struct btrfs_node *node; - - ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE, - bh->b_size - BTRFS_CSUM_SIZE, result); - if (ret) - return ret; - if (verify) { - if (memcmp(bh->b_data, result, BTRFS_CRC32_SIZE)) { - printk("btrfs: %s checksum verify failed on %llu\n", - root->fs_info->sb->s_id, - (unsigned long long)bh_blocknr(bh)); - return 1; - } - } else { - node = btrfs_buffer_node(bh); - memcpy(node->header.csum, result, BTRFS_CRC32_SIZE); - } - return 0; -} -#endif - #if 0 static int btree_writepage(struct page *page, struct writeback_control *wbc) { @@ -215,7 +266,7 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) if (!buf) return 0; read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, - buf, 0); + buf, 0, 0); free_extent_buffer(buf); return ret; } @@ -225,12 +276,29 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, { struct extent_buffer *buf = NULL; struct inode *btree_inode = root->fs_info->btree_inode; + struct extent_map_tree *extent_tree; + int ret; + + extent_tree = &BTRFS_I(btree_inode)->extent_tree; buf = btrfs_find_create_tree_block(root, bytenr, blocksize); if (!buf) return NULL; read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, - buf, 1); + buf, 0, 1); + if (buf->flags & EXTENT_CSUM) { + return buf; + } + if (test_range_bit(extent_tree, buf->start, buf->start + buf->len - 1, + EXTENT_CSUM, 1)) { + buf->flags |= EXTENT_CSUM; + return buf; + } + ret = csum_tree_block(root, buf, 1); + set_extent_bits(extent_tree, buf->start, + buf->start + buf->len - 1, + EXTENT_CSUM, GFP_NOFS); + buf->flags |= EXTENT_CSUM; return buf; } @@ -251,13 +319,6 @@ int wait_on_tree_block_writeback(struct btrfs_root *root, return 0; } -int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf) -{ - struct inode *btree_inode = root->fs_info->btree_inode; - set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); - return 0; -} - static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, struct btrfs_root *root, struct btrfs_fs_info *fs_info, @@ -416,7 +477,24 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, return root; } - +#if 0 +static int add_hasher(struct btrfs_fs_info *info, char *type) { + struct btrfs_hasher *hasher; + + hasher = kmalloc(sizeof(*hasher), GFP_NOFS); + if (!hasher) + return -ENOMEM; + hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); + if (!hasher->hash_tfm) { + kfree(hasher); + return -EINVAL; + } + spin_lock(&info->hash_lock); + list_add(&hasher->list, &info->hashers); + spin_unlock(&info->hash_lock); + return 0; +} +#endif struct btrfs_root *open_ctree(struct super_block *sb) { u32 sectorsize; @@ -440,6 +518,9 @@ struct btrfs_root *open_ctree(struct super_block *sb) INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); INIT_LIST_HEAD(&fs_info->trans_list); INIT_LIST_HEAD(&fs_info->dead_roots); + INIT_LIST_HEAD(&fs_info->hashers); + spin_lock_init(&fs_info->hash_lock); + memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); init_completion(&fs_info->kobj_unregister); sb_set_blocksize(sb, 4096); @@ -479,6 +560,14 @@ struct btrfs_root *open_ctree(struct super_block *sb) mutex_init(&fs_info->trans_mutex); mutex_init(&fs_info->fs_mutex); +#if 0 + ret = add_hasher(fs_info, "crc32c"); + if (ret) { + printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); + err = -ENOMEM; + goto fail_iput; + } +#endif __setup_root(512, 512, 512, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); @@ -509,25 +598,21 @@ struct btrfs_root *open_ctree(struct super_block *sb) i_size_write(fs_info->btree_inode, btrfs_super_total_bytes(disk_super)); - if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, sizeof(disk_super->magic))) { printk("btrfs: valid FS not found on %s\n", sb->s_id); goto fail_sb_buffer; } + blocksize = btrfs_level_size(tree_root, btrfs_super_root_level(disk_super)); + tree_root->node = read_tree_block(tree_root, btrfs_super_root(disk_super), blocksize); if (!tree_root->node) goto fail_sb_buffer; -#if 0 - btrfs_print_leaf(tree_root, tree_root->node); - err = -EIO; - goto fail_tree_root; -#endif mutex_lock(&fs_info->fs_mutex); ret = find_and_setup_root(tree_root, fs_info, @@ -634,9 +719,19 @@ int close_ctree(struct btrfs_root *root) btrfs_free_block_groups(root->fs_info); del_fs_roots(fs_info); - extent_map_tree_cleanup(&BTRFS_I(fs_info->btree_inode)->extent_tree); + extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree); truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); iput(fs_info->btree_inode); +#if 0 + while(!list_empty(&fs_info->hashers)) { + struct btrfs_hasher *hasher; + hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, + hashers); + list_del(&hasher->hashers); + crypto_free_hash(&fs_info->hash_tfm); + kfree(hasher); + } +#endif kfree(fs_info->extent_root); kfree(fs_info->tree_root); return 0; @@ -733,5 +828,5 @@ int btrfs_read_buffer(struct extent_buffer *buf) struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, - buf, 1); + buf, 0, 1); } diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index e4e68ea9648..6b2f2b414fa 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1178,6 +1178,10 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, btrfs_set_buffer_uptodate(buf); set_extent_dirty(&trans->transaction->dirty_pages, buf->start, buf->start + buf->len - 1, GFP_NOFS); + set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->extent_tree, + buf->start, buf->start + buf->len - 1, + EXTENT_CSUM, GFP_NOFS); + buf->flags |= EXTENT_CSUM; btrfs_set_buffer_defrag(buf); trans->blocks_used++; return buf; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index f8aaba8a30a..2a8bc4bd43a 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -81,7 +81,7 @@ void extent_map_tree_init(struct extent_map_tree *tree, } EXPORT_SYMBOL(extent_map_tree_init); -void extent_map_tree_cleanup(struct extent_map_tree *tree) +void extent_map_tree_empty_lru(struct extent_map_tree *tree) { struct extent_buffer *eb; while(!list_empty(&tree->buffer_lru)) { @@ -91,7 +91,7 @@ void extent_map_tree_cleanup(struct extent_map_tree *tree) free_extent_buffer(eb); } } -EXPORT_SYMBOL(extent_map_tree_cleanup); +EXPORT_SYMBOL(extent_map_tree_empty_lru); struct extent_map *alloc_extent_map(gfp_t mask) { @@ -1464,7 +1464,7 @@ void set_page_extent_mapped(struct page *page) if (!PagePrivate(page)) { SetPagePrivate(page); WARN_ON(!page->mapping->a_ops->invalidatepage); - set_page_private(page, 1); + set_page_private(page, EXTENT_PAGE_PRIVATE); page_cache_get(page); } } @@ -1979,8 +1979,9 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_map_tree *tree, spin_lock(&tree->lru_lock); eb = find_lru(tree, start, len); - if (eb) + if (eb) { goto lru_add; + } spin_unlock(&tree->lru_lock); if (eb) { @@ -2007,6 +2008,7 @@ static void __free_extent_buffer(struct extent_buffer *eb) struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, u64 start, unsigned long len, + struct page *page0, gfp_t mask) { unsigned long num_pages = num_extent_pages(start, len); @@ -2024,7 +2026,18 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, if (eb->flags & EXTENT_BUFFER_FILLED) return eb; - for (i = 0; i < num_pages; i++, index++) { + if (page0) { + eb->first_page = page0; + i = 1; + index++; + page_cache_get(page0); + set_page_extent_mapped(page0); + set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE | + len << 2); + } else { + i = 0; + } + for (; i < num_pages; i++, index++) { p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); if (!p) { WARN_ON(1); @@ -2036,8 +2049,13 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, goto fail; } set_page_extent_mapped(p); - if (i == 0) + if (i == 0) { eb->first_page = p; + set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE | + len << 2); + } else { + set_page_private(p, EXTENT_PAGE_PRIVATE); + } if (!PageUptodate(p)) uptodate = 0; unlock_page(p); @@ -2057,8 +2075,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, gfp_t mask) { unsigned long num_pages = num_extent_pages(start, len); - unsigned long i; - unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; struct page *p; struct address_space *mapping = tree->mapping; @@ -2082,8 +2099,15 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, goto fail; } set_page_extent_mapped(p); - if (i == 0) + + if (i == 0) { eb->first_page = p; + set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE | + len << 2); + } else { + set_page_private(p, EXTENT_PAGE_PRIVATE); + } + if (!PageUptodate(p)) uptodate = 0; unlock_page(p); @@ -2174,7 +2198,21 @@ int set_extent_buffer_dirty(struct extent_map_tree *tree, num_pages = num_extent_pages(eb->start, eb->len); for (i = 0; i < num_pages; i++) { + struct page *page = extent_buffer_page(eb, i); + /* writepage may need to do something special for the + * first page, we have to make sure page->private is + * properly set. releasepage may drop page->private + * on us if the page isn't already dirty. + */ + if (i == 0) { + lock_page(page); + set_page_private(page, + EXTENT_PAGE_PRIVATE_FIRST_PAGE | + eb->len << 2); + } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); + if (i == 0) + unlock_page(page); } return set_extent_dirty(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); @@ -2217,9 +2255,12 @@ int extent_buffer_uptodate(struct extent_map_tree *tree, EXPORT_SYMBOL(extent_buffer_uptodate); int read_extent_buffer_pages(struct extent_map_tree *tree, - struct extent_buffer *eb, int wait) + struct extent_buffer *eb, + u64 start, + int wait) { unsigned long i; + unsigned long start_i; struct page *page; int err; int ret = 0; @@ -2232,9 +2273,16 @@ int read_extent_buffer_pages(struct extent_map_tree *tree, EXTENT_UPTODATE, 1)) { return 0; } + if (start) { + WARN_ON(start < eb->start); + start_i = (start >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT); + } else { + start_i = 0; + } num_pages = num_extent_pages(eb->start, eb->len); - for (i = 0; i < num_pages; i++) { + for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); if (PageUptodate(page)) { continue; @@ -2260,7 +2308,7 @@ int read_extent_buffer_pages(struct extent_map_tree *tree, return ret; } - for (i = 0; i < num_pages; i++) { + for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); wait_on_page_locked(page); if (!PageUptodate(page)) { @@ -2314,7 +2362,7 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, } EXPORT_SYMBOL(read_extent_buffer); -static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start, +int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, unsigned long min_len, char **token, char **map, unsigned long *map_start, unsigned long *map_len, int km) @@ -2337,6 +2385,10 @@ static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start, offset = 0; *map_start = (i << PAGE_CACHE_SHIFT) - start_offset; } + if (start + min_len >= eb->len) { +printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); + WARN_ON(1); + } p = extent_buffer_page(eb, i); WARN_ON(!PageUptodate(p)); @@ -2346,6 +2398,7 @@ static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start, *map_len = PAGE_CACHE_SIZE - offset; return 0; } +EXPORT_SYMBOL(map_private_extent_buffer); int map_extent_buffer(struct extent_buffer *eb, unsigned long start, unsigned long min_len, @@ -2360,8 +2413,8 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start, eb->map_token = NULL; save = 1; } - err = __map_extent_buffer(eb, start, min_len, token, map, - map_start, map_len, km); + err = map_private_extent_buffer(eb, start, min_len, token, map, + map_start, map_len, km); if (!err && save) { eb->map_token = *token; eb->kaddr = *map; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index f1dc28d260e..39d78d31b6a 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -17,8 +17,17 @@ #define EXTENT_DEFRAG (1 << 6) #define EXTENT_DEFRAG_DONE (1 << 7) #define EXTENT_BUFFER_FILLED (1 << 8) +#define EXTENT_CSUM (1 << 9) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) +/* + * page->private values. Every page that is controlled by the extent + * map has page->private set to one. + */ + +#define EXTENT_PAGE_PRIVATE 1 +#define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3 + struct extent_map_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); @@ -89,7 +98,7 @@ typedef struct extent_map *(get_extent_t)(struct inode *inode, void extent_map_tree_init(struct extent_map_tree *tree, struct address_space *mapping, gfp_t mask); -void extent_map_tree_cleanup(struct extent_map_tree *tree); +void extent_map_tree_empty_lru(struct extent_map_tree *tree); struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, u64 start, u64 end); int add_extent_mapping(struct extent_map_tree *tree, @@ -143,13 +152,14 @@ void set_page_extent_mapped(struct page *page); struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, u64 start, unsigned long len, + struct page *page0, gfp_t mask); struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, u64 start, unsigned long len, gfp_t mask); void free_extent_buffer(struct extent_buffer *eb); int read_extent_buffer_pages(struct extent_map_tree *tree, - struct extent_buffer *eb, int wait); + struct extent_buffer *eb, u64 start, int wait); static inline void extent_buffer_get(struct extent_buffer *eb) { @@ -187,5 +197,9 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, unsigned long min_len, char **token, char **map, unsigned long *map_start, unsigned long *map_len, int km); +int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset, + unsigned long min_len, char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km); void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); #endif diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 84f496c838c..80f84b48a70 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1713,10 +1713,6 @@ static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) struct extent_map_tree *tree; int ret; - if (page->private != 1) { - WARN_ON(1); - return try_to_free_buffers(page); - } tree = &BTRFS_I(page->mapping->host)->extent_tree; ret = try_release_extent_mapping(tree, page); if (ret == 1) { -- cgit v1.2.3-70-g09d2 From ff79f8190b6e955ff7a71faf804a3017d526e657 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Oct 2007 16:22:25 -0400 Subject: Btrfs: Add back file data checksumming Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 1 + fs/btrfs/disk-io.h | 2 ++ fs/btrfs/extent_map.c | 3 +++ fs/btrfs/file-item.c | 17 +++++++++-------- fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 21 ++++++++++----------- 6 files changed, 26 insertions(+), 19 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3878772fcff..5accced2858 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -594,6 +594,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) tree_root->nodesize = nodesize; tree_root->leafsize = leafsize; tree_root->sectorsize = sectorsize; + sb_set_blocksize(sb, sectorsize); i_size_write(fs_info->btree_inode, btrfs_super_total_bytes(disk_super)); diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 190b07bc642..33d2ccfd74d 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -58,4 +58,6 @@ int btrfs_buffer_defrag_done(struct extent_buffer *buf); int btrfs_clear_buffer_defrag(struct extent_buffer *buf); int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf); int btrfs_read_buffer(struct extent_buffer *buf); +u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len); +void btrfs_csum_final(u32 crc, char *result); #endif diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 2a8bc4bd43a..e87e476dca9 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -2031,6 +2031,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, i = 1; index++; page_cache_get(page0); + mark_page_accessed(page0); set_page_extent_mapped(page0); set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); @@ -2049,6 +2050,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, goto fail; } set_page_extent_mapped(p); + mark_page_accessed(p); if (i == 0) { eb->first_page = p; set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE | @@ -2099,6 +2101,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, goto fail; } set_page_extent_mapped(p); + mark_page_accessed(p); if (i == 0) { eb->first_page = p; diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 9a65e97a4e2..10a4c740860 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -136,27 +136,27 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, u64 objectid, u64 offset, char *data, size_t len) { - return 0; -#if 0 int ret; struct btrfs_key file_key; struct btrfs_key found_key; struct btrfs_path *path; struct btrfs_csum_item *item; - struct extent_buffer *leaf; + struct extent_buffer *leaf = NULL; u64 csum_offset; + u32 csum_result = ~(u32)0; path = btrfs_alloc_path(); BUG_ON(!path); file_key.objectid = objectid; file_key.offset = offset; - file_key.flags = 0; btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); item = btrfs_lookup_csum(trans, root, path, objectid, offset, 1); - if (!IS_ERR(item)) + if (!IS_ERR(item)) { + leaf = path->nodes[0]; goto found; + } ret = PTR_ERR(item); if (ret == -EFBIG) { u32 item_size; @@ -226,14 +226,15 @@ csum: item = (struct btrfs_csum_item *)((unsigned char *)item + csum_offset * BTRFS_CRC32_SIZE); found: - /* FIXME!!!!!!!!!!!! */ - ret = btrfs_csum_data(root, data, len, &item->csum); + csum_result = btrfs_csum_data(root, data, csum_result, len); + btrfs_csum_final(csum_result, (char *)&csum_result); + write_extent_buffer(leaf, &csum_result, (unsigned long)item, + BTRFS_CRC32_SIZE); btrfs_mark_buffer_dirty(path->nodes[0]); fail: btrfs_release_path(root, path); btrfs_free_path(path); return ret; -#endif } int btrfs_csum_truncate(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index fe28404ae7f..1af2b6534da 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -289,6 +289,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, */ inline_size = end_pos; if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || + inline_size > 16384 || inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { u64 last_end; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 80f84b48a70..ac69f6926af 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -135,7 +135,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct btrfs_csum_item *item; struct btrfs_path *path = NULL; - u64 private; + u32 csum; mutex_lock(&root->fs_info->fs_mutex); path = btrfs_alloc_path(); @@ -145,11 +145,12 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) /* a csum that isn't present is a preallocated region. */ if (ret == -ENOENT || ret == -EFBIG) ret = 0; - private = 0; + csum = 0; goto out; } - memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE); - set_state_private(em_tree, start, private); + read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, + BTRFS_CRC32_SIZE); + set_state_private(em_tree, start, csum); out: if (path) btrfs_free_path(path); @@ -165,21 +166,19 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) char *kaddr; u64 private; int ret; + struct btrfs_root *root = BTRFS_I(inode)->root; + u32 csum = ~(u32)0; ret = get_state_private(em_tree, start, &private); kaddr = kmap_atomic(page, KM_IRQ0); if (ret) { goto zeroit; } - /* - struct btrfs_root *root = BTRFS_I(inode)->root; - char csum[BTRFS_CRC32_SIZE]; - ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum); - BUG_ON(ret); - if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) { + csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1); + btrfs_csum_final(csum, (char *)&csum); + if (csum != private) { goto zeroit; } - */ kunmap_atomic(kaddr, KM_IRQ0); return 0; -- cgit v1.2.3-70-g09d2 From ae2f5411c4ce7180cca8418853db50c8e52d40db Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 19 Oct 2007 09:22:59 -0400 Subject: btrfs: 32-bit type problems An assorted set of casts to get rid of the warnings on 32-bit archs. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 2 +- fs/btrfs/extent-tree.c | 8 ++++---- fs/btrfs/extent_map.c | 6 +++--- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 2 +- 5 files changed, 10 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 54967099a8e..17a322663f3 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1389,7 +1389,7 @@ int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf) ret = BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems); if (ret < 0) { printk("leaf free space ret %d, leaf data size %lu, used %d nritems %d\n", - ret, BTRFS_LEAF_DATA_SIZE(root), + ret, (unsigned long) BTRFS_LEAF_DATA_SIZE(root), leaf_space_used(leaf, 0, nritems), nritems); } return ret; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 6b2f2b414fa..7de7707d939 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -149,7 +149,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct if (ret) return NULL; - block_group = (struct btrfs_block_group_cache *)ptr; + block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; if (block_group->key.objectid <= bytenr && bytenr <= @@ -279,7 +279,7 @@ again: if (ret) break; - cache = (struct btrfs_block_group_cache *)ptr; + cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; last = cache->key.objectid + cache->key.offset; used = btrfs_block_group_used(&cache->item); @@ -537,7 +537,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, if (ret) break; - cache = (struct btrfs_block_group_cache *)ptr; + cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; err = write_one_cache_group(trans, root, path, cache); /* @@ -1541,7 +1541,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) found_key.objectid + found_key.offset - 1, bit | EXTENT_LOCKED, GFP_NOFS); set_state_private(block_group_cache, found_key.objectid, - (u64)cache); + (unsigned long)cache); if (key.objectid >= btrfs_super_total_bytes(&info->super_copy)) diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index e87e476dca9..89d0ebd0ea8 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -2654,8 +2654,8 @@ void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - src_off_in_page)); - cur = min(cur, (unsigned long)(PAGE_CACHE_SIZE - - dst_off_in_page)); + cur = min_t(unsigned long, cur, + (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page)); copy_pages(extent_buffer_page(dst, dst_i), extent_buffer_page(dst, src_i), @@ -2707,7 +2707,7 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, if (dst_i == 0) dst_off_in_page += start_offset; - cur = min(len, src_off_in_page + 1); + cur = min_t(unsigned long, len, src_off_in_page + 1); cur = min(cur, dst_off_in_page + 1); move_pages(extent_buffer_page(dst, dst_i), extent_buffer_page(dst, src_i), diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 4aacf99bd97..96df1b10cb6 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -188,7 +188,7 @@ insert: while (size > 0) { page = pages[i]; kaddr = kmap_atomic(page, KM_USER0); - cur_size = min(PAGE_CACHE_SIZE - page_offset, size); + cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size); write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); kunmap_atomic(kaddr, KM_USER0); page_offset = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index ac69f6926af..0e1ca423e22 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1606,7 +1606,7 @@ again: extent_start; ptr = btrfs_file_extent_inline_start(item) + extent_offset; map = kmap(page); - copy_size = min(PAGE_CACHE_SIZE - page_offset, + copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset, size - extent_offset); em->block_start = EXTENT_MAP_INLINE; -- cgit v1.2.3-70-g09d2 From bbf0d0062da2b3e1e239e381f8ac20971ef1b2ed Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 19 Oct 2007 09:23:07 -0400 Subject: Btrfs: KM_IRQ0 usage in end_io handling endio handling is typically called with interrupts disabled, but can also be called with it enabled. So save interrupts before using KM_IRQ0 to be completely safe. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0e1ca423e22..02d802cdd59 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -168,8 +168,10 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) int ret; struct btrfs_root *root = BTRFS_I(inode)->root; u32 csum = ~(u32)0; + unsigned long flags; ret = get_state_private(em_tree, start, &private); + local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); if (ret) { goto zeroit; @@ -180,6 +182,7 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) goto zeroit; } kunmap_atomic(kaddr, KM_IRQ0); + local_irq_restore(flags); return 0; zeroit: @@ -188,6 +191,7 @@ zeroit: memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); + local_irq_restore(flags); return 0; } -- cgit v1.2.3-70-g09d2 From f578d4bd7e141dd03ca7e8695c1cc118c326e69e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 25 Oct 2007 15:42:56 -0400 Subject: Btrfs: Optimize csum insertion to create larger items when possible This reduces the number of calls to btrfs_extend_item and greatly lowers the cpu usage while writing large files. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/file-item.c | 37 ++++++++++++++++++++++++++++++++++++- fs/btrfs/inode.c | 2 +- 3 files changed, 38 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 1d2b5415007..70e143b1357 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -991,6 +991,7 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, u64 bytenr, int mod); int btrfs_csum_file_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct inode *inode, u64 objectid, u64 offset, char *data, size_t len); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 10a4c740860..482a2b61532 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -133,17 +133,22 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, int btrfs_csum_file_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct inode *inode, u64 objectid, u64 offset, char *data, size_t len) { int ret; struct btrfs_key file_key; struct btrfs_key found_key; + u64 next_offset = (u64)-1; + int found_next = 0; struct btrfs_path *path; struct btrfs_csum_item *item; struct extent_buffer *leaf = NULL; u64 csum_offset; u32 csum_result = ~(u32)0; + u32 nritems; + u32 ins_size; path = btrfs_alloc_path(); BUG_ON(!path); @@ -168,7 +173,27 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, goto insert; } } else { + int slot = path->slots[0] + 1; /* we didn't find a csum item, insert one */ + nritems = btrfs_header_nritems(path->nodes[0]); + if (path->slots[0] >= nritems - 1) { + ret = btrfs_next_leaf(root, path); + if (ret == 1) { + found_next = 1; + } else if (ret == 0) { + slot = 0; + } else { + goto insert; + } + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot); + if (found_key.objectid != objectid || + found_key.type != BTRFS_CSUM_ITEM_KEY) { + found_next = 1; + goto insert; + } + next_offset = found_key.offset; + found_next = 1; goto insert; } @@ -211,8 +236,18 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans, insert: btrfs_release_path(root, path); csum_offset = 0; + if (found_next) { + u64 tmp = min((u64)i_size_read(inode), next_offset); + tmp -= offset + root->sectorsize - 1; + tmp >>= root->fs_info->sb->s_blocksize_bits; + tmp = max((u64)1, tmp); + tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root)); + ins_size = BTRFS_CRC32_SIZE * tmp; + } else { + ins_size = BTRFS_CRC32_SIZE; + } ret = btrfs_insert_empty_item(trans, root, path, &file_key, - BTRFS_CRC32_SIZE); + ins_size); if (ret < 0) goto fail; if (ret != 0) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 02d802cdd59..7efc0470abf 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -118,7 +118,7 @@ int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); kaddr = kmap(page); - btrfs_csum_file_block(trans, root, inode->i_ino, + btrfs_csum_file_block(trans, root, inode, inode->i_ino, start, kaddr + offset, end - start + 1); kunmap(page); ret = btrfs_end_transaction(trans, root); -- cgit v1.2.3-70-g09d2 From 65555a06b4d1ae116ce223dc4b82d6068b36df96 Mon Sep 17 00:00:00 2001 From: Yan Date: Thu, 25 Oct 2007 15:42:57 -0400 Subject: Btrfs: Off by one fixes in extent_map.c Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 10 +++++----- fs/btrfs/inode.c | 1 + 2 files changed, 6 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index a61379230dd..c44989a1e52 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -2045,7 +2045,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, struct extent_buffer *eb; struct page *p; struct address_space *mapping = tree->mapping; - int uptodate = 0; + int uptodate = 1; eb = __alloc_extent_buffer(tree, start, len, mask); if (!eb || IS_ERR(eb)) @@ -2197,7 +2197,7 @@ int clear_extent_buffer_dirty(struct extent_map_tree *tree, */ if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || ((i == num_pages - 1) && - ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) { + ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { start = page->index << PAGE_CACHE_SHIFT; end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, @@ -2265,7 +2265,7 @@ int set_extent_buffer_uptodate(struct extent_map_tree *tree, page = extent_buffer_page(eb, i); if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || ((i == num_pages - 1) && - ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) { + ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { check_page_uptodate(tree, page); continue; } @@ -2401,7 +2401,7 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, struct page *p; size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - unsigned long end_i = (start_offset + start + min_len) >> + unsigned long end_i = (start_offset + start + min_len - 1) >> PAGE_CACHE_SHIFT; if (i != end_i) @@ -2414,7 +2414,7 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, offset = 0; *map_start = (i << PAGE_CACHE_SHIFT) - start_offset; } - if (start + min_len >= eb->len) { + if (start + min_len > eb->len) { printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); WARN_ON(1); } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7efc0470abf..c0743307232 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -616,6 +616,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, break; } btrfs_set_key_type(&key, found_type); + btrfs_release_path(root, path); continue; } if (found_key.offset >= inode->i_size) -- cgit v1.2.3-70-g09d2 From 0d9f7f3e27a92b07212090a1772fdbb3a92fef81 Mon Sep 17 00:00:00 2001 From: Yan Date: Thu, 25 Oct 2007 15:48:28 -0400 Subject: btrfs_inode_by_name return random value. When inode is found, the return value is from the uninitialized variable 'ret'. -- Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c0743307232..b74a831fab1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -847,15 +847,16 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, struct btrfs_dir_item *di; struct btrfs_path *path; struct btrfs_root *root = BTRFS_I(dir)->root; - int ret; + int ret = 0; path = btrfs_alloc_path(); BUG_ON(!path); di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, namelen, 0); + if (IS_ERR(di)) + ret = PTR_ERR(di); if (!di || IS_ERR(di)) { location->objectid = 0; - ret = 0; goto out; } btrfs_dir_item_key_to_cpu(path->nodes[0], di, location); -- cgit v1.2.3-70-g09d2 From 134d45120164141c0847daf5222aaec77e04724a Mon Sep 17 00:00:00 2001 From: Yan Date: Thu, 25 Oct 2007 15:49:25 -0400 Subject: Fix ENOTEMPTY check in btrfs_rmdir The ENOTEMPTY check in btrfs_rmdir isn't reliable. It's possible that the backward search finds . or .. at first, then some other directory entry. In that case, btrfs_rmdir delete . or .. improperly. The patch also fixes a fs_mutex unlock issue in btrfs_rmdir. -- Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b74a831fab1..7e4b2c1b4a5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -448,6 +448,9 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) char *goodnames = ".."; unsigned long nr; + if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) + return -ENOTEMPTY; + path = btrfs_alloc_path(); BUG_ON(!path); mutex_lock(&root->fs_info->fs_mutex); @@ -501,9 +504,9 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) out: btrfs_release_path(root, path); btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); if (ret && !err) err = ret; -- cgit v1.2.3-70-g09d2 From 44ec0b7179b9bce7df1080370f51314a24a00216 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 29 Oct 2007 10:55:05 -0400 Subject: Btrfs: Compile fixes for 2.6.24-rc1 Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7e4b2c1b4a5..c88f13a8855 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2176,8 +2176,12 @@ void btrfs_destroy_inode(struct inode *inode) kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +static void init_once(struct kmem_cache * cachep, void *foo) +#else static void init_once(void * foo, struct kmem_cache * cachep, unsigned long flags) +#endif { struct btrfs_inode *ei = (struct btrfs_inode *) foo; @@ -2200,8 +2204,13 @@ void btrfs_destroy_cachep(void) struct kmem_cache *btrfs_cache_create(const char *name, size_t size, unsigned long extra_flags, +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) + void (*ctor)(struct kmem_cache *, void *) +#else void (*ctor)(void *, struct kmem_cache *, - unsigned long)) + unsigned long) +#endif + ) { return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | extra_flags), ctor -- cgit v1.2.3-70-g09d2 From 689f9346612d96ab93890160cf2fe0df5217ab73 Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 29 Oct 2007 11:41:07 -0400 Subject: Fix inline extent handling in btrfs_get_extent 1. Reorder kmap and the test for 'page != NULL' 2. Zero-fill rest area of a block when inline extent isn't big enough. 3. Do not insert extent_map into the map tree when page == NULL. (If insert the extent_map into the map tree, subsequent read requests will find it in the map tree directly and the corresponding inline extent data aren't copied into page by the the get_extent function. extent_read_full_page can't handle that case) Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 34 ++++++++++++++++++---------------- 1 file changed, 18 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c88f13a8855..0ba1cdbdf11 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1596,8 +1596,7 @@ again: size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, path->slots[0])); - - extent_end = (extent_start + size) | + extent_end = (extent_start + size - 1) | ((u64)root->sectorsize - 1); if (start < extent_start || start >= extent_end) { em->start = start; @@ -1610,29 +1609,32 @@ again: } goto not_found_em; } + em->block_start = EXTENT_MAP_INLINE; + em->block_end = EXTENT_MAP_INLINE; + + if (!page) { + em->start = extent_start; + em->end = extent_start + size - 1; + goto out; + } extent_offset = (page->index << PAGE_CACHE_SHIFT) - - extent_start; - ptr = btrfs_file_extent_inline_start(item) + extent_offset; - map = kmap(page); + extent_start + page_offset; copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset, size - extent_offset); - - em->block_start = EXTENT_MAP_INLINE; - em->block_end = EXTENT_MAP_INLINE; em->start = extent_start + extent_offset; em->end = (em->start + copy_size -1) | ((u64)root->sectorsize -1); - - if (!page) { - goto insert; + map = kmap(page); + ptr = btrfs_file_extent_inline_start(item) + extent_offset; + read_extent_buffer(leaf, map + page_offset, ptr, copy_size); + + if (em->start + copy_size <= em->end) { + size = min_t(u64, em->end + 1 - em->start, + PAGE_CACHE_SIZE - page_offset) - copy_size; + memset(map + page_offset + copy_size, 0, size); } - read_extent_buffer(leaf, map + page_offset, ptr, copy_size); - /* - memset(map + page_offset + copy_size, 0, - PAGE_CACHE_SIZE - copy_size - page_offset); - */ flush_dcache_page(page); kunmap(page); set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS); -- cgit v1.2.3-70-g09d2 From 35ebb934bd7fcc7ca991b155b7980c3c4ff9f1a5 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 30 Oct 2007 16:56:53 -0400 Subject: Btrfs: Fix PAGE_CACHE_SHIFT shifts on 32 bit machines Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 2 +- fs/btrfs/extent_map.c | 27 +++++++++++++++------------ fs/btrfs/file.c | 3 ++- fs/btrfs/inode.c | 14 +++++++------- fs/btrfs/transaction.c | 2 +- 5 files changed, 26 insertions(+), 22 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 5a7e60d7f44..8935eec31ee 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -159,7 +159,7 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { struct extent_map_tree *tree; - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 found_start; int found_level; unsigned long len; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index e4ee980cc13..238cb1d81d5 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1229,7 +1229,7 @@ EXPORT_SYMBOL(test_range_bit); static int check_page_uptodate(struct extent_map_tree *tree, struct page *page) { - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1)) SetPageUptodate(page); @@ -1243,7 +1243,7 @@ static int check_page_uptodate(struct extent_map_tree *tree, static int check_page_locked(struct extent_map_tree *tree, struct page *page) { - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0)) unlock_page(page); @@ -1257,7 +1257,7 @@ static int check_page_locked(struct extent_map_tree *tree, static int check_page_writeback(struct extent_map_tree *tree, struct page *page) { - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0)) end_page_writeback(page); @@ -1296,7 +1296,8 @@ static int end_bio_extent_writepage(struct bio *bio, do { struct page *page = bvec->bv_page; - start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; end = start + bvec->bv_len - 1; if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) @@ -1361,7 +1362,8 @@ static int end_bio_extent_readpage(struct bio *bio, do { struct page *page = bvec->bv_page; - start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; end = start + bvec->bv_len - 1; if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) @@ -1427,7 +1429,8 @@ static int end_bio_extent_preparewrite(struct bio *bio, do { struct page *page = bvec->bv_page; - start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; end = start + bvec->bv_len - 1; if (--bvec >= bio->bi_io_vec) @@ -1503,7 +1506,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, get_extent_t *get_extent) { struct inode *inode = page->mapping->host; - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = start + PAGE_CACHE_SIZE - 1; u64 end; u64 cur = start; @@ -1608,7 +1611,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, struct writeback_control *wbc) { struct inode *inode = page->mapping->host; - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = start + PAGE_CACHE_SIZE - 1; u64 end; u64 cur = start; @@ -1750,7 +1753,7 @@ EXPORT_SYMBOL(extent_write_full_page); int extent_invalidatepage(struct extent_map_tree *tree, struct page *page, unsigned long offset) { - u64 start = (page->index << PAGE_CACHE_SHIFT); + u64 start = ((u64)page->index << PAGE_CACHE_SHIFT); u64 end = start + PAGE_CACHE_SIZE - 1; size_t blocksize = page->mapping->host->i_sb->s_blocksize; @@ -1792,7 +1795,7 @@ int extent_prepare_write(struct extent_map_tree *tree, struct inode *inode, struct page *page, unsigned from, unsigned to, get_extent_t *get_extent) { - u64 page_start = page->index << PAGE_CACHE_SHIFT; + u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; u64 block_start; u64 orig_block_start; @@ -1890,7 +1893,7 @@ EXPORT_SYMBOL(extent_prepare_write); int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page) { struct extent_map *em; - u64 start = page->index << PAGE_CACHE_SHIFT; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 end = start + PAGE_CACHE_SIZE - 1; u64 orig_start = start; int ret = 1; @@ -2202,7 +2205,7 @@ int clear_extent_buffer_dirty(struct extent_map_tree *tree, if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || ((i == num_pages - 1) && ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { - start = page->index << PAGE_CACHE_SHIFT; + start = (u64)page->index << PAGE_CACHE_SHIFT; end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, EXTENT_DIRTY, 0)) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 843e920388e..abdd9caad94 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -301,7 +301,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, SetPageUptodate(p); set_page_dirty(p); } - last_end = pages[num_pages -1]->index << PAGE_CACHE_SHIFT; + last_end = (u64)(pages[num_pages -1]->index) << + PAGE_CACHE_SHIFT; last_end += PAGE_CACHE_SIZE - 1; set_extent_delalloc(em_tree, start_pos, end_of_last_block, GFP_NOFS); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0ba1cdbdf11..5d10b64e42d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -111,7 +111,7 @@ int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) struct btrfs_trans_handle *trans; char *kaddr; int ret; - u64 page_start = page->index << PAGE_CACHE_SHIFT; + u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; size_t offset = start - page_start; mutex_lock(&root->fs_info->fs_mutex); @@ -160,7 +160,7 @@ out: int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) { - size_t offset = start - (page->index << PAGE_CACHE_SHIFT); + size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT); struct inode *inode = page->mapping->host; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; char *kaddr; @@ -688,7 +688,7 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, char *kaddr; int ret = 0; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - u64 page_start = page->index << PAGE_CACHE_SHIFT; + u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; set_page_extent_mapped(page); @@ -739,7 +739,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) goto out; } } - page_start = page->index << PAGE_CACHE_SHIFT; + page_start = (u64)page->index << PAGE_CACHE_SHIFT; ret = btrfs_cow_one_page(inode, page, offset); @@ -1618,7 +1618,7 @@ again: goto out; } - extent_offset = (page->index << PAGE_CACHE_SHIFT) - + extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) - extent_start + page_offset; copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset, size - extent_offset); @@ -1769,7 +1769,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) lock_page(page); wait_on_page_writeback(page); size = i_size_read(inode); - page_start = page->index << PAGE_CACHE_SHIFT; + page_start = (u64)page->index << PAGE_CACHE_SHIFT; if ((page->mapping != inode->i_mapping) || (page_start > size)) { @@ -2065,7 +2065,7 @@ int btrfs_defrag_file(struct file *file) { goto out_unlock; } } - page_start = page->index << PAGE_CACHE_SHIFT; + page_start = (u64)page->index << PAGE_CACHE_SHIFT; page_end = page_start + PAGE_CACHE_SIZE - 1; lock_extent(em_tree, page_start, page_end, GFP_NOFS); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index bdfe05cf26c..4af1c0dab7f 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -169,7 +169,7 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, clear_extent_dirty(dirty_pages, start, end, GFP_NOFS); while(start <= end) { index = start >> PAGE_CACHE_SHIFT; - start = (index + 1) << PAGE_CACHE_SHIFT; + start = (u64)(index + 1) << PAGE_CACHE_SHIFT; page = find_lock_page(btree_inode->i_mapping, index); if (!page) continue; -- cgit v1.2.3-70-g09d2 From 179e29e488cc74f1e9bd67bc45f70b832740e9ec Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 1 Nov 2007 11:28:41 -0400 Subject: Btrfs: Fix a number of inline extent problems that Yan Zheng reported. The fixes do a number of things: 1) Most btrfs_drop_extent callers will try to leave the inline extents in place. It can truncate bytes off the beginning of the inline extent if required. 2) writepage can now update the inline extent, allowing mmap writes to go directly into the inline extent. 3) btrfs_truncate_in_transaction truncates inline extents 4) extent_map.c fixed to not merge inline extent mappings and hole mappings together Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 52 ++++++++++++++++++++++++++++++++++----- fs/btrfs/ctree.h | 2 +- fs/btrfs/dir-item.c | 2 +- fs/btrfs/extent_map.c | 11 ++++++--- fs/btrfs/extent_map.h | 1 - fs/btrfs/file-item.c | 2 +- fs/btrfs/file.c | 36 ++++++++++++++++++++++----- fs/btrfs/inode.c | 67 ++++++++++++++++++++++++++++++++++++--------------- 8 files changed, 135 insertions(+), 38 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 516b90224a1..eef9c92f86d 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1930,7 +1930,7 @@ again: int btrfs_truncate_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, - u32 new_size) + u32 new_size, int from_end) { int ret = 0; int slot; @@ -1946,13 +1946,17 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, slot_orig = path->slots[0]; leaf = path->nodes[0]; + slot = path->slots[0]; + + old_size = btrfs_item_size_nr(leaf, slot); + if (old_size == new_size) + return 0; nritems = btrfs_header_nritems(leaf); data_end = leaf_data_end(root, leaf); - slot = path->slots[0]; old_data_start = btrfs_item_offset_nr(leaf, slot); - old_size = btrfs_item_size_nr(leaf, slot); BUG_ON(old_size <= new_size); + size_diff = old_size - new_size; BUG_ON(slot < 0); @@ -1984,9 +1988,45 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, } /* shift the data */ - memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + - data_end + size_diff, btrfs_leaf_data(leaf) + - data_end, old_data_start + new_size - data_end); + if (from_end) { + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + + data_end + size_diff, btrfs_leaf_data(leaf) + + data_end, old_data_start + new_size - data_end); + } else { + struct btrfs_disk_key disk_key; + u64 offset; + + btrfs_item_key(leaf, &disk_key, slot); + + if (btrfs_disk_key_type(&disk_key) == BTRFS_EXTENT_DATA_KEY) { + unsigned long ptr; + struct btrfs_file_extent_item *fi; + + fi = btrfs_item_ptr(leaf, slot, + struct btrfs_file_extent_item); + fi = (struct btrfs_file_extent_item *)( + (unsigned long)fi - size_diff); + + if (btrfs_file_extent_type(leaf, fi) == + BTRFS_FILE_EXTENT_INLINE) { + ptr = btrfs_item_ptr_offset(leaf, slot); + memmove_extent_buffer(leaf, ptr, + (unsigned long)fi, + offsetof(struct btrfs_file_extent_item, + disk_bytenr)); + } + } + + memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + + data_end + size_diff, btrfs_leaf_data(leaf) + + data_end, old_data_start - data_end); + + offset = btrfs_disk_key_offset(&disk_key); + btrfs_set_disk_key_offset(&disk_key, offset + size_diff); + btrfs_set_item_key(leaf, &disk_key, slot); + if (slot == 0) + fixup_low_keys(trans, root, path, &disk_key, 1); + } item = btrfs_item_nr(leaf, slot); btrfs_set_item_size(leaf, item, new_size); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 70e143b1357..d82afb618bf 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -907,7 +907,7 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root int btrfs_truncate_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, - u32 new_size); + u32 new_size, int from_end); int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_path *p, int ins_len, int cow); diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 6f19de41b87..514a1dc337a 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -249,7 +249,7 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, memmove_extent_buffer(leaf, ptr, ptr + sub_item_len, item_len - (ptr + sub_item_len - start)); ret = btrfs_truncate_item(trans, root, path, - item_len - sub_item_len); + item_len - sub_item_len, 1); } return 0; } diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 238cb1d81d5..44be9cfd30e 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -263,7 +263,12 @@ int add_extent_mapping(struct extent_map_tree *tree, if (prev && prev->end + 1 == em->start && ((em->block_start == EXTENT_MAP_HOLE && prev->block_start == EXTENT_MAP_HOLE) || - (em->block_start == prev->block_end + 1))) { + (em->block_start == EXTENT_MAP_INLINE && + prev->block_start == EXTENT_MAP_INLINE) || + (em->block_start == EXTENT_MAP_DELALLOC && + prev->block_start == EXTENT_MAP_DELALLOC) || + (em->block_start < EXTENT_MAP_DELALLOC - 1 && + em->block_start == prev->block_end + 1))) { em->start = prev->start; em->block_start = prev->block_start; rb_erase(&prev->rb_node, &tree->map); @@ -1618,13 +1623,13 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, u64 extent_offset; u64 last_byte = i_size_read(inode); u64 block_start; + u64 iosize; sector_t sector; struct extent_map *em; struct block_device *bdev; int ret; int nr = 0; size_t page_offset = 0; - size_t iosize; size_t blocksize; loff_t i_size = i_size_read(inode); unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; @@ -1684,7 +1689,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, clear_extent_dirty(tree, cur, page_end, GFP_NOFS); break; } - em = get_extent(inode, page, page_offset, cur, end, 0); + em = get_extent(inode, page, page_offset, cur, end, 1); if (IS_ERR(em) || !em) { SetPageError(page); break; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 39d78d31b6a..b6f6519f9ba 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -24,7 +24,6 @@ * page->private values. Every page that is controlled by the extent * map has page->private set to one. */ - #define EXTENT_PAGE_PRIVATE 1 #define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3 diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 7eb9a5412e2..614176e5285 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -291,7 +291,7 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, new_item_size = blocks * BTRFS_CRC32_SIZE; if (new_item_size >= btrfs_item_size_nr(leaf, slot)) return 0; - ret = btrfs_truncate_item(trans, root, path, new_item_size); + ret = btrfs_truncate_item(trans, root, path, new_item_size, 1); BUG_ON(ret); return ret; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index abdd9caad94..4e52f7ec1cb 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -115,8 +115,20 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, goto fail; } if (ret == 1) { + struct btrfs_key found_key; + + if (path->slots[0] == 0) + goto insert; + path->slots[0]--; leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + if (found_key.objectid != inode->i_ino) + goto insert; + + if (found_key.type != BTRFS_EXTENT_DATA_KEY) + goto insert; ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); @@ -152,6 +164,7 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans, ret = btrfs_search_slot(trans, root, &key, path, offset + size - found_end, 1); BUG_ON(ret != 0); + ret = btrfs_extend_item(trans, root, path, offset + size - found_end); if (ret) { @@ -292,7 +305,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, */ inline_size = end_pos; if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || - inline_size > 8192 || + inline_size > 32768 || inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { u64 last_end; @@ -312,7 +325,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, aligned_end = (pos + write_bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); err = btrfs_drop_extents(trans, root, inode, start_pos, - aligned_end, end_pos, &hint_byte); + aligned_end, aligned_end, &hint_byte); if (err) goto failed; err = insert_inline_extent(trans, root, inode, start_pos, @@ -456,13 +469,15 @@ next_slot: goto next_slot; } - /* FIXME, there's only one inline extent allowed right now */ if (found_inline) { u64 mask = root->sectorsize - 1; search_start = (extent_end + mask) & ~mask; } else search_start = extent_end; + if (end <= extent_end && start >= key.offset && found_inline) { + *hint_byte = EXTENT_MAP_INLINE; + } if (end < extent_end && end >= key.offset) { if (found_extent) { u64 disk_bytenr = @@ -479,8 +494,10 @@ next_slot: BUG_ON(ret); } } - if (!found_inline) - bookend = 1; + bookend = 1; + if (found_inline && start <= key.offset && + inline_end < extent_end) + keep = 1; } /* truncate existing extent */ if (start > key.offset) { @@ -510,7 +527,7 @@ next_slot: new_size = btrfs_file_extent_calc_inline_size( inline_end - key.offset); btrfs_truncate_item(trans, root, path, - new_size); + new_size, 1); } } /* delete the entire extent */ @@ -551,6 +568,13 @@ next_slot: if (!bookend) continue; } + if (bookend && found_inline && start <= key.offset && + inline_end < extent_end) { + u32 new_size; + new_size = btrfs_file_extent_calc_inline_size( + extent_end - inline_end); + btrfs_truncate_item(trans, root, path, new_size, 0); + } /* create bookend, splitting the extent in two */ if (bookend && found_extent) { struct btrfs_key ins; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5d10b64e42d..0c65141b999 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -89,6 +89,9 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) ret = btrfs_drop_extents(trans, root, inode, start, start + num_bytes, start, &alloc_hint); + if (alloc_hint == EXTENT_MAP_INLINE) + goto out; + ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { @@ -558,6 +561,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, u64 item_end = 0; int found_extent; int del_item; + int extent_type = -1; btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1); path = btrfs_alloc_path(); @@ -597,10 +601,15 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, if (found_type == BTRFS_EXTENT_DATA_KEY) { fi = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - if (btrfs_file_extent_type(leaf, fi) != - BTRFS_FILE_EXTENT_INLINE) { + extent_type = btrfs_file_extent_type(leaf, fi); + if (extent_type != BTRFS_FILE_EXTENT_INLINE) { item_end += btrfs_file_extent_num_bytes(leaf, fi); + } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { + struct btrfs_item *item = btrfs_item_nr(leaf, + path->slots[0]); + item_end += btrfs_file_extent_inline_len(leaf, + item); } } if (found_type == BTRFS_CSUM_ITEM_KEY) { @@ -608,7 +617,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, inode->i_size); BUG_ON(ret); } - if (item_end < inode->i_size) { + if (item_end <= inode->i_size) { if (found_type == BTRFS_DIR_ITEM_KEY) { found_type = BTRFS_INODE_ITEM_KEY; } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { @@ -629,9 +638,10 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, found_extent = 0; /* FIXME, shrink the extent if the ref count is only 1 */ - if (found_type == BTRFS_EXTENT_DATA_KEY && - btrfs_file_extent_type(leaf, fi) != - BTRFS_FILE_EXTENT_INLINE) { + if (found_type != BTRFS_EXTENT_DATA_KEY) + goto delete; + + if (extent_type != BTRFS_FILE_EXTENT_INLINE) { u64 num_dec; extent_start = btrfs_file_extent_disk_bytenr(leaf, fi); if (!del_item) { @@ -659,7 +669,15 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, inode->i_blocks -= num_dec; } } + } else if (extent_type == BTRFS_FILE_EXTENT_INLINE && + !del_item) { + u32 newsize = inode->i_size - found_key.offset; + newsize = btrfs_file_extent_calc_inline_size(newsize); + ret = btrfs_truncate_item(trans, root, path, + newsize, 1); + BUG_ON(ret); } +delete: if (del_item) { ret = btrfs_del_item(trans, root, path); if (ret) @@ -769,7 +787,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) u64 pos = (inode->i_size + mask) & ~mask; u64 block_end = attr->ia_size | mask; u64 hole_size; - u64 alloc_hint; + u64 alloc_hint = 0; if (attr->ia_size <= pos) goto out; @@ -786,8 +804,11 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) pos, pos + hole_size, pos, &alloc_hint); - err = btrfs_insert_file_extent(trans, root, inode->i_ino, - pos, 0, 0, hole_size); + if (alloc_hint != EXTENT_MAP_INLINE) { + err = btrfs_insert_file_extent(trans, root, + inode->i_ino, + pos, 0, 0, hole_size); + } btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); unlock_extent(em_tree, pos, block_end, GFP_NOFS); @@ -1531,8 +1552,8 @@ again: em->end = EXTENT_MAP_HOLE; } em->bdev = inode->i_sb->s_bdev; - ret = btrfs_lookup_file_extent(NULL, root, path, - objectid, start, 0); + ret = btrfs_lookup_file_extent(trans, root, path, + objectid, start, trans != NULL); if (ret < 0) { err = ret; goto out; @@ -1627,15 +1648,23 @@ again: ((u64)root->sectorsize -1); map = kmap(page); ptr = btrfs_file_extent_inline_start(item) + extent_offset; - read_extent_buffer(leaf, map + page_offset, ptr, copy_size); - - if (em->start + copy_size <= em->end) { - size = min_t(u64, em->end + 1 - em->start, - PAGE_CACHE_SIZE - page_offset) - copy_size; - memset(map + page_offset + copy_size, 0, size); + if (create == 0 && !PageUptodate(page)) { + read_extent_buffer(leaf, map + page_offset, ptr, + copy_size); + flush_dcache_page(page); + } else if (create && PageUptodate(page)) { + if (!trans) { + kunmap(page); + free_extent_map(em); + em = NULL; + btrfs_release_path(root, path); + trans = btrfs_start_transaction(root, 1); + goto again; + } + write_extent_buffer(leaf, map + page_offset, ptr, + copy_size); + btrfs_mark_buffer_dirty(leaf); } - - flush_dcache_page(page); kunmap(page); set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS); goto insert; -- cgit v1.2.3-70-g09d2 From b293f02e1423f2099744f3ade23ddd83b65321fc Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 1 Nov 2007 19:45:34 -0400 Subject: Btrfs: Add writepages support Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 135 ++++++++++++++++++++++++++++++++++++++++---------- fs/btrfs/extent_map.h | 4 ++ fs/btrfs/inode.c | 10 ++++ 3 files changed, 124 insertions(+), 25 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index ff8881fb56d..b0677c84bb7 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -10,6 +10,7 @@ #include #include #include +#include #include "extent_map.h" /* temporary define until extent_map moves out of btrfs */ @@ -35,6 +36,12 @@ struct tree_entry { struct rb_node rb_node; }; +struct extent_page_data { + struct bio *bio; + struct extent_map_tree *tree; + get_extent_t *get_extent; +}; + void __init extent_map_init(void) { extent_map_cache = btrfs_cache_create("extent_map", @@ -1460,40 +1467,76 @@ static int end_bio_extent_preparewrite(struct bio *bio, #endif } -static int submit_extent_page(int rw, struct extent_map_tree *tree, - struct page *page, sector_t sector, - size_t size, unsigned long offset, - struct block_device *bdev, - bio_end_io_t end_io_func) +static struct bio * +extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, + gfp_t gfp_flags) { struct bio *bio; - int ret = 0; - bio = bio_alloc(GFP_NOIO, 1); + bio = bio_alloc(gfp_flags, nr_vecs); - bio->bi_sector = sector; - bio->bi_bdev = bdev; - bio->bi_io_vec[0].bv_page = page; - bio->bi_io_vec[0].bv_len = size; - bio->bi_io_vec[0].bv_offset = offset; - - bio->bi_vcnt = 1; - bio->bi_idx = 0; - bio->bi_size = size; + if (bio == NULL && (current->flags & PF_MEMALLOC)) { + while (!bio && (nr_vecs /= 2)) + bio = bio_alloc(gfp_flags, nr_vecs); + } - bio->bi_end_io = end_io_func; - bio->bi_private = tree; + if (bio) { + bio->bi_bdev = bdev; + bio->bi_sector = first_sector; + } + return bio; +} +static int submit_one_bio(int rw, struct bio *bio) +{ + int ret = 0; bio_get(bio); submit_bio(rw, bio); - if (bio_flagged(bio, BIO_EOPNOTSUPP)) ret = -EOPNOTSUPP; - bio_put(bio); return ret; } +static int submit_extent_page(int rw, struct extent_map_tree *tree, + struct page *page, sector_t sector, + size_t size, unsigned long offset, + struct block_device *bdev, + struct bio **bio_ret, + int max_pages, + bio_end_io_t end_io_func) +{ + int ret = 0; + struct bio *bio; + int nr; + + if (bio_ret && *bio_ret) { + bio = *bio_ret; + if (bio->bi_sector + (bio->bi_size >> 9) != sector || + bio_add_page(bio, page, size, offset) < size) { + ret = submit_one_bio(rw, bio); + bio = NULL; + } else { + return 0; + } + } + nr = min(max_pages, bio_get_nr_vecs(bdev)); + bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); + if (!bio) { + printk("failed to allocate bio nr %d\n", nr); + } + bio_add_page(bio, page, size, offset); + bio->bi_end_io = end_io_func; + bio->bi_private = tree; + if (bio_ret) { + *bio_ret = bio; + } else { + ret = submit_one_bio(rw, bio); + } + + return ret; +} + void set_page_extent_mapped(struct page *page) { if (!PagePrivate(page)) { @@ -1590,7 +1633,8 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, if (!ret) { ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, - bdev, end_bio_extent_readpage); + bdev, NULL, 1, + end_bio_extent_readpage); } if (ret) SetPageError(page); @@ -1613,11 +1657,12 @@ EXPORT_SYMBOL(extent_read_full_page); * are found, they are marked writeback. Then the lock bits are removed * and the end_io handler clears the writeback ranges */ -int extent_write_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent, - struct writeback_control *wbc) +static int __extent_writepage(struct page *page, struct writeback_control *wbc, + void *data) { struct inode *inode = page->mapping->host; + struct extent_page_data *epd = data; + struct extent_map_tree *tree = epd->tree; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = start + PAGE_CACHE_SIZE - 1; u64 end; @@ -1691,7 +1736,7 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, clear_extent_dirty(tree, cur, page_end, GFP_NOFS); break; } - em = get_extent(inode, page, page_offset, cur, end, 1); + em = epd->get_extent(inode, page, page_offset, cur, end, 1); if (IS_ERR(em) || !em) { SetPageError(page); break; @@ -1734,9 +1779,12 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, if (ret) SetPageError(page); else { + unsigned long nr = end_index + 1; set_range_writeback(tree, cur, cur + iosize - 1); + ret = submit_extent_page(WRITE, tree, page, sector, iosize, page_offset, bdev, + &epd->bio, nr, end_bio_extent_writepage); if (ret) SetPageError(page); @@ -1750,8 +1798,44 @@ done: unlock_page(page); return 0; } + +int extent_write_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + + ret = __extent_writepage(page, wbc, &epd); + if (epd.bio) + submit_one_bio(WRITE, epd.bio); + return ret; +} EXPORT_SYMBOL(extent_write_full_page); +int extent_writepages(struct extent_map_tree *tree, + struct address_space *mapping, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + + ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); + if (epd.bio) + submit_one_bio(WRITE, epd.bio); + return ret; +} +EXPORT_SYMBOL(extent_writepages); + /* * basic invalidatepage code, this waits on any locked or writeback * ranges corresponding to the page, and then deletes any extent state @@ -1869,6 +1953,7 @@ int extent_prepare_write(struct extent_map_tree *tree, EXTENT_LOCKED, 0, NULL, GFP_NOFS); ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, em->bdev, + NULL, 1, end_bio_extent_preparewrite); iocount++; block_start = block_start + iosize; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index b6f6519f9ba..0b0dcf19fe4 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -136,6 +136,10 @@ int extent_invalidatepage(struct extent_map_tree *tree, int extent_write_full_page(struct extent_map_tree *tree, struct page *page, get_extent_t *get_extent, struct writeback_control *wbc); +int extent_writepages(struct extent_map_tree *tree, + struct address_space *mapping, + get_extent_t *get_extent, + struct writeback_control *wbc); int extent_prepare_write(struct extent_map_tree *tree, struct inode *inode, struct page *page, unsigned from, unsigned to, get_extent_t *get_extent); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0c65141b999..25f32d7c7ee 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1747,6 +1747,15 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } + +static int btrfs_writepages(struct address_space *mapping, + struct writeback_control *wbc) +{ + struct extent_map_tree *tree; + tree = &BTRFS_I(mapping->host)->extent_tree; + return extent_writepages(tree, mapping, btrfs_get_extent, wbc); +} + static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) { struct extent_map_tree *tree; @@ -2526,6 +2535,7 @@ static struct extent_map_ops btrfs_extent_map_ops = { static struct address_space_operations btrfs_aops = { .readpage = btrfs_readpage, .writepage = btrfs_writepage, + .writepages = btrfs_writepages, .sync_page = block_sync_page, .prepare_write = btrfs_prepare_write, .commit_write = btrfs_commit_write, -- cgit v1.2.3-70-g09d2 From 008630c17cc5654e141c7fd24a3e737de643fec1 Mon Sep 17 00:00:00 2001 From: Yan Date: Wed, 7 Nov 2007 13:31:09 -0500 Subject: Properly delete csum item in btrfs_truncate_in_trans. When 'item_end' is equal to 'inode->i_size', 'found_type' is updated and current item is skipped. This behavior is correct for extent item, but incorrect for csum item. For example, there is a csum item with 'offset == 0'. When deleting the inode, 'inode->i_size' is set to 0, so the csum item isn't deleted. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 25f32d7c7ee..90cb71d6829 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -611,13 +611,14 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, item_end += btrfs_file_extent_inline_len(leaf, item); } + item_end--; } if (found_type == BTRFS_CSUM_ITEM_KEY) { ret = btrfs_csum_truncate(trans, root, path, inode->i_size); BUG_ON(ret); } - if (item_end <= inode->i_size) { + if (item_end < inode->i_size) { if (found_type == BTRFS_DIR_ITEM_KEY) { found_type = BTRFS_INODE_ITEM_KEY; } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { -- cgit v1.2.3-70-g09d2 From 3ab2fb5a8cb003897016b6eb38ddad916226c1b2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 8 Nov 2007 10:59:22 -0500 Subject: Btrfs: Add readpages support Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 70 +++++++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/extent_map.h | 4 +++ fs/btrfs/inode.c | 12 ++++++++- 3 files changed, 78 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 754bc42c162..0077c6c1d9f 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -11,6 +11,7 @@ #include #include #include +#include #include "extent_map.h" /* temporary define until extent_map moves out of btrfs */ @@ -1503,7 +1504,7 @@ static int submit_extent_page(int rw, struct extent_map_tree *tree, size_t size, unsigned long offset, struct block_device *bdev, struct bio **bio_ret, - int max_pages, + unsigned long max_pages, bio_end_io_t end_io_func) { int ret = 0; @@ -1520,7 +1521,7 @@ static int submit_extent_page(int rw, struct extent_map_tree *tree, return 0; } } - nr = min(max_pages, bio_get_nr_vecs(bdev)); + nr = min_t(int, max_pages, bio_get_nr_vecs(bdev)); bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); if (!bio) { printk("failed to allocate bio nr %d\n", nr); @@ -1552,8 +1553,10 @@ void set_page_extent_mapped(struct page *page) * into the tree that are removed when the IO is done (by the end_io * handlers) */ -int extent_read_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent) +static int __extent_read_full_page(struct extent_map_tree *tree, + struct page *page, + get_extent_t *get_extent, + struct bio **bio) { struct inode *inode = page->mapping->host; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -1631,10 +1634,12 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, cur + iosize - 1); } if (!ret) { + unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + nr -= page->index; ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, - bdev, NULL, 1, - end_bio_extent_readpage); + sector, iosize, page_offset, + bdev, bio, nr, + end_bio_extent_readpage); } if (ret) SetPageError(page); @@ -1649,6 +1654,18 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page, } return 0; } + +int extent_read_full_page(struct extent_map_tree *tree, struct page *page, + get_extent_t *get_extent) +{ + struct bio *bio = NULL; + int ret; + + ret = __extent_read_full_page(tree, page, get_extent, &bio); + if (bio) + submit_one_bio(READ, bio); + return ret; +} EXPORT_SYMBOL(extent_read_full_page); /* @@ -1836,6 +1853,45 @@ int extent_writepages(struct extent_map_tree *tree, } EXPORT_SYMBOL(extent_writepages); +int extent_readpages(struct extent_map_tree *tree, + struct address_space *mapping, + struct list_head *pages, unsigned nr_pages, + get_extent_t get_extent) +{ + struct bio *bio = NULL; + unsigned page_idx; + struct pagevec pvec; + + pagevec_init(&pvec, 0); + for (page_idx = 0; page_idx < nr_pages; page_idx++) { + struct page *page = list_entry(pages->prev, struct page, lru); + + prefetchw(&page->flags); + list_del(&page->lru); + /* + * what we want to do here is call add_to_page_cache_lru, + * but that isn't exported, so we reproduce it here + */ + if (!add_to_page_cache(page, mapping, + page->index, GFP_KERNEL)) { + + /* open coding of lru_cache_add, also not exported */ + page_cache_get(page); + if (!pagevec_add(&pvec, page)) + __pagevec_lru_add(&pvec); + __extent_read_full_page(tree, page, get_extent, &bio); + } + page_cache_release(page); + } + if (pagevec_count(&pvec)) + __pagevec_lru_add(&pvec); + BUG_ON(!list_empty(pages)); + if (bio) + submit_one_bio(READ, bio); + return 0; +} +EXPORT_SYMBOL(extent_readpages); + /* * basic invalidatepage code, this waits on any locked or writeback * ranges corresponding to the page, and then deletes any extent state diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 0b0dcf19fe4..fbd23a72f53 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -140,6 +140,10 @@ int extent_writepages(struct extent_map_tree *tree, struct address_space *mapping, get_extent_t *get_extent, struct writeback_control *wbc); +int extent_readpages(struct extent_map_tree *tree, + struct address_space *mapping, + struct list_head *pages, unsigned nr_pages, + get_extent_t get_extent); int extent_prepare_write(struct extent_map_tree *tree, struct inode *inode, struct page *page, unsigned from, unsigned to, get_extent_t *get_extent); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 90cb71d6829..d6d32465bc9 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1748,7 +1748,6 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } - static int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc) { @@ -1757,6 +1756,16 @@ static int btrfs_writepages(struct address_space *mapping, return extent_writepages(tree, mapping, btrfs_get_extent, wbc); } +static int +btrfs_readpages(struct file *file, struct address_space *mapping, + struct list_head *pages, unsigned nr_pages) +{ + struct extent_map_tree *tree; + tree = &BTRFS_I(mapping->host)->extent_tree; + return extent_readpages(tree, mapping, pages, nr_pages, + btrfs_get_extent); +} + static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) { struct extent_map_tree *tree; @@ -2537,6 +2546,7 @@ static struct address_space_operations btrfs_aops = { .readpage = btrfs_readpage, .writepage = btrfs_writepage, .writepages = btrfs_writepages, + .readpages = btrfs_readpages, .sync_page = block_sync_page, .prepare_write = btrfs_prepare_write, .commit_write = btrfs_commit_write, -- cgit v1.2.3-70-g09d2 From 5103e947b9b7ac18ddb21a04ee3486e94c6504d7 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 16 Nov 2007 11:45:54 -0500 Subject: xattr support for btrfs Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/acl.c | 66 +++++++ fs/btrfs/ctree.h | 22 ++- fs/btrfs/dir-item.c | 100 +++++++++- fs/btrfs/inode.c | 15 +- fs/btrfs/print-tree.c | 3 +- fs/btrfs/super.c | 2 + fs/btrfs/xattr.c | 497 ++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/xattr.h | 58 ++++++ 9 files changed, 751 insertions(+), 14 deletions(-) create mode 100644 fs/btrfs/acl.c create mode 100644 fs/btrfs/xattr.c create mode 100644 fs/btrfs/xattr.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 551743be5f0..d5804c5ca07 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -5,7 +5,7 @@ obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ - extent_map.o sysfs.o struct-funcs.o + extent_map.o sysfs.o struct-funcs.o xattr.o acl.o #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \ # root-tree.o dir-item.o hash.o file-item.o inode-item.o \ diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c new file mode 100644 index 00000000000..aee9f0657c3 --- /dev/null +++ b/fs/btrfs/acl.c @@ -0,0 +1,66 @@ +/* + * Copyright (C) 2007 Red Hat. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include +#include +#include "ctree.h" +#include "xattr.h" + +static int btrfs_xattr_acl_access_get(struct inode *inode, const char *name, + void *value, size_t size) +{ + return btrfs_xattr_get(inode, BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, name, + value, size); +} + +static int btrfs_xattr_acl_access_set(struct inode *inode, const char *name, + const void *value, size_t size, int flags) +{ + return btrfs_xattr_set(inode, BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, name, + value, size, flags); +} + +static int btrfs_xattr_acl_default_get(struct inode *inode, const char *name, + void *value, size_t size) +{ + return btrfs_xattr_get(inode, BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, + name, value, size); +} + +static int btrfs_xattr_acl_default_set(struct inode *inode, const char *name, + const void *value, size_t size, int flags) +{ + return btrfs_xattr_set(inode, BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, + name, value, size, flags); +} + +struct xattr_handler btrfs_xattr_acl_default_handler = { + .prefix = POSIX_ACL_XATTR_DEFAULT, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_acl_default_get, + .set = btrfs_xattr_acl_default_set, +}; + +struct xattr_handler btrfs_xattr_acl_access_handler = { + .prefix = POSIX_ACL_XATTR_ACCESS, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_acl_access_get, + .set = btrfs_xattr_acl_access_set, +}; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 030d21d7f98..27cadae1af6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -63,7 +63,8 @@ extern struct kmem_cache *btrfs_path_cachep; #define BTRFS_FT_FIFO 5 #define BTRFS_FT_SOCK 6 #define BTRFS_FT_SYMLINK 7 -#define BTRFS_FT_MAX 8 +#define BTRFS_FT_XATTR 8 +#define BTRFS_FT_MAX 9 /* * the key defines the order in the tree, and so it also defines (optimal) @@ -226,7 +227,7 @@ struct btrfs_inode_item { struct btrfs_dir_item { struct btrfs_disk_key location; - __le16 flags; + __le16 data_len; __le16 name_len; u8 type; } __attribute__ ((__packed__)); @@ -367,7 +368,7 @@ struct btrfs_root { * the FS */ #define BTRFS_INODE_ITEM_KEY 1 - +#define BTRFS_XATTR_ITEM_KEY 2 /* reserve 2-15 close to the inode for later flexibility */ /* @@ -621,7 +622,7 @@ static inline void btrfs_set_item_key(struct extent_buffer *eb, } /* struct btrfs_dir_item */ -BTRFS_SETGET_FUNCS(dir_flags, struct btrfs_dir_item, flags, 16); +BTRFS_SETGET_FUNCS(dir_data_len, struct btrfs_dir_item, data_len, 16); BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8); BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16); @@ -962,6 +963,15 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_dir_item *di); +int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, const char *name, + u16 name_len, const void *data, u16 data_len, + u64 dir); +struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 dir, + const char *name, u16 name_len, + int mod); /* inode-map.c */ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, struct btrfs_root *fs_root, @@ -1039,4 +1049,8 @@ int btrfs_sysfs_add_root(struct btrfs_root *root); void btrfs_sysfs_del_root(struct btrfs_root *root); void btrfs_sysfs_del_super(struct btrfs_fs_info *root); +/* xattr.c */ +ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size); +int btrfs_delete_xattrs(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode); #endif diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 514a1dc337a..ddbe12ae0d6 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -43,8 +43,6 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle return ERR_PTR(-EEXIST); ret = btrfs_extend_item(trans, root, path, data_size); WARN_ON(ret > 0); - if (ret) - return ERR_PTR(ret); } if (ret < 0) return ERR_PTR(ret); @@ -57,6 +55,57 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle return (struct btrfs_dir_item *)ptr; } +int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, const char *name, + u16 name_len, const void *data, u16 data_len, + u64 dir) +{ + int ret = 0; + struct btrfs_path *path; + struct btrfs_dir_item *dir_item; + unsigned long name_ptr, data_ptr; + struct btrfs_key key, location; + struct btrfs_disk_key disk_key; + struct extent_buffer *leaf; + u32 data_size; + + key.objectid = dir; + btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY); + ret = btrfs_name_hash(name, name_len, &key.offset); + BUG_ON(ret); + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + data_size = sizeof(*dir_item) + name_len + data_len; + dir_item = insert_with_overflow(trans, root, path, &key, data_size, + name, name_len); + /* + * FIXME: at some point we should handle xattr's that are larger than + * what we can fit in our leaf. We set location to NULL b/c we arent + * pointing at anything else, that will change if we store the xattr + * data in a separate inode. + */ + BUG_ON(IS_ERR(dir_item)); + memset(&location, 0, sizeof(location)); + + leaf = path->nodes[0]; + btrfs_cpu_key_to_disk(&disk_key, &location); + btrfs_set_dir_item_key(leaf, dir_item, &disk_key); + btrfs_set_dir_type(leaf, dir_item, BTRFS_FT_XATTR); + btrfs_set_dir_name_len(leaf, dir_item, name_len); + btrfs_set_dir_data_len(leaf, dir_item, data_len); + name_ptr = (unsigned long)(dir_item + 1); + data_ptr = (unsigned long)((char *)name_ptr + name_len); + + write_extent_buffer(leaf, name, name_ptr, name_len); + write_extent_buffer(leaf, data, data_ptr, data_len); + btrfs_mark_buffer_dirty(path->nodes[0]); + + btrfs_free_path(path); + return ret; +} + int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, u64 dir, struct btrfs_key *location, u8 type) @@ -90,7 +139,7 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_cpu_key_to_disk(&disk_key, location); btrfs_set_dir_item_key(leaf, dir_item, &disk_key); btrfs_set_dir_type(leaf, dir_item, type); - btrfs_set_dir_flags(leaf, dir_item, 0); + btrfs_set_dir_data_len(leaf, dir_item, 0); btrfs_set_dir_name_len(leaf, dir_item, name_len); name_ptr = (unsigned long)(dir_item + 1); @@ -117,7 +166,7 @@ second_insert: btrfs_cpu_key_to_disk(&disk_key, location); btrfs_set_dir_item_key(leaf, dir_item, &disk_key); btrfs_set_dir_type(leaf, dir_item, type); - btrfs_set_dir_flags(leaf, dir_item, 0); + btrfs_set_dir_data_len(leaf, dir_item, 0); btrfs_set_dir_name_len(leaf, dir_item, name_len); name_ptr = (unsigned long)(dir_item + 1); write_extent_buffer(leaf, name, name_ptr, name_len); @@ -194,6 +243,43 @@ btrfs_lookup_dir_index_item(struct btrfs_trans_handle *trans, return btrfs_match_dir_item_name(root, path, name, name_len); } +struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 dir, + const char *name, u16 name_len, + int mod) +{ + int ret; + struct btrfs_key key; + int ins_len = mod < 0 ? -1 : 0; + int cow = mod != 0; + struct btrfs_key found_key; + struct extent_buffer *leaf; + + key.objectid = dir; + btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY); + ret = btrfs_name_hash(name, name_len, &key.offset); + BUG_ON(ret); + ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow); + if (ret < 0) + return ERR_PTR(ret); + if (ret > 0) { + if (path->slots[0] == 0) + return NULL; + path->slots[0]--; + } + + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + if (found_key.objectid != dir || + btrfs_key_type(&found_key) != BTRFS_XATTR_ITEM_KEY || + found_key.offset != key.offset) + return NULL; + + return btrfs_match_dir_item_name(root, path, name, name_len); +} + struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path, const char *name, int name_len) @@ -210,7 +296,8 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, total_len = btrfs_item_size_nr(leaf, path->slots[0]); while(cur < total_len) { this_len = sizeof(*dir_item) + - btrfs_dir_name_len(leaf, dir_item); + btrfs_dir_name_len(leaf, dir_item) + + btrfs_dir_data_len(leaf, dir_item); name_ptr = (unsigned long)(dir_item + 1); if (btrfs_dir_name_len(leaf, dir_item) == name_len && @@ -236,7 +323,8 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, int ret = 0; leaf = path->nodes[0]; - sub_item_len = sizeof(*di) + btrfs_dir_name_len(leaf, di); + sub_item_len = sizeof(*di) + btrfs_dir_name_len(leaf, di) + + btrfs_dir_data_len(leaf, di); item_len = btrfs_item_size_nr(leaf, path->slots[0]); if (sub_item_len == item_len) { ret = btrfs_del_item(trans, root, path); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d6d32465bc9..37f8e0321af 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -32,6 +32,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -839,6 +840,9 @@ void btrfs_delete_inode(struct inode *inode) btrfs_set_trans_block_group(trans, inode); ret = btrfs_truncate_in_trans(trans, root, inode); + if (ret) + goto no_delete_lock; + ret = btrfs_delete_xattrs(trans, root, inode); if (ret) goto no_delete_lock; ret = btrfs_free_inode(trans, root, inode); @@ -1110,7 +1114,8 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) if (over) goto nopos; - di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di); + di_len = btrfs_dir_name_len(leaf, di) + + btrfs_dir_data_len(leaf, di) +sizeof(*di); di_cur += di_len; di = (struct btrfs_dir_item *)((char *)di + di_len); } @@ -2519,6 +2524,10 @@ static struct inode_operations btrfs_dir_inode_operations = { .symlink = btrfs_symlink, .setattr = btrfs_setattr, .mknod = btrfs_mknod, + .setxattr = generic_setxattr, + .getxattr = generic_getxattr, + .listxattr = btrfs_listxattr, + .removexattr = generic_removexattr, }; static struct inode_operations btrfs_dir_ro_inode_operations = { @@ -2567,6 +2576,10 @@ static struct inode_operations btrfs_file_inode_operations = { .truncate = btrfs_truncate, .getattr = btrfs_getattr, .setattr = btrfs_setattr, + .setxattr = generic_setxattr, + .getxattr = generic_getxattr, + .listxattr = btrfs_listxattr, + .removexattr = generic_removexattr, }; static struct inode_operations btrfs_special_inode_operations = { diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 9f8696c8a8e..030324febf6 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -58,9 +58,8 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) case BTRFS_DIR_ITEM_KEY: di = btrfs_item_ptr(l, i, struct btrfs_dir_item); btrfs_dir_item_key_to_cpu(l, di, &found_key); - printk("\t\tdir oid %llu flags %u type %u\n", + printk("\t\tdir oid %llu type %u\n", (unsigned long long)found_key.objectid, - btrfs_dir_flags(l, di), btrfs_dir_type(l, di)); break; case BTRFS_ROOT_ITEM_KEY: diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index f94aa1f97a0..c46bc391179 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -40,6 +40,7 @@ #include "btrfs_inode.h" #include "ioctl.h" #include "print-tree.h" +#include "xattr.h" #define BTRFS_SUPER_MAGIC 0x9123683E @@ -106,6 +107,7 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) sb->s_maxbytes = MAX_LFS_FILESIZE; sb->s_magic = BTRFS_SUPER_MAGIC; sb->s_op = &btrfs_super_ops; + sb->s_xattr = btrfs_xattr_handlers; sb->s_time_gran = 1; tree_root = open_ctree(sb); diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c new file mode 100644 index 00000000000..f4ac5e0bbad --- /dev/null +++ b/fs/btrfs/xattr.c @@ -0,0 +1,497 @@ +/* + * Copyright (C) 2007 Red Hat. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include +#include +#include +#include "ctree.h" +#include "btrfs_inode.h" +#include "transaction.h" +#include "xattr.h" +#include "disk-io.h" + +static struct xattr_handler *btrfs_xattr_handler_map[] = { + [BTRFS_XATTR_INDEX_USER] = &btrfs_xattr_user_handler, + [BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS] = &btrfs_xattr_acl_access_handler, + [BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT] = &btrfs_xattr_acl_default_handler, + [BTRFS_XATTR_INDEX_TRUSTED] = &btrfs_xattr_trusted_handler, + [BTRFS_XATTR_INDEX_SECURITY] = &btrfs_xattr_security_handler, + [BTRFS_XATTR_INDEX_SYSTEM] = &btrfs_xattr_system_handler, +}; + +struct xattr_handler *btrfs_xattr_handlers[] = { + &btrfs_xattr_user_handler, + &btrfs_xattr_acl_access_handler, + &btrfs_xattr_acl_default_handler, + &btrfs_xattr_trusted_handler, + &btrfs_xattr_security_handler, + &btrfs_xattr_system_handler, + NULL, +}; + +/* + * @param name - the xattr name + * @return - the xattr_handler for the xattr, NULL if its not found + * + * use this with listxattr where we don't already know the type of xattr we + * have + */ +static struct xattr_handler *find_btrfs_xattr_handler(struct extent_buffer *l, + unsigned long name_ptr, + u16 name_len) +{ + struct xattr_handler *handler = NULL; + int i = 0; + + for (handler = btrfs_xattr_handlers[i]; handler != NULL; i++, + handler = btrfs_xattr_handlers[i]) { + u16 prefix_len = strlen(handler->prefix); + + if (name_len < prefix_len) + continue; + + if (memcmp_extent_buffer(l, handler->prefix, name_ptr, + prefix_len) == 0) + break; + } + + return handler; +} + +/* + * @param name_index - the index for the xattr handler + * @return the xattr_handler if we found it, NULL otherwise + * + * use this if we know the type of the xattr already + */ +static struct xattr_handler *btrfs_xattr_handler(int name_index) +{ + struct xattr_handler *handler = NULL; + + if (name_index >= 0 && + name_index < ARRAY_SIZE(btrfs_xattr_handler_map)) + handler = btrfs_xattr_handler_map[name_index]; + + return handler; +} + +static inline char *get_name(const char *name, int name_index) +{ + char *ret = NULL; + struct xattr_handler *handler = btrfs_xattr_handler(name_index); + int prefix_len; + + if (!handler) + return ret; + + prefix_len = strlen(handler->prefix); + + ret = kmalloc(strlen(name) + prefix_len + 1, GFP_KERNEL); + if (!ret) + return ret; + + memcpy(ret, handler->prefix, prefix_len); + memcpy(ret+prefix_len, name, strlen(name)); + ret[prefix_len + strlen(name)] = '\0'; + + return ret; +} + +size_t btrfs_xattr_generic_list(struct inode *inode, char *list, + size_t list_size, const char *name, + size_t name_len) +{ + if (list && (name_len+1) <= list_size) { + memcpy(list, name, name_len); + list[name_len] = '\0'; + } else + return -ERANGE; + + return name_len+1; +} + +ssize_t btrfs_xattr_get(struct inode *inode, int name_index, + const char *attr_name, void *buffer, size_t size) +{ + struct btrfs_dir_item *di; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_path *path; + struct extent_buffer *leaf; + struct xattr_handler *handler = btrfs_xattr_handler(name_index); + int ret = 0; + unsigned long data_ptr; + char *name; + + if (!handler) + return -EOPNOTSUPP; + + /* just in case... */ + if (*attr_name == '\0') + return -EINVAL; + + name = get_name(attr_name, name_index); + if (!name) + return -ENOMEM; + + path = btrfs_alloc_path(); + if (!path) { + kfree(name); + return -ENOMEM; + } + + mutex_lock(&root->fs_info->fs_mutex); + /* lookup the xattr by name */ + di = btrfs_lookup_xattr(NULL, root, path, inode->i_ino, name, + strlen(name), 0); + if (!di || IS_ERR(di)) { + ret = -ENODATA; + goto out; + } + + leaf = path->nodes[0]; + /* if size is 0, that means we want the size of the attr */ + if (!size) { + ret = btrfs_dir_data_len(leaf, di); + goto out; + } + + /* now get the data out of our dir_item */ + if (btrfs_dir_data_len(leaf, di) > size) { + ret = -ERANGE; + goto out; + } + data_ptr = (unsigned long)((char *)(di + 1) + + btrfs_dir_name_len(leaf, di)); + read_extent_buffer(leaf, buffer, data_ptr, + btrfs_dir_name_len(leaf, di)); + ret = btrfs_dir_data_len(leaf, di); + +out: + mutex_unlock(&root->fs_info->fs_mutex); + kfree(name); + btrfs_free_path(path); + return ret; +} + +int btrfs_xattr_set(struct inode *inode, int name_index, + const char *attr_name, const void *value, size_t size, + int flags) +{ + struct btrfs_dir_item *di; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + struct btrfs_path *path; + struct xattr_handler *handler = btrfs_xattr_handler(name_index); + char *name; + int ret = 0, mod = 0; + + if (!handler) + return -EOPNOTSUPP; + + /* just in case... */ + if (*attr_name == '\0') + return -EINVAL; + + name = get_name(attr_name, name_index); + if (!name) + return -ENOMEM; + + path = btrfs_alloc_path(); + if (!path) { + kfree(name); + return -ENOMEM; + } + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + + /* first lets see if we already have this xattr */ + di = btrfs_lookup_xattr(trans, root, path, inode->i_ino, name, + strlen(name), -1); + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto out; + } + + /* ok we already have this xattr, lets remove it */ + if (di) { + /* if we want create only exit */ + if (flags & XATTR_CREATE) { + ret = -EEXIST; + goto out; + } + + ret = btrfs_delete_one_dir_name(trans, root, path, di); + if (ret) + goto out; + btrfs_release_path(root, path); + + /* if we don't have a value then we are removing the xattr */ + if (!value) { + mod = 1; + goto out; + } + } else if (flags & XATTR_REPLACE) { + /* we couldn't find the attr to replace, so error out */ + ret = -ENODATA; + goto out; + } + + /* ok we have to create a completely new xattr */ + ret = btrfs_insert_xattr_item(trans, root, name, strlen(name), + value, size, inode->i_ino); + if (ret) + goto out; + mod = 1; + +out: + if (mod) { + inode->i_ctime = CURRENT_TIME; + ret = btrfs_update_inode(trans, root, inode); + } + + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + kfree(name); + btrfs_free_path(path); + + return ret; +} + +ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size) +{ + struct btrfs_key key, found_key; + struct inode *inode = dentry->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_path *path; + struct btrfs_item *item; + struct extent_buffer *leaf; + struct btrfs_dir_item *di; + struct xattr_handler *handler; + int ret = 0, slot, advance; + size_t total_size = 0, size_left = size, written; + unsigned long name_ptr; + char *name; + u32 nritems; + + /* + * ok we want all objects associated with this id. + * NOTE: we set key.offset = 0; because we want to start with the + * first xattr that we find and walk forward + */ + key.objectid = inode->i_ino; + btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY); + key.offset = 0; + + path = btrfs_alloc_path(); + path->reada = 2; + if (!path) + return -ENOMEM; + + mutex_lock(&root->fs_info->fs_mutex); + + /* search for our xattrs */ + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto err; + ret = 0; + advance = 0; + while (1) { + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + slot = path->slots[0]; + + /* this is where we start walking through the path */ + if (advance || slot >= nritems) { + /* + * if we've reached the last slot in this leaf we need + * to go to the next leaf and reset everything + */ + if (slot >= nritems-1) { + ret = btrfs_next_leaf(root, path); + if (ret) + break; + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + slot = path->slots[0]; + } else { + /* + * just walking through the slots on this leaf + */ + slot++; + path->slots[0]++; + } + } + advance = 1; + + item = btrfs_item_nr(leaf, slot); + btrfs_item_key_to_cpu(leaf, &found_key, slot); + + /* check to make sure this item is what we want */ + if (found_key.objectid != key.objectid) + break; + if (btrfs_key_type(&found_key) != BTRFS_XATTR_ITEM_KEY) + break; + + di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); + + total_size += btrfs_dir_name_len(leaf, di)+1; + + /* we are just looking for how big our buffer needs to be */ + if (!size) + continue; + + /* find our handler for this xattr */ + name_ptr = (unsigned long)(di + 1); + handler = find_btrfs_xattr_handler(leaf, name_ptr, + btrfs_dir_name_len(leaf, di)); + if (!handler) { + printk(KERN_ERR "btrfs: unsupported xattr found\n"); + continue; + } + + name = kmalloc(btrfs_dir_name_len(leaf, di), GFP_KERNEL); + read_extent_buffer(leaf, name, name_ptr, + btrfs_dir_name_len(leaf, di)); + + /* call the list function associated with this xattr */ + written = handler->list(inode, buffer, size_left, name, + btrfs_dir_name_len(leaf, di)); + kfree(name); + + if (written < 0) { + ret = -ERANGE; + break; + } + + size_left -= written; + buffer += written; + } + ret = total_size; + +err: + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_free_path(path); + + return ret; +} + +/* + * delete all the xattrs associated with the inode. fs_mutex should be + * held when we come into here + */ +int btrfs_delete_xattrs(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode) +{ + struct btrfs_path *path; + struct btrfs_key key, found_key; + struct btrfs_item *item; + struct extent_buffer *leaf; + int ret; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + key.objectid = inode->i_ino; + btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY); + key.offset = (u64)-1; + + while(1) { + /* look for our next xattr */ + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) + goto out; + BUG_ON(ret == 0); + + if (path->slots[0] == 0) + break; + + path->slots[0]--; + leaf = path->nodes[0]; + item = btrfs_item_nr(leaf, path->slots[0]); + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + if (found_key.objectid != key.objectid) + break; + if (btrfs_key_type(&found_key) != BTRFS_XATTR_ITEM_KEY) + break; + + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + btrfs_release_path(root, path); + } + ret = 0; +out: + btrfs_free_path(path); + + return ret; +} + +/* + * Handler functions + */ +#define BTRFS_XATTR_SETGET_FUNCS(name, index) \ +static int btrfs_xattr_##name##_get(struct inode *inode, \ + const char *name, void *value, \ + size_t size) \ +{ \ + return btrfs_xattr_get(inode, index, name, value, size); \ +} \ +static int btrfs_xattr_##name##_set(struct inode *inode, \ + const char *name, const void *value,\ + size_t size, int flags) \ +{ \ + return btrfs_xattr_set(inode, index, name, value, size, flags); \ +} \ + +BTRFS_XATTR_SETGET_FUNCS(security, BTRFS_XATTR_INDEX_SECURITY); +BTRFS_XATTR_SETGET_FUNCS(system, BTRFS_XATTR_INDEX_SYSTEM); +BTRFS_XATTR_SETGET_FUNCS(user, BTRFS_XATTR_INDEX_USER); +BTRFS_XATTR_SETGET_FUNCS(trusted, BTRFS_XATTR_INDEX_TRUSTED); + +struct xattr_handler btrfs_xattr_security_handler = { + .prefix = XATTR_SECURITY_PREFIX, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_security_get, + .set = btrfs_xattr_security_set, +}; + +struct xattr_handler btrfs_xattr_system_handler = { + .prefix = XATTR_SYSTEM_PREFIX, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_system_get, + .set = btrfs_xattr_system_set, +}; + +struct xattr_handler btrfs_xattr_user_handler = { + .prefix = XATTR_USER_PREFIX, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_user_get, + .set = btrfs_xattr_user_set, +}; + +struct xattr_handler btrfs_xattr_trusted_handler = { + .prefix = XATTR_USER_PREFIX, + .list = btrfs_xattr_generic_list, + .get = btrfs_xattr_trusted_get, + .set = btrfs_xattr_trusted_set, +}; diff --git a/fs/btrfs/xattr.h b/fs/btrfs/xattr.h new file mode 100644 index 00000000000..b2e47e3f244 --- /dev/null +++ b/fs/btrfs/xattr.h @@ -0,0 +1,58 @@ +/* + * Copyright (C) 2007 Red Hat. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __XATTR__ +#define __XATTR__ + +#include +#include "ctree.h" + +/* Name indexes */ +enum { + BTRFS_XATTR_INDEX_USER, + BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, + BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, + BTRFS_XATTR_INDEX_TRUSTED, + BTRFS_XATTR_INDEX_SECURITY, + BTRFS_XATTR_INDEX_SYSTEM, + BTRFS_XATTR_INDEX_END, +}; + +extern struct xattr_handler btrfs_xattr_user_handler; +extern struct xattr_handler btrfs_xattr_trusted_handler; +extern struct xattr_handler btrfs_xattr_acl_access_handler; +extern struct xattr_handler btrfs_xattr_acl_default_handler; +extern struct xattr_handler btrfs_xattr_security_handler; +extern struct xattr_handler btrfs_xattr_system_handler; + +extern struct xattr_handler *btrfs_xattr_handlers[]; + +ssize_t btrfs_xattr_get(struct inode *inode, int name_index, const char *name, + void *buffer, size_t size); +int btrfs_xattr_set(struct inode *inode, int name_index, const char *name, + const void *value, size_t size, int flags); + +/* + * the only reason this is public is for acl.c. There may be a point where + * acl.c doesn't need it, and if thats the case we need to remove it and make + * it static in xattr.c + */ +size_t btrfs_xattr_generic_list(struct inode *inode, char *list, + size_t list_size, const char *name, + size_t name_len); +#endif /* __XATTR__ */ -- cgit v1.2.3-70-g09d2 From 9691975dd663b255fa52404dc7de4b18892bf244 Mon Sep 17 00:00:00 2001 From: Yan Date: Tue, 4 Dec 2007 13:20:20 -0500 Subject: Btrfs: Fix buffer get/release issue in create_snapshot btrfs_cow_block expects a reference to be held on the buffer being cow'd. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 37f8e0321af..d0af0807d9f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2034,11 +2034,11 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) key.objectid = objectid; key.offset = 1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); - + extent_buffer_get(root->node); btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); + free_extent_buffer(tmp); btrfs_set_root_bytenr(&new_root_item, root->node->start); btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node)); - ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); if (ret) -- cgit v1.2.3-70-g09d2 From 7bb86316c3961d1bc401ef184fd996f999556c7f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 11 Dec 2007 09:25:06 -0500 Subject: Btrfs: Add back pointers from extents to the btree or file referencing them Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 196 +++++++++++++++++++++++--- fs/btrfs/ctree.h | 41 ++++-- fs/btrfs/disk-io.c | 2 +- fs/btrfs/extent-tree.c | 373 +++++++++++++++++++++++++++++++++++++++++-------- fs/btrfs/file.c | 20 ++- fs/btrfs/inode.c | 38 +++-- fs/btrfs/print-tree.c | 10 ++ fs/btrfs/tree-defrag.c | 2 + 8 files changed, 577 insertions(+), 105 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 5697705f753..fd8233e05cf 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -77,13 +77,37 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct extent_buffer **cow_ret, u64 search_start, u64 empty_size) { + u64 root_gen; struct extent_buffer *cow; + u32 nritems; int ret = 0; int different_trans = 0; + int level; + struct btrfs_key first_key; + + if (root->ref_cows) { + root_gen = trans->transid; + } else { + root_gen = 0; + } + WARN_ON(root->ref_cows && trans->transid != + root->fs_info->running_transaction->transid); WARN_ON(root->ref_cows && trans->transid != root->last_trans); - cow = btrfs_alloc_free_block(trans, root, buf->len, + level = btrfs_header_level(buf); + nritems = btrfs_header_nritems(buf); + if (nritems) { + if (level == 0) + btrfs_item_key_to_cpu(buf, &first_key, 0); + else + btrfs_node_key_to_cpu(buf, &first_key, 0); + } else { + first_key.objectid = 0; + } + cow = __btrfs_alloc_free_block(trans, root, buf->len, + root->root_key.objectid, + root_gen, first_key.objectid, level, search_start, empty_size); if (IS_ERR(cow)) return PTR_ERR(cow); @@ -104,14 +128,17 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, } if (buf == root->node) { + root_gen = btrfs_header_generation(buf); root->node = cow; extent_buffer_get(cow); if (buf != root->commit_root) { btrfs_free_extent(trans, root, buf->start, - buf->len, 1); + buf->len, root->root_key.objectid, + root_gen, 0, 0, 1); } free_extent_buffer(buf); } else { + root_gen = btrfs_header_generation(parent); btrfs_set_node_blockptr(parent, parent_slot, cow->start); WARN_ON(trans->transid == 0); @@ -119,7 +146,9 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, trans->transid); btrfs_mark_buffer_dirty(parent); WARN_ON(btrfs_header_generation(parent) != trans->transid); - btrfs_free_extent(trans, root, buf->start, buf->len, 1); + btrfs_free_extent(trans, root, buf->start, buf->len, + btrfs_header_owner(parent), root_gen, + 0, 0, 1); } free_extent_buffer(buf); btrfs_mark_buffer_dirty(cow); @@ -606,6 +635,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root return 0; mid = path->nodes[level]; + WARN_ON(btrfs_header_generation(mid) != trans->transid); + orig_ptr = btrfs_node_blockptr(mid, orig_slot); if (level < BTRFS_MAX_LEVEL - 1) @@ -631,7 +662,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root wait_on_tree_block_writeback(root, mid); /* once for the path */ free_extent_buffer(mid); - ret = btrfs_free_extent(trans, root, mid->start, mid->len, 1); + ret = btrfs_free_extent(trans, root, mid->start, mid->len, + root->root_key.objectid, + btrfs_header_generation(mid), 0, 0, 1); /* once for the root ptr */ free_extent_buffer(mid); return ret; @@ -681,6 +714,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root ret = wret; if (btrfs_header_nritems(right) == 0) { u64 bytenr = right->start; + u64 generation = btrfs_header_generation(parent); u32 blocksize = right->len; clean_tree_block(trans, root, right); @@ -692,7 +726,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root if (wret) ret = wret; wret = btrfs_free_extent(trans, root, bytenr, - blocksize, 1); + blocksize, + btrfs_header_owner(parent), + generation, 0, 0, 1); if (wret) ret = wret; } else { @@ -722,6 +758,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root } if (btrfs_header_nritems(mid) == 0) { /* we've managed to empty the middle node, drop it */ + u64 root_gen = btrfs_header_generation(parent); u64 bytenr = mid->start; u32 blocksize = mid->len; clean_tree_block(trans, root, mid); @@ -731,7 +768,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root wret = del_ptr(trans, root, path, level + 1, pslot); if (wret) ret = wret; - wret = btrfs_free_extent(trans, root, bytenr, blocksize, 1); + wret = btrfs_free_extent(trans, root, bytenr, blocksize, + btrfs_header_owner(parent), + root_gen, 0, 0, 1); if (wret) ret = wret; } else { @@ -788,6 +827,7 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans, return 1; mid = path->nodes[level]; + WARN_ON(btrfs_header_generation(mid) != trans->transid); orig_ptr = btrfs_node_blockptr(mid, orig_slot); if (level < BTRFS_MAX_LEVEL - 1) @@ -1113,6 +1153,8 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root src_nritems = btrfs_header_nritems(src); dst_nritems = btrfs_header_nritems(dst); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; + WARN_ON(btrfs_header_generation(src) != trans->transid); + WARN_ON(btrfs_header_generation(dst) != trans->transid); if (push_items <= 0) { return 1; @@ -1159,6 +1201,9 @@ static int balance_node_right(struct btrfs_trans_handle *trans, int dst_nritems; int ret = 0; + WARN_ON(btrfs_header_generation(src) != trans->transid); + WARN_ON(btrfs_header_generation(dst) != trans->transid); + src_nritems = btrfs_header_nritems(src); dst_nritems = btrfs_header_nritems(dst); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; @@ -1202,6 +1247,8 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { + u64 root_gen; + u64 lower_gen; struct extent_buffer *lower; struct extent_buffer *c; struct btrfs_disk_key lower_key; @@ -1209,7 +1256,20 @@ static int insert_new_root(struct btrfs_trans_handle *trans, BUG_ON(path->nodes[level]); BUG_ON(path->nodes[level-1] != root->node); - c = btrfs_alloc_free_block(trans, root, root->nodesize, + if (root->ref_cows) + root_gen = trans->transid; + else + root_gen = 0; + + lower = path->nodes[level-1]; + if (level == 1) + btrfs_item_key(lower, &lower_key, 0); + else + btrfs_node_key(lower, &lower_key, 0); + + c = __btrfs_alloc_free_block(trans, root, root->nodesize, + root->root_key.objectid, + root_gen, lower_key.objectid, level, root->node->start, 0); if (IS_ERR(c)) return PTR_ERR(c); @@ -1219,19 +1279,16 @@ static int insert_new_root(struct btrfs_trans_handle *trans, btrfs_set_header_bytenr(c, c->start); btrfs_set_header_generation(c, trans->transid); btrfs_set_header_owner(c, root->root_key.objectid); - lower = path->nodes[level-1]; write_extent_buffer(c, root->fs_info->fsid, (unsigned long)btrfs_header_fsid(c), BTRFS_FSID_SIZE); - if (level == 1) - btrfs_item_key(lower, &lower_key, 0); - else - btrfs_node_key(lower, &lower_key, 0); btrfs_set_node_key(c, &lower_key, 0); btrfs_set_node_blockptr(c, 0, lower->start); - WARN_ON(btrfs_header_generation(lower) == 0); - btrfs_set_node_ptr_generation(c, 0, btrfs_header_generation(lower)); + lower_gen = btrfs_header_generation(lower); + WARN_ON(lower_gen == 0); + + btrfs_set_node_ptr_generation(c, 0, lower_gen); btrfs_mark_buffer_dirty(c); @@ -1241,6 +1298,18 @@ static int insert_new_root(struct btrfs_trans_handle *trans, extent_buffer_get(c); path->nodes[level] = c; path->slots[level] = 0; + + if (root->ref_cows && lower_gen != trans->transid) { + struct btrfs_path *back_path = btrfs_alloc_path(); + int ret; + ret = btrfs_insert_extent_backref(trans, + root->fs_info->extent_root, + path, lower->start, + root->root_key.objectid, + trans->transid, 0, 0); + BUG_ON(ret); + btrfs_free_path(back_path); + } return 0; } @@ -1294,6 +1363,7 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { + u64 root_gen; struct extent_buffer *c; struct extent_buffer *split; struct btrfs_disk_key disk_key; @@ -1303,6 +1373,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root u32 c_nritems; c = path->nodes[level]; + WARN_ON(btrfs_header_generation(c) != trans->transid); if (c == root->node) { /* trying to split the root, lets make a new one */ ret = insert_new_root(trans, root, path, level + 1); @@ -1319,8 +1390,17 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root } c_nritems = btrfs_header_nritems(c); - split = btrfs_alloc_free_block(trans, root, root->nodesize, - c->start, 0); + if (root->ref_cows) + root_gen = trans->transid; + else + root_gen = 0; + + btrfs_node_key(c, &disk_key, 0); + split = __btrfs_alloc_free_block(trans, root, root->nodesize, + root->root_key.objectid, + root_gen, + btrfs_disk_key_objectid(&disk_key), + level, c->start, 0); if (IS_ERR(split)) return PTR_ERR(split); @@ -1789,6 +1869,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *ins_key, struct btrfs_path *path, int data_size, int extend) { + u64 root_gen; struct extent_buffer *l; u32 nritems; int mid; @@ -1807,6 +1888,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root if (extend) space_needed = data_size; + if (root->ref_cows) + root_gen = trans->transid; + else + root_gen = 0; + /* first try to make some room by pushing left and right */ if (ins_key->type != BTRFS_DIR_ITEM_KEY) { wret = push_leaf_right(trans, root, path, data_size, 0); @@ -1837,8 +1923,12 @@ again: nritems = btrfs_header_nritems(l); mid = (nritems + 1)/ 2; - right = btrfs_alloc_free_block(trans, root, root->leafsize, - l->start, 0); + btrfs_item_key(l, &disk_key, 0); + + right = __btrfs_alloc_free_block(trans, root, root->leafsize, + root->root_key.objectid, + root_gen, disk_key.objectid, 0, + l->start, 0); if (IS_ERR(right)) return PTR_ERR(right); @@ -2413,13 +2503,16 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (leaf == root->node) { btrfs_set_header_level(leaf, 0); } else { + u64 root_gen = btrfs_header_generation(path->nodes[1]); clean_tree_block(trans, root, leaf); wait_on_tree_block_writeback(root, leaf); wret = del_ptr(trans, root, path, 1, path->slots[1]); if (wret) ret = wret; wret = btrfs_free_extent(trans, root, - leaf->start, leaf->len, 1); + leaf->start, leaf->len, + btrfs_header_owner(path->nodes[1]), + root_gen, 0, 0, 1); if (wret) ret = wret; } @@ -2456,9 +2549,13 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, } if (btrfs_header_nritems(leaf) == 0) { + u64 root_gen; u64 bytenr = leaf->start; u32 blocksize = leaf->len; + root_gen = btrfs_header_generation( + path->nodes[1]); + clean_tree_block(trans, root, leaf); wait_on_tree_block_writeback(root, leaf); @@ -2468,7 +2565,9 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, free_extent_buffer(leaf); wret = btrfs_free_extent(trans, root, bytenr, - blocksize, 1); + blocksize, + btrfs_header_owner(path->nodes[1]), + root_gen, 0, 0, 1); if (wret) ret = wret; } else { @@ -2482,6 +2581,61 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, return ret; } +/* + * walk up the tree as far as required to find the previous leaf. + * returns 0 if it found something or 1 if there are no lesser leaves. + * returns < 0 on io errors. + */ +int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) +{ + int slot; + int level = 1; + u64 bytenr; + struct extent_buffer *c; + struct extent_buffer *next = NULL; + + while(level < BTRFS_MAX_LEVEL) { + if (!path->nodes[level]) + return 1; + + slot = path->slots[level]; + c = path->nodes[level]; + if (slot == 0) { + level++; + if (level == BTRFS_MAX_LEVEL) + return 1; + continue; + } + slot--; + + bytenr = btrfs_node_blockptr(c, slot); + if (next) + free_extent_buffer(next); + + if (path->reada < 0) + reada_for_search(root, path, level, slot); + + next = read_tree_block(root, bytenr, + btrfs_level_size(root, level - 1)); + break; + } + path->slots[level] = slot; + while(1) { + level--; + c = path->nodes[level]; + free_extent_buffer(c); + path->nodes[level] = next; + path->slots[level] = 0; + if (!level) + break; + if (path->reada) + reada_for_search(root, path, level, 0); + next = read_tree_block(root, btrfs_node_blockptr(next, 0), + btrfs_level_size(root, level - 1)); + } + return 0; +} + /* * walk up the tree as far as required to find the next leaf. * returns 0 if it found something or 1 if there are no greater leaves. @@ -2503,6 +2657,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) c = path->nodes[level]; if (slot >= btrfs_header_nritems(c)) { level++; + if (level == BTRFS_MAX_LEVEL) + return 1; continue; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index fd58dd846e6..cb1b156d954 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -544,11 +544,12 @@ BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64); BTRFS_SETGET_FUNCS(ref_objectid, struct btrfs_extent_ref, objectid, 64); BTRFS_SETGET_FUNCS(ref_offset, struct btrfs_extent_ref, offset, 64); -BTRFS_SETGET_STACK_FUNCS(ref_root, struct btrfs_extent_ref, root, 64); -BTRFS_SETGET_STACK_FUNCS(ref_generation, struct btrfs_extent_ref, +BTRFS_SETGET_STACK_FUNCS(stack_ref_root, struct btrfs_extent_ref, root, 64); +BTRFS_SETGET_STACK_FUNCS(stack_ref_generation, struct btrfs_extent_ref, generation, 64); -BTRFS_SETGET_STACK_FUNCS(ref_objectid, struct btrfs_extent_ref, objectid, 64); -BTRFS_SETGET_STACK_FUNCS(ref_offset, struct btrfs_extent_ref, offset, 64); +BTRFS_SETGET_STACK_FUNCS(stack_ref_objectid, struct btrfs_extent_ref, + objectid, 64); +BTRFS_SETGET_STACK_FUNCS(stack_ref_offset, struct btrfs_extent_ref, offset, 64); BTRFS_SETGET_STACK_FUNCS(stack_extent_refs, struct btrfs_extent_item, refs, 32); @@ -914,24 +915,45 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, *hint, u64 search_start, int data, int owner); int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root); + struct btrfs_root *root, u64 owner_objectid); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u32 size, + u64 root_objectid, u64 hint, u64 empty_size); +struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u32 blocksize, + u64 root_objectid, + u64 ref_generation, + u64 first_objectid, + int level, + u64 hint, + u64 empty_size); +int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 owner, - u64 num_bytes, u64 empty_size, u64 search_start, + struct btrfs_root *root, + u64 num_bytes, u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, int data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, int pin); + *root, u64 bytenr, u64 num_bytes, + u64 root_objectid, u64 ref_generation, + u64 owner_objectid, u64 owner_offset, int pin); int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_map_tree *unpin); int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 bytenr, u64 num_bytes); + u64 bytenr, u64 num_bytes, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset); int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_free_block_groups(struct btrfs_fs_info *info); @@ -966,6 +988,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *cpu_key, u32 data_size); int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path); +int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path); int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf); int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root *root); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 60a30da6af0..0ac21e3aac8 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -210,7 +210,7 @@ static int btree_writepages(struct address_space *mapping, { struct extent_map_tree *tree; tree = &BTRFS_I(mapping->host)->extent_tree; - if (wbc->sync_mode == WB_SYNC_NONE) { + if (0 && wbc->sync_mode == WB_SYNC_NONE) { u64 num_dirty; u64 start = 0; unsigned long thresh = 96 * 1024 * 1024; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 0f1ebdd4e92..32991f73e9d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -17,6 +17,7 @@ */ #include +#include #include "hash.h" #include "ctree.h" #include "disk-io.h" @@ -89,7 +90,8 @@ static int cache_block_group(struct btrfs_root *root, btrfs_item_key_to_cpu(leaf, &key, slot); if (key.objectid < block_group->key.objectid) { - if (key.objectid + key.offset > first_free) + if (btrfs_key_type(&key) != BTRFS_EXTENT_REF_KEY && + key.objectid + key.offset > first_free) first_free = key.objectid + key.offset; goto next; } @@ -353,7 +355,7 @@ found: return found_group; } -static u64 hash_extent_ref(u64 root_objectid, u64 root_generation, +static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset) { u32 high_crc = ~(u32)0; @@ -362,53 +364,149 @@ static u64 hash_extent_ref(u64 root_objectid, u64 root_generation, lenum = cpu_to_le64(root_objectid); high_crc = crc32c(high_crc, &lenum, sizeof(lenum)); - lenum = cpu_to_le64(root_generation); - high_crc = crc32c(high_crc, &lenum, sizeof(lenum)); + lenum = cpu_to_le64(ref_generation); + low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); +#if 0 lenum = cpu_to_le64(owner); low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); - lenum = cpu_to_le64(owner_offset); low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); - +#endif return ((u64)high_crc << 32) | (u64)low_crc; } -int insert_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, - u64 bytenr, - u64 root_objectid, u64 root_generation, - u64 owner, u64 owner_offset) +static int match_extent_ref(struct extent_buffer *leaf, + struct btrfs_extent_ref *disk_ref, + struct btrfs_extent_ref *cpu_ref) +{ + int ret; + int len; + + if (cpu_ref->objectid) + len = sizeof(*cpu_ref); + else + len = 2 * sizeof(u64); + ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, + len); + return ret == 0; +} + +static int lookup_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, int del) { u64 hash; struct btrfs_key key; + struct btrfs_key found_key; struct btrfs_extent_ref ref; - struct extent_buffer *l; - struct btrfs_extent_item *item; + struct extent_buffer *leaf; + struct btrfs_extent_ref *disk_ref; + int ret; + int ret2; + + btrfs_set_stack_ref_root(&ref, root_objectid); + btrfs_set_stack_ref_generation(&ref, ref_generation); + btrfs_set_stack_ref_objectid(&ref, owner); + btrfs_set_stack_ref_offset(&ref, owner_offset); + + hash = hash_extent_ref(root_objectid, ref_generation, owner, + owner_offset); + key.offset = hash; + key.objectid = bytenr; + key.type = BTRFS_EXTENT_REF_KEY; + + while (1) { + ret = btrfs_search_slot(trans, root, &key, path, + del ? -1 : 0, del); + if (ret < 0) + goto out; + leaf = path->nodes[0]; + if (ret != 0) { + u32 nritems = btrfs_header_nritems(leaf); + if (path->slots[0] >= nritems) { + ret2 = btrfs_next_leaf(root, path); + if (ret2) + goto out; + leaf = path->nodes[0]; + } + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.objectid != bytenr || + found_key.type != BTRFS_EXTENT_REF_KEY) + goto out; + key.offset = found_key.offset; + if (del) { + btrfs_release_path(root, path); + continue; + } + } + disk_ref = btrfs_item_ptr(path->nodes[0], + path->slots[0], + struct btrfs_extent_ref); + if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { + ret = 0; + goto out; + } + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + key.offset = found_key.offset + 1; + btrfs_release_path(root, path); + } +out: + return ret; +} + +int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset) +{ + u64 hash; + struct btrfs_key key; + struct btrfs_extent_ref ref; + struct btrfs_extent_ref *disk_ref; int ret; btrfs_set_stack_ref_root(&ref, root_objectid); - btrfs_set_stack_ref_generation(&ref, root_generation); + btrfs_set_stack_ref_generation(&ref, ref_generation); btrfs_set_stack_ref_objectid(&ref, owner); btrfs_set_stack_ref_offset(&ref, owner_offset); - ret = btrfs_name_hash(&ref, sizeof(ref), &hash); + hash = hash_extent_ref(root_objectid, ref_generation, owner, + owner_offset); key.offset = hash; key.objectid = bytenr; key.type = BTRFS_EXTENT_REF_KEY; ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); while (ret == -EEXIST) { - + disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_extent_ref); + if (match_extent_ref(path->nodes[0], disk_ref, &ref)) + goto out; + key.offset++; + btrfs_release_path(root, path); + ret = btrfs_insert_empty_item(trans, root, path, &key, + sizeof(ref)); } - + if (ret) + goto out; + disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_extent_ref); + write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, + sizeof(ref)); + btrfs_mark_buffer_dirty(path->nodes[0]); +out: + btrfs_release_path(root, path); + return ret; } int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, - u64 root_objectid, u64 root_generation, + u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset) { struct btrfs_path *path; @@ -441,6 +539,11 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_release_path(root->fs_info->extent_root, path); + + ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, + path, bytenr, root_objectid, + ref_generation, owner, owner_offset); + BUG_ON(ret); finish_current_insert(trans, root->fs_info->extent_root); del_pending_extents(trans, root->fs_info->extent_root); @@ -489,10 +592,29 @@ out: } int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root) + struct btrfs_root *root, u64 owner_objectid) { + u64 generation; + u64 key_objectid; + u64 level; + u32 nritems; + struct btrfs_disk_key disk_key; + + level = btrfs_header_level(root->node); + generation = trans->transid; + nritems = btrfs_header_nritems(root->node); + if (nritems > 0) { + if (level == 0) + btrfs_item_key(root->node, &disk_key, 0); + else + btrfs_node_key(root->node, &disk_key, 0); + key_objectid = btrfs_disk_key_objectid(&disk_key); + } else { + key_objectid = 0; + } return btrfs_inc_extent_ref(trans, root, root->node->start, - root->node->len); + root->node->len, owner_objectid, + generation, 0, 0); } int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -506,7 +628,6 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, int level; int ret; int faili; - int err; if (!root->ref_cows) return 0; @@ -528,7 +649,9 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (disk_bytenr == 0) continue; ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, - btrfs_file_extent_disk_num_bytes(buf, fi)); + btrfs_file_extent_disk_num_bytes(buf, fi), + root->root_key.objectid, trans->transid, + key.objectid, key.offset); if (ret) { faili = i; goto fail; @@ -536,7 +659,9 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, } else { bytenr = btrfs_node_blockptr(buf, i); ret = btrfs_inc_extent_ref(trans, root, bytenr, - btrfs_level_size(root, level - 1)); + btrfs_level_size(root, level - 1), + root->root_key.objectid, + trans->transid, 0, 0); if (ret) { faili = i; goto fail; @@ -546,6 +671,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, return 0; fail: WARN_ON(1); +#if 0 for (i =0; i < faili; i++) { if (level == 0) { u64 disk_bytenr; @@ -571,6 +697,7 @@ fail: BUG_ON(err); } } +#endif return ret; } @@ -809,18 +936,18 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root) { + u64 start; + u64 end; + struct btrfs_fs_info *info = extent_root->fs_info; + struct btrfs_path *path; struct btrfs_key ins; struct btrfs_extent_item extent_item; int ret; int err = 0; - u64 start; - u64 end; - struct btrfs_fs_info *info = extent_root->fs_info; btrfs_set_stack_extent_refs(&extent_item, 1); btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); - btrfs_set_stack_extent_owner(&extent_item, - extent_root->root_key.objectid); + path = btrfs_alloc_path(); while(1) { ret = find_first_extent_bit(&info->extent_ins, 0, &start, @@ -834,7 +961,12 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct &extent_item, sizeof(extent_item)); clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, GFP_NOFS); + err = btrfs_insert_extent_backref(trans, extent_root, path, + start, extent_root->root_key.objectid, + 0, 0, 0); + BUG_ON(err); } + btrfs_free_path(path); return 0; } @@ -871,7 +1003,9 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, * remove an extent from the root, returns 0 on success */ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, int pin, + *root, u64 bytenr, u64 num_bytes, + u64 root_objectid, u64 ref_generation, + u64 owner_objectid, u64 owner_offset, int pin, int mark_free) { struct btrfs_path *path; @@ -891,6 +1025,24 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root if (!path) return -ENOMEM; + if (ref_generation && owner_objectid == 0 && root_objectid == 3) { +//printk("drop backref root %Lu gen %Lu byte %Lu\n", root_objectid, ref_generation, bytenr ); + } + ret = lookup_extent_backref(trans, extent_root, path, + bytenr, root_objectid, + ref_generation, + owner_objectid, owner_offset, 1); + if (ret == 0) { + ret = btrfs_del_item(trans, extent_root, path); + } else { + btrfs_print_leaf(extent_root, path->nodes[0]); + WARN_ON(1); + printk("Unable to find ref byte nr %Lu root %Lu " + " gen %Lu owner %Lu offset %Lu\n", bytenr, + root_objectid, ref_generation, owner_objectid, + owner_offset); + } + btrfs_release_path(extent_root, path); ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); if (ret < 0) return ret; @@ -965,7 +1117,9 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, GFP_NOFS); ret = __free_extent(trans, extent_root, - start, end + 1 - start, 0, 0); + start, end + 1 - start, + extent_root->root_key.objectid, + 0, 0, 0, 0, 0); if (ret) err = ret; } @@ -976,18 +1130,25 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct * remove an extent from the root, returns 0 on success */ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, int pin) + *root, u64 bytenr, u64 num_bytes, + u64 root_objectid, u64 ref_generation, + u64 owner_objectid, u64 owner_offset, int pin) { struct btrfs_root *extent_root = root->fs_info->extent_root; int pending_ret; int ret; WARN_ON(num_bytes < root->sectorsize); + if (!root->ref_cows) + ref_generation = 0; + if (root == extent_root) { pin_down_bytes(root, bytenr, num_bytes, 1); return 0; } - ret = __free_extent(trans, root, bytenr, num_bytes, pin, pin == 0); + ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, + ref_generation, owner_objectid, owner_offset, + pin, pin == 0); pending_ret = del_pending_extents(trans, root->fs_info->extent_root); return ret ? ret : pending_ret; } @@ -1080,23 +1241,26 @@ check_failed: btrfs_item_key_to_cpu(l, &key, path->slots[0]); /* - * a rare case, go back one key if we hit a block group item - * instead of an extent item + * walk backwards to find the first extent item key */ - if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY && - key.objectid + key.offset >= search_start) { - ins->objectid = key.objectid; - ins->offset = key.offset - 1; - btrfs_release_path(root, path); - ret = btrfs_search_slot(trans, root, ins, path, 0, 0); - if (ret < 0) - goto error; - - if (path->slots[0] > 0) { + while(btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) { + if (path->slots[0] == 0) { + ret = btrfs_prev_leaf(root, path); + if (ret != 0) { + ret = btrfs_search_slot(trans, root, ins, + path, 0, 0); + if (ret < 0) + goto error; + if (path->slots[0] > 0) + path->slots[0]--; + break; + } + } else { path->slots[0]--; } + l = path->nodes[0]; + btrfs_item_key_to_cpu(l, &key, path->slots[0]); } - while (1) { l = path->nodes[0]; slot = path->slots[0]; @@ -1146,7 +1310,8 @@ check_failed: } } if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) { - if (!start_found) { + if (!start_found && btrfs_key_type(&key) == + BTRFS_BLOCK_GROUP_ITEM_KEY) { last_byte = key.objectid; start_found = 1; } @@ -1244,8 +1409,10 @@ error: * returns 0 if everything worked, non-zero otherwise. */ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 owner, - u64 num_bytes, u64 empty_size, u64 hint_byte, + struct btrfs_root *root, + u64 num_bytes, u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, int data) { int ret; @@ -1255,9 +1422,9 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; struct btrfs_extent_item extent_item; + struct btrfs_path *path; btrfs_set_stack_extent_refs(&extent_item, 1); - btrfs_set_stack_extent_owner(&extent_item, owner); WARN_ON(num_bytes < root->sectorsize); ret = find_free_extent(trans, root, num_bytes, empty_size, @@ -1296,8 +1463,16 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, trans->alloc_exclude_start = 0; trans->alloc_exclude_nr = 0; + BUG_ON(ret); + + path = btrfs_alloc_path(); + BUG_ON(!path); + ret = btrfs_insert_extent_backref(trans, extent_root, path, + ins->objectid, root_objectid, + ref_generation, owner, owner_offset); BUG_ON(ret); + btrfs_free_path(path); finish_current_insert(trans, extent_root); pending_ret = del_pending_extents(trans, extent_root); @@ -1321,15 +1496,43 @@ update_block: */ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u32 blocksize, u64 hint, + u32 blocksize, + u64 root_objectid, u64 hint, + u64 empty_size) +{ + u64 ref_generation; + + if (root->ref_cows) + ref_generation = trans->transid; + else + ref_generation = 0; + + + return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, + ref_generation, 0, 0, hint, empty_size); +} + +/* + * helper function to allocate a block for a given tree + * returns the tree buffer or NULL. + */ +struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u32 blocksize, + u64 root_objectid, + u64 ref_generation, + u64 first_objectid, + int level, + u64 hint, u64 empty_size) { struct btrfs_key ins; int ret; struct extent_buffer *buf; - ret = btrfs_alloc_extent(trans, root, root->root_key.objectid, - blocksize, empty_size, hint, + ret = btrfs_alloc_extent(trans, root, blocksize, + root_objectid, ref_generation, + first_objectid, level, empty_size, hint, (u64)-1, &ins, 0); if (ret) { BUG_ON(ret > 0); @@ -1337,7 +1540,9 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, } buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); if (!buf) { - btrfs_free_extent(trans, root, ins.objectid, blocksize, 0); + btrfs_free_extent(trans, root, ins.objectid, blocksize, + root->root_key.objectid, ref_generation, + 0, 0, 0); return ERR_PTR(-ENOMEM); } btrfs_set_buffer_uptodate(buf); @@ -1355,6 +1560,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, static int drop_leaf_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *leaf) { + u64 leaf_owner; + u64 leaf_generation; struct btrfs_key key; struct btrfs_file_extent_item *fi; int i; @@ -1363,6 +1570,9 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans, BUG_ON(!btrfs_is_leaf(leaf)); nritems = btrfs_header_nritems(leaf); + leaf_owner = btrfs_header_owner(leaf); + leaf_generation = btrfs_header_generation(leaf); + for (i = 0; i < nritems; i++) { u64 disk_bytenr; @@ -1381,7 +1591,9 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans, if (disk_bytenr == 0) continue; ret = btrfs_free_extent(trans, root, disk_bytenr, - btrfs_file_extent_disk_num_bytes(leaf, fi), 0); + btrfs_file_extent_disk_num_bytes(leaf, fi), + leaf_owner, leaf_generation, + key.objectid, key.offset, 0); BUG_ON(ret); } return 0; @@ -1423,9 +1635,12 @@ static void reada_walk_down(struct btrfs_root *root, static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level) { + u64 root_owner; + u64 root_gen; + u64 bytenr; struct extent_buffer *next; struct extent_buffer *cur; - u64 bytenr; + struct extent_buffer *parent; u32 blocksize; int ret; u32 refs; @@ -1466,9 +1681,13 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); BUG_ON(ret); if (refs != 1) { + parent = path->nodes[*level]; + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); path->slots[*level]++; ret = btrfs_free_extent(trans, root, bytenr, - blocksize, 1); + blocksize, root_owner, + root_gen, 0, 0, 1); BUG_ON(ret); continue; } @@ -1484,10 +1703,16 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root blocksize, &refs); BUG_ON(ret); if (refs != 1) { + parent = path->nodes[*level]; + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); + path->slots[*level]++; free_extent_buffer(next); - ret = btrfs_free_extent(trans, root, - bytenr, blocksize, 1); + ret = btrfs_free_extent(trans, root, bytenr, + blocksize, + root_owner, + root_gen, 0, 0, 1); BUG_ON(ret); continue; } @@ -1502,8 +1727,19 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root out: WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); + + if (path->nodes[*level] == root->node) { + root_owner = root->root_key.objectid; + parent = path->nodes[*level]; + } else { + parent = path->nodes[*level + 1]; + root_owner = btrfs_header_owner(parent); + } + + root_gen = btrfs_header_generation(parent); ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, - path->nodes[*level]->len, 1); + path->nodes[*level]->len, + root_owner, root_gen, 0, 0, 1); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; @@ -1519,10 +1755,12 @@ out: static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level) { + u64 root_owner; + u64 root_gen; + struct btrfs_root_item *root_item = &root->root_item; int i; int slot; int ret; - struct btrfs_root_item *root_item = &root->root_item; for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { slot = path->slots[i]; @@ -1539,9 +1777,20 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root root_item->drop_level = i; return 0; } else { + if (path->nodes[*level] == root->node) { + root_owner = root->root_key.objectid; + root_gen = + btrfs_header_generation(path->nodes[*level]); + } else { + struct extent_buffer *node; + node = path->nodes[*level + 1]; + root_owner = btrfs_header_owner(node); + root_gen = btrfs_header_generation(node); + } ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, - path->nodes[*level]->len, 1); + path->nodes[*level]->len, + root_owner, root_gen, 0, 0, 1); BUG_ON(ret); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index b0d63778726..5b1f90f06e0 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -496,7 +496,10 @@ next_slot: sizeof(old)); if (disk_bytenr != 0) { ret = btrfs_inc_extent_ref(trans, root, - disk_bytenr, disk_num_bytes); + disk_bytenr, disk_num_bytes, + root->root_key.objectid, + trans->transid, + key.objectid, end); BUG_ON(ret); } } @@ -541,6 +544,14 @@ next_slot: u64 disk_bytenr = 0; u64 disk_num_bytes = 0; u64 extent_num_bytes = 0; + u64 root_gen; + + if (leaf != root->node) { + root_gen = + btrfs_header_generation(path->nodes[1]); + } else { + root_gen = btrfs_header_generation(leaf); + } if (found_extent) { disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, @@ -562,8 +573,11 @@ next_slot: if (found_extent && disk_bytenr != 0) { inode->i_blocks -= extent_num_bytes >> 9; ret = btrfs_free_extent(trans, root, - disk_bytenr, - disk_num_bytes, 0); + disk_bytenr, + disk_num_bytes, + root->root_key.objectid, + root_gen, inode->i_ino, + key.offset, 0); } BUG_ON(ret); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d0af0807d9f..bb70db0c9df 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -93,7 +93,9 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) if (alloc_hint == EXTENT_MAP_INLINE) goto out; - ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0, + ret = btrfs_alloc_extent(trans, root, num_bytes, + root->root_key.objectid, trans->transid, + inode->i_ino, start, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { WARN_ON(1); @@ -560,6 +562,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, u64 extent_start = 0; u64 extent_num_bytes = 0; u64 item_end = 0; + u64 root_gen = 0; int found_extent; int del_item; int extent_type = -1; @@ -670,6 +673,15 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, found_extent = 1; inode->i_blocks -= num_dec; } + if (leaf == root->node) { + root_gen = + btrfs_header_generation(leaf); + } else { + struct extent_buffer *parent; + parent = path->nodes[1]; + root_gen = + btrfs_header_generation(parent); + } } } else if (extent_type == BTRFS_FILE_EXTENT_INLINE && !del_item) { @@ -690,7 +702,10 @@ delete: btrfs_release_path(root, path); if (found_extent) { ret = btrfs_free_extent(trans, root, extent_start, - extent_num_bytes, 0); + extent_num_bytes, + root->root_key.objectid, + root_gen, inode->i_ino, + found_key.offset, 0); BUG_ON(ret); } } @@ -1900,7 +1915,14 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0); + ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, + 0, &objectid); + if (ret) + goto fail; + + leaf = __btrfs_alloc_free_block(trans, root, root->leafsize, + objectid, trans->transid, 0, 0, + 0, 0); if (IS_ERR(leaf)) return PTR_ERR(leaf); @@ -1908,7 +1930,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) btrfs_set_header_level(leaf, 0); btrfs_set_header_bytenr(leaf, leaf->start); btrfs_set_header_generation(leaf, trans->transid); - btrfs_set_header_owner(leaf, root->root_key.objectid); + btrfs_set_header_owner(leaf, objectid); + write_extent_buffer(leaf, root->fs_info->fsid, (unsigned long)btrfs_header_fsid(leaf), BTRFS_FSID_SIZE); @@ -1933,11 +1956,6 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) free_extent_buffer(leaf); leaf = NULL; - ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, - 0, &objectid); - if (ret) - goto fail; - btrfs_set_root_dirid(&root_item, new_dirid); key.objectid = objectid; @@ -2056,7 +2074,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; - ret = btrfs_inc_root_ref(trans, root); + ret = btrfs_inc_root_ref(trans, root, objectid); if (ret) goto fail; fail: diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 030324febf6..da0b4dcf361 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -33,6 +33,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) struct btrfs_file_extent_item *fi; struct btrfs_key key; struct btrfs_key found_key; + struct btrfs_extent_ref *ref; u32 type; printk("leaf %llu total ptrs %d free space %d\n", @@ -73,6 +74,15 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) printk("\t\textent data refs %u\n", btrfs_extent_refs(l, ei)); break; + case BTRFS_EXTENT_REF_KEY: + ref = btrfs_item_ptr(l, i, struct btrfs_extent_ref); + printk("\t\textent back ref root %llu gen %llu " + "owner %llu offset %llu\n", + (unsigned long long)btrfs_ref_root(l, ref), + (unsigned long long)btrfs_ref_generation(l, ref), + (unsigned long long)btrfs_ref_objectid(l, ref), + (unsigned long long)btrfs_ref_offset(l, ref)); + break; case BTRFS_EXTENT_DATA_KEY: fi = btrfs_item_ptr(l, i, diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 3994795edfe..5c58630dce0 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -78,6 +78,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, break; if (*level == 1) { + WARN_ON(btrfs_header_generation(path->nodes[*level]) != + trans->transid); ret = btrfs_realloc_node(trans, root, path->nodes[*level], path->slots[*level], -- cgit v1.2.3-70-g09d2 From d8d5f3e16d1ae4fe9b93312e083f2c04a95520f0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 11 Dec 2007 12:42:00 -0500 Subject: Btrfs: Add lowest key information to back refs for extent tree blocks as well. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 2 +- fs/btrfs/extent-tree.c | 102 ++++++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/file.c | 11 ++---- fs/btrfs/inode.c | 14 ++----- 4 files changed, 110 insertions(+), 19 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 0ac21e3aac8..60a30da6af0 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -210,7 +210,7 @@ static int btree_writepages(struct address_space *mapping, { struct extent_map_tree *tree; tree = &BTRFS_I(mapping->host)->extent_tree; - if (0 && wbc->sync_mode == WB_SYNC_NONE) { + if (wbc->sync_mode == WB_SYNC_NONE) { u64 num_dirty; u64 start = 0; unsigned long thresh = 96 * 1024 * 1024; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 32991f73e9d..187be401247 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -457,6 +457,94 @@ out: return ret; } +/* + * Back reference rules. Back refs have three main goals: + * + * 1) differentiate between all holders of references to an extent so that + * when a reference is dropped we can make sure it was a valid reference + * before freeing the extent. + * + * 2) Provide enough information to quickly find the holders of an extent + * if we notice a given block is corrupted or bad. + * + * 3) Make it easy to migrate blocks for FS shrinking or storage pool + * maintenance. This is actually the same as #2, but with a slightly + * different use case. + * + * File extents can be referenced by: + * + * - multiple snapshots, subvolumes, or different generations in one subvol + * - different files inside a single subvolume (in theory, not implemented yet) + * - different offsets inside a file (bookend extents in file.c) + * + * The extent ref structure has fields for: + * + * - Objectid of the subvolume root + * - Generation number of the tree holding the reference + * - objectid of the file holding the reference + * - offset in the file corresponding to the key holding the reference + * + * When a file extent is allocated the fields are filled in: + * (root_key.objectid, trans->transid, inode objectid, offset in file) + * + * When a leaf is cow'd new references are added for every file extent found + * in the leaf. It looks the same as the create case, but trans->transid + * will be different when the block is cow'd. + * + * (root_key.objectid, trans->transid, inode objectid, offset in file) + * + * When a file extent is removed either during snapshot deletion or file + * truncation, the corresponding back reference is found + * by searching for: + * + * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), + * inode objectid, offset in file) + * + * Btree extents can be referenced by: + * + * - Different subvolumes + * - Different generations of the same subvolume + * + * Storing sufficient information for a full reverse mapping of a btree + * block would require storing the lowest key of the block in the backref, + * and it would require updating that lowest key either before write out or + * every time it changed. Instead, the objectid of the lowest key is stored + * along with the level of the tree block. This provides a hint + * about where in the btree the block can be found. Searches through the + * btree only need to look for a pointer to that block, so they stop one + * level higher than the level recorded in the backref. + * + * Some btrees do not do reference counting on their extents. These + * include the extent tree and the tree of tree roots. Backrefs for these + * trees always have a generation of zero. + * + * When a tree block is created, back references are inserted: + * + * (root->root_key.objectid, trans->transid or zero, lowest_key_objectid, level) + * + * When a tree block is cow'd in a reference counted root, + * new back references are added for all the blocks it points to. + * These are of the form (trans->transid will have increased since creation): + * + * (root->root_key.objectid, trans->transid, lowest_key_objectid, level) + * + * Because the lowest_key_objectid and the level are just hints + * they are not used when backrefs are deleted. When a backref is deleted: + * + * if backref was for a tree root: + * root_objectid = root->root_key.objectid + * else + * root_objectid = btrfs_header_owner(parent) + * + * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) + * + * Back Reference Key hashing: + * + * Back references have four fields, each 64 bits long. Unfortunately, + * This is hashed into a single 64 bit number and placed into the key offset. + * The key objectid corresponds to the first byte in the extent, and the + * key type is set to BTRFS_EXTENT_REF_KEY + */ int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 bytenr, @@ -939,10 +1027,13 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct u64 start; u64 end; struct btrfs_fs_info *info = extent_root->fs_info; + struct extent_buffer *eb; struct btrfs_path *path; struct btrfs_key ins; + struct btrfs_disk_key first; struct btrfs_extent_item extent_item; int ret; + int level; int err = 0; btrfs_set_stack_extent_refs(&extent_item, 1); @@ -961,10 +1052,19 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct &extent_item, sizeof(extent_item)); clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, GFP_NOFS); + eb = read_tree_block(extent_root, ins.objectid, ins.offset); + level = btrfs_header_level(eb); + if (level == 0) { + btrfs_item_key(eb, &first, 0); + } else { + btrfs_node_key(eb, &first, 0); + } err = btrfs_insert_extent_backref(trans, extent_root, path, start, extent_root->root_key.objectid, - 0, 0, 0); + 0, btrfs_disk_key_objectid(&first), + level); BUG_ON(err); + free_extent_buffer(eb); } btrfs_free_path(path); return 0; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 5b1f90f06e0..1cc4d285951 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -545,13 +545,10 @@ next_slot: u64 disk_num_bytes = 0; u64 extent_num_bytes = 0; u64 root_gen; + u64 root_owner; - if (leaf != root->node) { - root_gen = - btrfs_header_generation(path->nodes[1]); - } else { - root_gen = btrfs_header_generation(leaf); - } + root_gen = btrfs_header_generation(leaf); + root_owner = btrfs_header_owner(leaf); if (found_extent) { disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, @@ -575,7 +572,7 @@ next_slot: ret = btrfs_free_extent(trans, root, disk_bytenr, disk_num_bytes, - root->root_key.objectid, + root_owner, root_gen, inode->i_ino, key.offset, 0); } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bb70db0c9df..03fea037667 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -563,6 +563,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, u64 extent_num_bytes = 0; u64 item_end = 0; u64 root_gen = 0; + u64 root_owner = 0; int found_extent; int del_item; int extent_type = -1; @@ -673,15 +674,8 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, found_extent = 1; inode->i_blocks -= num_dec; } - if (leaf == root->node) { - root_gen = - btrfs_header_generation(leaf); - } else { - struct extent_buffer *parent; - parent = path->nodes[1]; - root_gen = - btrfs_header_generation(parent); - } + root_gen = btrfs_header_generation(leaf); + root_owner = btrfs_header_owner(leaf); } } else if (extent_type == BTRFS_FILE_EXTENT_INLINE && !del_item) { @@ -703,7 +697,7 @@ delete: if (found_extent) { ret = btrfs_free_extent(trans, root, extent_start, extent_num_bytes, - root->root_key.objectid, + root_owner, root_gen, inode->i_ino, found_key.offset, 0); BUG_ON(ret); -- cgit v1.2.3-70-g09d2 From 3954401fa6013bb2f2c8758b903e9bffcf25b64b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 12 Dec 2007 14:38:19 -0500 Subject: Btrfs: Add back pointers from the inode to the directory that references it Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 21 +++- fs/btrfs/inode-item.c | 136 ++++++++++++++++++++++++++ fs/btrfs/inode.c | 258 +++++++++++++++++++++++--------------------------- 3 files changed, 272 insertions(+), 143 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index cb1b156d954..035fa155034 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -53,7 +53,7 @@ extern struct kmem_cache *btrfs_path_cachep; #define BTRFS_CSUM_SIZE 32 /* four bytes for CRC32 */ #define BTRFS_CRC32_SIZE 4 -#define BTRFS_EMPTY_DIR_SIZE 6 +#define BTRFS_EMPTY_DIR_SIZE 0 #define BTRFS_FT_UNKNOWN 0 #define BTRFS_FT_REG_FILE 1 @@ -206,6 +206,11 @@ struct btrfs_extent_ref { __le64 offset; } __attribute__ ((__packed__)); +struct btrfs_inode_ref { + __le16 name_len; + /* name goes here */ +} __attribute__ ((__packed__)); + struct btrfs_inode_timespec { __le64 sec; __le32 nsec; @@ -379,7 +384,8 @@ struct btrfs_root { * the FS */ #define BTRFS_INODE_ITEM_KEY 1 -#define BTRFS_XATTR_ITEM_KEY 2 +#define BTRFS_INODE_REF_KEY 2 +#define BTRFS_XATTR_ITEM_KEY 8 /* reserve 2-15 close to the inode for later flexibility */ /* @@ -486,6 +492,9 @@ BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item, BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item, used, 64); +/* struct btrfs_inode_ref */ +BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16); + /* struct btrfs_inode_item */ BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64); @@ -1043,6 +1052,14 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, int btrfs_find_highest_inode(struct btrfs_root *fs_root, u64 *objectid); /* inode-item.c */ +int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + u64 inode_objectid, u64 ref_objectid); +int btrfs_del_inode_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + u64 inode_objectid, u64 ref_objectid); int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid); diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index 35d2608f891..cba30b6cc6f 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -20,6 +20,142 @@ #include "disk-io.h" #include "transaction.h" +int find_name_in_backref(struct btrfs_path *path, const char * name, + int name_len, struct btrfs_inode_ref **ref_ret) +{ + struct extent_buffer *leaf; + struct btrfs_inode_ref *ref; + unsigned long ptr; + unsigned long name_ptr; + u32 item_size; + u32 cur_offset = 0; + int len; + + leaf = path->nodes[0]; + item_size = btrfs_item_size_nr(leaf, path->slots[0]); + ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); + while (cur_offset < item_size) { + ref = (struct btrfs_inode_ref *)(ptr + cur_offset); + len = btrfs_inode_ref_name_len(leaf, ref); + name_ptr = (unsigned long)(ref + 1); + cur_offset += len + sizeof(*ref); + if (len != name_len) + continue; + if (memcmp_extent_buffer(leaf, name, name_ptr, name_len) == 0) { + *ref_ret = ref; + return 1; + } + } + return 0; +} + +int btrfs_del_inode_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + u64 inode_objectid, u64 ref_objectid) +{ + struct btrfs_path *path; + struct btrfs_key key; + struct btrfs_inode_ref *ref; + struct extent_buffer *leaf; + unsigned long ptr; + unsigned long item_start; + u32 item_size; + u32 sub_item_len; + int ret; + int del_len = name_len + sizeof(*ref); + + key.objectid = inode_objectid; + key.offset = ref_objectid; + btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret > 0) { + ret = -ENOENT; + goto out; + } else if (ret < 0) { + goto out; + } + if (!find_name_in_backref(path, name, name_len, &ref)) { + ret = -ENOENT; + goto out; + } + leaf = path->nodes[0]; + item_size = btrfs_item_size_nr(leaf, path->slots[0]); + if (del_len == item_size) { + ret = btrfs_del_item(trans, root, path); + goto out; + } + ptr = (unsigned long)ref; + sub_item_len = name_len + sizeof(*ref); + item_start = btrfs_item_ptr_offset(leaf, path->slots[0]); + memmove_extent_buffer(leaf, ptr, ptr + sub_item_len, + item_size - (ptr + sub_item_len - item_start)); + ret = btrfs_truncate_item(trans, root, path, + item_size - sub_item_len, 1); + BUG_ON(ret); +out: + btrfs_free_path(path); + return ret; +} + +int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + u64 inode_objectid, u64 ref_objectid) +{ + struct btrfs_path *path; + struct btrfs_key key; + struct btrfs_inode_ref *ref; + unsigned long ptr; + int ret; + int ins_len = name_len + sizeof(*ref); + + key.objectid = inode_objectid; + key.offset = ref_objectid; + btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = btrfs_insert_empty_item(trans, root, path, &key, + ins_len); + if (ret == -EEXIST) { + u32 old_size; + + if (find_name_in_backref(path, name, name_len, &ref)) + goto out; + + old_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]); + ret = btrfs_extend_item(trans, root, path, ins_len); + BUG_ON(ret); + ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_inode_ref); + ref = (struct btrfs_inode_ref *)((unsigned long)ref + old_size); + btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + ptr = (unsigned long)(ref + 1); + ret = 0; + } else if (ret < 0) { + goto out; + } else { + ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_inode_ref); + btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + ptr = (unsigned long)(ref + 1); + } + write_extent_buffer(path->nodes[0], name, ptr, name_len); + btrfs_mark_buffer_dirty(path->nodes[0]); + +out: + btrfs_free_path(path); + return ret; +} + int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 03fea037667..cefe740b6c7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -404,6 +404,17 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, ret = btrfs_delete_one_dir_name(trans, root, path, di); dentry->d_inode->i_ctime = dir->i_ctime; + if (!S_ISLNK(dentry->d_inode->i_mode)) { + ret = btrfs_del_inode_ref(trans, root, name, name_len, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino); + if (ret) { + printk("failed to delete reference to %.*s, " + "inode %lu parent %lu\n", name_len, name, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino); + } + } err: btrfs_free_path(path); if (!ret) { @@ -445,75 +456,27 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) int err; int ret; struct btrfs_root *root = BTRFS_I(dir)->root; - struct btrfs_path *path; - struct btrfs_key key; struct btrfs_trans_handle *trans; - struct btrfs_key found_key; - int found_type; - struct extent_buffer *leaf; - char *goodnames = ".."; unsigned long nr; if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) return -ENOTEMPTY; - path = btrfs_alloc_path(); - BUG_ON(!path); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, dir); - key.objectid = inode->i_ino; - key.offset = (u64)-1; - key.type = (u8)-1; - while(1) { - ret = btrfs_search_slot(trans, root, &key, path, -1, 1); - if (ret < 0) { - err = ret; - goto out; - } - BUG_ON(ret == 0); - if (path->slots[0] == 0) { - err = -ENOENT; - goto out; - } - path->slots[0]--; - leaf = path->nodes[0]; - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - found_type = btrfs_key_type(&found_key); - if (found_key.objectid != inode->i_ino) { - err = -ENOENT; - goto out; - } - if ((found_type != BTRFS_DIR_ITEM_KEY && - found_type != BTRFS_DIR_INDEX_KEY) || - (!btrfs_match_dir_item_name(root, path, goodnames, 2) && - !btrfs_match_dir_item_name(root, path, goodnames, 1))) { - err = -ENOTEMPTY; - goto out; - } - ret = btrfs_del_item(trans, root, path); - BUG_ON(ret); - - if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1) - break; - btrfs_release_path(root, path); - } - ret = 0; - btrfs_release_path(root, path); /* now the directory is empty */ err = btrfs_unlink_trans(trans, root, dir, dentry); if (!err) { inode->i_size = 0; } -out: - btrfs_release_path(root, path); - btrfs_free_path(path); + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + if (ret && !err) err = ret; return err; @@ -887,21 +850,59 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = BTRFS_I(dir)->root; int ret = 0; + if (namelen == 1 && strcmp(name, ".") == 0) { + location->objectid = dir->i_ino; + location->type = BTRFS_INODE_ITEM_KEY; + location->offset = 0; + return 0; + } path = btrfs_alloc_path(); BUG_ON(!path); + + if (namelen == 1 && strcmp(name, "..") == 0) { + struct btrfs_key key; + struct extent_buffer *leaf; + u32 nritems; + int slot; + + key.objectid = dir->i_ino; + btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); + key.offset = 0; + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + BUG_ON(ret == 0); + ret = 0; + + leaf = path->nodes[0]; + slot = path->slots[0]; + nritems = btrfs_header_nritems(leaf); + if (slot >= nritems) + goto out_err; + + btrfs_item_key_to_cpu(leaf, &key, slot); + if (key.objectid != dir->i_ino || + key.type != BTRFS_INODE_REF_KEY) { + goto out_err; + } + location->objectid = key.offset; + location->type = BTRFS_INODE_ITEM_KEY; + location->offset = 0; + goto out; + } + di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, namelen, 0); if (IS_ERR(di)) ret = PTR_ERR(di); if (!di || IS_ERR(di)) { - location->objectid = 0; - goto out; + goto out_err; } btrfs_dir_item_key_to_cpu(path->nodes[0], di, location); out: - btrfs_release_path(root, path); btrfs_free_path(path); return ret; +out_err: + location->objectid = 0; + goto out; } /* @@ -1053,13 +1054,50 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) if (root->fs_info->tree_root == root) key_type = BTRFS_DIR_ITEM_KEY; + /* special case for "." */ + if (filp->f_pos == 0) { + over = filldir(dirent, ".", 1, + 1, inode->i_ino, + DT_DIR); + if (over) + return 0; + filp->f_pos = 1; + } + mutex_lock(&root->fs_info->fs_mutex); key.objectid = inode->i_ino; + path = btrfs_alloc_path(); + path->reada = 2; + + /* special case for .., just use the back ref */ + if (filp->f_pos == 1) { + btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); + key.offset = 0; + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + BUG_ON(ret == 0); + leaf = path->nodes[0]; + slot = path->slots[0]; + nritems = btrfs_header_nritems(leaf); + if (slot >= nritems) { + btrfs_release_path(root, path); + goto read_dir_items; + } + btrfs_item_key_to_cpu(leaf, &found_key, slot); + btrfs_release_path(root, path); + if (found_key.objectid != key.objectid || + found_key.type != BTRFS_INODE_REF_KEY) + goto read_dir_items; + over = filldir(dirent, "..", 2, + 2, found_key.offset, DT_DIR); + if (over) + goto nopos; + filp->f_pos = 2; + } + +read_dir_items: btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; - path = btrfs_alloc_path(); - path->reada = 2; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; @@ -1255,6 +1293,13 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_parent->d_inode->i_ino, &key, btrfs_inode_type(inode)); if (ret == 0) { + if (!S_ISLNK(inode->i_mode)) { + ret = btrfs_insert_inode_ref(trans, root, + dentry->d_name.name, + dentry->d_name.len, + inode->i_ino, + dentry->d_parent->d_inode->i_ino); + } parent_inode = dentry->d_parent->d_inode; parent_inode->i_size += dentry->d_name.len * 2; parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; @@ -1430,35 +1475,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, return err; } -static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 objectid, u64 dirid) -{ - int ret; - char buf[2]; - struct btrfs_key key; - - buf[0] = '.'; - buf[1] = '.'; - - key.objectid = objectid; - key.offset = 0; - btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); - - ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid, - &key, BTRFS_FT_DIR); - if (ret) - goto error; - - key.objectid = dirid; - ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid, - &key, BTRFS_FT_DIR); - if (ret) - goto error; -error: - return ret; -} - static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) { struct inode *inode; @@ -1496,11 +1512,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) inode->i_fop = &btrfs_dir_file_operations; btrfs_set_trans_block_group(trans, inode); - err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino); - if (err) - goto out_fail; - - inode->i_size = 6; + inode->i_size = 0; err = btrfs_update_inode(trans, root, inode); if (err) goto out_fail; @@ -1971,6 +1983,12 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; + ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, + name, namelen, objectid, + root->fs_info->sb->s_root->d_inode->i_ino); + if (ret) + goto fail; + ret = btrfs_commit_transaction(trans, root); if (ret) goto fail_commit; @@ -1989,12 +2007,10 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) inode->i_fop = &btrfs_dir_file_operations; new_root->inode = inode; - ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid); - if (ret) - goto fail; - + ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid, + new_dirid); inode->i_nlink = 1; - inode->i_size = 6; + inode->i_size = 0; ret = btrfs_update_inode(trans, new_root, inode); if (ret) goto fail; @@ -2068,6 +2084,13 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; + ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, + name, namelen, objectid, + root->fs_info->sb->s_root->d_inode->i_ino); + + if (ret) + goto fail; + ret = btrfs_inc_root_ref(trans, root, objectid); if (ret) goto fail; @@ -2338,7 +2361,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, struct inode *old_inode = old_dentry->d_inode; struct timespec ctime = CURRENT_TIME; struct btrfs_path *path; - struct btrfs_dir_item *di; int ret; if (S_ISDIR(old_inode->i_mode) && new_inode && @@ -2361,52 +2383,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, new_dir->i_ctime = new_dir->i_mtime = ctime; old_inode->i_ctime = ctime; - if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) { - struct btrfs_key *location = &BTRFS_I(new_dir)->location; - struct btrfs_key old_parent_key; - di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino, - "..", 2, -1); - if (IS_ERR(di)) { - ret = PTR_ERR(di); - goto out_fail; - } - if (!di) { - ret = -ENOENT; - goto out_fail; - } - btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key); - ret = btrfs_del_item(trans, root, path); - if (ret) { - goto out_fail; - } - btrfs_release_path(root, path); - - di = btrfs_lookup_dir_index_item(trans, root, path, - old_inode->i_ino, - old_parent_key.objectid, - "..", 2, -1); - if (IS_ERR(di)) { - ret = PTR_ERR(di); - goto out_fail; - } - if (!di) { - ret = -ENOENT; - goto out_fail; - } - ret = btrfs_del_item(trans, root, path); - if (ret) { - goto out_fail; - } - btrfs_release_path(root, path); - - ret = btrfs_insert_dir_item(trans, root, "..", 2, - old_inode->i_ino, location, - BTRFS_FT_DIR); - if (ret) - goto out_fail; - } - - ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry); if (ret) goto out_fail; -- cgit v1.2.3-70-g09d2 From 76fea00a0522185658b733d441d6ec049344456b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 13 Dec 2007 09:06:01 -0500 Subject: Btrfs: Add backrefs for symbolic link inodes Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 30 +++++++++++++----------------- 1 file changed, 13 insertions(+), 17 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index cefe740b6c7..06d92be1680 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -404,16 +404,14 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, ret = btrfs_delete_one_dir_name(trans, root, path, di); dentry->d_inode->i_ctime = dir->i_ctime; - if (!S_ISLNK(dentry->d_inode->i_mode)) { - ret = btrfs_del_inode_ref(trans, root, name, name_len, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino); - if (ret) { - printk("failed to delete reference to %.*s, " - "inode %lu parent %lu\n", name_len, name, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino); - } + ret = btrfs_del_inode_ref(trans, root, name, name_len, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino); + if (ret) { + printk("failed to delete reference to %.*s, " + "inode %lu parent %lu\n", name_len, name, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino); } err: btrfs_free_path(path); @@ -1293,13 +1291,11 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_parent->d_inode->i_ino, &key, btrfs_inode_type(inode)); if (ret == 0) { - if (!S_ISLNK(inode->i_mode)) { - ret = btrfs_insert_inode_ref(trans, root, - dentry->d_name.name, - dentry->d_name.len, - inode->i_ino, - dentry->d_parent->d_inode->i_ino); - } + ret = btrfs_insert_inode_ref(trans, root, + dentry->d_name.name, + dentry->d_name.len, + inode->i_ino, + dentry->d_parent->d_inode->i_ino); parent_inode = dentry->d_parent->d_inode; parent_inode->i_size += dentry->d_name.len * 2; parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; -- cgit v1.2.3-70-g09d2 From 7a7205367dfd9adf129f0ebc7d040a557fe0d1ba Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 13 Dec 2007 09:06:59 -0500 Subject: Btrfs: Fix typo in .. check (thanks Yan) Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 06d92be1680..41b0c7de7a5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -857,7 +857,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, path = btrfs_alloc_path(); BUG_ON(!path); - if (namelen == 1 && strcmp(name, "..") == 0) { + if (namelen == 2 && strcmp(name, "..") == 0) { struct btrfs_key key; struct extent_buffer *leaf; u32 nritems; -- cgit v1.2.3-70-g09d2 From e9906a98497c26a3c1a7532b51b81cc1fbdf9e1d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 14 Dec 2007 12:56:58 -0500 Subject: Fixes for loopback files in btrfs Signed-off-by: Chris Mason --- fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 15 +++++++++++++-- 2 files changed, 14 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index a280b956222..62fcd79d8ab 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -861,6 +861,7 @@ struct file_operations btrfs_file_operations = { .llseek = generic_file_llseek, .read = do_sync_read, .aio_read = generic_file_aio_read, + .splice_read = generic_file_splice_read, .write = btrfs_file_write, .mmap = btrfs_file_mmap, .open = generic_file_open, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 41b0c7de7a5..e535c50e669 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1893,8 +1893,19 @@ static void btrfs_truncate(struct inode *inode) int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to) { - return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree, - page->mapping->host, page, from, to); + loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; + struct inode *inode = page->mapping->host; + + btrfs_cow_one_page(inode, page, PAGE_CACHE_SIZE); + + set_page_extent_mapped(page); + set_page_dirty(page); + + if (pos > inode->i_size) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + return 0; } static int create_subvol(struct btrfs_root *root, char *name, int namelen) -- cgit v1.2.3-70-g09d2 From b6cda9bcb4df7544c67fc3548a53bc1607d59f46 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 14 Dec 2007 15:30:32 -0500 Subject: Btrfs: Add mount -o nodatasum to turn of file data checksumming Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 14 ++++++++------ fs/btrfs/disk-io.c | 1 + fs/btrfs/inode.c | 11 ++++++++++- fs/btrfs/super.c | 17 +++++++++++++++-- 4 files changed, 34 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 052555ced44..5e255cabfd1 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -321,6 +321,7 @@ struct btrfs_fs_info { u64 generation; u64 last_trans_committed; + unsigned long mount_opt; struct btrfs_transaction *running_transaction; struct btrfs_super_block super_copy; struct extent_buffer *sb_buffer; @@ -429,6 +430,13 @@ struct btrfs_root { */ #define BTRFS_STRING_ITEM_KEY 253 +#define BTRFS_MOUNT_NODATASUM 0x1 + +#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) +#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) +#define btrfs_test_opt(root, opt) ((root)->fs_info->mount_opt & \ + BTRFS_MOUNT_##opt) + /* some macros to generate set/get funcs for the struct fields. This * assumes there is a lefoo_to_cpu for every type, so lets make a simple * one for u8: @@ -906,12 +914,6 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { ((unsigned long)(btrfs_leaf_data(leaf) + \ btrfs_item_offset_nr(leaf, slot))) -/* mount option defines and helpers */ -#define BTRFS_MOUNT_SUBVOL 0x000001 -#define btrfs_clear_opt(o, opt) o &= ~BTRFS_MOUNT_##opt -#define btrfs_set_opt(o, opt) o |= BTRFS_MOUNT_##opt -#define btrfs_test_opt(sb, opt) (BTRFS_SB(sb)->s_mount_opt & \ - BTRFS_MOUNT_##opt) /* extent-tree.c */ int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 60a30da6af0..4338b00c2a2 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -568,6 +568,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; fs_info->sb = sb; + fs_info->mount_opt = 0; fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e535c50e669..11885cb114e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -116,10 +116,13 @@ int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; char *kaddr; - int ret; + int ret = 0; u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; size_t offset = start - page_start; + if (btrfs_test_opt(root, NODATASUM)) + return 0; + mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -143,6 +146,9 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) struct btrfs_path *path = NULL; u32 csum; + if (btrfs_test_opt(root, NODATASUM)) + return 0; + mutex_lock(&root->fs_info->fs_mutex); path = btrfs_alloc_path(); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); @@ -176,6 +182,9 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) u32 csum = ~(u32)0; unsigned long flags; + if (btrfs_test_opt(root, NODATASUM)) + return 0; + ret = get_state_private(em_tree, start, &private); local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index ffa02872c4d..ad4f280ec77 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -61,11 +61,12 @@ static void btrfs_put_super (struct super_block * sb) } enum { - Opt_subvol, Opt_err, + Opt_subvol, Opt_nodatasum, Opt_err, }; static match_table_t tokens = { {Opt_subvol, "subvol=%s"}, + {Opt_nodatasum, "nodatasum"}, {Opt_err, NULL} }; @@ -74,7 +75,12 @@ static int parse_options (char * options, char **subvol_name) { char * p; + struct btrfs_fs_info *info = NULL; substring_t args[MAX_OPT_ARGS]; + + if (root) + info = root->fs_info; + if (!options) return 1; @@ -86,7 +92,12 @@ static int parse_options (char * options, token = match_token(p, tokens, args); switch (token) { case Opt_subvol: - *subvol_name = match_strdup(&args[0]); + if (subvol_name) + *subvol_name = match_strdup(&args[0]); + break; + case Opt_nodatasum: + if (root) + btrfs_set_opt(info->mount_opt, NODATASUM); break; default: return 0; @@ -143,6 +154,8 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) goto fail_close; } + parse_options((char *)data, tree_root, NULL); + /* this does the super kobj at the same time */ err = btrfs_sysfs_add_super(tree_root->fs_info); if (err) -- cgit v1.2.3-70-g09d2 From be20aa9dbadc8c06283784ee12bbc0d97dea3418 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Dec 2007 20:14:01 -0500 Subject: Btrfs: Add mount option to turn off data cow A number of workloads do not require copy on write data or checksumming. mount -o nodatasum to disable checksums and -o nodatacow to disable both copy on write and checksumming. In nodatacow mode, copy on write is still performed when a given extent is under snapshot. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 53 ++++++++++++++++++++- fs/btrfs/ctree.h | 8 ++++ fs/btrfs/extent-tree.c | 101 ++++++++++++++++++++++++++++++++++++++-- fs/btrfs/inode.c | 122 +++++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/super.c | 34 +++++++++++--- 5 files changed, 293 insertions(+), 25 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index fd8233e05cf..585f279d111 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -70,7 +70,58 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) memset(p, 0, sizeof(*p)); } -static int __btrfs_cow_block(struct btrfs_trans_handle *trans, +int btrfs_copy_root(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *buf, + struct extent_buffer **cow_ret, u64 new_root_objectid) +{ + struct extent_buffer *cow; + u32 nritems; + int ret = 0; + int level; + struct btrfs_key first_key; + struct btrfs_root new_root; + + memcpy(&new_root, root, sizeof(new_root)); + new_root.root_key.objectid = new_root_objectid; + + WARN_ON(root->ref_cows && trans->transid != + root->fs_info->running_transaction->transid); + WARN_ON(root->ref_cows && trans->transid != root->last_trans); + + level = btrfs_header_level(buf); + nritems = btrfs_header_nritems(buf); + if (nritems) { + if (level == 0) + btrfs_item_key_to_cpu(buf, &first_key, 0); + else + btrfs_node_key_to_cpu(buf, &first_key, 0); + } else { + first_key.objectid = 0; + } + cow = __btrfs_alloc_free_block(trans, &new_root, buf->len, + new_root_objectid, + trans->transid, first_key.objectid, + level, buf->start, 0); + if (IS_ERR(cow)) + return PTR_ERR(cow); + + copy_extent_buffer(cow, buf, 0, 0, cow->len); + btrfs_set_header_bytenr(cow, cow->start); + btrfs_set_header_generation(cow, trans->transid); + btrfs_set_header_owner(cow, new_root_objectid); + + WARN_ON(btrfs_header_generation(buf) > trans->transid); + ret = btrfs_inc_ref(trans, &new_root, buf); + if (ret) + return ret; + + btrfs_mark_buffer_dirty(cow); + *cow_ret = cow; + return 0; +} + +int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5e255cabfd1..b51b021fff8 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -431,6 +431,7 @@ struct btrfs_root { #define BTRFS_STRING_ITEM_KEY 253 #define BTRFS_MOUNT_NODATASUM 0x1 +#define BTRFS_MOUNT_NODATACOW 0x2 #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) @@ -915,6 +916,9 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { btrfs_item_offset_nr(leaf, slot))) /* extent-tree.c */ +u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, + struct btrfs_path *count_path, + u64 first_extent); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy); @@ -974,6 +978,10 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, struct extent_buffer **cow_ret); +int btrfs_copy_root(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct extent_buffer *buf, + struct extent_buffer **cow_ret, u64 new_root_objectid); int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u32 data_size); int btrfs_truncate_item(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 00414836e96..1412d556313 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -679,6 +679,104 @@ out: return 0; } +u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, + struct btrfs_path *count_path, + u64 first_extent) +{ + struct btrfs_root *extent_root = root->fs_info->extent_root; + struct btrfs_path *path; + u64 bytenr; + u64 found_objectid; + u64 root_objectid = 0; + u32 total_count = 0; + u32 cur_count; + u32 refs; + u32 nritems; + int ret; + struct btrfs_key key; + struct btrfs_key found_key; + struct extent_buffer *l; + struct btrfs_extent_item *item; + struct btrfs_extent_ref *ref_item; + int level = -1; + + path = btrfs_alloc_path(); +again: + if (level == -1) + bytenr = first_extent; + else + bytenr = count_path->nodes[level]->start; + + cur_count = 0; + key.objectid = bytenr; + key.offset = 0; + + btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); + ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); + if (ret < 0) + goto out; + BUG_ON(ret == 0); + + l = path->nodes[0]; + btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); + + if (found_key.objectid != bytenr || + found_key.type != BTRFS_EXTENT_ITEM_KEY) { + goto out; + } + + item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); + refs = btrfs_extent_refs(l, item); + while (1) { + nritems = btrfs_header_nritems(l); + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(extent_root, path); + if (ret == 0) + continue; + break; + } + btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); + if (found_key.objectid != bytenr) + break; + if (found_key.type != BTRFS_EXTENT_REF_KEY) { + path->slots[0]++; + continue; + } + + cur_count++; + ref_item = btrfs_item_ptr(l, path->slots[0], + struct btrfs_extent_ref); + found_objectid = btrfs_ref_root(l, ref_item); + + if (found_objectid != root_objectid) + total_count++; + + if (total_count > 1) + goto out; + + if (root_objectid == 0) + root_objectid = found_objectid; + + path->slots[0]++; + } + if (cur_count == 0) { + total_count = 0; + goto out; + } + if (total_count > 1) + goto out; + if (level >= 0 && root->node == count_path->nodes[level]) + goto out; + level++; + btrfs_release_path(root, path); + goto again; + +out: + btrfs_free_path(path); + return total_count; + +} + int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 owner_objectid) { @@ -1127,9 +1225,6 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root if (!path) return -ENOMEM; - if (ref_generation && owner_objectid == 0 && root_objectid == 3) { -//printk("drop backref root %Lu gen %Lu byte %Lu\n", root_objectid, ref_generation, bytenr ); - } ret = lookup_extent_backref(trans, extent_root, path, bytenr, root_objectid, ref_generation, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 11885cb114e..91f3fc43e2a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -72,21 +72,22 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, }; -static int run_delalloc_range(struct inode *inode, u64 start, u64 end) +static int cow_file_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; - struct btrfs_key ins; u64 alloc_hint = 0; u64 num_bytes; - int ret; u64 blocksize = root->sectorsize; + struct btrfs_key ins; + int ret; - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); BUG_ON(!trans); + btrfs_set_trans_block_group(trans, inode); + num_bytes = (end - start + blocksize) & ~(blocksize - 1); + num_bytes = max(blocksize, num_bytes); ret = btrfs_drop_extents(trans, root, inode, start, start + num_bytes, start, &alloc_hint); @@ -106,6 +107,101 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) ins.offset); out: btrfs_end_transaction(trans, root); + return ret; +} + +static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) +{ + u64 extent_start; + u64 extent_end; + u64 bytenr; + u64 cow_end; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_buffer *leaf; + int found_type; + struct btrfs_path *path; + struct btrfs_file_extent_item *item; + int ret; + int err; + struct btrfs_key found_key; + + path = btrfs_alloc_path(); + BUG_ON(!path); +again: + ret = btrfs_lookup_file_extent(NULL, root, path, + inode->i_ino, start, 0); + if (ret < 0) { + btrfs_free_path(path); + return ret; + } + + cow_end = end; + if (ret != 0) { + if (path->slots[0] == 0) + goto not_found; + path->slots[0]--; + } + + leaf = path->nodes[0]; + item = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + + /* are we inside the extent that was found? */ + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + found_type = btrfs_key_type(&found_key); + if (found_key.objectid != inode->i_ino || + found_type != BTRFS_EXTENT_DATA_KEY) { + goto not_found; + } + + found_type = btrfs_file_extent_type(leaf, item); + extent_start = found_key.offset; + if (found_type == BTRFS_FILE_EXTENT_REG) { + extent_end = extent_start + + btrfs_file_extent_num_bytes(leaf, item); + err = 0; + + if (start < extent_start || start >= extent_end) + goto not_found; + + cow_end = min(end, extent_end - 1); + bytenr = btrfs_file_extent_disk_bytenr(leaf, item); + if (bytenr == 0) + goto not_found; + + bytenr += btrfs_file_extent_offset(leaf, item); + if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) { + goto not_found; + } + + start = extent_end; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + goto not_found; + } +loop: + if (start > end) { + btrfs_free_path(path); + return 0; + } + btrfs_release_path(root, path); + goto again; + +not_found: + cow_file_range(inode, start, cow_end); + start = cow_end + 1; + goto loop; +} + +static int run_delalloc_range(struct inode *inode, u64 start, u64 end) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret; + + mutex_lock(&root->fs_info->fs_mutex); + if (btrfs_test_opt(root, NODATACOW)) + ret = run_delalloc_nocow(inode, start, end); + else + ret = cow_file_range(inode, start, end); mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -1907,9 +2003,6 @@ int btrfs_commit_write(struct file *file, struct page *page, btrfs_cow_one_page(inode, page, PAGE_CACHE_SIZE); - set_page_extent_mapped(page); - set_page_dirty(page); - if (pos > inode->i_size) { i_size_write(inode, pos); mark_inode_dirty(inode); @@ -2078,13 +2171,18 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) key.objectid = objectid; key.offset = 1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + extent_buffer_get(root->node); btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); free_extent_buffer(tmp); - btrfs_set_root_bytenr(&new_root_item, root->node->start); - btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node)); + + btrfs_copy_root(trans, root, root->node, &tmp, objectid); + + btrfs_set_root_bytenr(&new_root_item, tmp->start); + btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); + free_extent_buffer(tmp); if (ret) goto fail; @@ -2106,10 +2204,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail; - - ret = btrfs_inc_root_ref(trans, root, objectid); - if (ret) - goto fail; fail: nr = trans->blocks_used; err = btrfs_commit_transaction(trans, root); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index ad4f280ec77..2116728d1f9 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -61,12 +61,13 @@ static void btrfs_put_super (struct super_block * sb) } enum { - Opt_subvol, Opt_nodatasum, Opt_err, + Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_err, }; static match_table_t tokens = { {Opt_subvol, "subvol=%s"}, {Opt_nodatasum, "nodatasum"}, + {Opt_nodatacow, "nodatacow"}, {Opt_err, NULL} }; @@ -78,12 +79,20 @@ static int parse_options (char * options, struct btrfs_fs_info *info = NULL; substring_t args[MAX_OPT_ARGS]; - if (root) - info = root->fs_info; - if (!options) return 1; + /* + * strsep changes the string, duplicate it because parse_options + * gets called twice + */ + options = kstrdup(options, GFP_NOFS); + if (!options) + return -ENOMEM; + + if (root) + info = root->fs_info; + while ((p = strsep (&options, ",")) != NULL) { int token; if (!*p) @@ -92,17 +101,28 @@ static int parse_options (char * options, token = match_token(p, tokens, args); switch (token) { case Opt_subvol: - if (subvol_name) + if (subvol_name) { *subvol_name = match_strdup(&args[0]); + } break; case Opt_nodatasum: - if (root) + if (info) { + printk("btrfs: setting nodatacsum\n"); btrfs_set_opt(info->mount_opt, NODATASUM); + } + break; + case Opt_nodatacow: + if (info) { + printk("btrfs: setting nodatacow\n"); + btrfs_set_opt(info->mount_opt, NODATACOW); + btrfs_set_opt(info->mount_opt, NODATASUM); + } break; default: - return 0; + break; } } + kfree(options); return 1; } -- cgit v1.2.3-70-g09d2 From c59f8951d48c5eb000926935f3ab063d8181d1ee Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Dec 2007 20:14:04 -0500 Subject: Btrfs: Add mount option to enforce a max extent size Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 1 + fs/btrfs/inode.c | 28 ++++++++++++++++++---------- fs/btrfs/super.c | 46 +++++++++++++++++++++++++++++++++++++++++++++- 4 files changed, 65 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b51b021fff8..32b24460ec8 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -322,6 +322,7 @@ struct btrfs_fs_info { u64 generation; u64 last_trans_committed; unsigned long mount_opt; + u64 max_extent; struct btrfs_transaction *running_transaction; struct btrfs_super_block super_copy; struct extent_buffer *sb_buffer; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 27856417899..7213012c27d 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -569,6 +569,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->extent_root = extent_root; fs_info->sb = sb; fs_info->mount_opt = 0; + fs_info->max_extent = (u64)-1; fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 91f3fc43e2a..686dd03f34f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -78,6 +78,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; u64 alloc_hint = 0; u64 num_bytes; + u64 cur_alloc_size; u64 blocksize = root->sectorsize; struct btrfs_key ins; int ret; @@ -94,17 +95,24 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) if (alloc_hint == EXTENT_MAP_INLINE) goto out; - ret = btrfs_alloc_extent(trans, root, num_bytes, - root->root_key.objectid, trans->transid, - inode->i_ino, start, 0, - alloc_hint, (u64)-1, &ins, 1); - if (ret) { - WARN_ON(1); - goto out; + while(num_bytes > 0) { + cur_alloc_size = min(num_bytes, root->fs_info->max_extent); + ret = btrfs_alloc_extent(trans, root, cur_alloc_size, + root->root_key.objectid, + trans->transid, + inode->i_ino, start, 0, + alloc_hint, (u64)-1, &ins, 1); + if (ret) { + WARN_ON(1); + goto out; + } + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + start, ins.objectid, ins.offset, + ins.offset); + num_bytes -= cur_alloc_size; + alloc_hint = ins.objectid + ins.offset; + start += cur_alloc_size; } - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - start, ins.objectid, ins.offset, - ins.offset); out: btrfs_end_transaction(trans, root); return ret; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 2116728d1f9..b7323914bf5 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -34,6 +34,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -61,16 +62,42 @@ static void btrfs_put_super (struct super_block * sb) } enum { - Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_err, + Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_err, }; static match_table_t tokens = { {Opt_subvol, "subvol=%s"}, {Opt_nodatasum, "nodatasum"}, {Opt_nodatacow, "nodatacow"}, + {Opt_max_extent, "max_extent=%s"}, {Opt_err, NULL} }; +static unsigned long parse_size(char *str) +{ + unsigned long res; + int mult = 1; + char *end; + char last; + + res = simple_strtoul(str, &end, 10); + + last = end[0]; + if (isalpha(last)) { + last = tolower(last); + switch (last) { + case 'g': + mult *= 1024; + case 'm': + mult *= 1024; + case 'k': + mult *= 1024; + } + res = res * mult; + } + return res; +} + static int parse_options (char * options, struct btrfs_root *root, char **subvol_name) @@ -118,6 +145,21 @@ static int parse_options (char * options, btrfs_set_opt(info->mount_opt, NODATASUM); } break; + case Opt_max_extent: + if (info) { + char *num = match_strdup(&args[0]); + if (num) { + info->max_extent = parse_size(num); + kfree(num); + + info->max_extent = max_t(u64, + info->max_extent, + root->sectorsize); + printk("btrfs: max_extent at %Lu\n", + info->max_extent); + } + } + break; default: break; } @@ -329,6 +371,8 @@ static int btrfs_get_sb(struct file_system_type *fs_type, ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, btrfs_fill_super, mnt, subvol_name ? subvol_name : "default"); + if (subvol_name) + kfree(subvol_name); return ret; } -- cgit v1.2.3-70-g09d2 From 6da6abae027e2dbc59bca5f4168b0760f25068c7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 18 Dec 2007 16:15:09 -0500 Subject: Btrfs: Back port to 2.6.18-el kernels Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 13 ++++ fs/btrfs/disk-io.c | 5 +- fs/btrfs/extent_map.c | 168 ++++++++++++++++++++++++++++++++++++++++++++++--- fs/btrfs/file.c | 15 +++-- fs/btrfs/inode.c | 21 +++++-- fs/btrfs/super.c | 1 + fs/btrfs/sysfs.c | 23 ++++++- fs/btrfs/transaction.c | 8 +++ fs/btrfs/transaction.h | 5 ++ 9 files changed, 239 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 32b24460ec8..2559f5d5d00 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -19,6 +19,7 @@ #ifndef __BTRFS__ #define __BTRFS__ +#include #include #include #include @@ -334,7 +335,11 @@ struct btrfs_fs_info { struct list_head trans_list; struct list_head hashers; struct list_head dead_roots; +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + struct work_struct trans_work; +#else struct delayed_work trans_work; +#endif struct kobject super_kobj; struct completion kobj_unregister; int do_barriers; @@ -916,6 +921,14 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { ((unsigned long)(btrfs_leaf_data(leaf) + \ btrfs_item_offset_nr(leaf, slot))) +static inline struct dentry *fdentry(struct file *file) { +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + return file->f_dentry; +#else + return file->f_path.dentry; +#endif +} + /* extent-tree.c */ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, struct btrfs_path *count_path, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 7213012c27d..ebb2db624fd 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -593,8 +593,11 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->do_barriers = 1; fs_info->closing = 0; fs_info->total_pinned = 0; - +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); +#else INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); +#endif BTRFS_I(fs_info->btree_inode)->root = tree_root; memset(&BTRFS_I(fs_info->btree_inode)->location, 0, sizeof(struct btrfs_key)); diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 8326a18413f..0d1e59a86e4 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1585,8 +1585,18 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, static int submit_one_bio(int rw, struct bio *bio) { + u64 maxsector; int ret = 0; + bio_get(bio); + + maxsector = bio->bi_bdev->bd_inode->i_size >> 9; + if (maxsector < bio->bi_sector) { + printk("sector too large max %Lu got %llu\n", maxsector, + (unsigned long long)bio->bi_sector); + WARN_ON(1); + } + submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) ret = -EOPNOTSUPP; @@ -1678,8 +1688,12 @@ static int __extent_read_full_page(struct extent_map_tree *tree, while (cur <= end) { if (cur >= last_byte) { + char *userpage; iosize = PAGE_CACHE_SIZE - page_offset; - zero_user_page(page, page_offset, iosize, KM_USER0); + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); set_extent_uptodate(tree, cur, cur + iosize - 1, GFP_NOFS); unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -1707,7 +1721,12 @@ static int __extent_read_full_page(struct extent_map_tree *tree, /* we've found a hole, just zero and go on */ if (block_start == EXTENT_MAP_HOLE) { - zero_user_page(page, page_offset, iosize, KM_USER0); + char *userpage; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + set_extent_uptodate(tree, cur, cur + iosize - 1, GFP_NOFS); unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -1804,9 +1823,14 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, } if (page->index == end_index) { + char *userpage; + size_t offset = i_size & (PAGE_CACHE_SIZE - 1); - zero_user_page(page, offset, - PAGE_CACHE_SIZE - offset, KM_USER0); + + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); } set_page_extent_mapped(page); @@ -1921,6 +1945,129 @@ done: return 0; } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + +/* Taken directly from 2.6.23 for 2.6.18 back port */ +typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, + void *data); + +/** + * write_cache_pages - walk the list of dirty pages of the given address space + * and write all of them. + * @mapping: address space structure to write + * @wbc: subtract the number of written pages from *@wbc->nr_to_write + * @writepage: function called for each page + * @data: data passed to writepage function + * + * If a page is already under I/O, write_cache_pages() skips it, even + * if it's dirty. This is desirable behaviour for memory-cleaning writeback, + * but it is INCORRECT for data-integrity system calls such as fsync(). fsync() + * and msync() need to guarantee that all the data which was dirty at the time + * the call was made get new I/O started against them. If wbc->sync_mode is + * WB_SYNC_ALL then we were called for data integrity and we must wait for + * existing IO to complete. + */ +static int write_cache_pages(struct address_space *mapping, + struct writeback_control *wbc, writepage_t writepage, + void *data) +{ + struct backing_dev_info *bdi = mapping->backing_dev_info; + int ret = 0; + int done = 0; + struct pagevec pvec; + int nr_pages; + pgoff_t index; + pgoff_t end; /* Inclusive */ + int scanned = 0; + int range_whole = 0; + + if (wbc->nonblocking && bdi_write_congested(bdi)) { + wbc->encountered_congestion = 1; + return 0; + } + + pagevec_init(&pvec, 0); + if (wbc->range_cyclic) { + index = mapping->writeback_index; /* Start from prev offset */ + end = -1; + } else { + index = wbc->range_start >> PAGE_CACHE_SHIFT; + end = wbc->range_end >> PAGE_CACHE_SHIFT; + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) + range_whole = 1; + scanned = 1; + } +retry: + while (!done && (index <= end) && + (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, + PAGECACHE_TAG_DIRTY, + min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { + unsigned i; + + scanned = 1; + for (i = 0; i < nr_pages; i++) { + struct page *page = pvec.pages[i]; + + /* + * At this point we hold neither mapping->tree_lock nor + * lock on the page itself: the page may be truncated or + * invalidated (changing page->mapping to NULL), or even + * swizzled back from swapper_space to tmpfs file + * mapping + */ + lock_page(page); + + if (unlikely(page->mapping != mapping)) { + unlock_page(page); + continue; + } + + if (!wbc->range_cyclic && page->index > end) { + done = 1; + unlock_page(page); + continue; + } + + if (wbc->sync_mode != WB_SYNC_NONE) + wait_on_page_writeback(page); + + if (PageWriteback(page) || + !clear_page_dirty_for_io(page)) { + unlock_page(page); + continue; + } + + ret = (*writepage)(page, wbc, data); + + if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) { + unlock_page(page); + ret = 0; + } + if (ret || (--(wbc->nr_to_write) <= 0)) + done = 1; + if (wbc->nonblocking && bdi_write_congested(bdi)) { + wbc->encountered_congestion = 1; + done = 1; + } + } + pagevec_release(&pvec); + cond_resched(); + } + if (!scanned && !done) { + /* + * We hit the last page and there is more work to be done: wrap + * back to the start of the file + */ + scanned = 1; + index = 0; + goto retry; + } + if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) + mapping->writeback_index = index; + return ret; +} +#endif + int extent_write_full_page(struct extent_map_tree *tree, struct page *page, get_extent_t *get_extent, struct writeback_control *wbc) @@ -1945,18 +2092,20 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page, ret = __extent_writepage(page, wbc, &epd); write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); - if (epd.bio) + if (epd.bio) { submit_one_bio(WRITE, epd.bio); + } return ret; } EXPORT_SYMBOL(extent_write_full_page); + int extent_writepages(struct extent_map_tree *tree, struct address_space *mapping, get_extent_t *get_extent, struct writeback_control *wbc) { - int ret; + int ret = 0; struct extent_page_data epd = { .bio = NULL, .tree = tree, @@ -1964,8 +2113,9 @@ int extent_writepages(struct extent_map_tree *tree, }; ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); - if (epd.bio) + if (epd.bio) { submit_one_bio(WRITE, epd.bio); + } return ret; } EXPORT_SYMBOL(extent_writepages); @@ -2106,7 +2256,9 @@ int extent_prepare_write(struct extent_map_tree *tree, flush_dcache_page(page); kunmap_atomic(kaddr, KM_USER0); } - if (!isnew && !PageUptodate(page) && + if ((em->block_start != EXTENT_MAP_HOLE && + em->block_start != EXTENT_MAP_INLINE) && + !isnew && !PageUptodate(page) && (block_off_end > to || block_off_start < from) && !test_range_bit(tree, block_start, cur_end, EXTENT_UPTODATE, 1)) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 62fcd79d8ab..461b09663fe 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -231,7 +231,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, { int err = 0; int i; - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = fdentry(file)->d_inode; struct extent_map *em; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; u64 hint_byte; @@ -652,7 +652,7 @@ static int prepare_pages(struct btrfs_root *root, { int i; unsigned long index = pos >> PAGE_CACHE_SHIFT; - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = fdentry(file)->d_inode; int err = 0; u64 start_pos; @@ -666,7 +666,11 @@ static int prepare_pages(struct btrfs_root *root, err = -ENOMEM; BUG_ON(1); } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + ClearPageDirty(pages[i]); +#else cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); +#endif wait_on_page_writeback(pages[i]); set_page_extent_mapped(pages[i]); WARN_ON(!PageLocked(pages[i])); @@ -682,7 +686,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, ssize_t num_written = 0; ssize_t err = 0; int ret = 0; - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; struct page **pages = NULL; int nrptrs; @@ -707,7 +711,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, goto out; if (count == 0) goto out; - err = remove_suid(file->f_path.dentry); + err = remove_suid(fdentry(file)); if (err) goto out; file_update_time(file); @@ -862,6 +866,9 @@ struct file_operations btrfs_file_operations = { .read = do_sync_read, .aio_read = generic_file_aio_read, .splice_read = generic_file_splice_read, +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + .sendfile = generic_file_sendfile, +#endif .write = btrfs_file_write, .mmap = btrfs_file_mmap, .open = generic_file_open, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 686dd03f34f..6a7d9160df2 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -532,7 +532,11 @@ err: dir->i_size -= name_len * 2; dir->i_mtime = dir->i_ctime = CURRENT_TIME; btrfs_update_inode(trans, root, dir); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + dentry->d_inode->i_nlink--; +#else drop_nlink(dentry->d_inode); +#endif ret = btrfs_update_inode(trans, root, dentry->d_inode); dir->i_sb->s_dirt = 1; } @@ -1139,7 +1143,7 @@ static unsigned char btrfs_filetype_table[] = { static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) { - struct inode *inode = filp->f_path.dentry->d_inode; + struct inode *inode = filp->f_dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_item *item; struct btrfs_dir_item *di; @@ -1554,7 +1558,11 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, if (inode->i_nlink == 0) return -ENOENT; +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + inode->i_nlink++; +#else inc_nlink(inode); +#endif mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); @@ -1825,6 +1833,9 @@ insert: if (ret == -EEXIST) { free_extent_map(em); em = NULL; + if (0 && failed_insert == 1) { + btrfs_drop_extent_cache(inode, start, end); + } failed_insert++; if (failed_insert > 5) { printk("failing to insert %Lu %Lu\n", start, end); @@ -1942,7 +1953,7 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) */ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) { - struct inode *inode = vma->vm_file->f_path.dentry->d_inode; + struct inode *inode = fdentry(vma->vm_file)->d_inode; unsigned long end; loff_t size; int ret = -EINVAL; @@ -2243,7 +2254,7 @@ static unsigned long force_ra(struct address_space *mapping, } int btrfs_defrag_file(struct file *file) { - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = fdentry(file)->d_inode; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct page *page; unsigned long last_index; @@ -2329,7 +2340,7 @@ static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) static int btrfs_ioctl_defrag(struct file *file) { - struct inode *inode = file->f_path.dentry->d_inode; + struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; switch (inode->i_mode & S_IFMT) { @@ -2350,7 +2361,7 @@ static int btrfs_ioctl_defrag(struct file *file) long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { - struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root; + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; switch (cmd) { case BTRFS_IOC_SNAP_CREATE: diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index b7323914bf5..7cab680418a 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -35,6 +35,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index 1671c731800..cd673ca9bb9 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -189,12 +189,29 @@ static struct kset btrfs_kset; int btrfs_sysfs_add_super(struct btrfs_fs_info *fs) { int error; + char *name; + char c; + int len = strlen(fs->sb->s_id) + 1; + int i; + + name = kmalloc(len, GFP_NOFS); + if (!name) { + error = -ENOMEM; + goto fail; + } + + for (i = 0; i < len; i++) { + c = fs->sb->s_id[i]; + if (c == '/' || c == '\\') + c = '!'; + name[i] = c; + } + name[len] = '\0'; fs->super_kobj.kset = &btrfs_kset; fs->super_kobj.ktype = &btrfs_super_ktype; - error = kobject_set_name(&fs->super_kobj, "%s", - fs->sb->s_id); + error = kobject_set_name(&fs->super_kobj, "%s", name); if (error) goto fail; @@ -202,9 +219,11 @@ int btrfs_sysfs_add_super(struct btrfs_fs_info *fs) if (error) goto fail; + kfree(name); return 0; fail: + kfree(name); printk(KERN_ERR "btrfs: sysfs creation for super failed\n"); return error; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 0229e69bd2f..1ad611b9f61 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -614,12 +614,20 @@ int btrfs_clean_old_snapshots(struct btrfs_root *root) } return 0; } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) +void btrfs_transaction_cleaner(void *p) +#else void btrfs_transaction_cleaner(struct work_struct *work) +#endif { +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + struct btrfs_fs_info *fs_info = p; +#else struct btrfs_fs_info *fs_info = container_of(work, struct btrfs_fs_info, trans_work.work); +#endif struct btrfs_root *root = fs_info->tree_root; struct btrfs_transaction *cur; struct btrfs_trans_handle *trans; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index ae39fcfc169..eef840bca91 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -73,7 +73,12 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, struct btrfs_root *root); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) +void btrfs_transaction_cleaner(void *p); +#else void btrfs_transaction_cleaner(struct work_struct *work); +#endif + void btrfs_transaction_flush_work(struct btrfs_root *root); void btrfs_transaction_queue_work(struct btrfs_root *root, int delay); void btrfs_init_transaction_sys(void); -- cgit v1.2.3-70-g09d2 From 4aec2b52327b3a3a20faa3f1af3102c0d97c7c92 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 18 Dec 2007 16:25:45 -0500 Subject: kmalloc a few large stack objects in the btrfs_ioctl path Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 20 +++++++++++++------ fs/btrfs/inode.c | 60 +++++++++++++++++++++++++++++++++++++++----------------- 2 files changed, 56 insertions(+), 24 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 585f279d111..7f764455f26 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -80,10 +80,14 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, int ret = 0; int level; struct btrfs_key first_key; - struct btrfs_root new_root; + struct btrfs_root *new_root; - memcpy(&new_root, root, sizeof(new_root)); - new_root.root_key.objectid = new_root_objectid; + new_root = kmalloc(sizeof(*new_root), GFP_NOFS); + if (!new_root) + return -ENOMEM; + + memcpy(new_root, root, sizeof(*new_root)); + new_root->root_key.objectid = new_root_objectid; WARN_ON(root->ref_cows && trans->transid != root->fs_info->running_transaction->transid); @@ -99,12 +103,14 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, } else { first_key.objectid = 0; } - cow = __btrfs_alloc_free_block(trans, &new_root, buf->len, + cow = __btrfs_alloc_free_block(trans, new_root, buf->len, new_root_objectid, trans->transid, first_key.objectid, level, buf->start, 0); - if (IS_ERR(cow)) + if (IS_ERR(cow)) { + kfree(new_root); return PTR_ERR(cow); + } copy_extent_buffer(cow, buf, 0, 0, cow->len); btrfs_set_header_bytenr(cow, cow->start); @@ -112,7 +118,9 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, btrfs_set_header_owner(cow, new_root_objectid); WARN_ON(btrfs_header_generation(buf) > trans->transid); - ret = btrfs_inc_ref(trans, &new_root, buf); + ret = btrfs_inc_ref(trans, new_root, buf); + kfree(new_root); + if (ret) return ret; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6a7d9160df2..1e725a48467 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2302,40 +2302,64 @@ out_unlock: static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) { - struct btrfs_ioctl_vol_args vol_args; + struct btrfs_ioctl_vol_args *vol_args; struct btrfs_dir_item *di; struct btrfs_path *path; - int namelen; u64 root_dirid; + int namelen; + int ret; - if (copy_from_user(&vol_args, arg, sizeof(vol_args))) - return -EFAULT; + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); - namelen = strlen(vol_args.name); - if (namelen > BTRFS_VOL_NAME_MAX) - return -EINVAL; - if (strchr(vol_args.name, '/')) - return -EINVAL; + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + + namelen = strlen(vol_args->name); + if (namelen > BTRFS_VOL_NAME_MAX) { + ret = -EINVAL; + goto out; + } + if (strchr(vol_args->name, '/')) { + ret = -EINVAL; + goto out; + } path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; + if (!path) { + ret = -ENOMEM; + goto out; + } root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, mutex_lock(&root->fs_info->fs_mutex); di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, root_dirid, - vol_args.name, namelen, 0); + vol_args->name, namelen, 0); mutex_unlock(&root->fs_info->fs_mutex); btrfs_free_path(path); - if (di && !IS_ERR(di)) - return -EEXIST; - if (IS_ERR(di)) - return PTR_ERR(di); + + if (di && !IS_ERR(di)) { + ret = -EEXIST; + goto out; + } + + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto out; + } if (root == root->fs_info->tree_root) - return create_subvol(root, vol_args.name, namelen); - return create_snapshot(root, vol_args.name, namelen); + ret = create_subvol(root, vol_args->name, namelen); + else + ret = create_snapshot(root, vol_args->name, namelen); +out: + kfree(vol_args); + return ret; } static int btrfs_ioctl_defrag(struct file *file) -- cgit v1.2.3-70-g09d2 From 190662b2128dd648749e197f5563e9f6bbb5e05c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 18 Dec 2007 16:25:45 -0500 Subject: Btrfs: Fix delayed allocation to avoid missing delalloc extents find_lock_delalloc_range could exit out too early Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 9 +++++++-- fs/btrfs/inode.c | 1 + 2 files changed, 8 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 0d1e59a86e4..a0dff34dd43 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1070,6 +1070,7 @@ u64 find_lock_delalloc_range(struct extent_map_tree *tree, search_again: node = tree_search(&tree->state, cur_start); if (!node || IS_ERR(node)) { + *end = (u64)-1; goto out; } @@ -1079,6 +1080,8 @@ search_again: goto out; } if (!(state->state & EXTENT_DELALLOC)) { + if (!found) + *end = state->end; goto out; } if (!found) { @@ -1841,8 +1844,10 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, &delalloc_end, 128 * 1024 * 1024); - if (nr_delalloc <= 0) - break; + if (nr_delalloc == 0) { + delalloc_start = delalloc_end + 1; + continue; + } tree->ops->fill_delalloc(inode, delalloc_start, delalloc_end); clear_extent_bit(tree, delalloc_start, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1e725a48467..55b2e142602 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -798,6 +798,7 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + WARN_ON(!PageLocked(page)); set_page_extent_mapped(page); lock_extent(em_tree, page_start, page_end, GFP_NOFS); -- cgit v1.2.3-70-g09d2 From 879c1cfc3104b632dea85d07490f1b55e54275a9 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 19 Dec 2007 09:11:48 -0500 Subject: Btrfs: Fix nodatacow extent lookup Yan Zheng noticed the offset into the extent was incorrectly being added to the extent start before trying to find it in the extent allocation tree. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 55b2e142602..1c61c85dc3d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -177,7 +177,6 @@ again: if (bytenr == 0) goto not_found; - bytenr += btrfs_file_extent_offset(leaf, item); if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) { goto not_found; } -- cgit v1.2.3-70-g09d2 From 1832a6d5ee3b1af61001cadba9e10da9e91af4a4 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 21 Dec 2007 16:27:21 -0500 Subject: Btrfs: Implement basic support for -ENOSPC This is intended to prevent accidentally filling the drive. A determined user can still make things oops. It includes some accounting of the current bytes under delayed allocation, but this will change as things get optimized Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 4 ++ fs/btrfs/disk-io.c | 5 +- fs/btrfs/extent_map.c | 20 +++++-- fs/btrfs/extent_map.h | 3 +- fs/btrfs/file.c | 28 +++++++-- fs/btrfs/inode.c | 161 +++++++++++++++++++++++++++++++++++++++++++++----- 6 files changed, 193 insertions(+), 28 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 2559f5d5d00..10129cc6656 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -346,6 +346,8 @@ struct btrfs_fs_info { int closing; u64 total_pinned; + spinlock_t delalloc_lock; + u64 delalloc_bytes; }; /* * in ram representation of the tree. extent_root is used for all allocations @@ -1115,6 +1117,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, + int for_del); int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); int btrfs_readpage(struct file *file, struct page *page); void btrfs_delete_inode(struct inode *inode); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index ebb2db624fd..eebb4fb65c6 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -223,7 +223,8 @@ static int btree_writepages(struct address_space *mapping, } else { thresh = 8 * 1024 * 1024; } - num_dirty = count_range_bits(tree, &start, thresh, EXTENT_DIRTY); + num_dirty = count_range_bits(tree, &start, (u64)-1, + thresh, EXTENT_DIRTY); if (num_dirty < thresh) { return 0; } @@ -559,6 +560,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) INIT_LIST_HEAD(&fs_info->dead_roots); INIT_LIST_HEAD(&fs_info->hashers); spin_lock_init(&fs_info->hash_lock); + spin_lock_init(&fs_info->delalloc_lock); memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); init_completion(&fs_info->kobj_unregister); @@ -570,6 +572,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->sb = sb; fs_info->mount_opt = 0; fs_info->max_extent = (u64)-1; + fs_info->delalloc_bytes = 0; fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index a0dff34dd43..2b92f107027 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1131,7 +1131,8 @@ out: } u64 count_range_bits(struct extent_map_tree *tree, - u64 *start, u64 max_bytes, unsigned long bits) + u64 *start, u64 search_end, u64 max_bytes, + unsigned long bits) { struct rb_node *node; struct extent_state *state; @@ -1139,9 +1140,14 @@ u64 count_range_bits(struct extent_map_tree *tree, u64 total_bytes = 0; int found = 0; + if (search_end <= cur_start) { + printk("search_end %Lu start %Lu\n", search_end, cur_start); + WARN_ON(1); + return 0; + } + write_lock_irq(&tree->lock); - if (bits == EXTENT_DIRTY) { - *start = 0; + if (cur_start == 0 && bits == EXTENT_DIRTY) { total_bytes = tree->dirty_bytes; goto out; } @@ -1156,8 +1162,11 @@ u64 count_range_bits(struct extent_map_tree *tree, while(1) { state = rb_entry(node, struct extent_state, rb_node); - if ((state->state & bits)) { - total_bytes += state->end - state->start + 1; + if (state->start > search_end) + break; + if (state->end >= cur_start && (state->state & bits)) { + total_bytes += min(search_end, state->end) + 1 - + max(cur_start, state->start); if (total_bytes >= max_bytes) break; if (!found) { @@ -1173,7 +1182,6 @@ out: write_unlock_irq(&tree->lock); return total_bytes; } - /* * helper function to lock both pages and extents in the tree. * pages must be locked first. diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 6e572d3e892..ea60f5447b5 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -115,7 +115,8 @@ int __init extent_map_init(void); void extent_map_exit(void); u64 count_range_bits(struct extent_map_tree *tree, - u64 *start, u64 max_bytes, unsigned long bits); + u64 *start, u64 search_end, + u64 max_bytes, unsigned long bits); int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits, int filled); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 461b09663fe..71dc2d33b6c 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -307,6 +307,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, inline_size > 32768 || inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { u64 last_end; + u64 existing_delalloc = 0; for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; @@ -316,8 +317,19 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans, last_end = (u64)(pages[num_pages -1]->index) << PAGE_CACHE_SHIFT; last_end += PAGE_CACHE_SIZE - 1; + if (start_pos < isize) { + u64 delalloc_start = start_pos; + existing_delalloc = count_range_bits(em_tree, + &delalloc_start, + end_of_last_block, (u64)-1, + EXTENT_DELALLOC); + } set_extent_delalloc(em_tree, start_pos, end_of_last_block, GFP_NOFS); + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes += (end_of_last_block + 1 - + start_pos) - existing_delalloc; + spin_unlock(&root->fs_info->delalloc_lock); } else { u64 aligned_end; /* step one, delete the existing extents in this range */ @@ -708,12 +720,12 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, current->backing_dev_info = inode->i_mapping->backing_dev_info; err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode)); if (err) - goto out; + goto out_nolock; if (count == 0) - goto out; + goto out_nolock; err = remove_suid(fdentry(file)); if (err) - goto out; + goto out_nolock; file_update_time(file); pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); @@ -758,6 +770,13 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, WARN_ON(num_pages > nrptrs); memset(pages, 0, sizeof(pages)); + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, write_bytes, 0); + mutex_unlock(&root->fs_info->fs_mutex); + if (ret) + goto out; + ret = prepare_pages(root, file, pages, num_pages, pos, first_index, last_index, write_bytes); @@ -787,8 +806,9 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, btrfs_btree_balance_dirty(root, 1); cond_resched(); } - mutex_unlock(&inode->i_mutex); out: + mutex_unlock(&inode->i_mutex); +out_nolock: kfree(pages); if (pinned[0]) page_cache_release(pinned[0]); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1c61c85dc3d..a9f5d6d417f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -72,6 +72,26 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, }; +int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, + int for_del) +{ + u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy); + u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy); + u64 thresh; + int ret = 0; + + if (for_del) + thresh = (total * 90) / 100; + else + thresh = (total * 85) / 100; + + spin_lock(&root->fs_info->delalloc_lock); + if (used + root->fs_info->delalloc_bytes + num_required > thresh) + ret = -ENOSPC; + spin_unlock(&root->fs_info->delalloc_lock); + return ret; +} + static int cow_file_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -124,6 +144,7 @@ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) u64 extent_end; u64 bytenr; u64 cow_end; + u64 loops = 0; struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_buffer *leaf; int found_type; @@ -169,6 +190,9 @@ again: btrfs_file_extent_num_bytes(leaf, item); err = 0; + if (loops && start != extent_start) + goto not_found; + if (start < extent_start || start >= extent_end) goto not_found; @@ -191,6 +215,7 @@ loop: return 0; } btrfs_release_path(root, path); + loops++; goto again; not_found: @@ -202,6 +227,7 @@ not_found: static int run_delalloc_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; + u64 num_bytes; int ret; mutex_lock(&root->fs_info->fs_mutex); @@ -209,6 +235,17 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) ret = run_delalloc_nocow(inode, start, end); else ret = cow_file_range(inode, start, end); + + spin_lock(&root->fs_info->delalloc_lock); + num_bytes = end + 1 - start; + if (root->fs_info->delalloc_bytes < num_bytes) { + printk("delalloc accounting error total %llu sub %llu\n", + root->fs_info->delalloc_bytes, num_bytes); + } else { + root->fs_info->delalloc_bytes -= num_bytes; + } + spin_unlock(&root->fs_info->delalloc_lock); + mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -547,10 +584,15 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) struct btrfs_root *root; struct btrfs_trans_handle *trans; int ret; - unsigned long nr; + unsigned long nr = 0; root = BTRFS_I(dir)->root; mutex_lock(&root->fs_info->fs_mutex); + + ret = btrfs_check_free_space(root, 1, 1); + if (ret) + goto fail; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -558,25 +600,29 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) nr = trans->blocks_used; btrfs_end_transaction(trans, root); +fail: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); - return ret; } static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) { struct inode *inode = dentry->d_inode; - int err; + int err = 0; int ret; struct btrfs_root *root = BTRFS_I(dir)->root; struct btrfs_trans_handle *trans; - unsigned long nr; + unsigned long nr = 0; if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) return -ENOTEMPTY; mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 1); + if (ret) + goto fail; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -588,6 +634,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); +fail: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); @@ -792,17 +839,29 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, size_t zero_start) { char *kaddr; - int ret = 0; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct btrfs_root *root = BTRFS_I(inode)->root; u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + u64 existing_delalloc; + u64 delalloc_start; + int ret = 0; WARN_ON(!PageLocked(page)); set_page_extent_mapped(page); lock_extent(em_tree, page_start, page_end, GFP_NOFS); + delalloc_start = page_start; + existing_delalloc = count_range_bits(&BTRFS_I(inode)->extent_tree, + &delalloc_start, page_end, + PAGE_CACHE_SIZE, EXTENT_DELALLOC); set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start, page_end, GFP_NOFS); + + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - existing_delalloc; + spin_unlock(&root->fs_info->delalloc_lock); + if (zero_start != PAGE_CACHE_SIZE) { kaddr = kmap(page); memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); @@ -881,6 +940,12 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) if (attr->ia_size <= pos) goto out; + mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + mutex_unlock(&root->fs_info->fs_mutex); + if (err) + goto fail; + btrfs_truncate_page(inode->i_mapping, inode->i_size); lock_extent(em_tree, pos, block_end, GFP_NOFS); @@ -906,7 +971,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) } out: err = inode_setattr(inode, attr); - +fail: return err; } void btrfs_delete_inode(struct inode *inode) @@ -1440,16 +1505,20 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; - struct inode *inode; + struct inode *inode = NULL; int err; int drop_inode = 0; u64 objectid; - unsigned long nr; + unsigned long nr = 0; if (!new_valid_dev(rdev)) return -EINVAL; mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + if (err) + goto fail; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -1480,6 +1549,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); +fail: mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { @@ -1495,13 +1565,16 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; - struct inode *inode; + struct inode *inode = NULL; int err; int drop_inode = 0; - unsigned long nr; + unsigned long nr = 0; u64 objectid; mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + if (err) + goto fail; trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -1535,6 +1608,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); +fail: mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { @@ -1551,7 +1625,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; struct inode *inode = old_dentry->d_inode; - unsigned long nr; + unsigned long nr = 0; int err; int drop_inode = 0; @@ -1564,6 +1638,9 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, inc_nlink(inode); #endif mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + if (err) + goto fail; trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -1582,6 +1659,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, nr = trans->blocks_used; btrfs_end_transaction(trans, root); +fail: mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { @@ -1603,6 +1681,10 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) unsigned long nr = 1; mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + if (err) + goto out_unlock; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -1869,6 +1951,15 @@ static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) static int btrfs_prepare_write(struct file *file, struct page *page, unsigned from, unsigned to) { + struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; + int err; + + mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); + mutex_lock(&root->fs_info->fs_mutex); + if (err) + return -ENOSPC; + return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree, page->mapping->host, page, from, to, btrfs_get_extent); @@ -1880,6 +1971,7 @@ int btrfs_readpage(struct file *file, struct page *page) tree = &BTRFS_I(page->mapping->host)->extent_tree; return extent_read_full_page(tree, page, btrfs_get_extent); } + static int btrfs_writepage(struct page *page, struct writeback_control *wbc) { struct extent_map_tree *tree; @@ -1954,11 +2046,20 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) { struct inode *inode = fdentry(vma->vm_file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; unsigned long end; loff_t size; - int ret = -EINVAL; + int ret; u64 page_start; + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); + mutex_lock(&root->fs_info->fs_mutex); + if (ret) + goto out; + + ret = -EINVAL; + down_read(&BTRFS_I(inode)->root->snap_sem); lock_page(page); wait_on_page_writeback(page); @@ -1982,6 +2083,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) out_unlock: up_read(&BTRFS_I(inode)->root->snap_sem); unlock_page(page); +out: return ret; } @@ -2046,6 +2148,10 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen) unsigned long nr = 1; mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 0); + if (ret) + goto fail_commit; + trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); @@ -2162,7 +2268,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) int ret; int err; u64 objectid; - unsigned long nr; + unsigned long nr = 0; if (!root->ref_cows) return -EINVAL; @@ -2172,6 +2278,10 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb); mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 0); + if (ret) + goto fail_unlock; + trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); @@ -2229,7 +2339,7 @@ fail: if (err && !ret) ret = err; - +fail_unlock: mutex_unlock(&root->fs_info->fs_mutex); up_write(&root->snap_sem); btrfs_btree_balance_dirty(root, nr); @@ -2255,6 +2365,7 @@ static unsigned long force_ra(struct address_space *mapping, int btrfs_defrag_file(struct file *file) { struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct page *page; unsigned long last_index; @@ -2262,6 +2373,13 @@ int btrfs_defrag_file(struct file *file) { u64 page_start; u64 page_end; unsigned long i; + int ret; + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, inode->i_size, 0); + mutex_unlock(&root->fs_info->fs_mutex); + if (ret) + return -ENOSPC; mutex_lock(&inode->i_mutex); last_index = inode->i_size >> PAGE_CACHE_SHIFT; @@ -2522,6 +2640,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, } mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 0); + if (ret) + goto out_unlock; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, new_dir); @@ -2553,6 +2675,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, out_fail: btrfs_free_path(path); btrfs_end_transaction(trans, root); +out_unlock: mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -2564,7 +2687,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = BTRFS_I(dir)->root; struct btrfs_path *path; struct btrfs_key key; - struct inode *inode; + struct inode *inode = NULL; int err; int drop_inode = 0; u64 objectid; @@ -2573,12 +2696,17 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, unsigned long ptr; struct btrfs_file_extent_item *ei; struct extent_buffer *leaf; - unsigned long nr; + unsigned long nr = 0; name_len = strlen(symname) + 1; if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) return -ENAMETOOLONG; + mutex_lock(&root->fs_info->fs_mutex); + err = btrfs_check_free_space(root, 1, 0); + if (err) + goto out_fail; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); @@ -2645,6 +2773,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); +out_fail: mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { inode_dec_link_count(inode); -- cgit v1.2.3-70-g09d2 From 5d4fb734b4cb89a3902d6de309af0f6f3e63fe73 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 21 Dec 2007 16:27:23 -0500 Subject: Btrfs: Fix an off by one in the extent_map prepare write code Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 4 ++-- fs/btrfs/inode.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 2b92f107027..f53d8c97621 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -2270,7 +2270,7 @@ int extent_prepare_write(struct extent_map_tree *tree, kunmap_atomic(kaddr, KM_USER0); } if ((em->block_start != EXTENT_MAP_HOLE && - em->block_start != EXTENT_MAP_INLINE) && + em->block_start != EXTENT_MAP_INLINE) && !isnew && !PageUptodate(page) && (block_off_end > to || block_off_start < from) && !test_range_bit(tree, block_start, cur_end, @@ -2279,7 +2279,7 @@ int extent_prepare_write(struct extent_map_tree *tree, u64 extent_offset = block_start - em->start; size_t iosize; sector = (em->block_start + extent_offset) >> 9; - iosize = (cur_end - block_start + blocksize - 1) & + iosize = (cur_end - block_start + blocksize) & ~((u64)blocksize - 1); /* * we've already got the extent locked, but we diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a9f5d6d417f..5f5b7b89b14 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1956,7 +1956,7 @@ static int btrfs_prepare_write(struct file *file, struct page *page, mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); - mutex_lock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->fs_mutex); if (err) return -ENOSPC; -- cgit v1.2.3-70-g09d2 From edbd8d4efe4ddaf29a175ae504e2c9a05a96ebee Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 21 Dec 2007 16:27:24 -0500 Subject: Btrfs: Support for online FS resize (grow and shrink) Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 10 ++ fs/btrfs/disk-io.c | 32 +++- fs/btrfs/disk-io.h | 2 + fs/btrfs/extent-tree.c | 480 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/inode.c | 105 ++++++++++- fs/btrfs/ioctl.h | 2 + fs/btrfs/root-tree.c | 5 + fs/btrfs/super.c | 7 +- 8 files changed, 629 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 10129cc6656..b83a1bec346 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -508,6 +508,8 @@ BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item, used, 64); BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item, used, 64); +BTRFS_SETGET_FUNCS(disk_block_group_flags, struct btrfs_block_group_item, + flags, 8); /* struct btrfs_inode_ref */ BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16); @@ -960,6 +962,9 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, int level, u64 hint, u64 empty_size); +int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 new_size); +int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size); int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 bytenr, @@ -1117,6 +1122,9 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +unsigned long btrfs_force_ra(struct address_space *mapping, + struct file_ra_state *ra, struct file *file, + pgoff_t offset, pgoff_t last_index); int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int for_del); int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); @@ -1162,4 +1170,6 @@ void btrfs_sysfs_del_super(struct btrfs_fs_info *root); ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size); int btrfs_delete_xattrs(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode); +/* super.c */ +u64 btrfs_parse_size(char *str); #endif diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index eebb4fb65c6..de026d9d9b2 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -471,13 +471,17 @@ insert: return root; } -struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, - struct btrfs_key *location, - const char *name, int namelen) +struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, + struct btrfs_key *location) { struct btrfs_root *root; int ret; + if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) + return fs_info->tree_root; + if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) + return fs_info->extent_root; + root = radix_tree_lookup(&fs_info->fs_roots_radix, (unsigned long)location->objectid); if (root) @@ -494,6 +498,23 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, kfree(root); return ERR_PTR(ret); } + ret = btrfs_find_dead_roots(fs_info->tree_root, + root->root_key.objectid, root); + BUG_ON(ret); + + return root; +} + +struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, + struct btrfs_key *location, + const char *name, int namelen) +{ + struct btrfs_root *root; + int ret; + + root = btrfs_read_fs_root_no_name(fs_info, location); + if (!root) + return NULL; ret = btrfs_set_root_name(root, name, namelen); if (ret) { @@ -509,11 +530,6 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, kfree(root); return ERR_PTR(ret); } - - ret = btrfs_find_dead_roots(fs_info->tree_root, - root->root_key.objectid, root); - BUG_ON(ret); - return root; } #if 0 diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 33d2ccfd74d..8c3cfd02901 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -39,6 +39,8 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, const char *name, int namelen); struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, struct btrfs_key *location); +struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, + struct btrfs_key *location); int btrfs_insert_dev_radix(struct btrfs_root *root, struct block_device *bdev, u64 device_id, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 1412d556313..de0fb0743cf 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -18,6 +18,7 @@ #include #include +#include #include "hash.h" #include "ctree.h" #include "disk-io.h" @@ -1622,6 +1623,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, int pending_ret; u64 super_used, root_used; u64 search_start = 0; + u64 new_hint; struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; struct btrfs_extent_item extent_item; @@ -1629,6 +1631,10 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, btrfs_set_stack_extent_refs(&extent_item, 1); + new_hint = max(hint_byte, 16ULL * 1024 * 1024 * 1024); + if (new_hint < btrfs_super_total_bytes(&info->super_copy)) + hint_byte = new_hint; + WARN_ON(num_bytes < root->sectorsize); ret = find_free_extent(trans, root, num_bytes, empty_size, search_start, search_end, hint_byte, ins, @@ -2100,6 +2106,480 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) return 0; } +static int relocate_inode_pages(struct inode *inode, u64 start, u64 len) +{ + u64 page_start; + u64 page_end; + u64 delalloc_start; + u64 existing_delalloc; + unsigned long last_index; + unsigned long first_index; + unsigned long i; + struct page *page; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct file_ra_state ra; + + mutex_lock(&inode->i_mutex); + first_index = start >> PAGE_CACHE_SHIFT; + last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; + + memset(&ra, 0, sizeof(ra)); + file_ra_state_init(&ra, inode->i_mapping); + btrfs_force_ra(inode->i_mapping, &ra, NULL, first_index, last_index); + + for (i = first_index; i <= last_index; i++) { + page = grab_cache_page(inode->i_mapping, i); + if (!page) + goto out_unlock; + if (!PageUptodate(page)) { + btrfs_readpage(NULL, page); + lock_page(page); + if (!PageUptodate(page)) { + unlock_page(page); + page_cache_release(page); + goto out_unlock; + } + } + page_start = (u64)page->index << PAGE_CACHE_SHIFT; + page_end = page_start + PAGE_CACHE_SIZE - 1; + + lock_extent(em_tree, page_start, page_end, GFP_NOFS); + + delalloc_start = page_start; + existing_delalloc = + count_range_bits(&BTRFS_I(inode)->extent_tree, + &delalloc_start, page_end, + PAGE_CACHE_SIZE, EXTENT_DELALLOC); + + set_extent_delalloc(em_tree, page_start, + page_end, GFP_NOFS); + + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - + existing_delalloc; + spin_unlock(&root->fs_info->delalloc_lock); + + unlock_extent(em_tree, page_start, page_end, GFP_NOFS); + set_page_dirty(page); + unlock_page(page); + page_cache_release(page); + } + +out_unlock: + mutex_unlock(&inode->i_mutex); + return 0; +} + +static int relocate_one_reference(struct btrfs_root *extent_root, + struct btrfs_path *path, + struct btrfs_key *extent_key, + u64 ref_root, u64 ref_gen, u64 ref_objectid, + u64 ref_offset) +{ + struct inode *inode; + struct btrfs_root *found_root; + struct btrfs_key root_location; + int ret; + + root_location.objectid = ref_root; + if (ref_gen == 0) + root_location.offset = 0; + else + root_location.offset = (u64)-1; + root_location.type = BTRFS_ROOT_ITEM_KEY; + + found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, + &root_location); + BUG_ON(!found_root); + + if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { + mutex_unlock(&extent_root->fs_info->fs_mutex); + inode = btrfs_iget_locked(extent_root->fs_info->sb, + ref_objectid, found_root); + if (inode->i_state & I_NEW) { + /* the inode and parent dir are two different roots */ + BTRFS_I(inode)->root = found_root; + BTRFS_I(inode)->location.objectid = ref_objectid; + BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; + BTRFS_I(inode)->location.offset = 0; + btrfs_read_locked_inode(inode); + unlock_new_inode(inode); + + } + /* this can happen if the reference is not against + * the latest version of the tree root + */ + if (is_bad_inode(inode)) { + mutex_lock(&extent_root->fs_info->fs_mutex); + goto out; + } + relocate_inode_pages(inode, ref_offset, extent_key->offset); + /* FIXME, data=ordered will help get rid of this */ + filemap_fdatawrite(inode->i_mapping); + iput(inode); + mutex_lock(&extent_root->fs_info->fs_mutex); + } else { + struct btrfs_trans_handle *trans; + struct btrfs_key found_key; + struct extent_buffer *eb; + int level; + int i; + + trans = btrfs_start_transaction(found_root, 1); + eb = read_tree_block(found_root, extent_key->objectid, + extent_key->offset); + level = btrfs_header_level(eb); + + if (level == 0) + btrfs_item_key_to_cpu(eb, &found_key, 0); + else + btrfs_node_key_to_cpu(eb, &found_key, 0); + + free_extent_buffer(eb); + + path->lowest_level = level; + path->reada = 0; + ret = btrfs_search_slot(trans, found_root, &found_key, path, + 0, 1); + path->lowest_level = 0; + for (i = level; i < BTRFS_MAX_LEVEL; i++) { + if (!path->nodes[i]) + break; + free_extent_buffer(path->nodes[i]); + path->nodes[i] = NULL; + } + btrfs_release_path(found_root, path); + btrfs_end_transaction(trans, found_root); + } + +out: + return 0; +} + +static int relocate_one_extent(struct btrfs_root *extent_root, + struct btrfs_path *path, + struct btrfs_key *extent_key) +{ + struct btrfs_key key; + struct btrfs_key found_key; + struct btrfs_extent_ref *ref; + struct extent_buffer *leaf; + u64 ref_root; + u64 ref_gen; + u64 ref_objectid; + u64 ref_offset; + u32 nritems; + u32 item_size; + int ret = 0; + + key.objectid = extent_key->objectid; + key.type = BTRFS_EXTENT_REF_KEY; + key.offset = 0; + + while(1) { + ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); + + BUG_ON(ret == 0); + + if (ret < 0) + goto out; + + ret = 0; + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + if (path->slots[0] == nritems) + goto out; + + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.objectid != extent_key->objectid) + break; + + if (found_key.type != BTRFS_EXTENT_REF_KEY) + break; + + key.offset = found_key.offset + 1; + item_size = btrfs_item_size_nr(leaf, path->slots[0]); + + ref = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_extent_ref); + ref_root = btrfs_ref_root(leaf, ref); + ref_gen = btrfs_ref_generation(leaf, ref); + ref_objectid = btrfs_ref_objectid(leaf, ref); + ref_offset = btrfs_ref_offset(leaf, ref); + btrfs_release_path(extent_root, path); + + ret = relocate_one_reference(extent_root, path, + extent_key, ref_root, ref_gen, + ref_objectid, ref_offset); + if (ret) + goto out; + } + ret = 0; +out: + btrfs_release_path(extent_root, path); + return ret; +} + +static int find_overlapping_extent(struct btrfs_root *root, + struct btrfs_path *path, u64 new_size) +{ + struct btrfs_key found_key; + struct extent_buffer *leaf; + int ret; + + while(1) { + if (path->slots[0] == 0) { + ret = btrfs_prev_leaf(root, path); + if (ret == 1) { + return 1; + } + if (ret < 0) + return ret; + } else { + path->slots[0]--; + } + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.type == BTRFS_EXTENT_ITEM_KEY) { + if (found_key.objectid + found_key.offset > new_size) + return 0; + else + return 1; + } + } + return 1; +} + +int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *tree_root = root->fs_info->tree_root; + struct btrfs_path *path; + u64 cur_byte; + u64 total_found; + u64 ptr; + struct btrfs_fs_info *info = root->fs_info; + struct extent_map_tree *block_group_cache; + struct btrfs_key key; + struct btrfs_key found_key = { 0, 0, 0 }; + struct extent_buffer *leaf; + u32 nritems; + int ret; + int slot; + + btrfs_set_super_total_bytes(&info->super_copy, new_size); + block_group_cache = &info->block_group_cache; + path = btrfs_alloc_path(); + root = root->fs_info->extent_root; + +again: + total_found = 0; + key.objectid = new_size; + cur_byte = key.objectid; + key.offset = 0; + key.type = 0; + while(1) { + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto out; +next: + leaf = path->nodes[0]; + if (key.objectid == new_size - 1) { + ret = find_overlapping_extent(root, path, new_size); + if (ret != 0) { + btrfs_release_path(root, path); + ret = btrfs_search_slot(NULL, root, &key, + path, 0, 0); + if (ret < 0) + goto out; + } + } + nritems = btrfs_header_nritems(leaf); + ret = 0; + slot = path->slots[0]; + if (slot < nritems) + btrfs_item_key_to_cpu(leaf, &found_key, slot); + if (slot == nritems || + btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY) { + path->slots[0]++; + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) + goto out; + if (ret == 1) { + ret = 0; + break; + } + } + goto next; + } + btrfs_item_key_to_cpu(leaf, &found_key, slot); + if (found_key.objectid + found_key.offset <= cur_byte) + continue; + total_found++; + cur_byte = found_key.objectid + found_key.offset; + key.objectid = cur_byte; + btrfs_release_path(root, path); + ret = relocate_one_extent(root, path, &found_key); + } + + btrfs_release_path(root, path); + + if (total_found > 0) { + trans = btrfs_start_transaction(tree_root, 1); + btrfs_commit_transaction(trans, tree_root); + + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_clean_old_snapshots(tree_root); + mutex_lock(&root->fs_info->fs_mutex); + + trans = btrfs_start_transaction(tree_root, 1); + btrfs_commit_transaction(trans, tree_root); + goto again; + } + + trans = btrfs_start_transaction(root, 1); + key.objectid = new_size; + key.offset = 0; + key.type = 0; + while(1) { + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) + goto out; +bg_next: + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + ret = 0; + slot = path->slots[0]; + if (slot < nritems) + btrfs_item_key_to_cpu(leaf, &found_key, slot); + if (slot == nritems || + btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) { + if (slot < nritems) { + printk("shrinker found key %Lu %u %Lu\n", + found_key.objectid, found_key.type, + found_key.offset); + path->slots[0]++; + } + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) + break; + if (ret == 1) { + ret = 0; + break; + } + } + goto bg_next; + } + btrfs_item_key_to_cpu(leaf, &found_key, slot); + ret = get_state_private(&info->block_group_cache, + found_key.objectid, &ptr); + if (!ret) + kfree((void *)(unsigned long)ptr); + + clear_extent_bits(&info->block_group_cache, found_key.objectid, + found_key.objectid + found_key.offset - 1, + (unsigned int)-1, GFP_NOFS); + + key.objectid = found_key.objectid + 1; + btrfs_del_item(trans, root, path); + btrfs_release_path(root, path); + } + clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, + GFP_NOFS); + btrfs_commit_transaction(trans, root); +out: + btrfs_free_path(path); + return ret; +} + +int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 new_size) +{ + struct btrfs_path *path; + u64 nr = 0; + u64 cur_byte; + u64 old_size; + struct btrfs_block_group_cache *cache; + struct btrfs_block_group_item *item; + struct btrfs_fs_info *info = root->fs_info; + struct extent_map_tree *block_group_cache; + struct btrfs_key key; + struct extent_buffer *leaf; + int ret; + int bit; + + old_size = btrfs_super_total_bytes(&info->super_copy); + block_group_cache = &info->block_group_cache; + + root = info->extent_root; + + cache = btrfs_lookup_block_group(root->fs_info, old_size - 1); + + cur_byte = cache->key.objectid + cache->key.offset; + if (cur_byte >= new_size) + goto set_size; + + key.offset = BTRFS_BLOCK_GROUP_SIZE; + btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + while(cur_byte < new_size) { + key.objectid = cur_byte; + ret = btrfs_insert_empty_item(trans, root, path, &key, + sizeof(struct btrfs_block_group_item)); + BUG_ON(ret); + leaf = path->nodes[0]; + item = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_block_group_item); + + btrfs_set_disk_block_group_used(leaf, item, 0); + if (nr % 3) { + btrfs_set_disk_block_group_flags(leaf, item, + BTRFS_BLOCK_GROUP_DATA); + } else { + btrfs_set_disk_block_group_flags(leaf, item, 0); + } + nr++; + + cache = kmalloc(sizeof(*cache), GFP_NOFS); + BUG_ON(!cache); + + read_extent_buffer(leaf, &cache->item, (unsigned long)item, + sizeof(cache->item)); + + memcpy(&cache->key, &key, sizeof(key)); + cache->cached = 0; + cache->pinned = 0; + cur_byte = key.objectid + key.offset; + btrfs_release_path(root, path); + + if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { + bit = BLOCK_GROUP_DATA; + cache->data = BTRFS_BLOCK_GROUP_DATA; + } else { + bit = BLOCK_GROUP_METADATA; + cache->data = 0; + } + + /* use EXTENT_LOCKED to prevent merging */ + set_extent_bits(block_group_cache, key.objectid, + key.objectid + key.offset - 1, + bit | EXTENT_LOCKED, GFP_NOFS); + set_state_private(block_group_cache, key.objectid, + (unsigned long)cache); + } + btrfs_free_path(path); +set_size: + btrfs_set_super_total_bytes(&info->super_copy, new_size); + return 0; +} + int btrfs_read_block_groups(struct btrfs_root *root) { struct btrfs_path *path; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5f5b7b89b14..f6a20112d9e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2346,7 +2346,7 @@ fail_unlock: return ret; } -static unsigned long force_ra(struct address_space *mapping, +unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, pgoff_t offset, pgoff_t last_index) { @@ -2372,6 +2372,8 @@ int btrfs_defrag_file(struct file *file) { unsigned long ra_index = 0; u64 page_start; u64 page_end; + u64 delalloc_start; + u64 existing_delalloc; unsigned long i; int ret; @@ -2385,8 +2387,9 @@ int btrfs_defrag_file(struct file *file) { last_index = inode->i_size >> PAGE_CACHE_SHIFT; for (i = 0; i <= last_index; i++) { if (i == ra_index) { - ra_index = force_ra(inode->i_mapping, &file->f_ra, - file, ra_index, last_index); + ra_index = btrfs_force_ra(inode->i_mapping, + &file->f_ra, + file, ra_index, last_index); } page = grab_cache_page(inode->i_mapping, i); if (!page) @@ -2404,8 +2407,19 @@ int btrfs_defrag_file(struct file *file) { page_end = page_start + PAGE_CACHE_SIZE - 1; lock_extent(em_tree, page_start, page_end, GFP_NOFS); + delalloc_start = page_start; + existing_delalloc = + count_range_bits(&BTRFS_I(inode)->extent_tree, + &delalloc_start, page_end, + PAGE_CACHE_SIZE, EXTENT_DELALLOC); set_extent_delalloc(em_tree, page_start, page_end, GFP_NOFS); + + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - + existing_delalloc; + spin_unlock(&root->fs_info->delalloc_lock); + unlock_extent(em_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); unlock_page(page); @@ -2418,6 +2432,89 @@ out_unlock: return 0; } +static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) +{ + u64 new_size; + u64 old_size; + struct btrfs_ioctl_vol_args *vol_args; + struct btrfs_trans_handle *trans; + char *sizestr; + int ret = 0; + int namelen; + int mod = 0; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + namelen = strlen(vol_args->name); + if (namelen > BTRFS_VOL_NAME_MAX) { + ret = -EINVAL; + goto out; + } + + sizestr = vol_args->name; + if (!strcmp(sizestr, "max")) + new_size = root->fs_info->sb->s_bdev->bd_inode->i_size; + else { + if (sizestr[0] == '-') { + mod = -1; + sizestr++; + } else if (sizestr[0] == '+') { + mod = 1; + sizestr++; + } + new_size = btrfs_parse_size(sizestr); + if (new_size == 0) { + ret = -EINVAL; + goto out; + } + } + + mutex_lock(&root->fs_info->fs_mutex); + old_size = btrfs_super_total_bytes(&root->fs_info->super_copy); + + if (mod < 0) { + if (new_size > old_size) { + ret = -EINVAL; + goto out_unlock; + } + new_size = old_size - new_size; + } else if (mod > 0) { + new_size = old_size + new_size; + } + + if (new_size < 256 * 1024 * 1024) { + ret = -EINVAL; + goto out_unlock; + } + if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) { + ret = -EFBIG; + goto out_unlock; + } + new_size = (new_size / root->sectorsize) * root->sectorsize; + +printk("new size is %Lu\n", new_size); + if (new_size > old_size) { + trans = btrfs_start_transaction(root, 1); + ret = btrfs_grow_extent_tree(trans, root, new_size); + btrfs_commit_transaction(trans, root); + } else { + ret = btrfs_shrink_extent_tree(root, new_size); + } + +out_unlock: + mutex_unlock(&root->fs_info->fs_mutex); +out: + kfree(vol_args); + return ret; +} + static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) { struct btrfs_ioctl_vol_args *vol_args; @@ -2510,6 +2607,8 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_snap_create(root, (void __user *)arg); case BTRFS_IOC_DEFRAG: return btrfs_ioctl_defrag(file); + case BTRFS_IOC_RESIZE: + return btrfs_ioctl_resize(root, (void __user *)arg); } return -ENOTTY; diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 8bc47dec286..8c6290665d4 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -30,4 +30,6 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_DEFRAG _IOW(BTRFS_IOCTL_MAGIC, 2, \ struct btrfs_ioctl_vol_args) +#define BTRFS_IOC_RESIZE _IOW(BTRFS_IOCTL_MAGIC, 3, \ + struct btrfs_ioctl_vol_args) #endif diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 88bcdd33f56..e2b9b86af69 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -179,6 +179,11 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, ret = btrfs_search_slot(trans, root, key, path, -1, 1); if (ret < 0) goto out; + if (ret) { +btrfs_print_leaf(root, path->nodes[0]); +printk("failed to del %Lu %u %Lu\n", key->objectid, key->type, key->offset); + + } BUG_ON(ret != 0); leaf = path->nodes[0]; ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 7cab680418a..3bbd7f66569 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -74,9 +74,9 @@ static match_table_t tokens = { {Opt_err, NULL} }; -static unsigned long parse_size(char *str) +u64 btrfs_parse_size(char *str) { - unsigned long res; + u64 res; int mult = 1; char *end; char last; @@ -150,7 +150,8 @@ static int parse_options (char * options, if (info) { char *num = match_strdup(&args[0]); if (num) { - info->max_extent = parse_size(num); + info->max_extent = + btrfs_parse_size(num); kfree(num); info->max_extent = max_t(u64, -- cgit v1.2.3-70-g09d2 From e52ec0eb62e76ea584c8438ef92933e9b815db80 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 21 Dec 2007 16:36:24 -0500 Subject: Btrfs: Fix NULL block groups on reading the inode Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 4 ++-- fs/btrfs/inode.c | 4 ++++ 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index de0fb0743cf..a421069e998 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1630,11 +1630,11 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; btrfs_set_stack_extent_refs(&extent_item, 1); - +#if 0 new_hint = max(hint_byte, 16ULL * 1024 * 1024 * 1024); if (new_hint < btrfs_super_total_bytes(&info->super_copy)) hint_byte = new_hint; - +#endif WARN_ON(num_bytes < root->sectorsize); ret = find_free_extent(trans, root, num_bytes, empty_size, search_start, search_end, hint_byte, ins, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f6a20112d9e..3c7bf3bddca 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -402,6 +402,10 @@ void btrfs_read_locked_inode(struct inode *inode) BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, alloc_group_block); + if (!BTRFS_I(inode)->block_group) { + BTRFS_I(inode)->block_group = btrfs_find_block_group(root, + NULL, 0, 0, 0); + } btrfs_free_path(path); inode_item = NULL; -- cgit v1.2.3-70-g09d2 From 8f662a76c6af8eb367fa519e9bb9766040d9cea8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 2 Jan 2008 10:01:11 -0500 Subject: Btrfs: Add readahead to the online shrinker, and a mount -o alloc_start= for testing Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 9 ++++++--- fs/btrfs/ctree.h | 1 + fs/btrfs/extent-tree.c | 9 +++++---- fs/btrfs/inode.c | 2 +- fs/btrfs/super.c | 16 +++++++++++++++- 5 files changed, 28 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 2b3ffa707ea..8fa92a2d981 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2654,9 +2654,9 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, */ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) { + u64 bytenr; int slot; int level = 1; - u64 bytenr; struct extent_buffer *c; struct extent_buffer *next = NULL; @@ -2687,11 +2687,14 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) level--; c = path->nodes[level]; free_extent_buffer(c); + slot = btrfs_header_nritems(next); + if (slot != 0) + slot--; path->nodes[level] = next; - path->slots[level] = 0; + path->slots[level] = slot; if (!level) break; - next = read_tree_block(root, btrfs_node_blockptr(next, 0), + next = read_tree_block(root, btrfs_node_blockptr(next, slot), btrfs_level_size(root, level - 1)); } return 0; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b83a1bec346..d6e5c19969b 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -324,6 +324,7 @@ struct btrfs_fs_info { u64 last_trans_committed; unsigned long mount_opt; u64 max_extent; + u64 alloc_start; struct btrfs_transaction *running_transaction; struct btrfs_super_block super_copy; struct extent_buffer *sb_buffer; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a421069e998..d6243c2afc7 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1630,11 +1630,11 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; btrfs_set_stack_extent_refs(&extent_item, 1); -#if 0 - new_hint = max(hint_byte, 16ULL * 1024 * 1024 * 1024); + + new_hint = max(hint_byte, root->fs_info->alloc_start); if (new_hint < btrfs_super_total_bytes(&info->super_copy)) hint_byte = new_hint; -#endif + WARN_ON(num_bytes < root->sectorsize); ret = find_free_extent(trans, root, num_bytes, empty_size, search_start, search_end, hint_byte, ins, @@ -2239,7 +2239,7 @@ static int relocate_one_reference(struct btrfs_root *extent_root, free_extent_buffer(eb); path->lowest_level = level; - path->reada = 0; + path->reada = 2; ret = btrfs_search_slot(trans, found_root, &found_key, path, 0, 1); path->lowest_level = 0; @@ -2372,6 +2372,7 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) block_group_cache = &info->block_group_cache; path = btrfs_alloc_path(); root = root->fs_info->extent_root; + path->reada = 2; again: total_found = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3c7bf3bddca..d6de4696188 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2058,7 +2058,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); - mutex_lock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->fs_mutex); if (ret) goto out; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 3bbd7f66569..f1b56eb77b1 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -63,7 +63,8 @@ static void btrfs_put_super (struct super_block * sb) } enum { - Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_err, + Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, + Opt_alloc_start, Opt_err, }; static match_table_t tokens = { @@ -71,6 +72,7 @@ static match_table_t tokens = { {Opt_nodatasum, "nodatasum"}, {Opt_nodatacow, "nodatacow"}, {Opt_max_extent, "max_extent=%s"}, + {Opt_alloc_start, "alloc_start=%s"}, {Opt_err, NULL} }; @@ -162,6 +164,18 @@ static int parse_options (char * options, } } break; + case Opt_alloc_start: + if (info) { + char *num = match_strdup(&args[0]); + if (num) { + info->alloc_start = + btrfs_parse_size(num); + kfree(num); + printk("btrfs: allocations start at " + "%Lu\n", info->alloc_start); + } + } + break; default: break; } -- cgit v1.2.3-70-g09d2 From 56b453c92fdf51fd3283a2dc2dfbedf36f516031 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Jan 2008 09:08:27 -0500 Subject: Btrfs: Explicitly send a root objectid to count_snapshots_in_path Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 16 ++++++---------- fs/btrfs/inode.c | 39 +++++---------------------------------- 2 files changed, 11 insertions(+), 44 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index d6243c2afc7..91c2b27d968 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -688,7 +688,7 @@ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, struct btrfs_path *path; u64 bytenr; u64 found_objectid; - u64 root_objectid = 0; + u64 root_objectid = root->root_key.objectid; u32 total_count = 0; u32 cur_count; u32 refs; @@ -749,15 +749,11 @@ again: struct btrfs_extent_ref); found_objectid = btrfs_ref_root(l, ref_item); - if (found_objectid != root_objectid) - total_count++; - - if (total_count > 1) - goto out; - - if (root_objectid == 0) - root_objectid = found_objectid; - + if (found_objectid != root_objectid) { + total_count = 2; + break; + } + total_count = 1; path->slots[0]++; } if (cur_count == 0) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d6de4696188..5003a86510a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1770,6 +1770,11 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, again: em = lookup_extent_mapping(em_tree, start, end); if (em) { + if (em->start > start) { + printk("get_extent start %Lu em start %Lu\n", + start, em->start); + WARN_ON(1); + } goto out; } if (!em) { @@ -1952,23 +1957,6 @@ static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) return extent_bmap(mapping, iblock, btrfs_get_extent); } -static int btrfs_prepare_write(struct file *file, struct page *page, - unsigned from, unsigned to) -{ - struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; - int err; - - mutex_lock(&root->fs_info->fs_mutex); - err = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); - mutex_unlock(&root->fs_info->fs_mutex); - if (err) - return -ENOSPC; - - return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree, - page->mapping->host, page, from, to, - btrfs_get_extent); -} - int btrfs_readpage(struct file *file, struct page *page) { struct extent_map_tree *tree; @@ -2120,21 +2108,6 @@ static void btrfs_truncate(struct inode *inode) btrfs_btree_balance_dirty(root, nr); } -int btrfs_commit_write(struct file *file, struct page *page, - unsigned from, unsigned to) -{ - loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; - struct inode *inode = page->mapping->host; - - btrfs_cow_one_page(inode, page, PAGE_CACHE_SIZE); - - if (pos > inode->i_size) { - i_size_write(inode, pos); - mark_inode_dirty(inode); - } - return 0; -} - static int create_subvol(struct btrfs_root *root, char *name, int namelen) { struct btrfs_trans_handle *trans; @@ -2930,8 +2903,6 @@ static struct address_space_operations btrfs_aops = { .writepages = btrfs_writepages, .readpages = btrfs_readpages, .sync_page = block_sync_page, - .prepare_write = btrfs_prepare_write, - .commit_write = btrfs_commit_write, .bmap = btrfs_bmap, .invalidatepage = btrfs_invalidatepage, .releasepage = btrfs_releasepage, -- cgit v1.2.3-70-g09d2 From 4313b3994d719fcdeb7e661473019ca3d62e829b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Jan 2008 09:08:48 -0500 Subject: Btrfs: Reduce stack usage in the resizer, fix 32 bit compiles Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 5 ++++ fs/btrfs/extent-tree.c | 74 ++++++++++++++++++++++++++------------------------ fs/btrfs/inode.c | 6 ++-- fs/btrfs/transaction.c | 20 ++++++++++---- 5 files changed, 63 insertions(+), 43 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index d6e5c19969b..9873975ce0e 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -386,6 +386,7 @@ struct btrfs_root { int defrag_running; int defrag_level; char *name; + int in_sysfs; }; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index de026d9d9b2..67d9fd72886 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -374,6 +374,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->highest_inode = 0; root->last_inode_alloc = 0; root->name = NULL; + root->in_sysfs = 0; memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); @@ -516,6 +517,9 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, if (!root) return NULL; + if (root->in_sysfs) + return root; + ret = btrfs_set_root_name(root, name, namelen); if (ret) { free_extent_buffer(root->node); @@ -530,6 +534,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, kfree(root); return ERR_PTR(ret); } + root->in_sysfs = 1; return root; } #if 0 diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 91c2b27d968..6137f06091e 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -751,7 +751,7 @@ again: if (found_objectid != root_objectid) { total_count = 2; - break; + goto out; } total_count = 1; path->slots[0]++; @@ -760,8 +760,6 @@ again: total_count = 0; goto out; } - if (total_count > 1) - goto out; if (level >= 0 && root->node == count_path->nodes[level]) goto out; level++; @@ -2109,22 +2107,23 @@ static int relocate_inode_pages(struct inode *inode, u64 start, u64 len) u64 delalloc_start; u64 existing_delalloc; unsigned long last_index; - unsigned long first_index; unsigned long i; struct page *page; struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - struct file_ra_state ra; + struct file_ra_state *ra; + + ra = kzalloc(sizeof(*ra), GFP_NOFS); mutex_lock(&inode->i_mutex); - first_index = start >> PAGE_CACHE_SHIFT; + i = start >> PAGE_CACHE_SHIFT; last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; - memset(&ra, 0, sizeof(ra)); - file_ra_state_init(&ra, inode->i_mapping); - btrfs_force_ra(inode->i_mapping, &ra, NULL, first_index, last_index); + file_ra_state_init(ra, inode->i_mapping); + btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index); + kfree(ra); - for (i = first_index; i <= last_index; i++) { + for (; i <= last_index; i++) { page = grab_cache_page(inode->i_mapping, i); if (!page) goto out_unlock; @@ -2167,27 +2166,43 @@ out_unlock: return 0; } +/* + * note, this releases the path + */ static int relocate_one_reference(struct btrfs_root *extent_root, struct btrfs_path *path, - struct btrfs_key *extent_key, - u64 ref_root, u64 ref_gen, u64 ref_objectid, - u64 ref_offset) + struct btrfs_key *extent_key) { struct inode *inode; struct btrfs_root *found_root; - struct btrfs_key root_location; + struct btrfs_key *root_location; + struct btrfs_extent_ref *ref; + u64 ref_root; + u64 ref_gen; + u64 ref_objectid; + u64 ref_offset; int ret; - root_location.objectid = ref_root; + ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_extent_ref); + ref_root = btrfs_ref_root(path->nodes[0], ref); + ref_gen = btrfs_ref_generation(path->nodes[0], ref); + ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); + ref_offset = btrfs_ref_offset(path->nodes[0], ref); + btrfs_release_path(extent_root, path); + + root_location = kmalloc(sizeof(*root_location), GFP_NOFS); + root_location->objectid = ref_root; if (ref_gen == 0) - root_location.offset = 0; + root_location->offset = 0; else - root_location.offset = (u64)-1; - root_location.type = BTRFS_ROOT_ITEM_KEY; + root_location->offset = (u64)-1; + root_location->type = BTRFS_ROOT_ITEM_KEY; found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, - &root_location); + root_location); BUG_ON(!found_root); + kfree(root_location); if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { mutex_unlock(&extent_root->fs_info->fs_mutex); @@ -2259,12 +2274,7 @@ static int relocate_one_extent(struct btrfs_root *extent_root, { struct btrfs_key key; struct btrfs_key found_key; - struct btrfs_extent_ref *ref; struct extent_buffer *leaf; - u64 ref_root; - u64 ref_gen; - u64 ref_objectid; - u64 ref_offset; u32 nritems; u32 item_size; int ret = 0; @@ -2297,17 +2307,7 @@ static int relocate_one_extent(struct btrfs_root *extent_root, key.offset = found_key.offset + 1; item_size = btrfs_item_size_nr(leaf, path->slots[0]); - ref = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_extent_ref); - ref_root = btrfs_ref_root(leaf, ref); - ref_gen = btrfs_ref_generation(leaf, ref); - ref_objectid = btrfs_ref_objectid(leaf, ref); - ref_offset = btrfs_ref_offset(leaf, ref); - btrfs_release_path(extent_root, path); - - ret = relocate_one_reference(extent_root, path, - extent_key, ref_root, ref_gen, - ref_objectid, ref_offset); + ret = relocate_one_reference(extent_root, path, extent_key); if (ret) goto out; } @@ -2354,7 +2354,6 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) struct btrfs_path *path; u64 cur_byte; u64 total_found; - u64 ptr; struct btrfs_fs_info *info = root->fs_info; struct extent_map_tree *block_group_cache; struct btrfs_key key; @@ -2377,6 +2376,7 @@ again: key.offset = 0; key.type = 0; while(1) { + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto out; @@ -2441,6 +2441,8 @@ next: key.offset = 0; key.type = 0; while(1) { + u64 ptr; + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) goto out; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5003a86510a..b62f35e862b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2108,7 +2108,8 @@ static void btrfs_truncate(struct inode *inode) btrfs_btree_balance_dirty(root, nr); } -static int create_subvol(struct btrfs_root *root, char *name, int namelen) +static int noinline create_subvol(struct btrfs_root *root, char *name, + int namelen) { struct btrfs_trans_handle *trans; struct btrfs_key key; @@ -2492,7 +2493,8 @@ out: return ret; } -static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg) +static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root, + void __user *arg) { struct btrfs_ioctl_vol_args *vol_args; struct btrfs_dir_item *di; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 1ad611b9f61..02721eea9a7 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -481,12 +481,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction *cur_trans; struct btrfs_transaction *prev_trans = NULL; struct list_head dirty_fs_roots; - struct extent_map_tree pinned_copy; + struct extent_map_tree *pinned_copy; DEFINE_WAIT(wait); int ret; - extent_map_tree_init(&pinned_copy, - root->fs_info->btree_inode->i_mapping, GFP_NOFS); INIT_LIST_HEAD(&dirty_fs_roots); mutex_lock(&root->fs_info->trans_mutex); @@ -507,6 +505,14 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->fs_mutex); return 0; } + + pinned_copy = kmalloc(sizeof(*pinned_copy), GFP_NOFS); + if (!pinned_copy) + return -ENOMEM; + + extent_map_tree_init(pinned_copy, + root->fs_info->btree_inode->i_mapping, GFP_NOFS); + trans->transaction->in_commit = 1; cur_trans = trans->transaction; if (cur_trans->list.prev != &root->fs_info->trans_list) { @@ -568,16 +574,20 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, &root->fs_info->super_copy, 0, sizeof(root->fs_info->super_copy)); - btrfs_copy_pinned(root, &pinned_copy); + btrfs_copy_pinned(root, pinned_copy); mutex_unlock(&root->fs_info->trans_mutex); mutex_unlock(&root->fs_info->fs_mutex); ret = btrfs_write_and_wait_transaction(trans, root); BUG_ON(ret); write_ctree_super(trans, root); + mutex_lock(&root->fs_info->fs_mutex); - btrfs_finish_extent_commit(trans, root, &pinned_copy); + btrfs_finish_extent_commit(trans, root, pinned_copy); mutex_lock(&root->fs_info->trans_mutex); + + kfree(pinned_copy); + cur_trans->commit_done = 1; root->fs_info->last_trans_committed = cur_trans->transid; wake_up(&cur_trans->commit_wait); -- cgit v1.2.3-70-g09d2 From f9ef6604ace23a6fcd698e08b58a883d6009157b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Jan 2008 09:22:38 -0500 Subject: Btrfs: 32 bit compile fixes for the resizer and enospc checks Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 4 +++- fs/btrfs/inode.c | 10 +++++++--- 2 files changed, 10 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 6137f06091e..54f2b1d8bdf 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2501,6 +2501,7 @@ int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, u64 nr = 0; u64 cur_byte; u64 old_size; + unsigned long rem; struct btrfs_block_group_cache *cache; struct btrfs_block_group_item *item; struct btrfs_fs_info *info = root->fs_info; @@ -2538,7 +2539,8 @@ int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, struct btrfs_block_group_item); btrfs_set_disk_block_group_used(leaf, item, 0); - if (nr % 3) { + div_long_long_rem(nr, 3, &rem); + if (rem) { btrfs_set_disk_block_group_flags(leaf, item, BTRFS_BLOCK_GROUP_DATA); } else { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b62f35e862b..8a2001bbf10 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -81,9 +81,11 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int ret = 0; if (for_del) - thresh = (total * 90) / 100; + thresh = total * 90; else - thresh = (total * 85) / 100; + thresh = total * 85; + + do_div(thresh, 100); spin_lock(&root->fs_info->delalloc_lock); if (used + root->fs_info->delalloc_bytes + num_required > thresh) @@ -2475,7 +2477,9 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) ret = -EFBIG; goto out_unlock; } - new_size = (new_size / root->sectorsize) * root->sectorsize; + + do_div(new_size, root->sectorsize); + new_size *= root->sectorsize; printk("new size is %Lu\n", new_size); if (new_size > old_size) { -- cgit v1.2.3-70-g09d2 From bd09835d9aad9c7b664cddc8435cc37b86077971 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Jan 2008 13:23:19 -0500 Subject: count_snapshots: Properly update the leaf pointer after btrfs_next_leaf Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 4 ++-- fs/btrfs/inode.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 4957cface9a..b801eb4b3ee 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -693,7 +693,6 @@ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, u64 root_objectid = root->root_key.objectid; u32 total_count = 0; u32 cur_count; - u32 refs; u32 nritems; int ret; struct btrfs_key key; @@ -729,8 +728,8 @@ again: } item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); - refs = btrfs_extent_refs(l, item); while (1) { + l = path->nodes[0]; nritems = btrfs_header_nritems(l); if (path->slots[0] >= nritems) { ret = btrfs_next_leaf(extent_root, path); @@ -741,6 +740,7 @@ again: btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); if (found_key.objectid != bytenr) break; + if (found_key.type != BTRFS_EXTENT_REF_KEY) { path->slots[0]++; continue; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8a2001bbf10..6e2d03626fb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -208,7 +208,7 @@ again: } start = extent_end; - } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + } else { goto not_found; } loop: -- cgit v1.2.3-70-g09d2 From d666746207a01546e55bdaa4b721d1890faaf6e8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Jan 2008 14:51:00 -0500 Subject: Btrfs: Change st_blocksize to 4k Some programs (python) do rwm cycles at the granularity returned by stat. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- fs/btrfs/root-tree.c | 9 ++++++++- 2 files changed, 9 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6e2d03626fb..6d0cd9a7a61 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2701,7 +2701,7 @@ static int btrfs_getattr(struct vfsmount *mnt, { struct inode *inode = dentry->d_inode; generic_fillattr(inode, stat); - stat->blksize = 256 * 1024; + stat->blksize = PAGE_CACHE_SIZE; return 0; } diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index e2b9b86af69..08f19ec8809 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -75,7 +75,14 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root ret = btrfs_search_slot(trans, root, key, path, 0, 1); if (ret < 0) goto out; - BUG_ON(ret != 0); + + if (ret != 0) { + btrfs_print_leaf(root, path->nodes[0]); + printk("unable to update root key %Lu %u %Lu\n", + key->objectid, key->type, key->offset); + BUG_ON(1); + } + l = path->nodes[0]; slot = path->slots[0]; ptr = btrfs_item_ptr_offset(l, slot); -- cgit v1.2.3-70-g09d2 From dc17ff8f11d129db9e83ab7244769e4eae05e14d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 8 Jan 2008 15:46:30 -0500 Subject: Btrfs: Add data=ordered support This forces file data extents down the disk along with the metadata that references them. The current implementation is fairly simple, and just writes out all of the dirty pages in an inode before the commit. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/ctree.c | 6 +- fs/btrfs/ctree.h | 7 +- fs/btrfs/disk-io.c | 16 +++- fs/btrfs/disk-io.h | 2 + fs/btrfs/extent-tree.c | 4 +- fs/btrfs/file.c | 5 +- fs/btrfs/inode.c | 46 ++++++---- fs/btrfs/ordered-data.c | 221 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/ordered-data.h | 39 +++++++++ fs/btrfs/transaction.c | 58 +++++++++++++ fs/btrfs/transaction.h | 8 +- 13 files changed, 387 insertions(+), 28 deletions(-) create mode 100644 fs/btrfs/ordered-data.c create mode 100644 fs/btrfs/ordered-data.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index d5804c5ca07..ab9a9f8edbf 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -5,7 +5,7 @@ obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ - extent_map.o sysfs.o struct-funcs.o xattr.o acl.o + extent_map.o sysfs.o struct-funcs.o xattr.o acl.o ordered-data.o #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \ # root-tree.o dir-item.o hash.o file-item.o inode-item.o \ diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index d1d5af471c3..f27e633f174 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -29,6 +29,7 @@ struct btrfs_inode { struct extent_map_tree extent_tree; struct inode vfs_inode; + u64 ordered_trans; /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 35c57074a37..43d23148a4f 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -221,7 +221,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, struct extent_buffer **cow_ret) { u64 search_start; + u64 header_trans; int ret; + if (trans->transaction != root->fs_info->running_transaction) { printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, root->fs_info->running_transaction->transid); @@ -232,7 +234,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, root->fs_info->generation); WARN_ON(1); } - if (btrfs_header_generation(buf) == trans->transid) { + + header_trans = btrfs_header_generation(buf); + if (header_trans == trans->transid) { *cow_ret = buf; return 0; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 9873975ce0e..b55dba58dfa 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -16,8 +16,8 @@ * Boston, MA 021110-1307, USA. */ -#ifndef __BTRFS__ -#define __BTRFS__ +#ifndef __BTRFS_CTREE__ +#define __BTRFS_CTREE__ #include #include @@ -363,7 +363,6 @@ struct btrfs_root { struct inode *inode; struct kobject root_kobj; struct completion kobj_unregister; - struct rw_semaphore snap_sem; u64 objectid; u64 last_trans; @@ -1142,6 +1141,8 @@ void btrfs_destroy_cachep(void); long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); +struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, + u64 root_objectid); int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to); struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index a6170ff19e7..34cf1f1f47b 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -406,7 +406,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); memset(&root->root_kobj, 0, sizeof(root->root_kobj)); init_completion(&root->kobj_unregister); - init_rwsem(&root->snap_sem); root->defrag_running = 0; root->defrag_level = 0; root->root_key.objectid = objectid; @@ -498,6 +497,21 @@ insert: return root; } +struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, + u64 root_objectid) +{ + struct btrfs_root *root; + + if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) + return fs_info->tree_root; + if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) + return fs_info->extent_root; + + root = radix_tree_lookup(&fs_info->fs_roots_radix, + (unsigned long)root_objectid); + return root; +} + struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, struct btrfs_key *location) { diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 8c3cfd02901..dae9fba8efc 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -34,6 +34,8 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize); +struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, + u64 root_objectid); struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_key *location, const char *name, int namelen); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c906bb19b21..68137cd8506 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1195,7 +1195,9 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, if (btrfs_buffer_uptodate(buf)) { u64 transid = root->fs_info->running_transaction->transid; - if (btrfs_header_generation(buf) == transid) { + u64 header_transid = + btrfs_header_generation(buf); + if (header_transid == transid) { free_extent_buffer(buf); return 1; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 94c93373cb7..0a5f4defe59 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -34,6 +34,7 @@ #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" +#include "ordered-data.h" #include "ioctl.h" #include "print-tree.h" @@ -329,6 +330,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, root->fs_info->delalloc_bytes += (end_of_last_block + 1 - start_pos) - existing_delalloc; spin_unlock(&root->fs_info->delalloc_lock); + btrfs_add_ordered_inode(inode); } else { u64 aligned_end; /* step one, delete the existing extents in this range */ @@ -724,8 +726,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); - down_read(&BTRFS_I(inode)->root->snap_sem); - mutex_lock(&inode->i_mutex); first_index = pos >> PAGE_CACHE_SHIFT; last_index = (pos + count) >> PAGE_CACHE_SHIFT; @@ -804,7 +804,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, } out: mutex_unlock(&inode->i_mutex); - up_read(&BTRFS_I(inode)->root->snap_sem); out_nolock: kfree(pages); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6d0cd9a7a61..6d6e1ac0a9a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -135,6 +135,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; } + btrfs_add_ordered_inode(inode); out: btrfs_end_transaction(trans, root); return ret; @@ -367,8 +368,8 @@ void btrfs_read_locked_inode(struct inode *inode) path = btrfs_alloc_path(); BUG_ON(!path); mutex_lock(&root->fs_info->fs_mutex); - memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); + ret = btrfs_lookup_inode(NULL, root, path, &location, 0); if (ret) goto make_bad; @@ -898,7 +899,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) if ((offset & (blocksize - 1)) == 0) goto out; - down_read(&root->snap_sem); ret = -ENOMEM; page = grab_cache_page(mapping, index); if (!page) @@ -917,7 +917,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) unlock_page(page); page_cache_release(page); - up_read(&BTRFS_I(inode)->root->snap_sem); out: return ret; } @@ -1146,6 +1145,19 @@ static int btrfs_find_actor(struct inode *inode, void *opaque) args->root == BTRFS_I(inode)->root); } +struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, + u64 root_objectid) +{ + struct btrfs_iget_args args; + args.ino = objectid; + args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid); + + if (!args.root) + return NULL; + + return ilookup5(s, objectid, btrfs_find_actor, (void *)&args); +} + struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root) { @@ -1336,7 +1348,6 @@ read_dir_items: d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)]; btrfs_dir_item_key_to_cpu(leaf, di, &location); - over = filldir(dirent, name_ptr, name_len, found_key.offset, location.objectid, @@ -2054,7 +2065,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) ret = -EINVAL; - down_read(&BTRFS_I(inode)->root->snap_sem); lock_page(page); wait_on_page_writeback(page); size = i_size_read(inode); @@ -2075,7 +2085,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) ret = btrfs_cow_one_page(inode, page, end); out_unlock: - up_read(&BTRFS_I(inode)->root->snap_sem); unlock_page(page); out: return ret; @@ -2118,7 +2127,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name, struct btrfs_root_item root_item; struct btrfs_inode_item *inode_item; struct extent_buffer *leaf; - struct btrfs_root *new_root; + struct btrfs_root *new_root = root; struct inode *inode; struct inode *dir; int ret; @@ -2230,7 +2239,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name, goto fail; fail: nr = trans->blocks_used; - err = btrfs_commit_transaction(trans, root); + err = btrfs_commit_transaction(trans, new_root); if (err && !ret) ret = err; fail_commit: @@ -2253,10 +2262,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (!root->ref_cows) return -EINVAL; - down_write(&root->snap_sem); - freeze_bdev(root->fs_info->sb->s_bdev); - thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb); - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 0); if (ret) @@ -2264,6 +2269,9 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); + err = btrfs_commit_transaction(trans, root); + + trans = btrfs_start_transaction(root, 1); ret = btrfs_update_inode(trans, root, root->inode); if (ret) @@ -2272,9 +2280,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, 0, &objectid); if (ret) - goto fail; - - memcpy(&new_root_item, &root->root_item, + goto fail; memcpy(&new_root_item, &root->root_item, sizeof(new_root_item)); key.objectid = objectid; @@ -2285,12 +2291,20 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); free_extent_buffer(tmp); + /* write the ordered inodes to force all delayed allocations to + * be filled. Once this is done, we can copy the root + */ + mutex_lock(&root->fs_info->trans_mutex); + btrfs_write_ordered_inodes(trans, root); + mutex_unlock(&root->fs_info->trans_mutex); + btrfs_copy_root(trans, root, root->node, &tmp, objectid); btrfs_set_root_bytenr(&new_root_item, tmp->start); btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &new_root_item); +printk("new root %Lu node %Lu\n", objectid, tmp->start); free_extent_buffer(tmp); if (ret) goto fail; @@ -2321,7 +2335,6 @@ fail: ret = err; fail_unlock: mutex_unlock(&root->fs_info->fs_mutex); - up_write(&root->snap_sem); btrfs_btree_balance_dirty(root, nr); return ret; } @@ -2608,6 +2621,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) if (!ei) return NULL; ei->last_trans = 0; + ei->ordered_trans = 0; return &ei->vfs_inode; } diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c new file mode 100644 index 00000000000..411aba84d30 --- /dev/null +++ b/fs/btrfs/ordered-data.c @@ -0,0 +1,221 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include "ctree.h" +#include "transaction.h" +#include "btrfs_inode.h" + +struct tree_entry { + u64 root_objectid; + u64 objectid; + struct rb_node rb_node; +}; + +/* + * returns > 0 if entry passed (root, objectid) is > entry, + * < 0 if (root, objectid) < entry and zero if they are equal + */ +static int comp_entry(struct tree_entry *entry, u64 root_objectid, + u64 objectid) +{ + if (root_objectid < entry->root_objectid) + return -1; + if (root_objectid > entry->root_objectid) + return 1; + if (objectid < entry->objectid) + return -1; + if (objectid > entry->objectid) + return 1; + return 0; +} + +static struct rb_node *tree_insert(struct rb_root *root, u64 root_objectid, + u64 objectid, struct rb_node *node) +{ + struct rb_node ** p = &root->rb_node; + struct rb_node * parent = NULL; + struct tree_entry *entry; + int comp; + + while(*p) { + parent = *p; + entry = rb_entry(parent, struct tree_entry, rb_node); + + comp = comp_entry(entry, root_objectid, objectid); + if (comp < 0) + p = &(*p)->rb_left; + else if (comp > 0) + p = &(*p)->rb_right; + else + return parent; + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct rb_node *__tree_search(struct rb_root *root, u64 root_objectid, + u64 objectid, struct rb_node **prev_ret) +{ + struct rb_node * n = root->rb_node; + struct rb_node *prev = NULL; + struct tree_entry *entry; + struct tree_entry *prev_entry = NULL; + int comp; + + while(n) { + entry = rb_entry(n, struct tree_entry, rb_node); + prev = n; + prev_entry = entry; + comp = comp_entry(entry, root_objectid, objectid); + + if (comp < 0) + n = n->rb_left; + else if (comp > 0) + n = n->rb_right; + else + return n; + } + if (!prev_ret) + return NULL; + + while(prev && comp_entry(prev_entry, root_objectid, objectid) >= 0) { + prev = rb_next(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *prev_ret = prev; + return NULL; +} + +static inline struct rb_node *tree_search(struct rb_root *root, + u64 root_objectid, u64 objectid) +{ + struct rb_node *prev; + struct rb_node *ret; + ret = __tree_search(root, root_objectid, objectid, &prev); + if (!ret) + return prev; + return ret; +} + +int btrfs_add_ordered_inode(struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 root_objectid = root->root_key.objectid; + u64 transid = root->fs_info->running_transaction->transid; + struct tree_entry *entry; + struct rb_node *node; + struct btrfs_ordered_inode_tree *tree; + + if (transid <= BTRFS_I(inode)->ordered_trans) + return 0; + + tree = &root->fs_info->running_transaction->ordered_inode_tree; + + read_lock(&tree->lock); + node = __tree_search(&tree->tree, root_objectid, inode->i_ino, NULL); + read_unlock(&tree->lock); + if (node) { + return 0; + } + + entry = kmalloc(sizeof(*entry), GFP_NOFS); + if (!entry) + return -ENOMEM; + + write_lock(&tree->lock); + entry->objectid = inode->i_ino; + entry->root_objectid = root_objectid; + + node = tree_insert(&tree->tree, root_objectid, + inode->i_ino, &entry->rb_node); + + BTRFS_I(inode)->ordered_trans = transid; + + write_unlock(&tree->lock); + if (node) + kfree(entry); + return 0; +} + +int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, + u64 *root_objectid, u64 *objectid) +{ + struct tree_entry *entry; + struct rb_node *node; + + write_lock(&tree->lock); + node = tree_search(&tree->tree, *root_objectid, *objectid); + if (!node) { + write_unlock(&tree->lock); + return 0; + } + entry = rb_entry(node, struct tree_entry, rb_node); + + while(comp_entry(entry, *root_objectid, *objectid) >= 0) { + node = rb_next(node); + if (!node) + break; + entry = rb_entry(node, struct tree_entry, rb_node); + } + if (!node) { + write_unlock(&tree->lock); + return 0; + } + + *root_objectid = entry->root_objectid; + *objectid = entry->objectid; + write_unlock(&tree->lock); + return 1; +} + +int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, + u64 *root_objectid, u64 *objectid) +{ + struct tree_entry *entry; + struct rb_node *node; + + write_lock(&tree->lock); + node = tree_search(&tree->tree, *root_objectid, *objectid); + if (!node) { + write_unlock(&tree->lock); + return 0; + } + + entry = rb_entry(node, struct tree_entry, rb_node); + while(comp_entry(entry, *root_objectid, *objectid) >= 0) { + node = rb_next(node); + if (!node) + break; + entry = rb_entry(node, struct tree_entry, rb_node); + } + if (!node) { + write_unlock(&tree->lock); + return 0; + } + + *root_objectid = entry->root_objectid; + *objectid = entry->objectid; + rb_erase(node, &tree->tree); + write_unlock(&tree->lock); + kfree(entry); + return 1; +} diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h new file mode 100644 index 00000000000..aaf9eb14271 --- /dev/null +++ b/fs/btrfs/ordered-data.h @@ -0,0 +1,39 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_ORDERED_DATA__ +#define __BTRFS_ORDERED_DATA__ + +struct btrfs_ordered_inode_tree { + rwlock_t lock; + struct rb_root tree; +}; + +static inline void +btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t) +{ + rwlock_init(&t->lock); + t->tree.rb_node = NULL; +} + +int btrfs_add_ordered_inode(struct inode *inode); +int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, + u64 *root_objectid, u64 *objectid); +int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, + u64 *root_objectid, u64 *objectid); +#endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 02721eea9a7..3ed5868e7c0 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -67,6 +67,7 @@ static int join_transaction(struct btrfs_root *root) cur_trans->commit_done = 0; cur_trans->start_time = get_seconds(); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); + btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree); extent_map_tree_init(&cur_trans->dirty_pages, root->fs_info->btree_inode->i_mapping, GFP_NOFS); @@ -473,6 +474,60 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, return ret; } +int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + struct btrfs_transaction *cur_trans = trans->transaction; + struct inode *inode; + u64 root_objectid = 0; + u64 objectid = 0; + u64 transid = trans->transid; + int ret; + +printk("write ordered trans %Lu\n", transid); + while(1) { + ret = btrfs_find_first_ordered_inode( + &cur_trans->ordered_inode_tree, + &root_objectid, &objectid); + if (!ret) + break; + + mutex_unlock(&root->fs_info->trans_mutex); + mutex_unlock(&root->fs_info->fs_mutex); + inode = btrfs_ilookup(root->fs_info->sb, objectid, + root_objectid); + if (inode) { + if (S_ISREG(inode->i_mode)) + filemap_fdatawrite(inode->i_mapping); + iput(inode); + } + mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->trans_mutex); + } + while(1) { + root_objectid = 0; + objectid = 0; + ret = btrfs_find_del_first_ordered_inode( + &cur_trans->ordered_inode_tree, + &root_objectid, &objectid); + if (!ret) + break; + mutex_unlock(&root->fs_info->trans_mutex); + mutex_unlock(&root->fs_info->fs_mutex); + inode = btrfs_ilookup(root->fs_info->sb, objectid, + root_objectid); + if (inode) { + if (S_ISREG(inode->i_mode)) + filemap_write_and_wait(inode->i_mapping); + iput(inode); + } + mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->trans_mutex); + } +printk("done write ordered trans %Lu\n", transid); + return 0; +} + int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) { @@ -550,10 +605,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); finish_wait(&cur_trans->writer_wait, &wait); + ret = btrfs_write_ordered_inodes(trans, root); + } while (cur_trans->num_writers > 1 || (cur_trans->num_joined != joined)); WARN_ON(cur_trans != trans->transaction); + ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, &dirty_fs_roots); BUG_ON(ret); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index eef840bca91..c157ddbe9d1 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -16,9 +16,10 @@ * Boston, MA 021110-1307, USA. */ -#ifndef __TRANSACTION__ -#define __TRANSACTION__ +#ifndef __BTRFS_TRANSACTION__ +#define __BTRFS_TRANSACTION__ #include "btrfs_inode.h" +#include "ordered-data.h" struct btrfs_transaction { u64 transid; @@ -30,6 +31,7 @@ struct btrfs_transaction { struct list_head list; struct extent_map_tree dirty_pages; unsigned long start_time; + struct btrfs_ordered_inode_tree ordered_inode_tree; wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; }; @@ -90,4 +92,6 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly); int btrfs_clean_old_snapshots(struct btrfs_root *root); int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); +int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, + struct btrfs_root *root); #endif -- cgit v1.2.3-70-g09d2 From 3063d29f2a4d4a4e9fa1ec77c124514f287c6da7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 8 Jan 2008 15:46:30 -0500 Subject: Btrfs: Move snapshot creation to commit time It is very difficult to create a consistent snapshot of the btree when other writers may update the btree before the commit is done. This changes the snapshot creation to happen during the commit, while no other updates are possible. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 79 ++++++++++-------------------------------------- fs/btrfs/transaction.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/transaction.h | 7 +++++ 3 files changed, 100 insertions(+), 67 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6d6e1ac0a9a..10cece11dbd 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2250,13 +2250,10 @@ fail_commit: static int create_snapshot(struct btrfs_root *root, char *name, int namelen) { + struct btrfs_pending_snapshot *pending_snapshot; struct btrfs_trans_handle *trans; - struct btrfs_key key; - struct btrfs_root_item new_root_item; - struct extent_buffer *tmp; int ret; int err; - u64 objectid; unsigned long nr = 0; if (!root->ref_cows) @@ -2267,72 +2264,26 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail_unlock; + pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); + if (!pending_snapshot) { + ret = -ENOMEM; + goto fail_unlock; + } + pending_snapshot->name = kstrndup(name, namelen, GFP_NOFS); + if (!pending_snapshot->name) { + ret = -ENOMEM; + kfree(pending_snapshot); + goto fail_unlock; + } trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - err = btrfs_commit_transaction(trans, root); - - trans = btrfs_start_transaction(root, 1); + pending_snapshot->root = root; + list_add(&pending_snapshot->list, + &trans->transaction->pending_snapshots); ret = btrfs_update_inode(trans, root, root->inode); - if (ret) - goto fail; - - ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, - 0, &objectid); - if (ret) - goto fail; memcpy(&new_root_item, &root->root_item, - sizeof(new_root_item)); - - key.objectid = objectid; - key.offset = 1; - btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); - - extent_buffer_get(root->node); - btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); - free_extent_buffer(tmp); - - /* write the ordered inodes to force all delayed allocations to - * be filled. Once this is done, we can copy the root - */ - mutex_lock(&root->fs_info->trans_mutex); - btrfs_write_ordered_inodes(trans, root); - mutex_unlock(&root->fs_info->trans_mutex); - - btrfs_copy_root(trans, root, root->node, &tmp, objectid); - - btrfs_set_root_bytenr(&new_root_item, tmp->start); - btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp)); - ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, - &new_root_item); -printk("new root %Lu node %Lu\n", objectid, tmp->start); - free_extent_buffer(tmp); - if (ret) - goto fail; - - /* - * insert the directory item - */ - key.offset = (u64)-1; - ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, - name, namelen, - root->fs_info->sb->s_root->d_inode->i_ino, - &key, BTRFS_FT_DIR); - - if (ret) - goto fail; - - ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, - name, namelen, objectid, - root->fs_info->sb->s_root->d_inode->i_ino); - - if (ret) - goto fail; -fail: - nr = trans->blocks_used; err = btrfs_commit_transaction(trans, root); - if (err && !ret) - ret = err; fail_unlock: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 3ed5868e7c0..dc9865323e3 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -66,6 +66,7 @@ static int join_transaction(struct btrfs_root *root) cur_trans->use_count = 1; cur_trans->commit_done = 0; cur_trans->start_time = get_seconds(); + INIT_LIST_HEAD(&cur_trans->pending_snapshots); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree); extent_map_tree_init(&cur_trans->dirty_pages, @@ -481,10 +482,8 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, struct inode *inode; u64 root_objectid = 0; u64 objectid = 0; - u64 transid = trans->transid; int ret; -printk("write ordered trans %Lu\n", transid); while(1) { ret = btrfs_find_first_ordered_inode( &cur_trans->ordered_inode_tree, @@ -524,7 +523,80 @@ printk("write ordered trans %Lu\n", transid); mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); } -printk("done write ordered trans %Lu\n", transid); + return 0; +} + +static int create_pending_snapshot(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info, + struct btrfs_pending_snapshot *pending) +{ + struct btrfs_key key; + struct btrfs_root_item new_root_item; + struct btrfs_root *tree_root = fs_info->tree_root; + struct btrfs_root *root = pending->root; + struct extent_buffer *tmp; + int ret; + u64 objectid; + + ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid); + if (ret) + goto fail; + + memcpy(&new_root_item, &root->root_item, sizeof(new_root_item)); + + key.objectid = objectid; + key.offset = 1; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + + extent_buffer_get(root->node); + btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); + free_extent_buffer(tmp); + + btrfs_copy_root(trans, root, root->node, &tmp, objectid); + + btrfs_set_root_bytenr(&new_root_item, tmp->start); + btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp)); + ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, + &new_root_item); + free_extent_buffer(tmp); + if (ret) + goto fail; + + /* + * insert the directory item + */ + key.offset = (u64)-1; + ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, + pending->name, strlen(pending->name), + root->fs_info->sb->s_root->d_inode->i_ino, + &key, BTRFS_FT_DIR); + + if (ret) + goto fail; + + ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, + pending->name, strlen(pending->name), objectid, + root->fs_info->sb->s_root->d_inode->i_ino); +fail: + return ret; +} + +static int create_pending_snapshots(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info) +{ + struct btrfs_pending_snapshot *pending; + struct list_head *head = &trans->transaction->pending_snapshots; + int ret; + + while(!list_empty(head)) { + pending = list_entry(head->next, + struct btrfs_pending_snapshot, list); + ret = create_pending_snapshot(trans, fs_info, pending); + BUG_ON(ret); + list_del(&pending->list); + kfree(pending->name); + kfree(pending); + } return 0; } @@ -610,6 +682,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, } while (cur_trans->num_writers > 1 || (cur_trans->num_joined != joined)); + ret = create_pending_snapshots(trans, root->fs_info); + BUG_ON(ret); + WARN_ON(cur_trans != trans->transaction); ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index c157ddbe9d1..fd52e9b2392 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -34,6 +34,7 @@ struct btrfs_transaction { struct btrfs_ordered_inode_tree ordered_inode_tree; wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; + struct list_head pending_snapshots; }; struct btrfs_trans_handle { @@ -46,6 +47,12 @@ struct btrfs_trans_handle { u64 alloc_exclude_nr; }; +struct btrfs_pending_snapshot { + struct btrfs_root *root; + char *name; + struct list_head list; +}; + static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans, struct inode *inode) -- cgit v1.2.3-70-g09d2 From e2008b61401ecb467a8ce1788fcd2116ae1cfbc1 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 8 Jan 2008 15:46:30 -0500 Subject: Btrfs: Add some simple throttling to wait for data=ordered and snapshot deletion Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 7 +++++++ fs/btrfs/disk-io.h | 1 + fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 12 ++++++++++++ fs/btrfs/transaction.c | 4 ++++ 6 files changed, 26 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b55dba58dfa..f0fb1978553 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -345,6 +345,7 @@ struct btrfs_fs_info { struct completion kobj_unregister; int do_barriers; int closing; + unsigned long throttles; u64 total_pinned; spinlock_t delalloc_lock; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 34cf1f1f47b..e0940a39ff0 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -631,6 +631,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; fs_info->sb = sb; + fs_info->throttles = 0; fs_info->mount_opt = 0; fs_info->max_extent = (u64)-1; fs_info->delalloc_bytes = 0; @@ -889,6 +890,12 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); } +void btrfs_throttle(struct btrfs_root *root) +{ + if (root->fs_info->throttles) + congestion_wait(WRITE, HZ/10); +} + void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) { balance_dirty_pages_ratelimited_nr( diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index dae9fba8efc..828f3a2081b 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -64,4 +64,5 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf); int btrfs_read_buffer(struct extent_buffer *buf); u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len); void btrfs_csum_final(u32 crc, char *result); +void btrfs_throttle(struct btrfs_root *root); #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 0a5f4defe59..897242e87fa 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -800,6 +800,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1) btrfs_btree_balance_dirty(root, 1); + btrfs_throttle(root); cond_resched(); } out: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 10cece11dbd..2cb2dd32407 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -610,6 +610,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) fail: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return ret; } @@ -644,6 +645,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) fail: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); if (ret && !err) err = ret; @@ -1010,6 +1012,7 @@ void btrfs_delete_inode(struct inode *inode) btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return; no_delete_lock: @@ -1017,6 +1020,7 @@ no_delete_lock: btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); no_delete: clear_inode(inode); } @@ -1574,6 +1578,7 @@ fail: iput(inode); } btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return err; } @@ -1633,6 +1638,7 @@ fail: iput(inode); } btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return err; } @@ -1684,6 +1690,7 @@ fail: iput(inode); } btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return err; } @@ -1752,6 +1759,7 @@ out_unlock: if (drop_on_err) iput(inode); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return err; } @@ -2117,6 +2125,7 @@ static void btrfs_truncate(struct inode *inode) BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); } static int noinline create_subvol(struct btrfs_root *root, char *name, @@ -2245,6 +2254,7 @@ fail: fail_commit: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return ret; } @@ -2287,6 +2297,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) fail_unlock: mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return ret; } @@ -2827,6 +2838,7 @@ out_fail: iput(inode); } btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); return err; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index dc9865323e3..614903f5c88 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -424,6 +424,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, num_bytes = btrfs_root_used(&dirty->root->root_item); root = dirty->latest_root; + root->fs_info->throttles++; while(1) { trans = btrfs_start_transaction(tree_root, 1); @@ -447,6 +448,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root, mutex_lock(&tree_root->fs_info->fs_mutex); } BUG_ON(ret); + root->fs_info->throttles--; num_bytes -= btrfs_root_used(&dirty->root->root_item); bytes_used = btrfs_root_used(&root->root_item); @@ -484,6 +486,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, u64 objectid = 0; int ret; + root->fs_info->throttles++; while(1) { ret = btrfs_find_first_ordered_inode( &cur_trans->ordered_inode_tree, @@ -523,6 +526,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); } + root->fs_info->throttles--; return 0; } -- cgit v1.2.3-70-g09d2 From c31f8830f0dfd31e7e196b85ca1b39aef8b77d57 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 8 Jan 2008 15:46:31 -0500 Subject: Btrfs: online shrinking fixes While shrinking the FS, the allocation functions need to make sure they don't try to allocate bytes past the end of the FS. nodatacow needed an extra check to force cows when the existing extents are past the end of the FS. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 32 +++++++++++++++++++++++--------- fs/btrfs/inode.c | 15 +++++++++++++-- 2 files changed, 36 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 660b05a4baf..99a8b0f0d31 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -192,11 +192,13 @@ static u64 noinline find_search_start(struct btrfs_root *root, u64 start = 0; u64 end = 0; u64 cache_miss = 0; + u64 total_fs_bytes; int wrapped = 0; if (!cache) { goto out; } + total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); again: ret = cache_block_group(root, cache); if (ret) @@ -223,6 +225,8 @@ again: if (data != BTRFS_BLOCK_GROUP_MIXED && start + num > cache->key.objectid + cache->key.offset) goto new_group; + if (start + num > total_fs_bytes) + goto new_group; return start; } out: @@ -239,7 +243,7 @@ new_group: last = cache->key.objectid + cache->key.offset; wrapped: cache = btrfs_lookup_block_group(root->fs_info, last); - if (!cache) { + if (!cache || cache->key.objectid >= total_fs_bytes) { no_cache: if (!wrapped) { wrapped = 1; @@ -287,6 +291,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, u64 end; u64 free_check; u64 ptr; + u64 total_fs_bytes; int bit; int ret; int full_search = 0; @@ -294,6 +299,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int data_swap = 0; block_group_cache = &info->block_group_cache; + total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); if (!owner) factor = 8; @@ -306,7 +312,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, else bit = BLOCK_GROUP_METADATA; - if (search_start) { + if (search_start && search_start < total_fs_bytes) { struct btrfs_block_group_cache *shint; shint = btrfs_lookup_block_group(info, search_start); if (shint && (shint->data == data || @@ -318,8 +324,8 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, } } } - if (hint && (hint->data == data || - hint->data == BTRFS_BLOCK_GROUP_MIXED)) { + if (hint && hint->key.objectid < total_fs_bytes && + (hint->data == data || hint->data == BTRFS_BLOCK_GROUP_MIXED)) { used = btrfs_block_group_used(&hint->item); if (used + hint->pinned < div_factor(hint->key.offset, factor)) { @@ -333,6 +339,8 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, else hint_last = search_start; + if (hint_last >= total_fs_bytes) + hint_last = search_start; last = hint_last; } again: @@ -350,6 +358,9 @@ again: last = cache->key.objectid + cache->key.offset; used = btrfs_block_group_used(&cache->item); + if (cache->key.objectid > total_fs_bytes) + break; + if (full_search) free_check = cache->key.offset; else @@ -1420,8 +1431,8 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, data = BTRFS_BLOCK_GROUP_MIXED; } - if (search_end == (u64)-1) - search_end = btrfs_super_total_bytes(&info->super_copy); + search_end = min(search_end, + btrfs_super_total_bytes(&info->super_copy)); if (hint_byte) { block_group = btrfs_lookup_block_group(info, hint_byte); if (!block_group) @@ -1617,7 +1628,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, { int ret; int pending_ret; - u64 super_used, root_used; + u64 super_used; + u64 root_used; u64 search_start = 0; u64 new_hint; struct btrfs_fs_info *info = root->fs_info; @@ -1636,6 +1648,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, search_start, search_end, hint_byte, ins, trans->alloc_exclude_start, trans->alloc_exclude_nr, data); +if (ret) +printk("find free extent returns %d\n", ret); BUG_ON(ret); if (ret) return ret; @@ -2292,8 +2306,6 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, while(1) { ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); - BUG_ON(ret == 0); - if (ret < 0) goto out; @@ -2340,6 +2352,8 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) int progress = 0; btrfs_set_super_total_bytes(&info->super_copy, new_size); + clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, + GFP_NOFS); block_group_cache = &info->block_group_cache; path = btrfs_alloc_path(); root = root->fs_info->extent_root; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2cb2dd32407..281757007dd 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -148,6 +148,7 @@ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) u64 bytenr; u64 cow_end; u64 loops = 0; + u64 total_fs_bytes; struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_buffer *leaf; int found_type; @@ -157,6 +158,7 @@ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) int err; struct btrfs_key found_key; + total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); path = btrfs_alloc_path(); BUG_ON(!path); again: @@ -189,8 +191,10 @@ again: found_type = btrfs_file_extent_type(leaf, item); extent_start = found_key.offset; if (found_type == BTRFS_FILE_EXTENT_REG) { - extent_end = extent_start + - btrfs_file_extent_num_bytes(leaf, item); + u64 extent_num_bytes; + + extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item); + extent_end = extent_start + extent_num_bytes; err = 0; if (loops && start != extent_start) @@ -204,6 +208,13 @@ again: if (bytenr == 0) goto not_found; + /* + * we may be called by the resizer, make sure we're inside + * the limits of the FS + */ + if (bytenr + extent_num_bytes > total_fs_bytes) + goto not_found; + if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) { goto not_found; } -- cgit v1.2.3-70-g09d2 From b98b6767a0f6f9baa1d00eff64a43d58412ecdeb Mon Sep 17 00:00:00 2001 From: Yan Date: Tue, 8 Jan 2008 15:54:37 -0500 Subject: Btrfs: Add inode flags support This patch adds NODATASUM & NODATACOW inode flags support. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/ctree.h | 13 ++++++++++++- fs/btrfs/inode.c | 24 +++++++++++------------- 3 files changed, 24 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index f27e633f174..9fc99883c70 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -34,6 +34,7 @@ struct btrfs_inode { * transid of the trans_handle that last modified this inode */ u64 last_trans; + u32 flags; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) { diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f0fb1978553..7eda51542d1 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -447,7 +447,18 @@ struct btrfs_root { #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) #define btrfs_test_opt(root, opt) ((root)->fs_info->mount_opt & \ BTRFS_MOUNT_##opt) - +/* + * Inode flags + */ +#define BTRFS_INODE_NODATASUM 0x1 +#define BTRFS_INODE_NODATACOW 0x2 + +#define btrfs_clear_flag(inode, flag) (BTRFS_I(inode)->flags &= \ + ~BTRFS_INODE_##flag) +#define btrfs_set_flag(inode, flag) (BTRFS_I(inode)->flags |= \ + BTRFS_INODE_##flag) +#define btrfs_test_flag(inode, flag) (BTRFS_I(inode)->flags & \ + BTRFS_INODE_##flag) /* some macros to generate set/get funcs for the struct fields. This * assumes there is a lefoo_to_cpu for every type, so lets make a simple * one for u8: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 281757007dd..fba0478a95f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -243,9 +243,9 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) struct btrfs_root *root = BTRFS_I(inode)->root; u64 num_bytes; int ret; - mutex_lock(&root->fs_info->fs_mutex); - if (btrfs_test_opt(root, NODATACOW)) + if (btrfs_test_opt(root, NODATACOW) || + btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, start, end); else ret = cow_file_range(inode, start, end); @@ -273,10 +273,9 @@ int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) int ret = 0; u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; size_t offset = start - page_start; - - if (btrfs_test_opt(root, NODATASUM)) + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) return 0; - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -299,10 +298,9 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) struct btrfs_csum_item *item; struct btrfs_path *path = NULL; u32 csum; - - if (btrfs_test_opt(root, NODATASUM)) + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) return 0; - mutex_lock(&root->fs_info->fs_mutex); path = btrfs_alloc_path(); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); @@ -335,10 +333,9 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) struct btrfs_root *root = BTRFS_I(inode)->root; u32 csum = ~(u32)0; unsigned long flags; - - if (btrfs_test_opt(root, NODATASUM)) + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) return 0; - ret = get_state_private(em_tree, start, &private); local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); @@ -415,7 +412,7 @@ void btrfs_read_locked_inode(struct inode *inode) alloc_group_block = btrfs_inode_block_group(leaf, inode_item); BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, alloc_group_block); - + BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); if (!BTRFS_I(inode)->block_group) { BTRFS_I(inode)->block_group = btrfs_find_block_group(root, NULL, 0, 0, 0); @@ -484,6 +481,7 @@ static void fill_inode_item(struct extent_buffer *leaf, btrfs_set_inode_nblocks(leaf, item, inode->i_blocks); btrfs_set_inode_generation(leaf, item, inode->i_generation); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); + btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group->key.objectid); } @@ -1454,7 +1452,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 1; group = btrfs_find_block_group(root, group, 0, 0, owner); BTRFS_I(inode)->block_group = group; - + BTRFS_I(inode)->flags = 0; ret = btrfs_insert_empty_inode(trans, root, path, objectid); if (ret) goto fail; -- cgit v1.2.3-70-g09d2 From fdebe2bd70047e057827cba85ba31b2545e31900 Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 14 Jan 2008 13:26:08 -0500 Subject: Btrfs: Add readonly inode flag This patch adds readonly inode flag support. A file with this flag can't be modified, but can be deleted. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 6 +++--- fs/btrfs/inode.c | 16 ++++++++++++---- 2 files changed, 15 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 0c0edacbc14..245952cf765 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -451,9 +451,9 @@ struct btrfs_root { /* * Inode flags */ -#define BTRFS_INODE_NODATASUM 0x1 -#define BTRFS_INODE_NODATACOW 0x2 - +#define BTRFS_INODE_NODATASUM (1 << 0) +#define BTRFS_INODE_NODATACOW (1 << 1) +#define BTRFS_INODE_READONLY (1 << 2) #define btrfs_clear_flag(inode, flag) (BTRFS_I(inode)->flags &= \ ~BTRFS_INODE_##flag) #define btrfs_set_flag(inode, flag) (BTRFS_I(inode)->flags |= \ diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index fba0478a95f..fd0e0d80333 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2850,6 +2850,13 @@ out_fail: btrfs_throttle(root); return err; } +static int btrfs_permission(struct inode *inode, int mask, + struct nameidata *nd) +{ + if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE)) + return -EACCES; + return generic_permission(inode, mask, NULL); +} static struct inode_operations btrfs_dir_inode_operations = { .lookup = btrfs_lookup, @@ -2866,12 +2873,12 @@ static struct inode_operations btrfs_dir_inode_operations = { .getxattr = generic_getxattr, .listxattr = btrfs_listxattr, .removexattr = generic_removexattr, + .permission = btrfs_permission, }; - static struct inode_operations btrfs_dir_ro_inode_operations = { .lookup = btrfs_lookup, + .permission = btrfs_permission, }; - static struct file_operations btrfs_dir_file_operations = { .llseek = generic_file_llseek, .read = generic_read_dir, @@ -2916,15 +2923,16 @@ static struct inode_operations btrfs_file_inode_operations = { .getxattr = generic_getxattr, .listxattr = btrfs_listxattr, .removexattr = generic_removexattr, + .permission = btrfs_permission, }; - static struct inode_operations btrfs_special_inode_operations = { .getattr = btrfs_getattr, .setattr = btrfs_setattr, + .permission = btrfs_permission, }; - static struct inode_operations btrfs_symlink_inode_operations = { .readlink = generic_readlink, .follow_link = page_follow_link_light, .put_link = page_put_link, + .permission = btrfs_permission, }; -- cgit v1.2.3-70-g09d2 From 8c416c9e0d35b0462e89b76b7581127efe03b8b2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 14 Jan 2008 15:10:26 -0500 Subject: Btrfs: Delete any remaining extent_maps before freeing the inode Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index fd0e0d80333..5260b470815 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2601,6 +2601,7 @@ void btrfs_destroy_inode(struct inode *inode) WARN_ON(!list_empty(&inode->i_dentry)); WARN_ON(inode->i_data.nrpages); + btrfs_drop_extent_cache(inode, 0, (u64)-1); kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } -- cgit v1.2.3-70-g09d2 From 61295eb8665e723e77af91d0a1e655a4bd28344f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 14 Jan 2008 16:24:38 -0500 Subject: Btrfs: Add drop inode func to avoid data=ordered deadlock Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/inode.c | 19 +++++++++++++++++++ fs/btrfs/super.c | 1 + 3 files changed, 21 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f65b258344d..a2c2d6d82c5 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1144,6 +1144,7 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); int btrfs_readpage(struct file *file, struct page *page); void btrfs_delete_inode(struct inode *inode); +void btrfs_drop_inode(struct inode *inode); void btrfs_read_locked_inode(struct inode *inode); int btrfs_write_inode(struct inode *inode, int wait); void btrfs_dirty_inode(struct inode *inode); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5260b470815..e53d2033164 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -990,6 +990,25 @@ out: fail: return err; } + +void btrfs_drop_inode(struct inode *inode) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + + if (!BTRFS_I(inode)->ordered_trans) { + generic_drop_inode(inode); + return; + } + /* nasty, but it prevents a deadlock with data=ordered by preventing + * a commit until after this inode is done + */ + trans = btrfs_start_transaction(root, 1); + generic_drop_inode(inode); + /* note, the inode is now untouchable */ + btrfs_end_transaction(trans, root); +} + void btrfs_delete_inode(struct inode *inode) { struct btrfs_trans_handle *trans; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 9ab4086d463..4deea393ca9 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -425,6 +425,7 @@ static struct file_system_type btrfs_fs_type = { static struct super_operations btrfs_super_ops = { .delete_inode = btrfs_delete_inode, + .drop_inode = btrfs_drop_inode, .put_super = btrfs_put_super, .read_inode = btrfs_read_locked_inode, .write_super = btrfs_write_super, -- cgit v1.2.3-70-g09d2 From cee36a03e8f7c6e14aefd497d3acf01bcd3ef153 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 15 Jan 2008 08:40:48 -0500 Subject: Rework btrfs_drop_inode to avoid scheduling Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 1 + fs/btrfs/inode.c | 13 +++---------- fs/btrfs/ordered-data.c | 36 ++++++++++++++++++++++++++++++++++++ fs/btrfs/ordered-data.h | 1 + fs/btrfs/transaction.c | 2 ++ 6 files changed, 44 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index a2c2d6d82c5..1e19f2d8633 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -349,6 +349,7 @@ struct btrfs_fs_info { u64 total_pinned; spinlock_t delalloc_lock; + spinlock_t new_trans_lock; u64 delalloc_bytes; }; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 87daed632f3..cd29922d407 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -639,6 +639,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) INIT_LIST_HEAD(&fs_info->hashers); spin_lock_init(&fs_info->hash_lock); spin_lock_init(&fs_info->delalloc_lock); + spin_lock_init(&fs_info->new_trans_lock); memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); init_completion(&fs_info->kobj_unregister); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e53d2033164..008e3445748 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -993,20 +993,13 @@ fail: void btrfs_drop_inode(struct inode *inode) { - struct btrfs_trans_handle *trans; - struct btrfs_root *root = BTRFS_I(inode)->root; - - if (!BTRFS_I(inode)->ordered_trans) { + if (!BTRFS_I(inode)->ordered_trans || inode->i_nlink) { generic_drop_inode(inode); return; } - /* nasty, but it prevents a deadlock with data=ordered by preventing - * a commit until after this inode is done - */ - trans = btrfs_start_transaction(root, 1); + /* FIXME, make sure this delete actually ends up in the transaction */ + btrfs_del_ordered_inode(inode); generic_drop_inode(inode); - /* note, the inode is now untouchable */ - btrfs_end_transaction(trans, root); } void btrfs_delete_inode(struct inode *inode) diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 411aba84d30..b56011baa17 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -219,3 +219,39 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, kfree(entry); return 1; } + +static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, + u64 root_objectid, u64 objectid) +{ + struct tree_entry *entry; + struct rb_node *node; + struct rb_node *prev; + + write_lock(&tree->lock); + node = __tree_search(&tree->tree, root_objectid, objectid, &prev); + if (!node) { + write_unlock(&tree->lock); + return 0; + } + rb_erase(node, &tree->tree); + write_unlock(&tree->lock); + entry = rb_entry(node, struct tree_entry, rb_node); + kfree(entry); + return 1; +} + +int btrfs_del_ordered_inode(struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 root_objectid = root->root_key.objectid; + + spin_lock(&root->fs_info->new_trans_lock); + if (root->fs_info->running_transaction) { + struct btrfs_ordered_inode_tree *tree; + tree = &root->fs_info->running_transaction->ordered_inode_tree; + __btrfs_del_ordered_inode(tree, root_objectid, inode->i_ino); + } + spin_unlock(&root->fs_info->new_trans_lock); + return 0; +} + diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index aaf9eb14271..26b26212865 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -36,4 +36,5 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, u64 *root_objectid, u64 *objectid); int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, u64 *root_objectid, u64 *objectid); +int btrfs_del_ordered_inode(struct inode *inode); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 614903f5c88..a3205808ab2 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -699,7 +699,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, BUG_ON(ret); cur_trans = root->fs_info->running_transaction; + spin_lock(&root->fs_info->new_trans_lock); root->fs_info->running_transaction = NULL; + spin_unlock(&root->fs_info->new_trans_lock); btrfs_set_super_generation(&root->fs_info->super_copy, cur_trans->transid); btrfs_set_super_root(&root->fs_info->super_copy, -- cgit v1.2.3-70-g09d2 From 9cce6c3bfca85bf92e8c9358542a18dfa6c232be Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 15 Jan 2008 08:44:06 -0500 Subject: Btrfs: Disable delalloc accounting for now Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 008e3445748..f83f88ca8ac 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -80,6 +80,8 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, u64 thresh; int ret = 0; + return 0; + if (for_del) thresh = total * 90; else -- cgit v1.2.3-70-g09d2 From 2da98f003f4788b0a72c5f87bc55b061f65f30fa Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 16 Jan 2008 11:44:43 -0500 Subject: Btrfs: Run igrab on data=ordered inodes to prevent deadlocks during writeout Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 +- fs/btrfs/inode.c | 32 ++++++++++++++++++++++++++------ fs/btrfs/ordered-data.c | 10 ++++++++-- fs/btrfs/super.c | 2 +- fs/btrfs/transaction.c | 1 + 5 files changed, 37 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 1e19f2d8633..fa65fe027e4 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1145,7 +1145,7 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page); int btrfs_readpage(struct file *file, struct page *page); void btrfs_delete_inode(struct inode *inode); -void btrfs_drop_inode(struct inode *inode); +void btrfs_put_inode(struct inode *inode); void btrfs_read_locked_inode(struct inode *inode); int btrfs_write_inode(struct inode *inode, int wait); void btrfs_dirty_inode(struct inode *inode); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f83f88ca8ac..c1ac0bcbb46 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -601,6 +601,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) { struct btrfs_root *root; struct btrfs_trans_handle *trans; + struct inode *inode = dentry->d_inode; int ret; unsigned long nr = 0; @@ -617,6 +618,18 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) ret = btrfs_unlink_trans(trans, root, dir, dentry); nr = trans->blocks_used; + if (inode->i_nlink == 0) { + int found; + /* if the inode isn't linked anywhere, + * we don't need to worry about + * data=ordered + */ + found = btrfs_del_ordered_inode(inode); + if (found == 1) { + atomic_dec(&inode->i_count); + } + } + btrfs_end_transaction(trans, root); fail: mutex_unlock(&root->fs_info->fs_mutex); @@ -993,15 +1006,22 @@ fail: return err; } -void btrfs_drop_inode(struct inode *inode) +void btrfs_put_inode(struct inode *inode) { - if (!BTRFS_I(inode)->ordered_trans || inode->i_nlink) { - generic_drop_inode(inode); + int ret; + + if (!BTRFS_I(inode)->ordered_trans) { + return; + } + + if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || + mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK)) return; + + ret = btrfs_del_ordered_inode(inode); + if (ret == 1) { + atomic_dec(&inode->i_count); } - /* FIXME, make sure this delete actually ends up in the transaction */ - btrfs_del_ordered_inode(inode); - generic_drop_inode(inode); } void btrfs_delete_inode(struct inode *inode) diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index b56011baa17..cba2b623d02 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -153,6 +153,8 @@ int btrfs_add_ordered_inode(struct inode *inode) write_unlock(&tree->lock); if (node) kfree(entry); + else + igrab(inode); return 0; } @@ -221,6 +223,7 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, } static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, + struct inode *inode, u64 root_objectid, u64 objectid) { struct tree_entry *entry; @@ -234,6 +237,7 @@ static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, return 0; } rb_erase(node, &tree->tree); + BTRFS_I(inode)->ordered_trans = 0; write_unlock(&tree->lock); entry = rb_entry(node, struct tree_entry, rb_node); kfree(entry); @@ -244,14 +248,16 @@ int btrfs_del_ordered_inode(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 root_objectid = root->root_key.objectid; + int ret = 0; spin_lock(&root->fs_info->new_trans_lock); if (root->fs_info->running_transaction) { struct btrfs_ordered_inode_tree *tree; tree = &root->fs_info->running_transaction->ordered_inode_tree; - __btrfs_del_ordered_inode(tree, root_objectid, inode->i_ino); + ret = __btrfs_del_ordered_inode(tree, inode, root_objectid, + inode->i_ino); } spin_unlock(&root->fs_info->new_trans_lock); - return 0; + return ret; } diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 4deea393ca9..e506de3168b 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -425,7 +425,7 @@ static struct file_system_type btrfs_fs_type = { static struct super_operations btrfs_super_ops = { .delete_inode = btrfs_delete_inode, - .drop_inode = btrfs_drop_inode, + .put_inode = btrfs_put_inode, .put_super = btrfs_put_super, .read_inode = btrfs_read_locked_inode, .write_super = btrfs_write_super, diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index a3205808ab2..08f7a188dc3 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -521,6 +521,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, if (inode) { if (S_ISREG(inode->i_mode)) filemap_write_and_wait(inode->i_mapping); + atomic_dec(&inode->i_count); iput(inode); } mutex_lock(&root->fs_info->fs_mutex); -- cgit v1.2.3-70-g09d2 From fb4bc1e0565a842d704846005ac0ec739e608ede Mon Sep 17 00:00:00 2001 From: Yan Date: Thu, 17 Jan 2008 11:59:51 -0500 Subject: Btrfs: Fix compile on 2.6.22 kernel This patch fixes compile error on kernel-2.6.22 Signed-off-by: Chris Mason --- fs/btrfs/acl.c | 5 +++++ fs/btrfs/inode.c | 5 +++-- 2 files changed, 8 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c index b4f85ebe1bd..27bb841fafe 100644 --- a/fs/btrfs/acl.c +++ b/fs/btrfs/acl.c @@ -22,6 +22,11 @@ #include #include "ctree.h" #include "xattr.h" +#ifndef is_owner_or_cap +#define is_owner_or_cap(inode) \ + ((current->fsuid == (inode)->i_uid) || capable(CAP_FOWNER)) +#endif + static int btrfs_xattr_set_acl(struct inode *inode, int type, const void *value, size_t size) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c1ac0bcbb46..51fc06dc4a5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2322,15 +2322,16 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) ret = -ENOMEM; goto fail_unlock; } - pending_snapshot->name = kstrndup(name, namelen, GFP_NOFS); + pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); if (!pending_snapshot->name) { ret = -ENOMEM; kfree(pending_snapshot); goto fail_unlock; } + memcpy(pending_snapshot->name, name, namelen); + pending_snapshot->name[namelen] = '\0'; trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); - pending_snapshot->root = root; list_add(&pending_snapshot->list, &trans->transaction->pending_snapshots); -- cgit v1.2.3-70-g09d2 From 5f56406aabdf5444d040c5955effc665b1d0dbaf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jan 2008 16:47:59 -0500 Subject: Btrfs: Fix hole insertion corner cases There were a few places that could cause duplicate extent insertion, this adjusts the code that creates holes to avoid it. lookup_extent_map is changed to correctly return all of the extents in a range, even when there are none matching at the start of the range. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/extent_map.c | 45 +++++++++++++++++++++++------ fs/btrfs/file.c | 78 ++++++++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/inode.c | 15 ++++++++-- 4 files changed, 126 insertions(+), 13 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 7a588ba2b74..b9f2975b55c 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1170,6 +1170,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, struct inode *inode); /* file.c */ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); +int btrfs_check_file(struct btrfs_root *root, struct inode *inode); extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 485cf0719b3..010a287fbd7 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -204,10 +204,12 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset, } static struct rb_node *__tree_search(struct rb_root *root, u64 offset, - struct rb_node **prev_ret) + struct rb_node **prev_ret, + struct rb_node **next_ret) { struct rb_node * n = root->rb_node; struct rb_node *prev = NULL; + struct rb_node *orig_prev = NULL; struct tree_entry *entry; struct tree_entry *prev_entry = NULL; @@ -223,13 +225,25 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, else return n; } - if (!prev_ret) - return NULL; - while(prev && offset > prev_entry->end) { - prev = rb_next(prev); + + if (prev_ret) { + orig_prev = prev; + while(prev && offset > prev_entry->end) { + prev = rb_next(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *prev_ret = prev; + prev = orig_prev; + } + + if (next_ret) { prev_entry = rb_entry(prev, struct tree_entry, rb_node); + while(prev && offset < prev_entry->start) { + prev = rb_prev(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *next_ret = prev; } - *prev_ret = prev; return NULL; } @@ -237,7 +251,7 @@ static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) { struct rb_node *prev; struct rb_node *ret; - ret = __tree_search(root, offset, &prev); + ret = __tree_search(root, offset, &prev, NULL); if (!ret) return prev; return ret; @@ -248,7 +262,7 @@ static int tree_delete(struct rb_root *root, u64 offset) struct rb_node *node; struct tree_entry *entry; - node = __tree_search(root, offset, NULL); + node = __tree_search(root, offset, NULL, NULL); if (!node) return -ENOENT; entry = rb_entry(node, struct tree_entry, rb_node); @@ -314,9 +328,21 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, { struct extent_map *em; struct rb_node *rb_node; + struct rb_node *prev = NULL; + struct rb_node *next = NULL; read_lock_irq(&tree->lock); - rb_node = tree_search(&tree->map, start); + rb_node = __tree_search(&tree->map, start, &prev, &next); + if (!rb_node && prev) { + em = rb_entry(prev, struct extent_map, rb_node); + if (em->start <= end && em->end >= start) + goto found; + } + if (!rb_node && next) { + em = rb_entry(next, struct extent_map, rb_node); + if (em->start <= end && em->end >= start) + goto found; + } if (!rb_node) { em = NULL; goto out; @@ -330,6 +356,7 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, em = NULL; goto out; } +found: atomic_inc(&em->refs); out: read_unlock_irq(&tree->lock); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 897242e87fa..1cd8c908811 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -278,7 +278,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 hole_size; u64 mask = root->sectorsize - 1; last_pos_in_file = (isize + mask) & ~mask; - hole_size = (start_pos - last_pos_in_file + mask) & ~mask; + hole_size = (end_pos - last_pos_in_file + mask) & ~mask; if (last_pos_in_file < start_pos) { err = btrfs_drop_extents(trans, root, inode, @@ -293,6 +293,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inode->i_ino, last_pos_in_file, 0, 0, hole_size); + btrfs_check_file(root, inode); } if (err) goto failed; @@ -378,6 +379,80 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) return 0; } +int btrfs_check_file(struct btrfs_root *root, struct inode *inode) +{ + return 0; +#if 0 + struct btrfs_path *path; + struct btrfs_key found_key; + struct extent_buffer *leaf; + struct btrfs_file_extent_item *extent; + u64 last_offset = 0; + int nritems; + int slot; + int found_type; + int ret; + int err = 0; + u64 extent_end = 0; + + path = btrfs_alloc_path(); + ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, + last_offset, 0); + while(1) { + nritems = btrfs_header_nritems(path->nodes[0]); + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret) + goto out; + nritems = btrfs_header_nritems(path->nodes[0]); + } + slot = path->slots[0]; + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, slot); + if (found_key.objectid != inode->i_ino) + break; + if (found_key.type != BTRFS_EXTENT_DATA_KEY) + goto out; + + if (found_key.offset != last_offset) { + WARN_ON(1); + btrfs_print_leaf(root, leaf); + printk("inode %lu found offset %Lu expected %Lu\n", + inode->i_ino, found_key.offset, last_offset); + err = 1; + goto out; + } + extent = btrfs_item_ptr(leaf, slot, + struct btrfs_file_extent_item); + found_type = btrfs_file_extent_type(leaf, extent); + if (found_type == BTRFS_FILE_EXTENT_REG) { + extent_end = found_key.offset + + btrfs_file_extent_num_bytes(leaf, extent); + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + struct btrfs_item *item; + item = btrfs_item_nr(leaf, slot); + extent_end = found_key.offset + + btrfs_file_extent_inline_len(leaf, item); + extent_end = (extent_end + root->sectorsize - 1) & + ~((u64)root->sectorsize -1 ); + } + last_offset = extent_end; + path->slots[0]++; + } + if (last_offset < inode->i_size) { + WARN_ON(1); + btrfs_print_leaf(root, leaf); + printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, + last_offset, inode->i_size); + err = 1; + + } +out: + btrfs_free_path(path); + return err; +#endif +} + /* * this is very complex, but the basic idea is to drop all extents * in the range start - end. hint_block is filled in with a block number @@ -436,6 +511,7 @@ next_slot: slot = path->slots[0]; ret = 0; btrfs_item_key_to_cpu(leaf, &key, slot); + if (key.offset >= end || key.objectid != inode->i_ino) { goto out; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 51fc06dc4a5..67005480e13 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -133,6 +133,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) ret = btrfs_insert_file_extent(trans, root, inode->i_ino, start, ins.objectid, ins.offset, ins.offset); + btrfs_check_file(root, inode); num_bytes -= cur_alloc_size; alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; @@ -965,12 +966,18 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) u64 mask = root->sectorsize - 1; u64 pos = (inode->i_size + mask) & ~mask; u64 block_end = attr->ia_size | mask; + u64 hole_start; u64 hole_size; u64 alloc_hint = 0; if (attr->ia_size <= pos) goto out; + if (pos != inode->i_size) + hole_start = pos + root->sectorsize; + else + hole_start = pos; + mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); mutex_unlock(&root->fs_info->fs_mutex); @@ -980,19 +987,21 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); lock_extent(em_tree, pos, block_end, GFP_NOFS); - hole_size = (attr->ia_size - pos + mask) & ~mask; + hole_size = block_end - hole_start; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); err = btrfs_drop_extents(trans, root, inode, - pos, pos + hole_size, pos, + pos, block_end, pos, &alloc_hint); if (alloc_hint != EXTENT_MAP_INLINE) { err = btrfs_insert_file_extent(trans, root, inode->i_ino, - pos, 0, 0, hole_size); + hole_start, 0, 0, + hole_size); + btrfs_check_file(root, inode); } btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); -- cgit v1.2.3-70-g09d2 From d1310b2e0cd98eb1348553e69b73827b436dca7b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 24 Jan 2008 16:13:08 -0500 Subject: Btrfs: Split the extent_map code into two parts There is now extent_map for mapping offsets in the file to disk and extent_io for state tracking, IO submission and extent_bufers. The new extent_map code shifts from [start,end] pairs to [start,len], and pushes the locking out into the caller. This allows a few performance optimizations and is easier to use. A number of extent_map usage bugs were fixed, mostly with failing to remove extent_map entries when changing the file. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 3 +- fs/btrfs/btrfs_inode.h | 2 + fs/btrfs/ctree.h | 15 +- fs/btrfs/disk-io.c | 123 +- fs/btrfs/extent-tree.c | 43 +- fs/btrfs/extent_io.c | 3089 +++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/extent_io.h | 193 +++ fs/btrfs/extent_map.c | 3119 ++---------------------------------------------- fs/btrfs/extent_map.h | 214 +--- fs/btrfs/file.c | 29 +- fs/btrfs/inode.c | 175 +-- fs/btrfs/super.c | 10 +- fs/btrfs/transaction.c | 8 +- fs/btrfs/transaction.h | 2 +- 14 files changed, 3612 insertions(+), 3413 deletions(-) create mode 100644 fs/btrfs/extent_io.c create mode 100644 fs/btrfs/extent_io.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 72cc3e16741..9c61609f371 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -5,7 +5,8 @@ obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ - extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o + extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ + extent_io.o ifeq ($(CONFIG_FS_POSIX_ACL),y) btrfs-y += acl.o diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 9fc99883c70..9a0647dc5a0 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -20,6 +20,7 @@ #define __BTRFS_I__ #include "extent_map.h" +#include "extent_io.h" /* in memory btrfs inode */ struct btrfs_inode { @@ -27,6 +28,7 @@ struct btrfs_inode { struct btrfs_block_group_cache *block_group; struct btrfs_key location; struct extent_map_tree extent_tree; + struct extent_io_tree io_tree; struct inode vfs_inode; u64 ordered_trans; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b9f2975b55c..6c65473e0fe 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -27,6 +27,7 @@ #include #include #include "bit-radix.h" +#include "extent_io.h" #include "extent_map.h" struct btrfs_trans_handle; @@ -314,11 +315,11 @@ struct btrfs_fs_info { struct btrfs_root *tree_root; struct radix_tree_root fs_roots_radix; - struct extent_map_tree free_space_cache; - struct extent_map_tree block_group_cache; - struct extent_map_tree pinned_extents; - struct extent_map_tree pending_del; - struct extent_map_tree extent_ins; + struct extent_io_tree free_space_cache; + struct extent_io_tree block_group_cache; + struct extent_io_tree pinned_extents; + struct extent_io_tree pending_del; + struct extent_io_tree extent_ins; u64 generation; u64 last_trans_committed; @@ -956,7 +957,7 @@ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, u64 first_extent); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); -int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy); +int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, u64 bytenr); @@ -1001,7 +1002,7 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root u64 owner_objectid, u64 owner_offset, int pin); int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_map_tree *unpin); + struct extent_io_tree *unpin); int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 5d1f9bca271..4c4ebea0b2a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -43,14 +43,14 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) } #endif -static struct extent_map_ops btree_extent_map_ops; +static struct extent_io_ops btree_extent_io_ops; struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) { struct inode *btree_inode = root->fs_info->btree_inode; struct extent_buffer *eb; - eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, + eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, bytenr, blocksize, GFP_NOFS); return eb; } @@ -61,13 +61,13 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, struct inode *btree_inode = root->fs_info->btree_inode; struct extent_buffer *eb; - eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree, + eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, bytenr, blocksize, NULL, GFP_NOFS); return eb; } struct extent_map *btree_get_extent(struct inode *inode, struct page *page, - size_t page_offset, u64 start, u64 end, + size_t page_offset, u64 start, u64 len, int create) { struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; @@ -75,7 +75,9 @@ struct extent_map *btree_get_extent(struct inode *inode, struct page *page, int ret; again: - em = lookup_extent_mapping(em_tree, start, end); + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, start, len); + spin_unlock(&em_tree->lock); if (em) { goto out; } @@ -85,11 +87,14 @@ again: goto out; } em->start = 0; - em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1; + em->len = i_size_read(inode); em->block_start = 0; - em->block_end = em->end; em->bdev = inode->i_sb->s_bdev; + + spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); + spin_unlock(&em_tree->lock); + if (ret == -EEXIST) { free_extent_map(em); em = NULL; @@ -175,13 +180,13 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { - struct extent_map_tree *tree; + struct extent_io_tree *tree; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; u64 found_start; int found_level; unsigned long len; struct extent_buffer *eb; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; if (page->private == EXTENT_PAGE_PRIVATE) goto out; @@ -230,16 +235,16 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) static int btree_writepage(struct page *page, struct writeback_control *wbc) { - struct extent_map_tree *tree; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; return extent_write_full_page(tree, page, btree_get_extent, wbc); } static int btree_writepages(struct address_space *mapping, struct writeback_control *wbc) { - struct extent_map_tree *tree; - tree = &BTRFS_I(mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(mapping->host)->io_tree; if (wbc->sync_mode == WB_SYNC_NONE) { u64 num_dirty; u64 start = 0; @@ -264,18 +269,20 @@ static int btree_writepages(struct address_space *mapping, int btree_readpage(struct file *file, struct page *page) { - struct extent_map_tree *tree; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; return extent_read_full_page(tree, page, btree_get_extent); } static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) { - struct extent_map_tree *tree; + struct extent_io_tree *tree; + struct extent_map_tree *map; int ret; - tree = &BTRFS_I(page->mapping->host)->extent_tree; - ret = try_release_extent_mapping(tree, page); + tree = &BTRFS_I(page->mapping->host)->io_tree; + map = &BTRFS_I(page->mapping->host)->extent_tree; + ret = try_release_extent_mapping(map, tree, page); if (ret == 1) { ClearPagePrivate(page); set_page_private(page, 0); @@ -286,8 +293,8 @@ static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) static void btree_invalidatepage(struct page *page, unsigned long offset) { - struct extent_map_tree *tree; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; extent_invalidatepage(tree, page, offset); btree_releasepage(page, GFP_NOFS); } @@ -331,7 +338,7 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) buf = btrfs_find_create_tree_block(root, bytenr, blocksize); if (!buf) return 0; - read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, + read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 0); free_extent_buffer(buf); return ret; @@ -342,40 +349,39 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, { struct extent_buffer *buf = NULL; struct inode *btree_inode = root->fs_info->btree_inode; - struct extent_map_tree *extent_tree; + struct extent_io_tree *io_tree; u64 end; int ret; - extent_tree = &BTRFS_I(btree_inode)->extent_tree; + io_tree = &BTRFS_I(btree_inode)->io_tree; buf = btrfs_find_create_tree_block(root, bytenr, blocksize); if (!buf) return NULL; - read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, - buf, 0, 1); + read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1); if (buf->flags & EXTENT_CSUM) return buf; end = buf->start + PAGE_CACHE_SIZE - 1; - if (test_range_bit(extent_tree, buf->start, end, EXTENT_CSUM, 1)) { + if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) { buf->flags |= EXTENT_CSUM; return buf; } - lock_extent(extent_tree, buf->start, end, GFP_NOFS); + lock_extent(io_tree, buf->start, end, GFP_NOFS); - if (test_range_bit(extent_tree, buf->start, end, EXTENT_CSUM, 1)) { + if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) { buf->flags |= EXTENT_CSUM; goto out_unlock; } ret = csum_tree_block(root, buf, 1); - set_extent_bits(extent_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS); + set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS); buf->flags |= EXTENT_CSUM; out_unlock: - unlock_extent(extent_tree, buf->start, end, GFP_NOFS); + unlock_extent(io_tree, buf->start, end, GFP_NOFS); return buf; } @@ -385,7 +391,7 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *btree_inode = root->fs_info->btree_inode; if (btrfs_header_generation(buf) == root->fs_info->running_transaction->transid) - clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, + clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); return 0; } @@ -394,7 +400,7 @@ int wait_on_tree_block_writeback(struct btrfs_root *root, struct extent_buffer *buf) { struct inode *btree_inode = root->fs_info->btree_inode; - wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree, + wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, buf); return 0; } @@ -659,20 +665,23 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->btree_inode->i_nlink = 1; fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size; fs_info->btree_inode->i_mapping->a_ops = &btree_aops; - extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, + extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, fs_info->btree_inode->i_mapping, GFP_NOFS); - BTRFS_I(fs_info->btree_inode)->extent_tree.ops = &btree_extent_map_ops; + extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, + GFP_NOFS); + + BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; - extent_map_tree_init(&fs_info->free_space_cache, + extent_io_tree_init(&fs_info->free_space_cache, fs_info->btree_inode->i_mapping, GFP_NOFS); - extent_map_tree_init(&fs_info->block_group_cache, + extent_io_tree_init(&fs_info->block_group_cache, fs_info->btree_inode->i_mapping, GFP_NOFS); - extent_map_tree_init(&fs_info->pinned_extents, + extent_io_tree_init(&fs_info->pinned_extents, fs_info->btree_inode->i_mapping, GFP_NOFS); - extent_map_tree_init(&fs_info->pending_del, + extent_io_tree_init(&fs_info->pending_del, fs_info->btree_inode->i_mapping, GFP_NOFS); - extent_map_tree_init(&fs_info->extent_ins, + extent_io_tree_init(&fs_info->extent_ins, fs_info->btree_inode->i_mapping, GFP_NOFS); fs_info->do_barriers = 1; fs_info->closing = 0; @@ -787,7 +796,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root if (!btrfs_test_opt(root, NOBARRIER)) blkdev_issue_flush(sb->s_bdev, NULL); - set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super); + set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super); ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping, super->start, super->len); if (!btrfs_test_opt(root, NOBARRIER)) @@ -864,12 +873,12 @@ int close_ctree(struct btrfs_root *root) filemap_write_and_wait(fs_info->btree_inode->i_mapping); - extent_map_tree_empty_lru(&fs_info->free_space_cache); - extent_map_tree_empty_lru(&fs_info->block_group_cache); - extent_map_tree_empty_lru(&fs_info->pinned_extents); - extent_map_tree_empty_lru(&fs_info->pending_del); - extent_map_tree_empty_lru(&fs_info->extent_ins); - extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree); + extent_io_tree_empty_lru(&fs_info->free_space_cache); + extent_io_tree_empty_lru(&fs_info->block_group_cache); + extent_io_tree_empty_lru(&fs_info->pinned_extents); + extent_io_tree_empty_lru(&fs_info->pending_del); + extent_io_tree_empty_lru(&fs_info->extent_ins); + extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); @@ -892,13 +901,13 @@ int close_ctree(struct btrfs_root *root) int btrfs_buffer_uptodate(struct extent_buffer *buf) { struct inode *btree_inode = buf->first_page->mapping->host; - return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf); + return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); } int btrfs_set_buffer_uptodate(struct extent_buffer *buf) { struct inode *btree_inode = buf->first_page->mapping->host; - return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, + return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); } @@ -914,7 +923,7 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) transid, root->fs_info->generation); WARN_ON(1); } - set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf); + set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); } void btrfs_throttle(struct btrfs_root *root) @@ -941,7 +950,7 @@ void btrfs_set_buffer_defrag(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start, + set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); } @@ -949,7 +958,7 @@ void btrfs_set_buffer_defrag_done(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start, + set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, GFP_NOFS); } @@ -958,7 +967,7 @@ int btrfs_buffer_defrag(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - return test_range_bit(&BTRFS_I(btree_inode)->extent_tree, + return test_range_bit(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0); } @@ -966,7 +975,7 @@ int btrfs_buffer_defrag_done(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - return test_range_bit(&BTRFS_I(btree_inode)->extent_tree, + return test_range_bit(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, 0); } @@ -975,7 +984,7 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree, + return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, GFP_NOFS); } @@ -984,7 +993,7 @@ int btrfs_clear_buffer_defrag(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree, + return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); } @@ -993,10 +1002,10 @@ int btrfs_read_buffer(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; struct inode *btree_inode = root->fs_info->btree_inode; - return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree, + return read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1); } -static struct extent_map_ops btree_extent_map_ops = { +static struct extent_io_ops btree_extent_io_ops = { .writepage_io_hook = btree_writepage_io_hook, }; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index b69a46691a9..1cf125ab782 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -63,7 +63,7 @@ static int cache_block_group(struct btrfs_root *root, int ret; struct btrfs_key key; struct extent_buffer *leaf; - struct extent_map_tree *free_space_cache; + struct extent_io_tree *free_space_cache; int slot; u64 last = 0; u64 hole_size; @@ -158,7 +158,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, u64 bytenr) { - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_block_group_cache *block_group = NULL; u64 ptr; u64 start; @@ -281,7 +281,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int data, int owner) { struct btrfs_block_group_cache *cache; - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_block_group_cache *found_group = NULL; struct btrfs_fs_info *info = root->fs_info; u64 used; @@ -951,7 +951,7 @@ fail: int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_block_group_cache *cache; int ret; int err = 0; @@ -1107,12 +1107,12 @@ static int update_pinned_extents(struct btrfs_root *root, return 0; } -int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy) +int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) { u64 last = 0; u64 start; u64 end; - struct extent_map_tree *pinned_extents = &root->fs_info->pinned_extents; + struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents; int ret; while(1) { @@ -1128,12 +1128,12 @@ int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy) int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_map_tree *unpin) + struct extent_io_tree *unpin) { u64 start; u64 end; int ret; - struct extent_map_tree *free_space_cache; + struct extent_io_tree *free_space_cache; free_space_cache = &root->fs_info->free_space_cache; while(1) { @@ -1329,8 +1329,8 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct int err = 0; u64 start; u64 end; - struct extent_map_tree *pending_del; - struct extent_map_tree *pinned_extents; + struct extent_io_tree *pending_del; + struct extent_io_tree *pinned_extents; pending_del = &extent_root->fs_info->pending_del; pinned_extents = &extent_root->fs_info->pinned_extents; @@ -1802,7 +1802,7 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, set_extent_dirty(&trans->transaction->dirty_pages, buf->start, buf->start + buf->len - 1, GFP_NOFS); - set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->extent_tree, + set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->io_tree, buf->start, buf->start + buf->len - 1, EXTENT_CSUM, GFP_NOFS); buf->flags |= EXTENT_CSUM; @@ -2166,7 +2166,7 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, unsigned long i; struct page *page; struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct file_ra_state *ra; ra = kzalloc(sizeof(*ra), GFP_NOFS); @@ -2195,15 +2195,14 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, page_start = (u64)page->index << PAGE_CACHE_SHIFT; page_end = page_start + PAGE_CACHE_SIZE - 1; - lock_extent(em_tree, page_start, page_end, GFP_NOFS); + lock_extent(io_tree, page_start, page_end, GFP_NOFS); delalloc_start = page_start; - existing_delalloc = - count_range_bits(&BTRFS_I(inode)->extent_tree, - &delalloc_start, page_end, - PAGE_CACHE_SIZE, EXTENT_DELALLOC); + existing_delalloc = count_range_bits(io_tree, + &delalloc_start, page_end, + PAGE_CACHE_SIZE, EXTENT_DELALLOC); - set_extent_delalloc(em_tree, page_start, + set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); spin_lock(&root->fs_info->delalloc_lock); @@ -2211,7 +2210,7 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, existing_delalloc; spin_unlock(&root->fs_info->delalloc_lock); - unlock_extent(em_tree, page_start, page_end, GFP_NOFS); + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); unlock_page(page); page_cache_release(page); @@ -2379,7 +2378,7 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) u64 cur_byte; u64 total_found; struct btrfs_fs_info *info = root->fs_info; - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; @@ -2561,7 +2560,7 @@ int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, struct btrfs_block_group_cache *cache; struct btrfs_block_group_item *item; struct btrfs_fs_info *info = root->fs_info; - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_key key; struct extent_buffer *leaf; int ret; @@ -2645,7 +2644,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) int bit; struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; - struct extent_map_tree *block_group_cache; + struct extent_io_tree *block_group_cache; struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c new file mode 100644 index 00000000000..15cc158a049 --- /dev/null +++ b/fs/btrfs/extent_io.c @@ -0,0 +1,3089 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "extent_io.h" +#include "extent_map.h" + +/* temporary define until extent_map moves out of btrfs */ +struct kmem_cache *btrfs_cache_create(const char *name, size_t size, + unsigned long extra_flags, + void (*ctor)(void *, struct kmem_cache *, + unsigned long)); + +static struct kmem_cache *extent_state_cache; +static struct kmem_cache *extent_buffer_cache; + +static LIST_HEAD(buffers); +static LIST_HEAD(states); + +static spinlock_t state_lock = SPIN_LOCK_UNLOCKED; +#define BUFFER_LRU_MAX 64 + +struct tree_entry { + u64 start; + u64 end; + int in_tree; + struct rb_node rb_node; +}; + +struct extent_page_data { + struct bio *bio; + struct extent_io_tree *tree; + get_extent_t *get_extent; +}; + +int __init extent_io_init(void) +{ + extent_state_cache = btrfs_cache_create("extent_state", + sizeof(struct extent_state), 0, + NULL); + if (!extent_state_cache) + return -ENOMEM; + + extent_buffer_cache = btrfs_cache_create("extent_buffers", + sizeof(struct extent_buffer), 0, + NULL); + if (!extent_buffer_cache) + goto free_state_cache; + return 0; + +free_state_cache: + kmem_cache_destroy(extent_state_cache); + return -ENOMEM; +} + +void extent_io_exit(void) +{ + struct extent_state *state; + + while (!list_empty(&states)) { + state = list_entry(states.next, struct extent_state, list); + printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs)); + list_del(&state->list); + kmem_cache_free(extent_state_cache, state); + + } + + if (extent_state_cache) + kmem_cache_destroy(extent_state_cache); + if (extent_buffer_cache) + kmem_cache_destroy(extent_buffer_cache); +} + +void extent_io_tree_init(struct extent_io_tree *tree, + struct address_space *mapping, gfp_t mask) +{ + tree->state.rb_node = NULL; + tree->ops = NULL; + tree->dirty_bytes = 0; + rwlock_init(&tree->lock); + spin_lock_init(&tree->lru_lock); + tree->mapping = mapping; + INIT_LIST_HEAD(&tree->buffer_lru); + tree->lru_size = 0; +} +EXPORT_SYMBOL(extent_io_tree_init); + +void extent_io_tree_empty_lru(struct extent_io_tree *tree) +{ + struct extent_buffer *eb; + while(!list_empty(&tree->buffer_lru)) { + eb = list_entry(tree->buffer_lru.next, struct extent_buffer, + lru); + list_del_init(&eb->lru); + free_extent_buffer(eb); + } +} +EXPORT_SYMBOL(extent_io_tree_empty_lru); + +struct extent_state *alloc_extent_state(gfp_t mask) +{ + struct extent_state *state; + unsigned long flags; + + state = kmem_cache_alloc(extent_state_cache, mask); + if (!state || IS_ERR(state)) + return state; + state->state = 0; + state->in_tree = 0; + state->private = 0; + + spin_lock_irqsave(&state_lock, flags); + list_add(&state->list, &states); + spin_unlock_irqrestore(&state_lock, flags); + + atomic_set(&state->refs, 1); + init_waitqueue_head(&state->wq); + return state; +} +EXPORT_SYMBOL(alloc_extent_state); + +void free_extent_state(struct extent_state *state) +{ + unsigned long flags; + if (!state) + return; + if (atomic_dec_and_test(&state->refs)) { + WARN_ON(state->in_tree); + spin_lock_irqsave(&state_lock, flags); + list_del(&state->list); + spin_unlock_irqrestore(&state_lock, flags); + kmem_cache_free(extent_state_cache, state); + } +} +EXPORT_SYMBOL(free_extent_state); + +static struct rb_node *tree_insert(struct rb_root *root, u64 offset, + struct rb_node *node) +{ + struct rb_node ** p = &root->rb_node; + struct rb_node * parent = NULL; + struct tree_entry *entry; + + while(*p) { + parent = *p; + entry = rb_entry(parent, struct tree_entry, rb_node); + + if (offset < entry->start) + p = &(*p)->rb_left; + else if (offset > entry->end) + p = &(*p)->rb_right; + else + return parent; + } + + entry = rb_entry(node, struct tree_entry, rb_node); + entry->in_tree = 1; + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct rb_node *__tree_search(struct rb_root *root, u64 offset, + struct rb_node **prev_ret, + struct rb_node **next_ret) +{ + struct rb_node * n = root->rb_node; + struct rb_node *prev = NULL; + struct rb_node *orig_prev = NULL; + struct tree_entry *entry; + struct tree_entry *prev_entry = NULL; + + while(n) { + entry = rb_entry(n, struct tree_entry, rb_node); + prev = n; + prev_entry = entry; + + if (offset < entry->start) + n = n->rb_left; + else if (offset > entry->end) + n = n->rb_right; + else + return n; + } + + if (prev_ret) { + orig_prev = prev; + while(prev && offset > prev_entry->end) { + prev = rb_next(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *prev_ret = prev; + prev = orig_prev; + } + + if (next_ret) { + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + while(prev && offset < prev_entry->start) { + prev = rb_prev(prev); + prev_entry = rb_entry(prev, struct tree_entry, rb_node); + } + *next_ret = prev; + } + return NULL; +} + +static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) +{ + struct rb_node *prev; + struct rb_node *ret; + ret = __tree_search(root, offset, &prev, NULL); + if (!ret) + return prev; + return ret; +} + +/* + * utility function to look for merge candidates inside a given range. + * Any extents with matching state are merged together into a single + * extent in the tree. Extents with EXTENT_IO in their state field + * are not merged because the end_io handlers need to be able to do + * operations on them without sleeping (or doing allocations/splits). + * + * This should be called with the tree lock held. + */ +static int merge_state(struct extent_io_tree *tree, + struct extent_state *state) +{ + struct extent_state *other; + struct rb_node *other_node; + + if (state->state & EXTENT_IOBITS) + return 0; + + other_node = rb_prev(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->end == state->start - 1 && + other->state == state->state) { + state->start = other->start; + other->in_tree = 0; + rb_erase(&other->rb_node, &tree->state); + free_extent_state(other); + } + } + other_node = rb_next(&state->rb_node); + if (other_node) { + other = rb_entry(other_node, struct extent_state, rb_node); + if (other->start == state->end + 1 && + other->state == state->state) { + other->start = state->start; + state->in_tree = 0; + rb_erase(&state->rb_node, &tree->state); + free_extent_state(state); + } + } + return 0; +} + +/* + * insert an extent_state struct into the tree. 'bits' are set on the + * struct before it is inserted. + * + * This may return -EEXIST if the extent is already there, in which case the + * state struct is freed. + * + * The tree lock is not taken internally. This is a utility function and + * probably isn't what you want to call (see set/clear_extent_bit). + */ +static int insert_state(struct extent_io_tree *tree, + struct extent_state *state, u64 start, u64 end, + int bits) +{ + struct rb_node *node; + + if (end < start) { + printk("end < start %Lu %Lu\n", end, start); + WARN_ON(1); + } + if (bits & EXTENT_DIRTY) + tree->dirty_bytes += end - start + 1; + state->state |= bits; + state->start = start; + state->end = end; + node = tree_insert(&tree->state, end, &state->rb_node); + if (node) { + struct extent_state *found; + found = rb_entry(node, struct extent_state, rb_node); + printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); + free_extent_state(state); + return -EEXIST; + } + merge_state(tree, state); + return 0; +} + +/* + * split a given extent state struct in two, inserting the preallocated + * struct 'prealloc' as the newly created second half. 'split' indicates an + * offset inside 'orig' where it should be split. + * + * Before calling, + * the tree has 'orig' at [orig->start, orig->end]. After calling, there + * are two extent state structs in the tree: + * prealloc: [orig->start, split - 1] + * orig: [ split, orig->end ] + * + * The tree locks are not taken by this function. They need to be held + * by the caller. + */ +static int split_state(struct extent_io_tree *tree, struct extent_state *orig, + struct extent_state *prealloc, u64 split) +{ + struct rb_node *node; + prealloc->start = orig->start; + prealloc->end = split - 1; + prealloc->state = orig->state; + orig->start = split; + + node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); + if (node) { + struct extent_state *found; + found = rb_entry(node, struct extent_state, rb_node); + printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); + free_extent_state(prealloc); + return -EEXIST; + } + return 0; +} + +/* + * utility function to clear some bits in an extent state struct. + * it will optionally wake up any one waiting on this state (wake == 1), or + * forcibly remove the state from the tree (delete == 1). + * + * If no bits are set on the state struct after clearing things, the + * struct is freed and removed from the tree + */ +static int clear_state_bit(struct extent_io_tree *tree, + struct extent_state *state, int bits, int wake, + int delete) +{ + int ret = state->state & bits; + + if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { + u64 range = state->end - state->start + 1; + WARN_ON(range > tree->dirty_bytes); + tree->dirty_bytes -= range; + } + state->state &= ~bits; + if (wake) + wake_up(&state->wq); + if (delete || state->state == 0) { + if (state->in_tree) { + rb_erase(&state->rb_node, &tree->state); + state->in_tree = 0; + free_extent_state(state); + } else { + WARN_ON(1); + } + } else { + merge_state(tree, state); + } + return ret; +} + +/* + * clear some bits on a range in the tree. This may require splitting + * or inserting elements in the tree, so the gfp mask is used to + * indicate which allocations or sleeping are allowed. + * + * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove + * the given range from the tree regardless of state (ie for truncate). + * + * the range [start, end] is inclusive. + * + * This takes the tree lock, and returns < 0 on error, > 0 if any of the + * bits were already set, or zero if none of the bits were already set. + */ +int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int wake, int delete, gfp_t mask) +{ + struct extent_state *state; + struct extent_state *prealloc = NULL; + struct rb_node *node; + unsigned long flags; + int err; + int set = 0; + +again: + if (!prealloc && (mask & __GFP_WAIT)) { + prealloc = alloc_extent_state(mask); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irqsave(&tree->lock, flags); + /* + * this search will find the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + goto out; + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > end) + goto out; + WARN_ON(state->end < start); + + /* + * | ---- desired range ---- | + * | state | or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip + * bits on second half. + * + * If the extent we found extends past our range, we + * just split and search again. It'll get split again + * the next time though. + * + * If the extent we found is inside our range, we clear + * the desired bit on it. + */ + + if (state->start < start) { + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + if (err) + goto out; + if (state->end <= end) { + start = state->end + 1; + set |= clear_state_bit(tree, state, bits, + wake, delete); + } else { + start = state->start; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and clear the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + err = split_state(tree, state, prealloc, end + 1); + BUG_ON(err == -EEXIST); + + if (wake) + wake_up(&state->wq); + set |= clear_state_bit(tree, prealloc, bits, + wake, delete); + prealloc = NULL; + goto out; + } + + start = state->end + 1; + set |= clear_state_bit(tree, state, bits, wake, delete); + goto search_again; + +out: + write_unlock_irqrestore(&tree->lock, flags); + if (prealloc) + free_extent_state(prealloc); + + return set; + +search_again: + if (start > end) + goto out; + write_unlock_irqrestore(&tree->lock, flags); + if (mask & __GFP_WAIT) + cond_resched(); + goto again; +} +EXPORT_SYMBOL(clear_extent_bit); + +static int wait_on_state(struct extent_io_tree *tree, + struct extent_state *state) +{ + DEFINE_WAIT(wait); + prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); + read_unlock_irq(&tree->lock); + schedule(); + read_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + return 0; +} + +/* + * waits for one or more bits to clear on a range in the state tree. + * The range [start, end] is inclusive. + * The tree lock is taken by this function + */ +int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) +{ + struct extent_state *state; + struct rb_node *node; + + read_lock_irq(&tree->lock); +again: + while (1) { + /* + * this search will find all the extents that end after + * our range starts + */ + node = tree_search(&tree->state, start); + if (!node) + break; + + state = rb_entry(node, struct extent_state, rb_node); + + if (state->start > end) + goto out; + + if (state->state & bits) { + start = state->start; + atomic_inc(&state->refs); + wait_on_state(tree, state); + free_extent_state(state); + goto again; + } + start = state->end + 1; + + if (start > end) + break; + + if (need_resched()) { + read_unlock_irq(&tree->lock); + cond_resched(); + read_lock_irq(&tree->lock); + } + } +out: + read_unlock_irq(&tree->lock); + return 0; +} +EXPORT_SYMBOL(wait_extent_bit); + +static void set_state_bits(struct extent_io_tree *tree, + struct extent_state *state, + int bits) +{ + if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) { + u64 range = state->end - state->start + 1; + tree->dirty_bytes += range; + } + state->state |= bits; +} + +/* + * set some bits on a range in the tree. This may require allocations + * or sleeping, so the gfp mask is used to indicate what is allowed. + * + * If 'exclusive' == 1, this will fail with -EEXIST if some part of the + * range already has the desired bits set. The start of the existing + * range is returned in failed_start in this case. + * + * [start, end] is inclusive + * This takes the tree lock. + */ +int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, + int exclusive, u64 *failed_start, gfp_t mask) +{ + struct extent_state *state; + struct extent_state *prealloc = NULL; + struct rb_node *node; + unsigned long flags; + int err = 0; + int set; + u64 last_start; + u64 last_end; +again: + if (!prealloc && (mask & __GFP_WAIT)) { + prealloc = alloc_extent_state(mask); + if (!prealloc) + return -ENOMEM; + } + + write_lock_irqsave(&tree->lock, flags); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node) { + err = insert_state(tree, prealloc, start, end, bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + goto out; + } + + state = rb_entry(node, struct extent_state, rb_node); + last_start = state->start; + last_end = state->end; + + /* + * | ---- desired range ---- | + * | state | + * + * Just lock what we found and keep going + */ + if (state->start == start && state->end <= end) { + set = state->state & bits; + if (set && exclusive) { + *failed_start = state->start; + err = -EEXIST; + goto out; + } + set_state_bits(tree, state, bits); + start = state->end + 1; + merge_state(tree, state); + goto search_again; + } + + /* + * | ---- desired range ---- | + * | state | + * or + * | ------------- state -------------- | + * + * We need to split the extent we found, and may flip bits on + * second half. + * + * If the extent we found extends past our + * range, we just split and search again. It'll get split + * again the next time though. + * + * If the extent we found is inside our range, we set the + * desired bit on it. + */ + if (state->start < start) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, start); + BUG_ON(err == -EEXIST); + prealloc = NULL; + if (err) + goto out; + if (state->end <= end) { + set_state_bits(tree, state, bits); + start = state->end + 1; + merge_state(tree, state); + } else { + start = state->start; + } + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | or | state | + * + * There's a hole, we need to insert something in it and + * ignore the extent we found. + */ + if (state->start > start) { + u64 this_end; + if (end < last_start) + this_end = end; + else + this_end = last_start -1; + err = insert_state(tree, prealloc, start, this_end, + bits); + prealloc = NULL; + BUG_ON(err == -EEXIST); + if (err) + goto out; + start = this_end + 1; + goto search_again; + } + /* + * | ---- desired range ---- | + * | state | + * We need to split the extent, and set the bit + * on the first half + */ + if (state->start <= end && state->end > end) { + set = state->state & bits; + if (exclusive && set) { + *failed_start = start; + err = -EEXIST; + goto out; + } + err = split_state(tree, state, prealloc, end + 1); + BUG_ON(err == -EEXIST); + + set_state_bits(tree, prealloc, bits); + merge_state(tree, prealloc); + prealloc = NULL; + goto out; + } + + goto search_again; + +out: + write_unlock_irqrestore(&tree->lock, flags); + if (prealloc) + free_extent_state(prealloc); + + return err; + +search_again: + if (start > end) + goto out; + write_unlock_irqrestore(&tree->lock, flags); + if (mask & __GFP_WAIT) + cond_resched(); + goto again; +} +EXPORT_SYMBOL(set_extent_bit); + +/* wrappers around set/clear extent bit */ +int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_dirty); + +int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, + int bits, gfp_t mask) +{ + return set_extent_bit(tree, start, end, bits, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_bits); + +int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, + int bits, gfp_t mask) +{ + return clear_extent_bit(tree, start, end, bits, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_bits); + +int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, + EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_delalloc); + +int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, + EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_dirty); + +int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_new); + +int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_new); + +int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, + mask); +} +EXPORT_SYMBOL(set_extent_uptodate); + +int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); +} +EXPORT_SYMBOL(clear_extent_uptodate); + +int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_WRITEBACK, + 0, NULL, mask); +} +EXPORT_SYMBOL(set_extent_writeback); + +int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); +} +EXPORT_SYMBOL(clear_extent_writeback); + +int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) +{ + return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK); +} +EXPORT_SYMBOL(wait_on_extent_writeback); + +/* + * locks a range in ascending order, waiting for any locked regions + * it hits on the way. [start,end] are inclusive, and this will sleep. + */ +int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) +{ + int err; + u64 failed_start; + while (1) { + err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, + &failed_start, mask); + if (err == -EEXIST && (mask & __GFP_WAIT)) { + wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); + start = failed_start; + } else { + break; + } + WARN_ON(start > end); + } + return err; +} +EXPORT_SYMBOL(lock_extent); + +int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask); +} +EXPORT_SYMBOL(unlock_extent); + +/* + * helper function to set pages and extents in the tree dirty + */ +int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + BUG_ON(!page); + __set_page_dirty_nobuffers(page); + page_cache_release(page); + index++; + } + set_extent_dirty(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(set_range_dirty); + +/* + * helper function to set both pages and extents in the tree writeback + */ +int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + BUG_ON(!page); + set_page_writeback(page); + page_cache_release(page); + index++; + } + set_extent_writeback(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(set_range_writeback); + +int find_first_extent_bit(struct extent_io_tree *tree, u64 start, + u64 *start_ret, u64 *end_ret, int bits) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 1; + + read_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->end >= start && (state->state & bits)) { + *start_ret = state->start; + *end_ret = state->end; + ret = 0; + break; + } + node = rb_next(node); + if (!node) + break; + } +out: + read_unlock_irq(&tree->lock); + return ret; +} +EXPORT_SYMBOL(find_first_extent_bit); + +u64 find_lock_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) +{ + struct rb_node *node; + struct extent_state *state; + u64 cur_start = *start; + u64 found = 0; + u64 total_bytes = 0; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ +search_again: + node = tree_search(&tree->state, cur_start); + if (!node || IS_ERR(node)) { + *end = (u64)-1; + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (found && state->start != cur_start) { + goto out; + } + if (!(state->state & EXTENT_DELALLOC)) { + if (!found) + *end = state->end; + goto out; + } + if (!found) { + struct extent_state *prev_state; + struct rb_node *prev_node = node; + while(1) { + prev_node = rb_prev(prev_node); + if (!prev_node) + break; + prev_state = rb_entry(prev_node, + struct extent_state, + rb_node); + if (!(prev_state->state & EXTENT_DELALLOC)) + break; + state = prev_state; + node = prev_node; + } + } + if (state->state & EXTENT_LOCKED) { + DEFINE_WAIT(wait); + atomic_inc(&state->refs); + prepare_to_wait(&state->wq, &wait, + TASK_UNINTERRUPTIBLE); + write_unlock_irq(&tree->lock); + schedule(); + write_lock_irq(&tree->lock); + finish_wait(&state->wq, &wait); + free_extent_state(state); + goto search_again; + } + state->state |= EXTENT_LOCKED; + if (!found) + *start = state->start; + found++; + *end = state->end; + cur_start = state->end + 1; + node = rb_next(node); + if (!node) + break; + total_bytes += state->end - state->start + 1; + if (total_bytes >= max_bytes) + break; + } +out: + write_unlock_irq(&tree->lock); + return found; +} + +u64 count_range_bits(struct extent_io_tree *tree, + u64 *start, u64 search_end, u64 max_bytes, + unsigned long bits) +{ + struct rb_node *node; + struct extent_state *state; + u64 cur_start = *start; + u64 total_bytes = 0; + int found = 0; + + if (search_end <= cur_start) { + printk("search_end %Lu start %Lu\n", search_end, cur_start); + WARN_ON(1); + return 0; + } + + write_lock_irq(&tree->lock); + if (cur_start == 0 && bits == EXTENT_DIRTY) { + total_bytes = tree->dirty_bytes; + goto out; + } + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, cur_start); + if (!node || IS_ERR(node)) { + goto out; + } + + while(1) { + state = rb_entry(node, struct extent_state, rb_node); + if (state->start > search_end) + break; + if (state->end >= cur_start && (state->state & bits)) { + total_bytes += min(search_end, state->end) + 1 - + max(cur_start, state->start); + if (total_bytes >= max_bytes) + break; + if (!found) { + *start = state->start; + found = 1; + } + } + node = rb_next(node); + if (!node) + break; + } +out: + write_unlock_irq(&tree->lock); + return total_bytes; +} +/* + * helper function to lock both pages and extents in the tree. + * pages must be locked first. + */ +int lock_range(struct extent_io_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + int err; + + while (index <= end_index) { + page = grab_cache_page(tree->mapping, index); + if (!page) { + err = -ENOMEM; + goto failed; + } + if (IS_ERR(page)) { + err = PTR_ERR(page); + goto failed; + } + index++; + } + lock_extent(tree, start, end, GFP_NOFS); + return 0; + +failed: + /* + * we failed above in getting the page at 'index', so we undo here + * up to but not including the page at 'index' + */ + end_index = index; + index = start >> PAGE_CACHE_SHIFT; + while (index < end_index) { + page = find_get_page(tree->mapping, index); + unlock_page(page); + page_cache_release(page); + index++; + } + return err; +} +EXPORT_SYMBOL(lock_range); + +/* + * helper function to unlock both pages and extents in the tree. + */ +int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + struct page *page; + + while (index <= end_index) { + page = find_get_page(tree->mapping, index); + unlock_page(page); + page_cache_release(page); + index++; + } + unlock_extent(tree, start, end, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(unlock_range); + +int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 0; + + write_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + state->private = private; +out: + write_unlock_irq(&tree->lock); + return ret; +} + +int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) +{ + struct rb_node *node; + struct extent_state *state; + int ret = 0; + + read_lock_irq(&tree->lock); + /* + * this search will find all the extents that end after + * our range starts. + */ + node = tree_search(&tree->state, start); + if (!node || IS_ERR(node)) { + ret = -ENOENT; + goto out; + } + state = rb_entry(node, struct extent_state, rb_node); + if (state->start != start) { + ret = -ENOENT; + goto out; + } + *private = state->private; +out: + read_unlock_irq(&tree->lock); + return ret; +} + +/* + * searches a range in the state tree for a given mask. + * If 'filled' == 1, this returns 1 only if ever extent in the tree + * has the bits set. Otherwise, 1 is returned if any bit in the + * range is found set. + */ +int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int filled) +{ + struct extent_state *state = NULL; + struct rb_node *node; + int bitset = 0; + unsigned long flags; + + read_lock_irqsave(&tree->lock, flags); + node = tree_search(&tree->state, start); + while (node && start <= end) { + state = rb_entry(node, struct extent_state, rb_node); + + if (filled && state->start > start) { + bitset = 0; + break; + } + + if (state->start > end) + break; + + if (state->state & bits) { + bitset = 1; + if (!filled) + break; + } else if (filled) { + bitset = 0; + break; + } + start = state->end + 1; + if (start > end) + break; + node = rb_next(node); + if (!node) { + if (filled) + bitset = 0; + break; + } + } + read_unlock_irqrestore(&tree->lock, flags); + return bitset; +} +EXPORT_SYMBOL(test_range_bit); + +/* + * helper function to set a given page up to date if all the + * extents in the tree for that page are up to date + */ +static int check_page_uptodate(struct extent_io_tree *tree, + struct page *page) +{ + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1)) + SetPageUptodate(page); + return 0; +} + +/* + * helper function to unlock a page if all the extents in the tree + * for that page are unlocked + */ +static int check_page_locked(struct extent_io_tree *tree, + struct page *page) +{ + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0)) + unlock_page(page); + return 0; +} + +/* + * helper function to end page writeback if all the extents + * in the tree for that page are done with writeback + */ +static int check_page_writeback(struct extent_io_tree *tree, + struct page *page) +{ + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0)) + end_page_writeback(page); + return 0; +} + +/* lots and lots of room for performance fixes in the end_bio funcs */ + +/* + * after a writepage IO is done, we need to: + * clear the uptodate bits on error + * clear the writeback bits in the extent tree for this IO + * end_page_writeback if the page has no more pending IO + * + * Scheduling is not allowed, so the extent state tree is expected + * to have one and only one object corresponding to this IO. + */ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +static void end_bio_extent_writepage(struct bio *bio, int err) +#else +static int end_bio_extent_writepage(struct bio *bio, + unsigned int bytes_done, int err) +#endif +{ + const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_io_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + + do { + struct page *page = bvec->bv_page; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (!uptodate) { + clear_extent_uptodate(tree, start, end, GFP_ATOMIC); + ClearPageUptodate(page); + SetPageError(page); + } + clear_extent_writeback(tree, start, end, GFP_ATOMIC); + + if (whole_page) + end_page_writeback(page); + else + check_page_writeback(tree, page); + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, end); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +/* + * after a readpage IO is done, we need to: + * clear the uptodate bits on error + * set the uptodate bits if things worked + * set the page up to date if all extents in the tree are uptodate + * clear the lock bit in the extent tree + * unlock the page if there are no other extents locked for it + * + * Scheduling is not allowed, so the extent state tree is expected + * to have one and only one object corresponding to this IO. + */ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +static void end_bio_extent_readpage(struct bio *bio, int err) +#else +static int end_bio_extent_readpage(struct bio *bio, + unsigned int bytes_done, int err) +#endif +{ + int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_io_tree *tree = bio->bi_private; + u64 start; + u64 end; + int whole_page; + int ret; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + + do { + struct page *page = bvec->bv_page; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) + whole_page = 1; + else + whole_page = 0; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { + ret = tree->ops->readpage_end_io_hook(page, start, end); + if (ret) + uptodate = 0; + } + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + if (whole_page) + SetPageUptodate(page); + else + check_page_uptodate(tree, page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + if (whole_page) + unlock_page(page); + else + check_page_locked(tree, page); + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +/* + * IO done from prepare_write is pretty simple, we just unlock + * the structs in the extent tree when done, and set the uptodate bits + * as appropriate. + */ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +static void end_bio_extent_preparewrite(struct bio *bio, int err) +#else +static int end_bio_extent_preparewrite(struct bio *bio, + unsigned int bytes_done, int err) +#endif +{ + const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct extent_io_tree *tree = bio->bi_private; + u64 start; + u64 end; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + + do { + struct page *page = bvec->bv_page; + start = ((u64)page->index << PAGE_CACHE_SHIFT) + + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + if (--bvec >= bio->bi_io_vec) + prefetchw(&bvec->bv_page->flags); + + if (uptodate) { + set_extent_uptodate(tree, start, end, GFP_ATOMIC); + } else { + ClearPageUptodate(page); + SetPageError(page); + } + + unlock_extent(tree, start, end, GFP_ATOMIC); + + } while (bvec >= bio->bi_io_vec); + + bio_put(bio); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + +static struct bio * +extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, + gfp_t gfp_flags) +{ + struct bio *bio; + + bio = bio_alloc(gfp_flags, nr_vecs); + + if (bio == NULL && (current->flags & PF_MEMALLOC)) { + while (!bio && (nr_vecs /= 2)) + bio = bio_alloc(gfp_flags, nr_vecs); + } + + if (bio) { + bio->bi_bdev = bdev; + bio->bi_sector = first_sector; + } + return bio; +} + +static int submit_one_bio(int rw, struct bio *bio) +{ + u64 maxsector; + int ret = 0; + + bio_get(bio); + + maxsector = bio->bi_bdev->bd_inode->i_size >> 9; + if (maxsector < bio->bi_sector) { + printk("sector too large max %Lu got %llu\n", maxsector, + (unsigned long long)bio->bi_sector); + WARN_ON(1); + } + + submit_bio(rw, bio); + if (bio_flagged(bio, BIO_EOPNOTSUPP)) + ret = -EOPNOTSUPP; + bio_put(bio); + return ret; +} + +static int submit_extent_page(int rw, struct extent_io_tree *tree, + struct page *page, sector_t sector, + size_t size, unsigned long offset, + struct block_device *bdev, + struct bio **bio_ret, + unsigned long max_pages, + bio_end_io_t end_io_func) +{ + int ret = 0; + struct bio *bio; + int nr; + + if (bio_ret && *bio_ret) { + bio = *bio_ret; + if (bio->bi_sector + (bio->bi_size >> 9) != sector || + bio_add_page(bio, page, size, offset) < size) { + ret = submit_one_bio(rw, bio); + bio = NULL; + } else { + return 0; + } + } + nr = min_t(int, max_pages, bio_get_nr_vecs(bdev)); + bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); + if (!bio) { + printk("failed to allocate bio nr %d\n", nr); + } + bio_add_page(bio, page, size, offset); + bio->bi_end_io = end_io_func; + bio->bi_private = tree; + if (bio_ret) { + *bio_ret = bio; + } else { + ret = submit_one_bio(rw, bio); + } + + return ret; +} + +void set_page_extent_mapped(struct page *page) +{ + if (!PagePrivate(page)) { + SetPagePrivate(page); + WARN_ON(!page->mapping->a_ops->invalidatepage); + set_page_private(page, EXTENT_PAGE_PRIVATE); + page_cache_get(page); + } +} + +void set_page_extent_head(struct page *page, unsigned long len) +{ + set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); +} + +/* + * basic readpage implementation. Locked extent state structs are inserted + * into the tree that are removed when the IO is done (by the end_io + * handlers) + */ +static int __extent_read_full_page(struct extent_io_tree *tree, + struct page *page, + get_extent_t *get_extent, + struct bio **bio) +{ + struct inode *inode = page->mapping->host; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + u64 cur_end; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_offset = 0; + size_t iosize; + size_t blocksize = inode->i_sb->s_blocksize; + + set_page_extent_mapped(page); + + end = page_end; + lock_extent(tree, start, end, GFP_NOFS); + + while (cur <= end) { + if (cur >= last_byte) { + char *userpage; + iosize = PAGE_CACHE_SIZE - page_offset; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + break; + } + em = get_extent(inode, page, page_offset, cur, + end - cur + 1, 0); + if (IS_ERR(em) || !em) { + SetPageError(page); + unlock_extent(tree, cur, end, GFP_NOFS); + break; + } + + extent_offset = cur - em->start; + BUG_ON(extent_map_end(em) <= cur); + BUG_ON(end < cur); + + iosize = min(extent_map_end(em) - cur, end - cur + 1); + cur_end = min(extent_map_end(em) - 1, end); + iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + sector = (em->block_start + extent_offset) >> 9; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + /* we've found a hole, just zero and go on */ + if (block_start == EXTENT_MAP_HOLE) { + char *userpage; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + page_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + + set_extent_uptodate(tree, cur, cur + iosize - 1, + GFP_NOFS); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + /* the get_extent function already copied into the page */ + if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + ret = 0; + if (tree->ops && tree->ops->readpage_io_hook) { + ret = tree->ops->readpage_io_hook(page, cur, + cur + iosize - 1); + } + if (!ret) { + unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + nr -= page->index; + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, + bdev, bio, nr, + end_bio_extent_readpage); + } + if (ret) + SetPageError(page); + cur = cur + iosize; + page_offset += iosize; + nr++; + } + if (!nr) { + if (!PageError(page)) + SetPageUptodate(page); + unlock_page(page); + } + return 0; +} + +int extent_read_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent) +{ + struct bio *bio = NULL; + int ret; + + ret = __extent_read_full_page(tree, page, get_extent, &bio); + if (bio) + submit_one_bio(READ, bio); + return ret; +} +EXPORT_SYMBOL(extent_read_full_page); + +/* + * the writepage semantics are similar to regular writepage. extent + * records are inserted to lock ranges in the tree, and as dirty areas + * are found, they are marked writeback. Then the lock bits are removed + * and the end_io handler clears the writeback ranges + */ +static int __extent_writepage(struct page *page, struct writeback_control *wbc, + void *data) +{ + struct inode *inode = page->mapping->host; + struct extent_page_data *epd = data; + struct extent_io_tree *tree = epd->tree; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 delalloc_start; + u64 page_end = start + PAGE_CACHE_SIZE - 1; + u64 end; + u64 cur = start; + u64 extent_offset; + u64 last_byte = i_size_read(inode); + u64 block_start; + u64 iosize; + sector_t sector; + struct extent_map *em; + struct block_device *bdev; + int ret; + int nr = 0; + size_t page_offset = 0; + size_t blocksize; + loff_t i_size = i_size_read(inode); + unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; + u64 nr_delalloc; + u64 delalloc_end; + + WARN_ON(!PageLocked(page)); + if (page->index > end_index) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; + } + + if (page->index == end_index) { + char *userpage; + + size_t offset = i_size & (PAGE_CACHE_SIZE - 1); + + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + } + + set_page_extent_mapped(page); + + delalloc_start = start; + delalloc_end = 0; + while(delalloc_end < page_end) { + nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, + &delalloc_end, + 128 * 1024 * 1024); + if (nr_delalloc == 0) { + delalloc_start = delalloc_end + 1; + continue; + } + tree->ops->fill_delalloc(inode, delalloc_start, + delalloc_end); + clear_extent_bit(tree, delalloc_start, + delalloc_end, + EXTENT_LOCKED | EXTENT_DELALLOC, + 1, 0, GFP_NOFS); + delalloc_start = delalloc_end + 1; + } + lock_extent(tree, start, page_end, GFP_NOFS); + + end = page_end; + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { + printk("found delalloc bits after lock_extent\n"); + } + + if (last_byte <= start) { + clear_extent_dirty(tree, start, page_end, GFP_NOFS); + goto done; + } + + set_extent_uptodate(tree, start, page_end, GFP_NOFS); + blocksize = inode->i_sb->s_blocksize; + + while (cur <= end) { + if (cur >= last_byte) { + clear_extent_dirty(tree, cur, page_end, GFP_NOFS); + break; + } + em = epd->get_extent(inode, page, page_offset, cur, + end - cur + 1, 1); + if (IS_ERR(em) || !em) { + SetPageError(page); + break; + } + + extent_offset = cur - em->start; + BUG_ON(extent_map_end(em) <= cur); + BUG_ON(end < cur); + iosize = min(extent_map_end(em) - cur, end - cur + 1); + iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); + sector = (em->block_start + extent_offset) >> 9; + bdev = em->bdev; + block_start = em->block_start; + free_extent_map(em); + em = NULL; + + if (block_start == EXTENT_MAP_HOLE || + block_start == EXTENT_MAP_INLINE) { + clear_extent_dirty(tree, cur, + cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } + + /* leave this out until we have a page_mkwrite call */ + if (0 && !test_range_bit(tree, cur, cur + iosize - 1, + EXTENT_DIRTY, 0)) { + cur = cur + iosize; + page_offset += iosize; + continue; + } + clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); + if (tree->ops && tree->ops->writepage_io_hook) { + ret = tree->ops->writepage_io_hook(page, cur, + cur + iosize - 1); + } else { + ret = 0; + } + if (ret) + SetPageError(page); + else { + unsigned long max_nr = end_index + 1; + set_range_writeback(tree, cur, cur + iosize - 1); + if (!PageWriteback(page)) { + printk("warning page %lu not writeback, " + "cur %llu end %llu\n", page->index, + (unsigned long long)cur, + (unsigned long long)end); + } + + ret = submit_extent_page(WRITE, tree, page, sector, + iosize, page_offset, bdev, + &epd->bio, max_nr, + end_bio_extent_writepage); + if (ret) + SetPageError(page); + } + cur = cur + iosize; + page_offset += iosize; + nr++; + } +done: + if (nr == 0) { + /* make sure the mapping tag for page dirty gets cleared */ + set_page_writeback(page); + end_page_writeback(page); + } + unlock_extent(tree, start, page_end, GFP_NOFS); + unlock_page(page); + return 0; +} + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + +/* Taken directly from 2.6.23 for 2.6.18 back port */ +typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, + void *data); + +/** + * write_cache_pages - walk the list of dirty pages of the given address space + * and write all of them. + * @mapping: address space structure to write + * @wbc: subtract the number of written pages from *@wbc->nr_to_write + * @writepage: function called for each page + * @data: data passed to writepage function + * + * If a page is already under I/O, write_cache_pages() skips it, even + * if it's dirty. This is desirable behaviour for memory-cleaning writeback, + * but it is INCORRECT for data-integrity system calls such as fsync(). fsync() + * and msync() need to guarantee that all the data which was dirty at the time + * the call was made get new I/O started against them. If wbc->sync_mode is + * WB_SYNC_ALL then we were called for data integrity and we must wait for + * existing IO to complete. + */ +static int write_cache_pages(struct address_space *mapping, + struct writeback_control *wbc, writepage_t writepage, + void *data) +{ + struct backing_dev_info *bdi = mapping->backing_dev_info; + int ret = 0; + int done = 0; + struct pagevec pvec; + int nr_pages; + pgoff_t index; + pgoff_t end; /* Inclusive */ + int scanned = 0; + int range_whole = 0; + + if (wbc->nonblocking && bdi_write_congested(bdi)) { + wbc->encountered_congestion = 1; + return 0; + } + + pagevec_init(&pvec, 0); + if (wbc->range_cyclic) { + index = mapping->writeback_index; /* Start from prev offset */ + end = -1; + } else { + index = wbc->range_start >> PAGE_CACHE_SHIFT; + end = wbc->range_end >> PAGE_CACHE_SHIFT; + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) + range_whole = 1; + scanned = 1; + } +retry: + while (!done && (index <= end) && + (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, + PAGECACHE_TAG_DIRTY, + min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { + unsigned i; + + scanned = 1; + for (i = 0; i < nr_pages; i++) { + struct page *page = pvec.pages[i]; + + /* + * At this point we hold neither mapping->tree_lock nor + * lock on the page itself: the page may be truncated or + * invalidated (changing page->mapping to NULL), or even + * swizzled back from swapper_space to tmpfs file + * mapping + */ + lock_page(page); + + if (unlikely(page->mapping != mapping)) { + unlock_page(page); + continue; + } + + if (!wbc->range_cyclic && page->index > end) { + done = 1; + unlock_page(page); + continue; + } + + if (wbc->sync_mode != WB_SYNC_NONE) + wait_on_page_writeback(page); + + if (PageWriteback(page) || + !clear_page_dirty_for_io(page)) { + unlock_page(page); + continue; + } + + ret = (*writepage)(page, wbc, data); + + if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) { + unlock_page(page); + ret = 0; + } + if (ret || (--(wbc->nr_to_write) <= 0)) + done = 1; + if (wbc->nonblocking && bdi_write_congested(bdi)) { + wbc->encountered_congestion = 1; + done = 1; + } + } + pagevec_release(&pvec); + cond_resched(); + } + if (!scanned && !done) { + /* + * We hit the last page and there is more work to be done: wrap + * back to the start of the file + */ + scanned = 1; + index = 0; + goto retry; + } + if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) + mapping->writeback_index = index; + return ret; +} +#endif + +int extent_write_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret; + struct address_space *mapping = page->mapping; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + struct writeback_control wbc_writepages = { + .bdi = wbc->bdi, + .sync_mode = WB_SYNC_NONE, + .older_than_this = NULL, + .nr_to_write = 64, + .range_start = page_offset(page) + PAGE_CACHE_SIZE, + .range_end = (loff_t)-1, + }; + + + ret = __extent_writepage(page, wbc, &epd); + + write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); + if (epd.bio) { + submit_one_bio(WRITE, epd.bio); + } + return ret; +} +EXPORT_SYMBOL(extent_write_full_page); + + +int extent_writepages(struct extent_io_tree *tree, + struct address_space *mapping, + get_extent_t *get_extent, + struct writeback_control *wbc) +{ + int ret = 0; + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + }; + + ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); + if (epd.bio) { + submit_one_bio(WRITE, epd.bio); + } + return ret; +} +EXPORT_SYMBOL(extent_writepages); + +int extent_readpages(struct extent_io_tree *tree, + struct address_space *mapping, + struct list_head *pages, unsigned nr_pages, + get_extent_t get_extent) +{ + struct bio *bio = NULL; + unsigned page_idx; + struct pagevec pvec; + + pagevec_init(&pvec, 0); + for (page_idx = 0; page_idx < nr_pages; page_idx++) { + struct page *page = list_entry(pages->prev, struct page, lru); + + prefetchw(&page->flags); + list_del(&page->lru); + /* + * what we want to do here is call add_to_page_cache_lru, + * but that isn't exported, so we reproduce it here + */ + if (!add_to_page_cache(page, mapping, + page->index, GFP_KERNEL)) { + + /* open coding of lru_cache_add, also not exported */ + page_cache_get(page); + if (!pagevec_add(&pvec, page)) + __pagevec_lru_add(&pvec); + __extent_read_full_page(tree, page, get_extent, &bio); + } + page_cache_release(page); + } + if (pagevec_count(&pvec)) + __pagevec_lru_add(&pvec); + BUG_ON(!list_empty(pages)); + if (bio) + submit_one_bio(READ, bio); + return 0; +} +EXPORT_SYMBOL(extent_readpages); + +/* + * basic invalidatepage code, this waits on any locked or writeback + * ranges corresponding to the page, and then deletes any extent state + * records from the tree + */ +int extent_invalidatepage(struct extent_io_tree *tree, + struct page *page, unsigned long offset) +{ + u64 start = ((u64)page->index << PAGE_CACHE_SHIFT); + u64 end = start + PAGE_CACHE_SIZE - 1; + size_t blocksize = page->mapping->host->i_sb->s_blocksize; + + start += (offset + blocksize -1) & ~(blocksize - 1); + if (start > end) + return 0; + + lock_extent(tree, start, end, GFP_NOFS); + wait_on_extent_writeback(tree, start, end); + clear_extent_bit(tree, start, end, + EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, + 1, 1, GFP_NOFS); + return 0; +} +EXPORT_SYMBOL(extent_invalidatepage); + +/* + * simple commit_write call, set_range_dirty is used to mark both + * the pages and the extent records as dirty + */ +int extent_commit_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to) +{ + loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; + + set_page_extent_mapped(page); + set_page_dirty(page); + + if (pos > inode->i_size) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + return 0; +} +EXPORT_SYMBOL(extent_commit_write); + +int extent_prepare_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to, get_extent_t *get_extent) +{ + u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + u64 block_start; + u64 orig_block_start; + u64 block_end; + u64 cur_end; + struct extent_map *em; + unsigned blocksize = 1 << inode->i_blkbits; + size_t page_offset = 0; + size_t block_off_start; + size_t block_off_end; + int err = 0; + int iocount = 0; + int ret = 0; + int isnew; + + set_page_extent_mapped(page); + + block_start = (page_start + from) & ~((u64)blocksize - 1); + block_end = (page_start + to - 1) | (blocksize - 1); + orig_block_start = block_start; + + lock_extent(tree, page_start, page_end, GFP_NOFS); + while(block_start <= block_end) { + em = get_extent(inode, page, page_offset, block_start, + block_end - block_start + 1, 1); + if (IS_ERR(em) || !em) { + goto err; + } + cur_end = min(block_end, extent_map_end(em) - 1); + block_off_start = block_start & (PAGE_CACHE_SIZE - 1); + block_off_end = block_off_start + blocksize; + isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS); + + if (!PageUptodate(page) && isnew && + (block_off_end > to || block_off_start < from)) { + void *kaddr; + + kaddr = kmap_atomic(page, KM_USER0); + if (block_off_end > to) + memset(kaddr + to, 0, block_off_end - to); + if (block_off_start < from) + memset(kaddr + block_off_start, 0, + from - block_off_start); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + } + if ((em->block_start != EXTENT_MAP_HOLE && + em->block_start != EXTENT_MAP_INLINE) && + !isnew && !PageUptodate(page) && + (block_off_end > to || block_off_start < from) && + !test_range_bit(tree, block_start, cur_end, + EXTENT_UPTODATE, 1)) { + u64 sector; + u64 extent_offset = block_start - em->start; + size_t iosize; + sector = (em->block_start + extent_offset) >> 9; + iosize = (cur_end - block_start + blocksize) & + ~((u64)blocksize - 1); + /* + * we've already got the extent locked, but we + * need to split the state such that our end_bio + * handler can clear the lock. + */ + set_extent_bit(tree, block_start, + block_start + iosize - 1, + EXTENT_LOCKED, 0, NULL, GFP_NOFS); + ret = submit_extent_page(READ, tree, page, + sector, iosize, page_offset, em->bdev, + NULL, 1, + end_bio_extent_preparewrite); + iocount++; + block_start = block_start + iosize; + } else { + set_extent_uptodate(tree, block_start, cur_end, + GFP_NOFS); + unlock_extent(tree, block_start, cur_end, GFP_NOFS); + block_start = cur_end + 1; + } + page_offset = block_start & (PAGE_CACHE_SIZE - 1); + free_extent_map(em); + } + if (iocount) { + wait_extent_bit(tree, orig_block_start, + block_end, EXTENT_LOCKED); + } + check_page_uptodate(tree, page); +err: + /* FIXME, zero out newly allocated blocks on error */ + return err; +} +EXPORT_SYMBOL(extent_prepare_write); + +/* + * a helper for releasepage. As long as there are no locked extents + * in the range corresponding to the page, both state records and extent + * map records are removed + */ +int try_release_extent_mapping(struct extent_map_tree *map, + struct extent_io_tree *tree, struct page *page) +{ + struct extent_map *em; + u64 start = (u64)page->index << PAGE_CACHE_SHIFT; + u64 end = start + PAGE_CACHE_SIZE - 1; + u64 orig_start = start; + int ret = 1; + + while (start <= end) { + spin_lock(&map->lock); + em = lookup_extent_mapping(map, start, end); + if (!em || IS_ERR(em)) { + spin_unlock(&map->lock); + break; + } + if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, + EXTENT_LOCKED, 0)) { + remove_extent_mapping(map, em); + /* once for the rb tree */ + free_extent_map(em); + } + start = extent_map_end(em); + spin_unlock(&map->lock); + + /* once for us */ + free_extent_map(em); + } + if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) + ret = 0; + else + clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, + 1, 1, GFP_NOFS); + return ret; +} +EXPORT_SYMBOL(try_release_extent_mapping); + +sector_t extent_bmap(struct address_space *mapping, sector_t iblock, + get_extent_t *get_extent) +{ + struct inode *inode = mapping->host; + u64 start = iblock << inode->i_blkbits; + sector_t sector = 0; + struct extent_map *em; + + em = get_extent(inode, NULL, 0, start, (1 << inode->i_blkbits), 0); + if (!em || IS_ERR(em)) + return 0; + + if (em->block_start == EXTENT_MAP_INLINE || + em->block_start == EXTENT_MAP_HOLE) + goto out; + + sector = (em->block_start + start - em->start) >> inode->i_blkbits; +printk("bmap finds %Lu %Lu block %Lu\n", em->start, em->len, em->block_start); +out: + free_extent_map(em); + return sector; +} + +static int add_lru(struct extent_io_tree *tree, struct extent_buffer *eb) +{ + if (list_empty(&eb->lru)) { + extent_buffer_get(eb); + list_add(&eb->lru, &tree->buffer_lru); + tree->lru_size++; + if (tree->lru_size >= BUFFER_LRU_MAX) { + struct extent_buffer *rm; + rm = list_entry(tree->buffer_lru.prev, + struct extent_buffer, lru); + tree->lru_size--; + list_del_init(&rm->lru); + free_extent_buffer(rm); + } + } else + list_move(&eb->lru, &tree->buffer_lru); + return 0; +} +static struct extent_buffer *find_lru(struct extent_io_tree *tree, + u64 start, unsigned long len) +{ + struct list_head *lru = &tree->buffer_lru; + struct list_head *cur = lru->next; + struct extent_buffer *eb; + + if (list_empty(lru)) + return NULL; + + do { + eb = list_entry(cur, struct extent_buffer, lru); + if (eb->start == start && eb->len == len) { + extent_buffer_get(eb); + return eb; + } + cur = cur->next; + } while (cur != lru); + return NULL; +} + +static inline unsigned long num_extent_pages(u64 start, u64 len) +{ + return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT); +} + +static inline struct page *extent_buffer_page(struct extent_buffer *eb, + unsigned long i) +{ + struct page *p; + struct address_space *mapping; + + if (i == 0) + return eb->first_page; + i += eb->start >> PAGE_CACHE_SHIFT; + mapping = eb->first_page->mapping; + read_lock_irq(&mapping->tree_lock); + p = radix_tree_lookup(&mapping->page_tree, i); + read_unlock_irq(&mapping->tree_lock); + return p; +} + +static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, + u64 start, + unsigned long len, + gfp_t mask) +{ + struct extent_buffer *eb = NULL; + + spin_lock(&tree->lru_lock); + eb = find_lru(tree, start, len); + spin_unlock(&tree->lru_lock); + if (eb) { + return eb; + } + + eb = kmem_cache_zalloc(extent_buffer_cache, mask); + INIT_LIST_HEAD(&eb->lru); + eb->start = start; + eb->len = len; + atomic_set(&eb->refs, 1); + + return eb; +} + +static void __free_extent_buffer(struct extent_buffer *eb) +{ + kmem_cache_free(extent_buffer_cache, eb); +} + +struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, + u64 start, unsigned long len, + struct page *page0, + gfp_t mask) +{ + unsigned long num_pages = num_extent_pages(start, len); + unsigned long i; + unsigned long index = start >> PAGE_CACHE_SHIFT; + struct extent_buffer *eb; + struct page *p; + struct address_space *mapping = tree->mapping; + int uptodate = 1; + + eb = __alloc_extent_buffer(tree, start, len, mask); + if (!eb || IS_ERR(eb)) + return NULL; + + if (eb->flags & EXTENT_BUFFER_FILLED) + goto lru_add; + + if (page0) { + eb->first_page = page0; + i = 1; + index++; + page_cache_get(page0); + mark_page_accessed(page0); + set_page_extent_mapped(page0); + WARN_ON(!PageUptodate(page0)); + set_page_extent_head(page0, len); + } else { + i = 0; + } + for (; i < num_pages; i++, index++) { + p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); + if (!p) { + WARN_ON(1); + goto fail; + } + set_page_extent_mapped(p); + mark_page_accessed(p); + if (i == 0) { + eb->first_page = p; + set_page_extent_head(p, len); + } else { + set_page_private(p, EXTENT_PAGE_PRIVATE); + } + if (!PageUptodate(p)) + uptodate = 0; + unlock_page(p); + } + if (uptodate) + eb->flags |= EXTENT_UPTODATE; + eb->flags |= EXTENT_BUFFER_FILLED; + +lru_add: + spin_lock(&tree->lru_lock); + add_lru(tree, eb); + spin_unlock(&tree->lru_lock); + return eb; + +fail: + spin_lock(&tree->lru_lock); + list_del_init(&eb->lru); + spin_unlock(&tree->lru_lock); + if (!atomic_dec_and_test(&eb->refs)) + return NULL; + for (index = 1; index < i; index++) { + page_cache_release(extent_buffer_page(eb, index)); + } + if (i > 0) + page_cache_release(extent_buffer_page(eb, 0)); + __free_extent_buffer(eb); + return NULL; +} +EXPORT_SYMBOL(alloc_extent_buffer); + +struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, + u64 start, unsigned long len, + gfp_t mask) +{ + unsigned long num_pages = num_extent_pages(start, len); + unsigned long i; + unsigned long index = start >> PAGE_CACHE_SHIFT; + struct extent_buffer *eb; + struct page *p; + struct address_space *mapping = tree->mapping; + int uptodate = 1; + + eb = __alloc_extent_buffer(tree, start, len, mask); + if (!eb || IS_ERR(eb)) + return NULL; + + if (eb->flags & EXTENT_BUFFER_FILLED) + goto lru_add; + + for (i = 0; i < num_pages; i++, index++) { + p = find_lock_page(mapping, index); + if (!p) { + goto fail; + } + set_page_extent_mapped(p); + mark_page_accessed(p); + + if (i == 0) { + eb->first_page = p; + set_page_extent_head(p, len); + } else { + set_page_private(p, EXTENT_PAGE_PRIVATE); + } + + if (!PageUptodate(p)) + uptodate = 0; + unlock_page(p); + } + if (uptodate) + eb->flags |= EXTENT_UPTODATE; + eb->flags |= EXTENT_BUFFER_FILLED; + +lru_add: + spin_lock(&tree->lru_lock); + add_lru(tree, eb); + spin_unlock(&tree->lru_lock); + return eb; +fail: + spin_lock(&tree->lru_lock); + list_del_init(&eb->lru); + spin_unlock(&tree->lru_lock); + if (!atomic_dec_and_test(&eb->refs)) + return NULL; + for (index = 1; index < i; index++) { + page_cache_release(extent_buffer_page(eb, index)); + } + if (i > 0) + page_cache_release(extent_buffer_page(eb, 0)); + __free_extent_buffer(eb); + return NULL; +} +EXPORT_SYMBOL(find_extent_buffer); + +void free_extent_buffer(struct extent_buffer *eb) +{ + unsigned long i; + unsigned long num_pages; + + if (!eb) + return; + + if (!atomic_dec_and_test(&eb->refs)) + return; + + WARN_ON(!list_empty(&eb->lru)); + num_pages = num_extent_pages(eb->start, eb->len); + + for (i = 1; i < num_pages; i++) { + page_cache_release(extent_buffer_page(eb, i)); + } + page_cache_release(extent_buffer_page(eb, 0)); + __free_extent_buffer(eb); +} +EXPORT_SYMBOL(free_extent_buffer); + +int clear_extent_buffer_dirty(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + int set; + unsigned long i; + unsigned long num_pages; + struct page *page; + + u64 start = eb->start; + u64 end = start + eb->len - 1; + + set = clear_extent_dirty(tree, start, end, GFP_NOFS); + num_pages = num_extent_pages(eb->start, eb->len); + + for (i = 0; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + lock_page(page); + if (i == 0) + set_page_extent_head(page, eb->len); + else + set_page_private(page, EXTENT_PAGE_PRIVATE); + + /* + * if we're on the last page or the first page and the + * block isn't aligned on a page boundary, do extra checks + * to make sure we don't clean page that is partially dirty + */ + if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || + ((i == num_pages - 1) && + ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { + start = (u64)page->index << PAGE_CACHE_SHIFT; + end = start + PAGE_CACHE_SIZE - 1; + if (test_range_bit(tree, start, end, + EXTENT_DIRTY, 0)) { + unlock_page(page); + continue; + } + } + clear_page_dirty_for_io(page); + write_lock_irq(&page->mapping->tree_lock); + if (!PageDirty(page)) { + radix_tree_tag_clear(&page->mapping->page_tree, + page_index(page), + PAGECACHE_TAG_DIRTY); + } + write_unlock_irq(&page->mapping->tree_lock); + unlock_page(page); + } + return 0; +} +EXPORT_SYMBOL(clear_extent_buffer_dirty); + +int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + return wait_on_extent_writeback(tree, eb->start, + eb->start + eb->len - 1); +} +EXPORT_SYMBOL(wait_on_extent_buffer_writeback); + +int set_extent_buffer_dirty(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + unsigned long num_pages; + + num_pages = num_extent_pages(eb->start, eb->len); + for (i = 0; i < num_pages; i++) { + struct page *page = extent_buffer_page(eb, i); + /* writepage may need to do something special for the + * first page, we have to make sure page->private is + * properly set. releasepage may drop page->private + * on us if the page isn't already dirty. + */ + if (i == 0) { + lock_page(page); + set_page_extent_head(page, eb->len); + } else if (PagePrivate(page) && + page->private != EXTENT_PAGE_PRIVATE) { + lock_page(page); + set_page_extent_mapped(page); + unlock_page(page); + } + __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); + if (i == 0) + unlock_page(page); + } + return set_extent_dirty(tree, eb->start, + eb->start + eb->len - 1, GFP_NOFS); +} +EXPORT_SYMBOL(set_extent_buffer_dirty); + +int set_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + struct page *page; + unsigned long num_pages; + + num_pages = num_extent_pages(eb->start, eb->len); + + set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, + GFP_NOFS); + for (i = 0; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || + ((i == num_pages - 1) && + ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { + check_page_uptodate(tree, page); + continue; + } + SetPageUptodate(page); + } + return 0; +} +EXPORT_SYMBOL(set_extent_buffer_uptodate); + +int extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + if (eb->flags & EXTENT_UPTODATE) + return 1; + return test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1); +} +EXPORT_SYMBOL(extent_buffer_uptodate); + +int read_extent_buffer_pages(struct extent_io_tree *tree, + struct extent_buffer *eb, + u64 start, + int wait) +{ + unsigned long i; + unsigned long start_i; + struct page *page; + int err; + int ret = 0; + unsigned long num_pages; + + if (eb->flags & EXTENT_UPTODATE) + return 0; + + if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1, + EXTENT_UPTODATE, 1)) { + return 0; + } + + if (start) { + WARN_ON(start < eb->start); + start_i = (start >> PAGE_CACHE_SHIFT) - + (eb->start >> PAGE_CACHE_SHIFT); + } else { + start_i = 0; + } + + num_pages = num_extent_pages(eb->start, eb->len); + for (i = start_i; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + if (PageUptodate(page)) { + continue; + } + if (!wait) { + if (TestSetPageLocked(page)) { + continue; + } + } else { + lock_page(page); + } + if (!PageUptodate(page)) { + err = page->mapping->a_ops->readpage(NULL, page); + if (err) { + ret = err; + } + } else { + unlock_page(page); + } + } + + if (ret || !wait) { + return ret; + } + + for (i = start_i; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + wait_on_page_locked(page); + if (!PageUptodate(page)) { + ret = -EIO; + } + } + if (!ret) + eb->flags |= EXTENT_UPTODATE; + return ret; +} +EXPORT_SYMBOL(read_extent_buffer_pages); + +void read_extent_buffer(struct extent_buffer *eb, void *dstv, + unsigned long start, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *dst = (char *)dstv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + unsigned long num_pages = num_extent_pages(eb->start, eb->len); + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + if (!PageUptodate(page)) { + printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len); + WARN_ON(1); + } + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + kaddr = kmap_atomic(page, KM_USER1); + memcpy(dst, kaddr + offset, cur); + kunmap_atomic(kaddr, KM_USER1); + + dst += cur; + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(read_extent_buffer); + +int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, + unsigned long min_len, char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + size_t offset = start & (PAGE_CACHE_SIZE - 1); + char *kaddr; + struct page *p; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + unsigned long end_i = (start_offset + start + min_len - 1) >> + PAGE_CACHE_SHIFT; + + if (i != end_i) + return -EINVAL; + + if (i == 0) { + offset = start_offset; + *map_start = 0; + } else { + offset = 0; + *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; + } + if (start + min_len > eb->len) { +printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); + WARN_ON(1); + } + + p = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(p)); + kaddr = kmap_atomic(p, km); + *token = kaddr; + *map = kaddr + offset; + *map_len = PAGE_CACHE_SIZE - offset; + return 0; +} +EXPORT_SYMBOL(map_private_extent_buffer); + +int map_extent_buffer(struct extent_buffer *eb, unsigned long start, + unsigned long min_len, + char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km) +{ + int err; + int save = 0; + if (eb->map_token) { + unmap_extent_buffer(eb, eb->map_token, km); + eb->map_token = NULL; + save = 1; + } + err = map_private_extent_buffer(eb, start, min_len, token, map, + map_start, map_len, km); + if (!err && save) { + eb->map_token = *token; + eb->kaddr = *map; + eb->map_start = *map_start; + eb->map_len = *map_len; + } + return err; +} +EXPORT_SYMBOL(map_extent_buffer); + +void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) +{ + kunmap_atomic(token, km); +} +EXPORT_SYMBOL(unmap_extent_buffer); + +int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, + unsigned long start, + unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *ptr = (char *)ptrv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + int ret = 0; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, (PAGE_CACHE_SIZE - offset)); + + kaddr = kmap_atomic(page, KM_USER0); + ret = memcmp(ptr, kaddr + offset, cur); + kunmap_atomic(kaddr, KM_USER0); + if (ret) + break; + + ptr += cur; + len -= cur; + offset = 0; + i++; + } + return ret; +} +EXPORT_SYMBOL(memcmp_extent_buffer); + +void write_extent_buffer(struct extent_buffer *eb, const void *srcv, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + char *src = (char *)srcv; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + kaddr = kmap_atomic(page, KM_USER1); + memcpy(kaddr + offset, src, cur); + kunmap_atomic(kaddr, KM_USER1); + + src += cur; + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(write_extent_buffer); + +void memset_extent_buffer(struct extent_buffer *eb, char c, + unsigned long start, unsigned long len) +{ + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; + + WARN_ON(start > eb->len); + WARN_ON(start + len > eb->start + eb->len); + + offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(eb, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, PAGE_CACHE_SIZE - offset); + kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr + offset, c, cur); + kunmap_atomic(kaddr, KM_USER0); + + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(memset_extent_buffer); + +void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, + unsigned long dst_offset, unsigned long src_offset, + unsigned long len) +{ + u64 dst_len = dst->len; + size_t cur; + size_t offset; + struct page *page; + char *kaddr; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + + WARN_ON(src->len != dst_len); + + offset = (start_offset + dst_offset) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + while(len > 0) { + page = extent_buffer_page(dst, i); + WARN_ON(!PageUptodate(page)); + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset)); + + kaddr = kmap_atomic(page, KM_USER0); + read_extent_buffer(src, kaddr + offset, src_offset, cur); + kunmap_atomic(kaddr, KM_USER0); + + src_offset += cur; + len -= cur; + offset = 0; + i++; + } +} +EXPORT_SYMBOL(copy_extent_buffer); + +static void move_pages(struct page *dst_page, struct page *src_page, + unsigned long dst_off, unsigned long src_off, + unsigned long len) +{ + char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); + if (dst_page == src_page) { + memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len); + } else { + char *src_kaddr = kmap_atomic(src_page, KM_USER1); + char *p = dst_kaddr + dst_off + len; + char *s = src_kaddr + src_off + len; + + while (len--) + *--p = *--s; + + kunmap_atomic(src_kaddr, KM_USER1); + } + kunmap_atomic(dst_kaddr, KM_USER0); +} + +static void copy_pages(struct page *dst_page, struct page *src_page, + unsigned long dst_off, unsigned long src_off, + unsigned long len) +{ + char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); + char *src_kaddr; + + if (dst_page != src_page) + src_kaddr = kmap_atomic(src_page, KM_USER1); + else + src_kaddr = dst_kaddr; + + memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len); + kunmap_atomic(dst_kaddr, KM_USER0); + if (dst_page != src_page) + kunmap_atomic(src_kaddr, KM_USER1); +} + +void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + unsigned long src_i; + + if (src_offset + len > dst->len) { + printk("memmove bogus src_offset %lu move len %lu len %lu\n", + src_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset + len > dst->len) { + printk("memmove bogus dst_offset %lu move len %lu len %lu\n", + dst_offset, len, dst->len); + BUG_ON(1); + } + + while(len > 0) { + dst_off_in_page = (start_offset + dst_offset) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + src_off_in_page = (start_offset + src_offset) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT; + + cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - + src_off_in_page)); + cur = min_t(unsigned long, cur, + (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page)); + + copy_pages(extent_buffer_page(dst, dst_i), + extent_buffer_page(dst, src_i), + dst_off_in_page, src_off_in_page, cur); + + src_offset += cur; + dst_offset += cur; + len -= cur; + } +} +EXPORT_SYMBOL(memcpy_extent_buffer); + +void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len) +{ + size_t cur; + size_t dst_off_in_page; + size_t src_off_in_page; + unsigned long dst_end = dst_offset + len - 1; + unsigned long src_end = src_offset + len - 1; + size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); + unsigned long dst_i; + unsigned long src_i; + + if (src_offset + len > dst->len) { + printk("memmove bogus src_offset %lu move len %lu len %lu\n", + src_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset + len > dst->len) { + printk("memmove bogus dst_offset %lu move len %lu len %lu\n", + dst_offset, len, dst->len); + BUG_ON(1); + } + if (dst_offset < src_offset) { + memcpy_extent_buffer(dst, dst_offset, src_offset, len); + return; + } + while(len > 0) { + dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; + src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; + + dst_off_in_page = (start_offset + dst_end) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + src_off_in_page = (start_offset + src_end) & + ((unsigned long)PAGE_CACHE_SIZE - 1); + + cur = min_t(unsigned long, len, src_off_in_page + 1); + cur = min(cur, dst_off_in_page + 1); + move_pages(extent_buffer_page(dst, dst_i), + extent_buffer_page(dst, src_i), + dst_off_in_page - cur + 1, + src_off_in_page - cur + 1, cur); + + dst_end -= cur; + src_end -= cur; + len -= cur; + } +} +EXPORT_SYMBOL(memmove_extent_buffer); diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h new file mode 100644 index 00000000000..06be1fe84b2 --- /dev/null +++ b/fs/btrfs/extent_io.h @@ -0,0 +1,193 @@ +#ifndef __EXTENTIO__ +#define __EXTENTIO__ + +#include + +/* bits for the extent state */ +#define EXTENT_DIRTY 1 +#define EXTENT_WRITEBACK (1 << 1) +#define EXTENT_UPTODATE (1 << 2) +#define EXTENT_LOCKED (1 << 3) +#define EXTENT_NEW (1 << 4) +#define EXTENT_DELALLOC (1 << 5) +#define EXTENT_DEFRAG (1 << 6) +#define EXTENT_DEFRAG_DONE (1 << 7) +#define EXTENT_BUFFER_FILLED (1 << 8) +#define EXTENT_CSUM (1 << 9) +#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) + +/* + * page->private values. Every page that is controlled by the extent + * map has page->private set to one. + */ +#define EXTENT_PAGE_PRIVATE 1 +#define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3 + +struct extent_io_ops { + int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); + int (*writepage_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end); + void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end); +}; + +struct extent_io_tree { + struct rb_root state; + struct address_space *mapping; + u64 dirty_bytes; + rwlock_t lock; + struct extent_io_ops *ops; + spinlock_t lru_lock; + struct list_head buffer_lru; + int lru_size; +}; + +struct extent_state { + u64 start; + u64 end; /* inclusive */ + int in_tree; + struct rb_node rb_node; + wait_queue_head_t wq; + atomic_t refs; + unsigned long state; + + /* for use by the FS */ + u64 private; + + struct list_head list; +}; + +struct extent_buffer { + u64 start; + unsigned long len; + char *map_token; + char *kaddr; + unsigned long map_start; + unsigned long map_len; + struct page *first_page; + struct list_head lru; + atomic_t refs; + int flags; +}; + +struct extent_map_tree; + +typedef struct extent_map *(get_extent_t)(struct inode *inode, + struct page *page, + size_t page_offset, + u64 start, u64 len, + int create); + +void extent_io_tree_init(struct extent_io_tree *tree, + struct address_space *mapping, gfp_t mask); +void extent_io_tree_empty_lru(struct extent_io_tree *tree); +int try_release_extent_mapping(struct extent_map_tree *map, + struct extent_io_tree *tree, struct page *page); +int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); +int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); +int extent_read_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent); +int __init extent_io_init(void); +void extent_io_exit(void); + +u64 count_range_bits(struct extent_io_tree *tree, + u64 *start, u64 search_end, + u64 max_bytes, unsigned long bits); + +int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int filled); +int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, + int bits, gfp_t mask); +int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, + int bits, gfp_t mask); +int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int find_first_extent_bit(struct extent_io_tree *tree, u64 start, + u64 *start_ret, u64 *end_ret, int bits); +int extent_invalidatepage(struct extent_io_tree *tree, + struct page *page, unsigned long offset); +int extent_write_full_page(struct extent_io_tree *tree, struct page *page, + get_extent_t *get_extent, + struct writeback_control *wbc); +int extent_writepages(struct extent_io_tree *tree, + struct address_space *mapping, + get_extent_t *get_extent, + struct writeback_control *wbc); +int extent_readpages(struct extent_io_tree *tree, + struct address_space *mapping, + struct list_head *pages, unsigned nr_pages, + get_extent_t get_extent); +int extent_prepare_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to, get_extent_t *get_extent); +int extent_commit_write(struct extent_io_tree *tree, + struct inode *inode, struct page *page, + unsigned from, unsigned to); +sector_t extent_bmap(struct address_space *mapping, sector_t iblock, + get_extent_t *get_extent); +int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end); +int set_state_private(struct extent_io_tree *tree, u64 start, u64 private); +int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private); +void set_page_extent_mapped(struct page *page); + +struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, + u64 start, unsigned long len, + struct page *page0, + gfp_t mask); +struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, + u64 start, unsigned long len, + gfp_t mask); +void free_extent_buffer(struct extent_buffer *eb); +int read_extent_buffer_pages(struct extent_io_tree *tree, + struct extent_buffer *eb, u64 start, int wait); + +static inline void extent_buffer_get(struct extent_buffer *eb) +{ + atomic_inc(&eb->refs); +} + +int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, + unsigned long start, + unsigned long len); +void read_extent_buffer(struct extent_buffer *eb, void *dst, + unsigned long start, + unsigned long len); +void write_extent_buffer(struct extent_buffer *eb, const void *src, + unsigned long start, unsigned long len); +void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, + unsigned long dst_offset, unsigned long src_offset, + unsigned long len); +void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len); +void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, + unsigned long src_offset, unsigned long len); +void memset_extent_buffer(struct extent_buffer *eb, char c, + unsigned long start, unsigned long len); +int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, + struct extent_buffer *eb); +int clear_extent_buffer_dirty(struct extent_io_tree *tree, + struct extent_buffer *eb); +int set_extent_buffer_dirty(struct extent_io_tree *tree, + struct extent_buffer *eb); +int set_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb); +int extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb); +int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, + unsigned long min_len, char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km); +int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset, + unsigned long min_len, char **token, char **map, + unsigned long *map_start, + unsigned long *map_len, int km); +void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); +#endif diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 010a287fbd7..268ad8facf6 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -1,17 +1,10 @@ -#include -#include -#include -#include +#include #include -#include -#include +#include #include #include -#include -#include #include -#include -#include +#include #include "extent_map.h" /* temporary define until extent_map moves out of btrfs */ @@ -21,27 +14,6 @@ struct kmem_cache *btrfs_cache_create(const char *name, size_t size, unsigned long)); static struct kmem_cache *extent_map_cache; -static struct kmem_cache *extent_state_cache; -static struct kmem_cache *extent_buffer_cache; - -static LIST_HEAD(buffers); -static LIST_HEAD(states); - -static spinlock_t state_lock = SPIN_LOCK_UNLOCKED; -#define BUFFER_LRU_MAX 64 - -struct tree_entry { - u64 start; - u64 end; - int in_tree; - struct rb_node rb_node; -}; - -struct extent_page_data { - struct bio *bio; - struct extent_map_tree *tree; - get_extent_t *get_extent; -}; int __init extent_map_init(void) { @@ -50,72 +22,23 @@ int __init extent_map_init(void) NULL); if (!extent_map_cache) return -ENOMEM; - extent_state_cache = btrfs_cache_create("extent_state", - sizeof(struct extent_state), 0, - NULL); - if (!extent_state_cache) - goto free_map_cache; - extent_buffer_cache = btrfs_cache_create("extent_buffers", - sizeof(struct extent_buffer), 0, - NULL); - if (!extent_buffer_cache) - goto free_state_cache; return 0; - -free_state_cache: - kmem_cache_destroy(extent_state_cache); -free_map_cache: - kmem_cache_destroy(extent_map_cache); - return -ENOMEM; } void extent_map_exit(void) { - struct extent_state *state; - - while (!list_empty(&states)) { - state = list_entry(states.next, struct extent_state, list); - printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs)); - list_del(&state->list); - kmem_cache_free(extent_state_cache, state); - - } - if (extent_map_cache) kmem_cache_destroy(extent_map_cache); - if (extent_state_cache) - kmem_cache_destroy(extent_state_cache); - if (extent_buffer_cache) - kmem_cache_destroy(extent_buffer_cache); } -void extent_map_tree_init(struct extent_map_tree *tree, - struct address_space *mapping, gfp_t mask) +void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask) { tree->map.rb_node = NULL; - tree->state.rb_node = NULL; - tree->ops = NULL; - tree->dirty_bytes = 0; - rwlock_init(&tree->lock); - spin_lock_init(&tree->lru_lock); - tree->mapping = mapping; - INIT_LIST_HEAD(&tree->buffer_lru); - tree->lru_size = 0; + tree->last = NULL; + spin_lock_init(&tree->lock); } EXPORT_SYMBOL(extent_map_tree_init); -void extent_map_tree_empty_lru(struct extent_map_tree *tree) -{ - struct extent_buffer *eb; - while(!list_empty(&tree->buffer_lru)) { - eb = list_entry(tree->buffer_lru.next, struct extent_buffer, - lru); - list_del_init(&eb->lru); - free_extent_buffer(eb); - } -} -EXPORT_SYMBOL(extent_map_tree_empty_lru); - struct extent_map *alloc_extent_map(gfp_t mask) { struct extent_map *em; @@ -123,6 +46,7 @@ struct extent_map *alloc_extent_map(gfp_t mask) if (!em || IS_ERR(em)) return em; em->in_tree = 0; + em->flags = 0; atomic_set(&em->refs, 1); return em; } @@ -132,6 +56,7 @@ void free_extent_map(struct extent_map *em) { if (!em) return; + WARN_ON(atomic_read(&em->refs) == 0); if (atomic_dec_and_test(&em->refs)) { WARN_ON(em->in_tree); kmem_cache_free(extent_map_cache, em); @@ -139,64 +64,28 @@ void free_extent_map(struct extent_map *em) } EXPORT_SYMBOL(free_extent_map); - -struct extent_state *alloc_extent_state(gfp_t mask) -{ - struct extent_state *state; - unsigned long flags; - - state = kmem_cache_alloc(extent_state_cache, mask); - if (!state || IS_ERR(state)) - return state; - state->state = 0; - state->in_tree = 0; - state->private = 0; - - spin_lock_irqsave(&state_lock, flags); - list_add(&state->list, &states); - spin_unlock_irqrestore(&state_lock, flags); - - atomic_set(&state->refs, 1); - init_waitqueue_head(&state->wq); - return state; -} -EXPORT_SYMBOL(alloc_extent_state); - -void free_extent_state(struct extent_state *state) -{ - unsigned long flags; - if (!state) - return; - if (atomic_dec_and_test(&state->refs)) { - WARN_ON(state->in_tree); - spin_lock_irqsave(&state_lock, flags); - list_del(&state->list); - spin_unlock_irqrestore(&state_lock, flags); - kmem_cache_free(extent_state_cache, state); - } -} -EXPORT_SYMBOL(free_extent_state); - static struct rb_node *tree_insert(struct rb_root *root, u64 offset, struct rb_node *node) { struct rb_node ** p = &root->rb_node; struct rb_node * parent = NULL; - struct tree_entry *entry; + struct extent_map *entry; while(*p) { parent = *p; - entry = rb_entry(parent, struct tree_entry, rb_node); + entry = rb_entry(parent, struct extent_map, rb_node); + + WARN_ON(!entry->in_tree); if (offset < entry->start) p = &(*p)->rb_left; - else if (offset > entry->end) + else if (offset >= extent_map_end(entry)) p = &(*p)->rb_right; else return parent; } - entry = rb_entry(node, struct tree_entry, rb_node); + entry = rb_entry(node, struct extent_map, rb_node); entry->in_tree = 1; rb_link_node(node, parent, p); rb_insert_color(node, root); @@ -210,17 +99,19 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, struct rb_node * n = root->rb_node; struct rb_node *prev = NULL; struct rb_node *orig_prev = NULL; - struct tree_entry *entry; - struct tree_entry *prev_entry = NULL; + struct extent_map *entry; + struct extent_map *prev_entry = NULL; while(n) { - entry = rb_entry(n, struct tree_entry, rb_node); + entry = rb_entry(n, struct extent_map, rb_node); prev = n; prev_entry = entry; + WARN_ON(!entry->in_tree); + if (offset < entry->start) n = n->rb_left; - else if (offset > entry->end) + else if (offset >= extent_map_end(entry)) n = n->rb_right; else return n; @@ -228,19 +119,19 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, if (prev_ret) { orig_prev = prev; - while(prev && offset > prev_entry->end) { + while(prev && offset >= extent_map_end(prev_entry)) { prev = rb_next(prev); - prev_entry = rb_entry(prev, struct tree_entry, rb_node); + prev_entry = rb_entry(prev, struct extent_map, rb_node); } *prev_ret = prev; prev = orig_prev; } if (next_ret) { - prev_entry = rb_entry(prev, struct tree_entry, rb_node); + prev_entry = rb_entry(prev, struct extent_map, rb_node); while(prev && offset < prev_entry->start) { prev = rb_prev(prev); - prev_entry = rb_entry(prev, struct tree_entry, rb_node); + prev_entry = rb_entry(prev, struct extent_map, rb_node); } *next_ret = prev; } @@ -257,22 +148,26 @@ static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) return ret; } -static int tree_delete(struct rb_root *root, u64 offset) +static int mergable_maps(struct extent_map *prev, struct extent_map *next) { - struct rb_node *node; - struct tree_entry *entry; - - node = __tree_search(root, offset, NULL, NULL); - if (!node) - return -ENOENT; - entry = rb_entry(node, struct tree_entry, rb_node); - entry->in_tree = 0; - rb_erase(node, root); + if (extent_map_end(prev) == next->start && + prev->flags == next->flags && + prev->bdev == next->bdev && + ((next->block_start == EXTENT_MAP_HOLE && + prev->block_start == EXTENT_MAP_HOLE) || + (next->block_start == EXTENT_MAP_INLINE && + prev->block_start == EXTENT_MAP_INLINE) || + (next->block_start == EXTENT_MAP_DELALLOC && + prev->block_start == EXTENT_MAP_DELALLOC) || + (next->block_start < EXTENT_MAP_LAST_BYTE - 1 && + next->block_start == extent_map_block_end(prev)))) { + return 1; + } return 0; } /* - * add_extent_mapping tries a simple backward merge with existing + * add_extent_mapping tries a simple forward/backward merge with existing * mappings. The extent_map struct passed in will be inserted into * the tree directly (no copies made, just a reference taken). */ @@ -280,13 +175,12 @@ int add_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) { int ret = 0; - struct extent_map *prev = NULL; + struct extent_map *merge = NULL; struct rb_node *rb; - write_lock_irq(&tree->lock); - rb = tree_insert(&tree->map, em->end, &em->rb_node); + rb = tree_insert(&tree->map, em->start, &em->rb_node); if (rb) { - prev = rb_entry(rb, struct extent_map, rb_node); + merge = rb_entry(rb, struct extent_map, rb_node); ret = -EEXIST; goto out; } @@ -294,53 +188,60 @@ int add_extent_mapping(struct extent_map_tree *tree, if (em->start != 0) { rb = rb_prev(&em->rb_node); if (rb) - prev = rb_entry(rb, struct extent_map, rb_node); - if (prev && prev->end + 1 == em->start && - ((em->block_start == EXTENT_MAP_HOLE && - prev->block_start == EXTENT_MAP_HOLE) || - (em->block_start == EXTENT_MAP_INLINE && - prev->block_start == EXTENT_MAP_INLINE) || - (em->block_start == EXTENT_MAP_DELALLOC && - prev->block_start == EXTENT_MAP_DELALLOC) || - (em->block_start < EXTENT_MAP_DELALLOC - 1 && - em->block_start == prev->block_end + 1))) { - em->start = prev->start; - em->block_start = prev->block_start; - rb_erase(&prev->rb_node, &tree->map); - prev->in_tree = 0; - free_extent_map(prev); + merge = rb_entry(rb, struct extent_map, rb_node); + if (rb && mergable_maps(merge, em)) { + em->start = merge->start; + em->len += merge->len; + em->block_start = merge->block_start; + merge->in_tree = 0; + rb_erase(&merge->rb_node, &tree->map); + free_extent_map(merge); } } + rb = rb_next(&em->rb_node); + if (rb) + merge = rb_entry(rb, struct extent_map, rb_node); + if (rb && mergable_maps(em, merge)) { + em->len += merge->len; + rb_erase(&merge->rb_node, &tree->map); + merge->in_tree = 0; + free_extent_map(merge); + } + tree->last = em; out: - write_unlock_irq(&tree->lock); return ret; } EXPORT_SYMBOL(add_extent_mapping); +static u64 range_end(u64 start, u64 len) +{ + if (start + len < start) + return (u64)-1; + return start + len; +} + /* * lookup_extent_mapping returns the first extent_map struct in the - * tree that intersects the [start, end] (inclusive) range. There may + * tree that intersects the [start, len] range. There may * be additional objects in the tree that intersect, so check the object * returned carefully to make sure you don't need additional lookups. */ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, - u64 start, u64 end) + u64 start, u64 len) { struct extent_map *em; struct rb_node *rb_node; - struct rb_node *prev = NULL; - struct rb_node *next = NULL; + struct rb_node *prev = NULL; struct rb_node *next = NULL; u64 end = range_end(start, len); em = tree->last; if (em && end > em->start && start < extent_map_end(em)) goto found; - read_lock_irq(&tree->lock); rb_node = __tree_search(&tree->map, start, &prev, &next); if (!rb_node && prev) { em = rb_entry(prev, struct extent_map, rb_node); - if (em->start <= end && em->end >= start) + if (end > em->start && start < extent_map_end(em)) goto found; } if (!rb_node && next) { em = rb_entry(next, struct extent_map, rb_node); - if (em->start <= end && em->end >= start) + if (end > em->start && start < extent_map_end(em)) goto found; } if (!rb_node) { @@ -352,14 +253,16 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, goto out; } em = rb_entry(rb_node, struct extent_map, rb_node); - if (em->end < start || em->start > end) { - em = NULL; - goto out; - } + if (end > em->start && start < extent_map_end(em)) + goto found; + + em = NULL; + goto out; + found: atomic_inc(&em->refs); + tree->last = em; out: - read_unlock_irq(&tree->lock); return em; } EXPORT_SYMBOL(lookup_extent_mapping); @@ -370,2866 +273,12 @@ EXPORT_SYMBOL(lookup_extent_mapping); */ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) { - int ret; + int ret = 0; - write_lock_irq(&tree->lock); - ret = tree_delete(&tree->map, em->end); - write_unlock_irq(&tree->lock); + rb_erase(&em->rb_node, &tree->map); + em->in_tree = 0; + if (tree->last == em) + tree->last = NULL; return ret; } EXPORT_SYMBOL(remove_extent_mapping); - -/* - * utility function to look for merge candidates inside a given range. - * Any extents with matching state are merged together into a single - * extent in the tree. Extents with EXTENT_IO in their state field - * are not merged because the end_io handlers need to be able to do - * operations on them without sleeping (or doing allocations/splits). - * - * This should be called with the tree lock held. - */ -static int merge_state(struct extent_map_tree *tree, - struct extent_state *state) -{ - struct extent_state *other; - struct rb_node *other_node; - - if (state->state & EXTENT_IOBITS) - return 0; - - other_node = rb_prev(&state->rb_node); - if (other_node) { - other = rb_entry(other_node, struct extent_state, rb_node); - if (other->end == state->start - 1 && - other->state == state->state) { - state->start = other->start; - other->in_tree = 0; - rb_erase(&other->rb_node, &tree->state); - free_extent_state(other); - } - } - other_node = rb_next(&state->rb_node); - if (other_node) { - other = rb_entry(other_node, struct extent_state, rb_node); - if (other->start == state->end + 1 && - other->state == state->state) { - other->start = state->start; - state->in_tree = 0; - rb_erase(&state->rb_node, &tree->state); - free_extent_state(state); - } - } - return 0; -} - -/* - * insert an extent_state struct into the tree. 'bits' are set on the - * struct before it is inserted. - * - * This may return -EEXIST if the extent is already there, in which case the - * state struct is freed. - * - * The tree lock is not taken internally. This is a utility function and - * probably isn't what you want to call (see set/clear_extent_bit). - */ -static int insert_state(struct extent_map_tree *tree, - struct extent_state *state, u64 start, u64 end, - int bits) -{ - struct rb_node *node; - - if (end < start) { - printk("end < start %Lu %Lu\n", end, start); - WARN_ON(1); - } - if (bits & EXTENT_DIRTY) - tree->dirty_bytes += end - start + 1; - state->state |= bits; - state->start = start; - state->end = end; - node = tree_insert(&tree->state, end, &state->rb_node); - if (node) { - struct extent_state *found; - found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); - free_extent_state(state); - return -EEXIST; - } - merge_state(tree, state); - return 0; -} - -/* - * split a given extent state struct in two, inserting the preallocated - * struct 'prealloc' as the newly created second half. 'split' indicates an - * offset inside 'orig' where it should be split. - * - * Before calling, - * the tree has 'orig' at [orig->start, orig->end]. After calling, there - * are two extent state structs in the tree: - * prealloc: [orig->start, split - 1] - * orig: [ split, orig->end ] - * - * The tree locks are not taken by this function. They need to be held - * by the caller. - */ -static int split_state(struct extent_map_tree *tree, struct extent_state *orig, - struct extent_state *prealloc, u64 split) -{ - struct rb_node *node; - prealloc->start = orig->start; - prealloc->end = split - 1; - prealloc->state = orig->state; - orig->start = split; - - node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); - if (node) { - struct extent_state *found; - found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); - free_extent_state(prealloc); - return -EEXIST; - } - return 0; -} - -/* - * utility function to clear some bits in an extent state struct. - * it will optionally wake up any one waiting on this state (wake == 1), or - * forcibly remove the state from the tree (delete == 1). - * - * If no bits are set on the state struct after clearing things, the - * struct is freed and removed from the tree - */ -static int clear_state_bit(struct extent_map_tree *tree, - struct extent_state *state, int bits, int wake, - int delete) -{ - int ret = state->state & bits; - - if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { - u64 range = state->end - state->start + 1; - WARN_ON(range > tree->dirty_bytes); - tree->dirty_bytes -= range; - } - state->state &= ~bits; - if (wake) - wake_up(&state->wq); - if (delete || state->state == 0) { - if (state->in_tree) { - rb_erase(&state->rb_node, &tree->state); - state->in_tree = 0; - free_extent_state(state); - } else { - WARN_ON(1); - } - } else { - merge_state(tree, state); - } - return ret; -} - -/* - * clear some bits on a range in the tree. This may require splitting - * or inserting elements in the tree, so the gfp mask is used to - * indicate which allocations or sleeping are allowed. - * - * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove - * the given range from the tree regardless of state (ie for truncate). - * - * the range [start, end] is inclusive. - * - * This takes the tree lock, and returns < 0 on error, > 0 if any of the - * bits were already set, or zero if none of the bits were already set. - */ -int clear_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, - int bits, int wake, int delete, gfp_t mask) -{ - struct extent_state *state; - struct extent_state *prealloc = NULL; - struct rb_node *node; - unsigned long flags; - int err; - int set = 0; - -again: - if (!prealloc && (mask & __GFP_WAIT)) { - prealloc = alloc_extent_state(mask); - if (!prealloc) - return -ENOMEM; - } - - write_lock_irqsave(&tree->lock, flags); - /* - * this search will find the extents that end after - * our range starts - */ - node = tree_search(&tree->state, start); - if (!node) - goto out; - state = rb_entry(node, struct extent_state, rb_node); - if (state->start > end) - goto out; - WARN_ON(state->end < start); - - /* - * | ---- desired range ---- | - * | state | or - * | ------------- state -------------- | - * - * We need to split the extent we found, and may flip - * bits on second half. - * - * If the extent we found extends past our range, we - * just split and search again. It'll get split again - * the next time though. - * - * If the extent we found is inside our range, we clear - * the desired bit on it. - */ - - if (state->start < start) { - err = split_state(tree, state, prealloc, start); - BUG_ON(err == -EEXIST); - prealloc = NULL; - if (err) - goto out; - if (state->end <= end) { - start = state->end + 1; - set |= clear_state_bit(tree, state, bits, - wake, delete); - } else { - start = state->start; - } - goto search_again; - } - /* - * | ---- desired range ---- | - * | state | - * We need to split the extent, and clear the bit - * on the first half - */ - if (state->start <= end && state->end > end) { - err = split_state(tree, state, prealloc, end + 1); - BUG_ON(err == -EEXIST); - - if (wake) - wake_up(&state->wq); - set |= clear_state_bit(tree, prealloc, bits, - wake, delete); - prealloc = NULL; - goto out; - } - - start = state->end + 1; - set |= clear_state_bit(tree, state, bits, wake, delete); - goto search_again; - -out: - write_unlock_irqrestore(&tree->lock, flags); - if (prealloc) - free_extent_state(prealloc); - - return set; - -search_again: - if (start > end) - goto out; - write_unlock_irqrestore(&tree->lock, flags); - if (mask & __GFP_WAIT) - cond_resched(); - goto again; -} -EXPORT_SYMBOL(clear_extent_bit); - -static int wait_on_state(struct extent_map_tree *tree, - struct extent_state *state) -{ - DEFINE_WAIT(wait); - prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - read_unlock_irq(&tree->lock); - schedule(); - read_lock_irq(&tree->lock); - finish_wait(&state->wq, &wait); - return 0; -} - -/* - * waits for one or more bits to clear on a range in the state tree. - * The range [start, end] is inclusive. - * The tree lock is taken by this function - */ -int wait_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits) -{ - struct extent_state *state; - struct rb_node *node; - - read_lock_irq(&tree->lock); -again: - while (1) { - /* - * this search will find all the extents that end after - * our range starts - */ - node = tree_search(&tree->state, start); - if (!node) - break; - - state = rb_entry(node, struct extent_state, rb_node); - - if (state->start > end) - goto out; - - if (state->state & bits) { - start = state->start; - atomic_inc(&state->refs); - wait_on_state(tree, state); - free_extent_state(state); - goto again; - } - start = state->end + 1; - - if (start > end) - break; - - if (need_resched()) { - read_unlock_irq(&tree->lock); - cond_resched(); - read_lock_irq(&tree->lock); - } - } -out: - read_unlock_irq(&tree->lock); - return 0; -} -EXPORT_SYMBOL(wait_extent_bit); - -static void set_state_bits(struct extent_map_tree *tree, - struct extent_state *state, - int bits) -{ - if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) { - u64 range = state->end - state->start + 1; - tree->dirty_bytes += range; - } - state->state |= bits; -} - -/* - * set some bits on a range in the tree. This may require allocations - * or sleeping, so the gfp mask is used to indicate what is allowed. - * - * If 'exclusive' == 1, this will fail with -EEXIST if some part of the - * range already has the desired bits set. The start of the existing - * range is returned in failed_start in this case. - * - * [start, end] is inclusive - * This takes the tree lock. - */ -int set_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits, - int exclusive, u64 *failed_start, gfp_t mask) -{ - struct extent_state *state; - struct extent_state *prealloc = NULL; - struct rb_node *node; - unsigned long flags; - int err = 0; - int set; - u64 last_start; - u64 last_end; -again: - if (!prealloc && (mask & __GFP_WAIT)) { - prealloc = alloc_extent_state(mask); - if (!prealloc) - return -ENOMEM; - } - - write_lock_irqsave(&tree->lock, flags); - /* - * this search will find all the extents that end after - * our range starts. - */ - node = tree_search(&tree->state, start); - if (!node) { - err = insert_state(tree, prealloc, start, end, bits); - prealloc = NULL; - BUG_ON(err == -EEXIST); - goto out; - } - - state = rb_entry(node, struct extent_state, rb_node); - last_start = state->start; - last_end = state->end; - - /* - * | ---- desired range ---- | - * | state | - * - * Just lock what we found and keep going - */ - if (state->start == start && state->end <= end) { - set = state->state & bits; - if (set && exclusive) { - *failed_start = state->start; - err = -EEXIST; - goto out; - } - set_state_bits(tree, state, bits); - start = state->end + 1; - merge_state(tree, state); - goto search_again; - } - - /* - * | ---- desired range ---- | - * | state | - * or - * | ------------- state -------------- | - * - * We need to split the extent we found, and may flip bits on - * second half. - * - * If the extent we found extends past our - * range, we just split and search again. It'll get split - * again the next time though. - * - * If the extent we found is inside our range, we set the - * desired bit on it. - */ - if (state->start < start) { - set = state->state & bits; - if (exclusive && set) { - *failed_start = start; - err = -EEXIST; - goto out; - } - err = split_state(tree, state, prealloc, start); - BUG_ON(err == -EEXIST); - prealloc = NULL; - if (err) - goto out; - if (state->end <= end) { - set_state_bits(tree, state, bits); - start = state->end + 1; - merge_state(tree, state); - } else { - start = state->start; - } - goto search_again; - } - /* - * | ---- desired range ---- | - * | state | or | state | - * - * There's a hole, we need to insert something in it and - * ignore the extent we found. - */ - if (state->start > start) { - u64 this_end; - if (end < last_start) - this_end = end; - else - this_end = last_start -1; - err = insert_state(tree, prealloc, start, this_end, - bits); - prealloc = NULL; - BUG_ON(err == -EEXIST); - if (err) - goto out; - start = this_end + 1; - goto search_again; - } - /* - * | ---- desired range ---- | - * | state | - * We need to split the extent, and set the bit - * on the first half - */ - if (state->start <= end && state->end > end) { - set = state->state & bits; - if (exclusive && set) { - *failed_start = start; - err = -EEXIST; - goto out; - } - err = split_state(tree, state, prealloc, end + 1); - BUG_ON(err == -EEXIST); - - set_state_bits(tree, prealloc, bits); - merge_state(tree, prealloc); - prealloc = NULL; - goto out; - } - - goto search_again; - -out: - write_unlock_irqrestore(&tree->lock, flags); - if (prealloc) - free_extent_state(prealloc); - - return err; - -search_again: - if (start > end) - goto out; - write_unlock_irqrestore(&tree->lock, flags); - if (mask & __GFP_WAIT) - cond_resched(); - goto again; -} -EXPORT_SYMBOL(set_extent_bit); - -/* wrappers around set/clear extent bit */ -int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_dirty); - -int set_extent_bits(struct extent_map_tree *tree, u64 start, u64 end, - int bits, gfp_t mask) -{ - return set_extent_bit(tree, start, end, bits, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_bits); - -int clear_extent_bits(struct extent_map_tree *tree, u64 start, u64 end, - int bits, gfp_t mask) -{ - return clear_extent_bit(tree, start, end, bits, 0, 0, mask); -} -EXPORT_SYMBOL(clear_extent_bits); - -int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, - EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_delalloc); - -int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, - EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask); -} -EXPORT_SYMBOL(clear_extent_dirty); - -int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_new); - -int clear_extent_new(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask); -} -EXPORT_SYMBOL(clear_extent_new); - -int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, - mask); -} -EXPORT_SYMBOL(set_extent_uptodate); - -int clear_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); -} -EXPORT_SYMBOL(clear_extent_uptodate); - -int set_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return set_extent_bit(tree, start, end, EXTENT_WRITEBACK, - 0, NULL, mask); -} -EXPORT_SYMBOL(set_extent_writeback); - -int clear_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); -} -EXPORT_SYMBOL(clear_extent_writeback); - -int wait_on_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end) -{ - return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK); -} -EXPORT_SYMBOL(wait_on_extent_writeback); - -/* - * locks a range in ascending order, waiting for any locked regions - * it hits on the way. [start,end] are inclusive, and this will sleep. - */ -int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask) -{ - int err; - u64 failed_start; - while (1) { - err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, - &failed_start, mask); - if (err == -EEXIST && (mask & __GFP_WAIT)) { - wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); - start = failed_start; - } else { - break; - } - WARN_ON(start > end); - } - return err; -} -EXPORT_SYMBOL(lock_extent); - -int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask) -{ - return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask); -} -EXPORT_SYMBOL(unlock_extent); - -/* - * helper function to set pages and extents in the tree dirty - */ -int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - - while (index <= end_index) { - page = find_get_page(tree->mapping, index); - BUG_ON(!page); - __set_page_dirty_nobuffers(page); - page_cache_release(page); - index++; - } - set_extent_dirty(tree, start, end, GFP_NOFS); - return 0; -} -EXPORT_SYMBOL(set_range_dirty); - -/* - * helper function to set both pages and extents in the tree writeback - */ -int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - - while (index <= end_index) { - page = find_get_page(tree->mapping, index); - BUG_ON(!page); - set_page_writeback(page); - page_cache_release(page); - index++; - } - set_extent_writeback(tree, start, end, GFP_NOFS); - return 0; -} -EXPORT_SYMBOL(set_range_writeback); - -int find_first_extent_bit(struct extent_map_tree *tree, u64 start, - u64 *start_ret, u64 *end_ret, int bits) -{ - struct rb_node *node; - struct extent_state *state; - int ret = 1; - - read_lock_irq(&tree->lock); - /* - * this search will find all the extents that end after - * our range starts. - */ - node = tree_search(&tree->state, start); - if (!node || IS_ERR(node)) { - goto out; - } - - while(1) { - state = rb_entry(node, struct extent_state, rb_node); - if (state->end >= start && (state->state & bits)) { - *start_ret = state->start; - *end_ret = state->end; - ret = 0; - break; - } - node = rb_next(node); - if (!node) - break; - } -out: - read_unlock_irq(&tree->lock); - return ret; -} -EXPORT_SYMBOL(find_first_extent_bit); - -u64 find_lock_delalloc_range(struct extent_map_tree *tree, - u64 *start, u64 *end, u64 max_bytes) -{ - struct rb_node *node; - struct extent_state *state; - u64 cur_start = *start; - u64 found = 0; - u64 total_bytes = 0; - - write_lock_irq(&tree->lock); - /* - * this search will find all the extents that end after - * our range starts. - */ -search_again: - node = tree_search(&tree->state, cur_start); - if (!node || IS_ERR(node)) { - *end = (u64)-1; - goto out; - } - - while(1) { - state = rb_entry(node, struct extent_state, rb_node); - if (found && state->start != cur_start) { - goto out; - } - if (!(state->state & EXTENT_DELALLOC)) { - if (!found) - *end = state->end; - goto out; - } - if (!found) { - struct extent_state *prev_state; - struct rb_node *prev_node = node; - while(1) { - prev_node = rb_prev(prev_node); - if (!prev_node) - break; - prev_state = rb_entry(prev_node, - struct extent_state, - rb_node); - if (!(prev_state->state & EXTENT_DELALLOC)) - break; - state = prev_state; - node = prev_node; - } - } - if (state->state & EXTENT_LOCKED) { - DEFINE_WAIT(wait); - atomic_inc(&state->refs); - prepare_to_wait(&state->wq, &wait, - TASK_UNINTERRUPTIBLE); - write_unlock_irq(&tree->lock); - schedule(); - write_lock_irq(&tree->lock); - finish_wait(&state->wq, &wait); - free_extent_state(state); - goto search_again; - } - state->state |= EXTENT_LOCKED; - if (!found) - *start = state->start; - found++; - *end = state->end; - cur_start = state->end + 1; - node = rb_next(node); - if (!node) - break; - total_bytes += state->end - state->start + 1; - if (total_bytes >= max_bytes) - break; - } -out: - write_unlock_irq(&tree->lock); - return found; -} - -u64 count_range_bits(struct extent_map_tree *tree, - u64 *start, u64 search_end, u64 max_bytes, - unsigned long bits) -{ - struct rb_node *node; - struct extent_state *state; - u64 cur_start = *start; - u64 total_bytes = 0; - int found = 0; - - if (search_end <= cur_start) { - printk("search_end %Lu start %Lu\n", search_end, cur_start); - WARN_ON(1); - return 0; - } - - write_lock_irq(&tree->lock); - if (cur_start == 0 && bits == EXTENT_DIRTY) { - total_bytes = tree->dirty_bytes; - goto out; - } - /* - * this search will find all the extents that end after - * our range starts. - */ - node = tree_search(&tree->state, cur_start); - if (!node || IS_ERR(node)) { - goto out; - } - - while(1) { - state = rb_entry(node, struct extent_state, rb_node); - if (state->start > search_end) - break; - if (state->end >= cur_start && (state->state & bits)) { - total_bytes += min(search_end, state->end) + 1 - - max(cur_start, state->start); - if (total_bytes >= max_bytes) - break; - if (!found) { - *start = state->start; - found = 1; - } - } - node = rb_next(node); - if (!node) - break; - } -out: - write_unlock_irq(&tree->lock); - return total_bytes; -} -/* - * helper function to lock both pages and extents in the tree. - * pages must be locked first. - */ -int lock_range(struct extent_map_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - int err; - - while (index <= end_index) { - page = grab_cache_page(tree->mapping, index); - if (!page) { - err = -ENOMEM; - goto failed; - } - if (IS_ERR(page)) { - err = PTR_ERR(page); - goto failed; - } - index++; - } - lock_extent(tree, start, end, GFP_NOFS); - return 0; - -failed: - /* - * we failed above in getting the page at 'index', so we undo here - * up to but not including the page at 'index' - */ - end_index = index; - index = start >> PAGE_CACHE_SHIFT; - while (index < end_index) { - page = find_get_page(tree->mapping, index); - unlock_page(page); - page_cache_release(page); - index++; - } - return err; -} -EXPORT_SYMBOL(lock_range); - -/* - * helper function to unlock both pages and extents in the tree. - */ -int unlock_range(struct extent_map_tree *tree, u64 start, u64 end) -{ - unsigned long index = start >> PAGE_CACHE_SHIFT; - unsigned long end_index = end >> PAGE_CACHE_SHIFT; - struct page *page; - - while (index <= end_index) { - page = find_get_page(tree->mapping, index); - unlock_page(page); - page_cache_release(page); - index++; - } - unlock_extent(tree, start, end, GFP_NOFS); - return 0; -} -EXPORT_SYMBOL(unlock_range); - -int set_state_private(struct extent_map_tree *tree, u64 start, u64 private) -{ - struct rb_node *node; - struct extent_state *state; - int ret = 0; - - write_lock_irq(&tree->lock); - /* - * this search will find all the extents that end after - * our range starts. - */ - node = tree_search(&tree->state, start); - if (!node || IS_ERR(node)) { - ret = -ENOENT; - goto out; - } - state = rb_entry(node, struct extent_state, rb_node); - if (state->start != start) { - ret = -ENOENT; - goto out; - } - state->private = private; -out: - write_unlock_irq(&tree->lock); - return ret; -} - -int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private) -{ - struct rb_node *node; - struct extent_state *state; - int ret = 0; - - read_lock_irq(&tree->lock); - /* - * this search will find all the extents that end after - * our range starts. - */ - node = tree_search(&tree->state, start); - if (!node || IS_ERR(node)) { - ret = -ENOENT; - goto out; - } - state = rb_entry(node, struct extent_state, rb_node); - if (state->start != start) { - ret = -ENOENT; - goto out; - } - *private = state->private; -out: - read_unlock_irq(&tree->lock); - return ret; -} - -/* - * searches a range in the state tree for a given mask. - * If 'filled' == 1, this returns 1 only if ever extent in the tree - * has the bits set. Otherwise, 1 is returned if any bit in the - * range is found set. - */ -int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end, - int bits, int filled) -{ - struct extent_state *state = NULL; - struct rb_node *node; - int bitset = 0; - - read_lock_irq(&tree->lock); - node = tree_search(&tree->state, start); - while (node && start <= end) { - state = rb_entry(node, struct extent_state, rb_node); - - if (filled && state->start > start) { - bitset = 0; - break; - } - - if (state->start > end) - break; - - if (state->state & bits) { - bitset = 1; - if (!filled) - break; - } else if (filled) { - bitset = 0; - break; - } - start = state->end + 1; - if (start > end) - break; - node = rb_next(node); - if (!node) { - if (filled) - bitset = 0; - break; - } - } - read_unlock_irq(&tree->lock); - return bitset; -} -EXPORT_SYMBOL(test_range_bit); - -/* - * helper function to set a given page up to date if all the - * extents in the tree for that page are up to date - */ -static int check_page_uptodate(struct extent_map_tree *tree, - struct page *page) -{ - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 end = start + PAGE_CACHE_SIZE - 1; - if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1)) - SetPageUptodate(page); - return 0; -} - -/* - * helper function to unlock a page if all the extents in the tree - * for that page are unlocked - */ -static int check_page_locked(struct extent_map_tree *tree, - struct page *page) -{ - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 end = start + PAGE_CACHE_SIZE - 1; - if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0)) - unlock_page(page); - return 0; -} - -/* - * helper function to end page writeback if all the extents - * in the tree for that page are done with writeback - */ -static int check_page_writeback(struct extent_map_tree *tree, - struct page *page) -{ - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 end = start + PAGE_CACHE_SIZE - 1; - if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0)) - end_page_writeback(page); - return 0; -} - -/* lots and lots of room for performance fixes in the end_bio funcs */ - -/* - * after a writepage IO is done, we need to: - * clear the uptodate bits on error - * clear the writeback bits in the extent tree for this IO - * end_page_writeback if the page has no more pending IO - * - * Scheduling is not allowed, so the extent state tree is expected - * to have one and only one object corresponding to this IO. - */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) -static void end_bio_extent_writepage(struct bio *bio, int err) -#else -static int end_bio_extent_writepage(struct bio *bio, - unsigned int bytes_done, int err) -#endif -{ - const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); - struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_map_tree *tree = bio->bi_private; - u64 start; - u64 end; - int whole_page; - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - - do { - struct page *page = bvec->bv_page; - start = ((u64)page->index << PAGE_CACHE_SHIFT) + - bvec->bv_offset; - end = start + bvec->bv_len - 1; - - if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) - whole_page = 1; - else - whole_page = 0; - - if (--bvec >= bio->bi_io_vec) - prefetchw(&bvec->bv_page->flags); - - if (!uptodate) { - clear_extent_uptodate(tree, start, end, GFP_ATOMIC); - ClearPageUptodate(page); - SetPageError(page); - } - clear_extent_writeback(tree, start, end, GFP_ATOMIC); - - if (whole_page) - end_page_writeback(page); - else - check_page_writeback(tree, page); - if (tree->ops && tree->ops->writepage_end_io_hook) - tree->ops->writepage_end_io_hook(page, start, end); - } while (bvec >= bio->bi_io_vec); - - bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif -} - -/* - * after a readpage IO is done, we need to: - * clear the uptodate bits on error - * set the uptodate bits if things worked - * set the page up to date if all extents in the tree are uptodate - * clear the lock bit in the extent tree - * unlock the page if there are no other extents locked for it - * - * Scheduling is not allowed, so the extent state tree is expected - * to have one and only one object corresponding to this IO. - */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) -static void end_bio_extent_readpage(struct bio *bio, int err) -#else -static int end_bio_extent_readpage(struct bio *bio, - unsigned int bytes_done, int err) -#endif -{ - int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); - struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_map_tree *tree = bio->bi_private; - u64 start; - u64 end; - int whole_page; - int ret; - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - - do { - struct page *page = bvec->bv_page; - start = ((u64)page->index << PAGE_CACHE_SHIFT) + - bvec->bv_offset; - end = start + bvec->bv_len - 1; - - if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) - whole_page = 1; - else - whole_page = 0; - - if (--bvec >= bio->bi_io_vec) - prefetchw(&bvec->bv_page->flags); - - if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { - ret = tree->ops->readpage_end_io_hook(page, start, end); - if (ret) - uptodate = 0; - } - if (uptodate) { - set_extent_uptodate(tree, start, end, GFP_ATOMIC); - if (whole_page) - SetPageUptodate(page); - else - check_page_uptodate(tree, page); - } else { - ClearPageUptodate(page); - SetPageError(page); - } - - unlock_extent(tree, start, end, GFP_ATOMIC); - - if (whole_page) - unlock_page(page); - else - check_page_locked(tree, page); - } while (bvec >= bio->bi_io_vec); - - bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif -} - -/* - * IO done from prepare_write is pretty simple, we just unlock - * the structs in the extent tree when done, and set the uptodate bits - * as appropriate. - */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) -static void end_bio_extent_preparewrite(struct bio *bio, int err) -#else -static int end_bio_extent_preparewrite(struct bio *bio, - unsigned int bytes_done, int err) -#endif -{ - const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); - struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_map_tree *tree = bio->bi_private; - u64 start; - u64 end; - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - - do { - struct page *page = bvec->bv_page; - start = ((u64)page->index << PAGE_CACHE_SHIFT) + - bvec->bv_offset; - end = start + bvec->bv_len - 1; - - if (--bvec >= bio->bi_io_vec) - prefetchw(&bvec->bv_page->flags); - - if (uptodate) { - set_extent_uptodate(tree, start, end, GFP_ATOMIC); - } else { - ClearPageUptodate(page); - SetPageError(page); - } - - unlock_extent(tree, start, end, GFP_ATOMIC); - - } while (bvec >= bio->bi_io_vec); - - bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif -} - -static struct bio * -extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, - gfp_t gfp_flags) -{ - struct bio *bio; - - bio = bio_alloc(gfp_flags, nr_vecs); - - if (bio == NULL && (current->flags & PF_MEMALLOC)) { - while (!bio && (nr_vecs /= 2)) - bio = bio_alloc(gfp_flags, nr_vecs); - } - - if (bio) { - bio->bi_bdev = bdev; - bio->bi_sector = first_sector; - } - return bio; -} - -static int submit_one_bio(int rw, struct bio *bio) -{ - u64 maxsector; - int ret = 0; - - bio_get(bio); - - maxsector = bio->bi_bdev->bd_inode->i_size >> 9; - if (maxsector < bio->bi_sector) { - printk("sector too large max %Lu got %llu\n", maxsector, - (unsigned long long)bio->bi_sector); - WARN_ON(1); - } - - submit_bio(rw, bio); - if (bio_flagged(bio, BIO_EOPNOTSUPP)) - ret = -EOPNOTSUPP; - bio_put(bio); - return ret; -} - -static int submit_extent_page(int rw, struct extent_map_tree *tree, - struct page *page, sector_t sector, - size_t size, unsigned long offset, - struct block_device *bdev, - struct bio **bio_ret, - unsigned long max_pages, - bio_end_io_t end_io_func) -{ - int ret = 0; - struct bio *bio; - int nr; - - if (bio_ret && *bio_ret) { - bio = *bio_ret; - if (bio->bi_sector + (bio->bi_size >> 9) != sector || - bio_add_page(bio, page, size, offset) < size) { - ret = submit_one_bio(rw, bio); - bio = NULL; - } else { - return 0; - } - } - nr = min_t(int, max_pages, bio_get_nr_vecs(bdev)); - bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); - if (!bio) { - printk("failed to allocate bio nr %d\n", nr); - } - bio_add_page(bio, page, size, offset); - bio->bi_end_io = end_io_func; - bio->bi_private = tree; - if (bio_ret) { - *bio_ret = bio; - } else { - ret = submit_one_bio(rw, bio); - } - - return ret; -} - -void set_page_extent_mapped(struct page *page) -{ - if (!PagePrivate(page)) { - SetPagePrivate(page); - WARN_ON(!page->mapping->a_ops->invalidatepage); - set_page_private(page, EXTENT_PAGE_PRIVATE); - page_cache_get(page); - } -} - -void set_page_extent_head(struct page *page, unsigned long len) -{ - set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); -} - -/* - * basic readpage implementation. Locked extent state structs are inserted - * into the tree that are removed when the IO is done (by the end_io - * handlers) - */ -static int __extent_read_full_page(struct extent_map_tree *tree, - struct page *page, - get_extent_t *get_extent, - struct bio **bio) -{ - struct inode *inode = page->mapping->host; - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 page_end = start + PAGE_CACHE_SIZE - 1; - u64 end; - u64 cur = start; - u64 extent_offset; - u64 last_byte = i_size_read(inode); - u64 block_start; - u64 cur_end; - sector_t sector; - struct extent_map *em; - struct block_device *bdev; - int ret; - int nr = 0; - size_t page_offset = 0; - size_t iosize; - size_t blocksize = inode->i_sb->s_blocksize; - - set_page_extent_mapped(page); - - end = page_end; - lock_extent(tree, start, end, GFP_NOFS); - - while (cur <= end) { - if (cur >= last_byte) { - char *userpage; - iosize = PAGE_CACHE_SIZE - page_offset; - userpage = kmap_atomic(page, KM_USER0); - memset(userpage + page_offset, 0, iosize); - flush_dcache_page(page); - kunmap_atomic(userpage, KM_USER0); - set_extent_uptodate(tree, cur, cur + iosize - 1, - GFP_NOFS); - unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); - break; - } - em = get_extent(inode, page, page_offset, cur, end, 0); - if (IS_ERR(em) || !em) { - SetPageError(page); - unlock_extent(tree, cur, end, GFP_NOFS); - break; - } - - extent_offset = cur - em->start; - BUG_ON(em->end < cur); - BUG_ON(end < cur); - - iosize = min(em->end - cur, end - cur) + 1; - cur_end = min(em->end, end); - iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); - sector = (em->block_start + extent_offset) >> 9; - bdev = em->bdev; - block_start = em->block_start; - free_extent_map(em); - em = NULL; - - /* we've found a hole, just zero and go on */ - if (block_start == EXTENT_MAP_HOLE) { - char *userpage; - userpage = kmap_atomic(page, KM_USER0); - memset(userpage + page_offset, 0, iosize); - flush_dcache_page(page); - kunmap_atomic(userpage, KM_USER0); - - set_extent_uptodate(tree, cur, cur + iosize - 1, - GFP_NOFS); - unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); - cur = cur + iosize; - page_offset += iosize; - continue; - } - /* the get_extent function already copied into the page */ - if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) { - unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); - cur = cur + iosize; - page_offset += iosize; - continue; - } - - ret = 0; - if (tree->ops && tree->ops->readpage_io_hook) { - ret = tree->ops->readpage_io_hook(page, cur, - cur + iosize - 1); - } - if (!ret) { - unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; - nr -= page->index; - ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, - bdev, bio, nr, - end_bio_extent_readpage); - } - if (ret) - SetPageError(page); - cur = cur + iosize; - page_offset += iosize; - nr++; - } - if (!nr) { - if (!PageError(page)) - SetPageUptodate(page); - unlock_page(page); - } - return 0; -} - -int extent_read_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent) -{ - struct bio *bio = NULL; - int ret; - - ret = __extent_read_full_page(tree, page, get_extent, &bio); - if (bio) - submit_one_bio(READ, bio); - return ret; -} -EXPORT_SYMBOL(extent_read_full_page); - -/* - * the writepage semantics are similar to regular writepage. extent - * records are inserted to lock ranges in the tree, and as dirty areas - * are found, they are marked writeback. Then the lock bits are removed - * and the end_io handler clears the writeback ranges - */ -static int __extent_writepage(struct page *page, struct writeback_control *wbc, - void *data) -{ - struct inode *inode = page->mapping->host; - struct extent_page_data *epd = data; - struct extent_map_tree *tree = epd->tree; - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 delalloc_start; - u64 page_end = start + PAGE_CACHE_SIZE - 1; - u64 end; - u64 cur = start; - u64 extent_offset; - u64 last_byte = i_size_read(inode); - u64 block_start; - u64 iosize; - sector_t sector; - struct extent_map *em; - struct block_device *bdev; - int ret; - int nr = 0; - size_t page_offset = 0; - size_t blocksize; - loff_t i_size = i_size_read(inode); - unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; - u64 nr_delalloc; - u64 delalloc_end; - - WARN_ON(!PageLocked(page)); - if (page->index > end_index) { - clear_extent_dirty(tree, start, page_end, GFP_NOFS); - unlock_page(page); - return 0; - } - - if (page->index == end_index) { - char *userpage; - - size_t offset = i_size & (PAGE_CACHE_SIZE - 1); - - userpage = kmap_atomic(page, KM_USER0); - memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); - flush_dcache_page(page); - kunmap_atomic(userpage, KM_USER0); - } - - set_page_extent_mapped(page); - - delalloc_start = start; - delalloc_end = 0; - while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, - &delalloc_end, - 128 * 1024 * 1024); - if (nr_delalloc == 0) { - delalloc_start = delalloc_end + 1; - continue; - } - tree->ops->fill_delalloc(inode, delalloc_start, - delalloc_end); - clear_extent_bit(tree, delalloc_start, - delalloc_end, - EXTENT_LOCKED | EXTENT_DELALLOC, - 1, 0, GFP_NOFS); - delalloc_start = delalloc_end + 1; - } - lock_extent(tree, start, page_end, GFP_NOFS); - - end = page_end; - if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { - printk("found delalloc bits after lock_extent\n"); - } - - if (last_byte <= start) { - clear_extent_dirty(tree, start, page_end, GFP_NOFS); - goto done; - } - - set_extent_uptodate(tree, start, page_end, GFP_NOFS); - blocksize = inode->i_sb->s_blocksize; - - while (cur <= end) { - if (cur >= last_byte) { - clear_extent_dirty(tree, cur, page_end, GFP_NOFS); - break; - } - em = epd->get_extent(inode, page, page_offset, cur, end, 1); - if (IS_ERR(em) || !em) { - SetPageError(page); - break; - } - - extent_offset = cur - em->start; - BUG_ON(em->end < cur); - BUG_ON(end < cur); - iosize = min(em->end - cur, end - cur) + 1; - iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); - sector = (em->block_start + extent_offset) >> 9; - bdev = em->bdev; - block_start = em->block_start; - free_extent_map(em); - em = NULL; - - if (block_start == EXTENT_MAP_HOLE || - block_start == EXTENT_MAP_INLINE) { - clear_extent_dirty(tree, cur, - cur + iosize - 1, GFP_NOFS); - cur = cur + iosize; - page_offset += iosize; - continue; - } - - /* leave this out until we have a page_mkwrite call */ - if (0 && !test_range_bit(tree, cur, cur + iosize - 1, - EXTENT_DIRTY, 0)) { - cur = cur + iosize; - page_offset += iosize; - continue; - } - clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); - if (tree->ops && tree->ops->writepage_io_hook) { - ret = tree->ops->writepage_io_hook(page, cur, - cur + iosize - 1); - } else { - ret = 0; - } - if (ret) - SetPageError(page); - else { - unsigned long max_nr = end_index + 1; - set_range_writeback(tree, cur, cur + iosize - 1); - if (!PageWriteback(page)) { - printk("warning page %lu not writeback, " - "cur %llu end %llu\n", page->index, - (unsigned long long)cur, - (unsigned long long)end); - } - - ret = submit_extent_page(WRITE, tree, page, sector, - iosize, page_offset, bdev, - &epd->bio, max_nr, - end_bio_extent_writepage); - if (ret) - SetPageError(page); - } - cur = cur + iosize; - page_offset += iosize; - nr++; - } -done: - if (nr == 0) { - /* make sure the mapping tag for page dirty gets cleared */ - set_page_writeback(page); - end_page_writeback(page); - } - unlock_extent(tree, start, page_end, GFP_NOFS); - unlock_page(page); - return 0; -} - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - -/* Taken directly from 2.6.23 for 2.6.18 back port */ -typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, - void *data); - -/** - * write_cache_pages - walk the list of dirty pages of the given address space - * and write all of them. - * @mapping: address space structure to write - * @wbc: subtract the number of written pages from *@wbc->nr_to_write - * @writepage: function called for each page - * @data: data passed to writepage function - * - * If a page is already under I/O, write_cache_pages() skips it, even - * if it's dirty. This is desirable behaviour for memory-cleaning writeback, - * but it is INCORRECT for data-integrity system calls such as fsync(). fsync() - * and msync() need to guarantee that all the data which was dirty at the time - * the call was made get new I/O started against them. If wbc->sync_mode is - * WB_SYNC_ALL then we were called for data integrity and we must wait for - * existing IO to complete. - */ -static int write_cache_pages(struct address_space *mapping, - struct writeback_control *wbc, writepage_t writepage, - void *data) -{ - struct backing_dev_info *bdi = mapping->backing_dev_info; - int ret = 0; - int done = 0; - struct pagevec pvec; - int nr_pages; - pgoff_t index; - pgoff_t end; /* Inclusive */ - int scanned = 0; - int range_whole = 0; - - if (wbc->nonblocking && bdi_write_congested(bdi)) { - wbc->encountered_congestion = 1; - return 0; - } - - pagevec_init(&pvec, 0); - if (wbc->range_cyclic) { - index = mapping->writeback_index; /* Start from prev offset */ - end = -1; - } else { - index = wbc->range_start >> PAGE_CACHE_SHIFT; - end = wbc->range_end >> PAGE_CACHE_SHIFT; - if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) - range_whole = 1; - scanned = 1; - } -retry: - while (!done && (index <= end) && - (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, - PAGECACHE_TAG_DIRTY, - min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { - unsigned i; - - scanned = 1; - for (i = 0; i < nr_pages; i++) { - struct page *page = pvec.pages[i]; - - /* - * At this point we hold neither mapping->tree_lock nor - * lock on the page itself: the page may be truncated or - * invalidated (changing page->mapping to NULL), or even - * swizzled back from swapper_space to tmpfs file - * mapping - */ - lock_page(page); - - if (unlikely(page->mapping != mapping)) { - unlock_page(page); - continue; - } - - if (!wbc->range_cyclic && page->index > end) { - done = 1; - unlock_page(page); - continue; - } - - if (wbc->sync_mode != WB_SYNC_NONE) - wait_on_page_writeback(page); - - if (PageWriteback(page) || - !clear_page_dirty_for_io(page)) { - unlock_page(page); - continue; - } - - ret = (*writepage)(page, wbc, data); - - if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) { - unlock_page(page); - ret = 0; - } - if (ret || (--(wbc->nr_to_write) <= 0)) - done = 1; - if (wbc->nonblocking && bdi_write_congested(bdi)) { - wbc->encountered_congestion = 1; - done = 1; - } - } - pagevec_release(&pvec); - cond_resched(); - } - if (!scanned && !done) { - /* - * We hit the last page and there is more work to be done: wrap - * back to the start of the file - */ - scanned = 1; - index = 0; - goto retry; - } - if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) - mapping->writeback_index = index; - return ret; -} -#endif - -int extent_write_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent, - struct writeback_control *wbc) -{ - int ret; - struct address_space *mapping = page->mapping; - struct extent_page_data epd = { - .bio = NULL, - .tree = tree, - .get_extent = get_extent, - }; - struct writeback_control wbc_writepages = { - .bdi = wbc->bdi, - .sync_mode = WB_SYNC_NONE, - .older_than_this = NULL, - .nr_to_write = 64, - .range_start = page_offset(page) + PAGE_CACHE_SIZE, - .range_end = (loff_t)-1, - }; - - - ret = __extent_writepage(page, wbc, &epd); - - write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); - if (epd.bio) { - submit_one_bio(WRITE, epd.bio); - } - return ret; -} -EXPORT_SYMBOL(extent_write_full_page); - - -int extent_writepages(struct extent_map_tree *tree, - struct address_space *mapping, - get_extent_t *get_extent, - struct writeback_control *wbc) -{ - int ret = 0; - struct extent_page_data epd = { - .bio = NULL, - .tree = tree, - .get_extent = get_extent, - }; - - ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); - if (epd.bio) { - submit_one_bio(WRITE, epd.bio); - } - return ret; -} -EXPORT_SYMBOL(extent_writepages); - -int extent_readpages(struct extent_map_tree *tree, - struct address_space *mapping, - struct list_head *pages, unsigned nr_pages, - get_extent_t get_extent) -{ - struct bio *bio = NULL; - unsigned page_idx; - struct pagevec pvec; - - pagevec_init(&pvec, 0); - for (page_idx = 0; page_idx < nr_pages; page_idx++) { - struct page *page = list_entry(pages->prev, struct page, lru); - - prefetchw(&page->flags); - list_del(&page->lru); - /* - * what we want to do here is call add_to_page_cache_lru, - * but that isn't exported, so we reproduce it here - */ - if (!add_to_page_cache(page, mapping, - page->index, GFP_KERNEL)) { - - /* open coding of lru_cache_add, also not exported */ - page_cache_get(page); - if (!pagevec_add(&pvec, page)) - __pagevec_lru_add(&pvec); - __extent_read_full_page(tree, page, get_extent, &bio); - } - page_cache_release(page); - } - if (pagevec_count(&pvec)) - __pagevec_lru_add(&pvec); - BUG_ON(!list_empty(pages)); - if (bio) - submit_one_bio(READ, bio); - return 0; -} -EXPORT_SYMBOL(extent_readpages); - -/* - * basic invalidatepage code, this waits on any locked or writeback - * ranges corresponding to the page, and then deletes any extent state - * records from the tree - */ -int extent_invalidatepage(struct extent_map_tree *tree, - struct page *page, unsigned long offset) -{ - u64 start = ((u64)page->index << PAGE_CACHE_SHIFT); - u64 end = start + PAGE_CACHE_SIZE - 1; - size_t blocksize = page->mapping->host->i_sb->s_blocksize; - - start += (offset + blocksize -1) & ~(blocksize - 1); - if (start > end) - return 0; - - lock_extent(tree, start, end, GFP_NOFS); - wait_on_extent_writeback(tree, start, end); - clear_extent_bit(tree, start, end, - EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, - 1, 1, GFP_NOFS); - return 0; -} -EXPORT_SYMBOL(extent_invalidatepage); - -/* - * simple commit_write call, set_range_dirty is used to mark both - * the pages and the extent records as dirty - */ -int extent_commit_write(struct extent_map_tree *tree, - struct inode *inode, struct page *page, - unsigned from, unsigned to) -{ - loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; - - set_page_extent_mapped(page); - set_page_dirty(page); - - if (pos > inode->i_size) { - i_size_write(inode, pos); - mark_inode_dirty(inode); - } - return 0; -} -EXPORT_SYMBOL(extent_commit_write); - -int extent_prepare_write(struct extent_map_tree *tree, - struct inode *inode, struct page *page, - unsigned from, unsigned to, get_extent_t *get_extent) -{ - u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 page_end = page_start + PAGE_CACHE_SIZE - 1; - u64 block_start; - u64 orig_block_start; - u64 block_end; - u64 cur_end; - struct extent_map *em; - unsigned blocksize = 1 << inode->i_blkbits; - size_t page_offset = 0; - size_t block_off_start; - size_t block_off_end; - int err = 0; - int iocount = 0; - int ret = 0; - int isnew; - - set_page_extent_mapped(page); - - block_start = (page_start + from) & ~((u64)blocksize - 1); - block_end = (page_start + to - 1) | (blocksize - 1); - orig_block_start = block_start; - - lock_extent(tree, page_start, page_end, GFP_NOFS); - while(block_start <= block_end) { - em = get_extent(inode, page, page_offset, block_start, - block_end, 1); - if (IS_ERR(em) || !em) { - goto err; - } - cur_end = min(block_end, em->end); - block_off_start = block_start & (PAGE_CACHE_SIZE - 1); - block_off_end = block_off_start + blocksize; - isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS); - - if (!PageUptodate(page) && isnew && - (block_off_end > to || block_off_start < from)) { - void *kaddr; - - kaddr = kmap_atomic(page, KM_USER0); - if (block_off_end > to) - memset(kaddr + to, 0, block_off_end - to); - if (block_off_start < from) - memset(kaddr + block_off_start, 0, - from - block_off_start); - flush_dcache_page(page); - kunmap_atomic(kaddr, KM_USER0); - } - if ((em->block_start != EXTENT_MAP_HOLE && - em->block_start != EXTENT_MAP_INLINE) && - !isnew && !PageUptodate(page) && - (block_off_end > to || block_off_start < from) && - !test_range_bit(tree, block_start, cur_end, - EXTENT_UPTODATE, 1)) { - u64 sector; - u64 extent_offset = block_start - em->start; - size_t iosize; - sector = (em->block_start + extent_offset) >> 9; - iosize = (cur_end - block_start + blocksize) & - ~((u64)blocksize - 1); - /* - * we've already got the extent locked, but we - * need to split the state such that our end_bio - * handler can clear the lock. - */ - set_extent_bit(tree, block_start, - block_start + iosize - 1, - EXTENT_LOCKED, 0, NULL, GFP_NOFS); - ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, em->bdev, - NULL, 1, - end_bio_extent_preparewrite); - iocount++; - block_start = block_start + iosize; - } else { - set_extent_uptodate(tree, block_start, cur_end, - GFP_NOFS); - unlock_extent(tree, block_start, cur_end, GFP_NOFS); - block_start = cur_end + 1; - } - page_offset = block_start & (PAGE_CACHE_SIZE - 1); - free_extent_map(em); - } - if (iocount) { - wait_extent_bit(tree, orig_block_start, - block_end, EXTENT_LOCKED); - } - check_page_uptodate(tree, page); -err: - /* FIXME, zero out newly allocated blocks on error */ - return err; -} -EXPORT_SYMBOL(extent_prepare_write); - -/* - * a helper for releasepage. As long as there are no locked extents - * in the range corresponding to the page, both state records and extent - * map records are removed - */ -int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page) -{ - struct extent_map *em; - u64 start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 end = start + PAGE_CACHE_SIZE - 1; - u64 orig_start = start; - int ret = 1; - - while (start <= end) { - em = lookup_extent_mapping(tree, start, end); - if (!em || IS_ERR(em)) - break; - if (!test_range_bit(tree, em->start, em->end, - EXTENT_LOCKED, 0)) { - remove_extent_mapping(tree, em); - /* once for the rb tree */ - free_extent_map(em); - } - start = em->end + 1; - /* once for us */ - free_extent_map(em); - } - if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) - ret = 0; - else - clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, - 1, 1, GFP_NOFS); - return ret; -} -EXPORT_SYMBOL(try_release_extent_mapping); - -sector_t extent_bmap(struct address_space *mapping, sector_t iblock, - get_extent_t *get_extent) -{ - struct inode *inode = mapping->host; - u64 start = iblock << inode->i_blkbits; - u64 end = start + (1 << inode->i_blkbits) - 1; - sector_t sector = 0; - struct extent_map *em; - - em = get_extent(inode, NULL, 0, start, end, 0); - if (!em || IS_ERR(em)) - return 0; - - if (em->block_start == EXTENT_MAP_INLINE || - em->block_start == EXTENT_MAP_HOLE) - goto out; - - sector = (em->block_start + start - em->start) >> inode->i_blkbits; -out: - free_extent_map(em); - return sector; -} - -static int add_lru(struct extent_map_tree *tree, struct extent_buffer *eb) -{ - if (list_empty(&eb->lru)) { - extent_buffer_get(eb); - list_add(&eb->lru, &tree->buffer_lru); - tree->lru_size++; - if (tree->lru_size >= BUFFER_LRU_MAX) { - struct extent_buffer *rm; - rm = list_entry(tree->buffer_lru.prev, - struct extent_buffer, lru); - tree->lru_size--; - list_del_init(&rm->lru); - free_extent_buffer(rm); - } - } else - list_move(&eb->lru, &tree->buffer_lru); - return 0; -} -static struct extent_buffer *find_lru(struct extent_map_tree *tree, - u64 start, unsigned long len) -{ - struct list_head *lru = &tree->buffer_lru; - struct list_head *cur = lru->next; - struct extent_buffer *eb; - - if (list_empty(lru)) - return NULL; - - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start == start && eb->len == len) { - extent_buffer_get(eb); - return eb; - } - cur = cur->next; - } while (cur != lru); - return NULL; -} - -static inline unsigned long num_extent_pages(u64 start, u64 len) -{ - return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - - (start >> PAGE_CACHE_SHIFT); -} - -static inline struct page *extent_buffer_page(struct extent_buffer *eb, - unsigned long i) -{ - struct page *p; - struct address_space *mapping; - - if (i == 0) - return eb->first_page; - i += eb->start >> PAGE_CACHE_SHIFT; - mapping = eb->first_page->mapping; - read_lock_irq(&mapping->tree_lock); - p = radix_tree_lookup(&mapping->page_tree, i); - read_unlock_irq(&mapping->tree_lock); - return p; -} - -static struct extent_buffer *__alloc_extent_buffer(struct extent_map_tree *tree, - u64 start, - unsigned long len, - gfp_t mask) -{ - struct extent_buffer *eb = NULL; - - spin_lock(&tree->lru_lock); - eb = find_lru(tree, start, len); - spin_unlock(&tree->lru_lock); - if (eb) { - return eb; - } - - eb = kmem_cache_zalloc(extent_buffer_cache, mask); - INIT_LIST_HEAD(&eb->lru); - eb->start = start; - eb->len = len; - atomic_set(&eb->refs, 1); - - return eb; -} - -static void __free_extent_buffer(struct extent_buffer *eb) -{ - kmem_cache_free(extent_buffer_cache, eb); -} - -struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, - u64 start, unsigned long len, - struct page *page0, - gfp_t mask) -{ - unsigned long num_pages = num_extent_pages(start, len); - unsigned long i; - unsigned long index = start >> PAGE_CACHE_SHIFT; - struct extent_buffer *eb; - struct page *p; - struct address_space *mapping = tree->mapping; - int uptodate = 1; - - eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb || IS_ERR(eb)) - return NULL; - - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - - if (page0) { - eb->first_page = page0; - i = 1; - index++; - page_cache_get(page0); - mark_page_accessed(page0); - set_page_extent_mapped(page0); - WARN_ON(!PageUptodate(page0)); - set_page_extent_head(page0, len); - } else { - i = 0; - } - for (; i < num_pages; i++, index++) { - p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); - if (!p) { - WARN_ON(1); - goto fail; - } - set_page_extent_mapped(p); - mark_page_accessed(p); - if (i == 0) { - eb->first_page = p; - set_page_extent_head(p, len); - } else { - set_page_private(p, EXTENT_PAGE_PRIVATE); - } - if (!PageUptodate(p)) - uptodate = 0; - unlock_page(p); - } - if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; - -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); - return eb; - -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); - if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { - page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); - return NULL; -} -EXPORT_SYMBOL(alloc_extent_buffer); - -struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, - u64 start, unsigned long len, - gfp_t mask) -{ - unsigned long num_pages = num_extent_pages(start, len); - unsigned long i; - unsigned long index = start >> PAGE_CACHE_SHIFT; - struct extent_buffer *eb; - struct page *p; - struct address_space *mapping = tree->mapping; - int uptodate = 1; - - eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb || IS_ERR(eb)) - return NULL; - - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - - for (i = 0; i < num_pages; i++, index++) { - p = find_lock_page(mapping, index); - if (!p) { - goto fail; - } - set_page_extent_mapped(p); - mark_page_accessed(p); - - if (i == 0) { - eb->first_page = p; - set_page_extent_head(p, len); - } else { - set_page_private(p, EXTENT_PAGE_PRIVATE); - } - - if (!PageUptodate(p)) - uptodate = 0; - unlock_page(p); - } - if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; - -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); - return eb; -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); - if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { - page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); - return NULL; -} -EXPORT_SYMBOL(find_extent_buffer); - -void free_extent_buffer(struct extent_buffer *eb) -{ - unsigned long i; - unsigned long num_pages; - - if (!eb) - return; - - if (!atomic_dec_and_test(&eb->refs)) - return; - - WARN_ON(!list_empty(&eb->lru)); - num_pages = num_extent_pages(eb->start, eb->len); - - for (i = 1; i < num_pages; i++) { - page_cache_release(extent_buffer_page(eb, i)); - } - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); -} -EXPORT_SYMBOL(free_extent_buffer); - -int clear_extent_buffer_dirty(struct extent_map_tree *tree, - struct extent_buffer *eb) -{ - int set; - unsigned long i; - unsigned long num_pages; - struct page *page; - - u64 start = eb->start; - u64 end = start + eb->len - 1; - - set = clear_extent_dirty(tree, start, end, GFP_NOFS); - num_pages = num_extent_pages(eb->start, eb->len); - - for (i = 0; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - lock_page(page); - if (i == 0) - set_page_extent_head(page, eb->len); - else - set_page_private(page, EXTENT_PAGE_PRIVATE); - - /* - * if we're on the last page or the first page and the - * block isn't aligned on a page boundary, do extra checks - * to make sure we don't clean page that is partially dirty - */ - if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || - ((i == num_pages - 1) && - ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { - start = (u64)page->index << PAGE_CACHE_SHIFT; - end = start + PAGE_CACHE_SIZE - 1; - if (test_range_bit(tree, start, end, - EXTENT_DIRTY, 0)) { - unlock_page(page); - continue; - } - } - clear_page_dirty_for_io(page); - write_lock_irq(&page->mapping->tree_lock); - if (!PageDirty(page)) { - radix_tree_tag_clear(&page->mapping->page_tree, - page_index(page), - PAGECACHE_TAG_DIRTY); - } - write_unlock_irq(&page->mapping->tree_lock); - unlock_page(page); - } - return 0; -} -EXPORT_SYMBOL(clear_extent_buffer_dirty); - -int wait_on_extent_buffer_writeback(struct extent_map_tree *tree, - struct extent_buffer *eb) -{ - return wait_on_extent_writeback(tree, eb->start, - eb->start + eb->len - 1); -} -EXPORT_SYMBOL(wait_on_extent_buffer_writeback); - -int set_extent_buffer_dirty(struct extent_map_tree *tree, - struct extent_buffer *eb) -{ - unsigned long i; - unsigned long num_pages; - - num_pages = num_extent_pages(eb->start, eb->len); - for (i = 0; i < num_pages; i++) { - struct page *page = extent_buffer_page(eb, i); - /* writepage may need to do something special for the - * first page, we have to make sure page->private is - * properly set. releasepage may drop page->private - * on us if the page isn't already dirty. - */ - if (i == 0) { - lock_page(page); - set_page_extent_head(page, eb->len); - } else if (PagePrivate(page) && - page->private != EXTENT_PAGE_PRIVATE) { - lock_page(page); - set_page_extent_mapped(page); - unlock_page(page); - } - __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); - if (i == 0) - unlock_page(page); - } - return set_extent_dirty(tree, eb->start, - eb->start + eb->len - 1, GFP_NOFS); -} -EXPORT_SYMBOL(set_extent_buffer_dirty); - -int set_extent_buffer_uptodate(struct extent_map_tree *tree, - struct extent_buffer *eb) -{ - unsigned long i; - struct page *page; - unsigned long num_pages; - - num_pages = num_extent_pages(eb->start, eb->len); - - set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, - GFP_NOFS); - for (i = 0; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || - ((i == num_pages - 1) && - ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { - check_page_uptodate(tree, page); - continue; - } - SetPageUptodate(page); - } - return 0; -} -EXPORT_SYMBOL(set_extent_buffer_uptodate); - -int extent_buffer_uptodate(struct extent_map_tree *tree, - struct extent_buffer *eb) -{ - if (eb->flags & EXTENT_UPTODATE) - return 1; - return test_range_bit(tree, eb->start, eb->start + eb->len - 1, - EXTENT_UPTODATE, 1); -} -EXPORT_SYMBOL(extent_buffer_uptodate); - -int read_extent_buffer_pages(struct extent_map_tree *tree, - struct extent_buffer *eb, - u64 start, - int wait) -{ - unsigned long i; - unsigned long start_i; - struct page *page; - int err; - int ret = 0; - unsigned long num_pages; - - if (eb->flags & EXTENT_UPTODATE) - return 0; - - if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1, - EXTENT_UPTODATE, 1)) { - return 0; - } - - if (start) { - WARN_ON(start < eb->start); - start_i = (start >> PAGE_CACHE_SHIFT) - - (eb->start >> PAGE_CACHE_SHIFT); - } else { - start_i = 0; - } - - num_pages = num_extent_pages(eb->start, eb->len); - for (i = start_i; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - if (PageUptodate(page)) { - continue; - } - if (!wait) { - if (TestSetPageLocked(page)) { - continue; - } - } else { - lock_page(page); - } - if (!PageUptodate(page)) { - err = page->mapping->a_ops->readpage(NULL, page); - if (err) { - ret = err; - } - } else { - unlock_page(page); - } - } - - if (ret || !wait) { - return ret; - } - - for (i = start_i; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - wait_on_page_locked(page); - if (!PageUptodate(page)) { - ret = -EIO; - } - } - if (!ret) - eb->flags |= EXTENT_UPTODATE; - return ret; -} -EXPORT_SYMBOL(read_extent_buffer_pages); - -void read_extent_buffer(struct extent_buffer *eb, void *dstv, - unsigned long start, - unsigned long len) -{ - size_t cur; - size_t offset; - struct page *page; - char *kaddr; - char *dst = (char *)dstv; - size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - unsigned long num_pages = num_extent_pages(eb->start, eb->len); - - WARN_ON(start > eb->len); - WARN_ON(start + len > eb->start + eb->len); - - offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - - while(len > 0) { - page = extent_buffer_page(eb, i); - if (!PageUptodate(page)) { - printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len); - WARN_ON(1); - } - WARN_ON(!PageUptodate(page)); - - cur = min(len, (PAGE_CACHE_SIZE - offset)); - kaddr = kmap_atomic(page, KM_USER1); - memcpy(dst, kaddr + offset, cur); - kunmap_atomic(kaddr, KM_USER1); - - dst += cur; - len -= cur; - offset = 0; - i++; - } -} -EXPORT_SYMBOL(read_extent_buffer); - -int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, - unsigned long min_len, char **token, char **map, - unsigned long *map_start, - unsigned long *map_len, int km) -{ - size_t offset = start & (PAGE_CACHE_SIZE - 1); - char *kaddr; - struct page *p; - size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - unsigned long end_i = (start_offset + start + min_len - 1) >> - PAGE_CACHE_SHIFT; - - if (i != end_i) - return -EINVAL; - - if (i == 0) { - offset = start_offset; - *map_start = 0; - } else { - offset = 0; - *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; - } - if (start + min_len > eb->len) { -printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); - WARN_ON(1); - } - - p = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(p)); - kaddr = kmap_atomic(p, km); - *token = kaddr; - *map = kaddr + offset; - *map_len = PAGE_CACHE_SIZE - offset; - return 0; -} -EXPORT_SYMBOL(map_private_extent_buffer); - -int map_extent_buffer(struct extent_buffer *eb, unsigned long start, - unsigned long min_len, - char **token, char **map, - unsigned long *map_start, - unsigned long *map_len, int km) -{ - int err; - int save = 0; - if (eb->map_token) { - unmap_extent_buffer(eb, eb->map_token, km); - eb->map_token = NULL; - save = 1; - } - err = map_private_extent_buffer(eb, start, min_len, token, map, - map_start, map_len, km); - if (!err && save) { - eb->map_token = *token; - eb->kaddr = *map; - eb->map_start = *map_start; - eb->map_len = *map_len; - } - return err; -} -EXPORT_SYMBOL(map_extent_buffer); - -void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) -{ - kunmap_atomic(token, km); -} -EXPORT_SYMBOL(unmap_extent_buffer); - -int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, - unsigned long start, - unsigned long len) -{ - size_t cur; - size_t offset; - struct page *page; - char *kaddr; - char *ptr = (char *)ptrv; - size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - int ret = 0; - - WARN_ON(start > eb->len); - WARN_ON(start + len > eb->start + eb->len); - - offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - - while(len > 0) { - page = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(page)); - - cur = min(len, (PAGE_CACHE_SIZE - offset)); - - kaddr = kmap_atomic(page, KM_USER0); - ret = memcmp(ptr, kaddr + offset, cur); - kunmap_atomic(kaddr, KM_USER0); - if (ret) - break; - - ptr += cur; - len -= cur; - offset = 0; - i++; - } - return ret; -} -EXPORT_SYMBOL(memcmp_extent_buffer); - -void write_extent_buffer(struct extent_buffer *eb, const void *srcv, - unsigned long start, unsigned long len) -{ - size_t cur; - size_t offset; - struct page *page; - char *kaddr; - char *src = (char *)srcv; - size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - - WARN_ON(start > eb->len); - WARN_ON(start + len > eb->start + eb->len); - - offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - - while(len > 0) { - page = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(page)); - - cur = min(len, PAGE_CACHE_SIZE - offset); - kaddr = kmap_atomic(page, KM_USER1); - memcpy(kaddr + offset, src, cur); - kunmap_atomic(kaddr, KM_USER1); - - src += cur; - len -= cur; - offset = 0; - i++; - } -} -EXPORT_SYMBOL(write_extent_buffer); - -void memset_extent_buffer(struct extent_buffer *eb, char c, - unsigned long start, unsigned long len) -{ - size_t cur; - size_t offset; - struct page *page; - char *kaddr; - size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; - - WARN_ON(start > eb->len); - WARN_ON(start + len > eb->start + eb->len); - - offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - - while(len > 0) { - page = extent_buffer_page(eb, i); - WARN_ON(!PageUptodate(page)); - - cur = min(len, PAGE_CACHE_SIZE - offset); - kaddr = kmap_atomic(page, KM_USER0); - memset(kaddr + offset, c, cur); - kunmap_atomic(kaddr, KM_USER0); - - len -= cur; - offset = 0; - i++; - } -} -EXPORT_SYMBOL(memset_extent_buffer); - -void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, - unsigned long dst_offset, unsigned long src_offset, - unsigned long len) -{ - u64 dst_len = dst->len; - size_t cur; - size_t offset; - struct page *page; - char *kaddr; - size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; - - WARN_ON(src->len != dst_len); - - offset = (start_offset + dst_offset) & - ((unsigned long)PAGE_CACHE_SIZE - 1); - - while(len > 0) { - page = extent_buffer_page(dst, i); - WARN_ON(!PageUptodate(page)); - - cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset)); - - kaddr = kmap_atomic(page, KM_USER0); - read_extent_buffer(src, kaddr + offset, src_offset, cur); - kunmap_atomic(kaddr, KM_USER0); - - src_offset += cur; - len -= cur; - offset = 0; - i++; - } -} -EXPORT_SYMBOL(copy_extent_buffer); - -static void move_pages(struct page *dst_page, struct page *src_page, - unsigned long dst_off, unsigned long src_off, - unsigned long len) -{ - char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); - if (dst_page == src_page) { - memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len); - } else { - char *src_kaddr = kmap_atomic(src_page, KM_USER1); - char *p = dst_kaddr + dst_off + len; - char *s = src_kaddr + src_off + len; - - while (len--) - *--p = *--s; - - kunmap_atomic(src_kaddr, KM_USER1); - } - kunmap_atomic(dst_kaddr, KM_USER0); -} - -static void copy_pages(struct page *dst_page, struct page *src_page, - unsigned long dst_off, unsigned long src_off, - unsigned long len) -{ - char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); - char *src_kaddr; - - if (dst_page != src_page) - src_kaddr = kmap_atomic(src_page, KM_USER1); - else - src_kaddr = dst_kaddr; - - memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len); - kunmap_atomic(dst_kaddr, KM_USER0); - if (dst_page != src_page) - kunmap_atomic(src_kaddr, KM_USER1); -} - -void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, - unsigned long src_offset, unsigned long len) -{ - size_t cur; - size_t dst_off_in_page; - size_t src_off_in_page; - size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long dst_i; - unsigned long src_i; - - if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); - BUG_ON(1); - } - if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); - BUG_ON(1); - } - - while(len > 0) { - dst_off_in_page = (start_offset + dst_offset) & - ((unsigned long)PAGE_CACHE_SIZE - 1); - src_off_in_page = (start_offset + src_offset) & - ((unsigned long)PAGE_CACHE_SIZE - 1); - - dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; - src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT; - - cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - - src_off_in_page)); - cur = min_t(unsigned long, cur, - (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page)); - - copy_pages(extent_buffer_page(dst, dst_i), - extent_buffer_page(dst, src_i), - dst_off_in_page, src_off_in_page, cur); - - src_offset += cur; - dst_offset += cur; - len -= cur; - } -} -EXPORT_SYMBOL(memcpy_extent_buffer); - -void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, - unsigned long src_offset, unsigned long len) -{ - size_t cur; - size_t dst_off_in_page; - size_t src_off_in_page; - unsigned long dst_end = dst_offset + len - 1; - unsigned long src_end = src_offset + len - 1; - size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); - unsigned long dst_i; - unsigned long src_i; - - if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); - BUG_ON(1); - } - if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); - BUG_ON(1); - } - if (dst_offset < src_offset) { - memcpy_extent_buffer(dst, dst_offset, src_offset, len); - return; - } - while(len > 0) { - dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; - src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; - - dst_off_in_page = (start_offset + dst_end) & - ((unsigned long)PAGE_CACHE_SIZE - 1); - src_off_in_page = (start_offset + src_end) & - ((unsigned long)PAGE_CACHE_SIZE - 1); - - cur = min_t(unsigned long, len, src_off_in_page + 1); - cur = min(cur, dst_off_in_page + 1); - move_pages(extent_buffer_page(dst, dst_i), - extent_buffer_page(dst, src_i), - dst_off_in_page - cur + 1, - src_off_in_page - cur + 1, cur); - - dst_end -= cur; - src_end -= cur; - len -= cur; - } -} -EXPORT_SYMBOL(memmove_extent_buffer); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index ea60f5447b5..56314217cfc 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -3,215 +3,53 @@ #include +#define EXTENT_MAP_LAST_BYTE (u64)-4 #define EXTENT_MAP_HOLE (u64)-3 #define EXTENT_MAP_INLINE (u64)-2 #define EXTENT_MAP_DELALLOC (u64)-1 -/* bits for the extent state */ -#define EXTENT_DIRTY 1 -#define EXTENT_WRITEBACK (1 << 1) -#define EXTENT_UPTODATE (1 << 2) -#define EXTENT_LOCKED (1 << 3) -#define EXTENT_NEW (1 << 4) -#define EXTENT_DELALLOC (1 << 5) -#define EXTENT_DEFRAG (1 << 6) -#define EXTENT_DEFRAG_DONE (1 << 7) -#define EXTENT_BUFFER_FILLED (1 << 8) -#define EXTENT_CSUM (1 << 9) -#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) - -/* - * page->private values. Every page that is controlled by the extent - * map has page->private set to one. - */ -#define EXTENT_PAGE_PRIVATE 1 -#define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3 - - -struct extent_map_ops { - int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); - int (*writepage_io_hook)(struct page *page, u64 start, u64 end); - int (*readpage_io_hook)(struct page *page, u64 start, u64 end); - int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end); - void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end); -}; - -struct extent_map_tree { - struct rb_root map; - struct rb_root state; - struct address_space *mapping; - u64 dirty_bytes; - rwlock_t lock; - struct extent_map_ops *ops; - spinlock_t lru_lock; - struct list_head buffer_lru; - int lru_size; -}; - -/* note, this must start with the same fields as fs/extent_map.c:tree_entry */ struct extent_map { - u64 start; - u64 end; /* inclusive */ - int in_tree; struct rb_node rb_node; - /* block_start and block_end are in bytes */ + + /* all of these are in bytes */ + u64 start; + u64 len; u64 block_start; - u64 block_end; /* inclusive */ + unsigned long flags; struct block_device *bdev; atomic_t refs; -}; - -/* note, this must start with the same fields as fs/extent_map.c:tree_entry */ -struct extent_state { - u64 start; - u64 end; /* inclusive */ int in_tree; - struct rb_node rb_node; - wait_queue_head_t wq; - atomic_t refs; - unsigned long state; - - /* for use by the FS */ - u64 private; - - struct list_head list; }; -struct extent_buffer { - u64 start; - unsigned long len; - char *map_token; - char *kaddr; - unsigned long map_start; - unsigned long map_len; - struct page *first_page; - struct list_head lru; - atomic_t refs; - int flags; +struct extent_map_tree { + struct rb_root map; + struct extent_map *last; + spinlock_t lock; }; -typedef struct extent_map *(get_extent_t)(struct inode *inode, - struct page *page, - size_t page_offset, - u64 start, u64 end, - int create); +static inline u64 extent_map_end(struct extent_map *em) +{ + if (em->start + em->len < em->start) + return (u64)-1; + return em->start + em->len; +} + +static inline u64 extent_map_block_end(struct extent_map *em) +{ + if (em->block_start + em->len < em->block_start) + return (u64)-1; + return em->block_start + em->len; +} -void extent_map_tree_init(struct extent_map_tree *tree, - struct address_space *mapping, gfp_t mask); -void extent_map_tree_empty_lru(struct extent_map_tree *tree); +void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask); struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, - u64 start, u64 end); + u64 start, u64 len); int add_extent_mapping(struct extent_map_tree *tree, struct extent_map *em); int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em); -int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page); -int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); -int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask); + struct extent_map *alloc_extent_map(gfp_t mask); void free_extent_map(struct extent_map *em); -int extent_read_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent); int __init extent_map_init(void); void extent_map_exit(void); - -u64 count_range_bits(struct extent_map_tree *tree, - u64 *start, u64 search_end, - u64 max_bytes, unsigned long bits); - -int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end, - int bits, int filled); -int clear_extent_bits(struct extent_map_tree *tree, u64 start, u64 end, - int bits, gfp_t mask); -int set_extent_bits(struct extent_map_tree *tree, u64 start, u64 end, - int bits, gfp_t mask); -int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask); -int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask); -int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask); -int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask); -int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end, - gfp_t mask); -int find_first_extent_bit(struct extent_map_tree *tree, u64 start, - u64 *start_ret, u64 *end_ret, int bits); -int extent_invalidatepage(struct extent_map_tree *tree, - struct page *page, unsigned long offset); -int extent_write_full_page(struct extent_map_tree *tree, struct page *page, - get_extent_t *get_extent, - struct writeback_control *wbc); -int extent_writepages(struct extent_map_tree *tree, - struct address_space *mapping, - get_extent_t *get_extent, - struct writeback_control *wbc); -int extent_readpages(struct extent_map_tree *tree, - struct address_space *mapping, - struct list_head *pages, unsigned nr_pages, - get_extent_t get_extent); -int extent_prepare_write(struct extent_map_tree *tree, - struct inode *inode, struct page *page, - unsigned from, unsigned to, get_extent_t *get_extent); -int extent_commit_write(struct extent_map_tree *tree, - struct inode *inode, struct page *page, - unsigned from, unsigned to); -sector_t extent_bmap(struct address_space *mapping, sector_t iblock, - get_extent_t *get_extent); -int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end); -int set_state_private(struct extent_map_tree *tree, u64 start, u64 private); -int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private); -void set_page_extent_mapped(struct page *page); - -struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree, - u64 start, unsigned long len, - struct page *page0, - gfp_t mask); -struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree, - u64 start, unsigned long len, - gfp_t mask); -void free_extent_buffer(struct extent_buffer *eb); -int read_extent_buffer_pages(struct extent_map_tree *tree, - struct extent_buffer *eb, u64 start, int wait); - -static inline void extent_buffer_get(struct extent_buffer *eb) -{ - atomic_inc(&eb->refs); -} - -int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, - unsigned long start, - unsigned long len); -void read_extent_buffer(struct extent_buffer *eb, void *dst, - unsigned long start, - unsigned long len); -void write_extent_buffer(struct extent_buffer *eb, const void *src, - unsigned long start, unsigned long len); -void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, - unsigned long dst_offset, unsigned long src_offset, - unsigned long len); -void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, - unsigned long src_offset, unsigned long len); -void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, - unsigned long src_offset, unsigned long len); -void memset_extent_buffer(struct extent_buffer *eb, char c, - unsigned long start, unsigned long len); -int wait_on_extent_buffer_writeback(struct extent_map_tree *tree, - struct extent_buffer *eb); -int clear_extent_buffer_dirty(struct extent_map_tree *tree, - struct extent_buffer *eb); -int set_extent_buffer_dirty(struct extent_map_tree *tree, - struct extent_buffer *eb); -int set_extent_buffer_uptodate(struct extent_map_tree *tree, - struct extent_buffer *eb); -int extent_buffer_uptodate(struct extent_map_tree *tree, - struct extent_buffer *eb); -int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, - unsigned long min_len, char **token, char **map, - unsigned long *map_start, - unsigned long *map_len, int km); -int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset, - unsigned long min_len, char **token, char **map, - unsigned long *map_start, - unsigned long *map_len, int km); -void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1cd8c908811..c5bb00f9239 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -233,8 +233,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, int err = 0; int i; struct inode *inode = fdentry(file)->d_inode; - struct extent_map *em; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; u64 hint_byte; u64 num_bytes; u64 start_pos; @@ -242,11 +241,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 end_pos = pos + write_bytes; u64 inline_size; loff_t isize = i_size_read(inode); - em = alloc_extent_map(GFP_NOFS); - if (!em) - return -ENOMEM; - - em->bdev = inode->i_sb->s_bdev; start_pos = pos & ~((u64)root->sectorsize - 1); num_bytes = (write_bytes + pos - start_pos + @@ -254,7 +248,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, end_of_last_block = start_pos + num_bytes - 1; - lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); + lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); if (!trans) { @@ -268,7 +262,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, if ((end_of_last_block & 4095) == 0) { printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); } - set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS); + set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); /* FIXME...EIEIO, ENOSPC and more */ @@ -293,6 +287,8 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inode->i_ino, last_pos_in_file, 0, 0, hole_size); + btrfs_drop_extent_cache(inode, last_pos_in_file, + last_pos_in_file + hole_size -1); btrfs_check_file(root, inode); } if (err) @@ -320,12 +316,12 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, last_end += PAGE_CACHE_SIZE - 1; if (start_pos < isize) { u64 delalloc_start = start_pos; - existing_delalloc = count_range_bits(em_tree, + existing_delalloc = count_range_bits(io_tree, &delalloc_start, end_of_last_block, (u64)-1, EXTENT_DELALLOC); } - set_extent_delalloc(em_tree, start_pos, end_of_last_block, + set_extent_delalloc(io_tree, start_pos, end_of_last_block, GFP_NOFS); spin_lock(&root->fs_info->delalloc_lock); root->fs_info->delalloc_bytes += (end_of_last_block + 1 - @@ -346,6 +342,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inline_size -= start_pos; err = insert_inline_extent(trans, root, inode, start_pos, inline_size, pages, 0, num_pages); + btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1); BUG_ON(err); } if (end_pos > isize) { @@ -356,8 +353,7 @@ failed: err = btrfs_end_transaction(trans, root); out_unlock: mutex_unlock(&root->fs_info->fs_mutex); - unlock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS); - free_extent_map(em); + unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); return err; } @@ -367,10 +363,15 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; while(1) { + spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, end); - if (!em) + if (!em) { + spin_unlock(&em_tree->lock); break; + } remove_extent_mapping(em_tree, em); + spin_unlock(&em_tree->lock); + /* once for us */ free_extent_map(em); /* once for the tree*/ diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 67005480e13..16d3aef45d1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -53,7 +53,7 @@ static struct inode_operations btrfs_file_inode_operations; static struct address_space_operations btrfs_aops; static struct address_space_operations btrfs_symlink_aops; static struct file_operations btrfs_dir_file_operations; -static struct extent_map_ops btrfs_extent_map_ops; +static struct extent_io_ops btrfs_extent_io_ops; static struct kmem_cache *btrfs_inode_cachep; struct kmem_cache *btrfs_trans_handle_cachep; @@ -104,6 +104,8 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) u64 num_bytes; u64 cur_alloc_size; u64 blocksize = root->sectorsize; + u64 orig_start = start; + u64 orig_num_bytes; struct btrfs_key ins; int ret; @@ -115,6 +117,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) num_bytes = max(blocksize, num_bytes); ret = btrfs_drop_extents(trans, root, inode, start, start + num_bytes, start, &alloc_hint); + orig_num_bytes = num_bytes; if (alloc_hint == EXTENT_MAP_INLINE) goto out; @@ -138,6 +141,8 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; } + btrfs_drop_extent_cache(inode, orig_start, + orig_start + orig_num_bytes - 1); btrfs_add_ordered_inode(inode); out: btrfs_end_transaction(trans, root); @@ -297,7 +302,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) int ret = 0; struct inode *inode = page->mapping->host; struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct btrfs_csum_item *item; struct btrfs_path *path = NULL; u32 csum; @@ -317,7 +322,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) } read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, BTRFS_CRC32_SIZE); - set_state_private(em_tree, start, csum); + set_state_private(io_tree, start, csum); out: if (path) btrfs_free_path(path); @@ -329,17 +334,19 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) { size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT); struct inode *inode = page->mapping->host; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; char *kaddr; u64 private; int ret; struct btrfs_root *root = BTRFS_I(inode)->root; u32 csum = ~(u32)0; unsigned long flags; + if (btrfs_test_opt(root, NODATASUM) || btrfs_test_flag(inode, NODATASUM)) return 0; - ret = get_state_private(em_tree, start, &private); + + ret = get_state_private(io_tree, start, &private); local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); if (ret) { @@ -428,7 +435,7 @@ void btrfs_read_locked_inode(struct inode *inode) switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; - BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; + BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; break; @@ -873,7 +880,7 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, size_t zero_start) { char *kaddr; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct btrfs_root *root = BTRFS_I(inode)->root; u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; @@ -884,12 +891,12 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, WARN_ON(!PageLocked(page)); set_page_extent_mapped(page); - lock_extent(em_tree, page_start, page_end, GFP_NOFS); + lock_extent(io_tree, page_start, page_end, GFP_NOFS); delalloc_start = page_start; - existing_delalloc = count_range_bits(&BTRFS_I(inode)->extent_tree, + existing_delalloc = count_range_bits(&BTRFS_I(inode)->io_tree, &delalloc_start, page_end, PAGE_CACHE_SIZE, EXTENT_DELALLOC); - set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start, + set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); spin_lock(&root->fs_info->delalloc_lock); @@ -903,7 +910,7 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, kunmap(page); } set_page_dirty(page); - unlock_extent(em_tree, page_start, page_end, GFP_NOFS); + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); return ret; } @@ -961,7 +968,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; u64 mask = root->sectorsize - 1; u64 pos = (inode->i_size + mask) & ~mask; @@ -986,7 +993,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); - lock_extent(em_tree, pos, block_end, GFP_NOFS); + lock_extent(io_tree, pos, block_end, GFP_NOFS); hole_size = block_end - hole_start; mutex_lock(&root->fs_info->fs_mutex); @@ -1001,11 +1008,13 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) inode->i_ino, hole_start, 0, 0, hole_size); + btrfs_drop_extent_cache(inode, hole_start, + hole_size - 1); btrfs_check_file(root, inode); } btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - unlock_extent(em_tree, pos, block_end, GFP_NOFS); + unlock_extent(io_tree, pos, block_end, GFP_NOFS); if (err) return err; } @@ -1189,7 +1198,8 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) struct btrfs_iget_args *args = p; inode->i_ino = args->ino; BTRFS_I(inode)->root = args->root; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); return 0; } @@ -1485,7 +1495,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, if (!inode) return ERR_PTR(-ENOMEM); - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->root = root; @@ -1672,9 +1683,10 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping->a_ops = &btrfs_aops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); - BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; + BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -1816,7 +1828,7 @@ out_unlock: } struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, - size_t page_offset, u64 start, u64 end, + size_t page_offset, u64 start, u64 len, int create) { int ret; @@ -1826,7 +1838,6 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, u64 extent_end = 0; u64 objectid = inode->i_ino; u32 found_type; - int failed_insert = 0; struct btrfs_path *path; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_file_extent_item *item; @@ -1834,6 +1845,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, struct btrfs_key found_key; struct extent_map *em = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct btrfs_trans_handle *trans = NULL; path = btrfs_alloc_path(); @@ -1841,24 +1853,26 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, mutex_lock(&root->fs_info->fs_mutex); again: - em = lookup_extent_mapping(em_tree, start, end); + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, start, len); + spin_unlock(&em_tree->lock); + if (em) { if (em->start > start) { - printk("get_extent start %Lu em start %Lu\n", - start, em->start); + printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n", + start, len, em->start, em->len); WARN_ON(1); } goto out; } + em = alloc_extent_map(GFP_NOFS); if (!em) { - em = alloc_extent_map(GFP_NOFS); - if (!em) { - err = -ENOMEM; - goto out; - } - em->start = EXTENT_MAP_HOLE; - em->end = EXTENT_MAP_HOLE; + err = -ENOMEM; + goto out; } + + em->start = EXTENT_MAP_HOLE; + em->len = (u64)-1; em->bdev = inode->i_sb->s_bdev; ret = btrfs_lookup_file_extent(trans, root, path, objectid, start, trans != NULL); @@ -1893,28 +1907,25 @@ again: if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { - if (end < extent_start) + if (start + len <= extent_start) goto not_found; - em->end = extent_end - 1; + em->len = extent_end - extent_start; } else { - em->end = end; + em->len = len; } goto not_found_em; } bytenr = btrfs_file_extent_disk_bytenr(leaf, item); if (bytenr == 0) { em->start = extent_start; - em->end = extent_end - 1; + em->len = extent_end - extent_start; em->block_start = EXTENT_MAP_HOLE; - em->block_end = EXTENT_MAP_HOLE; goto insert; } bytenr += btrfs_file_extent_offset(leaf, item); em->block_start = bytenr; - em->block_end = em->block_start + - btrfs_file_extent_num_bytes(leaf, item) - 1; em->start = extent_start; - em->end = extent_end - 1; + em->len = extent_end - extent_start; goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { unsigned long ptr; @@ -1925,25 +1936,24 @@ again: size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, path->slots[0])); - extent_end = (extent_start + size - 1) | - ((u64)root->sectorsize - 1); + extent_end = (extent_start + size + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); if (start < extent_start || start >= extent_end) { em->start = start; if (start < extent_start) { - if (end < extent_start) + if (start + len <= extent_start) goto not_found; - em->end = extent_end; + em->len = extent_end - extent_start; } else { - em->end = end; + em->len = len; } goto not_found_em; } em->block_start = EXTENT_MAP_INLINE; - em->block_end = EXTENT_MAP_INLINE; if (!page) { em->start = extent_start; - em->end = extent_start + size - 1; + em->len = size; goto out; } @@ -1952,8 +1962,7 @@ again: copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset, size - extent_offset); em->start = extent_start + extent_offset; - em->end = (em->start + copy_size -1) | - ((u64)root->sectorsize -1); + em->len = copy_size; map = kmap(page); ptr = btrfs_file_extent_inline_start(item) + extent_offset; if (create == 0 && !PageUptodate(page)) { @@ -1974,7 +1983,8 @@ again: btrfs_mark_buffer_dirty(leaf); } kunmap(page); - set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS); + set_extent_uptodate(io_tree, em->start, + extent_map_end(em) - 1, GFP_NOFS); goto insert; } else { printk("unkknown found_type %d\n", found_type); @@ -1982,33 +1992,29 @@ again: } not_found: em->start = start; - em->end = end; + em->len = len; not_found_em: em->block_start = EXTENT_MAP_HOLE; - em->block_end = EXTENT_MAP_HOLE; insert: btrfs_release_path(root, path); - if (em->start > start || em->end < start) { - printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end); + if (em->start > start || extent_map_end(em) <= start) { + printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len); err = -EIO; goto out; } + + err = 0; + spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); if (ret == -EEXIST) { free_extent_map(em); - em = NULL; - if (0 && failed_insert == 1) { - btrfs_drop_extent_cache(inode, start, end); - } - failed_insert++; - if (failed_insert > 5) { - printk("failing to insert %Lu %Lu\n", start, end); + em = lookup_extent_mapping(em_tree, start, len); + if (!em) { err = -EIO; - goto out; + printk("failing to insert %Lu %Lu\n", start, len); } - goto again; } - err = 0; + spin_unlock(&em_tree->lock); out: btrfs_free_path(path); if (trans) { @@ -2032,14 +2038,14 @@ static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) int btrfs_readpage(struct file *file, struct page *page) { - struct extent_map_tree *tree; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; return extent_read_full_page(tree, page, btrfs_get_extent); } static int btrfs_writepage(struct page *page, struct writeback_control *wbc) { - struct extent_map_tree *tree; + struct extent_io_tree *tree; if (current->flags & PF_MEMALLOC) { @@ -2047,15 +2053,15 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) unlock_page(page); return 0; } - tree = &BTRFS_I(page->mapping->host)->extent_tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } static int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc) { - struct extent_map_tree *tree; - tree = &BTRFS_I(mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(mapping->host)->io_tree; return extent_writepages(tree, mapping, btrfs_get_extent, wbc); } @@ -2063,19 +2069,21 @@ static int btrfs_readpages(struct file *file, struct address_space *mapping, struct list_head *pages, unsigned nr_pages) { - struct extent_map_tree *tree; - tree = &BTRFS_I(mapping->host)->extent_tree; + struct extent_io_tree *tree; + tree = &BTRFS_I(mapping->host)->io_tree; return extent_readpages(tree, mapping, pages, nr_pages, btrfs_get_extent); } static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) { - struct extent_map_tree *tree; + struct extent_io_tree *tree; + struct extent_map_tree *map; int ret; - tree = &BTRFS_I(page->mapping->host)->extent_tree; - ret = try_release_extent_mapping(tree, page); + tree = &BTRFS_I(page->mapping->host)->io_tree; + map = &BTRFS_I(page->mapping->host)->extent_tree; + ret = try_release_extent_mapping(map, tree, page); if (ret == 1) { ClearPagePrivate(page); set_page_private(page, 0); @@ -2086,9 +2094,9 @@ static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) static void btrfs_invalidatepage(struct page *page, unsigned long offset) { - struct extent_map_tree *tree; + struct extent_io_tree *tree; - tree = &BTRFS_I(page->mapping->host)->extent_tree; + tree = &BTRFS_I(page->mapping->host)->io_tree; extent_invalidatepage(tree, page, offset); btrfs_releasepage(page, GFP_NOFS); } @@ -2374,7 +2382,7 @@ unsigned long btrfs_force_ra(struct address_space *mapping, int btrfs_defrag_file(struct file *file) { struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct page *page; unsigned long last_index; unsigned long ra_index = 0; @@ -2414,13 +2422,13 @@ int btrfs_defrag_file(struct file *file) { page_start = (u64)page->index << PAGE_CACHE_SHIFT; page_end = page_start + PAGE_CACHE_SIZE - 1; - lock_extent(em_tree, page_start, page_end, GFP_NOFS); + lock_extent(io_tree, page_start, page_end, GFP_NOFS); delalloc_start = page_start; existing_delalloc = - count_range_bits(&BTRFS_I(inode)->extent_tree, + count_range_bits(&BTRFS_I(inode)->io_tree, &delalloc_start, page_end, PAGE_CACHE_SIZE, EXTENT_DELALLOC); - set_extent_delalloc(em_tree, page_start, + set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); spin_lock(&root->fs_info->delalloc_lock); @@ -2428,7 +2436,7 @@ int btrfs_defrag_file(struct file *file) { existing_delalloc; spin_unlock(&root->fs_info->delalloc_lock); - unlock_extent(em_tree, page_start, page_end, GFP_NOFS); + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); unlock_page(page); page_cache_release(page); @@ -2842,9 +2850,10 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping->a_ops = &btrfs_aops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); - BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops; + BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2934,7 +2943,7 @@ static struct file_operations btrfs_dir_file_operations = { #endif }; -static struct extent_map_ops btrfs_extent_map_ops = { +static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, .writepage_io_hook = btrfs_writepage_io_hook, .readpage_io_hook = btrfs_readpage_io_hook, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 8b52c69fda2..f8a1016600b 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -468,10 +468,15 @@ static int __init init_btrfs_fs(void) err = btrfs_init_cachep(); if (err) goto free_transaction_sys; - err = extent_map_init(); + + err = extent_io_init(); if (err) goto free_cachep; + err = extent_map_init(); + if (err) + goto free_extent_io; + err = register_filesystem(&btrfs_fs_type); if (err) goto free_extent_map; @@ -479,6 +484,8 @@ static int __init init_btrfs_fs(void) free_extent_map: extent_map_exit(); +free_extent_io: + extent_io_exit(); free_cachep: btrfs_destroy_cachep(); free_transaction_sys: @@ -492,6 +499,7 @@ static void __exit exit_btrfs_fs(void) btrfs_exit_transaction_sys(); btrfs_destroy_cachep(); extent_map_exit(); + extent_io_exit(); unregister_filesystem(&btrfs_fs_type); btrfs_exit_sysfs(); } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 163c01a2449..b4a1bc62a78 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -70,7 +70,7 @@ static int join_transaction(struct btrfs_root *root) INIT_LIST_HEAD(&cur_trans->pending_snapshots); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree); - extent_map_tree_init(&cur_trans->dirty_pages, + extent_io_tree_init(&cur_trans->dirty_pages, root->fs_info->btree_inode->i_mapping, GFP_NOFS); } else { @@ -153,7 +153,7 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, int ret; int err; int werr = 0; - struct extent_map_tree *dirty_pages; + struct extent_io_tree *dirty_pages; struct page *page; struct inode *btree_inode = root->fs_info->btree_inode; u64 start; @@ -610,7 +610,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction *cur_trans; struct btrfs_transaction *prev_trans = NULL; struct list_head dirty_fs_roots; - struct extent_map_tree *pinned_copy; + struct extent_io_tree *pinned_copy; DEFINE_WAIT(wait); int ret; @@ -639,7 +639,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, if (!pinned_copy) return -ENOMEM; - extent_map_tree_init(pinned_copy, + extent_io_tree_init(pinned_copy, root->fs_info->btree_inode->i_mapping, GFP_NOFS); trans->transaction->in_commit = 1; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index fd52e9b2392..c3172ddb332 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -29,7 +29,7 @@ struct btrfs_transaction { int use_count; int commit_done; struct list_head list; - struct extent_map_tree dirty_pages; + struct extent_io_tree dirty_pages; unsigned long start_time; struct btrfs_ordered_inode_tree ordered_inode_tree; wait_queue_head_t writer_wait; -- cgit v1.2.3-70-g09d2 From c2a8b6e11009398ca9363d8ba8d4e7e40fb897fd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 09:10:26 -0500 Subject: Btrfs: Force f_pos to the max when a readdir hits the end of the directory. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 16d3aef45d1..026bc9c4c90 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1430,7 +1430,7 @@ read_dir_items: di = (struct btrfs_dir_item *)((char *)di + di_len); } } - filp->f_pos++; + filp->f_pos = INT_LIMIT(typeof(filp->f_pos)); nopos: ret = 0; err: -- cgit v1.2.3-70-g09d2 From aadfeb6e39ad6bde080cb3ab23f4da57ccb25f4a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 09:10:27 -0500 Subject: Btrfs: Add some extra debugging around file data checksum failures Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 4 ++++ fs/btrfs/inode.c | 8 +++++--- 2 files changed, 9 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 614176e5285..3ebbc058d08 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -261,6 +261,10 @@ csum: found: csum_result = btrfs_csum_data(root, data, csum_result, len); btrfs_csum_final(csum_result, (char *)&csum_result); + if (csum_result == 0) { + printk("csum result is 0 for inode %lu offset %Lu\n", inode->i_ino, offset); + } + write_extent_buffer(leaf, &csum_result, (unsigned long)item, BTRFS_CRC32_SIZE); btrfs_mark_buffer_dirty(path->nodes[0]); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 026bc9c4c90..48455127410 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -318,6 +318,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) if (ret == -ENOENT || ret == -EFBIG) ret = 0; csum = 0; + printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start); goto out; } read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, @@ -336,7 +337,7 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) struct inode *inode = page->mapping->host; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; char *kaddr; - u64 private; + u64 private = ~(u32)0; int ret; struct btrfs_root *root = BTRFS_I(inode)->root; u32 csum = ~(u32)0; @@ -362,8 +363,9 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) return 0; zeroit: - printk("btrfs csum failed ino %lu off %llu\n", - page->mapping->host->i_ino, (unsigned long long)start); + printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n", + page->mapping->host->i_ino, (unsigned long long)start, csum, + private); memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); -- cgit v1.2.3-70-g09d2 From 70dec8079d78691e476cc6c7cede40656078ad30 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 09:59:12 -0500 Subject: Btrfs: extent_io and extent_state optimizations The end_bio routines are changed to take a pointer to the extent state struct, and the state tree is walked in order to set/clear appropriate bits as IO completes. This greatly reduces the number of rbtree searches done by the end_bio handlers, and reduces lock contention. The extent_io releasepage function is changed to avoid expensive searches for locked state. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 4 +- fs/btrfs/extent_io.c | 358 +++++++++++++++++++++++++++++++++++++-------------- fs/btrfs/extent_io.h | 15 ++- fs/btrfs/inode.c | 35 +++-- 4 files changed, 298 insertions(+), 114 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 4c4ebea0b2a..e29a0716f6a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -274,7 +274,7 @@ int btree_readpage(struct file *file, struct page *page) return extent_read_full_page(tree, page, btree_get_extent); } -static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) +static int btree_releasepage(struct page *page, gfp_t gfp_flags) { struct extent_io_tree *tree; struct extent_map_tree *map; @@ -282,7 +282,7 @@ static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags) tree = &BTRFS_I(page->mapping->host)->io_tree; map = &BTRFS_I(page->mapping->host)->extent_tree; - ret = try_release_extent_mapping(map, tree, page); + ret = try_release_extent_mapping(map, tree, page, gfp_flags); if (ret == 1) { ClearPagePrivate(page); set_page_private(page, 0); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 15cc158a049..1f734c34dc2 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -27,13 +27,11 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); -static spinlock_t state_lock = SPIN_LOCK_UNLOCKED; #define BUFFER_LRU_MAX 64 struct tree_entry { u64 start; u64 end; - int in_tree; struct rb_node rb_node; }; @@ -69,7 +67,7 @@ void extent_io_exit(void) while (!list_empty(&states)) { state = list_entry(states.next, struct extent_state, list); - printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs)); + printk("state leak: start %Lu end %Lu state %lu in tree %p refs %d\n", state->start, state->end, state->state, state->tree, atomic_read(&state->refs)); list_del(&state->list); kmem_cache_free(extent_state_cache, state); @@ -87,7 +85,7 @@ void extent_io_tree_init(struct extent_io_tree *tree, tree->state.rb_node = NULL; tree->ops = NULL; tree->dirty_bytes = 0; - rwlock_init(&tree->lock); + spin_lock_init(&tree->lock); spin_lock_init(&tree->lru_lock); tree->mapping = mapping; INIT_LIST_HEAD(&tree->buffer_lru); @@ -110,18 +108,13 @@ EXPORT_SYMBOL(extent_io_tree_empty_lru); struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; - unsigned long flags; state = kmem_cache_alloc(extent_state_cache, mask); if (!state || IS_ERR(state)) return state; state->state = 0; - state->in_tree = 0; state->private = 0; - - spin_lock_irqsave(&state_lock, flags); - list_add(&state->list, &states); - spin_unlock_irqrestore(&state_lock, flags); + state->tree = NULL; atomic_set(&state->refs, 1); init_waitqueue_head(&state->wq); @@ -131,14 +124,10 @@ EXPORT_SYMBOL(alloc_extent_state); void free_extent_state(struct extent_state *state) { - unsigned long flags; if (!state) return; if (atomic_dec_and_test(&state->refs)) { - WARN_ON(state->in_tree); - spin_lock_irqsave(&state_lock, flags); - list_del(&state->list); - spin_unlock_irqrestore(&state_lock, flags); + WARN_ON(state->tree); kmem_cache_free(extent_state_cache, state); } } @@ -164,7 +153,6 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset, } entry = rb_entry(node, struct tree_entry, rb_node); - entry->in_tree = 1; rb_link_node(node, parent, p); rb_insert_color(node, root); return NULL; @@ -216,8 +204,9 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) { - struct rb_node *prev; + struct rb_node *prev = NULL; struct rb_node *ret; + ret = __tree_search(root, offset, &prev, NULL); if (!ret) return prev; @@ -248,7 +237,7 @@ static int merge_state(struct extent_io_tree *tree, if (other->end == state->start - 1 && other->state == state->state) { state->start = other->start; - other->in_tree = 0; + other->tree = NULL; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -259,7 +248,7 @@ static int merge_state(struct extent_io_tree *tree, if (other->start == state->end + 1 && other->state == state->state) { other->start = state->start; - state->in_tree = 0; + state->tree = NULL; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); } @@ -300,6 +289,7 @@ static int insert_state(struct extent_io_tree *tree, free_extent_state(state); return -EEXIST; } + state->tree = tree; merge_state(tree, state); return 0; } @@ -335,6 +325,7 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, free_extent_state(prealloc); return -EEXIST; } + prealloc->tree = tree; return 0; } @@ -361,9 +352,9 @@ static int clear_state_bit(struct extent_io_tree *tree, if (wake) wake_up(&state->wq); if (delete || state->state == 0) { - if (state->in_tree) { + if (state->tree) { rb_erase(&state->rb_node, &tree->state); - state->in_tree = 0; + state->tree = NULL; free_extent_state(state); } else { WARN_ON(1); @@ -404,7 +395,7 @@ again: return -ENOMEM; } - write_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); /* * this search will find the extents that end after * our range starts @@ -434,6 +425,8 @@ again: */ if (state->start < start) { + if (!prealloc) + prealloc = alloc_extent_state(GFP_ATOMIC); err = split_state(tree, state, prealloc, start); BUG_ON(err == -EEXIST); prealloc = NULL; @@ -455,6 +448,8 @@ again: * on the first half */ if (state->start <= end && state->end > end) { + if (!prealloc) + prealloc = alloc_extent_state(GFP_ATOMIC); err = split_state(tree, state, prealloc, end + 1); BUG_ON(err == -EEXIST); @@ -471,7 +466,7 @@ again: goto search_again; out: - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (prealloc) free_extent_state(prealloc); @@ -480,7 +475,7 @@ out: search_again: if (start > end) goto out; - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -492,9 +487,9 @@ static int wait_on_state(struct extent_io_tree *tree, { DEFINE_WAIT(wait); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); schedule(); - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); finish_wait(&state->wq, &wait); return 0; } @@ -509,7 +504,7 @@ int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) struct extent_state *state; struct rb_node *node; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); again: while (1) { /* @@ -538,13 +533,13 @@ again: break; if (need_resched()) { - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); cond_resched(); - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); } } out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return 0; } EXPORT_SYMBOL(wait_extent_bit); @@ -589,7 +584,7 @@ again: return -ENOMEM; } - write_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); /* * this search will find all the extents that end after * our range starts. @@ -709,7 +704,7 @@ again: goto search_again; out: - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (prealloc) free_extent_state(prealloc); @@ -718,7 +713,7 @@ out: search_again: if (start > end) goto out; - write_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -817,10 +812,6 @@ int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(wait_on_extent_writeback); -/* - * locks a range in ascending order, waiting for any locked regions - * it hits on the way. [start,end] are inclusive, and this will sleep. - */ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { int err; @@ -896,7 +887,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, struct extent_state *state; int ret = 1; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -919,7 +910,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, break; } out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } EXPORT_SYMBOL(find_first_extent_bit); @@ -933,7 +924,7 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, u64 found = 0; u64 total_bytes = 0; - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -976,9 +967,9 @@ search_again: atomic_inc(&state->refs); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); schedule(); - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); finish_wait(&state->wq, &wait); free_extent_state(state); goto search_again; @@ -997,7 +988,7 @@ search_again: break; } out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return found; } @@ -1017,7 +1008,7 @@ u64 count_range_bits(struct extent_io_tree *tree, return 0; } - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); if (cur_start == 0 && bits == EXTENT_DIRTY) { total_bytes = tree->dirty_bytes; goto out; @@ -1050,7 +1041,7 @@ u64 count_range_bits(struct extent_io_tree *tree, break; } out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return total_bytes; } /* @@ -1122,7 +1113,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) struct extent_state *state; int ret = 0; - write_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1139,7 +1130,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) } state->private = private; out: - write_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } @@ -1149,7 +1140,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) struct extent_state *state; int ret = 0; - read_lock_irq(&tree->lock); + spin_lock_irq(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1166,13 +1157,13 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) } *private = state->private; out: - read_unlock_irq(&tree->lock); + spin_unlock_irq(&tree->lock); return ret; } /* * searches a range in the state tree for a given mask. - * If 'filled' == 1, this returns 1 only if ever extent in the tree + * If 'filled' == 1, this returns 1 only if every extent in the tree * has the bits set. Otherwise, 1 is returned if any bit in the * range is found set. */ @@ -1184,7 +1175,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, int bitset = 0; unsigned long flags; - read_lock_irqsave(&tree->lock, flags); + spin_lock_irqsave(&tree->lock, flags); node = tree_search(&tree->state, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1215,7 +1206,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, break; } } - read_unlock_irqrestore(&tree->lock, flags); + spin_unlock_irqrestore(&tree->lock, flags); return bitset; } EXPORT_SYMBOL(test_range_bit); @@ -1282,16 +1273,19 @@ static int end_bio_extent_writepage(struct bio *bio, { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; + struct rb_node *node; u64 start; u64 end; + u64 cur; int whole_page; + unsigned long flags; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) if (bio->bi_size) return 1; #endif - do { struct page *page = bvec->bv_page; start = ((u64)page->index << PAGE_CACHE_SHIFT) + @@ -1311,16 +1305,80 @@ static int end_bio_extent_writepage(struct bio *bio, ClearPageUptodate(page); SetPageError(page); } - clear_extent_writeback(tree, start, end, GFP_ATOMIC); + + if (tree->ops && tree->ops->writepage_end_io_hook) { + tree->ops->writepage_end_io_hook(page, start, end, + state); + } + + /* + * bios can get merged in funny ways, and so we need to + * be careful with the state variable. We know the + * state won't be merged with others because it has + * WRITEBACK set, but we can't be sure each biovec is + * sequential in the file. So, if our cached state + * doesn't match the expected end, search the tree + * for the correct one. + */ + + spin_lock_irqsave(&tree->lock, flags); + if (!state || state->end != end) { + state = NULL; + node = __tree_search(&tree->state, start, NULL, NULL); + if (node) { + state = rb_entry(node, struct extent_state, + rb_node); + if (state->end != end || + !(state->state & EXTENT_WRITEBACK)) + state = NULL; + } + if (!state) { + spin_unlock_irqrestore(&tree->lock, flags); + clear_extent_writeback(tree, start, + end, GFP_ATOMIC); + goto next_io; + } + } + cur = end; + while(1) { + struct extent_state *clear = state; + cur = state->start; + node = rb_prev(&state->rb_node); + if (node) { + state = rb_entry(node, + struct extent_state, + rb_node); + } else { + state = NULL; + } + + clear_state_bit(tree, clear, EXTENT_WRITEBACK, + 1, 0); + if (cur == start) + break; + if (cur < start) { + WARN_ON(1); + break; + } + if (!node) + break; + } + /* before releasing the lock, make sure the next state + * variable has the expected bits set and corresponds + * to the correct offsets in the file + */ + if (state && (state->end + 1 != start || + !state->state & EXTENT_WRITEBACK)) { + state = NULL; + } + spin_unlock_irqrestore(&tree->lock, flags); +next_io: if (whole_page) end_page_writeback(page); else check_page_writeback(tree, page); - if (tree->ops && tree->ops->writepage_end_io_hook) - tree->ops->writepage_end_io_hook(page, start, end); } while (bvec >= bio->bi_io_vec); - bio_put(bio); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) return 0; @@ -1347,9 +1405,13 @@ static int end_bio_extent_readpage(struct bio *bio, { int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; + struct rb_node *node; u64 start; u64 end; + u64 cur; + unsigned long flags; int whole_page; int ret; @@ -1373,27 +1435,83 @@ static int end_bio_extent_readpage(struct bio *bio, prefetchw(&bvec->bv_page->flags); if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { - ret = tree->ops->readpage_end_io_hook(page, start, end); + ret = tree->ops->readpage_end_io_hook(page, start, end, + state); if (ret) uptodate = 0; } - if (uptodate) { - set_extent_uptodate(tree, start, end, GFP_ATOMIC); - if (whole_page) - SetPageUptodate(page); - else - check_page_uptodate(tree, page); - } else { - ClearPageUptodate(page); - SetPageError(page); - } - unlock_extent(tree, start, end, GFP_ATOMIC); + spin_lock_irqsave(&tree->lock, flags); + if (!state || state->end != end) { + state = NULL; + node = __tree_search(&tree->state, start, NULL, NULL); + if (node) { + state = rb_entry(node, struct extent_state, + rb_node); + if (state->end != end || + !(state->state & EXTENT_LOCKED)) + state = NULL; + } + if (!state) { + spin_unlock_irqrestore(&tree->lock, flags); + set_extent_uptodate(tree, start, end, + GFP_ATOMIC); + unlock_extent(tree, start, end, GFP_ATOMIC); + goto next_io; + } + } - if (whole_page) + cur = end; + while(1) { + struct extent_state *clear = state; + cur = state->start; + node = rb_prev(&state->rb_node); + if (node) { + state = rb_entry(node, + struct extent_state, + rb_node); + } else { + state = NULL; + } + clear->state |= EXTENT_UPTODATE; + clear_state_bit(tree, clear, EXTENT_LOCKED, + 1, 0); + if (cur == start) + break; + if (cur < start) { + WARN_ON(1); + break; + } + if (!node) + break; + } + /* before releasing the lock, make sure the next state + * variable has the expected bits set and corresponds + * to the correct offsets in the file + */ + if (state && (state->end + 1 != start || + !state->state & EXTENT_WRITEBACK)) { + state = NULL; + } + spin_unlock_irqrestore(&tree->lock, flags); +next_io: + if (whole_page) { + if (uptodate) { + SetPageUptodate(page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } unlock_page(page); - else + } else { + if (uptodate) { + check_page_uptodate(tree, page); + } else { + ClearPageUptodate(page); + SetPageError(page); + } check_page_locked(tree, page); + } } while (bvec >= bio->bi_io_vec); bio_put(bio); @@ -1416,7 +1534,8 @@ static int end_bio_extent_preparewrite(struct bio *bio, { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; - struct extent_io_tree *tree = bio->bi_private; + struct extent_state *state = bio->bi_private; + struct extent_io_tree *tree = state->tree; u64 start; u64 end; @@ -1475,6 +1594,29 @@ static int submit_one_bio(int rw, struct bio *bio) { u64 maxsector; int ret = 0; + struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct page *page = bvec->bv_page; + struct extent_io_tree *tree = bio->bi_private; + struct rb_node *node; + struct extent_state *state; + u64 start; + u64 end; + + start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; + end = start + bvec->bv_len - 1; + + spin_lock_irq(&tree->lock); + node = __tree_search(&tree->state, start, NULL, NULL); + BUG_ON(!node); + state = rb_entry(node, struct extent_state, rb_node); + while(state->end < end) { + node = rb_next(node); + state = rb_entry(node, struct extent_state, rb_node); + } + BUG_ON(state->end != end); + spin_unlock_irq(&tree->lock); + + bio->bi_private = state; bio_get(bio); @@ -1519,9 +1661,12 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (!bio) { printk("failed to allocate bio nr %d\n", nr); } + + bio_add_page(bio, page, size, offset); bio->bi_end_io = end_io_func; bio->bi_private = tree; + if (bio_ret) { *bio_ret = bio; } else { @@ -1635,6 +1780,16 @@ static int __extent_read_full_page(struct extent_io_tree *tree, page_offset += iosize; continue; } + /* we have an inline extent but it didn't get marked up + * to date. Error out + */ + if (block_start == EXTENT_MAP_INLINE) { + SetPageError(page); + unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); + cur = cur + iosize; + page_offset += iosize; + continue; + } ret = 0; if (tree->ops && tree->ops->readpage_io_hook) { @@ -2205,7 +2360,8 @@ EXPORT_SYMBOL(extent_prepare_write); * map records are removed */ int try_release_extent_mapping(struct extent_map_tree *map, - struct extent_io_tree *tree, struct page *page) + struct extent_io_tree *tree, struct page *page, + gfp_t mask) { struct extent_map *em; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -2213,30 +2369,42 @@ int try_release_extent_mapping(struct extent_map_tree *map, u64 orig_start = start; int ret = 1; - while (start <= end) { - spin_lock(&map->lock); - em = lookup_extent_mapping(map, start, end); - if (!em || IS_ERR(em)) { + if ((mask & __GFP_WAIT) && + page->mapping->host->i_size > 16 * 1024 * 1024) { + while (start <= end) { + spin_lock(&map->lock); + em = lookup_extent_mapping(map, start, end); + if (!em || IS_ERR(em)) { + spin_unlock(&map->lock); + break; + } + if (em->start != start) { + spin_unlock(&map->lock); + free_extent_map(em); + break; + } + if (!test_range_bit(tree, em->start, + extent_map_end(em) - 1, + EXTENT_LOCKED, 0)) { + remove_extent_mapping(map, em); + /* once for the rb tree */ + free_extent_map(em); + } + start = extent_map_end(em); spin_unlock(&map->lock); - break; - } - if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, - EXTENT_LOCKED, 0)) { - remove_extent_mapping(map, em); - /* once for the rb tree */ + + /* once for us */ free_extent_map(em); } - start = extent_map_end(em); - spin_unlock(&map->lock); - - /* once for us */ - free_extent_map(em); } - if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0)) + if (test_range_bit(tree, orig_start, end, EXTENT_IOBITS, 0)) ret = 0; - else + else { + if ((mask & GFP_NOFS) == GFP_NOFS) + mask = GFP_NOFS; clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE, - 1, 1, GFP_NOFS); + 1, 1, mask); + } return ret; } EXPORT_SYMBOL(try_release_extent_mapping); @@ -2553,13 +2721,13 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); - write_lock_irq(&page->mapping->tree_lock); + read_lock_irq(&page->mapping->tree_lock); if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } - write_unlock_irq(&page->mapping->tree_lock); + read_unlock_irq(&page->mapping->tree_lock); unlock_page(page); } return 0; diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 06be1fe84b2..f1e8bf251f3 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -23,19 +23,23 @@ #define EXTENT_PAGE_PRIVATE 1 #define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3 +struct extent_state; + struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); - int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end); - void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, + struct extent_state *state); + void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, + struct extent_state *state); }; struct extent_io_tree { struct rb_root state; struct address_space *mapping; u64 dirty_bytes; - rwlock_t lock; + spinlock_t lock; struct extent_io_ops *ops; spinlock_t lru_lock; struct list_head buffer_lru; @@ -45,8 +49,8 @@ struct extent_io_tree { struct extent_state { u64 start; u64 end; /* inclusive */ - int in_tree; struct rb_node rb_node; + struct extent_io_tree *tree; wait_queue_head_t wq; atomic_t refs; unsigned long state; @@ -82,7 +86,8 @@ void extent_io_tree_init(struct extent_io_tree *tree, struct address_space *mapping, gfp_t mask); void extent_io_tree_empty_lru(struct extent_io_tree *tree); int try_release_extent_mapping(struct extent_map_tree *map, - struct extent_io_tree *tree, struct page *page); + struct extent_io_tree *tree, struct page *page, + gfp_t mask); int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); int extent_read_full_page(struct extent_io_tree *tree, struct page *page, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 48455127410..bac8722e14e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -331,7 +331,8 @@ out: return ret; } -int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) +int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, + struct extent_state *state) { size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT); struct inode *inode = page->mapping->host; @@ -347,7 +348,12 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end) btrfs_test_flag(inode, NODATASUM)) return 0; - ret = get_state_private(io_tree, start, &private); + if (state->start == start) { + private = state->private; + ret = 0; + } else { + ret = get_state_private(io_tree, start, &private); + } local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); if (ret) { @@ -1830,7 +1836,7 @@ out_unlock: } struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, - size_t page_offset, u64 start, u64 len, + size_t pg_offset, u64 start, u64 len, int create) { int ret; @@ -1865,7 +1871,10 @@ again: start, len, em->start, em->len); WARN_ON(1); } - goto out; + if (em->block_start == EXTENT_MAP_INLINE && page) + free_extent_map(em); + else + goto out; } em = alloc_extent_map(GFP_NOFS); if (!em) { @@ -1930,6 +1939,7 @@ again: em->len = extent_end - extent_start; goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + u64 page_start; unsigned long ptr; char *map; size_t size; @@ -1959,16 +1969,17 @@ again: goto out; } - extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) - - extent_start + page_offset; - copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset, + page_start = page_offset(page) + pg_offset; + extent_offset = page_start - extent_start; + copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset, size - extent_offset); em->start = extent_start + extent_offset; - em->len = copy_size; + em->len = (copy_size + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); map = kmap(page); ptr = btrfs_file_extent_inline_start(item) + extent_offset; if (create == 0 && !PageUptodate(page)) { - read_extent_buffer(leaf, map + page_offset, ptr, + read_extent_buffer(leaf, map + pg_offset, ptr, copy_size); flush_dcache_page(page); } else if (create && PageUptodate(page)) { @@ -1980,7 +1991,7 @@ again: trans = btrfs_start_transaction(root, 1); goto again; } - write_extent_buffer(leaf, map + page_offset, ptr, + write_extent_buffer(leaf, map + pg_offset, ptr, copy_size); btrfs_mark_buffer_dirty(leaf); } @@ -2077,7 +2088,7 @@ btrfs_readpages(struct file *file, struct address_space *mapping, btrfs_get_extent); } -static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) +static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) { struct extent_io_tree *tree; struct extent_map_tree *map; @@ -2085,7 +2096,7 @@ static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags) tree = &BTRFS_I(page->mapping->host)->io_tree; map = &BTRFS_I(page->mapping->host)->extent_tree; - ret = try_release_extent_mapping(map, tree, page); + ret = try_release_extent_mapping(map, tree, page, gfp_flags); if (ret == 1) { ClearPagePrivate(page); set_page_private(page, 0); -- cgit v1.2.3-70-g09d2 From 85e21bac165b4ba1f6f90431ad6fc658ffcbaf3a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 15:11:36 -0500 Subject: Btrfs: During deletes and truncate, remove many items at once from the tree Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 38 +++++++++-------- fs/btrfs/ctree.h | 12 +++++- fs/btrfs/extent_io.c | 1 - fs/btrfs/inode.c | 114 +++++++++++++++++++++++++++++---------------------- 4 files changed, 95 insertions(+), 70 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 43d23148a4f..84ad53e06b3 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2514,34 +2514,36 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, * delete the item at the leaf level in path. If that empties * the leaf, remove it from the tree */ -int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct btrfs_path *path) +int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct btrfs_path *path, int slot, int nr) { - int slot; struct extent_buffer *leaf; struct btrfs_item *item; - int doff; - int dsize; + int last_off; + int dsize = 0; int ret = 0; int wret; + int i; u32 nritems; leaf = path->nodes[0]; - slot = path->slots[0]; - doff = btrfs_item_offset_nr(leaf, slot); - dsize = btrfs_item_size_nr(leaf, slot); + last_off = btrfs_item_offset_nr(leaf, slot + nr - 1); + + for (i = 0; i < nr; i++) + dsize += btrfs_item_size_nr(leaf, slot + i); + nritems = btrfs_header_nritems(leaf); - if (slot != nritems - 1) { + if (slot + nr != nritems) { int i; int data_end = leaf_data_end(root, leaf); memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + data_end + dsize, btrfs_leaf_data(leaf) + data_end, - doff - data_end); + last_off - data_end); - for (i = slot + 1; i < nritems; i++) { + for (i = slot + nr; i < nritems; i++) { u32 ioff; item = btrfs_item_nr(leaf, i); @@ -2562,12 +2564,12 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, } memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot), - btrfs_item_nr_offset(slot + 1), + btrfs_item_nr_offset(slot + nr), sizeof(struct btrfs_item) * - (nritems - slot - 1)); + (nritems - slot - nr)); } - btrfs_set_header_nritems(leaf, nritems - 1); - nritems--; + btrfs_set_header_nritems(leaf, nritems - nr); + nritems -= nr; /* delete the leaf if we've emptied it */ if (nritems == 0) { @@ -2600,7 +2602,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, } /* delete the leaf if it is mostly empty */ - if (used < BTRFS_LEAF_DATA_SIZE(root) / 3) { + if (used < BTRFS_LEAF_DATA_SIZE(root) / 4) { /* push_leaf_left fixes the path. * make sure the path still points to our leaf * for possible call to del_ptr below @@ -2608,13 +2610,13 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, slot = path->slots[1]; extent_buffer_get(leaf); - wret = push_leaf_right(trans, root, path, 1, 1); + wret = push_leaf_left(trans, root, path, 1, 1); if (wret < 0 && wret != -ENOSPC) ret = wret; if (path->nodes[0] == leaf && btrfs_header_nritems(leaf)) { - wret = push_leaf_left(trans, root, path, 1, 1); + wret = push_leaf_right(trans, root, path, 1, 1); if (wret < 0 && wret != -ENOSPC) ret = wret; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 6c65473e0fe..098cf088315 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1038,8 +1038,16 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); void btrfs_free_path(struct btrfs_path *p); void btrfs_init_path(struct btrfs_path *p); -int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct btrfs_path *path); +int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct btrfs_path *path, int slot, int nr); + +static inline int btrfs_del_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path) +{ + return btrfs_del_items(trans, root, path, path->slots[0], 1); +} + int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, void *data, u32 data_size); int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 1f734c34dc2..8aec72253a1 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2863,7 +2863,6 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (ret || !wait) { return ret; } - for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); wait_on_page_locked(page); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bac8722e14e..0a2fe51c412 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -692,27 +692,6 @@ fail: return err; } -static int btrfs_free_inode(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode) -{ - struct btrfs_path *path; - int ret; - - clear_inode(inode); - - path = btrfs_alloc_path(); - BUG_ON(!path); - ret = btrfs_lookup_inode(trans, root, path, - &BTRFS_I(inode)->location, -1); - if (ret > 0) - ret = -ENOENT; - if (!ret) - ret = btrfs_del_item(trans, root, path); - btrfs_free_path(path); - return ret; -} - /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find @@ -723,7 +702,8 @@ static int btrfs_free_inode(struct btrfs_trans_handle *trans, */ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct inode *inode) + struct inode *inode, + u32 min_type) { int ret; struct btrfs_path *path; @@ -739,6 +719,8 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, u64 root_owner = 0; int found_extent; int del_item; + int pending_del_nr = 0; + int pending_del_slot = 0; int extent_type = -1; btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1); @@ -751,17 +733,19 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, key.offset = (u64)-1; key.type = (u8)-1; + btrfs_init_path(path); +search_again: + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) { + goto error; + } + if (ret > 0) { + BUG_ON(path->slots[0] == 0); + path->slots[0]--; + } + while(1) { - btrfs_init_path(path); fi = NULL; - ret = btrfs_search_slot(trans, root, &key, path, -1, 1); - if (ret < 0) { - goto error; - } - if (ret > 0) { - BUG_ON(path->slots[0] == 0); - path->slots[0]--; - } leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); found_type = btrfs_key_type(&found_key); @@ -769,10 +753,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, if (found_key.objectid != inode->i_ino) break; - if (found_type != BTRFS_CSUM_ITEM_KEY && - found_type != BTRFS_DIR_ITEM_KEY && - found_type != BTRFS_DIR_INDEX_KEY && - found_type != BTRFS_EXTENT_DATA_KEY) + if (found_type < min_type) break; item_end = found_key.offset; @@ -801,14 +782,17 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, found_type = BTRFS_INODE_ITEM_KEY; } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { found_type = BTRFS_CSUM_ITEM_KEY; + } else if (found_type == BTRFS_EXTENT_DATA_KEY) { + found_type = BTRFS_XATTR_ITEM_KEY; + } else if (found_type == BTRFS_XATTR_ITEM_KEY) { + found_type = BTRFS_INODE_REF_KEY; } else if (found_type) { found_type--; } else { break; } btrfs_set_key_type(&key, found_type); - btrfs_release_path(root, path); - continue; + goto next; } if (found_key.offset >= inode->i_size) del_item = 1; @@ -860,13 +844,21 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, } delete: if (del_item) { - ret = btrfs_del_item(trans, root, path); - if (ret) - goto error; + if (!pending_del_nr) { + /* no pending yet, add ourselves */ + pending_del_slot = path->slots[0]; + pending_del_nr = 1; + } else if (pending_del_nr && + path->slots[0] + 1 == pending_del_slot) { + /* hop on the pending chunk */ + pending_del_nr++; + pending_del_slot = path->slots[0]; + } else { + printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot); + } } else { break; } - btrfs_release_path(root, path); if (found_extent) { ret = btrfs_free_extent(trans, root, extent_start, extent_num_bytes, @@ -875,9 +867,36 @@ delete: found_key.offset, 0); BUG_ON(ret); } +next: + if (path->slots[0] == 0) { + if (pending_del_nr) + goto del_pending; + btrfs_release_path(root, path); + goto search_again; + } + + path->slots[0]--; + if (pending_del_nr && + path->slots[0] + 1 != pending_del_slot) { + struct btrfs_key debug; +del_pending: + btrfs_item_key_to_cpu(path->nodes[0], &debug, + pending_del_slot); + ret = btrfs_del_items(trans, root, path, + pending_del_slot, + pending_del_nr); + BUG_ON(ret); + pending_del_nr = 0; + btrfs_release_path(root, path); + goto search_again; + } } ret = 0; error: + if (pending_del_nr) { + ret = btrfs_del_items(trans, root, path, pending_del_slot, + pending_del_nr); + } btrfs_release_path(root, path); btrfs_free_path(path); inode->i_sb->s_dirt = 1; @@ -1067,16 +1086,12 @@ void btrfs_delete_inode(struct inode *inode) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); - ret = btrfs_truncate_in_trans(trans, root, inode); - if (ret) - goto no_delete_lock; - ret = btrfs_delete_xattrs(trans, root, inode); - if (ret) - goto no_delete_lock; - ret = btrfs_free_inode(trans, root, inode); + ret = btrfs_truncate_in_trans(trans, root, inode, 0); if (ret) goto no_delete_lock; + nr = trans->blocks_used; + clear_inode(inode); btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); @@ -2190,7 +2205,8 @@ static void btrfs_truncate(struct inode *inode) btrfs_set_trans_block_group(trans, inode); /* FIXME, add redo link to tree so we don't leak on crash */ - ret = btrfs_truncate_in_trans(trans, root, inode); + ret = btrfs_truncate_in_trans(trans, root, inode, + BTRFS_EXTENT_DATA_KEY); btrfs_update_inode(trans, root, inode); nr = trans->blocks_used; -- cgit v1.2.3-70-g09d2 From 9c58309d6cf22471dacbcb6de54d00cef9ca20d4 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 15:15:18 -0500 Subject: Btrfs: Add inode item and backref in one insert, reducing cpu usage Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 39 ++++++++++++++++---------- fs/btrfs/ctree.h | 17 ++++++++++-- fs/btrfs/inode.c | 85 +++++++++++++++++++++++++++++++++++++++++--------------- 3 files changed, 102 insertions(+), 39 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 84ad53e06b3..fb2e2bd506c 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2332,27 +2332,34 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, * Given a key and some data, insert an item into the tree. * This does all the path init required, making room in the tree if needed. */ -int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, +int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, - struct btrfs_key *cpu_key, u32 data_size) + struct btrfs_key *cpu_key, u32 *data_size, + int nr) { struct extent_buffer *leaf; struct btrfs_item *item; int ret = 0; int slot; int slot_orig; + int i; u32 nritems; + u32 total_size = 0; + u32 total_data = 0; unsigned int data_end; struct btrfs_disk_key disk_key; - btrfs_cpu_key_to_disk(&disk_key, cpu_key); + for (i = 0; i < nr; i++) { + total_data += data_size[i]; + } /* create a root if there isn't one */ if (!root->node) BUG(); - ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1); + total_size = total_data + (nr - 1) * sizeof(struct btrfs_item); + ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); if (ret == 0) { return -EEXIST; } @@ -2366,10 +2373,10 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, data_end = leaf_data_end(root, leaf); if (btrfs_leaf_free_space(root, leaf) < - sizeof(struct btrfs_item) + data_size) { + sizeof(struct btrfs_item) + total_size) { btrfs_print_leaf(root, leaf); printk("not enough freespace need %u have %d\n", - data_size, btrfs_leaf_free_space(root, leaf)); + total_size, btrfs_leaf_free_space(root, leaf)); BUG(); } @@ -2404,7 +2411,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, } ioff = btrfs_item_offset(leaf, item); - btrfs_set_item_offset(leaf, item, ioff - data_size); + btrfs_set_item_offset(leaf, item, ioff - total_data); } if (leaf->map_token) { unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); @@ -2412,23 +2419,27 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, } /* shift the items */ - memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1), + memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr), btrfs_item_nr_offset(slot), (nritems - slot) * sizeof(struct btrfs_item)); /* shift the data */ memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) + - data_end - data_size, btrfs_leaf_data(leaf) + + data_end - total_data, btrfs_leaf_data(leaf) + data_end, old_data - data_end); data_end = old_data; } /* setup the item for the new data */ - btrfs_set_item_key(leaf, &disk_key, slot); - item = btrfs_item_nr(leaf, slot); - btrfs_set_item_offset(leaf, item, data_end - data_size); - btrfs_set_item_size(leaf, item, data_size); - btrfs_set_header_nritems(leaf, nritems + 1); + for (i = 0; i < nr; i++) { + btrfs_cpu_key_to_disk(&disk_key, cpu_key + i); + btrfs_set_item_key(leaf, &disk_key, slot + i); + item = btrfs_item_nr(leaf, slot + i); + btrfs_set_item_offset(leaf, item, data_end - data_size[i]); + data_end -= data_size[i]; + btrfs_set_item_size(leaf, item, data_size[i]); + } + btrfs_set_header_nritems(leaf, nritems + nr); btrfs_mark_buffer_dirty(leaf); ret = 0; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 098cf088315..127c86f795d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1050,9 +1050,20 @@ static inline int btrfs_del_item(struct btrfs_trans_handle *trans, int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, void *data, u32 data_size); -int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, struct btrfs_key - *cpu_key, u32 data_size); +int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct btrfs_key *cpu_key, u32 *data_size, int nr); + +static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct btrfs_key *key, + u32 data_size) +{ + return btrfs_insert_empty_items(trans, root, path, key, &data_size, 1); +} + int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path); int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path); int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0a2fe51c412..413b1012de5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1500,6 +1500,8 @@ void btrfs_dirty_inode(struct inode *inode) static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, + const char *name, int name_len, + u64 ref_objectid, u64 objectid, struct btrfs_block_group_cache *group, int mode) @@ -1508,6 +1510,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, struct btrfs_inode_item *inode_item; struct btrfs_key *location; struct btrfs_path *path; + struct btrfs_inode_ref *ref; + struct btrfs_key key[2]; + u32 sizes[2]; + unsigned long ptr; int ret; int owner; @@ -1530,10 +1536,25 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, group = btrfs_find_block_group(root, group, 0, 0, owner); BTRFS_I(inode)->block_group = group; BTRFS_I(inode)->flags = 0; - ret = btrfs_insert_empty_inode(trans, root, path, objectid); - if (ret) + + key[0].objectid = objectid; + btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); + key[0].offset = 0; + + key[1].objectid = objectid; + btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY); + key[1].offset = ref_objectid; + + sizes[0] = sizeof(struct btrfs_inode_item); + sizes[1] = name_len + sizeof(*ref); + + ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2); + if (ret != 0) goto fail; + if (objectid > root->highest_inode) + root->highest_inode = objectid; + inode->i_uid = current->fsuid; inode->i_gid = current->fsgid; inode->i_mode = mode; @@ -1543,6 +1564,13 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_inode_item); fill_inode_item(path->nodes[0], inode_item, inode); + + ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, + struct btrfs_inode_ref); + btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + ptr = (unsigned long)(ref + 1); + write_extent_buffer(path->nodes[0], name, ptr, name_len); + btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_free_path(path); @@ -1564,7 +1592,8 @@ static inline u8 btrfs_inode_type(struct inode *inode) } static int btrfs_add_link(struct btrfs_trans_handle *trans, - struct dentry *dentry, struct inode *inode) + struct dentry *dentry, struct inode *inode, + int add_backref) { int ret; struct btrfs_key key; @@ -1580,11 +1609,13 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_parent->d_inode->i_ino, &key, btrfs_inode_type(inode)); if (ret == 0) { - ret = btrfs_insert_inode_ref(trans, root, - dentry->d_name.name, - dentry->d_name.len, - inode->i_ino, - dentry->d_parent->d_inode->i_ino); + if (add_backref) { + ret = btrfs_insert_inode_ref(trans, root, + dentry->d_name.name, + dentry->d_name.len, + inode->i_ino, + dentry->d_parent->d_inode->i_ino); + } parent_inode = dentry->d_parent->d_inode; parent_inode->i_size += dentry->d_name.len * 2; parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; @@ -1595,9 +1626,10 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, } static int btrfs_add_nondir(struct btrfs_trans_handle *trans, - struct dentry *dentry, struct inode *inode) + struct dentry *dentry, struct inode *inode, + int backref) { - int err = btrfs_add_link(trans, dentry, inode); + int err = btrfs_add_link(trans, dentry, inode, backref); if (!err) { d_instantiate(dentry, inode); return 0; @@ -1635,14 +1667,16 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, objectid, + inode = btrfs_new_inode(trans, root, dentry->d_name.name, + dentry->d_name.len, + dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, mode); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode); + err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) drop_inode = 1; else { @@ -1692,14 +1726,16 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, objectid, - BTRFS_I(dir)->block_group, mode); + inode = btrfs_new_inode(trans, root, dentry->d_name.name, + dentry->d_name.len, + dentry->d_parent->d_inode->i_ino, + objectid, BTRFS_I(dir)->block_group, mode); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode); + err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) drop_inode = 1; else { @@ -1755,7 +1791,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, btrfs_set_trans_block_group(trans, dir); atomic_inc(&inode->i_count); - err = btrfs_add_nondir(trans, dentry, inode); + err = btrfs_add_nondir(trans, dentry, inode, 1); if (err) drop_inode = 1; @@ -1810,7 +1846,9 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) goto out_unlock; } - inode = btrfs_new_inode(trans, root, objectid, + inode = btrfs_new_inode(trans, root, dentry->d_name.name, + dentry->d_name.len, + dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, S_IFDIR | mode); if (IS_ERR(inode)) { err = PTR_ERR(inode); @@ -1827,7 +1865,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) if (err) goto out_fail; - err = btrfs_add_link(trans, dentry, inode); + err = btrfs_add_link(trans, dentry, inode, 0); if (err) goto out_fail; @@ -2320,7 +2358,8 @@ static int noinline create_subvol(struct btrfs_root *root, char *name, trans = btrfs_start_transaction(new_root, 1); BUG_ON(!trans); - inode = btrfs_new_inode(trans, new_root, new_dirid, + inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid, + new_dirid, BTRFS_I(dir)->block_group, S_IFDIR | 0700); if (IS_ERR(inode)) goto fail; @@ -2817,7 +2856,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, if (ret) goto out_fail; } - ret = btrfs_add_link(trans, new_dentry, old_inode); + ret = btrfs_add_link(trans, new_dentry, old_inode, 1); if (ret) goto out_fail; @@ -2865,14 +2904,16 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, objectid, + inode = btrfs_new_inode(trans, root, dentry->d_name.name, + dentry->d_name.len, + dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode); + err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) drop_inode = 1; else { -- cgit v1.2.3-70-g09d2 From 291d673e6a22d9c6834e939f66c7cfef90669021 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jan 2008 15:55:23 -0500 Subject: Btrfs: Do delalloc accounting via hooks in the extent_state code Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 6 ----- fs/btrfs/extent_io.c | 25 +++++++++++++++++++++ fs/btrfs/extent_io.h | 4 ++++ fs/btrfs/file.c | 4 ---- fs/btrfs/inode.c | 60 ++++++++++++++++++++------------------------------ 5 files changed, 53 insertions(+), 46 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c4fe37791d3..90718f77bbc 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2180,7 +2180,6 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, unsigned long last_index; unsigned long i; struct page *page; - struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct file_ra_state *ra; @@ -2220,11 +2219,6 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); - spin_lock(&root->fs_info->delalloc_lock); - root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - - existing_delalloc; - spin_unlock(&root->fs_info->delalloc_lock); - unlock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); unlock_page(page); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8aec72253a1..e81b35b0618 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -256,6 +256,26 @@ static int merge_state(struct extent_io_tree *tree, return 0; } +static void set_state_cb(struct extent_io_tree *tree, + struct extent_state *state, + unsigned long bits) +{ + if (tree->ops && tree->ops->set_bit_hook) { + tree->ops->set_bit_hook(tree->mapping->host, state->start, + state->end, bits); + } +} + +static void clear_state_cb(struct extent_io_tree *tree, + struct extent_state *state, + unsigned long bits) +{ + if (tree->ops && tree->ops->set_bit_hook) { + tree->ops->clear_bit_hook(tree->mapping->host, state->start, + state->end, bits); + } +} + /* * insert an extent_state struct into the tree. 'bits' are set on the * struct before it is inserted. @@ -281,6 +301,7 @@ static int insert_state(struct extent_io_tree *tree, state->state |= bits; state->start = start; state->end = end; + set_state_cb(tree, state, bits); node = tree_insert(&tree->state, end, &state->rb_node); if (node) { struct extent_state *found; @@ -349,6 +370,7 @@ static int clear_state_bit(struct extent_io_tree *tree, tree->dirty_bytes -= range; } state->state &= ~bits; + clear_state_cb(tree, state, bits); if (wake) wake_up(&state->wq); if (delete || state->state == 0) { @@ -553,6 +575,7 @@ static void set_state_bits(struct extent_io_tree *tree, tree->dirty_bytes += range; } state->state |= bits; + set_state_cb(tree, state, bits); } /* @@ -975,6 +998,7 @@ search_again: goto search_again; } state->state |= EXTENT_LOCKED; + set_state_cb(tree, state, EXTENT_LOCKED); if (!found) *start = state->start; found++; @@ -1474,6 +1498,7 @@ static int end_bio_extent_readpage(struct bio *bio, state = NULL; } clear->state |= EXTENT_UPTODATE; + set_state_cb(tree, clear, EXTENT_UPTODATE); clear_state_bit(tree, clear, EXTENT_LOCKED, 1, 0); if (cur == start) diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index f1e8bf251f3..a96c5a14134 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -33,6 +33,10 @@ struct extent_io_ops { struct extent_state *state); void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); + int (*set_bit_hook)(struct inode *inode, u64 start, u64 end, + unsigned long bits); + int (*clear_bit_hook)(struct inode *inode, u64 start, u64 end, + unsigned long bits); }; struct extent_io_tree { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index c5bb00f9239..8e210616d70 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -323,10 +323,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, } set_extent_delalloc(io_tree, start_pos, end_of_last_block, GFP_NOFS); - spin_lock(&root->fs_info->delalloc_lock); - root->fs_info->delalloc_bytes += (end_of_last_block + 1 - - start_pos) - existing_delalloc; - spin_unlock(&root->fs_info->delalloc_lock); btrfs_add_ordered_inode(inode); } else { u64 aligned_end; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 413b1012de5..5a38443a24e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -80,8 +80,6 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, u64 thresh; int ret = 0; - return 0; - if (for_del) thresh = total * 90; else @@ -249,7 +247,6 @@ not_found: static int run_delalloc_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; - u64 num_bytes; int ret; mutex_lock(&root->fs_info->fs_mutex); if (btrfs_test_opt(root, NODATACOW) || @@ -258,20 +255,34 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) else ret = cow_file_range(inode, start, end); - spin_lock(&root->fs_info->delalloc_lock); - num_bytes = end + 1 - start; - if (root->fs_info->delalloc_bytes < num_bytes) { - printk("delalloc accounting error total %llu sub %llu\n", - root->fs_info->delalloc_bytes, num_bytes); - } else { - root->fs_info->delalloc_bytes -= num_bytes; - } - spin_unlock(&root->fs_info->delalloc_lock); - mutex_unlock(&root->fs_info->fs_mutex); return ret; } +int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, + unsigned long bits) +{ + if ((bits & EXTENT_DELALLOC)) { + struct btrfs_root *root = BTRFS_I(inode)->root; + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes += end - start + 1; + spin_unlock(&root->fs_info->delalloc_lock); + } + return 0; +} + +int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, + unsigned long bits) +{ + if ((bits & EXTENT_DELALLOC)) { + struct btrfs_root *root = BTRFS_I(inode)->root; + spin_lock(&root->fs_info->delalloc_lock); + root->fs_info->delalloc_bytes -= end - start + 1; + spin_unlock(&root->fs_info->delalloc_lock); + } + return 0; +} + int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) { struct inode *inode = page->mapping->host; @@ -908,28 +919,17 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page, { char *kaddr; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct btrfs_root *root = BTRFS_I(inode)->root; u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; u64 page_end = page_start + PAGE_CACHE_SIZE - 1; - u64 existing_delalloc; - u64 delalloc_start; int ret = 0; WARN_ON(!PageLocked(page)); set_page_extent_mapped(page); lock_extent(io_tree, page_start, page_end, GFP_NOFS); - delalloc_start = page_start; - existing_delalloc = count_range_bits(&BTRFS_I(inode)->io_tree, - &delalloc_start, page_end, - PAGE_CACHE_SIZE, EXTENT_DELALLOC); set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); - spin_lock(&root->fs_info->delalloc_lock); - root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - existing_delalloc; - spin_unlock(&root->fs_info->delalloc_lock); - if (zero_start != PAGE_CACHE_SIZE) { kaddr = kmap(page); memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); @@ -2456,8 +2456,6 @@ int btrfs_defrag_file(struct file *file) { unsigned long ra_index = 0; u64 page_start; u64 page_end; - u64 delalloc_start; - u64 existing_delalloc; unsigned long i; int ret; @@ -2491,19 +2489,9 @@ int btrfs_defrag_file(struct file *file) { page_end = page_start + PAGE_CACHE_SIZE - 1; lock_extent(io_tree, page_start, page_end, GFP_NOFS); - delalloc_start = page_start; - existing_delalloc = - count_range_bits(&BTRFS_I(inode)->io_tree, - &delalloc_start, page_end, - PAGE_CACHE_SIZE, EXTENT_DELALLOC); set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); - spin_lock(&root->fs_info->delalloc_lock); - root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - - existing_delalloc; - spin_unlock(&root->fs_info->delalloc_lock); - unlock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); unlock_page(page); -- cgit v1.2.3-70-g09d2 From b1632b10c0133992fd0290b4836c47661bed4396 Mon Sep 17 00:00:00 2001 From: Yan Date: Wed, 30 Jan 2008 11:54:04 -0500 Subject: Btrfs: Align extent length to sectorsize in --- Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5a38443a24e..de17c61dfa0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -823,6 +823,8 @@ search_again: btrfs_file_extent_num_bytes(leaf, fi); extent_num_bytes = inode->i_size - found_key.offset + root->sectorsize - 1; + extent_num_bytes = extent_num_bytes & + ~((u64)root->sectorsize - 1); btrfs_set_file_extent_num_bytes(leaf, fi, extent_num_bytes); num_dec = (orig_num_bytes - -- cgit v1.2.3-70-g09d2 From f392a938f3cc36a2abe68db9286bd864c3b2f7dd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 30 Jan 2008 11:54:05 -0500 Subject: Properly align the hole size in btrfs_setattr Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index de17c61dfa0..515f8b01ab0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1001,7 +1001,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) u64 mask = root->sectorsize - 1; u64 pos = (inode->i_size + mask) & ~mask; - u64 block_end = attr->ia_size | mask; + u64 block_end = (attr->ia_size + mask) & ~mask; u64 hole_start; u64 hole_size; u64 alloc_hint = 0; @@ -1022,7 +1022,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); - lock_extent(io_tree, pos, block_end, GFP_NOFS); + lock_extent(io_tree, pos, block_end - 1, GFP_NOFS); hole_size = block_end - hole_start; mutex_lock(&root->fs_info->fs_mutex); @@ -1043,7 +1043,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) } btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - unlock_extent(io_tree, pos, block_end, GFP_NOFS); + unlock_extent(io_tree, pos, block_end - 1, GFP_NOFS); if (err) return err; } -- cgit v1.2.3-70-g09d2 From 1b0f7c29e2f4f41e1367e7581d3a9c8c70f0394a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 30 Jan 2008 14:33:02 -0500 Subject: Fix hole start calculation in btrfs_settar Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 515f8b01ab0..bcf3b35fb65 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1000,20 +1000,14 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; u64 mask = root->sectorsize - 1; - u64 pos = (inode->i_size + mask) & ~mask; + u64 hole_start = (inode->i_size + mask) & ~mask; u64 block_end = (attr->ia_size + mask) & ~mask; - u64 hole_start; u64 hole_size; u64 alloc_hint = 0; - if (attr->ia_size <= pos) + if (attr->ia_size <= hole_start) goto out; - if (pos != inode->i_size) - hole_start = pos + root->sectorsize; - else - hole_start = pos; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); mutex_unlock(&root->fs_info->fs_mutex); @@ -1022,14 +1016,14 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); - lock_extent(io_tree, pos, block_end - 1, GFP_NOFS); + lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); hole_size = block_end - hole_start; mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); err = btrfs_drop_extents(trans, root, inode, - pos, block_end, pos, + hole_start, block_end, hole_start, &alloc_hint); if (alloc_hint != EXTENT_MAP_INLINE) { @@ -1043,7 +1037,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) } btrfs_end_transaction(trans, root); mutex_unlock(&root->fs_info->fs_mutex); - unlock_extent(io_tree, pos, block_end - 1, GFP_NOFS); + unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); if (err) return err; } -- cgit v1.2.3-70-g09d2 From b0c68f8bed058d9f2023b067b16ed06a8c439544 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 31 Jan 2008 11:05:37 -0500 Subject: Btrfs: Enable delalloc accounting Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 4 ++++ fs/btrfs/extent_io.c | 14 +++++++------- fs/btrfs/extent_io.h | 4 ++-- fs/btrfs/inode.c | 18 +++++++++++++----- 4 files changed, 26 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 6647b90ef74..ad72e219201 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -861,6 +861,10 @@ int close_ctree(struct btrfs_root *root) write_ctree_super(NULL, root); mutex_unlock(&fs_info->fs_mutex); + if (fs_info->delalloc_bytes) { + printk("btrfs: at unmount delalloc count %Lu\n", + fs_info->delalloc_bytes); + } if (fs_info->extent_root->node) free_extent_buffer(fs_info->extent_root->node); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 6999b819db6..624aabc3e6f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -262,7 +262,7 @@ static void set_state_cb(struct extent_io_tree *tree, { if (tree->ops && tree->ops->set_bit_hook) { tree->ops->set_bit_hook(tree->mapping->host, state->start, - state->end, bits); + state->end, state->state, bits); } } @@ -272,7 +272,7 @@ static void clear_state_cb(struct extent_io_tree *tree, { if (tree->ops && tree->ops->set_bit_hook) { tree->ops->clear_bit_hook(tree->mapping->host, state->start, - state->end, bits); + state->end, state->state, bits); } } @@ -298,10 +298,10 @@ static int insert_state(struct extent_io_tree *tree, } if (bits & EXTENT_DIRTY) tree->dirty_bytes += end - start + 1; + set_state_cb(tree, state, bits); state->state |= bits; state->start = start; state->end = end; - set_state_cb(tree, state, bits); node = tree_insert(&tree->state, end, &state->rb_node); if (node) { struct extent_state *found; @@ -369,8 +369,8 @@ static int clear_state_bit(struct extent_io_tree *tree, WARN_ON(range > tree->dirty_bytes); tree->dirty_bytes -= range; } - state->state &= ~bits; clear_state_cb(tree, state, bits); + state->state &= ~bits; if (wake) wake_up(&state->wq); if (delete || state->state == 0) { @@ -574,8 +574,8 @@ static void set_state_bits(struct extent_io_tree *tree, u64 range = state->end - state->start + 1; tree->dirty_bytes += range; } - state->state |= bits; set_state_cb(tree, state, bits); + state->state |= bits; } /* @@ -997,8 +997,8 @@ search_again: free_extent_state(state); goto search_again; } - state->state |= EXTENT_LOCKED; set_state_cb(tree, state, EXTENT_LOCKED); + state->state |= EXTENT_LOCKED; if (!found) *start = state->start; found++; @@ -1497,8 +1497,8 @@ static int end_bio_extent_readpage(struct bio *bio, } else { state = NULL; } - clear->state |= EXTENT_UPTODATE; set_state_cb(tree, clear, EXTENT_UPTODATE); + clear->state |= EXTENT_UPTODATE; clear_state_bit(tree, clear, EXTENT_LOCKED, 1, 0); if (cur == start) diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index a96c5a14134..6fd5e2c9061 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -34,9 +34,9 @@ struct extent_io_ops { void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); int (*set_bit_hook)(struct inode *inode, u64 start, u64 end, - unsigned long bits); + unsigned long old, unsigned long bits); int (*clear_bit_hook)(struct inode *inode, u64 start, u64 end, - unsigned long bits); + unsigned long old, unsigned long bits); }; struct extent_io_tree { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bcf3b35fb65..f13b1dbc061 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -260,9 +260,9 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) } int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, - unsigned long bits) + unsigned long old, unsigned long bits) { - if ((bits & EXTENT_DELALLOC)) { + if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; spin_lock(&root->fs_info->delalloc_lock); root->fs_info->delalloc_bytes += end - start + 1; @@ -272,12 +272,18 @@ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, } int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, - unsigned long bits) + unsigned long old, unsigned long bits) { - if ((bits & EXTENT_DELALLOC)) { + if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; spin_lock(&root->fs_info->delalloc_lock); - root->fs_info->delalloc_bytes -= end - start + 1; + if (end - start + 1 > root->fs_info->delalloc_bytes) { + printk("warning: delalloc account %Lu %Lu\n", + end - start + 1, root->fs_info->delalloc_bytes); + root->fs_info->delalloc_bytes = 0; + } else { + root->fs_info->delalloc_bytes -= end - start + 1; + } spin_unlock(&root->fs_info->delalloc_lock); } return 0; @@ -3002,6 +3008,8 @@ static struct extent_io_ops btrfs_extent_io_ops = { .writepage_io_hook = btrfs_writepage_io_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, + .set_bit_hook = btrfs_set_bit_hook, + .clear_bit_hook = btrfs_clear_bit_hook, }; static struct address_space_operations btrfs_aops = { -- cgit v1.2.3-70-g09d2 From c2e639f02cff721172569add21ba5d058fd1a81c Mon Sep 17 00:00:00 2001 From: Yan Date: Mon, 4 Feb 2008 08:57:25 -0500 Subject: Btrfs: Fix typo in extent_io.c --- Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 4 ++-- fs/btrfs/inode.c | 3 +-- 2 files changed, 3 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5d2fb527fd9..36a943e5180 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1416,7 +1416,7 @@ static int end_bio_extent_writepage(struct bio *bio, * to the correct offsets in the file */ if (state && (state->end + 1 != start || - !state->state & EXTENT_WRITEBACK)) { + !(state->state & EXTENT_WRITEBACK))) { state = NULL; } spin_unlock_irqrestore(&tree->lock, flags); @@ -1539,7 +1539,7 @@ static int end_bio_extent_readpage(struct bio *bio, * to the correct offsets in the file */ if (state && (state->end + 1 != start || - !state->state & EXTENT_WRITEBACK)) { + !(state->state & EXTENT_LOCKED))) { state = NULL; } spin_unlock_irqrestore(&tree->lock, flags); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f13b1dbc061..44fc94fee2c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -364,8 +364,7 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, if (btrfs_test_opt(root, NODATASUM) || btrfs_test_flag(inode, NODATASUM)) return 0; - - if (state->start == start) { + if (state && state->start == start) { private = state->private; ret = 0; } else { -- cgit v1.2.3-70-g09d2 From 9069218d448ea547dbad5f1cbd537e88d6519d66 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 8 Feb 2008 13:49:28 -0500 Subject: Btrfs: Fix i_blocks accounting Now that delayed allocation accounting works, i_blocks accounting is changed to only modify i_blocks when extents inserted or removed. The fillattr call is changed to include the delayed allocation byte count in the i_blocks result. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/ctree.h | 9 +++++++++ fs/btrfs/file.c | 17 +++++++++++------ fs/btrfs/inode.c | 43 +++++++++++++++++++++++++++++-------------- 4 files changed, 50 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 9a0647dc5a0..9b9db9cbc01 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -36,6 +36,7 @@ struct btrfs_inode { * transid of the trans_handle that last modified this inode */ u64 last_trans; + u64 delalloc_bytes; u32 flags; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index dfb700efcd6..52144b04f1a 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1161,6 +1161,15 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +static inline void dec_i_blocks(struct inode *inode, u64 dec) +{ + dec = dec >> 9; + if (dec <= inode->i_blocks) + inode->i_blocks -= dec; + else + inode->i_blocks = 0; +} + unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, pgoff_t offset, pgoff_t last_index); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index bfa4149c053..f8939608254 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -175,6 +175,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, leaf = path->nodes[0]; ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); + inode->i_blocks += (offset + size - found_end) >> 9; } if (found_end < offset) { ptr = btrfs_file_extent_inline_start(ei) + found_size; @@ -184,6 +185,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, insert: btrfs_release_path(root, path); datasize = offset + size - key.offset; + inode->i_blocks += datasize >> 9; datasize = btrfs_file_extent_calc_inline_size(datasize); ret = btrfs_insert_empty_item(trans, root, path, &key, datasize); @@ -256,7 +258,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, goto out_unlock; } btrfs_set_trans_block_group(trans, inode); - inode->i_blocks += num_bytes >> 9; hint_byte = 0; if ((end_of_last_block & 4095) == 0) { @@ -410,7 +411,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) if (found_key.type != BTRFS_EXTENT_DATA_KEY) goto out; - if (found_key.offset != last_offset) { + if (found_key.offset < last_offset) { WARN_ON(1); btrfs_print_leaf(root, leaf); printk("inode %lu found offset %Lu expected %Lu\n", @@ -435,7 +436,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) last_offset = extent_end; path->slots[0]++; } - if (last_offset < inode->i_size) { + if (0 && last_offset < inode->i_size) { WARN_ON(1); btrfs_print_leaf(root, leaf); printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, @@ -608,8 +609,7 @@ next_slot: extent); if (btrfs_file_extent_disk_bytenr(leaf, extent)) { - inode->i_blocks -= - (old_num - new_num) >> 9; + dec_i_blocks(inode, old_num - new_num); } btrfs_set_file_extent_num_bytes(leaf, extent, new_num); @@ -620,6 +620,8 @@ next_slot: u32 new_size; new_size = btrfs_file_extent_calc_inline_size( inline_limit - key.offset); + dec_i_blocks(inode, (extent_end - key.offset) - + (inline_limit - key.offset)); btrfs_truncate_item(trans, root, path, new_size, 1); } @@ -653,7 +655,7 @@ next_slot: btrfs_release_path(root, path); extent = NULL; if (found_extent && disk_bytenr != 0) { - inode->i_blocks -= extent_num_bytes >> 9; + dec_i_blocks(inode, extent_num_bytes); ret = btrfs_free_extent(trans, root, disk_bytenr, disk_num_bytes, @@ -674,6 +676,8 @@ next_slot: u32 new_size; new_size = btrfs_file_extent_calc_inline_size( extent_end - end); + dec_i_blocks(inode, (extent_end - key.offset) - + (extent_end - end)); btrfs_truncate_item(trans, root, path, new_size, 0); } /* create bookend, splitting the extent in two */ @@ -718,6 +722,7 @@ next_slot: } out: btrfs_free_path(path); + btrfs_check_file(root, inode); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 44fc94fee2c..913ab128eee 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -134,6 +134,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) ret = btrfs_insert_file_extent(trans, root, inode->i_ino, start, ins.objectid, ins.offset, ins.offset); + inode->i_blocks += ins.offset >> 9; btrfs_check_file(root, inode); num_bytes -= cur_alloc_size; alloc_hint = ins.objectid + ins.offset; @@ -142,6 +143,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) btrfs_drop_extent_cache(inode, orig_start, orig_start + orig_num_bytes - 1); btrfs_add_ordered_inode(inode); + btrfs_update_inode(trans, root, inode); out: btrfs_end_transaction(trans, root); return ret; @@ -265,6 +267,7 @@ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; spin_lock(&root->fs_info->delalloc_lock); + BTRFS_I(inode)->delalloc_bytes += end - start + 1; root->fs_info->delalloc_bytes += end - start + 1; spin_unlock(&root->fs_info->delalloc_lock); } @@ -281,8 +284,10 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, printk("warning: delalloc account %Lu %Lu\n", end - start + 1, root->fs_info->delalloc_bytes); root->fs_info->delalloc_bytes = 0; + BTRFS_I(inode)->delalloc_bytes = 0; } else { root->fs_info->delalloc_bytes -= end - start + 1; + BTRFS_I(inode)->delalloc_bytes -= end - start + 1; } spin_unlock(&root->fs_info->delalloc_lock); } @@ -833,32 +838,37 @@ search_again: btrfs_set_file_extent_num_bytes(leaf, fi, extent_num_bytes); num_dec = (orig_num_bytes - - extent_num_bytes) >> 9; - if (extent_start != 0) { - inode->i_blocks -= num_dec; - } + extent_num_bytes); + if (extent_start != 0) + dec_i_blocks(inode, num_dec); btrfs_mark_buffer_dirty(leaf); } else { extent_num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi); /* FIXME blocksize != 4096 */ - num_dec = btrfs_file_extent_num_bytes(leaf, - fi) >> 9; + num_dec = btrfs_file_extent_num_bytes(leaf, fi); if (extent_start != 0) { found_extent = 1; - inode->i_blocks -= num_dec; + dec_i_blocks(inode, num_dec); } root_gen = btrfs_header_generation(leaf); root_owner = btrfs_header_owner(leaf); } - } else if (extent_type == BTRFS_FILE_EXTENT_INLINE && - !del_item) { - u32 newsize = inode->i_size - found_key.offset; - newsize = btrfs_file_extent_calc_inline_size(newsize); - ret = btrfs_truncate_item(trans, root, path, - newsize, 1); - BUG_ON(ret); + } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { + if (!del_item) { + u32 newsize = inode->i_size - found_key.offset; + dec_i_blocks(inode, item_end + 1 - + found_key.offset - newsize); + newsize = + btrfs_file_extent_calc_inline_size(newsize); + ret = btrfs_truncate_item(trans, root, path, + newsize, 1); + BUG_ON(ret); + } else { + dec_i_blocks(inode, item_end + 1 - + found_key.offset); + } } delete: if (del_item) { @@ -1222,6 +1232,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) struct btrfs_iget_args *args = p; inode->i_ino = args->ino; BTRFS_I(inode)->root = args->root; + BTRFS_I(inode)->delalloc_bytes = 0; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); @@ -1528,6 +1539,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->root = root; if (mode & S_IFDIR) @@ -1746,6 +1758,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -2797,6 +2810,7 @@ static int btrfs_getattr(struct vfsmount *mnt, struct inode *inode = dentry->d_inode; generic_fillattr(inode, stat); stat->blksize = PAGE_CACHE_SIZE; + stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9); return 0; } @@ -2912,6 +2926,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; -- cgit v1.2.3-70-g09d2 From 5e591a0703bc03047d37938586f97fd8ca960c08 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Tue, 19 Feb 2008 11:41:02 -0500 Subject: Btrfs: Fix looping on readdir of the subvol roots Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 913ab128eee..3e79572a5cb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1465,7 +1465,10 @@ read_dir_items: di = (struct btrfs_dir_item *)((char *)di + di_len); } } - filp->f_pos = INT_LIMIT(typeof(filp->f_pos)); + if (key_type == BTRFS_DIR_INDEX_KEY) + filp->f_pos = INT_LIMIT(typeof(filp->f_pos)); + else + filp->f_pos++; nopos: ret = 0; err: -- cgit v1.2.3-70-g09d2 From 065631f6dccea07bfad48d8981369f6d9cfd6e2b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Feb 2008 12:07:25 -0500 Subject: Btrfs: checksum file data at bio submission time instead of during writepage When we checkum file data during writepage, the checksumming is done one page at a time, making it difficult to do bulk metadata modifications to insert checksums for large ranges of the file at once. This patch changes btrfs to checksum on a per-bio basis instead. The bios are checksummed before they are handed off to the block layer, so each bio is contiguous and only has pages from the same inode. Checksumming on a bio basis allows us to insert and modify the file checksum items in large groups. It also allows the checksumming to be done more easily by async worker threads. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 8 +++----- fs/btrfs/extent_io.c | 2 ++ fs/btrfs/extent_io.h | 1 + fs/btrfs/file-item.c | 50 +++++++++++++++++++++++++++++++++++++++----------- fs/btrfs/inode.c | 32 ++++++++++++++++++++++++++++++-- 5 files changed, 75 insertions(+), 18 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 086cb0525cd..92d892f9207 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1142,11 +1142,9 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid, u64 bytenr, int mod); -int btrfs_csum_file_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode, - u64 objectid, u64 offset, - char *data, size_t len); +int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + struct bio *bio); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index fb6400895ed..e8130c87633 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1705,6 +1705,8 @@ static int submit_one_bio(int rw, struct bio *bio) (unsigned long long)bio->bi_sector); WARN_ON(1); } + if (tree->ops && tree->ops->submit_bio_hook) + tree->ops->submit_bio_hook(rw, bio); submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index fcc4bb078c2..9d665466708 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -28,6 +28,7 @@ struct extent_state; struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); + int (*submit_bio_hook)(int rw, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 3ebbc058d08..3f0e71b0e5d 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -16,6 +16,9 @@ * Boston, MA 021110-1307, USA. */ +#include +#include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -131,28 +134,35 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } -int btrfs_csum_file_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode, - u64 objectid, u64 offset, - char *data, size_t len) +int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + struct bio *bio) { + u64 objectid = inode->i_ino; + u64 offset; int ret; struct btrfs_key file_key; struct btrfs_key found_key; - u64 next_offset = (u64)-1; - int found_next = 0; + u64 next_offset; + int found_next; struct btrfs_path *path; struct btrfs_csum_item *item; + struct btrfs_csum_item *item_end; struct extent_buffer *leaf = NULL; u64 csum_offset; - u32 csum_result = ~(u32)0; + u32 csum_result; u32 nritems; u32 ins_size; + int bio_index = 0; + struct bio_vec *bvec = bio->bi_io_vec; + char *data; path = btrfs_alloc_path(); BUG_ON(!path); - +again: + next_offset = (u64)-1; + found_next = 0; + offset = page_offset(bvec->bv_page) + bvec->bv_offset; file_key.objectid = objectid; file_key.offset = offset; btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); @@ -259,7 +269,15 @@ csum: item = (struct btrfs_csum_item *)((unsigned char *)item + csum_offset * BTRFS_CRC32_SIZE); found: - csum_result = btrfs_csum_data(root, data, csum_result, len); + item_end = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item); + item_end = (struct btrfs_csum_item *)((unsigned char *)item_end + + btrfs_item_size_nr(leaf, path->slots[0])); +next_bvec: + data = kmap_atomic(bvec->bv_page, KM_IRQ0); + csum_result = ~(u32)0; + csum_result = btrfs_csum_data(root, data + bvec->bv_offset, + csum_result, bvec->bv_len); + kunmap_atomic(data, KM_IRQ0); btrfs_csum_final(csum_result, (char *)&csum_result); if (csum_result == 0) { printk("csum result is 0 for inode %lu offset %Lu\n", inode->i_ino, offset); @@ -267,9 +285,19 @@ found: write_extent_buffer(leaf, &csum_result, (unsigned long)item, BTRFS_CRC32_SIZE); + bio_index++; + bvec++; + if (bio_index < bio->bi_vcnt) { + item = (struct btrfs_csum_item *)((char *)item + BTRFS_CRC32_SIZE); + if (item < item_end) + goto next_bvec; + } btrfs_mark_buffer_dirty(path->nodes[0]); + if (bio_index < bio->bi_vcnt) { + btrfs_release_path(root, path); + goto again; + } fail: - btrfs_release_path(root, path); btrfs_free_path(path); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3e79572a5cb..bbb71fad861 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -16,6 +16,7 @@ * Boston, MA 021110-1307, USA. */ +#include #include #include #include @@ -294,6 +295,32 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, return 0; } +int btrfs_submit_bio_hook(int rw, struct bio *bio) +{ + // struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; + struct bio_vec *bvec = bio->bi_io_vec; + struct inode *inode = bvec->bv_page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + if (rw != WRITE) + return 0; + + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) + return 0; + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + btrfs_csum_file_blocks(trans, root, inode, bio); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} +#if 0 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) { struct inode *inode = page->mapping->host; @@ -318,7 +345,7 @@ int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) mutex_unlock(&root->fs_info->fs_mutex); return ret; } - +#endif int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) { int ret = 0; @@ -3022,7 +3049,8 @@ static struct file_operations btrfs_dir_file_operations = { static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, - .writepage_io_hook = btrfs_writepage_io_hook, + // .writepage_io_hook = btrfs_writepage_io_hook, + .submit_bio_hook = btrfs_submit_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .set_bit_hook = btrfs_set_bit_hook, -- cgit v1.2.3-70-g09d2 From 6885f308b557020efce6ef9c1fcc96adaabecbb6 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Feb 2008 16:11:05 -0500 Subject: Btrfs: Misc 2.6.25 updates Remove the btrfs read_inode method, and use save_mount_options Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 29 +---------------------------- fs/btrfs/super.c | 11 ++++++++++- 2 files changed, 11 insertions(+), 29 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bbb71fad861..315dd550d56 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -297,7 +297,6 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, int btrfs_submit_bio_hook(int rw, struct bio *bio) { - // struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct bio_vec *bvec = bio->bi_io_vec; struct inode *inode = bvec->bv_page->mapping->host; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -320,32 +319,7 @@ int btrfs_submit_bio_hook(int rw, struct bio *bio) mutex_unlock(&root->fs_info->fs_mutex); return ret; } -#if 0 -int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end) -{ - struct inode *inode = page->mapping->host; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; - char *kaddr; - int ret = 0; - u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; - size_t offset = start - page_start; - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) - return 0; - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - kaddr = kmap(page); - btrfs_csum_file_block(trans, root, inode, inode->i_ino, - start, kaddr + offset, end - start + 1); - kunmap(page); - ret = btrfs_end_transaction(trans, root); - BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); - return ret; -} -#endif + int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) { int ret = 0; @@ -3049,7 +3023,6 @@ static struct file_operations btrfs_dir_file_operations = { static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, - // .writepage_io_hook = btrfs_writepage_io_hook, .submit_bio_hook = btrfs_submit_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index a46300c4753..4423a91206a 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -272,6 +272,11 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent) sb->s_root = root_dentry; btrfs_transaction_queue_work(tree_root, HZ * 30); + +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25) + save_mount_options(sb, data); +#endif + return 0; fail_close: @@ -462,9 +467,13 @@ static struct super_operations btrfs_super_ops = { .delete_inode = btrfs_delete_inode, .put_inode = btrfs_put_inode, .put_super = btrfs_put_super, - .read_inode = btrfs_read_locked_inode, .write_super = btrfs_write_super, .sync_fs = btrfs_sync_fs, +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25) + .read_inode = btrfs_read_locked_inode, +#else + .show_options = generic_show_options, +#endif .write_inode = btrfs_write_inode, .dirty_inode = btrfs_dirty_inode, .alloc_inode = btrfs_alloc_inode, -- cgit v1.2.3-70-g09d2 From 0b86a832a1f38abec695864ec2eaedc9d2383f1b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 24 Mar 2008 15:01:56 -0400 Subject: Btrfs: Add support for multiple devices per filesystem Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.c | 38 ++- fs/btrfs/ctree.h | 313 ++++++++++++++++-- fs/btrfs/disk-io.c | 97 +++++- fs/btrfs/disk-io.h | 2 + fs/btrfs/extent-tree.c | 429 +++++++------------------ fs/btrfs/extent_io.c | 6 +- fs/btrfs/extent_io.h | 2 +- fs/btrfs/inode.c | 26 +- fs/btrfs/print-tree.c | 48 +++ fs/btrfs/transaction.c | 50 ++- fs/btrfs/volumes.c | 852 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/volumes.h | 78 +++++ 13 files changed, 1563 insertions(+), 380 deletions(-) create mode 100644 fs/btrfs/volumes.c create mode 100644 fs/btrfs/volumes.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 9c61609f371..5ac6f2946c2 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ - extent_io.o + extent_io.o volumes.o ifeq ($(CONFIG_FS_POSIX_ACL),y) btrfs-y += acl.o diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index ef8809c1e69..70b6ddfe15a 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -70,6 +70,14 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) memset(p, 0, sizeof(*p)); } +static void add_root_to_dirty_list(struct btrfs_root *root) +{ + if (root->track_dirty && list_empty(&root->dirty_list)) { + list_add(&root->dirty_list, + &root->fs_info->dirty_cowonly_roots); + } +} + int btrfs_copy_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, @@ -196,6 +204,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, root_gen, 0, 0, 1); } free_extent_buffer(buf); + add_root_to_dirty_list(root); } else { root_gen = btrfs_header_generation(parent); btrfs_set_node_blockptr(parent, parent_slot, @@ -241,7 +250,7 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, return 0; } - search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1); + search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1); ret = __btrfs_cow_block(trans, root, buf, parent, parent_slot, cow_ret, search_start, 0); return ret; @@ -724,6 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans, BUG_ON(ret); root->node = child; + add_root_to_dirty_list(root); path->nodes[level] = NULL; clean_tree_block(trans, root, mid); wait_on_tree_block_writeback(root, mid); @@ -1369,6 +1379,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, /* the super has an extra ref to root->node */ free_extent_buffer(root->node); root->node = c; + add_root_to_dirty_list(root); extent_buffer_get(c); path->nodes[level] = c; path->slots[level] = 0; @@ -2777,3 +2788,28 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) } return 0; } + +int btrfs_previous_item(struct btrfs_root *root, + struct btrfs_path *path, u64 min_objectid, + int type) +{ + struct btrfs_key found_key; + struct extent_buffer *leaf; + int ret; + + while(1) { + if (path->slots[0] == 0) { + ret = btrfs_prev_leaf(root, path); + if (ret != 0) + return ret; + } else { + path->slots[0]--; + } + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.type == type) + return 0; + } + return 1; +} + diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 92d892f9207..1453d995fef 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -40,12 +40,44 @@ extern struct kmem_cache *btrfs_path_cachep; #define BTRFS_MAGIC "_B4RfS_M" #define BTRFS_MAX_LEVEL 8 + +/* holds pointers to all of the tree roots */ #define BTRFS_ROOT_TREE_OBJECTID 1ULL + +/* stores information about which extents are in use, and reference counts */ #define BTRFS_EXTENT_TREE_OBJECTID 2ULL + +/* one per subvolume, storing files and directories */ #define BTRFS_FS_TREE_OBJECTID 3ULL + +/* directory objectid inside the root tree */ #define BTRFS_ROOT_TREE_DIR_OBJECTID 4ULL + + +/* + * chunk tree stores translations from logical -> physical block numbering + * the super block points to the chunk tree + */ +#define BTRFS_CHUNK_TREE_OBJECTID 5ULL + +/* + * stores information about which areas of a given device are in use. + * one per device. The tree of tree roots points to the device tree + */ +#define BTRFS_DEV_TREE_OBJECTID 6ULL + +/* + * All files have objectids higher than this. + */ #define BTRFS_FIRST_FREE_OBJECTID 256ULL + +/* + * the device items go into the chunk tree. The key is in the form + * [ 1 BTRFS_DEV_ITEM_KEY device_id ] + */ +#define BTRFS_DEV_ITEMS_OBJECTID 1ULL + /* * we can actually store much bigger names, but lets not confuse the rest * of linux @@ -95,6 +127,81 @@ struct btrfs_key { u64 offset; } __attribute__ ((__packed__)); +struct btrfs_mapping_tree { + struct extent_map_tree map_tree; +}; + +#define BTRFS_DEV_UUID_SIZE 16 +struct btrfs_dev_item { + /* the internal btrfs device id */ + __le64 devid; + + /* size of the device */ + __le64 total_bytes; + + /* bytes used */ + __le64 bytes_used; + + /* optimal io alignment for this device */ + __le32 io_align; + + /* optimal io width for this device */ + __le32 io_width; + + /* minimal io size for this device */ + __le32 sector_size; + + /* the kernel device number */ + __le64 rdev; + + /* type and info about this device */ + __le64 type; + + /* partition number, 0 for whole dev */ + __le32 partition; + + /* length of the name data at the end of the item */ + __le16 name_len; + + /* physical drive uuid (or lvm uuid) */ + u8 uuid[BTRFS_DEV_UUID_SIZE]; + /* name goes here */ +} __attribute__ ((__packed__)); + +struct btrfs_stripe { + __le64 devid; + __le64 offset; +} __attribute__ ((__packed__)); + +struct btrfs_chunk { + __le64 owner; + __le64 stripe_len; + __le64 type; + + /* optimal io alignment for this chunk */ + __le32 io_align; + + /* optimal io width for this chunk */ + __le32 io_width; + + /* minimal io size for this chunk */ + __le32 sector_size; + + /* 2^16 stripes is quite a lot, a second limit is the size of a single + * item in the btree + */ + __le16 num_stripes; + struct btrfs_stripe stripe; + /* additional stripes go here */ +} __attribute__ ((__packed__)); + +static inline unsigned long btrfs_chunk_item_size(int num_stripes) +{ + BUG_ON(num_stripes == 0); + return sizeof(struct btrfs_chunk) + + sizeof(struct btrfs_stripe) * (num_stripes - 1); +} + #define BTRFS_FSID_SIZE 16 /* * every tree block (leaf or node) starts with this header. @@ -119,6 +226,13 @@ struct btrfs_header { sizeof(struct btrfs_item) - \ sizeof(struct btrfs_file_extent_item)) + +/* + * this is a very generous portion of the super block, giving us + * room to translate 14 chunks with 3 stripes each. + */ +#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048 + /* * the super block basically lists the main trees of the FS * it currently lacks any block count etc etc @@ -131,6 +245,7 @@ struct btrfs_super_block { __le64 magic; __le64 generation; __le64 root; + __le64 chunk_root; __le64 total_bytes; __le64 bytes_used; __le64 root_dir_objectid; @@ -138,7 +253,10 @@ struct btrfs_super_block { __le32 nodesize; __le32 leafsize; __le32 stripesize; + __le32 sys_chunk_array_size; u8 root_level; + u8 chunk_root_level; + u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE]; } __attribute__ ((__packed__)); /* @@ -208,12 +326,22 @@ struct btrfs_extent_ref { __le64 offset; } __attribute__ ((__packed__)); +/* dev extents record free space on individual devices. The owner + * field points back to the chunk allocation mapping tree that allocated + * the extent + */ +struct btrfs_dev_extent { + __le64 owner; + __le64 length; +} __attribute__ ((__packed__)); + + struct btrfs_inode_ref { __le16 name_len; /* name goes here */ } __attribute__ ((__packed__)); -struct btrfs_inode_timespec { +struct btrfs_timespec { __le64 sec; __le32 nsec; } __attribute__ ((__packed__)); @@ -231,13 +359,13 @@ struct btrfs_inode_item { __le32 uid; __le32 gid; __le32 mode; - __le32 rdev; + __le64 rdev; __le16 flags; __le16 compat_flags; - struct btrfs_inode_timespec atime; - struct btrfs_inode_timespec ctime; - struct btrfs_inode_timespec mtime; - struct btrfs_inode_timespec otime; + struct btrfs_timespec atime; + struct btrfs_timespec ctime; + struct btrfs_timespec mtime; + struct btrfs_timespec otime; } __attribute__ ((__packed__)); struct btrfs_dir_item { @@ -290,29 +418,34 @@ struct btrfs_csum_item { u8 csum; } __attribute__ ((__packed__)); -/* tag for the radix tree of block groups in ram */ -#define BTRFS_BLOCK_GROUP_SIZE (256 * 1024 * 1024) - +/* different types of block groups (and chunks) */ +#define BTRFS_BLOCK_GROUP_DATA (1 << 0) +#define BTRFS_BLOCK_GROUP_SYSTEM (1 << 1) +#define BTRFS_BLOCK_GROUP_METADATA (1 << 2) -#define BTRFS_BLOCK_GROUP_DATA 1 -#define BTRFS_BLOCK_GROUP_MIXED 2 struct btrfs_block_group_item { __le64 used; - u8 flags; + __le64 chunk_tree; + __le64 chunk_objectid; + __le64 flags; } __attribute__ ((__packed__)); struct btrfs_block_group_cache { struct btrfs_key key; struct btrfs_block_group_item item; - int data; - int cached; u64 pinned; + u64 flags; + int cached; }; + +struct btrfs_device; struct btrfs_fs_info { u8 fsid[BTRFS_FSID_SIZE]; struct btrfs_root *extent_root; struct btrfs_root *tree_root; + struct btrfs_root *chunk_root; + struct btrfs_root *dev_root; struct radix_tree_root fs_roots_radix; struct extent_io_tree free_space_cache; @@ -321,6 +454,9 @@ struct btrfs_fs_info { struct extent_io_tree pending_del; struct extent_io_tree extent_ins; + /* logical->physical extent mapping */ + struct btrfs_mapping_tree mapping_tree; + u64 generation; u64 last_trans_committed; unsigned long mount_opt; @@ -330,6 +466,7 @@ struct btrfs_fs_info { struct btrfs_transaction *running_transaction; struct btrfs_super_block super_copy; struct extent_buffer *sb_buffer; + struct block_device *__bdev; struct super_block *sb; struct inode *btree_inode; spinlock_t hash_lock; @@ -350,12 +487,17 @@ struct btrfs_fs_info { unsigned long throttles; u64 total_pinned; + struct list_head dirty_cowonly_roots; + + struct list_head devices; + struct list_head *last_device; spinlock_t delalloc_lock; spinlock_t new_trans_lock; u64 delalloc_bytes; u64 last_alloc; u64 last_data_alloc; }; + /* * in ram representation of the tree. extent_root is used for all allocations * and for the extent tree extent_root root. @@ -387,14 +529,19 @@ struct btrfs_root { u64 highest_inode; u64 last_inode_alloc; int ref_cows; + int track_dirty; struct btrfs_key defrag_progress; int defrag_running; int defrag_level; char *name; int in_sysfs; + + /* the dirty list is only used by non-reference counted roots */ + struct list_head dirty_list; }; /* + * inode items have the data typically returned from stat and store other * info about object characteristics. There is one for every file and dir in * the FS @@ -439,6 +586,10 @@ struct btrfs_root { */ #define BTRFS_BLOCK_GROUP_ITEM_KEY 50 +#define BTRFS_DEV_EXTENT_KEY 75 +#define BTRFS_DEV_ITEM_KEY 76 +#define BTRFS_CHUNK_ITEM_KEY 77 + /* * string items are for debugging. They just store a short string of * data in the FS @@ -518,13 +669,104 @@ static inline void btrfs_set_##name(type *s, u##bits val) \ s->member = cpu_to_le##bits(val); \ } +BTRFS_SETGET_FUNCS(device_type, struct btrfs_dev_item, type, 64); +BTRFS_SETGET_FUNCS(device_total_bytes, struct btrfs_dev_item, total_bytes, 64); +BTRFS_SETGET_FUNCS(device_bytes_used, struct btrfs_dev_item, bytes_used, 64); +BTRFS_SETGET_FUNCS(device_io_align, struct btrfs_dev_item, io_align, 32); +BTRFS_SETGET_FUNCS(device_io_width, struct btrfs_dev_item, io_width, 32); +BTRFS_SETGET_FUNCS(device_sector_size, struct btrfs_dev_item, sector_size, 32); +BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64); +BTRFS_SETGET_FUNCS(device_rdev, struct btrfs_dev_item, rdev, 64); +BTRFS_SETGET_FUNCS(device_partition, struct btrfs_dev_item, partition, 32); +BTRFS_SETGET_FUNCS(device_name_len, struct btrfs_dev_item, name_len, 16); + +static inline char *btrfs_device_uuid(struct btrfs_dev_item *d) +{ + return (char *)d + offsetof(struct btrfs_dev_item, uuid); +} + +static inline char *btrfs_device_name(struct btrfs_dev_item *d) +{ + return (char *)(d + 1); +} + +BTRFS_SETGET_FUNCS(chunk_owner, struct btrfs_chunk, owner, 64); +BTRFS_SETGET_FUNCS(chunk_stripe_len, struct btrfs_chunk, stripe_len, 64); +BTRFS_SETGET_FUNCS(chunk_io_align, struct btrfs_chunk, io_align, 32); +BTRFS_SETGET_FUNCS(chunk_io_width, struct btrfs_chunk, io_width, 32); +BTRFS_SETGET_FUNCS(chunk_sector_size, struct btrfs_chunk, sector_size, 32); +BTRFS_SETGET_FUNCS(chunk_type, struct btrfs_chunk, type, 64); +BTRFS_SETGET_FUNCS(chunk_num_stripes, struct btrfs_chunk, num_stripes, 16); +BTRFS_SETGET_FUNCS(stripe_devid, struct btrfs_stripe, devid, 64); +BTRFS_SETGET_FUNCS(stripe_offset, struct btrfs_stripe, offset, 64); + +BTRFS_SETGET_STACK_FUNCS(stack_chunk_owner, struct btrfs_chunk, owner, 64); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_stripe_len, struct btrfs_chunk, + stripe_len, 64); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_align, struct btrfs_chunk, + io_align, 32); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_width, struct btrfs_chunk, + io_width, 32); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_sector_size, struct btrfs_chunk, + sector_size, 32); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_type, struct btrfs_chunk, type, 64); +BTRFS_SETGET_STACK_FUNCS(stack_chunk_num_stripes, struct btrfs_chunk, + num_stripes, 16); +BTRFS_SETGET_STACK_FUNCS(stack_stripe_devid, struct btrfs_stripe, devid, 64); +BTRFS_SETGET_STACK_FUNCS(stack_stripe_offset, struct btrfs_stripe, offset, 64); + +static inline struct btrfs_stripe *btrfs_stripe_nr(struct btrfs_chunk *c, + int nr) +{ + unsigned long offset = (unsigned long)c; + offset += offsetof(struct btrfs_chunk, stripe); + offset += nr * sizeof(struct btrfs_stripe); + return (struct btrfs_stripe *)offset; +} + +static inline u64 btrfs_stripe_offset_nr(struct extent_buffer *eb, + struct btrfs_chunk *c, int nr) +{ + return btrfs_stripe_offset(eb, btrfs_stripe_nr(c, nr)); +} + +static inline void btrfs_set_stripe_offset_nr(struct extent_buffer *eb, + struct btrfs_chunk *c, int nr, + u64 val) +{ + btrfs_set_stripe_offset(eb, btrfs_stripe_nr(c, nr), val); +} + +static inline u64 btrfs_stripe_devid_nr(struct extent_buffer *eb, + struct btrfs_chunk *c, int nr) +{ + return btrfs_stripe_devid(eb, btrfs_stripe_nr(c, nr)); +} + +static inline void btrfs_set_stripe_devid_nr(struct extent_buffer *eb, + struct btrfs_chunk *c, int nr, + u64 val) +{ + btrfs_set_stripe_devid(eb, btrfs_stripe_nr(c, nr), val); +} + /* struct btrfs_block_group_item */ BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item, used, 64); BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item, used, 64); -BTRFS_SETGET_FUNCS(disk_block_group_flags, struct btrfs_block_group_item, - flags, 8); +BTRFS_SETGET_STACK_FUNCS(block_group_chunk_tree, struct btrfs_block_group_item, + chunk_tree, 64); +BTRFS_SETGET_FUNCS(disk_block_group_chunk_tree, struct btrfs_block_group_item, + chunk_tree, 64); +BTRFS_SETGET_STACK_FUNCS(block_group_chunk_objectid, + struct btrfs_block_group_item, chunk_objectid, 64); +BTRFS_SETGET_FUNCS(disk_block_group_chunk_objecitd, + struct btrfs_block_group_item, chunk_objectid, 64); +BTRFS_SETGET_FUNCS(disk_block_group_flags, + struct btrfs_block_group_item, flags, 64); +BTRFS_SETGET_STACK_FUNCS(block_group_flags, + struct btrfs_block_group_item, flags, 64); /* struct btrfs_inode_ref */ BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16); @@ -538,49 +780,53 @@ BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32); BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32); BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32); BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32); -BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 32); +BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 64); BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 16); BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item, compat_flags, 16); -static inline struct btrfs_inode_timespec * +static inline struct btrfs_timespec * btrfs_inode_atime(struct btrfs_inode_item *inode_item) { unsigned long ptr = (unsigned long)inode_item; ptr += offsetof(struct btrfs_inode_item, atime); - return (struct btrfs_inode_timespec *)ptr; + return (struct btrfs_timespec *)ptr; } -static inline struct btrfs_inode_timespec * +static inline struct btrfs_timespec * btrfs_inode_mtime(struct btrfs_inode_item *inode_item) { unsigned long ptr = (unsigned long)inode_item; ptr += offsetof(struct btrfs_inode_item, mtime); - return (struct btrfs_inode_timespec *)ptr; + return (struct btrfs_timespec *)ptr; } -static inline struct btrfs_inode_timespec * +static inline struct btrfs_timespec * btrfs_inode_ctime(struct btrfs_inode_item *inode_item) { unsigned long ptr = (unsigned long)inode_item; ptr += offsetof(struct btrfs_inode_item, ctime); - return (struct btrfs_inode_timespec *)ptr; + return (struct btrfs_timespec *)ptr; } -static inline struct btrfs_inode_timespec * +static inline struct btrfs_timespec * btrfs_inode_otime(struct btrfs_inode_item *inode_item) { unsigned long ptr = (unsigned long)inode_item; ptr += offsetof(struct btrfs_inode_item, otime); - return (struct btrfs_inode_timespec *)ptr; + return (struct btrfs_timespec *)ptr; } -BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_inode_timespec, sec, 64); -BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_inode_timespec, nsec, 32); +BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64); +BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32); /* struct btrfs_extent_item */ BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32); +/* struct btrfs_dev_extent */ +BTRFS_SETGET_FUNCS(dev_extent_owner, struct btrfs_dev_extent, owner, 64); +BTRFS_SETGET_FUNCS(dev_extent_length, struct btrfs_dev_extent, length, 64); + /* struct btrfs_extent_ref */ BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64); BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64); @@ -846,8 +1092,14 @@ BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64); BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block, generation, 64); BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64); +BTRFS_SETGET_STACK_FUNCS(super_sys_array_size, + struct btrfs_super_block, sys_chunk_array_size, 32); BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block, root_level, 8); +BTRFS_SETGET_STACK_FUNCS(super_chunk_root, struct btrfs_super_block, + chunk_root, 64); +BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block, + chunk_root_level, 64); BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block, total_bytes, 64); BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block, @@ -1009,7 +1261,14 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_free_block_groups(struct btrfs_fs_info *info); int btrfs_read_block_groups(struct btrfs_root *root); +int btrfs_make_block_group(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytes_used, + u64 type, u64 chunk_tree, u64 chunk_objectid, + u64 size); /* ctree.c */ +int btrfs_previous_item(struct btrfs_root *root, + struct btrfs_path *path, u64 min_objectid, + int type); int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 88e21bdbc47..8e37fa120cc 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -28,6 +28,7 @@ #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" +#include "volumes.h" #include "print-tree.h" #if 0 @@ -234,6 +235,19 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) return 0; } +static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 offset; + offset = bio->bi_sector << 9; + if (offset == BTRFS_SUPER_INFO_OFFSET) { + bio->bi_bdev = root->fs_info->sb->s_bdev; + submit_bio(rw, bio); + return 0; + } + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio); +} + static int btree_writepage(struct page *page, struct writeback_control *wbc) { struct extent_io_tree *tree; @@ -345,6 +359,23 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) return ret; } +static int close_all_devices(struct btrfs_fs_info *fs_info) +{ + struct list_head *list; + struct list_head *next; + struct btrfs_device *device; + + list = &fs_info->devices; + while(!list_empty(list)) { + next = list->next; + list_del(next); + device = list_entry(next, struct btrfs_device, dev_list); + kfree(device->name); + kfree(device); + } + return 0; +} + struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) { @@ -420,6 +451,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->leafsize = leafsize; root->stripesize = stripesize; root->ref_cows = 0; + root->track_dirty = 0; + root->fs_info = fs_info; root->objectid = objectid; root->last_trans = 0; @@ -427,6 +460,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->last_inode_alloc = 0; root->name = NULL; root->in_sysfs = 0; + + INIT_LIST_HEAD(&root->dirty_list); memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); @@ -634,6 +669,10 @@ struct btrfs_root *open_ctree(struct super_block *sb) GFP_NOFS); struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info), GFP_NOFS); + struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), + GFP_NOFS); + struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), + GFP_NOFS); int ret; int err = -EIO; struct btrfs_super_block *disk_super; @@ -657,6 +696,12 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->last_trans_committed = 0; fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; + fs_info->chunk_root = chunk_root; + fs_info->dev_root = dev_root; + INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); + INIT_LIST_HEAD(&fs_info->devices); + btrfs_mapping_init(&fs_info->mapping_tree); + fs_info->last_device = &fs_info->devices; fs_info->sb = sb; fs_info->throttles = 0; fs_info->mount_opt = 0; @@ -714,12 +759,12 @@ struct btrfs_root *open_ctree(struct super_block *sb) goto fail_iput; } #endif - __setup_root(512, 512, 512, 512, tree_root, + __setup_root(4096, 4096, 4096, 4096, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); fs_info->sb_buffer = read_tree_block(tree_root, BTRFS_SUPER_INFO_OFFSET, - 512); + 4096); if (!fs_info->sb_buffer) goto fail_iput; @@ -730,6 +775,7 @@ struct btrfs_root *open_ctree(struct super_block *sb) read_extent_buffer(fs_info->sb_buffer, fs_info->fsid, (unsigned long)btrfs_super_fsid(fs_info->sb_buffer), BTRFS_FSID_SIZE); + disk_super = &fs_info->super_copy; if (!btrfs_super_root(disk_super)) goto fail_sb_buffer; @@ -753,23 +799,47 @@ struct btrfs_root *open_ctree(struct super_block *sb) goto fail_sb_buffer; } + mutex_lock(&fs_info->fs_mutex); + ret = btrfs_read_sys_array(tree_root); + BUG_ON(ret); + + blocksize = btrfs_level_size(tree_root, + btrfs_super_chunk_root_level(disk_super)); + + __setup_root(nodesize, leafsize, sectorsize, stripesize, + chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); + + chunk_root->node = read_tree_block(chunk_root, + btrfs_super_chunk_root(disk_super), + blocksize); + BUG_ON(!chunk_root->node); + + ret = btrfs_read_chunk_tree(chunk_root); + BUG_ON(ret); + blocksize = btrfs_level_size(tree_root, btrfs_super_root_level(disk_super)); + tree_root->node = read_tree_block(tree_root, btrfs_super_root(disk_super), blocksize); if (!tree_root->node) goto fail_sb_buffer; - mutex_lock(&fs_info->fs_mutex); ret = find_and_setup_root(tree_root, fs_info, BTRFS_EXTENT_TREE_OBJECTID, extent_root); - if (ret) { - mutex_unlock(&fs_info->fs_mutex); + if (ret) goto fail_tree_root; - } + extent_root->track_dirty = 1; + + ret = find_and_setup_root(tree_root, fs_info, + BTRFS_DEV_TREE_OBJECTID, dev_root); + dev_root->track_dirty = 1; + + if (ret) + goto fail_extent_root; btrfs_read_block_groups(extent_root); @@ -777,7 +847,10 @@ struct btrfs_root *open_ctree(struct super_block *sb) mutex_unlock(&fs_info->fs_mutex); return tree_root; +fail_extent_root: + free_extent_buffer(extent_root->node); fail_tree_root: + mutex_unlock(&fs_info->fs_mutex); free_extent_buffer(tree_root->node); fail_sb_buffer: free_extent_buffer(fs_info->sb_buffer); @@ -874,6 +947,12 @@ int close_ctree(struct btrfs_root *root) if (fs_info->tree_root->node) free_extent_buffer(fs_info->tree_root->node); + if (root->fs_info->chunk_root->node); + free_extent_buffer(root->fs_info->chunk_root->node); + + if (root->fs_info->dev_root->node); + free_extent_buffer(root->fs_info->dev_root->node); + free_extent_buffer(fs_info->sb_buffer); btrfs_free_block_groups(root->fs_info); @@ -901,8 +980,13 @@ int close_ctree(struct btrfs_root *root) kfree(hasher); } #endif + close_all_devices(fs_info); + btrfs_mapping_tree_free(&fs_info->mapping_tree); + kfree(fs_info->extent_root); kfree(fs_info->tree_root); + kfree(fs_info->chunk_root); + kfree(fs_info->dev_root); return 0; } @@ -1016,4 +1100,5 @@ int btrfs_read_buffer(struct extent_buffer *buf) static struct extent_io_ops btree_extent_io_ops = { .writepage_io_hook = btree_writepage_io_hook, + .submit_bio_hook = btree_submit_bio_hook, }; diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 828f3a2081b..206cb48638f 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -20,6 +20,7 @@ #define __DISKIO__ #define BTRFS_SUPER_INFO_OFFSET (16 * 1024) +struct btrfs_device; struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize); @@ -65,4 +66,5 @@ int btrfs_read_buffer(struct extent_buffer *buf); u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len); void btrfs_csum_final(u32 crc, char *result); void btrfs_throttle(struct btrfs_root *root); +int btrfs_open_device(struct btrfs_device *dev); #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index ebfd304138c..2cd957d6e8d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -24,37 +24,19 @@ #include "disk-io.h" #include "print-tree.h" #include "transaction.h" +#include "volumes.h" -#define BLOCK_GROUP_DATA EXTENT_WRITEBACK +#define BLOCK_GROUP_DATA EXTENT_WRITEBACK #define BLOCK_GROUP_METADATA EXTENT_UPTODATE +#define BLOCK_GROUP_SYSTEM EXTENT_NEW + #define BLOCK_GROUP_DIRTY EXTENT_DIRTY static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); static int del_pending_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); -static int find_previous_extent(struct btrfs_root *root, - struct btrfs_path *path) -{ - struct btrfs_key found_key; - struct extent_buffer *leaf; - int ret; - while(1) { - if (path->slots[0] == 0) { - ret = btrfs_prev_leaf(root, path); - if (ret != 0) - return ret; - } else { - path->slots[0]--; - } - leaf = path->nodes[0]; - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (found_key.type == BTRFS_EXTENT_ITEM_KEY) - return 0; - } - return 1; -} static int cache_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *block_group) @@ -91,7 +73,7 @@ static int cache_block_group(struct btrfs_root *root, ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) return ret; - ret = find_previous_extent(root, path); + ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); if (ret < 0) return ret; if (ret == 0) { @@ -168,7 +150,8 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct block_group_cache = &info->block_group_cache; ret = find_first_extent_bit(block_group_cache, bytenr, &start, &end, - BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA); + BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | + BLOCK_GROUP_SYSTEM); if (ret) { return NULL; } @@ -182,23 +165,38 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct return block_group; return NULL; } -static u64 noinline find_search_start(struct btrfs_root *root, + +static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) +{ + if ((bits & BLOCK_GROUP_DATA) && + (cache->flags & BTRFS_BLOCK_GROUP_DATA)) + return 1; + if ((bits & BLOCK_GROUP_METADATA) && + (cache->flags & BTRFS_BLOCK_GROUP_METADATA)) + return 1; + if ((bits & BLOCK_GROUP_SYSTEM) && + (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM)) + return 1; + return 0; +} + +static int noinline find_search_start(struct btrfs_root *root, struct btrfs_block_group_cache **cache_ret, - u64 search_start, int num, int data) + u64 *start_ret, int num, int data) { int ret; struct btrfs_block_group_cache *cache = *cache_ret; struct extent_io_tree *free_space_cache; - struct extent_state *state; u64 last; u64 start = 0; + u64 end = 0; u64 cache_miss = 0; u64 total_fs_bytes; + u64 search_start = *start_ret; int wrapped = 0; - if (!cache) { + if (!cache) goto out; - } total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); free_space_cache = &root->fs_info->free_space_cache; @@ -208,6 +206,9 @@ again: goto out; last = max(search_start, cache->key.objectid); + if (!block_group_bits(cache, data)) { + goto new_group; + } while(1) { ret = find_first_extent_bit(&root->fs_info->free_space_cache, @@ -225,22 +226,20 @@ again: cache_miss = start; continue; } - if (data != BTRFS_BLOCK_GROUP_MIXED && - start + num > cache->key.objectid + cache->key.offset) + if (start + num > cache->key.objectid + cache->key.offset) goto new_group; if (start + num > total_fs_bytes) goto new_group; - return start; + *start_ret = start; + return 0; } out: cache = btrfs_lookup_block_group(root->fs_info, search_start); if (!cache) { - printk("Unable to find block group for %Lu\n", - search_start); + printk("Unable to find block group for %Lu\n", search_start); WARN_ON(1); - return search_start; } - return search_start; + return -ENOSPC; new_group: last = cache->key.objectid + cache->key.offset; @@ -251,7 +250,6 @@ no_cache: if (!wrapped) { wrapped = 1; last = search_start; - data = BTRFS_BLOCK_GROUP_MIXED; goto wrapped; } goto out; @@ -299,7 +297,6 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int ret; int full_search = 0; int factor = 8; - int data_swap = 0; block_group_cache = &info->block_group_cache; total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); @@ -307,19 +304,12 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, if (!owner) factor = 8; - if (data == BTRFS_BLOCK_GROUP_MIXED) { - bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; - factor = 10; - } else if (data) - bit = BLOCK_GROUP_DATA; - else - bit = BLOCK_GROUP_METADATA; + bit = data; if (search_start && search_start < total_fs_bytes) { struct btrfs_block_group_cache *shint; shint = btrfs_lookup_block_group(info, search_start); - if (shint && (shint->data == data || - shint->data == BTRFS_BLOCK_GROUP_MIXED)) { + if (shint && block_group_bits(shint, data)) { used = btrfs_block_group_used(&shint->item); if (used + shint->pinned < div_factor(shint->key.offset, factor)) { @@ -327,8 +317,8 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, } } } - if (hint && hint->key.objectid < total_fs_bytes && - (hint->data == data || hint->data == BTRFS_BLOCK_GROUP_MIXED)) { + if (hint && block_group_bits(hint, data) && + hint->key.objectid < total_fs_bytes) { used = btrfs_block_group_used(&hint->item); if (used + hint->pinned < div_factor(hint->key.offset, factor)) { @@ -379,12 +369,6 @@ again: full_search = 1; goto again; } - if (!data_swap) { - data_swap = 1; - bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; - last = search_start; - goto again; - } found: return found_group; } @@ -1002,7 +986,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, static int update_block_group(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, int alloc, - int mark_free, int data) + int mark_free) { struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; @@ -1027,41 +1011,6 @@ static int update_block_group(struct btrfs_trans_handle *trans, old_val = btrfs_block_group_used(&cache->item); num_bytes = min(total, cache->key.offset - byte_in_group); if (alloc) { - if (cache->data != data && - old_val < (cache->key.offset >> 1)) { - int bit_to_clear; - int bit_to_set; - cache->data = data; - if (data) { - bit_to_clear = BLOCK_GROUP_METADATA; - bit_to_set = BLOCK_GROUP_DATA; - cache->item.flags &= - ~BTRFS_BLOCK_GROUP_MIXED; - cache->item.flags |= - BTRFS_BLOCK_GROUP_DATA; - } else { - bit_to_clear = BLOCK_GROUP_DATA; - bit_to_set = BLOCK_GROUP_METADATA; - cache->item.flags &= - ~BTRFS_BLOCK_GROUP_MIXED; - cache->item.flags &= - ~BTRFS_BLOCK_GROUP_DATA; - } - clear_extent_bits(&info->block_group_cache, - start, end, bit_to_clear, - GFP_NOFS); - set_extent_bits(&info->block_group_cache, - start, end, bit_to_set, - GFP_NOFS); - } else if (cache->data != data && - cache->data != BTRFS_BLOCK_GROUP_MIXED) { - cache->data = BTRFS_BLOCK_GROUP_MIXED; - set_extent_bits(&info->block_group_cache, - start, end, - BLOCK_GROUP_DATA | - BLOCK_GROUP_METADATA, - GFP_NOFS); - } old_val += num_bytes; } else { old_val -= num_bytes; @@ -1357,7 +1306,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root return ret; } ret = update_block_group(trans, root, bytenr, num_bytes, 0, - mark_free, 0); + mark_free); BUG_ON(ret); } btrfs_free_path(path); @@ -1450,38 +1399,21 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, u64 exclude_start, u64 exclude_nr, int data) { - struct btrfs_path *path; - struct btrfs_key key; - u64 hole_size = 0; - u64 aligned; int ret; - int slot = 0; - u64 last_byte = 0; - u64 *last_ptr = NULL; u64 orig_search_start = search_start; - int start_found; - struct extent_buffer *l; struct btrfs_root * root = orig_root->fs_info->extent_root; struct btrfs_fs_info *info = root->fs_info; u64 total_needed = num_bytes; - int level; struct btrfs_block_group_cache *block_group; int full_scan = 0; int wrapped = 0; - int empty_cluster; - u64 cached_start; WARN_ON(num_bytes < root->sectorsize); btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); - level = btrfs_header_level(root->node); - - if (num_bytes >= 32 * 1024 * 1024 && hint_byte) { - data = BTRFS_BLOCK_GROUP_MIXED; - } - if (search_end == (u64)-1) search_end = btrfs_super_total_bytes(&info->super_copy); + if (hint_byte) { block_group = btrfs_lookup_block_group(info, hint_byte); if (!block_group) @@ -1495,7 +1427,7 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, } total_needed += empty_size; - path = btrfs_alloc_path(); + check_failed: if (!block_group) { block_group = btrfs_lookup_block_group(info, search_start); @@ -1503,135 +1435,49 @@ check_failed: block_group = btrfs_lookup_block_group(info, orig_search_start); } - search_start = find_search_start(root, &block_group, search_start, - total_needed, data); - search_start = stripe_align(root, search_start); - cached_start = search_start; - btrfs_init_path(path); - ins->objectid = search_start; - ins->offset = 0; - start_found = 0; - path->reada = 2; - - ret = btrfs_search_slot(trans, root, ins, path, 0, 0); - if (ret < 0) - goto error; - ret = find_previous_extent(root, path); - if (ret < 0) + ret = find_search_start(root, &block_group, &search_start, + total_needed, data); + if (ret) goto error; - l = path->nodes[0]; - btrfs_item_key_to_cpu(l, &key, path->slots[0]); - while (1) { - l = path->nodes[0]; - slot = path->slots[0]; - if (slot >= btrfs_header_nritems(l)) { - ret = btrfs_next_leaf(root, path); - if (ret == 0) - continue; - if (ret < 0) - goto error; - search_start = max(search_start, - block_group->key.objectid); - if (!start_found) { - aligned = stripe_align(root, search_start); - ins->objectid = aligned; - if (aligned >= search_end) { - ret = -ENOSPC; - goto error; - } - ins->offset = search_end - aligned; - start_found = 1; - goto check_pending; - } - ins->objectid = stripe_align(root, - last_byte > search_start ? - last_byte : search_start); - if (search_end <= ins->objectid) { - ret = -ENOSPC; - goto error; - } - ins->offset = search_end - ins->objectid; - BUG_ON(ins->objectid >= search_end); - goto check_pending; - } - btrfs_item_key_to_cpu(l, &key, slot); - - if (key.objectid >= search_start && key.objectid > last_byte && - start_found) { - if (last_byte < search_start) - last_byte = search_start; - aligned = stripe_align(root, last_byte); - hole_size = key.objectid - aligned; - if (key.objectid > aligned && hole_size >= num_bytes) { - ins->objectid = aligned; - ins->offset = hole_size; - goto check_pending; - } - } - if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) { - if (!start_found && btrfs_key_type(&key) == - BTRFS_BLOCK_GROUP_ITEM_KEY) { - last_byte = key.objectid; - start_found = 1; - } - goto next; - } - - - start_found = 1; - last_byte = key.objectid + key.offset; - - if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED && - last_byte >= block_group->key.objectid + - block_group->key.offset) { - btrfs_release_path(root, path); - search_start = block_group->key.objectid + - block_group->key.offset; - goto new_group; - } -next: - path->slots[0]++; - cond_resched(); - } -check_pending: - /* we have to make sure we didn't find an extent that has already - * been allocated by the map tree or the original allocation - */ - btrfs_release_path(root, path); - BUG_ON(ins->objectid < search_start); + search_start = stripe_align(root, search_start); + ins->objectid = search_start; + ins->offset = num_bytes; if (ins->objectid + num_bytes >= search_end) goto enospc; - if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED && - ins->objectid + num_bytes > block_group-> - key.objectid + block_group->key.offset) { + + if (ins->objectid + num_bytes > + block_group->key.objectid + block_group->key.offset) { search_start = block_group->key.objectid + block_group->key.offset; goto new_group; } + if (test_range_bit(&info->extent_ins, ins->objectid, ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { search_start = ins->objectid + num_bytes; goto new_group; } + if (test_range_bit(&info->pinned_extents, ins->objectid, ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { search_start = ins->objectid + num_bytes; goto new_group; } + if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && ins->objectid < exclude_start + exclude_nr)) { search_start = exclude_start + exclude_nr; goto new_group; } - if (!data) { + + if (!(data & BLOCK_GROUP_DATA)) { block_group = btrfs_lookup_block_group(info, ins->objectid); if (block_group) trans->block_group = block_group; } ins->offset = num_bytes; - btrfs_free_path(path); return 0; new_group: @@ -1646,7 +1492,6 @@ enospc: if (!full_scan) total_needed -= empty_size; full_scan = 1; - data = BTRFS_BLOCK_GROUP_MIXED; } else wrapped = 1; } @@ -1657,8 +1502,6 @@ enospc: goto check_failed; error: - btrfs_release_path(root, path); - btrfs_free_path(path); return ret; } /* @@ -1689,6 +1532,13 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; struct btrfs_key keys[2]; + if (data) + data = BLOCK_GROUP_DATA; + else if (root == root->fs_info->chunk_root) + data = BLOCK_GROUP_SYSTEM; + else + data = BLOCK_GROUP_METADATA; + new_hint = max(hint_byte, root->fs_info->alloc_start); if (new_hint < btrfs_super_total_bytes(&info->super_copy)) hint_byte = new_hint; @@ -1718,7 +1568,6 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, set_extent_bits(&root->fs_info->extent_ins, ins->objectid, ins->objectid + ins->offset - 1, EXTENT_LOCKED, GFP_NOFS); - WARN_ON(data == 1); goto update_block; } @@ -1768,8 +1617,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, } update_block: - ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0, - data); + ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); if (ret) { printk("update block group failed for %Lu %Lu\n", ins->objectid, ins->offset); @@ -2457,7 +2305,7 @@ again: if (ret < 0) goto out; - ret = find_previous_extent(root, path); + ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); if (ret < 0) goto out; if (ret == 0) { @@ -2604,95 +2452,48 @@ out: int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 new_size) { - struct btrfs_path *path; - u64 nr = 0; - u64 cur_byte; - u64 old_size; - unsigned long rem; - struct btrfs_block_group_cache *cache; - struct btrfs_block_group_item *item; - struct btrfs_fs_info *info = root->fs_info; - struct extent_io_tree *block_group_cache; - struct btrfs_key key; - struct extent_buffer *leaf; - int ret; - int bit; - - old_size = btrfs_super_total_bytes(&info->super_copy); - block_group_cache = &info->block_group_cache; - - root = info->extent_root; - - cache = btrfs_lookup_block_group(root->fs_info, old_size - 1); - - cur_byte = cache->key.objectid + cache->key.offset; - if (cur_byte >= new_size) - goto set_size; - - key.offset = BTRFS_BLOCK_GROUP_SIZE; - btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); + btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size); + return 0; +} - path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; +int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, + struct btrfs_key *key) +{ + int ret; + struct btrfs_key found_key; + struct extent_buffer *leaf; + int slot; - while(cur_byte < new_size) { - key.objectid = cur_byte; - ret = btrfs_insert_empty_item(trans, root, path, &key, - sizeof(struct btrfs_block_group_item)); - BUG_ON(ret); + ret = btrfs_search_slot(NULL, root, key, path, 0, 0); + if (ret < 0) + return ret; + while(1) { + slot = path->slots[0]; leaf = path->nodes[0]; - item = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_block_group_item); - - btrfs_set_disk_block_group_used(leaf, item, 0); - div_long_long_rem(nr, 3, &rem); - if (rem) { - btrfs_set_disk_block_group_flags(leaf, item, - BTRFS_BLOCK_GROUP_DATA); - } else { - btrfs_set_disk_block_group_flags(leaf, item, 0); - } - nr++; - - cache = kmalloc(sizeof(*cache), GFP_NOFS); - BUG_ON(!cache); - - read_extent_buffer(leaf, &cache->item, (unsigned long)item, - sizeof(cache->item)); - - memcpy(&cache->key, &key, sizeof(key)); - cache->cached = 0; - cache->pinned = 0; - cur_byte = key.objectid + key.offset; - btrfs_release_path(root, path); - - if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { - bit = BLOCK_GROUP_DATA; - cache->data = BTRFS_BLOCK_GROUP_DATA; - } else { - bit = BLOCK_GROUP_METADATA; - cache->data = 0; + if (slot >= btrfs_header_nritems(leaf)) { + ret = btrfs_next_leaf(root, path); + if (ret == 0) + continue; + if (ret < 0) + goto error; + break; } + btrfs_item_key_to_cpu(leaf, &found_key, slot); - /* use EXTENT_LOCKED to prevent merging */ - set_extent_bits(block_group_cache, key.objectid, - key.objectid + key.offset - 1, - bit | EXTENT_LOCKED, GFP_NOFS); - set_state_private(block_group_cache, key.objectid, - (unsigned long)cache); + if (found_key.objectid >= key->objectid && + found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) + return 0; + path->slots[0]++; } - btrfs_free_path(path); -set_size: - btrfs_set_super_total_bytes(&info->super_copy, new_size); - return 0; + ret = -ENOENT; +error: + return ret; } int btrfs_read_block_groups(struct btrfs_root *root) { struct btrfs_path *path; int ret; - int err = 0; int bit; struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; @@ -2702,28 +2503,28 @@ int btrfs_read_block_groups(struct btrfs_root *root) struct extent_buffer *leaf; block_group_cache = &info->block_group_cache; - root = info->extent_root; key.objectid = 0; - key.offset = BTRFS_BLOCK_GROUP_SIZE; + key.offset = 0; btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); - path = btrfs_alloc_path(); if (!path) return -ENOMEM; while(1) { - ret = btrfs_search_slot(NULL, info->extent_root, - &key, path, 0, 0); - if (ret != 0) { - err = ret; - break; + ret = find_first_block_group(root, path, &key); + if (ret > 0) { + ret = 0; + goto error; } + if (ret != 0) + goto error; + leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); cache = kmalloc(sizeof(*cache), GFP_NOFS); if (!cache) { - err = -1; + ret = -ENOMEM; break; } @@ -2733,18 +2534,17 @@ int btrfs_read_block_groups(struct btrfs_root *root) memcpy(&cache->key, &found_key, sizeof(found_key)); cache->cached = 0; cache->pinned = 0; + key.objectid = found_key.objectid + found_key.offset; btrfs_release_path(root, path); - - if (cache->item.flags & BTRFS_BLOCK_GROUP_MIXED) { - bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; - cache->data = BTRFS_BLOCK_GROUP_MIXED; - } else if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { + cache->flags = btrfs_block_group_flags(&cache->item); + bit = 0; + if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { bit = BLOCK_GROUP_DATA; - cache->data = BTRFS_BLOCK_GROUP_DATA; - } else { + } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { + bit = BLOCK_GROUP_SYSTEM; + } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { bit = BLOCK_GROUP_METADATA; - cache->data = 0; } /* use EXTENT_LOCKED to prevent merging */ @@ -2758,7 +2558,8 @@ int btrfs_read_block_groups(struct btrfs_root *root) btrfs_super_total_bytes(&info->super_copy)) break; } - + ret = 0; +error: btrfs_free_path(path); - return 0; + return ret; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e8130c87633..7e3a1ebde9f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1706,9 +1706,9 @@ static int submit_one_bio(int rw, struct bio *bio) WARN_ON(1); } if (tree->ops && tree->ops->submit_bio_hook) - tree->ops->submit_bio_hook(rw, bio); - - submit_bio(rw, bio); + tree->ops->submit_bio_hook(page->mapping->host, rw, bio); + else + submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) ret = -EOPNOTSUPP; bio_put(bio); diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 9d665466708..0dca89328f9 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -28,7 +28,7 @@ struct extent_state; struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); - int (*submit_bio_hook)(int rw, struct bio *bio); + int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 315dd550d56..17063cd2cb7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -40,6 +40,7 @@ #include "btrfs_inode.h" #include "ioctl.h" #include "print-tree.h" +#include "volumes.h" struct btrfs_iget_args { u64 ino; @@ -295,20 +296,20 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, return 0; } -int btrfs_submit_bio_hook(int rw, struct bio *bio) +int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) { - struct bio_vec *bvec = bio->bi_io_vec; - struct inode *inode = bvec->bv_page->mapping->host; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; int ret = 0; - if (rw != WRITE) - return 0; + if (rw != WRITE) { + goto mapit; + } if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) - return 0; + btrfs_test_flag(inode, NODATASUM)) { + goto mapit; + } mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); @@ -317,7 +318,8 @@ int btrfs_submit_bio_hook(int rw, struct bio *bio) ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); - return ret; +mapit: + return btrfs_map_bio(root, rw, bio); } int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) @@ -406,7 +408,7 @@ void btrfs_read_locked_inode(struct inode *inode) struct btrfs_path *path; struct extent_buffer *leaf; struct btrfs_inode_item *inode_item; - struct btrfs_inode_timespec *tspec; + struct btrfs_timespec *tspec; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_key location; u64 alloc_group_block; @@ -455,7 +457,8 @@ void btrfs_read_locked_inode(struct inode *inode) BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); if (!BTRFS_I(inode)->block_group) { BTRFS_I(inode)->block_group = btrfs_find_block_group(root, - NULL, 0, 0, 0); + NULL, 0, + BTRFS_BLOCK_GROUP_METADATA, 0); } btrfs_free_path(path); inode_item = NULL; @@ -1550,7 +1553,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 0; else owner = 1; - group = btrfs_find_block_group(root, group, 0, 0, owner); + group = btrfs_find_block_group(root, group, 0, + BTRFS_BLOCK_GROUP_METADATA, owner); BTRFS_I(inode)->block_group = group; BTRFS_I(inode)->flags = 0; diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index da0b4dcf361..9c1335dad40 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -20,6 +20,40 @@ #include "disk-io.h" #include "print-tree.h" +static void print_chunk(struct extent_buffer *eb, struct btrfs_chunk *chunk) +{ + int num_stripes = btrfs_chunk_num_stripes(eb, chunk); + int i; + printk("\t\tchunk owner %llu type %llu num_stripes %d\n", + (unsigned long long)btrfs_chunk_owner(eb, chunk), + (unsigned long long)btrfs_chunk_type(eb, chunk), + num_stripes); + for (i = 0 ; i < num_stripes ; i++) { + printk("\t\t\tstripe %d devid %llu offset %llu\n", i, + (unsigned long long)btrfs_stripe_devid_nr(eb, chunk, i), + (unsigned long long)btrfs_stripe_offset_nr(eb, chunk, i)); + } +} +static void print_dev_item(struct extent_buffer *eb, + struct btrfs_dev_item *dev_item) +{ + char *name; + int name_len; + + name_len = btrfs_device_name_len(eb, dev_item); + name = kmalloc(name_len, GFP_NOFS); + if (name) { + read_extent_buffer(eb, name, + (unsigned long)btrfs_device_name(dev_item), + name_len); + } + printk("\t\tdev item name %.*s devid %llu " + "total_bytes %llu bytes used %Lu\n", name_len, name, + (unsigned long long)btrfs_device_id(eb, dev_item), + (unsigned long long)btrfs_device_total_bytes(eb, dev_item), + (unsigned long long)btrfs_device_bytes_used(eb, dev_item)); + kfree(name); +} void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) { int i; @@ -34,6 +68,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) struct btrfs_key key; struct btrfs_key found_key; struct btrfs_extent_ref *ref; + struct btrfs_dev_extent *dev_extent; u32 type; printk("leaf %llu total ptrs %d free space %d\n", @@ -106,6 +141,19 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) printk("\t\tblock group used %llu\n", (unsigned long long)btrfs_disk_block_group_used(l, bi)); break; + case BTRFS_CHUNK_ITEM_KEY: + print_chunk(l, btrfs_item_ptr(l, i, struct btrfs_chunk)); + break; + case BTRFS_DEV_ITEM_KEY: + print_dev_item(l, btrfs_item_ptr(l, i, + struct btrfs_dev_item)); + break; + case BTRFS_DEV_EXTENT_KEY: + dev_extent = btrfs_item_ptr(l, i, + struct btrfs_dev_extent); + printk("\t\tdev extent owner %llu length %llu\n", + (unsigned long long)btrfs_dev_extent_owner(l, dev_extent), + (unsigned long long)btrfs_dev_extent_length(l, dev_extent)); }; } } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index e9a0983897f..5e9f69244f9 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -198,29 +198,42 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, return werr; } -int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, - struct btrfs_root *root) +static int update_cowonly_root(struct btrfs_trans_handle *trans, + struct btrfs_root *root) { int ret; - u64 old_extent_block; - struct btrfs_fs_info *fs_info = root->fs_info; - struct btrfs_root *tree_root = fs_info->tree_root; - struct btrfs_root *extent_root = fs_info->extent_root; + u64 old_root_bytenr; + struct btrfs_root *tree_root = root->fs_info->tree_root; - btrfs_write_dirty_block_groups(trans, extent_root); + btrfs_write_dirty_block_groups(trans, root); while(1) { - old_extent_block = btrfs_root_bytenr(&extent_root->root_item); - if (old_extent_block == extent_root->node->start) + old_root_bytenr = btrfs_root_bytenr(&root->root_item); + if (old_root_bytenr == root->node->start) break; - btrfs_set_root_bytenr(&extent_root->root_item, - extent_root->node->start); - btrfs_set_root_level(&extent_root->root_item, - btrfs_header_level(extent_root->node)); + btrfs_set_root_bytenr(&root->root_item, + root->node->start); + btrfs_set_root_level(&root->root_item, + btrfs_header_level(root->node)); ret = btrfs_update_root(trans, tree_root, - &extent_root->root_key, - &extent_root->root_item); + &root->root_key, + &root->root_item); BUG_ON(ret); - btrfs_write_dirty_block_groups(trans, extent_root); + btrfs_write_dirty_block_groups(trans, root); + } + return 0; +} + +int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + struct btrfs_fs_info *fs_info = root->fs_info; + struct list_head *next; + + while(!list_empty(&fs_info->dirty_cowonly_roots)) { + next = fs_info->dirty_cowonly_roots.next; + list_del_init(next); + root = list_entry(next, struct btrfs_root, dirty_list); + update_cowonly_root(trans, root); } return 0; } @@ -616,6 +629,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, unsigned long timeout = 1; struct btrfs_transaction *cur_trans; struct btrfs_transaction *prev_trans = NULL; + struct btrfs_root *chunk_root = root->fs_info->chunk_root; struct list_head dirty_fs_roots; struct extent_io_tree *pinned_copy; DEFINE_WAIT(wait); @@ -714,6 +728,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_set_super_root_level(&root->fs_info->super_copy, btrfs_header_level(root->fs_info->tree_root->node)); + btrfs_set_super_chunk_root(&root->fs_info->super_copy, + chunk_root->node->start); + btrfs_set_super_chunk_root_level(&root->fs_info->super_copy, + btrfs_header_level(chunk_root->node)); write_extent_buffer(root->fs_info->sb_buffer, &root->fs_info->super_copy, 0, sizeof(root->fs_info->super_copy)); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c new file mode 100644 index 00000000000..90a8d45dc6d --- /dev/null +++ b/fs/btrfs/volumes.c @@ -0,0 +1,852 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ +#include +#include +#include "ctree.h" +#include "extent_map.h" +#include "disk-io.h" +#include "transaction.h" +#include "print-tree.h" +#include "volumes.h" + +struct map_lookup { + struct btrfs_device *dev; + u64 physical; +}; + +/* + * this uses a pretty simple search, the expectation is that it is + * called very infrequently and that a given device has a small number + * of extents + */ +static int find_free_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, + struct btrfs_path *path, + u64 num_bytes, u64 *start) +{ + struct btrfs_key key; + struct btrfs_root *root = device->dev_root; + struct btrfs_dev_extent *dev_extent = NULL; + u64 hole_size = 0; + u64 last_byte = 0; + u64 search_start = 0; + u64 search_end = device->total_bytes; + int ret; + int slot = 0; + int start_found; + struct extent_buffer *l; + + start_found = 0; + path->reada = 2; + + /* FIXME use last free of some kind */ + + key.objectid = device->devid; + key.offset = search_start; + key.type = BTRFS_DEV_EXTENT_KEY; + ret = btrfs_search_slot(trans, root, &key, path, 0, 0); + if (ret < 0) + goto error; + ret = btrfs_previous_item(root, path, 0, key.type); + if (ret < 0) + goto error; + l = path->nodes[0]; + btrfs_item_key_to_cpu(l, &key, path->slots[0]); + while (1) { + l = path->nodes[0]; + slot = path->slots[0]; + if (slot >= btrfs_header_nritems(l)) { + ret = btrfs_next_leaf(root, path); + if (ret == 0) + continue; + if (ret < 0) + goto error; +no_more_items: + if (!start_found) { + if (search_start >= search_end) { + ret = -ENOSPC; + goto error; + } + *start = search_start; + start_found = 1; + goto check_pending; + } + *start = last_byte > search_start ? + last_byte : search_start; + if (search_end <= *start) { + ret = -ENOSPC; + goto error; + } + goto check_pending; + } + btrfs_item_key_to_cpu(l, &key, slot); + + if (key.objectid < device->devid) + goto next; + + if (key.objectid > device->devid) + goto no_more_items; + + if (key.offset >= search_start && key.offset > last_byte && + start_found) { + if (last_byte < search_start) + last_byte = search_start; + hole_size = key.offset - last_byte; + if (key.offset > last_byte && + hole_size >= num_bytes) { + *start = last_byte; + goto check_pending; + } + } + if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { + goto next; + } + + start_found = 1; + dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); + last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); +next: + path->slots[0]++; + cond_resched(); + } +check_pending: + /* we have to make sure we didn't find an extent that has already + * been allocated by the map tree or the original allocation + */ + btrfs_release_path(root, path); + BUG_ON(*start < search_start); + + if (*start + num_bytes >= search_end) { + ret = -ENOSPC; + goto error; + } + /* check for pending inserts here */ + return 0; + +error: + btrfs_release_path(root, path); + return ret; +} + +int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, + u64 owner, u64 num_bytes, u64 *start) +{ + int ret; + struct btrfs_path *path; + struct btrfs_root *root = device->dev_root; + struct btrfs_dev_extent *extent; + struct extent_buffer *leaf; + struct btrfs_key key; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = find_free_dev_extent(trans, device, path, num_bytes, start); + if (ret) + goto err; + + key.objectid = device->devid; + key.offset = *start; + key.type = BTRFS_DEV_EXTENT_KEY; + ret = btrfs_insert_empty_item(trans, root, path, &key, + sizeof(*extent)); + BUG_ON(ret); + + leaf = path->nodes[0]; + extent = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_dev_extent); + btrfs_set_dev_extent_owner(leaf, extent, owner); + btrfs_set_dev_extent_length(leaf, extent, num_bytes); + btrfs_mark_buffer_dirty(leaf); +err: + btrfs_free_path(path); + return ret; +} + +static int find_next_chunk(struct btrfs_root *root, u64 *objectid) +{ + struct btrfs_path *path; + int ret; + struct btrfs_key key; + struct btrfs_key found_key; + + path = btrfs_alloc_path(); + BUG_ON(!path); + + key.objectid = (u64)-1; + key.offset = (u64)-1; + key.type = BTRFS_CHUNK_ITEM_KEY; + + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto error; + + BUG_ON(ret == 0); + + ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); + if (ret) { + *objectid = 0; + } else { + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + *objectid = found_key.objectid + found_key.offset; + } + ret = 0; +error: + btrfs_free_path(path); + return ret; +} + +static struct btrfs_device *next_device(struct list_head *head, + struct list_head *last) +{ + struct list_head *next = last->next; + struct btrfs_device *dev; + + if (list_empty(head)) + return NULL; + + if (next == head) + next = next->next; + + dev = list_entry(next, struct btrfs_device, dev_list); + return dev; +} + +static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, + u64 *objectid) +{ + int ret; + struct btrfs_key key; + struct btrfs_key found_key; + + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.type = BTRFS_DEV_ITEM_KEY; + key.offset = (u64)-1; + + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto error; + + BUG_ON(ret == 0); + + ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, + BTRFS_DEV_ITEM_KEY); + if (ret) { + *objectid = 1; + } else { + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + *objectid = found_key.offset + 1; + } + ret = 0; +error: + btrfs_release_path(root, path); + return ret; +} + +/* + * the device information is stored in the chunk root + * the btrfs_device struct should be fully filled in + */ +int btrfs_add_device(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_device *device) +{ + int ret; + struct btrfs_path *path; + struct btrfs_dev_item *dev_item; + struct extent_buffer *leaf; + struct btrfs_key key; + unsigned long ptr; + u64 free_devid; + + root = root->fs_info->chunk_root; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = find_next_devid(root, path, &free_devid); + if (ret) + goto out; + + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.type = BTRFS_DEV_ITEM_KEY; + key.offset = free_devid; + + ret = btrfs_insert_empty_item(trans, root, path, &key, + sizeof(*dev_item) + device->name_len); + if (ret) + goto out; + + leaf = path->nodes[0]; + dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); + + btrfs_set_device_id(leaf, dev_item, device->devid); + btrfs_set_device_type(leaf, dev_item, device->type); + btrfs_set_device_io_align(leaf, dev_item, device->io_align); + btrfs_set_device_io_width(leaf, dev_item, device->io_width); + btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); + btrfs_set_device_rdev(leaf, dev_item, device->rdev); + btrfs_set_device_partition(leaf, dev_item, device->partition); + btrfs_set_device_name_len(leaf, dev_item, device->name_len); + btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); + btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); + + ptr = (unsigned long)btrfs_device_name(dev_item); + write_extent_buffer(leaf, device->name, ptr, device->name_len); + + ptr = (unsigned long)btrfs_device_uuid(dev_item); + write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); + btrfs_mark_buffer_dirty(leaf); + ret = 0; + +out: + btrfs_free_path(path); + return ret; +} +int btrfs_update_device(struct btrfs_trans_handle *trans, + struct btrfs_device *device) +{ + int ret; + struct btrfs_path *path; + struct btrfs_root *root; + struct btrfs_dev_item *dev_item; + struct extent_buffer *leaf; + struct btrfs_key key; + + root = device->dev_root->fs_info->chunk_root; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.type = BTRFS_DEV_ITEM_KEY; + key.offset = device->devid; + + ret = btrfs_search_slot(trans, root, &key, path, 0, 1); + if (ret < 0) + goto out; + + if (ret > 0) { + ret = -ENOENT; + goto out; + } + + leaf = path->nodes[0]; + dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); + + btrfs_set_device_id(leaf, dev_item, device->devid); + btrfs_set_device_type(leaf, dev_item, device->type); + btrfs_set_device_io_align(leaf, dev_item, device->io_align); + btrfs_set_device_io_width(leaf, dev_item, device->io_width); + btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); + btrfs_set_device_rdev(leaf, dev_item, device->rdev); + btrfs_set_device_partition(leaf, dev_item, device->partition); + btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); + btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); + btrfs_mark_buffer_dirty(leaf); + +out: + btrfs_free_path(path); + return ret; +} + +int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_key *key, + struct btrfs_chunk *chunk, int item_size) +{ + struct btrfs_super_block *super_copy = &root->fs_info->super_copy; + struct btrfs_disk_key disk_key; + u32 array_size; + u8 *ptr; + + array_size = btrfs_super_sys_array_size(super_copy); + if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) + return -EFBIG; + + ptr = super_copy->sys_chunk_array + array_size; + btrfs_cpu_key_to_disk(&disk_key, key); + memcpy(ptr, &disk_key, sizeof(disk_key)); + ptr += sizeof(disk_key); + memcpy(ptr, chunk, item_size); + item_size += sizeof(disk_key); + btrfs_set_super_sys_array_size(super_copy, array_size + item_size); + return 0; +} + +int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, u64 *start, + u64 *num_bytes, u32 type) +{ + u64 dev_offset; + struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; + struct btrfs_stripe *stripes; + struct btrfs_device *device = NULL; + struct btrfs_chunk *chunk; + struct list_head *dev_list = &extent_root->fs_info->devices; + struct list_head *last_dev = extent_root->fs_info->last_device; + struct extent_map_tree *em_tree; + struct map_lookup *map; + struct extent_map *em; + u64 physical; + u64 calc_size = 1024 * 1024 * 1024; + int num_stripes; + int ret; + int index = 0; + struct btrfs_key key; + + + ret = find_next_chunk(chunk_root, &key.objectid); + if (ret) + return ret; + + num_stripes = 1; + chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); + if (!chunk) + return -ENOMEM; + + stripes = &chunk->stripe; + + *num_bytes = calc_size; + while(index < num_stripes) { + device = next_device(dev_list, last_dev); + BUG_ON(!device); + last_dev = &device->dev_list; + extent_root->fs_info->last_device = last_dev; + + ret = btrfs_alloc_dev_extent(trans, device, + key.objectid, + calc_size, &dev_offset); + BUG_ON(ret); + + device->bytes_used += calc_size; + ret = btrfs_update_device(trans, device); + BUG_ON(ret); + + btrfs_set_stack_stripe_devid(stripes + index, device->devid); + btrfs_set_stack_stripe_offset(stripes + index, dev_offset); + physical = dev_offset; + index++; + } + + /* key.objectid was set above */ + key.offset = *num_bytes; + key.type = BTRFS_CHUNK_ITEM_KEY; + btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); + btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024); + btrfs_set_stack_chunk_type(chunk, type); + btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); + btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize); + btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize); + btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); + + ret = btrfs_insert_item(trans, chunk_root, &key, chunk, + btrfs_chunk_item_size(num_stripes)); + BUG_ON(ret); + *start = key.objectid; + + em = alloc_extent_map(GFP_NOFS); + if (!em) + return -ENOMEM; + map = kmalloc(sizeof(*map), GFP_NOFS); + if (!map) { + free_extent_map(em); + return -ENOMEM; + } + + em->bdev = (struct block_device *)map; + em->start = key.objectid; + em->len = key.offset; + em->block_start = 0; + + map->physical = physical; + map->dev = device; + + if (!map->dev) { + kfree(map); + free_extent_map(em); + return -EIO; + } + kfree(chunk); + + em_tree = &extent_root->fs_info->mapping_tree.map_tree; + spin_lock(&em_tree->lock); + ret = add_extent_mapping(em_tree, em); + BUG_ON(ret); + spin_unlock(&em_tree->lock); + free_extent_map(em); + return ret; +} + +void btrfs_mapping_init(struct btrfs_mapping_tree *tree) +{ + extent_map_tree_init(&tree->map_tree, GFP_NOFS); +} + +void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) +{ + struct extent_map *em; + + while(1) { + spin_lock(&tree->map_tree.lock); + em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); + if (em) + remove_extent_mapping(&tree->map_tree, em); + spin_unlock(&tree->map_tree.lock); + if (!em) + break; + kfree(em->bdev); + /* once for us */ + free_extent_map(em); + /* once for the tree */ + free_extent_map(em); + } +} + +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, + u64 logical, u64 *phys, u64 *length, + struct btrfs_device **dev) +{ + struct extent_map *em; + struct map_lookup *map; + struct extent_map_tree *em_tree = &map_tree->map_tree; + u64 offset; + + + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, logical, *length); + BUG_ON(!em); + + BUG_ON(em->start > logical || em->start + em->len < logical); + map = (struct map_lookup *)em->bdev; + offset = logical - em->start; + *phys = map->physical + offset; + *length = em->len - offset; + *dev = map->dev; + free_extent_map(em); + spin_unlock(&em_tree->lock); + return 0; +} + +int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) +{ + struct btrfs_mapping_tree *map_tree; + struct btrfs_device *dev; + u64 logical = bio->bi_sector << 9; + u64 physical; + u64 length = 0; + u64 map_length; + struct bio_vec *bvec; + int i; + int ret; + + bio_for_each_segment(bvec, bio, i) { + length += bvec->bv_len; + } + map_tree = &root->fs_info->mapping_tree; + map_length = length; + ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); + BUG_ON(map_length < length); + bio->bi_sector = physical >> 9; + bio->bi_bdev = dev->bdev; + submit_bio(rw, bio); + return 0; +} + +struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid) +{ + struct btrfs_device *dev; + struct list_head *cur = root->fs_info->devices.next; + struct list_head *head = &root->fs_info->devices; + + while(cur != head) { + dev = list_entry(cur, struct btrfs_device, dev_list); + if (dev->devid == devid) + return dev; + cur = cur->next; + } + return NULL; +} + +static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, + struct extent_buffer *leaf, + struct btrfs_chunk *chunk) +{ + struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; + struct map_lookup *map; + struct extent_map *em; + u64 logical; + u64 length; + u64 devid; + int ret; + + logical = key->objectid; + length = key->offset; + spin_lock(&map_tree->map_tree.lock); + em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); + + /* already mapped? */ + if (em && em->start <= logical && em->start + em->len > logical) { + free_extent_map(em); + spin_unlock(&map_tree->map_tree.lock); + return 0; + } else if (em) { + free_extent_map(em); + } + spin_unlock(&map_tree->map_tree.lock); + + map = kzalloc(sizeof(*map), GFP_NOFS); + if (!map) + return -ENOMEM; + + em = alloc_extent_map(GFP_NOFS); + if (!em) + return -ENOMEM; + map = kmalloc(sizeof(*map), GFP_NOFS); + if (!map) { + free_extent_map(em); + return -ENOMEM; + } + + em->bdev = (struct block_device *)map; + em->start = logical; + em->len = length; + em->block_start = 0; + + map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0); + devid = btrfs_stripe_devid_nr(leaf, chunk, 0); + map->dev = btrfs_find_device(root, devid); + if (!map->dev) { + kfree(map); + free_extent_map(em); + return -EIO; + } + + spin_lock(&map_tree->map_tree.lock); + ret = add_extent_mapping(&map_tree->map_tree, em); + BUG_ON(ret); + spin_unlock(&map_tree->map_tree.lock); + free_extent_map(em); + + return 0; +} + +static int fill_device_from_item(struct extent_buffer *leaf, + struct btrfs_dev_item *dev_item, + struct btrfs_device *device) +{ + unsigned long ptr; + char *name; + + device->devid = btrfs_device_id(leaf, dev_item); + device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); + device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); + device->type = btrfs_device_type(leaf, dev_item); + device->io_align = btrfs_device_io_align(leaf, dev_item); + device->io_width = btrfs_device_io_width(leaf, dev_item); + device->sector_size = btrfs_device_sector_size(leaf, dev_item); + device->rdev = btrfs_device_rdev(leaf, dev_item); + device->partition = btrfs_device_partition(leaf, dev_item); + device->name_len = btrfs_device_name_len(leaf, dev_item); + + ptr = (unsigned long)btrfs_device_uuid(dev_item); + read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); + + name = kmalloc(device->name_len + 1, GFP_NOFS); + if (!name) + return -ENOMEM; + device->name = name; + ptr = (unsigned long)btrfs_device_name(dev_item); + read_extent_buffer(leaf, name, ptr, device->name_len); + name[device->name_len] = '\0'; + return 0; +} + +static int read_one_dev(struct btrfs_root *root, struct btrfs_key *key, + struct extent_buffer *leaf, + struct btrfs_dev_item *dev_item) +{ + struct btrfs_device *device; + u64 devid; + int ret; + + devid = btrfs_device_id(leaf, dev_item); + if (btrfs_find_device(root, devid)) + return 0; + + device = kmalloc(sizeof(*device), GFP_NOFS); + if (!device) + return -ENOMEM; + + fill_device_from_item(leaf, dev_item, device); + device->dev_root = root->fs_info->dev_root; + device->bdev = root->fs_info->sb->s_bdev; + list_add(&device->dev_list, &root->fs_info->devices); + memcpy(&device->dev_key, key, sizeof(*key)); + ret = 0; +#if 0 + ret = btrfs_open_device(device); + if (ret) { + kfree(device); + } +#endif + return ret; +} + +int btrfs_read_sys_array(struct btrfs_root *root) +{ + struct btrfs_super_block *super_copy = &root->fs_info->super_copy; + struct extent_buffer *sb = root->fs_info->sb_buffer; + struct btrfs_disk_key *disk_key; + struct btrfs_dev_item *dev_item; + struct btrfs_chunk *chunk; + struct btrfs_key key; + u32 num_stripes; + u32 array_size; + u32 len = 0; + u8 *ptr; + unsigned long sb_ptr; + u32 cur; + int ret; + int dev_only = 1; + + array_size = btrfs_super_sys_array_size(super_copy); + + /* + * we do this loop twice, once for the device items and + * once for all of the chunks. This way there are device + * structs filled in for every chunk + */ +again: + ptr = super_copy->sys_chunk_array; + sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); + cur = 0; + + while (cur < array_size) { + disk_key = (struct btrfs_disk_key *)ptr; + btrfs_disk_key_to_cpu(&key, disk_key); + + len = sizeof(*disk_key); + ptr += len; + sb_ptr += len; + cur += len; + + if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID && + key.type == BTRFS_DEV_ITEM_KEY) { + dev_item = (struct btrfs_dev_item *)sb_ptr; + if (dev_only) { + ret = read_one_dev(root, &key, sb, dev_item); + BUG_ON(ret); + } + len = sizeof(*dev_item); + len += btrfs_device_name_len(sb, dev_item); + } else if (key.type == BTRFS_CHUNK_ITEM_KEY) { + + chunk = (struct btrfs_chunk *)sb_ptr; + if (!dev_only) { + ret = read_one_chunk(root, &key, sb, chunk); + BUG_ON(ret); + } + num_stripes = btrfs_chunk_num_stripes(sb, chunk); + len = btrfs_chunk_item_size(num_stripes); + } else { + BUG(); + } + ptr += len; + sb_ptr += len; + cur += len; + } + if (dev_only == 1) { + dev_only = 0; + goto again; + } + return 0; +} + +int btrfs_read_chunk_tree(struct btrfs_root *root) +{ + struct btrfs_path *path; + struct extent_buffer *leaf; + struct btrfs_key key; + struct btrfs_key found_key; + int ret; + int slot; + + root = root->fs_info->chunk_root; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + /* first we search for all of the device items, and then we + * read in all of the chunk items. This way we can create chunk + * mappings that reference all of the devices that are afound + */ + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.offset = 0; + key.type = 0; +again: + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + while(1) { + leaf = path->nodes[0]; + slot = path->slots[0]; + if (slot >= btrfs_header_nritems(leaf)) { + ret = btrfs_next_leaf(root, path); + if (ret == 0) + continue; + if (ret < 0) + goto error; + break; + } + btrfs_item_key_to_cpu(leaf, &found_key, slot); + if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { + if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) + break; + if (found_key.type == BTRFS_DEV_ITEM_KEY) { + struct btrfs_dev_item *dev_item; + dev_item = btrfs_item_ptr(leaf, slot, + struct btrfs_dev_item); + ret = read_one_dev(root, &found_key, leaf, + dev_item); + BUG_ON(ret); + } + } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { + struct btrfs_chunk *chunk; + chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); + ret = read_one_chunk(root, &found_key, leaf, chunk); + } + path->slots[0]++; + } + if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { + key.objectid = 0; + btrfs_release_path(root, path); + goto again; + } + + btrfs_free_path(path); + ret = 0; +error: + return ret; +} + diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h new file mode 100644 index 00000000000..4a47dcb00a7 --- /dev/null +++ b/fs/btrfs/volumes.h @@ -0,0 +1,78 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_VOLUMES_ +#define __BTRFS_VOLUMES_ +struct btrfs_device { + struct list_head dev_list; + struct btrfs_root *dev_root; + struct btrfs_key dev_key; + + struct block_device *bdev; + + /* the internal btrfs device id */ + u64 devid; + + /* size of the device */ + u64 total_bytes; + + /* bytes used */ + u64 bytes_used; + + /* optimal io alignment for this device */ + u32 io_align; + + /* optimal io width for this device */ + u32 io_width; + + /* minimal io size for this device */ + u32 sector_size; + + /* the kernel device number */ + u64 rdev; + + /* type and info about this device */ + u64 type; + + /* partition number, 0 for whole dev */ + int partition; + + /* length of the name data at the end of the item */ + int name_len; + + /* physical drive uuid (or lvm uuid) */ + u8 uuid[BTRFS_DEV_UUID_SIZE]; + + char *name; +}; + +int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, + u64 owner, u64 num_bytes, u64 *start); +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, + u64 logical, u64 *phys, u64 *length, + struct btrfs_device **dev); +int btrfs_read_sys_array(struct btrfs_root *root); +int btrfs_read_chunk_tree(struct btrfs_root *root); +int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, u64 *start, + u64 *num_bytes, u32 type); +void btrfs_mapping_init(struct btrfs_mapping_tree *tree); +void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); +int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio); +#endif -- cgit v1.2.3-70-g09d2 From 6324fbf334f4586325057197da7752f4ffa409d3 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 24 Mar 2008 15:01:59 -0400 Subject: Btrfs: Dynamic chunk and block group allocation Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 12 ++- fs/btrfs/disk-io.c | 2 +- fs/btrfs/extent-tree.c | 208 ++++++++++++++++++++++++++++++++++++++++++++----- fs/btrfs/inode.c | 9 ++- fs/btrfs/volumes.c | 87 +++++++++++++-------- fs/btrfs/volumes.h | 2 +- 6 files changed, 263 insertions(+), 57 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 1453d995fef..a877105f5c4 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -431,9 +431,19 @@ struct btrfs_block_group_item { __le64 flags; } __attribute__ ((__packed__)); +struct btrfs_space_info { + u64 flags; + u64 total_bytes; + u64 bytes_used; + u64 bytes_pinned; + int full; + struct list_head list; +}; + struct btrfs_block_group_cache { struct btrfs_key key; struct btrfs_block_group_item item; + struct btrfs_space_info *space_info; u64 pinned; u64 flags; int cached; @@ -490,7 +500,7 @@ struct btrfs_fs_info { struct list_head dirty_cowonly_roots; struct list_head devices; - struct list_head *last_device; + struct list_head space_info; spinlock_t delalloc_lock; spinlock_t new_trans_lock; u64 delalloc_bytes; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 8e37fa120cc..2a239ae49f7 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -700,8 +700,8 @@ struct btrfs_root *open_ctree(struct super_block *sb) fs_info->dev_root = dev_root; INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); INIT_LIST_HEAD(&fs_info->devices); + INIT_LIST_HEAD(&fs_info->space_info); btrfs_mapping_init(&fs_info->mapping_tree); - fs_info->last_device = &fs_info->devices; fs_info->sb = sb; fs_info->throttles = 0; fs_info->mount_opt = 0; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 2cd957d6e8d..15082b1087b 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -36,6 +36,10 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); static int del_pending_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); +int btrfs_make_block_group(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytes_used, + u64 type, u64 chunk_tree, u64 chunk_objectid, + u64 size); static int cache_block_group(struct btrfs_root *root, @@ -168,16 +172,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) { - if ((bits & BLOCK_GROUP_DATA) && - (cache->flags & BTRFS_BLOCK_GROUP_DATA)) - return 1; - if ((bits & BLOCK_GROUP_METADATA) && - (cache->flags & BTRFS_BLOCK_GROUP_METADATA)) - return 1; - if ((bits & BLOCK_GROUP_SYSTEM) && - (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM)) - return 1; - return 0; + return (cache->flags & bits); } static int noinline find_search_start(struct btrfs_root *root, @@ -276,6 +271,18 @@ static u64 div_factor(u64 num, int factor) return num; } +static int block_group_state_bits(u64 flags) +{ + int bits = 0; + if (flags & BTRFS_BLOCK_GROUP_DATA) + bits |= BLOCK_GROUP_DATA; + if (flags & BTRFS_BLOCK_GROUP_METADATA) + bits |= BLOCK_GROUP_METADATA; + if (flags & BTRFS_BLOCK_GROUP_SYSTEM) + bits |= BLOCK_GROUP_SYSTEM; + return bits; +} + struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *hint, u64 search_start, @@ -304,7 +311,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, if (!owner) factor = 8; - bit = data; + bit = block_group_state_bits(data); if (search_start && search_start < total_fs_bytes) { struct btrfs_block_group_cache *shint; @@ -358,10 +365,15 @@ again: free_check = cache->key.offset; else free_check = div_factor(cache->key.offset, factor); + if (used + cache->pinned < free_check) { found_group = cache; goto found; } + if (full_search) { + printk("failed on cache %Lu used %Lu total %Lu\n", + cache->key.objectid, used, cache->key.offset); + } cond_resched(); } if (!full_search) { @@ -983,6 +995,58 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, return werr; } +static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, + u64 flags) +{ + struct list_head *head = &info->space_info; + struct list_head *cur; + struct btrfs_space_info *found; + list_for_each(cur, head) { + found = list_entry(cur, struct btrfs_space_info, list); + if (found->flags == flags) + return found; + } + return NULL; + +} + +static int do_chunk_alloc(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, u64 alloc_bytes, + u64 flags) +{ + struct btrfs_space_info *space_info; + u64 thresh; + u64 start; + u64 num_bytes; + int ret; + + space_info = __find_space_info(extent_root->fs_info, flags); + BUG_ON(!space_info); + + if (space_info->full) + return 0; + + thresh = div_factor(space_info->total_bytes, 7); + if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < + thresh) + return 0; + + ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); + if (ret == -ENOSPC) { +printk("space info full %Lu\n", flags); + space_info->full = 1; + return 0; + } + + BUG_ON(ret); + + ret = btrfs_make_block_group(trans, extent_root, 0, flags, + extent_root->fs_info->chunk_root->root_key.objectid, + start, num_bytes); + BUG_ON(ret); + return 0; +} + static int update_block_group(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, int alloc, @@ -1012,8 +1076,10 @@ static int update_block_group(struct btrfs_trans_handle *trans, num_bytes = min(total, cache->key.offset - byte_in_group); if (alloc) { old_val += num_bytes; + cache->space_info->bytes_used += num_bytes; } else { old_val -= num_bytes; + cache->space_info->bytes_used -= num_bytes; if (mark_free) { set_extent_dirty(&info->free_space_cache, bytenr, bytenr + num_bytes - 1, @@ -1026,6 +1092,7 @@ static int update_block_group(struct btrfs_trans_handle *trans, } return 0; } + static int update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin) { @@ -1047,9 +1114,11 @@ static int update_pinned_extents(struct btrfs_root *root, (bytenr - cache->key.objectid)); if (pin) { cache->pinned += len; + cache->space_info->bytes_pinned += len; fs_info->total_pinned += len; } else { cache->pinned -= len; + cache->space_info->bytes_pinned -= len; fs_info->total_pinned -= len; } bytenr += len; @@ -1472,7 +1541,7 @@ check_failed: goto new_group; } - if (!(data & BLOCK_GROUP_DATA)) { + if (!(data & BTRFS_BLOCK_GROUP_DATA)) { block_group = btrfs_lookup_block_group(info, ins->objectid); if (block_group) trans->block_group = block_group; @@ -1532,12 +1601,25 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; struct btrfs_key keys[2]; - if (data) - data = BLOCK_GROUP_DATA; - else if (root == root->fs_info->chunk_root) - data = BLOCK_GROUP_SYSTEM; - else - data = BLOCK_GROUP_METADATA; + if (data) { + data = BTRFS_BLOCK_GROUP_DATA; + } else if (root == root->fs_info->chunk_root) { + data = BTRFS_BLOCK_GROUP_SYSTEM; + } else { + data = BTRFS_BLOCK_GROUP_METADATA; + } + + if (root->ref_cows) { + if (data != BTRFS_BLOCK_GROUP_METADATA) { + ret = do_chunk_alloc(trans, root->fs_info->extent_root, + num_bytes, + BTRFS_BLOCK_GROUP_METADATA); + BUG_ON(ret); + } + ret = do_chunk_alloc(trans, root->fs_info->extent_root, + num_bytes, data); + BUG_ON(ret); + } new_hint = max(hint_byte, root->fs_info->alloc_start); if (new_hint < btrfs_super_total_bytes(&info->super_copy)) @@ -2490,6 +2572,34 @@ error: return ret; } +static int update_space_info(struct btrfs_fs_info *info, u64 flags, + u64 total_bytes, u64 bytes_used, + struct btrfs_space_info **space_info) +{ + struct btrfs_space_info *found; + + found = __find_space_info(info, flags); + if (found) { + found->total_bytes += total_bytes; + found->bytes_used += bytes_used; + WARN_ON(found->total_bytes < found->bytes_used); + *space_info = found; + return 0; + } + found = kmalloc(sizeof(*found), GFP_NOFS); + if (!found) + return -ENOMEM; + + list_add(&found->list, &info->space_info); + found->flags = flags; + found->total_bytes = total_bytes; + found->bytes_used = bytes_used; + found->bytes_pinned = 0; + found->full = 0; + *space_info = found; + return 0; +} + int btrfs_read_block_groups(struct btrfs_root *root) { struct btrfs_path *path; @@ -2497,6 +2607,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) int bit; struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; + struct btrfs_space_info *space_info; struct extent_io_tree *block_group_cache; struct btrfs_key key; struct btrfs_key found_key; @@ -2547,6 +2658,12 @@ int btrfs_read_block_groups(struct btrfs_root *root) bit = BLOCK_GROUP_METADATA; } + ret = update_space_info(info, cache->flags, found_key.offset, + btrfs_block_group_used(&cache->item), + &space_info); + BUG_ON(ret); + cache->space_info = space_info; + /* use EXTENT_LOCKED to prevent merging */ set_extent_bits(block_group_cache, found_key.objectid, found_key.objectid + found_key.offset - 1, @@ -2563,3 +2680,58 @@ error: btrfs_free_path(path); return ret; } + +int btrfs_make_block_group(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytes_used, + u64 type, u64 chunk_tree, u64 chunk_objectid, + u64 size) +{ + int ret; + int bit = 0; + struct btrfs_root *extent_root; + struct btrfs_block_group_cache *cache; + struct extent_io_tree *block_group_cache; + + extent_root = root->fs_info->extent_root; + block_group_cache = &root->fs_info->block_group_cache; + + cache = kmalloc(sizeof(*cache), GFP_NOFS); + BUG_ON(!cache); + cache->key.objectid = chunk_objectid; + cache->key.offset = size; + cache->cached = 0; + cache->pinned = 0; + btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); + memset(&cache->item, 0, sizeof(cache->item)); + btrfs_set_block_group_used(&cache->item, bytes_used); + btrfs_set_block_group_chunk_tree(&cache->item, chunk_tree); + btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); + cache->flags = type; + btrfs_set_block_group_flags(&cache->item, type); + + ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, + &cache->space_info); + BUG_ON(ret); + + if (type & BTRFS_BLOCK_GROUP_DATA) { + bit = BLOCK_GROUP_DATA; + } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) { + bit = BLOCK_GROUP_SYSTEM; + } else if (type & BTRFS_BLOCK_GROUP_METADATA) { + bit = BLOCK_GROUP_METADATA; + } + set_extent_bits(block_group_cache, chunk_objectid, + chunk_objectid + size - 1, + bit | EXTENT_LOCKED, GFP_NOFS); + set_state_private(block_group_cache, chunk_objectid, + (unsigned long)cache); + + ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, + sizeof(cache->item)); + BUG_ON(ret); + + finish_current_insert(trans, extent_root); + ret = del_pending_extents(trans, extent_root); + BUG_ON(ret); + return 0; +} diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 17063cd2cb7..109576b57f6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1527,6 +1527,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, { struct inode *inode; struct btrfs_inode_item *inode_item; + struct btrfs_block_group_cache *new_inode_group; struct btrfs_key *location; struct btrfs_path *path; struct btrfs_inode_ref *ref; @@ -1553,9 +1554,13 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 0; else owner = 1; - group = btrfs_find_block_group(root, group, 0, + new_inode_group = btrfs_find_block_group(root, group, 0, BTRFS_BLOCK_GROUP_METADATA, owner); - BTRFS_I(inode)->block_group = group; + if (!new_inode_group) { + printk("find_block group failed\n"); + new_inode_group = group; + } + BTRFS_I(inode)->block_group = new_inode_group; BTRFS_I(inode)->flags = 0; key[0].objectid = objectid; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 90a8d45dc6d..a52a13f365d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -131,7 +131,7 @@ check_pending: btrfs_release_path(root, path); BUG_ON(*start < search_start); - if (*start + num_bytes >= search_end) { + if (*start + num_bytes > search_end) { ret = -ENOSPC; goto error; } @@ -159,8 +159,9 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, return -ENOMEM; ret = find_free_dev_extent(trans, device, path, num_bytes, start); - if (ret) + if (ret) { goto err; + } key.objectid = device->devid; key.offset = *start; @@ -214,22 +215,6 @@ error: return ret; } -static struct btrfs_device *next_device(struct list_head *head, - struct list_head *last) -{ - struct list_head *next = last->next; - struct btrfs_device *dev; - - if (list_empty(head)) - return NULL; - - if (next == head) - next = next->next; - - dev = list_entry(next, struct btrfs_device, dev_list); - return dev; -} - static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, u64 *objectid) { @@ -397,31 +382,63 @@ int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 *start, - u64 *num_bytes, u32 type) + u64 *num_bytes, u64 type) { u64 dev_offset; struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; struct btrfs_stripe *stripes; struct btrfs_device *device = NULL; struct btrfs_chunk *chunk; + struct list_head private_devs; struct list_head *dev_list = &extent_root->fs_info->devices; - struct list_head *last_dev = extent_root->fs_info->last_device; + struct list_head *cur; struct extent_map_tree *em_tree; struct map_lookup *map; struct extent_map *em; u64 physical; u64 calc_size = 1024 * 1024 * 1024; - int num_stripes; + u64 avail; + u64 max_avail = 0; + int num_stripes = 1; + int looped = 0; int ret; - int index = 0; + int index; struct btrfs_key key; + if (list_empty(dev_list)) + return -ENOSPC; +again: + INIT_LIST_HEAD(&private_devs); + cur = dev_list->next; + index = 0; + /* build a private list of devices we will allocate from */ + while(index < num_stripes) { + device = list_entry(cur, struct btrfs_device, dev_list); + avail = device->total_bytes - device->bytes_used; + cur = cur->next; + if (avail > max_avail) + max_avail = avail; + if (avail >= calc_size) { + list_move_tail(&device->dev_list, &private_devs); + index++; + } + if (cur == dev_list) + break; + } + if (index < num_stripes) { + list_splice(&private_devs, dev_list); + if (!looped && max_avail > 0) { + looped = 1; + calc_size = max_avail; + goto again; + } + return -ENOSPC; + } ret = find_next_chunk(chunk_root, &key.objectid); if (ret) return ret; - num_stripes = 1; chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); if (!chunk) return -ENOMEM; @@ -429,11 +446,12 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, stripes = &chunk->stripe; *num_bytes = calc_size; + index = 0; while(index < num_stripes) { - device = next_device(dev_list, last_dev); - BUG_ON(!device); - last_dev = &device->dev_list; - extent_root->fs_info->last_device = last_dev; + BUG_ON(list_empty(&private_devs)); + cur = private_devs.next; + device = list_entry(cur, struct btrfs_device, dev_list); + list_move_tail(&device->dev_list, dev_list); ret = btrfs_alloc_dev_extent(trans, device, key.objectid, @@ -449,6 +467,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, physical = dev_offset; index++; } + BUG_ON(!list_empty(&private_devs)); /* key.objectid was set above */ key.offset = *num_bytes; @@ -692,17 +711,17 @@ static int read_one_dev(struct btrfs_root *root, struct btrfs_key *key, int ret; devid = btrfs_device_id(leaf, dev_item); - if (btrfs_find_device(root, devid)) - return 0; - - device = kmalloc(sizeof(*device), GFP_NOFS); - if (!device) - return -ENOMEM; + device = btrfs_find_device(root, devid); + if (!device) { + device = kmalloc(sizeof(*device), GFP_NOFS); + if (!device) + return -ENOMEM; + list_add(&device->dev_list, &root->fs_info->devices); + } fill_device_from_item(leaf, dev_item, device); device->dev_root = root->fs_info->dev_root; device->bdev = root->fs_info->sb->s_bdev; - list_add(&device->dev_list, &root->fs_info->devices); memcpy(&device->dev_key, key, sizeof(*key)); ret = 0; #if 0 diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 4a47dcb00a7..4ab9891878a 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -71,7 +71,7 @@ int btrfs_read_sys_array(struct btrfs_root *root); int btrfs_read_chunk_tree(struct btrfs_root *root); int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 *start, - u64 *num_bytes, u32 type); + u64 *num_bytes, u64 type); void btrfs_mapping_init(struct btrfs_mapping_tree *tree); void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio); -- cgit v1.2.3-70-g09d2 From 239b14b32dc39232ebf9cce29ff77c4c564355fd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 24 Mar 2008 15:02:07 -0400 Subject: Btrfs: Bring back mount -o ssd optimizations Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 +++ fs/btrfs/disk-io.c | 2 ++ fs/btrfs/extent-tree.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/extent_io.c | 2 ++ fs/btrfs/extent_io.h | 2 ++ fs/btrfs/inode.c | 29 +++++++++++++++++++++++++ fs/btrfs/volumes.c | 5 +++++ fs/btrfs/volumes.h | 3 +++ 8 files changed, 103 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 96a49321786..acf22ad6115 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1405,6 +1405,9 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +int btrfs_merge_bio_hook(struct page *page, unsigned long offset, + size_t size, struct bio *bio); + static inline void dec_i_blocks(struct inode *inode, u64 dec) { dec = dec >> 9; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 26185d46712..4890151cd68 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1103,4 +1103,6 @@ int btrfs_read_buffer(struct extent_buffer *buf) static struct extent_io_ops btree_extent_io_ops = { .writepage_io_hook = btree_writepage_io_hook, .submit_bio_hook = btree_submit_bio_hook, + /* note we're sharing with inode.c for the merge bio hook */ + .merge_bio_hook = btrfs_merge_bio_hook, }; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 14eb8fc8701..e9ef644ff56 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1473,13 +1473,31 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root * root = orig_root->fs_info->extent_root; struct btrfs_fs_info *info = root->fs_info; u64 total_needed = num_bytes; + u64 *last_ptr = NULL; struct btrfs_block_group_cache *block_group; int full_scan = 0; int wrapped = 0; + int empty_cluster = 2 * 1024 * 1024; WARN_ON(num_bytes < root->sectorsize); btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); + if (data & BTRFS_BLOCK_GROUP_METADATA) { + last_ptr = &root->fs_info->last_alloc; + } + + if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { + last_ptr = &root->fs_info->last_data_alloc; + } + + if (last_ptr) { + if (*last_ptr) + hint_byte = *last_ptr; + else { + empty_size += empty_cluster; + } + } + if (search_end == (u64)-1) search_end = btrfs_super_total_bytes(&info->super_copy); @@ -1489,11 +1507,14 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, hint_byte = search_start; block_group = btrfs_find_block_group(root, block_group, hint_byte, data, 1); + if (last_ptr && *last_ptr == 0 && block_group) + hint_byte = block_group->key.objectid; } else { block_group = btrfs_find_block_group(root, trans->block_group, search_start, data, 1); } + search_start = max(search_start, hint_byte); total_needed += empty_size; @@ -1506,9 +1527,36 @@ check_failed: } ret = find_search_start(root, &block_group, &search_start, total_needed, data); + if (ret == -ENOSPC && last_ptr && *last_ptr) { + *last_ptr = 0; + block_group = btrfs_lookup_block_group(info, + orig_search_start); + search_start = orig_search_start; + ret = find_search_start(root, &block_group, &search_start, + total_needed, data); + } + if (ret == -ENOSPC) + goto enospc; if (ret) goto error; + if (last_ptr && *last_ptr && search_start != *last_ptr) { + *last_ptr = 0; + if (!empty_size) { + empty_size += empty_cluster; + total_needed += empty_size; + } + block_group = btrfs_lookup_block_group(info, + orig_search_start); + search_start = orig_search_start; + ret = find_search_start(root, &block_group, + &search_start, total_needed, data); + if (ret == -ENOSPC) + goto enospc; + if (ret) + goto error; + } + search_start = stripe_align(root, search_start); ins->objectid = search_start; ins->offset = num_bytes; @@ -1547,6 +1595,13 @@ check_failed: trans->block_group = block_group; } ins->offset = num_bytes; + if (last_ptr) { + *last_ptr = ins->objectid + ins->offset; + if (*last_ptr == + btrfs_super_total_bytes(&root->fs_info->super_copy)) { + *last_ptr = 0; + } + } return 0; new_group: @@ -1612,12 +1667,12 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, if (root->ref_cows) { if (data != BTRFS_BLOCK_GROUP_METADATA) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, - num_bytes, + 2 * 1024 * 1024, BTRFS_BLOCK_GROUP_METADATA); BUG_ON(ret); } ret = do_chunk_alloc(trans, root->fs_info->extent_root, - num_bytes, data); + num_bytes + 2 * 1024 * 1024, data); BUG_ON(ret); } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7e3a1ebde9f..6dab664529c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1730,6 +1730,8 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (bio_ret && *bio_ret) { bio = *bio_ret; if (bio->bi_sector + (bio->bi_size >> 9) != sector || + (tree->ops && tree->ops->merge_bio_hook && + tree->ops->merge_bio_hook(page, offset, size, bio)) || bio_add_page(bio, page, size, offset) < size) { ret = submit_one_bio(rw, bio); bio = NULL; diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 0dca89328f9..8b5319db251 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -29,6 +29,8 @@ struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio); + int (*merge_bio_hook)(struct page *page, unsigned long offset, + size_t size, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 109576b57f6..5140d680184 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -296,6 +296,34 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, return 0; } +int btrfs_merge_bio_hook(struct page *page, unsigned long offset, + size_t size, struct bio *bio) +{ + struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; + struct btrfs_mapping_tree *map_tree; + struct btrfs_device *dev; + u64 logical = bio->bi_sector << 9; + u64 physical; + u64 length = 0; + u64 map_length; + struct bio_vec *bvec; + int i; + int ret; + + bio_for_each_segment(bvec, bio, i) { + length += bvec->bv_len; + } + map_tree = &root->fs_info->mapping_tree; + map_length = length; + ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); + if (map_length < length + size) { + printk("merge bio hook logical %Lu bio len %Lu physical %Lu " + "len %Lu\n", logical, length, physical, map_length); + return 1; + } + return 0; +} + int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -3033,6 +3061,7 @@ static struct file_operations btrfs_dir_file_operations = { static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, .submit_bio_hook = btrfs_submit_bio_hook, + .merge_bio_hook = btrfs_merge_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .set_bit_hook = btrfs_set_bit_hook, diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index ae22d01ecf5..16fb6bbe6e2 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -578,6 +578,11 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) map_tree = &root->fs_info->mapping_tree; map_length = length; ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); + if (map_length < length) { + printk("mapping failed logical %Lu bio len %Lu physical %Lu " + "len %Lu\n", logical, length, physical, map_length); + BUG(); + } BUG_ON(map_length < length); bio->bi_sector = physical >> 9; bio->bi_bdev = dev->bdev; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 77fa6efd79c..20259128152 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -64,4 +64,7 @@ void btrfs_mapping_init(struct btrfs_mapping_tree *tree); void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio); int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf); +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, + u64 logical, u64 *phys, u64 *length, + struct btrfs_device **dev); #endif -- cgit v1.2.3-70-g09d2 From 593060d756e0c2382d59cf00d5f3b9e3a336c408 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 25 Mar 2008 16:50:33 -0400 Subject: Btrfs: Implement raid0 when multiple devices are present Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 ++ fs/btrfs/disk-io.c | 2 + fs/btrfs/extent-tree.c | 102 +++++++++++++++++++++++++------------- fs/btrfs/inode.c | 2 - fs/btrfs/volumes.c | 130 +++++++++++++++++++++++++++++++++++++------------ 5 files changed, 173 insertions(+), 66 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 7556f8319c6..f43dfe7ffe6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -413,6 +413,7 @@ struct btrfs_csum_item { #define BTRFS_BLOCK_GROUP_DATA (1 << 0) #define BTRFS_BLOCK_GROUP_SYSTEM (1 << 1) #define BTRFS_BLOCK_GROUP_METADATA (1 << 2) +#define BTRFS_BLOCK_GROUP_RAID0 (1 << 3) struct btrfs_block_group_item { @@ -498,6 +499,8 @@ struct btrfs_fs_info { u64 delalloc_bytes; u64 last_alloc; u64 last_data_alloc; + int extra_data_alloc_bits; + int extra_alloc_bits; }; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index f971a29e4f2..d8474796dea 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -736,6 +736,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->total_pinned = 0; fs_info->last_alloc = 0; fs_info->last_data_alloc = 0; + fs_info->extra_alloc_bits = 0; + fs_info->extra_data_alloc_bits = 0; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index e9ef644ff56..c226656f29b 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -172,7 +172,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) { - return (cache->flags & bits); + return (cache->flags & bits) == bits; } static int noinline find_search_start(struct btrfs_root *root, @@ -1010,6 +1010,35 @@ static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, } +static int update_space_info(struct btrfs_fs_info *info, u64 flags, + u64 total_bytes, u64 bytes_used, + struct btrfs_space_info **space_info) +{ + struct btrfs_space_info *found; + + found = __find_space_info(info, flags); + if (found) { + found->total_bytes += total_bytes; + found->bytes_used += bytes_used; + WARN_ON(found->total_bytes < found->bytes_used); + *space_info = found; + return 0; + } + found = kmalloc(sizeof(*found), GFP_NOFS); + if (!found) + return -ENOMEM; + + list_add(&found->list, &info->space_info); + found->flags = flags; + found->total_bytes = total_bytes; + found->bytes_used = bytes_used; + found->bytes_pinned = 0; + found->full = 0; + *space_info = found; + return 0; +} + + static int do_chunk_alloc(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 alloc_bytes, u64 flags) @@ -1021,6 +1050,11 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, int ret; space_info = __find_space_info(extent_root->fs_info, flags); + if (!space_info) { + ret = update_space_info(extent_root->fs_info, flags, + 0, 0, &space_info); + BUG_ON(ret); + } BUG_ON(!space_info); if (space_info->full) @@ -1044,6 +1078,17 @@ printk("space info full %Lu\n", flags); extent_root->fs_info->chunk_root->root_key.objectid, start, num_bytes); BUG_ON(ret); + + if (flags & BTRFS_BLOCK_GROUP_RAID0) { + if (flags & BTRFS_BLOCK_GROUP_DATA) { + extent_root->fs_info->extra_data_alloc_bits = + BTRFS_BLOCK_GROUP_RAID0; + } + if (flags & BTRFS_BLOCK_GROUP_METADATA) { + extent_root->fs_info->extra_alloc_bits = + BTRFS_BLOCK_GROUP_RAID0; + } + } return 0; } @@ -1655,24 +1700,31 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_extent_ref *ref; struct btrfs_path *path; struct btrfs_key keys[2]; + int extra_chunk_alloc_bits = 0; if (data) { - data = BTRFS_BLOCK_GROUP_DATA; + data = BTRFS_BLOCK_GROUP_DATA | info->extra_data_alloc_bits; } else if (root == root->fs_info->chunk_root) { data = BTRFS_BLOCK_GROUP_SYSTEM; } else { - data = BTRFS_BLOCK_GROUP_METADATA; + data = BTRFS_BLOCK_GROUP_METADATA | info->extra_alloc_bits; } + if (btrfs_super_num_devices(&info->super_copy) > 1 && + !(data & BTRFS_BLOCK_GROUP_SYSTEM)) + extra_chunk_alloc_bits = BTRFS_BLOCK_GROUP_RAID0; if (root->ref_cows) { - if (data != BTRFS_BLOCK_GROUP_METADATA) { + if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, 2 * 1024 * 1024, - BTRFS_BLOCK_GROUP_METADATA); + BTRFS_BLOCK_GROUP_METADATA | + info->extra_alloc_bits | + extra_chunk_alloc_bits); BUG_ON(ret); } ret = do_chunk_alloc(trans, root->fs_info->extent_root, - num_bytes + 2 * 1024 * 1024, data); + num_bytes + 2 * 1024 * 1024, data | + extra_chunk_alloc_bits); BUG_ON(ret); } @@ -2627,34 +2679,6 @@ error: return ret; } -static int update_space_info(struct btrfs_fs_info *info, u64 flags, - u64 total_bytes, u64 bytes_used, - struct btrfs_space_info **space_info) -{ - struct btrfs_space_info *found; - - found = __find_space_info(info, flags); - if (found) { - found->total_bytes += total_bytes; - found->bytes_used += bytes_used; - WARN_ON(found->total_bytes < found->bytes_used); - *space_info = found; - return 0; - } - found = kmalloc(sizeof(*found), GFP_NOFS); - if (!found) - return -ENOMEM; - - list_add(&found->list, &info->space_info); - found->flags = flags; - found->total_bytes = total_bytes; - found->bytes_used = bytes_used; - found->bytes_pinned = 0; - found->full = 0; - *space_info = found; - return 0; -} - int btrfs_read_block_groups(struct btrfs_root *root) { struct btrfs_path *path; @@ -2712,6 +2736,16 @@ int btrfs_read_block_groups(struct btrfs_root *root) } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { bit = BLOCK_GROUP_METADATA; } + if (cache->flags & BTRFS_BLOCK_GROUP_RAID0) { + if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { + info->extra_data_alloc_bits = + BTRFS_BLOCK_GROUP_RAID0; + } + if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { + info->extra_alloc_bits = + BTRFS_BLOCK_GROUP_RAID0; + } + } ret = update_space_info(info, cache->flags, found_key.offset, btrfs_block_group_used(&cache->item), diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5140d680184..db60d85598c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -317,8 +317,6 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, map_length = length; ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); if (map_length < length + size) { - printk("merge bio hook logical %Lu bio len %Lu physical %Lu " - "len %Lu\n", logical, length, physical, map_length); return 1; } return 0; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 263f01cc3db..d8fce32a3bb 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -18,6 +18,7 @@ #include #include #include +#include #include "ctree.h" #include "extent_map.h" #include "disk-io.h" @@ -25,10 +26,24 @@ #include "print-tree.h" #include "volumes.h" -struct map_lookup { +struct stripe { struct btrfs_device *dev; u64 physical; }; + +struct map_lookup { + u64 type; + int io_align; + int io_width; + int stripe_len; + int sector_size; + int num_stripes; + struct stripe stripes[]; +}; + +#define map_lookup_size(n) (sizeof(struct map_lookup) + \ + (sizeof(struct stripe) * (n))) + static DEFINE_MUTEX(uuid_mutex); static LIST_HEAD(fs_uuids); @@ -592,6 +607,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 *num_bytes, u64 type) { u64 dev_offset; + struct btrfs_fs_info *info = extent_root->fs_info; struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; struct btrfs_stripe *stripes; struct btrfs_device *device = NULL; @@ -610,10 +626,18 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, int looped = 0; int ret; int index; + int stripe_len = 64 * 1024; struct btrfs_key key; if (list_empty(dev_list)) return -ENOSPC; + + if (type & BTRFS_BLOCK_GROUP_RAID0) + num_stripes = btrfs_super_num_devices(&info->super_copy); + if (type & BTRFS_BLOCK_GROUP_DATA) + stripe_len = 64 * 1024; + if (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_SYSTEM)) + stripe_len = 32 * 1024; again: INIT_LIST_HEAD(&private_devs); cur = dev_list->next; @@ -650,9 +674,15 @@ again: if (!chunk) return -ENOMEM; + map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); + if (!map) { + kfree(chunk); + return -ENOMEM; + } + stripes = &chunk->stripe; - *num_bytes = calc_size; + *num_bytes = calc_size * num_stripes; index = 0; while(index < num_stripes) { BUG_ON(list_empty(&private_devs)); @@ -669,6 +699,8 @@ printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid); ret = btrfs_update_device(trans, device); BUG_ON(ret); + map->stripes[index].dev = device; + map->stripes[index].physical = dev_offset; btrfs_set_stack_stripe_devid(stripes + index, device->devid); btrfs_set_stack_stripe_offset(stripes + index, dev_offset); physical = dev_offset; @@ -680,12 +712,18 @@ printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid); key.offset = *num_bytes; key.type = BTRFS_CHUNK_ITEM_KEY; btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); - btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024); + btrfs_set_stack_chunk_stripe_len(chunk, stripe_len); btrfs_set_stack_chunk_type(chunk, type); btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); - btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize); - btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize); + btrfs_set_stack_chunk_io_align(chunk, stripe_len); + btrfs_set_stack_chunk_io_width(chunk, stripe_len); btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); + map->sector_size = extent_root->sectorsize; + map->stripe_len = stripe_len; + map->io_align = stripe_len; + map->io_width = stripe_len; + map->type = type; + map->num_stripes = num_stripes; ret = btrfs_insert_item(trans, chunk_root, &key, chunk, btrfs_chunk_item_size(num_stripes)); @@ -695,25 +733,11 @@ printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid); em = alloc_extent_map(GFP_NOFS); if (!em) return -ENOMEM; - map = kmalloc(sizeof(*map), GFP_NOFS); - if (!map) { - free_extent_map(em); - return -ENOMEM; - } - em->bdev = (struct block_device *)map; em->start = key.objectid; em->len = key.offset; em->block_start = 0; - map->physical = physical; - map->dev = device; - - if (!map->dev) { - kfree(map); - free_extent_map(em); - return -EIO; - } kfree(chunk); em_tree = &extent_root->fs_info->mapping_tree.map_tree; @@ -758,6 +782,9 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, struct map_lookup *map; struct extent_map_tree *em_tree = &map_tree->map_tree; u64 offset; + u64 stripe_offset; + u64 stripe_nr; + int stripe_index; spin_lock(&em_tree->lock); @@ -767,9 +794,40 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, BUG_ON(em->start > logical || em->start + em->len < logical); map = (struct map_lookup *)em->bdev; offset = logical - em->start; - *phys = map->physical + offset; - *length = em->len - offset; - *dev = map->dev; + + stripe_nr = offset; + /* + * stripe_nr counts the total number of stripes we have to stride + * to get to this block + */ + do_div(stripe_nr, map->stripe_len); + + stripe_offset = stripe_nr * map->stripe_len; + BUG_ON(offset < stripe_offset); + + /* stripe_offset is the offset of this block in its stripe*/ + stripe_offset = offset - stripe_offset; + + /* + * after this do_div call, stripe_nr is the number of stripes + * on this device we have to walk to find the data, and + * stripe_index is the number of our device in the stripe array + */ + stripe_index = do_div(stripe_nr, map->num_stripes); + + BUG_ON(stripe_index >= map->num_stripes); + + *phys = map->stripes[stripe_index].physical + stripe_offset + + stripe_nr * map->stripe_len; + + if (map->type & BTRFS_BLOCK_GROUP_RAID0) { + /* we limit the length of each bio to what fits in a stripe */ + *length = min_t(u64, em->len - offset, + map->stripe_len - stripe_offset); + } else { + *length = em->len - offset; + } + *dev = map->stripes[stripe_index].dev; free_extent_map(em); spin_unlock(&em_tree->lock); return 0; @@ -822,7 +880,9 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, u64 logical; u64 length; u64 devid; + int num_stripes; int ret; + int i; logical = key->objectid; length = key->offset; @@ -846,7 +906,8 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, em = alloc_extent_map(GFP_NOFS); if (!em) return -ENOMEM; - map = kmalloc(sizeof(*map), GFP_NOFS); + num_stripes = btrfs_chunk_num_stripes(leaf, chunk); + map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); if (!map) { free_extent_map(em); return -ENOMEM; @@ -857,13 +918,22 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, em->len = length; em->block_start = 0; - map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0); - devid = btrfs_stripe_devid_nr(leaf, chunk, 0); - map->dev = btrfs_find_device(root, devid); - if (!map->dev) { - kfree(map); - free_extent_map(em); - return -EIO; + map->num_stripes = num_stripes; + map->io_width = btrfs_chunk_io_width(leaf, chunk); + map->io_align = btrfs_chunk_io_align(leaf, chunk); + map->sector_size = btrfs_chunk_sector_size(leaf, chunk); + map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); + map->type = btrfs_chunk_type(leaf, chunk); + for (i = 0; i < num_stripes; i++) { + map->stripes[i].physical = + btrfs_stripe_offset_nr(leaf, chunk, i); + devid = btrfs_stripe_devid_nr(leaf, chunk, i); + map->stripes[i].dev = btrfs_find_device(root, devid); + if (!map->stripes[i].dev) { + kfree(map); + free_extent_map(em); + return -EIO; + } } spin_lock(&map_tree->map_tree.lock); -- cgit v1.2.3-70-g09d2 From 04160088141a6be06d5cb9e53e34076b51803ffc Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 26 Mar 2008 10:28:07 -0400 Subject: Create a btrfs backing dev info This allows intelligent versions of unplug and congestion functions Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 ++ fs/btrfs/disk-io.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/inode.c | 5 +++++ 3 files changed, 60 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f43dfe7ffe6..fec96ba7c23 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -25,6 +25,7 @@ #include #include #include +#include #include #include "bit-radix.h" #include "extent_io.h" @@ -472,6 +473,7 @@ struct btrfs_fs_info { struct block_device *__bdev; struct super_block *sb; struct inode *btree_inode; + struct backing_dev_info bdi; spinlock_t hash_lock; struct mutex trans_mutex; struct mutex fs_mutex; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index d8474796dea..6a70e4916ab 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -655,6 +655,55 @@ static int add_hasher(struct btrfs_fs_info *info, char *type) { return 0; } #endif + +static int btrfs_congested_fn(void *congested_data, int bdi_bits) +{ + struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; + int ret = 0; + struct list_head *cur; + struct btrfs_device *device; + struct backing_dev_info *bdi; + + list_for_each(cur, &info->fs_devices->devices) { + device = list_entry(cur, struct btrfs_device, dev_list); + bdi = blk_get_backing_dev_info(device->bdev); + if (bdi && bdi_congested(bdi, bdi_bits)) { + ret = 1; + break; + } + } + return ret; +} + +void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) +{ + struct list_head *cur; + struct btrfs_device *device; + struct btrfs_fs_info *info; + + info = (struct btrfs_fs_info *)bdi->unplug_io_data; + list_for_each(cur, &info->fs_devices->devices) { + device = list_entry(cur, struct btrfs_device, dev_list); + bdi = blk_get_backing_dev_info(device->bdev); + if (bdi->unplug_io_fn) { + bdi->unplug_io_fn(bdi, page); + } + } +} + +static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) +{ + bdi_init(bdi); + bdi->ra_pages = default_backing_dev_info.ra_pages; + bdi->state = 0; + bdi->capabilities = default_backing_dev_info.capabilities; + bdi->unplug_io_fn = btrfs_unplug_io_fn; + bdi->unplug_io_data = info; + bdi->congested_fn = btrfs_congested_fn; + bdi->congested_data = info; + return 0; +} + struct btrfs_root *open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices) { @@ -708,11 +757,14 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->max_extent = (u64)-1; fs_info->max_inline = 8192 * 1024; fs_info->delalloc_bytes = 0; + setup_bdi(fs_info, &fs_info->bdi); fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size; fs_info->btree_inode->i_mapping->a_ops = &btree_aops; + fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; + extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, fs_info->btree_inode->i_mapping, GFP_NOFS); @@ -992,6 +1044,7 @@ int close_ctree(struct btrfs_root *root) #endif close_all_devices(fs_info); btrfs_mapping_tree_free(&fs_info->mapping_tree); + bdi_destroy(&fs_info->bdi); kfree(fs_info->extent_root); kfree(fs_info->tree_root); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index db60d85598c..0fa7cf227f1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -494,6 +494,7 @@ void btrfs_read_locked_inode(struct inode *inode) switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; + inode->i_mapping->backing_dev_info = &root->fs_info->bdi; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; @@ -508,6 +509,7 @@ void btrfs_read_locked_inode(struct inode *inode) case S_IFLNK: inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; + inode->i_mapping->backing_dev_info = &root->fs_info->bdi; break; default: init_special_inode(inode, inode->i_mode, rdev); @@ -1792,6 +1794,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, drop_inode = 1; else { inode->i_mapping->a_ops = &btrfs_aops; + inode->i_mapping->backing_dev_info = &root->fs_info->bdi; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); @@ -2960,6 +2963,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, drop_inode = 1; else { inode->i_mapping->a_ops = &btrfs_aops; + inode->i_mapping->backing_dev_info = &root->fs_info->bdi; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); @@ -2999,6 +3003,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; + inode->i_mapping->backing_dev_info = &root->fs_info->bdi; inode->i_size = name_len - 1; err = btrfs_update_inode(trans, root, inode); if (err) -- cgit v1.2.3-70-g09d2 From 8790d502e4401a4a3a4175b83a3a47e8d595c771 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 3 Apr 2008 16:29:03 -0400 Subject: Btrfs: Add support for mirroring across drives Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 9 ++- fs/btrfs/disk-io.c | 20 +++---- fs/btrfs/extent-tree.c | 93 +++++++++++++++-------------- fs/btrfs/inode.c | 4 +- fs/btrfs/volumes.c | 154 ++++++++++++++++++++++++++++++++++++++++--------- fs/btrfs/volumes.h | 11 ++-- 6 files changed, 198 insertions(+), 93 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 67d533cf8f4..0a207861472 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -418,6 +418,7 @@ struct btrfs_csum_item { #define BTRFS_BLOCK_GROUP_SYSTEM (1 << 1) #define BTRFS_BLOCK_GROUP_METADATA (1 << 2) #define BTRFS_BLOCK_GROUP_RAID0 (1 << 3) +#define BTRFS_BLOCK_GROUP_RAID1 (1 << 4) struct btrfs_block_group_item { @@ -504,8 +505,12 @@ struct btrfs_fs_info { u64 delalloc_bytes; u64 last_alloc; u64 last_data_alloc; - int extra_data_alloc_bits; - int extra_alloc_bits; + int avail_data_alloc_bits; + int avail_metadata_alloc_bits; + int avail_system_alloc_bits; + int data_alloc_profile; + int metadata_alloc_profile; + int system_alloc_profile; }; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e40fb318ad9..ff75ad58676 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -735,7 +735,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, GFP_NOFS); struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), GFP_NOFS); - struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info), + struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), GFP_NOFS); struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), GFP_NOFS); @@ -744,6 +744,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, int ret; int err = -EINVAL; struct btrfs_super_block *disk_super; + if (!extent_root || !tree_root || !fs_info) { err = -ENOMEM; goto fail; @@ -756,11 +757,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, spin_lock_init(&fs_info->delalloc_lock); spin_lock_init(&fs_info->new_trans_lock); - memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj)); init_completion(&fs_info->kobj_unregister); sb_set_blocksize(sb, 4096); - fs_info->running_transaction = NULL; - fs_info->last_trans_committed = 0; fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; fs_info->chunk_root = chunk_root; @@ -770,11 +768,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->space_info); btrfs_mapping_init(&fs_info->mapping_tree); fs_info->sb = sb; - fs_info->throttles = 0; - fs_info->mount_opt = 0; fs_info->max_extent = (u64)-1; fs_info->max_inline = 8192 * 1024; - fs_info->delalloc_bytes = 0; setup_bdi(fs_info, &fs_info->bdi); fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; @@ -802,12 +797,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, extent_io_tree_init(&fs_info->extent_ins, fs_info->btree_inode->i_mapping, GFP_NOFS); fs_info->do_barriers = 1; - fs_info->closing = 0; - fs_info->total_pinned = 0; - fs_info->last_alloc = 0; - fs_info->last_data_alloc = 0; - fs_info->extra_alloc_bits = 0; - fs_info->extra_data_alloc_bits = 0; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); @@ -923,6 +912,11 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_read_block_groups(extent_root); fs_info->generation = btrfs_super_generation(disk_super) + 1; + if (btrfs_super_num_devices(disk_super) > 0) { + fs_info->data_alloc_profile = BTRFS_BLOCK_GROUP_RAID0; + fs_info->metadata_alloc_profile = BTRFS_BLOCK_GROUP_RAID1; + fs_info->system_alloc_profile = BTRFS_BLOCK_GROUP_RAID0; + } mutex_unlock(&fs_info->fs_mutex); return tree_root; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a34c289aec2..4ab98d8b73f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -230,9 +230,13 @@ again: goto new_group; if (start + num > total_fs_bytes) goto new_group; + if (!block_group_bits(cache, data)) { + printk("block group bits don't match %Lu %Lu\n", cache->flags, data); + } *start_ret = start; return 0; - } out: + } +out: cache = btrfs_lookup_block_group(root->fs_info, search_start); if (!cache) { printk("Unable to find block group for %Lu\n", search_start); @@ -365,14 +369,17 @@ again: if (cache->key.objectid > total_fs_bytes) break; - if (full_search) - free_check = cache->key.offset; - else - free_check = div_factor(cache->key.offset, factor); + if (block_group_bits(cache, data)) { + if (full_search) + free_check = cache->key.offset; + else + free_check = div_factor(cache->key.offset, + factor); - if (used + cache->pinned < free_check) { - found_group = cache; - goto found; + if (used + cache->pinned < free_check) { + found_group = cache; + goto found; + } } cond_resched(); } @@ -1038,6 +1045,19 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, return 0; } +static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) +{ + u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID1); + if (extra_flags) { + if (flags & BTRFS_BLOCK_GROUP_DATA) + fs_info->avail_data_alloc_bits |= extra_flags; + if (flags & BTRFS_BLOCK_GROUP_METADATA) + fs_info->avail_metadata_alloc_bits |= extra_flags; + if (flags & BTRFS_BLOCK_GROUP_SYSTEM) + fs_info->avail_system_alloc_bits |= extra_flags; + } +} static int do_chunk_alloc(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 alloc_bytes, @@ -1060,7 +1080,7 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, if (space_info->full) return 0; - thresh = div_factor(space_info->total_bytes, 7); + thresh = div_factor(space_info->total_bytes, 6); if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < thresh) return 0; @@ -1079,16 +1099,7 @@ printk("space info full %Lu\n", flags); start, num_bytes); BUG_ON(ret); - if (flags & BTRFS_BLOCK_GROUP_RAID0) { - if (flags & BTRFS_BLOCK_GROUP_DATA) { - extent_root->fs_info->extra_data_alloc_bits = - BTRFS_BLOCK_GROUP_RAID0; - } - if (flags & BTRFS_BLOCK_GROUP_METADATA) { - extent_root->fs_info->extra_alloc_bits = - BTRFS_BLOCK_GROUP_RAID0; - } - } + set_avail_alloc_bits(extent_root->fs_info, flags); return 0; } @@ -1529,6 +1540,7 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, if (data & BTRFS_BLOCK_GROUP_METADATA) { last_ptr = &root->fs_info->last_alloc; + empty_cluster = 256 * 1024; } if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { @@ -1693,6 +1705,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 root_used; u64 search_start = 0; u64 new_hint; + u64 alloc_profile; u32 sizes[2]; struct btrfs_fs_info *info = root->fs_info; struct btrfs_root *extent_root = info->extent_root; @@ -1700,31 +1713,32 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_extent_ref *ref; struct btrfs_path *path; struct btrfs_key keys[2]; - int extra_chunk_alloc_bits = 0; if (data) { - data = BTRFS_BLOCK_GROUP_DATA | info->extra_data_alloc_bits; + alloc_profile = info->avail_data_alloc_bits & + info->data_alloc_profile; + data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; } else if (root == root->fs_info->chunk_root) { - data = BTRFS_BLOCK_GROUP_SYSTEM; + alloc_profile = info->avail_system_alloc_bits & + info->system_alloc_profile; + data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; } else { - data = BTRFS_BLOCK_GROUP_METADATA | info->extra_alloc_bits; + alloc_profile = info->avail_metadata_alloc_bits & + info->metadata_alloc_profile; + data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } - if (btrfs_super_num_devices(&info->super_copy) > 1 && - !(data & BTRFS_BLOCK_GROUP_SYSTEM)) - extra_chunk_alloc_bits = BTRFS_BLOCK_GROUP_RAID0; if (root->ref_cows) { if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, 2 * 1024 * 1024, BTRFS_BLOCK_GROUP_METADATA | - info->extra_alloc_bits | - extra_chunk_alloc_bits); + (info->metadata_alloc_profile & + info->avail_metadata_alloc_bits)); BUG_ON(ret); } ret = do_chunk_alloc(trans, root->fs_info->extent_root, - num_bytes + 2 * 1024 * 1024, data | - extra_chunk_alloc_bits); + num_bytes + 2 * 1024 * 1024, data); BUG_ON(ret); } @@ -2046,12 +2060,12 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, if (!next || !btrfs_buffer_uptodate(next)) { free_extent_buffer(next); reada_walk_down(root, cur, path->slots[*level]); + + mutex_unlock(&root->fs_info->fs_mutex); next = read_tree_block(root, bytenr, blocksize); + mutex_lock(&root->fs_info->fs_mutex); - /* we used to drop the lock above, keep the - * code to double check so that we won't forget - * when we drop the lock again in the future - */ + /* we've dropped the lock, double check */ ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); BUG_ON(ret); @@ -2739,16 +2753,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { bit = BLOCK_GROUP_METADATA; } - if (cache->flags & BTRFS_BLOCK_GROUP_RAID0) { - if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { - info->extra_data_alloc_bits = - BTRFS_BLOCK_GROUP_RAID0; - } - if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { - info->extra_alloc_bits = - BTRFS_BLOCK_GROUP_RAID0; - } - } + set_avail_alloc_bits(info, cache->flags); ret = update_space_info(info, cache->flags, found_key.offset, btrfs_block_group_used(&cache->item), diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0fa7cf227f1..a8ae68c6fbb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -306,6 +306,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, u64 physical; u64 length = 0; u64 map_length; + int total_devs; struct bio_vec *bvec; int i; int ret; @@ -315,7 +316,8 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, } map_tree = &root->fs_info->mapping_tree; map_length = length; - ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); + ret = btrfs_map_block(map_tree, READ, 0, logical, &physical, + &map_length, &dev, &total_devs); if (map_length < length + size) { return 1; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 169be0f7285..bc3c0b97588 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -31,6 +31,13 @@ struct stripe { u64 physical; }; +struct multi_bio { + atomic_t stripes; + bio_end_io_t *end_io; + void *private; + int error; +}; + struct map_lookup { u64 type; int io_align; @@ -632,12 +639,12 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, if (list_empty(dev_list)) return -ENOSPC; - if (type & BTRFS_BLOCK_GROUP_RAID0) + if (type & (BTRFS_BLOCK_GROUP_RAID0)) num_stripes = btrfs_super_num_devices(&info->super_copy); - if (type & BTRFS_BLOCK_GROUP_DATA) - stripe_len = 64 * 1024; - if (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_SYSTEM)) - stripe_len = 32 * 1024; + if (type & (BTRFS_BLOCK_GROUP_RAID1)) { + num_stripes = min_t(u64, 2, + btrfs_super_num_devices(&info->super_copy)); + } again: INIT_LIST_HEAD(&private_devs); cur = dev_list->next; @@ -682,7 +689,11 @@ again: stripes = &chunk->stripe; - *num_bytes = calc_size * num_stripes; + if (type & BTRFS_BLOCK_GROUP_RAID1) + *num_bytes = calc_size; + else + *num_bytes = calc_size * num_stripes; + index = 0; while(index < num_stripes) { BUG_ON(list_empty(&private_devs)); @@ -694,7 +705,7 @@ again: key.objectid, calc_size, &dev_offset); BUG_ON(ret); -printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid); +printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.objectid, calc_size, device->devid, type); device->bytes_used += calc_size; ret = btrfs_update_device(trans, device); BUG_ON(ret); @@ -774,9 +785,9 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) } } -int btrfs_map_block(struct btrfs_mapping_tree *map_tree, - u64 logical, u64 *phys, u64 *length, - struct btrfs_device **dev) +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, + int dev_nr, u64 logical, u64 *phys, u64 *length, + struct btrfs_device **dev, int *total_devs) { struct extent_map *em; struct map_lookup *map; @@ -808,19 +819,39 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, /* stripe_offset is the offset of this block in its stripe*/ stripe_offset = offset - stripe_offset; - /* - * after this do_div call, stripe_nr is the number of stripes - * on this device we have to walk to find the data, and - * stripe_index is the number of our device in the stripe array - */ - stripe_index = do_div(stripe_nr, map->num_stripes); - + if (map->type & BTRFS_BLOCK_GROUP_RAID1) { + stripe_index = dev_nr; + if (rw & (1 << BIO_RW)) + *total_devs = map->num_stripes; + else { + int i; + u64 least = (u64)-1; + struct btrfs_device *cur; + + for (i = 0; i < map->num_stripes; i++) { + cur = map->stripes[i].dev; + spin_lock(&cur->io_lock); + if (cur->total_ios < least) { + least = cur->total_ios; + stripe_index = i; + } + spin_unlock(&cur->io_lock); + } + *total_devs = 1; + } + } else { + /* + * after this do_div call, stripe_nr is the number of stripes + * on this device we have to walk to find the data, and + * stripe_index is the number of our device in the stripe array + */ + stripe_index = do_div(stripe_nr, map->num_stripes); + } BUG_ON(stripe_index >= map->num_stripes); - *phys = map->stripes[stripe_index].physical + stripe_offset + stripe_nr * map->stripe_len; - if (map->type & BTRFS_BLOCK_GROUP_RAID0) { + if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1)) { /* we limit the length of each bio to what fits in a stripe */ *length = min_t(u64, em->len - offset, map->stripe_len - stripe_offset); @@ -833,33 +864,98 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, return 0; } +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +static void end_bio_multi_stripe(struct bio *bio, int err) +#else +static int end_bio_multi_stripe(struct bio *bio, + unsigned int bytes_done, int err) +#endif +{ + struct multi_bio *multi = bio->bi_private; + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + if (bio->bi_size) + return 1; +#endif + if (err) + multi->error = err; + + if (atomic_dec_and_test(&multi->stripes)) { + bio->bi_private = multi->private; + bio->bi_end_io = multi->end_io; + + if (!err && multi->error) + err = multi->error; + kfree(multi); + + bio_endio(bio, err); + } else { + bio_put(bio); + } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + return 0; +#endif +} + int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) { struct btrfs_mapping_tree *map_tree; struct btrfs_device *dev; + struct bio *first_bio = bio; u64 logical = bio->bi_sector << 9; u64 physical; u64 length = 0; u64 map_length; struct bio_vec *bvec; + struct multi_bio *multi = NULL; int i; int ret; + int dev_nr = 0; + int total_devs = 1; bio_for_each_segment(bvec, bio, i) { length += bvec->bv_len; } + map_tree = &root->fs_info->mapping_tree; map_length = length; - ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); - if (map_length < length) { - printk("mapping failed logical %Lu bio len %Lu physical %Lu " - "len %Lu\n", logical, length, physical, map_length); - BUG(); + while(dev_nr < total_devs) { + ret = btrfs_map_block(map_tree, rw, dev_nr, logical, + &physical, &map_length, &dev, + &total_devs); + if (map_length < length) { + printk("mapping failed logical %Lu bio len %Lu physical %Lu " + "len %Lu\n", logical, length, physical, map_length); + BUG(); + } + BUG_ON(map_length < length); + if (total_devs > 1) { + if (!multi) { + multi = kmalloc(sizeof(*multi), GFP_NOFS); + atomic_set(&multi->stripes, 1); + multi->end_io = bio->bi_end_io; + multi->private = first_bio->bi_private; + multi->error = 0; + } else { + atomic_inc(&multi->stripes); + } + if (dev_nr < total_devs - 1) { + bio = bio_clone(first_bio, GFP_NOFS); + BUG_ON(!bio); + } else { + bio = first_bio; + } + bio->bi_private = multi; + bio->bi_end_io = end_bio_multi_stripe; + } + bio->bi_sector = physical >> 9; + bio->bi_bdev = dev->bdev; + spin_lock(&dev->io_lock); + dev->total_ios++; + spin_unlock(&dev->io_lock); + submit_bio(rw, bio); + dev_nr++; } - BUG_ON(map_length < length); - bio->bi_sector = physical >> 9; - bio->bi_bdev = dev->bdev; - submit_bio(rw, bio); return 0; } @@ -982,6 +1078,8 @@ static int read_one_dev(struct btrfs_root *root, return -ENOMEM; list_add(&device->dev_list, &root->fs_info->fs_devices->devices); + device->total_ios = 0; + spin_lock_init(&device->io_lock); } fill_device_from_item(leaf, dev_item, device); diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 12f297eb055..35dec3efd78 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -18,12 +18,16 @@ #ifndef __BTRFS_VOLUMES_ #define __BTRFS_VOLUMES_ + struct btrfs_device { struct list_head dev_list; struct btrfs_root *dev_root; + spinlock_t io_lock; struct block_device *bdev; + u64 total_ios; + char *name; /* the internal btrfs device id */ @@ -68,9 +72,9 @@ struct btrfs_fs_devices { int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 owner, u64 num_bytes, u64 *start); -int btrfs_map_block(struct btrfs_mapping_tree *map_tree, +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, int stripe_nr, u64 logical, u64 *phys, u64 *length, - struct btrfs_device **dev); + struct btrfs_device **dev, int *total_stripes); int btrfs_read_sys_array(struct btrfs_root *root); int btrfs_read_chunk_tree(struct btrfs_root *root); int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, @@ -80,9 +84,6 @@ void btrfs_mapping_init(struct btrfs_mapping_tree *tree); void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio); int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf); -int btrfs_map_block(struct btrfs_mapping_tree *map_tree, - u64 logical, u64 *phys, u64 *length, - struct btrfs_device **dev); int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags, void *holder); int btrfs_scan_one_device(const char *path, int flags, void *holder, -- cgit v1.2.3-70-g09d2 From cea9e4452ebaf18dd0951e90dc84d82a5dee40b2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Change btrfs_map_block to return a structure with mappings for all stripes Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 4 +- fs/btrfs/inode.c | 8 ++-- fs/btrfs/volumes.c | 135 +++++++++++++++++++++++++++++------------------------ fs/btrfs/volumes.h | 25 ++++++++-- 4 files changed, 103 insertions(+), 69 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 3c7891cf0d7..c92c6b0ee58 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2025,8 +2025,10 @@ again: root->root_key.objectid, root_gen, disk_key.objectid, 0, l->start, 0); - if (IS_ERR(right)) + if (IS_ERR(right)) { + BUG_ON(1); return PTR_ERR(right); + } memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header)); btrfs_set_header_bytenr(right, right->start); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a8ae68c6fbb..7ae677d8a6d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -301,12 +301,9 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, { struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; struct btrfs_mapping_tree *map_tree; - struct btrfs_device *dev; u64 logical = bio->bi_sector << 9; - u64 physical; u64 length = 0; u64 map_length; - int total_devs; struct bio_vec *bvec; int i; int ret; @@ -316,8 +313,9 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, } map_tree = &root->fs_info->mapping_tree; map_length = length; - ret = btrfs_map_block(map_tree, READ, 0, logical, &physical, - &map_length, &dev, &total_devs); + ret = btrfs_map_block(map_tree, READ, logical, + &map_length, NULL); + if (map_length < length + size) { return 1; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index b9294e3c05f..008d3640e8c 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -26,18 +26,6 @@ #include "print-tree.h" #include "volumes.h" -struct stripe { - struct btrfs_device *dev; - u64 physical; -}; - -struct multi_bio { - atomic_t stripes; - bio_end_io_t *end_io; - void *private; - int error; -}; - struct map_lookup { u64 type; int io_align; @@ -45,11 +33,11 @@ struct map_lookup { int stripe_len; int sector_size; int num_stripes; - struct stripe stripes[]; + struct btrfs_bio_stripe stripes[]; }; #define map_lookup_size(n) (sizeof(struct map_lookup) + \ - (sizeof(struct stripe) * (n))) + (sizeof(struct btrfs_bio_stripe) * (n))) static DEFINE_MUTEX(uuid_mutex); static LIST_HEAD(fs_uuids); @@ -801,8 +789,8 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) } int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, - int dev_nr, u64 logical, u64 *phys, u64 *length, - struct btrfs_device **dev, int *total_devs) + u64 logical, u64 *length, + struct btrfs_multi_bio **multi_ret) { struct extent_map *em; struct map_lookup *map; @@ -810,8 +798,21 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, u64 offset; u64 stripe_offset; u64 stripe_nr; + int stripes_allocated = 8; int stripe_index; + int i; + struct btrfs_multi_bio *multi = NULL; + if (multi_ret && !(rw & (1 << BIO_RW))) { + stripes_allocated = 1; + } +again: + if (multi_ret) { + multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), + GFP_NOFS); + if (!multi) + return -ENOMEM; + } spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, logical, *length); @@ -821,6 +822,17 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, map = (struct map_lookup *)em->bdev; offset = logical - em->start; + /* if our multi bio struct is too small, back off and try again */ + if (multi_ret && (rw & (1 << BIO_RW)) && + stripes_allocated < map->num_stripes && + ((map->type & BTRFS_BLOCK_GROUP_RAID1) || + (map->type & BTRFS_BLOCK_GROUP_DUP))) { + stripes_allocated = map->num_stripes; + spin_unlock(&em_tree->lock); + free_extent_map(em); + kfree(multi); + goto again; + } stripe_nr = offset; /* * stripe_nr counts the total number of stripes we have to stride @@ -834,10 +846,22 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, /* stripe_offset is the offset of this block in its stripe*/ stripe_offset = offset - stripe_offset; + if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_DUP)) { + /* we limit the length of each bio to what fits in a stripe */ + *length = min_t(u64, em->len - offset, + map->stripe_len - stripe_offset); + } else { + *length = em->len - offset; + } + if (!multi_ret) + goto out; + + multi->num_stripes = 1; + stripe_index = 0; if (map->type & BTRFS_BLOCK_GROUP_RAID1) { - stripe_index = dev_nr; if (rw & (1 << BIO_RW)) - *total_devs = map->num_stripes; + multi->num_stripes = map->num_stripes; else { int i; u64 least = (u64)-1; @@ -852,16 +876,10 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, } spin_unlock(&cur->io_lock); } - *total_devs = 1; } } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { - if (rw == WRITE) { - *total_devs = map->num_stripes; - stripe_index = dev_nr; - } else { - stripe_index = 0; - *total_devs = 1; - } + if (rw & (1 << BIO_RW)) + multi->num_stripes = map->num_stripes; } else { /* * after this do_div call, stripe_nr is the number of stripes @@ -871,18 +889,17 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, stripe_index = do_div(stripe_nr, map->num_stripes); } BUG_ON(stripe_index >= map->num_stripes); - *phys = map->stripes[stripe_index].physical + stripe_offset + - stripe_nr * map->stripe_len; - - if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_DUP)) { - /* we limit the length of each bio to what fits in a stripe */ - *length = min_t(u64, em->len - offset, - map->stripe_len - stripe_offset); - } else { - *length = em->len - offset; + BUG_ON(stripe_index != 0 && multi->num_stripes > 1); + + for (i = 0; i < multi->num_stripes; i++) { + multi->stripes[i].physical = + map->stripes[stripe_index].physical + stripe_offset + + stripe_nr * map->stripe_len; + multi->stripes[i].dev = map->stripes[stripe_index].dev; + stripe_index++; } - *dev = map->stripes[stripe_index].dev; + *multi_ret = multi; +out: free_extent_map(em); spin_unlock(&em_tree->lock); return 0; @@ -895,7 +912,7 @@ static int end_bio_multi_stripe(struct bio *bio, unsigned int bytes_done, int err) #endif { - struct multi_bio *multi = bio->bi_private; + struct btrfs_multi_bio *multi = bio->bi_private; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) if (bio->bi_size) @@ -904,7 +921,7 @@ static int end_bio_multi_stripe(struct bio *bio, if (err) multi->error = err; - if (atomic_dec_and_test(&multi->stripes)) { + if (atomic_dec_and_test(&multi->stripes_pending)) { bio->bi_private = multi->private; bio->bi_end_io = multi->end_io; @@ -927,11 +944,10 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) struct btrfs_device *dev; struct bio *first_bio = bio; u64 logical = bio->bi_sector << 9; - u64 physical; u64 length = 0; u64 map_length; struct bio_vec *bvec; - struct multi_bio *multi = NULL; + struct btrfs_multi_bio *multi = NULL; int i; int ret; int dev_nr = 0; @@ -943,26 +959,22 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) map_tree = &root->fs_info->mapping_tree; map_length = length; + + ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi); + BUG_ON(ret); + + total_devs = multi->num_stripes; + if (map_length < length) { + printk("mapping failed logical %Lu bio len %Lu " + "len %Lu\n", logical, length, map_length); + BUG(); + } + multi->end_io = first_bio->bi_end_io; + multi->private = first_bio->bi_private; + atomic_set(&multi->stripes_pending, multi->num_stripes); + while(dev_nr < total_devs) { - ret = btrfs_map_block(map_tree, rw, dev_nr, logical, - &physical, &map_length, &dev, - &total_devs); - if (map_length < length) { - printk("mapping failed logical %Lu bio len %Lu physical %Lu " - "len %Lu\n", logical, length, physical, map_length); - BUG(); - } - BUG_ON(map_length < length); if (total_devs > 1) { - if (!multi) { - multi = kmalloc(sizeof(*multi), GFP_NOFS); - atomic_set(&multi->stripes, 1); - multi->end_io = bio->bi_end_io; - multi->private = first_bio->bi_private; - multi->error = 0; - } else { - atomic_inc(&multi->stripes); - } if (dev_nr < total_devs - 1) { bio = bio_clone(first_bio, GFP_NOFS); BUG_ON(!bio); @@ -972,7 +984,8 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) bio->bi_private = multi; bio->bi_end_io = end_bio_multi_stripe; } - bio->bi_sector = physical >> 9; + bio->bi_sector = multi->stripes[dev_nr].physical >> 9; + dev = multi->stripes[dev_nr].dev; bio->bi_bdev = dev->bdev; spin_lock(&dev->io_lock); dev->total_ios++; @@ -980,6 +993,8 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) submit_bio(rw, bio); dev_nr++; } + if (total_devs == 1) + kfree(multi); return 0; } diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 35dec3efd78..10ca0104750 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -19,6 +19,8 @@ #ifndef __BTRFS_VOLUMES_ #define __BTRFS_VOLUMES_ +#include + struct btrfs_device { struct list_head dev_list; struct btrfs_root *dev_root; @@ -69,12 +71,29 @@ struct btrfs_fs_devices { struct list_head list; }; +struct btrfs_bio_stripe { + struct btrfs_device *dev; + u64 physical; +}; + +struct btrfs_multi_bio { + atomic_t stripes_pending; + bio_end_io_t *end_io; + void *private; + int error; + int num_stripes; + struct btrfs_bio_stripe stripes[]; +}; + +#define btrfs_multi_bio_size(n) (sizeof(struct btrfs_multi_bio) + \ + (sizeof(struct btrfs_bio_stripe) * (n))) + int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 owner, u64 num_bytes, u64 *start); -int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, int stripe_nr, - u64 logical, u64 *phys, u64 *length, - struct btrfs_device **dev, int *total_stripes); +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, + u64 logical, u64 *length, + struct btrfs_multi_bio **multi_ret); int btrfs_read_sys_array(struct btrfs_root *root); int btrfs_read_chunk_tree(struct btrfs_root *root); int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, -- cgit v1.2.3-70-g09d2 From 22c599485b1fdd95e4476a4752596a6cf6c6629a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Handle data block end_io through the async work queue Before it was done by the bio end_io routine, the work queue code is able to scale much better with faster IO subsystems. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 34 +++++++++++++++++++++++----------- fs/btrfs/disk-io.h | 2 ++ fs/btrfs/inode.c | 4 +++- 3 files changed, 28 insertions(+), 12 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 82109204788..94ff87d0eae 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -54,6 +54,7 @@ struct end_io_wq { void *private; struct btrfs_fs_info *info; int error; + int metadata; struct list_head list; }; @@ -308,29 +309,40 @@ static int end_workqueue_bio(struct bio *bio, #endif } -static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) +int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, + int metadata) { - struct btrfs_root *root = BTRFS_I(inode)->root; struct end_io_wq *end_io_wq; - u64 offset; - offset = bio->bi_sector << 9; - - if (rw & (1 << BIO_RW)) { - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio); - } - end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); if (!end_io_wq) return -ENOMEM; end_io_wq->private = bio->bi_private; end_io_wq->end_io = bio->bi_end_io; - end_io_wq->info = root->fs_info; + end_io_wq->info = info; end_io_wq->error = 0; end_io_wq->bio = bio; + end_io_wq->metadata = metadata; bio->bi_private = end_io_wq; bio->bi_end_io = end_workqueue_bio; + return 0; +} + +static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 offset; + int ret; + + offset = bio->bi_sector << 9; + + if (rw & (1 << BIO_RW)) { + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio); + } + + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); + BUG_ON(ret); if (offset == BTRFS_SUPER_INFO_OFFSET) { bio->bi_bdev = root->fs_info->sb->s_bdev; @@ -880,7 +892,7 @@ void btrfs_end_io_csum(struct work_struct *work) end_io_wq = list_entry(next, struct end_io_wq, list); bio = end_io_wq->bio; - if (!bio_ready_for_csum(bio)) { + if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { spin_lock_irqsave(&fs_info->end_io_work_lock, flags); was_empty = list_empty(&fs_info->end_io_work_list); list_add_tail(&end_io_wq->list, diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 05b88d0e75e..4fac0ccbf8f 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -71,4 +71,6 @@ void btrfs_throttle(struct btrfs_root *root); int btrfs_open_device(struct btrfs_device *dev); int btrfs_verify_block_csum(struct btrfs_root *root, struct extent_buffer *buf); +int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, + int metadata); #endif diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7ae677d8a6d..e1ef1acdb35 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -328,7 +328,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) struct btrfs_trans_handle *trans; int ret = 0; - if (rw != WRITE) { + if (!(rw & (1 << BIO_RW))) { + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); + BUG_ON(ret); goto mapit; } -- cgit v1.2.3-70-g09d2 From f188591e987e21b6f7f8864c66a02858b95b530e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Retry metadata reads in the face of checksum failures Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 11 +++++++- fs/btrfs/disk-io.c | 72 ++++++++++++++++++++++++++++++++++++++-------------- fs/btrfs/extent_io.c | 50 +++++++++++++++++++++--------------- fs/btrfs/extent_io.h | 5 ++-- fs/btrfs/inode.c | 7 ++--- fs/btrfs/volumes.c | 39 +++++++++++++++++++++++++--- fs/btrfs/volumes.h | 6 +++-- 7 files changed, 138 insertions(+), 52 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index efce173a935..ff4e9c6859d 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -569,7 +569,16 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, static int noinline check_block(struct btrfs_root *root, struct btrfs_path *path, int level) { - return 0; + u64 found_start; + if (btrfs_header_level(path->nodes[level]) != level) + printk("warning: bad level %Lu wanted %d found %d\n", + path->nodes[level]->start, level, + btrfs_header_level(path->nodes[level])); + found_start = btrfs_header_bytenr(path->nodes[level]); + if (found_start != path->nodes[level]->start) { + printk("warning: bad bytentr %Lu found %Lu\n", + path->nodes[level]->start, found_start); + } #if 0 struct extent_buffer *buf = path->nodes[level]; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 94ff87d0eae..59bdf0474be 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -156,7 +156,6 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, memcpy(&found, result, BTRFS_CRC32_SIZE); read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); - WARN_ON(1); printk("btrfs: %s checksum verify failed on %llu " "wanted %X found %X from_this_trans %d " "level %d\n", @@ -171,6 +170,40 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, return 0; } +static int btree_read_extent_buffer_pages(struct btrfs_root *root, + struct extent_buffer *eb, + u64 start) +{ + struct extent_io_tree *io_tree; + int ret; + int num_copies = 0; + int mirror_num = 0; + + io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; + while (1) { + ret = read_extent_buffer_pages(io_tree, eb, start, 1, + btree_get_extent, mirror_num); + if (!ret) { + if (mirror_num) +printk("good read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies); + return ret; + } + num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, + eb->start, eb->len); +printk("failed to read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies); + if (num_copies == 1) { +printk("reading %Lu failed only one copy\n", eb->start); + return ret; + } + mirror_num++; + if (mirror_num > num_copies) { +printk("bailing at mirror %d of %d\n", mirror_num, num_copies); + return ret; + } + } +printk("read extent buffer page last\n"); + return -EIO; +} int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { @@ -180,6 +213,8 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page) int found_level; unsigned long len; struct extent_buffer *eb; + int ret; + tree = &BTRFS_I(page->mapping->host)->io_tree; if (page->private == EXTENT_PAGE_PRIVATE) @@ -191,8 +226,8 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page) WARN_ON(1); } eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); - read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1, - btree_get_extent); + ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE); + BUG_ON(ret); btrfs_clear_buffer_defrag(eb); found_start = btrfs_header_bytenr(eb); if (found_start != start) { @@ -240,7 +275,7 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, unsigned long len; struct extent_buffer *eb; struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; - int ret; + int ret = 0; tree = &BTRFS_I(page->mapping->host)->io_tree; if (page->private == EXTENT_PAGE_PRIVATE) @@ -252,25 +287,26 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, WARN_ON(1); } eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); - read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1, - btree_get_extent); + btrfs_clear_buffer_defrag(eb); found_start = btrfs_header_bytenr(eb); if (found_start != start) { - printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", - start, found_start, len); - WARN_ON(1); +printk("bad start on %Lu found %Lu\n", eb->start, found_start); + ret = -EIO; goto err; } if (eb->first_page != page) { printk("bad first page %lu %lu\n", eb->first_page->index, page->index); WARN_ON(1); + ret = -EIO; goto err; } found_level = btrfs_header_level(eb); ret = csum_tree_block(root, eb, 1); + if (ret) + ret = -EIO; end = min_t(u64, eb->len, PAGE_CACHE_SIZE); end = eb->start + end - 1; @@ -278,7 +314,7 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, err: free_extent_buffer(eb); out: - return 0; + return ret; } #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) @@ -329,7 +365,8 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, return 0; } -static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) +static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, + int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 offset; @@ -338,7 +375,7 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) offset = bio->bi_sector << 9; if (rw & (1 << BIO_RW)) { - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio); + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); } ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); @@ -349,7 +386,7 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) submit_bio(rw, bio); return 0; } - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio); + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); } static int btree_writepage(struct page *page, struct writeback_control *wbc) @@ -459,7 +496,7 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize) if (!buf) return 0; read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, - buf, 0, 0, btree_get_extent); + buf, 0, 0, btree_get_extent, 0); free_extent_buffer(buf); return ret; } @@ -522,8 +559,7 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, if (!buf) return NULL; - ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, - 1, btree_get_extent); + ret = btree_read_extent_buffer_pages(root, buf, 0); if (ret == 0) { buf->flags |= EXTENT_UPTODATE; @@ -1366,10 +1402,8 @@ int btrfs_clear_buffer_defrag(struct extent_buffer *buf) int btrfs_read_buffer(struct extent_buffer *buf) { struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; - struct inode *btree_inode = root->fs_info->btree_inode; int ret; - ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, - buf, 0, 1, btree_get_extent); + ret = btree_read_extent_buffer_pages(root, buf, 0); if (ret == 0) { buf->flags |= EXTENT_UPTODATE; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index cfc383c17a3..2f159375c87 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1546,7 +1546,7 @@ static int end_bio_extent_readpage(struct bio *bio, !(state->state & EXTENT_LOCKED)) state = NULL; } - if (!state) { + if (!state && uptodate) { spin_unlock_irqrestore(&tree->lock, flags); set_extent_uptodate(tree, start, end, GFP_ATOMIC); @@ -1567,8 +1567,10 @@ static int end_bio_extent_readpage(struct bio *bio, } else { state = NULL; } - set_state_cb(tree, clear, EXTENT_UPTODATE); - clear->state |= EXTENT_UPTODATE; + if (uptodate) { + set_state_cb(tree, clear, EXTENT_UPTODATE); + clear->state |= EXTENT_UPTODATE; + } clear_state_bit(tree, clear, EXTENT_LOCKED, 1, 0); if (cur == start) @@ -1685,7 +1687,7 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, return bio; } -static int submit_one_bio(int rw, struct bio *bio) +static int submit_one_bio(int rw, struct bio *bio, int mirror_num) { u64 maxsector; int ret = 0; @@ -1722,7 +1724,8 @@ static int submit_one_bio(int rw, struct bio *bio) WARN_ON(1); } if (tree->ops && tree->ops->submit_bio_hook) - tree->ops->submit_bio_hook(page->mapping->host, rw, bio); + tree->ops->submit_bio_hook(page->mapping->host, rw, bio, + mirror_num); else submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) @@ -1737,7 +1740,8 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, struct block_device *bdev, struct bio **bio_ret, unsigned long max_pages, - bio_end_io_t end_io_func) + bio_end_io_t end_io_func, + int mirror_num) { int ret = 0; struct bio *bio; @@ -1749,7 +1753,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, (tree->ops && tree->ops->merge_bio_hook && tree->ops->merge_bio_hook(page, offset, size, bio)) || bio_add_page(bio, page, size, offset) < size) { - ret = submit_one_bio(rw, bio); + ret = submit_one_bio(rw, bio, mirror_num); bio = NULL; } else { return 0; @@ -1769,7 +1773,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, if (bio_ret) { *bio_ret = bio; } else { - ret = submit_one_bio(rw, bio); + ret = submit_one_bio(rw, bio, mirror_num); } return ret; @@ -1798,7 +1802,7 @@ void set_page_extent_head(struct page *page, unsigned long len) static int __extent_read_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, - struct bio **bio) + struct bio **bio, int mirror_num) { struct inode *inode = page->mapping->host; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -1901,7 +1905,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree, ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, bdev, bio, nr, - end_bio_extent_readpage); + end_bio_extent_readpage, mirror_num); } if (ret) SetPageError(page); @@ -1923,9 +1927,9 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page, struct bio *bio = NULL; int ret; - ret = __extent_read_full_page(tree, page, get_extent, &bio); + ret = __extent_read_full_page(tree, page, get_extent, &bio, 0); if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, 0); return ret; } EXPORT_SYMBOL(extent_read_full_page); @@ -2077,7 +2081,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, ret = submit_extent_page(WRITE, tree, page, sector, iosize, page_offset, bdev, &epd->bio, max_nr, - end_bio_extent_writepage); + end_bio_extent_writepage, 0); if (ret) SetPageError(page); } @@ -2244,7 +2248,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio); + submit_one_bio(WRITE, epd.bio, 0); } return ret; } @@ -2265,7 +2269,7 @@ int extent_writepages(struct extent_io_tree *tree, ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio); + submit_one_bio(WRITE, epd.bio, 0); } return ret; } @@ -2297,7 +2301,8 @@ int extent_readpages(struct extent_io_tree *tree, page_cache_get(page); if (!pagevec_add(&pvec, page)) __pagevec_lru_add(&pvec); - __extent_read_full_page(tree, page, get_extent, &bio); + __extent_read_full_page(tree, page, get_extent, + &bio, 0); } page_cache_release(page); } @@ -2305,7 +2310,7 @@ int extent_readpages(struct extent_io_tree *tree, __pagevec_lru_add(&pvec); BUG_ON(!list_empty(pages)); if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, 0); return 0; } EXPORT_SYMBOL(extent_readpages); @@ -2430,7 +2435,7 @@ int extent_prepare_write(struct extent_io_tree *tree, ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, em->bdev, NULL, 1, - end_bio_extent_preparewrite); + end_bio_extent_preparewrite, 0); iocount++; block_start = block_start + iosize; } else { @@ -2696,6 +2701,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, mark_page_accessed(page0); set_page_extent_mapped(page0); set_page_extent_head(page0, len); + uptodate = PageUptodate(page0); } else { i = 0; } @@ -3006,7 +3012,7 @@ EXPORT_SYMBOL(extent_buffer_uptodate); int read_extent_buffer_pages(struct extent_io_tree *tree, struct extent_buffer *eb, u64 start, int wait, - get_extent_t *get_extent) + get_extent_t *get_extent, int mirror_num) { unsigned long i; unsigned long start_i; @@ -3062,8 +3068,10 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (!PageUptodate(page)) { if (start_i == 0) inc_all_pages = 1; + ClearPageError(page); err = __extent_read_full_page(tree, page, - get_extent, &bio); + get_extent, &bio, + mirror_num); if (err) { ret = err; } @@ -3073,7 +3081,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, } if (bio) - submit_one_bio(READ, bio); + submit_one_bio(READ, bio, mirror_num); if (ret || !wait) { return ret; diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 9d2991d1d3c..8d6b8a14cc3 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -27,7 +27,8 @@ struct extent_state; struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); - int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio); + int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio, + int mirror_num); int (*merge_bio_hook)(struct page *page, unsigned long offset, size_t size, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); @@ -172,7 +173,7 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, void free_extent_buffer(struct extent_buffer *eb); int read_extent_buffer_pages(struct extent_io_tree *tree, struct extent_buffer *eb, u64 start, int wait, - get_extent_t *get_extent); + get_extent_t *get_extent, int mirror_num); static inline void extent_buffer_get(struct extent_buffer *eb) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e1ef1acdb35..8c2d5d036bd 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -314,7 +314,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, map_tree = &root->fs_info->mapping_tree; map_length = length; ret = btrfs_map_block(map_tree, READ, logical, - &map_length, NULL); + &map_length, NULL, 0); if (map_length < length + size) { return 1; @@ -322,7 +322,8 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, return 0; } -int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) +int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, + int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; @@ -347,7 +348,7 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio) BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); mapit: - return btrfs_map_bio(root, rw, bio); + return btrfs_map_bio(root, rw, bio, mirror_num); } int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 008d3640e8c..3b927f69832 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -788,9 +788,31 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) } } +int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) +{ + struct extent_map *em; + struct map_lookup *map; + struct extent_map_tree *em_tree = &map_tree->map_tree; + int ret; + + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, logical, len); + BUG_ON(!em); + + BUG_ON(em->start > logical || em->start + em->len < logical); + map = (struct map_lookup *)em->bdev; + if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) + ret = map->num_stripes; + else + ret = 1; + free_extent_map(em); + spin_unlock(&em_tree->lock); + return ret; +} + int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, u64 logical, u64 *length, - struct btrfs_multi_bio **multi_ret) + struct btrfs_multi_bio **multi_ret, int mirror_num) { struct extent_map *em; struct map_lookup *map; @@ -822,6 +844,9 @@ again: map = (struct map_lookup *)em->bdev; offset = logical - em->start; + if (mirror_num > map->num_stripes) + mirror_num = 0; + /* if our multi bio struct is too small, back off and try again */ if (multi_ret && (rw & (1 << BIO_RW)) && stripes_allocated < map->num_stripes && @@ -862,7 +887,9 @@ again: if (map->type & BTRFS_BLOCK_GROUP_RAID1) { if (rw & (1 << BIO_RW)) multi->num_stripes = map->num_stripes; - else { + else if (mirror_num) { + stripe_index = mirror_num - 1; + } else { int i; u64 least = (u64)-1; struct btrfs_device *cur; @@ -880,6 +907,8 @@ again: } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { if (rw & (1 << BIO_RW)) multi->num_stripes = map->num_stripes; + else if (mirror_num) + stripe_index = mirror_num - 1; } else { /* * after this do_div call, stripe_nr is the number of stripes @@ -938,7 +967,8 @@ static int end_bio_multi_stripe(struct bio *bio, #endif } -int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) +int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, + int mirror_num) { struct btrfs_mapping_tree *map_tree; struct btrfs_device *dev; @@ -960,7 +990,8 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) map_tree = &root->fs_info->mapping_tree; map_length = length; - ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi); + ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, + mirror_num); BUG_ON(ret); total_devs = multi->num_stripes; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 10ca0104750..3d5d0a9cb82 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -93,7 +93,7 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, u64 owner, u64 num_bytes, u64 *start); int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, u64 logical, u64 *length, - struct btrfs_multi_bio **multi_ret); + struct btrfs_multi_bio **multi_ret, int mirror_num); int btrfs_read_sys_array(struct btrfs_root *root); int btrfs_read_chunk_tree(struct btrfs_root *root); int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, @@ -101,7 +101,8 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 *num_bytes, u64 type); void btrfs_mapping_init(struct btrfs_mapping_tree *tree); void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); -int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio); +int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, + int mirror_num); int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf); int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags, void *holder); @@ -112,4 +113,5 @@ int btrfs_add_device(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_device *device); int btrfs_cleanup_fs_uuids(void); +int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len); #endif -- cgit v1.2.3-70-g09d2 From 7e38326f5b125288a2baea1d815c500502ab9fc0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 9 Apr 2008 16:28:12 -0400 Subject: Btrfs: Handle checksumming errors while reading data blocks Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/extent_io.c | 9 ++++ fs/btrfs/extent_io.h | 3 ++ fs/btrfs/inode.c | 114 ++++++++++++++++++++++++++++++++++++++++++++++++- 4 files changed, 126 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 9b9db9cbc01..fe6ef8e3416 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -29,6 +29,7 @@ struct btrfs_inode { struct btrfs_key location; struct extent_map_tree extent_tree; struct extent_io_tree io_tree; + struct extent_io_tree io_failure_tree; struct inode vfs_inode; u64 ordered_trans; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 2f159375c87..866460c3d72 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1534,6 +1534,15 @@ static int end_bio_extent_readpage(struct bio *bio, if (ret) uptodate = 0; } + if (!uptodate && tree->ops && + tree->ops->readpage_io_failed_hook) { + ret = tree->ops->readpage_io_failed_hook(bio, page, + start, end, state); + if (ret == 0) { + state = NULL; + continue; + } + } spin_lock_irqsave(&tree->lock, flags); if (!state || state->end != end) { diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 8d6b8a14cc3..b47859ccd78 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -32,6 +32,9 @@ struct extent_io_ops { int (*merge_bio_hook)(struct page *page, unsigned long offset, size_t size, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); + int (*readpage_io_failed_hook)(struct bio *bio, struct page *page, + u64 start, u64 end, + struct extent_state *state); int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8c2d5d036bd..48f1d1b9645 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -385,6 +385,86 @@ out: return ret; } +struct io_failure_record { + struct page *page; + u64 start; + u64 len; + u64 logical; + int last_mirror; +}; + +int btrfs_readpage_io_failed_hook(struct bio *failed_bio, + struct page *page, u64 start, u64 end, + struct extent_state *state) +{ + struct io_failure_record *failrec = NULL; + u64 private; + struct extent_map *em; + struct inode *inode = page->mapping->host; + struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree; + struct bio *bio; + int num_copies; + int ret; + u64 logical; + + ret = get_state_private(failure_tree, start, &private); + if (ret) { + size_t pg_offset = start - page_offset(page); + failrec = kmalloc(sizeof(*failrec), GFP_NOFS); + if (!failrec) + return -ENOMEM; + failrec->start = start; + failrec->len = end - start + 1; + failrec->last_mirror = 0; + + em = btrfs_get_extent(inode, NULL, pg_offset, start, + failrec->len, 0); + + if (!em || IS_ERR(em)) { + kfree(failrec); + return -EIO; + } + logical = start - em->start; + logical = em->block_start + logical; + failrec->logical = logical; + free_extent_map(em); + set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | + EXTENT_DIRTY, GFP_NOFS); + set_state_private(failure_tree, start, (u64)failrec); + } else { + failrec = (struct io_failure_record *)private; + } + num_copies = btrfs_num_copies( + &BTRFS_I(inode)->root->fs_info->mapping_tree, + failrec->logical, failrec->len); + failrec->last_mirror++; + if (!state) { + spin_lock_irq(&BTRFS_I(inode)->io_tree.lock); + state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree, + failrec->start, + EXTENT_LOCKED); + if (state && state->start != failrec->start) + state = NULL; + spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock); + } + if (!state || failrec->last_mirror > num_copies) { + set_state_private(failure_tree, failrec->start, 0); + clear_extent_bits(failure_tree, failrec->start, + failrec->start + failrec->len - 1, + EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS); + kfree(failrec); + return -EIO; + } + bio = bio_alloc(GFP_NOFS, 1); + bio->bi_private = state; + bio->bi_end_io = failed_bio->bi_end_io; + bio->bi_sector = failrec->logical >> 9; + bio->bi_bdev = failed_bio->bi_bdev; + bio_add_page(bio, page, failrec->len, start - page_offset(page)); + btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror); + return 0; +} + int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state) { @@ -419,6 +499,29 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, } kunmap_atomic(kaddr, KM_IRQ0); local_irq_restore(flags); + + /* if the io failure tree for this inode is non-empty, + * check to see if we've recovered from a failed IO + */ + private = 0; + if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private, + (u64)-1, 1, EXTENT_DIRTY)) { + u64 private_failure; + struct io_failure_record *failure; + ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, + start, &private_failure); + if (ret == 0) { + failure = (struct io_failure_record *)private_failure; + set_state_private(&BTRFS_I(inode)->io_failure_tree, + failure->start, 0); + clear_extent_bits(&BTRFS_I(inode)->io_failure_tree, + failure->start, + failure->start + failure->len - 1, + EXTENT_DIRTY | EXTENT_LOCKED, + GFP_NOFS); + kfree(failure); + } + } return 0; zeroit: @@ -429,7 +532,7 @@ zeroit: flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); local_irq_restore(flags); - return 0; + return -EIO; } void btrfs_read_locked_inode(struct inode *inode) @@ -1271,6 +1374,8 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, + inode->i_mapping, GFP_NOFS); return 0; } @@ -1578,6 +1683,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, + inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->root = root; @@ -1803,6 +1910,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, + inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } @@ -2972,6 +3081,8 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); + extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, + inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } @@ -3070,6 +3181,7 @@ static struct extent_io_ops btrfs_extent_io_ops = { .merge_bio_hook = btrfs_merge_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, + .readpage_io_failed_hook = btrfs_readpage_io_failed_hook, .set_bit_hook = btrfs_set_bit_hook, .clear_bit_hook = btrfs_clear_bit_hook, }; -- cgit v1.2.3-70-g09d2 From 16432985920f3c45af82da214e2498f3e2f9066b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 10 Apr 2008 10:23:21 -0400 Subject: Btrfs: Add O_DIRECT read and write (writes == buffered + cache flush) This adds basic O_DIRECT read and write support. In the write case, we just do a normal buffered write followed by a cache flush. O_DIRECT + O_SYNC are required to trigger metadata syncs. In the read case, there is a basic btrfs_get_block call for use by the generic O_DIRECT code. This does honor multi-volume mapping rules but it skips all checksumming. Signed-off-by: Chris Mason --- fs/btrfs/file.c | 11 ++++++++-- fs/btrfs/inode.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 75 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 5fa450452f9..5b938645221 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -796,8 +796,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, PAGE_CACHE_SIZE / (sizeof(struct page *))); pinned[0] = NULL; pinned[1] = NULL; - if (file->f_flags & O_DIRECT) - return -EINVAL; pos = *ppos; start_pos = pos; @@ -909,6 +907,15 @@ out_nolock: start_pos, num_written); if (err < 0) num_written = err; + } else if (num_written > 0 && (file->f_flags & O_DIRECT)) { + do_sync_mapping_range(inode->i_mapping, start_pos, + start_pos + num_written - 1, + SYNC_FILE_RANGE_WRITE | + SYNC_FILE_RANGE_WAIT_AFTER); + + invalidate_mapping_pages(inode->i_mapping, + start_pos >> PAGE_CACHE_SHIFT, + (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); } current->backing_dev_info = NULL; return num_written ? num_written : err; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 48f1d1b9645..32e778fb014 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2265,6 +2265,70 @@ out: return em; } +static int btrfs_get_block(struct inode *inode, sector_t iblock, + struct buffer_head *bh_result, int create) +{ + struct extent_map *em; + u64 start = (u64)iblock << inode->i_blkbits; + struct btrfs_multi_bio *multi = NULL; + struct btrfs_root *root = BTRFS_I(inode)->root; + u64 len; + u64 logical; + u64 map_length; + int ret = 0; + + em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0); + + if (!em || IS_ERR(em)) + goto out; + + if (em->start > start || em->start + em->len <= start) + goto out; + + if (em->block_start == EXTENT_MAP_INLINE) { + ret = -EINVAL; + goto out; + } + + if (em->block_start == EXTENT_MAP_HOLE || + em->block_start == EXTENT_MAP_DELALLOC) { + goto out; + } + + len = em->start + em->len - start; + len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size))); + + logical = start - em->start; + logical = em->block_start + logical; + + map_length = len; + ret = btrfs_map_block(&root->fs_info->mapping_tree, READ, + logical, &map_length, &multi, 0); + BUG_ON(ret); + bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits; + bh_result->b_size = min(map_length, len); + bh_result->b_bdev = multi->stripes[0].dev->bdev; + set_buffer_mapped(bh_result); + kfree(multi); +out: + free_extent_map(em); + return ret; +} + +static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, + const struct iovec *iov, loff_t offset, + unsigned long nr_segs) +{ + struct file *file = iocb->ki_filp; + struct inode *inode = file->f_mapping->host; + + if (rw == WRITE) + return -EINVAL; + + return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, + offset, nr_segs, btrfs_get_block, NULL); +} + static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) { return extent_bmap(mapping, iblock, btrfs_get_extent); @@ -3136,6 +3200,7 @@ out_fail: btrfs_throttle(root); return err; } + static int btrfs_permission(struct inode *inode, int mask, struct nameidata *nd) { @@ -3193,6 +3258,7 @@ static struct address_space_operations btrfs_aops = { .readpages = btrfs_readpages, .sync_page = block_sync_page, .bmap = btrfs_bmap, + .direct_IO = btrfs_direct_IO, .invalidatepage = btrfs_invalidatepage, .releasepage = btrfs_releasepage, .set_page_dirty = __set_page_dirty_nobuffers, -- cgit v1.2.3-70-g09d2 From 587f77043a1c86e2a7900ff2ce86bef3c1f4e075 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 11 Apr 2008 12:16:46 -0400 Subject: Btrfs: Fixup a few u64<->pointer casts for 32 bit Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 32e778fb014..46f54dfa2b0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -430,9 +430,10 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, free_extent_map(em); set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS); - set_state_private(failure_tree, start, (u64)failrec); + set_state_private(failure_tree, start, + (u64)(unsigned long)failrec); } else { - failrec = (struct io_failure_record *)private; + failrec = (struct io_failure_record *)(unsigned long)private; } num_copies = btrfs_num_copies( &BTRFS_I(inode)->root->fs_info->mapping_tree, @@ -511,7 +512,8 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, start, &private_failure); if (ret == 0) { - failure = (struct io_failure_record *)private_failure; + failure = (struct io_failure_record *)(unsigned long) + private_failure; set_state_private(&BTRFS_I(inode)->io_failure_tree, failure->start, 0); clear_extent_bits(&BTRFS_I(inode)->io_failure_tree, -- cgit v1.2.3-70-g09d2 From 98d20f67cf99ccda638dbcdf7b3a9ee0a428d932 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 14 Apr 2008 09:46:10 -0400 Subject: Add a min size parameter to btrfs_alloc_extent On huge machines, delayed allocation may try to allocate massive extents. This change allows btrfs_alloc_extent to return something smaller than the caller asked for, and the data allocation routines will loop over the allocations until it fills the whole delayed alloc. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 ++- fs/btrfs/extent-tree.c | 12 +++++++++--- fs/btrfs/inode.c | 2 ++ 3 files changed, 13 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index e803c4daad2..09d614fcafb 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1286,7 +1286,8 @@ int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, u64 owner, u64 owner_offset); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 num_bytes, u64 root_objectid, u64 ref_generation, + u64 num_bytes, u64 min_bytes, + u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, int data); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a056a2df689..e49147e767d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1694,7 +1694,8 @@ error: */ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 num_bytes, u64 root_objectid, u64 ref_generation, + u64 num_bytes, u64 min_alloc_size, + u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, int data) @@ -1727,7 +1728,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, info->metadata_alloc_profile; data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } - +again: if (root->ref_cows) { if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, @@ -1751,6 +1752,11 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, search_start, search_end, hint_byte, ins, trans->alloc_exclude_start, trans->alloc_exclude_nr, data); + if (ret == -ENOSPC && num_bytes > min_alloc_size) { + num_bytes = num_bytes >> 1; + num_bytes = max(num_bytes, min_alloc_size); + goto again; + } BUG_ON(ret); if (ret) return ret; @@ -1869,7 +1875,7 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, int ret; struct extent_buffer *buf; - ret = btrfs_alloc_extent(trans, root, blocksize, + ret = btrfs_alloc_extent(trans, root, blocksize, blocksize, root_objectid, ref_generation, level, first_objectid, empty_size, hint, (u64)-1, &ins, 0); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 46f54dfa2b0..7869c0157a1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -125,6 +125,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) while(num_bytes > 0) { cur_alloc_size = min(num_bytes, root->fs_info->max_extent); ret = btrfs_alloc_extent(trans, root, cur_alloc_size, + root->sectorsize, root->root_key.objectid, trans->transid, inode->i_ino, start, 0, @@ -133,6 +134,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) WARN_ON(1); goto out; } + cur_alloc_size = ins.offset; ret = btrfs_insert_file_extent(trans, root, inode->i_ino, start, ins.objectid, ins.offset, ins.offset); -- cgit v1.2.3-70-g09d2 From 44b8bd7edda4f63de180d0f7325c9fb704b3806b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 16 Apr 2008 11:14:51 -0400 Subject: Btrfs: Create a work queue for bio writes This allows checksumming to happen in parallel among many cpus, and keeps us from bogging down pdflush with the checksumming code. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 ++ fs/btrfs/disk-io.c | 93 ++++++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/disk-io.h | 3 ++ fs/btrfs/extent_io.h | 5 +-- fs/btrfs/inode.c | 28 +++++++++++----- 5 files changed, 119 insertions(+), 13 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index a22edcf4917..ff15b8513f9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -514,8 +514,11 @@ struct btrfs_fs_info { struct list_head hashers; struct list_head dead_roots; struct list_head end_io_work_list; + struct list_head async_submit_work_list; struct work_struct end_io_work; + struct work_struct async_submit_work; spinlock_t end_io_work_lock; + spinlock_t async_submit_work_lock; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) struct work_struct trans_work; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 79c284c8728..9e41ea93ebc 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -47,6 +47,7 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) static struct extent_io_ops btree_extent_io_ops; static struct workqueue_struct *end_io_workqueue; +static struct workqueue_struct *async_submit_workqueue; struct end_io_wq { struct bio *bio; @@ -58,6 +59,15 @@ struct end_io_wq { struct list_head list; }; +struct async_submit_bio { + struct inode *inode; + struct bio *bio; + struct list_head list; + extent_submit_bio_hook_t *submit_bio_hook; + int rw; + int mirror_num; +}; + struct extent_map *btree_get_extent(struct inode *inode, struct page *page, size_t page_offset, u64 start, u64 len, int create) @@ -365,7 +375,31 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, return 0; } -static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, +int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, + int rw, struct bio *bio, int mirror_num, + extent_submit_bio_hook_t *submit_bio_hook) +{ + struct async_submit_bio *async; + + async = kmalloc(sizeof(*async), GFP_NOFS); + if (!async) + return -ENOMEM; + + async->inode = inode; + async->rw = rw; + async->bio = bio; + async->mirror_num = mirror_num; + async->submit_bio_hook = submit_bio_hook; + + spin_lock(&fs_info->async_submit_work_lock); + list_add_tail(&async->list, &fs_info->async_submit_work_list); + spin_unlock(&fs_info->async_submit_work_lock); + + queue_work(async_submit_workqueue, &fs_info->async_submit_work); + return 0; +} + +static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -389,6 +423,17 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); } +static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, + int mirror_num) +{ + if (!(rw & (1 << BIO_RW))) { + return __btree_submit_bio_hook(inode, rw, bio, mirror_num); + } + return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, + inode, rw, bio, mirror_num, + __btree_submit_bio_hook); +} + static int btree_writepage(struct page *page, struct writeback_control *wbc) { struct extent_io_tree *tree; @@ -903,9 +948,9 @@ static int bio_ready_for_csum(struct bio *bio) } #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) -void btrfs_end_io_csum(void *p) +static void btrfs_end_io_csum(void *p) #else -void btrfs_end_io_csum(struct work_struct *work) +static void btrfs_end_io_csum(struct work_struct *work) #endif { #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) @@ -959,6 +1004,39 @@ void btrfs_end_io_csum(struct work_struct *work) } } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) +static void btrfs_async_submit_work(void *p) +#else +static void btrfs_async_submit_work(struct work_struct *work) +#endif +{ +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + struct btrfs_fs_info *fs_info = p; +#else + struct btrfs_fs_info *fs_info = container_of(work, + struct btrfs_fs_info, + async_submit_work); +#endif + struct async_submit_bio *async; + struct list_head *next; + + while(1) { + spin_lock(&fs_info->async_submit_work_lock); + if (list_empty(&fs_info->async_submit_work_list)) { + spin_unlock(&fs_info->async_submit_work_lock); + return; + } + next = fs_info->async_submit_work_list.next; + list_del(next); + spin_unlock(&fs_info->async_submit_work_lock); + + async = list_entry(next, struct async_submit_bio, list); + async->submit_bio_hook(async->inode, async->rw, async->bio, + async->mirror_num); + kfree(async); + } +} + struct btrfs_root *open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices) { @@ -987,14 +1065,17 @@ struct btrfs_root *open_ctree(struct super_block *sb, } end_io_workqueue = create_workqueue("btrfs-end-io"); BUG_ON(!end_io_workqueue); + async_submit_workqueue = create_workqueue("btrfs-async-submit"); INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); INIT_LIST_HEAD(&fs_info->trans_list); INIT_LIST_HEAD(&fs_info->dead_roots); INIT_LIST_HEAD(&fs_info->hashers); INIT_LIST_HEAD(&fs_info->end_io_work_list); + INIT_LIST_HEAD(&fs_info->async_submit_work_list); spin_lock_init(&fs_info->hash_lock); spin_lock_init(&fs_info->end_io_work_lock); + spin_lock_init(&fs_info->async_submit_work_lock); spin_lock_init(&fs_info->delalloc_lock); spin_lock_init(&fs_info->new_trans_lock); @@ -1041,9 +1122,12 @@ struct btrfs_root *open_ctree(struct super_block *sb, #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info); + INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work, + fs_info); INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); #else INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum); + INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work); INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); #endif BTRFS_I(fs_info->btree_inode)->root = tree_root; @@ -1403,6 +1487,9 @@ int close_ctree(struct btrfs_root *root) flush_workqueue(end_io_workqueue); destroy_workqueue(end_io_workqueue); + flush_workqueue(async_submit_workqueue); + destroy_workqueue(async_submit_workqueue); + iput(fs_info->btree_inode); #if 0 while(!list_empty(&fs_info->hashers)) { diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 60b01902db7..5d5f6979c41 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -74,4 +74,7 @@ int btrfs_verify_block_csum(struct btrfs_root *root, struct extent_buffer *buf); int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, int metadata); +int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, + int rw, struct bio *bio, int mirror_num, + extent_submit_bio_hook_t *submit_bio_hook); #endif diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index b47859ccd78..232300d13be 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -24,11 +24,12 @@ struct extent_state; +typedef int (extent_submit_bio_hook_t)(struct inode *inode, int rw, + struct bio *bio, int mirror_num); struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); - int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio, - int mirror_num); + extent_submit_bio_hook_t *submit_bio_hook; int (*merge_bio_hook)(struct page *page, unsigned long offset, size_t size, struct bio *bio); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7869c0157a1..457daf725d2 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -324,13 +324,29 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, return 0; } -int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, +int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; int ret = 0; + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); + btrfs_csum_file_blocks(trans, root, inode, bio); + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); + mutex_unlock(&root->fs_info->fs_mutex); + return btrfs_map_bio(root, rw, bio, mirror_num); +} + +int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, + int mirror_num) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret = 0; + if (!(rw & (1 << BIO_RW))) { ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); @@ -342,13 +358,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, goto mapit; } - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - btrfs_csum_file_blocks(trans, root, inode, bio); - ret = btrfs_end_transaction(trans, root); - BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); + return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, + inode, rw, bio, mirror_num, + __btrfs_submit_bio_hook); mapit: return btrfs_map_bio(root, rw, bio, mirror_num); } -- cgit v1.2.3-70-g09d2 From e015640f9c4fa2417dcc3bbbb3b2b61ad4059ab0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 16 Apr 2008 11:15:20 -0400 Subject: Btrfs: Write bio checksumming outside the FS mutex This significantly improves streaming write performance by allowing concurrency in the data checksumming. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 4 +++- fs/btrfs/file-item.c | 46 ++++++++++++++++++++++++++++++++-------------- fs/btrfs/inode.c | 11 ++++++++++- 3 files changed, 45 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ff15b8513f9..b56ae195065 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1515,7 +1515,9 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, u64 bytenr, int mod); int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - struct bio *bio); + struct bio *bio, char *sums); +int btrfs_csum_one_bio(struct btrfs_root *root, + struct bio *bio, char **sums_ret); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 374d0245b29..9259aece6ed 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -134,9 +134,36 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } +int btrfs_csum_one_bio(struct btrfs_root *root, + struct bio *bio, char **sums_ret) +{ + u32 *sums; + char *data; + struct bio_vec *bvec = bio->bi_io_vec; + int bio_index = 0; + + sums = kmalloc(bio->bi_vcnt * BTRFS_CRC32_SIZE, GFP_NOFS); + if (!sums) + return -ENOMEM; + *sums_ret = (char *)sums; + + while(bio_index < bio->bi_vcnt) { + data = kmap_atomic(bvec->bv_page, KM_USER0); + *sums = ~(u32)0; + *sums = btrfs_csum_data(root, data + bvec->bv_offset, + *sums, bvec->bv_len); + kunmap_atomic(data, KM_USER0); + btrfs_csum_final(*sums, (char *)sums); + sums++; + bio_index++; + bvec++; + } + return 0; +} + int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - struct bio *bio) + struct bio *bio, char *sums) { u64 objectid = inode->i_ino; u64 offset; @@ -150,12 +177,11 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_csum_item *item_end; struct extent_buffer *leaf = NULL; u64 csum_offset; - u32 csum_result; + u32 *sums32 = (u32 *)sums; u32 nritems; u32 ins_size; int bio_index = 0; struct bio_vec *bvec = bio->bi_io_vec; - char *data; char *eb_map; char *eb_token; unsigned long map_len; @@ -278,15 +304,6 @@ found: btrfs_item_size_nr(leaf, path->slots[0])); eb_token = NULL; next_bvec: - data = kmap_atomic(bvec->bv_page, KM_USER0); - csum_result = ~(u32)0; - csum_result = btrfs_csum_data(root, data + bvec->bv_offset, - csum_result, bvec->bv_len); - kunmap_atomic(data, KM_USER0); - btrfs_csum_final(csum_result, (char *)&csum_result); - if (csum_result == 0) { - printk("csum result is 0 for inode %lu offset %Lu\n", inode->i_ino, offset); - } if (!eb_token || (unsigned long)item + BTRFS_CRC32_SIZE >= map_start + map_len) { @@ -304,13 +321,14 @@ next_bvec: } if (eb_token) { memcpy(eb_token + ((unsigned long)item & (PAGE_CACHE_SIZE - 1)), - &csum_result, BTRFS_CRC32_SIZE); + sums32, BTRFS_CRC32_SIZE); } else { - write_extent_buffer(leaf, &csum_result, (unsigned long)item, + write_extent_buffer(leaf, sums32, (unsigned long)item, BTRFS_CRC32_SIZE); } bio_index++; bvec++; + sums32++; if (bio_index < bio->bi_vcnt) { item = (struct btrfs_csum_item *)((char *)item + BTRFS_CRC32_SIZE); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 457daf725d2..3e1f21c632c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -330,14 +330,23 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; int ret = 0; + char *sums = NULL; + + ret = btrfs_csum_one_bio(root, bio, &sums); + BUG_ON(ret); mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); - btrfs_csum_file_blocks(trans, root, inode, bio); + btrfs_csum_file_blocks(trans, root, inode, bio, sums); + ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); + + kfree(sums); + return btrfs_map_bio(root, rw, bio, mirror_num); } -- cgit v1.2.3-70-g09d2 From 699122f55904bec252ea4f8a64228ce98e3d9fda Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 16 Apr 2008 12:59:22 -0400 Subject: Btrfs: Don't wait on tree block writeback before freeing them anymore This isn't required anymore because we don't reallocate blocks that have already been written in this transaction. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 5 ----- fs/btrfs/extent-tree.c | 1 - fs/btrfs/inode.c | 2 ++ 3 files changed, 2 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 618e526c904..1c3d9d6fbda 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -751,7 +751,6 @@ static int balance_level(struct btrfs_trans_handle *trans, add_root_to_dirty_list(root); path->nodes[level] = NULL; clean_tree_block(trans, root, mid); - wait_on_tree_block_writeback(root, mid); /* once for the path */ free_extent_buffer(mid); ret = btrfs_free_extent(trans, root, mid->start, mid->len, @@ -810,7 +809,6 @@ static int balance_level(struct btrfs_trans_handle *trans, u32 blocksize = right->len; clean_tree_block(trans, root, right); - wait_on_tree_block_writeback(root, right); free_extent_buffer(right); right = NULL; wret = del_ptr(trans, root, path, level + 1, pslot + @@ -854,7 +852,6 @@ static int balance_level(struct btrfs_trans_handle *trans, u64 bytenr = mid->start; u32 blocksize = mid->len; clean_tree_block(trans, root, mid); - wait_on_tree_block_writeback(root, mid); free_extent_buffer(mid); mid = NULL; wret = del_ptr(trans, root, path, level + 1, pslot); @@ -2638,7 +2635,6 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, } else { u64 root_gen = btrfs_header_generation(path->nodes[1]); clean_tree_block(trans, root, leaf); - wait_on_tree_block_writeback(root, leaf); wret = del_ptr(trans, root, path, 1, path->slots[1]); if (wret) ret = wret; @@ -2690,7 +2686,6 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, path->nodes[1]); clean_tree_block(trans, root, leaf); - wait_on_tree_block_writeback(root, leaf); wret = del_ptr(trans, root, path, 1, slot); if (wret) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 4e5bd62e6e1..76fd5d7146e 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1887,7 +1887,6 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, } btrfs_set_header_generation(buf, trans->transid); clean_tree_block(trans, root, buf); - wait_on_tree_block_writeback(root, buf); btrfs_set_buffer_uptodate(buf); if (PageDirty(buf->first_page)) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3e1f21c632c..5632ea76007 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -383,9 +383,11 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) struct btrfs_csum_item *item; struct btrfs_path *path = NULL; u32 csum; + if (btrfs_test_opt(root, NODATASUM) || btrfs_test_flag(inode, NODATASUM)) return 0; + mutex_lock(&root->fs_info->fs_mutex); path = btrfs_alloc_path(); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); -- cgit v1.2.3-70-g09d2 From 3b951516ed703af0f6d82053937655ad69b60864 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Apr 2008 11:29:12 -0400 Subject: Btrfs: Use the extent map cache to find the logical disk block during data retries The data read retry code needs to find the logical disk block before it can resubmit new bios. But, finding this block isn't allowed to take the fs_mutex because that will deadlock with a number of different callers. This changes the retry code to use the extent map cache instead, but that requires the extent map cache to have the extent we're looking for. This is a problem because btrfs_drop_extent_cache just drops the entire extent instead of the little tiny part it is invalidating. The bulk of the code in this patch changes btrfs_drop_extent_cache to invalidate only a portion of the extent cache, and changes btrfs_get_extent to deal with the results. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 1 + fs/btrfs/extent_io.c | 12 +++-- fs/btrfs/file.c | 47 ++++++++++++++++++- fs/btrfs/inode.c | 120 ++++++++++++++++++++++++++++++++++++++++++++----- fs/btrfs/volumes.c | 3 ++ 5 files changed, 168 insertions(+), 15 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 76fd5d7146e..593011e5d45 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1747,6 +1747,7 @@ again: search_start, search_end, hint_byte, ins, trans->alloc_exclude_start, trans->alloc_exclude_nr, data); + if (ret == -ENOSPC && num_bytes > min_alloc_size) { num_bytes = num_bytes >> 1; num_bytes = max(num_bytes, min_alloc_size); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 88322684be6..21597bea21f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1025,7 +1025,8 @@ u64 find_lock_delalloc_range(struct extent_io_tree *tree, search_again: node = tree_search(tree, cur_start); if (!node) { - *end = (u64)-1; + if (!found) + *end = (u64)-1; goto out; } @@ -1540,6 +1541,8 @@ static int end_bio_extent_readpage(struct bio *bio, start, end, state); if (ret == 0) { state = NULL; + uptodate = + test_bit(BIO_UPTODATE, &bio->bi_flags); continue; } } @@ -1555,10 +1558,11 @@ static int end_bio_extent_readpage(struct bio *bio, !(state->state & EXTENT_LOCKED)) state = NULL; } - if (!state && uptodate) { + if (!state) { spin_unlock_irqrestore(&tree->lock, flags); - set_extent_uptodate(tree, start, end, - GFP_ATOMIC); + if (uptodate) + set_extent_uptodate(tree, start, end, + GFP_ATOMIC); unlock_extent(tree, start, end, GFP_ATOMIC); goto next_io; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 9fbda655206..3f5525f0834 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -356,12 +356,23 @@ out_unlock: int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) { struct extent_map *em; + struct extent_map *split = NULL; + struct extent_map *split2 = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; u64 len = end - start + 1; + int ret; + int testend = 1; - if (end == (u64)-1) + if (end == (u64)-1) { len = (u64)-1; + testend = 0; + } while(1) { + if (!split) + split = alloc_extent_map(GFP_NOFS); + if (!split2) + split2 = alloc_extent_map(GFP_NOFS); + spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, len); if (!em) { @@ -369,6 +380,36 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) break; } remove_extent_mapping(em_tree, em); + + if (em->block_start < EXTENT_MAP_LAST_BYTE && + em->start < start) { + split->start = em->start; + split->len = start - em->start; + split->block_start = em->block_start; + split->bdev = em->bdev; + split->flags = em->flags; + ret = add_extent_mapping(em_tree, split); + BUG_ON(ret); + free_extent_map(split); + split = split2; + split2 = NULL; + } + if (em->block_start < EXTENT_MAP_LAST_BYTE && + testend && em->start + em->len > start + len) { + u64 diff = start + len - em->start; + + split->start = start + len; + split->len = em->start + em->len - (start + len); + split->bdev = em->bdev; + split->flags = em->flags; + + split->block_start = em->block_start + diff; + + ret = add_extent_mapping(em_tree, split); + BUG_ON(ret); + free_extent_map(split); + split = NULL; + } spin_unlock(&em_tree->lock); /* once for us */ @@ -376,6 +417,10 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) /* once for the tree*/ free_extent_map(em); } + if (split) + free_extent_map(split); + if (split2) + free_extent_map(split2); return 0; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5632ea76007..40f8da88409 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -122,6 +122,8 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) if (alloc_hint == EXTENT_MAP_INLINE) goto out; + BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); + while(num_bytes > 0) { cur_alloc_size = min(num_bytes, root->fs_info->max_extent); ret = btrfs_alloc_extent(trans, root, cur_alloc_size, @@ -140,6 +142,11 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) ins.offset); inode->i_blocks += ins.offset >> 9; btrfs_check_file(root, inode); + if (num_bytes < cur_alloc_size) { + printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes, + cur_alloc_size); + break; + } num_bytes -= cur_alloc_size; alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; @@ -427,6 +434,7 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, struct extent_map *em; struct inode *inode = page->mapping->host; struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct bio *bio; int num_copies; int ret; @@ -434,7 +442,6 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, ret = get_state_private(failure_tree, start, &private); if (ret) { - size_t pg_offset = start - page_offset(page); failrec = kmalloc(sizeof(*failrec), GFP_NOFS); if (!failrec) return -ENOMEM; @@ -442,8 +449,13 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, failrec->len = end - start + 1; failrec->last_mirror = 0; - em = btrfs_get_extent(inode, NULL, pg_offset, start, - failrec->len, 0); + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, start, failrec->len); + if (em->start > start || em->start + em->len < start) { + free_extent_map(em); + em = NULL; + } + spin_unlock(&em_tree->lock); if (!em || IS_ERR(em)) { kfree(failrec); @@ -559,6 +571,8 @@ zeroit: flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); local_irq_restore(flags); + if (private == 0) + return 0; return -EIO; } @@ -908,8 +922,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int pending_del_nr = 0; int pending_del_slot = 0; int extent_type = -1; + u64 mask = root->sectorsize - 1; - btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1); + btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1); path = btrfs_alloc_path(); path->reada = -1; BUG_ON(!path); @@ -1212,7 +1227,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) hole_start, 0, 0, hole_size); btrfs_drop_extent_cache(inode, hole_start, - hole_size - 1); + (u64)-1); btrfs_check_file(root, inode); } btrfs_end_transaction(trans, root); @@ -2083,6 +2098,68 @@ out_unlock: return err; } +static int merge_extent_mapping(struct extent_map_tree *em_tree, + struct extent_map *existing, + struct extent_map *em) +{ + u64 start_diff; + u64 new_end; + int ret = 0; + int real_blocks = existing->block_start < EXTENT_MAP_LAST_BYTE; + + if (real_blocks && em->block_start >= EXTENT_MAP_LAST_BYTE) + goto invalid; + + if (!real_blocks && em->block_start != existing->block_start) + goto invalid; + + new_end = max(existing->start + existing->len, em->start + em->len); + + if (existing->start >= em->start) { + if (em->start + em->len < existing->start) + goto invalid; + + start_diff = existing->start - em->start; + if (real_blocks && em->block_start + start_diff != + existing->block_start) + goto invalid; + + em->len = new_end - em->start; + + remove_extent_mapping(em_tree, existing); + /* free for the tree */ + free_extent_map(existing); + ret = add_extent_mapping(em_tree, em); + + } else if (em->start > existing->start) { + + if (existing->start + existing->len < em->start) + goto invalid; + + start_diff = em->start - existing->start; + if (real_blocks && existing->block_start + start_diff != + em->block_start) + goto invalid; + + remove_extent_mapping(em_tree, existing); + em->block_start = existing->block_start; + em->start = existing->start; + em->len = new_end - existing->start; + free_extent_map(existing); + + ret = add_extent_mapping(em_tree, em); + } else { + goto invalid; + } + return ret; + +invalid: + printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n", + existing->start, existing->len, existing->block_start, + em->start, em->len, em->block_start); + return -EIO; +} + struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, size_t pg_offset, u64 start, u64 len, int create) @@ -2267,12 +2344,35 @@ insert: err = 0; spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); + + /* it is possible that someone inserted the extent into the tree + * while we had the lock dropped. It is also possible that + * an overlapping map exists in the tree + */ if (ret == -EEXIST) { - free_extent_map(em); - em = lookup_extent_mapping(em_tree, start, len); - if (!em) { - err = -EIO; - printk("failing to insert %Lu %Lu\n", start, len); + struct extent_map *existing; + existing = lookup_extent_mapping(em_tree, start, len); + if (!existing) { + existing = lookup_extent_mapping(em_tree, em->start, + em->len); + if (existing) { + err = merge_extent_mapping(em_tree, existing, + em); + free_extent_map(existing); + if (err) { + free_extent_map(em); + em = NULL; + } + } else { + err = -EIO; + printk("failing to insert %Lu %Lu\n", + start, len); + free_extent_map(em); + em = NULL; + } + } else { + free_extent_map(em); + em = existing; } } spin_unlock(&em_tree->lock); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index e6417a573d4..0e658c1d821 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -883,6 +883,9 @@ again: spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, logical, *length); spin_unlock(&em_tree->lock); + if (!em) { + printk("unable to find logical %Lu\n", logical); + } BUG_ON(!em); BUG_ON(em->start > logical || em->start + em->len < logical); -- cgit v1.2.3-70-g09d2 From 9ad6b7bc2e00ba02f915cffd5b6bcd6564bb2c75 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Apr 2008 16:11:30 -0400 Subject: Force page->private removal in btrfs_invalidatepage btrfs_invalidatepage is not allowed to leave pages around on the lru. Any such pages will trigger an oops later on because the VM will see page->private and assume it is a buffer head. This also forces extra flushes of the async work queues before dropping all the pages on the btree inode during unmount. Left over items on the work queues are one possible cause of busy state ranges during truncate_inode_pages. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 12 ++++++++++++ fs/btrfs/inode.c | 6 ++++++ 2 files changed, 18 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 9d648f2f41a..b479cdfe3ee 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -533,6 +533,12 @@ static void btree_invalidatepage(struct page *page, unsigned long offset) tree = &BTRFS_I(page->mapping->host)->io_tree; extent_invalidatepage(tree, page, offset); btree_releasepage(page, GFP_NOFS); + if (PagePrivate(page)) { + printk("2invalidate page cleaning up after releasepage\n"); + ClearPagePrivate(page); + set_page_private(page, 0); + page_cache_release(page); + } } #if 0 @@ -1484,6 +1490,8 @@ int close_ctree(struct btrfs_root *root) write_ctree_super(NULL, root); mutex_unlock(&fs_info->fs_mutex); + btrfs_transaction_flush_work(root); + if (fs_info->delalloc_bytes) { printk("btrfs: at unmount delalloc count %Lu\n", fs_info->delalloc_bytes); @@ -1514,7 +1522,11 @@ int close_ctree(struct btrfs_root *root) extent_io_tree_empty_lru(&fs_info->extent_ins); extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); + flush_workqueue(end_io_workqueue); + flush_workqueue(async_submit_workqueue); + truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); + flush_workqueue(end_io_workqueue); destroy_workqueue(end_io_workqueue); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 40f8da88409..347cd85db12 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2524,6 +2524,12 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) tree = &BTRFS_I(page->mapping->host)->io_tree; extent_invalidatepage(tree, page, offset); btrfs_releasepage(page, GFP_NOFS); + if (PagePrivate(page)) { + printk("invalidate page cleaning up after releasepage\n"); + ClearPagePrivate(page); + set_page_private(page, 0); + page_cache_release(page); + } } /* -- cgit v1.2.3-70-g09d2 From 4ef64eae288a3644d1f8b748eb075426701e96d1 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 21 Apr 2008 08:52:50 -0400 Subject: Btrfs: Remove debugging statements from the invalidatepage calls Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 2 +- fs/btrfs/inode.c | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 46188ee1662..c829612c797 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -534,7 +534,7 @@ static void btree_invalidatepage(struct page *page, unsigned long offset) extent_invalidatepage(tree, page, offset); btree_releasepage(page, GFP_NOFS); if (PagePrivate(page)) { - printk("2invalidate page cleaning up after releasepage\n"); + invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 347cd85db12..5b1514e58de 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2510,6 +2510,7 @@ static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) map = &BTRFS_I(page->mapping->host)->extent_tree; ret = try_release_extent_mapping(map, tree, page, gfp_flags); if (ret == 1) { + invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); @@ -2525,7 +2526,7 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) extent_invalidatepage(tree, page, offset); btrfs_releasepage(page, GFP_NOFS); if (PagePrivate(page)) { - printk("invalidate page cleaning up after releasepage\n"); + invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); -- cgit v1.2.3-70-g09d2 From f2d8d74d7874f8f81222363cd6459a365796e35a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 21 Apr 2008 10:03:05 -0400 Subject: Btrfs: Make an unplug function that doesn't unplug every spindle Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 26 ++++++++++-------- fs/btrfs/inode.c | 6 +---- fs/btrfs/volumes.c | 79 +++++++++++++++++++++++++++++++++++++++--------------- fs/btrfs/volumes.h | 2 ++ 4 files changed, 75 insertions(+), 38 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index c829612c797..7f5aca35494 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -913,18 +913,22 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits) void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) { - struct list_head *cur; - struct btrfs_device *device; - struct btrfs_fs_info *info; + struct inode *inode = page->mapping->host; + struct extent_map_tree *em_tree; + struct extent_map *em; + u64 offset = page_offset(page); - info = (struct btrfs_fs_info *)bdi->unplug_io_data; - list_for_each(cur, &info->fs_devices->devices) { - device = list_entry(cur, struct btrfs_device, dev_list); - bdi = blk_get_backing_dev_info(device->bdev); - if (bdi->unplug_io_fn) { - bdi->unplug_io_fn(bdi, page); - } - } + em_tree = &BTRFS_I(inode)->extent_tree; + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); + spin_unlock(&em_tree->lock); + if (!em) + return; + + offset = offset - em->start; + btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, + em->block_start + offset, page); + free_extent_map(em); } static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5b1514e58de..e875c7c8a64 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -313,13 +313,9 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, u64 logical = bio->bi_sector << 9; u64 length = 0; u64 map_length; - struct bio_vec *bvec; - int i; int ret; - bio_for_each_segment(bvec, bio, i) { - length += bvec->bv_len; - } + length = bio->bi_size; map_tree = &root->fs_info->mapping_tree; map_length = length; ret = btrfs_map_block(map_tree, READ, logical, diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 4fc92cf813d..46024070650 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -18,6 +18,7 @@ #include #include #include +#include #include #include "ctree.h" #include "extent_map.h" @@ -930,9 +931,10 @@ int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) return ret; } -int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, - u64 logical, u64 *length, - struct btrfs_multi_bio **multi_ret, int mirror_num) +static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, + u64 logical, u64 *length, + struct btrfs_multi_bio **multi_ret, + int mirror_num, struct page *unplug_page) { struct extent_map *em; struct map_lookup *map; @@ -944,6 +946,7 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, int stripes_required = 1; int stripe_index; int i; + int num_stripes; struct btrfs_multi_bio *multi = NULL; if (multi_ret && !(rw & (1 << BIO_RW))) { @@ -960,10 +963,14 @@ again: spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, logical, *length); spin_unlock(&em_tree->lock); + + if (!em && unplug_page) + return 0; + if (!em) { printk("unable to find logical %Lu\n", logical); + BUG(); } - BUG_ON(!em); BUG_ON(em->start > logical || em->start + em->len < logical); map = (struct map_lookup *)em->bdev; @@ -1010,14 +1017,15 @@ again: } else { *length = em->len - offset; } - if (!multi_ret) + + if (!multi_ret && !unplug_page) goto out; - multi->num_stripes = 1; + num_stripes = 1; stripe_index = 0; if (map->type & BTRFS_BLOCK_GROUP_RAID1) { - if (rw & (1 << BIO_RW)) - multi->num_stripes = map->num_stripes; + if (unplug_page || (rw & (1 << BIO_RW))) + num_stripes = map->num_stripes; else if (mirror_num) { stripe_index = mirror_num - 1; } else { @@ -1037,7 +1045,7 @@ again: } } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { if (rw & (1 << BIO_RW)) - multi->num_stripes = map->num_stripes; + num_stripes = map->num_stripes; else if (mirror_num) stripe_index = mirror_num - 1; } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { @@ -1047,8 +1055,8 @@ again: stripe_index = do_div(stripe_nr, factor); stripe_index *= map->sub_stripes; - if (rw & (1 << BIO_RW)) - multi->num_stripes = map->sub_stripes; + if (unplug_page || (rw & (1 << BIO_RW))) + num_stripes = map->sub_stripes; else if (mirror_num) stripe_index += mirror_num - 1; else @@ -1063,19 +1071,50 @@ again: } BUG_ON(stripe_index >= map->num_stripes); - for (i = 0; i < multi->num_stripes; i++) { - multi->stripes[i].physical = - map->stripes[stripe_index].physical + stripe_offset + - stripe_nr * map->stripe_len; - multi->stripes[i].dev = map->stripes[stripe_index].dev; + for (i = 0; i < num_stripes; i++) { + if (unplug_page) { + struct btrfs_device *device; + struct backing_dev_info *bdi; + + device = map->stripes[stripe_index].dev; + bdi = blk_get_backing_dev_info(device->bdev); + if (bdi->unplug_io_fn) { + bdi->unplug_io_fn(bdi, unplug_page); + } + } else { + multi->stripes[i].physical = + map->stripes[stripe_index].physical + + stripe_offset + stripe_nr * map->stripe_len; + multi->stripes[i].dev = map->stripes[stripe_index].dev; + } stripe_index++; } - *multi_ret = multi; + if (multi_ret) { + *multi_ret = multi; + multi->num_stripes = num_stripes; + } out: free_extent_map(em); return 0; } +int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, + u64 logical, u64 *length, + struct btrfs_multi_bio **multi_ret, int mirror_num) +{ + return __btrfs_map_block(map_tree, rw, logical, length, multi_ret, + mirror_num, NULL); +} + +int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, + u64 logical, struct page *page) +{ + u64 length = PAGE_CACHE_SIZE; + return __btrfs_map_block(map_tree, READ, logical, &length, + NULL, 0, page); +} + + #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_multi_stripe(struct bio *bio, int err) #else @@ -1122,16 +1161,12 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, u64 logical = bio->bi_sector << 9; u64 length = 0; u64 map_length; - struct bio_vec *bvec; struct btrfs_multi_bio *multi = NULL; - int i; int ret; int dev_nr = 0; int total_devs = 1; - bio_for_each_segment(bvec, bio, i) { - length += bvec->bv_len; - } + length = bio->bi_size; map_tree = &root->fs_info->mapping_tree; map_length = length; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index f9cae307217..dfe7e8a9507 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -119,4 +119,6 @@ int btrfs_add_device(struct btrfs_trans_handle *trans, struct btrfs_device *device); int btrfs_cleanup_fs_uuids(void); int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len); +int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, + u64 logical, struct page *page); #endif -- cgit v1.2.3-70-g09d2 From e1c4b7451e22f5b0a9fbccfa560ee7b80c35b8cd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Apr 2008 13:26:46 -0400 Subject: Fix btrfs_get_extent and get_block corner cases, and disable O_DIRECT reads The generic O_DIRECT code assumes all the bios have the same bdev, which isn't true for multi-device btrfs. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 1 + fs/btrfs/inode.c | 32 +++++++++++++++++++++----------- fs/btrfs/volumes.c | 2 +- 3 files changed, 23 insertions(+), 12 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 81f8b4fd069..6e4bf029c6d 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1694,6 +1694,7 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, } if (bio) { + bio->bi_size = 0; bio->bi_bdev = bdev; bio->bi_sector = first_sector; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e875c7c8a64..abfe86df02d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -494,6 +494,7 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, bio->bi_end_io = failed_bio->bi_end_io; bio->bi_sector = failrec->logical >> 9; bio->bi_bdev = failed_bio->bi_bdev; + bio->bi_size = 0; bio_add_page(bio, page, failrec->len, start - page_offset(page)); btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror); return 0; @@ -2187,12 +2188,9 @@ again: spin_unlock(&em_tree->lock); if (em) { - if (em->start > start) { - printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n", - start, len, em->start, em->len); - WARN_ON(1); - } - if (em->block_start == EXTENT_MAP_INLINE && page) + if (em->start > start || em->start + em->len <= start) + free_extent_map(em); + else if (em->block_start == EXTENT_MAP_INLINE && page) free_extent_map(em); else goto out; @@ -2340,7 +2338,6 @@ insert: err = 0; spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); - /* it is possible that someone inserted the extent into the tree * while we had the lock dropped. It is also possible that * an overlapping map exists in the tree @@ -2348,6 +2345,11 @@ insert: if (ret == -EEXIST) { struct extent_map *existing; existing = lookup_extent_mapping(em_tree, start, len); + if (existing && (existing->start > start || + existing->start + existing->len <= start)) { + free_extent_map(existing); + existing = NULL; + } if (!existing) { existing = lookup_extent_mapping(em_tree, em->start, em->len); @@ -2388,6 +2390,7 @@ out: return em; } +#if 0 /* waiting for O_DIRECT reads */ static int btrfs_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) { @@ -2405,22 +2408,24 @@ static int btrfs_get_block(struct inode *inode, sector_t iblock, if (!em || IS_ERR(em)) goto out; - if (em->start > start || em->start + em->len <= start) + if (em->start > start || em->start + em->len <= start) { goto out; + } if (em->block_start == EXTENT_MAP_INLINE) { ret = -EINVAL; goto out; } + len = em->start + em->len - start; + len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size))); + if (em->block_start == EXTENT_MAP_HOLE || em->block_start == EXTENT_MAP_DELALLOC) { + bh_result->b_size = len; goto out; } - len = em->start + em->len - start; - len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size))); - logical = start - em->start; logical = em->block_start + logical; @@ -2430,6 +2435,7 @@ static int btrfs_get_block(struct inode *inode, sector_t iblock, BUG_ON(ret); bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits; bh_result->b_size = min(map_length, len); + bh_result->b_bdev = multi->stripes[0].dev->bdev; set_buffer_mapped(bh_result); kfree(multi); @@ -2437,11 +2443,14 @@ out: free_extent_map(em); return ret; } +#endif static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t offset, unsigned long nr_segs) { + return -EINVAL; +#if 0 struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; @@ -2450,6 +2459,7 @@ static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, offset, nr_segs, btrfs_get_block, NULL); +#endif } static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5619e50583e..bccb5566fd8 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1161,7 +1161,6 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, int total_devs = 1; length = bio->bi_size; - map_tree = &root->fs_info->mapping_tree; map_length = length; @@ -1192,6 +1191,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, } bio->bi_sector = multi->stripes[dev_nr].physical >> 9; dev = multi->stripes[dev_nr].dev; + bio->bi_bdev = dev->bdev; spin_lock(&dev->io_lock); dev->total_ios++; -- cgit v1.2.3-70-g09d2 From bcbfce8abd5f8d3f84eab60a9df1ec147f81c34f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Apr 2008 13:26:47 -0400 Subject: Btrfs: Fix the unplug_io_fn to grab a consistent copy of page->mapping Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 13 ++++++++++++- fs/btrfs/inode.c | 16 ++++++++++------ 2 files changed, 22 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 19c258d2c7e..1281c393c7e 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -936,14 +936,25 @@ void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) struct inode *inode; struct extent_map_tree *em_tree; struct extent_map *em; + struct address_space *mapping; u64 offset; + /* the generic O_DIRECT read code does this */ if (!page) { __unplug_io_fn(bdi, page); return; } - inode = page->mapping->host; + /* + * page->mapping may change at any time. Get a consistent copy + * and use that for everything below + */ + smp_mb(); + mapping = page->mapping; + if (!mapping) + return; + + inode = mapping->host; offset = page_offset(page); em_tree = &BTRFS_I(inode)->extent_tree; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index abfe86df02d..d58b4d5a1fe 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -80,6 +80,7 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy); u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy); u64 thresh; + unsigned long flags; int ret = 0; if (for_del) @@ -89,10 +90,10 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, do_div(thresh, 100); - spin_lock(&root->fs_info->delalloc_lock); + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); if (used + root->fs_info->delalloc_bytes + num_required > thresh) ret = -ENOSPC; - spin_unlock(&root->fs_info->delalloc_lock); + spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); return ret; } @@ -275,12 +276,13 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { + unsigned long flags; if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; - spin_lock(&root->fs_info->delalloc_lock); + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); BTRFS_I(inode)->delalloc_bytes += end - start + 1; root->fs_info->delalloc_bytes += end - start + 1; - spin_unlock(&root->fs_info->delalloc_lock); + spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); } return 0; } @@ -290,7 +292,9 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, { if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; - spin_lock(&root->fs_info->delalloc_lock); + unsigned long flags; + + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); if (end - start + 1 > root->fs_info->delalloc_bytes) { printk("warning: delalloc account %Lu %Lu\n", end - start + 1, root->fs_info->delalloc_bytes); @@ -300,7 +304,7 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, root->fs_info->delalloc_bytes -= end - start + 1; BTRFS_I(inode)->delalloc_bytes -= end - start + 1; } - spin_unlock(&root->fs_info->delalloc_lock); + spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); } return 0; } -- cgit v1.2.3-70-g09d2 From 81d7ed29ff6bdec903c36c26b386e16c014993b2 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 25 Apr 2008 08:51:48 -0400 Subject: Btrfs: Throttle file_write when data=ordered is flushing the inode Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 4 ++++ fs/btrfs/ordered-data.c | 13 +++++++++++++ fs/btrfs/ordered-data.h | 1 + fs/btrfs/transaction.c | 10 ++++++++-- 6 files changed, 28 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index fe6ef8e3416..5ba83894c8b 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -31,6 +31,7 @@ struct btrfs_inode { struct extent_io_tree io_tree; struct extent_io_tree io_failure_tree; struct inode vfs_inode; + atomic_t ordered_writeback; u64 ordered_trans; /* diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 5d537f26dc8..8effdf4f5d6 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -977,6 +977,7 @@ out_nolock: (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); } current->backing_dev_info = NULL; + btrfs_ordered_throttle(root, inode); return num_written ? num_written : err; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d58b4d5a1fe..b31f52d4f2c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1419,6 +1419,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); return 0; } @@ -1728,6 +1729,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->root = root; @@ -1956,6 +1958,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->delalloc_bytes = 0; + atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -3292,6 +3295,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); BTRFS_I(inode)->delalloc_bytes = 0; + atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 3ee51e10c18..b474902c90e 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -269,3 +269,16 @@ int btrfs_del_ordered_inode(struct inode *inode) return ret; } +int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode) +{ + struct btrfs_transaction *cur = root->fs_info->running_transaction; + while(cur == root->fs_info->running_transaction && + atomic_read(&BTRFS_I(inode)->ordered_writeback)) { +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) + congestion_wait(WRITE, HZ/20); +#else + blk_congestion_wait(WRITE, HZ/20); +#endif + } + return 0; +} diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index f25c6771ec6..29047e0abaa 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -39,4 +39,5 @@ int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, u64 *root_objectid, u64 *objectid, struct inode **inode); int btrfs_del_ordered_inode(struct inode *inode); +int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 1ed179c020a..c85cb48d95e 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -512,8 +512,11 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->trans_mutex); mutex_unlock(&root->fs_info->fs_mutex); - if (S_ISREG(inode->i_mode)) + if (S_ISREG(inode->i_mode)) { + atomic_inc(&BTRFS_I(inode)->ordered_writeback); filemap_fdatawrite(inode->i_mapping); + atomic_dec(&BTRFS_I(inode)->ordered_writeback); + } iput(inode); mutex_lock(&root->fs_info->fs_mutex); @@ -530,8 +533,11 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->trans_mutex); mutex_unlock(&root->fs_info->fs_mutex); - if (S_ISREG(inode->i_mode)) + if (S_ISREG(inode->i_mode)) { + atomic_inc(&BTRFS_I(inode)->ordered_writeback); filemap_write_and_wait(inode->i_mapping); + atomic_dec(&BTRFS_I(inode)->ordered_writeback); + } atomic_dec(&inode->i_count); iput(inode); -- cgit v1.2.3-70-g09d2 From 8f18cf13396caae5a3d7ae91201cfb15181a9642 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 25 Apr 2008 16:53:30 -0400 Subject: Btrfs: Make the resizer work based on shrinking and growing devices Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 4 + fs/btrfs/extent-tree.c | 145 +++++++++------------- fs/btrfs/inode.c | 36 ++++-- fs/btrfs/volumes.c | 324 +++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/volumes.h | 5 + 6 files changed, 407 insertions(+), 108 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f00c4be59ad..ac7106ec535 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -474,6 +474,7 @@ struct btrfs_block_group_cache { u64 pinned; u64 flags; int cached; + int ro; }; struct btrfs_device; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 71838264ca6..a9ce491d279 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -818,6 +818,10 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, return fs_info->tree_root; if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) return fs_info->extent_root; + if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) + return fs_info->chunk_root; + if (location->objectid == BTRFS_DEV_TREE_OBJECTID) + return fs_info->dev_root; root = radix_tree_lookup(&fs_info->fs_roots_radix, (unsigned long)location->objectid); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c49592c5127..6540095544e 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -187,6 +187,7 @@ static int noinline find_search_start(struct btrfs_root *root, if (!cache) goto out; + total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); free_space_cache = &root->fs_info->free_space_cache; @@ -196,7 +197,7 @@ again: goto out; last = max(search_start, cache->key.objectid); - if (!block_group_bits(cache, data)) { + if (!block_group_bits(cache, data) || cache->ro) { goto new_group; } @@ -221,6 +222,8 @@ again: continue; } spin_unlock_irq(&free_space_cache->lock); + if (cache->ro) + goto new_group; if (start + num > cache->key.objectid + cache->key.offset) goto new_group; if (start + num > total_fs_bytes) @@ -319,7 +322,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, if (search_start && search_start < total_fs_bytes) { struct btrfs_block_group_cache *shint; shint = btrfs_lookup_block_group(info, search_start); - if (shint && block_group_bits(shint, data)) { + if (shint && block_group_bits(shint, data) && !shint->ro) { used = btrfs_block_group_used(&shint->item); if (used + shint->pinned < div_factor(shint->key.offset, factor)) { @@ -327,7 +330,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, } } } - if (hint && block_group_bits(hint, data) && + if (hint && !hint->ro && block_group_bits(hint, data) && hint->key.objectid < total_fs_bytes) { used = btrfs_block_group_used(&hint->item); if (used + hint->pinned < @@ -364,7 +367,7 @@ again: if (cache->key.objectid > total_fs_bytes) break; - if (block_group_bits(cache, data)) { + if (!cache->ro && block_group_bits(cache, data)) { if (full_search) free_check = cache->key.offset; else @@ -1020,6 +1023,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, if (found) { found->total_bytes += total_bytes; found->bytes_used += bytes_used; + found->full = 0; WARN_ON(found->total_bytes < found->bytes_used); *space_info = found; return 0; @@ -1700,7 +1704,6 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 super_used; u64 root_used; u64 search_start = 0; - u64 new_hint; u64 alloc_profile; u32 sizes[2]; struct btrfs_fs_info *info = root->fs_info; @@ -1724,7 +1727,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } again: - if (root->ref_cows) { + if (root != root->fs_info->extent_root) { if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, 2 * 1024 * 1024, @@ -1738,10 +1741,6 @@ again: BUG_ON(ret); } - new_hint = max(hint_byte, root->fs_info->alloc_start); - if (new_hint < btrfs_super_total_bytes(&info->super_copy)) - hint_byte = new_hint; - WARN_ON(num_bytes < root->sectorsize); ret = find_free_extent(trans, root, num_bytes, empty_size, search_start, search_end, hint_byte, ins, @@ -2473,15 +2472,16 @@ out: return ret; } -int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) +int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) { struct btrfs_trans_handle *trans; struct btrfs_root *tree_root = root->fs_info->tree_root; struct btrfs_path *path; u64 cur_byte; u64 total_found; + u64 shrink_last_byte; + struct btrfs_block_group_cache *shrink_block_group; struct btrfs_fs_info *info = root->fs_info; - struct extent_io_tree *block_group_cache; struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; @@ -2489,17 +2489,29 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) int ret; int progress = 0; - btrfs_set_super_total_bytes(&info->super_copy, new_size); - clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, - GFP_NOFS); - block_group_cache = &info->block_group_cache; + shrink_block_group = btrfs_lookup_block_group(root->fs_info, + shrink_start); + BUG_ON(!shrink_block_group); + + shrink_last_byte = shrink_start + shrink_block_group->key.offset; + + shrink_block_group->space_info->total_bytes -= + shrink_block_group->key.offset; +printk("shrink_extent_tree %Lu -> %Lu type %Lu\n", shrink_start, shrink_last_byte, shrink_block_group->flags); path = btrfs_alloc_path(); root = root->fs_info->extent_root; path->reada = 2; again: + trans = btrfs_start_transaction(root, 1); + do_chunk_alloc(trans, root->fs_info->extent_root, + btrfs_block_group_used(&shrink_block_group->item) + + 2 * 1024 * 1024, shrink_block_group->flags); + btrfs_end_transaction(trans, root); + shrink_block_group->ro = 1; + total_found = 0; - key.objectid = new_size; + key.objectid = shrink_start; key.offset = 0; key.type = 0; cur_byte = key.objectid; @@ -2511,10 +2523,12 @@ again: ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); if (ret < 0) goto out; + if (ret == 0) { leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (found_key.objectid + found_key.offset > new_size) { + if (found_key.objectid + found_key.offset > shrink_start && + found_key.objectid < shrink_last_byte) { cur_byte = found_key.objectid; key.objectid = cur_byte; } @@ -2543,6 +2557,9 @@ next: btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.objectid >= shrink_last_byte) + break; + if (progress && need_resched()) { memcpy(&key, &found_key, sizeof(key)); mutex_unlock(&root->fs_info->fs_mutex); @@ -2583,68 +2600,31 @@ next: goto again; } + /* + * we've freed all the extents, now remove the block + * group item from the tree + */ trans = btrfs_start_transaction(root, 1); - key.objectid = new_size; - key.offset = 0; - key.type = 0; - while(1) { - u64 ptr; - - ret = btrfs_search_slot(trans, root, &key, path, -1, 1); - if (ret < 0) - goto out; - - leaf = path->nodes[0]; - nritems = btrfs_header_nritems(leaf); -bg_next: - if (path->slots[0] >= nritems) { - ret = btrfs_next_leaf(root, path); - if (ret < 0) - break; - if (ret == 1) { - ret = 0; - break; - } - leaf = path->nodes[0]; - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + memcpy(&key, &shrink_block_group->key, sizeof(key)); - /* - * btrfs_next_leaf doesn't cow buffers, we have to - * do the search again - */ - memcpy(&key, &found_key, sizeof(key)); - btrfs_release_path(root, path); - goto resched_check; - } + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret > 0) + ret = -EIO; + if (ret < 0) + goto out; - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) { - printk("shrinker found key %Lu %u %Lu\n", - found_key.objectid, found_key.type, - found_key.offset); - path->slots[0]++; - goto bg_next; - } - ret = get_state_private(&info->block_group_cache, - found_key.objectid, &ptr); - if (!ret) - kfree((void *)(unsigned long)ptr); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + kfree(shrink_block_group); - clear_extent_bits(&info->block_group_cache, found_key.objectid, - found_key.objectid + found_key.offset - 1, - (unsigned int)-1, GFP_NOFS); + clear_extent_bits(&info->block_group_cache, found_key.objectid, + found_key.objectid + found_key.offset - 1, + (unsigned int)-1, GFP_NOFS); - key.objectid = found_key.objectid + 1; - btrfs_del_item(trans, root, path); - btrfs_release_path(root, path); -resched_check: - if (need_resched()) { - mutex_unlock(&root->fs_info->fs_mutex); - cond_resched(); - mutex_lock(&root->fs_info->fs_mutex); - } - } - clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, + btrfs_del_item(trans, root, path); + clear_extent_dirty(&info->free_space_cache, + shrink_start, shrink_last_byte - 1, GFP_NOFS); btrfs_commit_transaction(trans, root); out: @@ -2652,13 +2632,6 @@ out: return ret; } -int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 new_size) -{ - btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size); - return 0; -} - int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *key) { @@ -2726,7 +2699,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - cache = kmalloc(sizeof(*cache), GFP_NOFS); + cache = kzalloc(sizeof(*cache), GFP_NOFS); if (!cache) { ret = -ENOMEM; break; @@ -2736,8 +2709,6 @@ int btrfs_read_block_groups(struct btrfs_root *root) btrfs_item_ptr_offset(leaf, path->slots[0]), sizeof(cache->item)); memcpy(&cache->key, &found_key, sizeof(found_key)); - cache->cached = 0; - cache->pinned = 0; key.objectid = found_key.objectid + found_key.offset; btrfs_release_path(root, path); @@ -2789,12 +2760,10 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, extent_root = root->fs_info->extent_root; block_group_cache = &root->fs_info->block_group_cache; - cache = kmalloc(sizeof(*cache), GFP_NOFS); + cache = kzalloc(sizeof(*cache), GFP_NOFS); BUG_ON(!cache); cache->key.objectid = chunk_offset; cache->key.offset = size; - cache->cached = 0; - cache->pinned = 0; btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); memset(&cache->item, 0, sizeof(cache->item)); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b31f52d4f2c..4d12aa532c5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -16,6 +16,7 @@ * Boston, MA 021110-1307, USA. */ +#include #include #include #include @@ -2887,9 +2888,12 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) { u64 new_size; u64 old_size; + u64 devid = 1; struct btrfs_ioctl_vol_args *vol_args; struct btrfs_trans_handle *trans; + struct btrfs_device *device = NULL; char *sizestr; + char *devstr = NULL; int ret = 0; int namelen; int mod = 0; @@ -2909,9 +2913,25 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) goto out; } + mutex_lock(&root->fs_info->fs_mutex); sizestr = vol_args->name; + devstr = strchr(sizestr, ':'); + if (devstr) { + char *end; + sizestr = devstr + 1; + *devstr = '\0'; + devstr = vol_args->name; + devid = simple_strtoull(devstr, &end, 10); +printk("resizing devid %Lu\n", devid); + } + device = btrfs_find_device(root, devid, NULL); + if (!device) { + printk("resizer unable to find device %Lu\n", devid); + ret = -EINVAL; + goto out_unlock; + } if (!strcmp(sizestr, "max")) - new_size = root->fs_info->sb->s_bdev->bd_inode->i_size; + new_size = device->bdev->bd_inode->i_size; else { if (sizestr[0] == '-') { mod = -1; @@ -2923,12 +2943,11 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) new_size = btrfs_parse_size(sizestr); if (new_size == 0) { ret = -EINVAL; - goto out; + goto out_unlock; } } - mutex_lock(&root->fs_info->fs_mutex); - old_size = btrfs_super_total_bytes(&root->fs_info->super_copy); + old_size = device->total_bytes; if (mod < 0) { if (new_size > old_size) { @@ -2944,7 +2963,7 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) ret = -EINVAL; goto out_unlock; } - if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) { + if (new_size > device->bdev->bd_inode->i_size) { ret = -EFBIG; goto out_unlock; } @@ -2952,13 +2971,14 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) do_div(new_size, root->sectorsize); new_size *= root->sectorsize; -printk("new size is %Lu\n", new_size); +printk("new size for %s is %llu\n", device->name, (unsigned long long)new_size); + if (new_size > old_size) { trans = btrfs_start_transaction(root, 1); - ret = btrfs_grow_extent_tree(trans, root, new_size); + ret = btrfs_grow_device(trans, device, new_size); btrfs_commit_transaction(trans, root); } else { - ret = btrfs_shrink_extent_tree(root, new_size); + ret = btrfs_shrink_device(device, new_size); } out_unlock: diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c63a982e31d..a2c56de1548 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -77,7 +77,7 @@ static struct btrfs_device *__find_device(struct list_head *head, u64 devid, list_for_each(cur, head) { dev = list_entry(cur, struct btrfs_device, dev_list); if (dev->devid == devid && - !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE)) { + (!uuid || !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE))) { return dev; } } @@ -293,6 +293,10 @@ static int find_free_dev_extent(struct btrfs_trans_handle *trans, * so we make sure to start at an offset of at least 1MB */ search_start = max((u64)1024 * 1024, search_start); + + if (root->fs_info->alloc_start + num_bytes <= device->total_bytes) + search_start = max(root->fs_info->alloc_start, search_start); + key.objectid = device->devid; key.offset = search_start; key.type = BTRFS_DEV_EXTENT_KEY; @@ -380,6 +384,33 @@ error: return ret; } +int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, + struct btrfs_device *device, + u64 start) +{ + int ret; + struct btrfs_path *path; + struct btrfs_root *root = device->dev_root; + struct btrfs_key key; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + key.objectid = device->devid; + key.offset = start; + key.type = BTRFS_DEV_EXTENT_KEY; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + BUG_ON(ret); + + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + + btrfs_free_path(path); + return ret; +} + int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 chunk_tree, u64 chunk_objectid, @@ -560,6 +591,7 @@ out: btrfs_free_path(path); return ret; } + int btrfs_update_device(struct btrfs_trans_handle *trans, struct btrfs_device *device) { @@ -606,6 +638,254 @@ out: return ret; } +int btrfs_grow_device(struct btrfs_trans_handle *trans, + struct btrfs_device *device, u64 new_size) +{ + struct btrfs_super_block *super_copy = + &device->dev_root->fs_info->super_copy; + u64 old_total = btrfs_super_total_bytes(super_copy); + u64 diff = new_size - device->total_bytes; + + btrfs_set_super_total_bytes(super_copy, old_total + diff); + return btrfs_update_device(trans, device); +} + +static int btrfs_free_chunk(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 chunk_tree, u64 chunk_objectid, + u64 chunk_offset) +{ + int ret; + struct btrfs_path *path; + struct btrfs_key key; + + root = root->fs_info->chunk_root; + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + key.objectid = chunk_objectid; + key.offset = chunk_offset; + key.type = BTRFS_CHUNK_ITEM_KEY; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + BUG_ON(ret); + + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + + btrfs_free_path(path); + return 0; +} + +int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 + chunk_offset) +{ + struct btrfs_super_block *super_copy = &root->fs_info->super_copy; + struct btrfs_disk_key *disk_key; + struct btrfs_chunk *chunk; + u8 *ptr; + int ret = 0; + u32 num_stripes; + u32 array_size; + u32 len = 0; + u32 cur; + struct btrfs_key key; + + array_size = btrfs_super_sys_array_size(super_copy); + + ptr = super_copy->sys_chunk_array; + cur = 0; + + while (cur < array_size) { + disk_key = (struct btrfs_disk_key *)ptr; + btrfs_disk_key_to_cpu(&key, disk_key); + + len = sizeof(*disk_key); + + if (key.type == BTRFS_CHUNK_ITEM_KEY) { + chunk = (struct btrfs_chunk *)(ptr + len); + num_stripes = btrfs_stack_chunk_num_stripes(chunk); + len += btrfs_chunk_item_size(num_stripes); + } else { + ret = -EIO; + break; + } + if (key.objectid == chunk_objectid && + key.offset == chunk_offset) { + memmove(ptr, ptr + len, array_size - (cur + len)); + array_size -= len; + btrfs_set_super_sys_array_size(super_copy, array_size); + } else { + ptr += len; + cur += len; + } + } + return ret; +} + + +int btrfs_relocate_chunk(struct btrfs_root *root, + u64 chunk_tree, u64 chunk_objectid, + u64 chunk_offset) +{ + struct extent_map_tree *em_tree; + struct btrfs_root *extent_root; + struct btrfs_trans_handle *trans; + struct extent_map *em; + struct map_lookup *map; + int ret; + int i; + + root = root->fs_info->chunk_root; + extent_root = root->fs_info->extent_root; + em_tree = &root->fs_info->mapping_tree.map_tree; + + /* step one, relocate all the extents inside this chunk */ + ret = btrfs_shrink_extent_tree(extent_root, chunk_offset); + BUG_ON(ret); + + trans = btrfs_start_transaction(root, 1); + BUG_ON(!trans); + + /* + * step two, delete the device extents and the + * chunk tree entries + */ + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, chunk_offset, 1); + spin_unlock(&em_tree->lock); + + BUG_ON(em->start > chunk_offset || em->start + em->len < chunk_offset); + map = (struct map_lookup *)em->bdev; + + for (i = 0; i < map->num_stripes; i++) { + ret = btrfs_free_dev_extent(trans, map->stripes[i].dev, + map->stripes[i].physical); + BUG_ON(ret); + } + ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid, + chunk_offset); + + BUG_ON(ret); + + if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) { + ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset); + BUG_ON(ret); + goto out; + } + + + + spin_lock(&em_tree->lock); + remove_extent_mapping(em_tree, em); + kfree(map); + em->bdev = NULL; + + /* once for the tree */ + free_extent_map(em); + spin_unlock(&em_tree->lock); + +out: + /* once for us */ + free_extent_map(em); + + btrfs_end_transaction(trans, root); + return 0; +} + +/* + * shrinking a device means finding all of the device extents past + * the new size, and then following the back refs to the chunks. + * The chunk relocation code actually frees the device extent + */ +int btrfs_shrink_device(struct btrfs_device *device, u64 new_size) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = device->dev_root; + struct btrfs_dev_extent *dev_extent = NULL; + struct btrfs_path *path; + u64 length; + u64 chunk_tree; + u64 chunk_objectid; + u64 chunk_offset; + int ret; + int slot; + struct extent_buffer *l; + struct btrfs_key key; + struct btrfs_super_block *super_copy = &root->fs_info->super_copy; + u64 old_total = btrfs_super_total_bytes(super_copy); + u64 diff = device->total_bytes - new_size; + + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + trans = btrfs_start_transaction(root, 1); + if (!trans) { + ret = -ENOMEM; + goto done; + } + + path->reada = 2; + + device->total_bytes = new_size; + ret = btrfs_update_device(trans, device); + if (ret) { + btrfs_end_transaction(trans, root); + goto done; + } + WARN_ON(diff > old_total); + btrfs_set_super_total_bytes(super_copy, old_total - diff); + btrfs_end_transaction(trans, root); + + key.objectid = device->devid; + key.offset = (u64)-1; + key.type = BTRFS_DEV_EXTENT_KEY; + + while (1) { + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto done; + + ret = btrfs_previous_item(root, path, 0, key.type); + if (ret < 0) + goto done; + if (ret) { + ret = 0; + goto done; + } + + l = path->nodes[0]; + slot = path->slots[0]; + btrfs_item_key_to_cpu(l, &key, path->slots[0]); + + if (key.objectid != device->devid) + goto done; + + dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); + length = btrfs_dev_extent_length(l, dev_extent); + + if (key.offset + length <= new_size) + goto done; + + chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent); + chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent); + chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent); + btrfs_release_path(root, path); + + ret = btrfs_relocate_chunk(root, chunk_tree, chunk_objectid, + chunk_offset); + if (ret) + goto done; + } + +done: + btrfs_free_path(path); + return ret; +} + int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, @@ -658,6 +938,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, u64 dev_offset; struct btrfs_fs_info *info = extent_root->fs_info; struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; + struct btrfs_path *path; struct btrfs_stripe *stripes; struct btrfs_device *device = NULL; struct btrfs_chunk *chunk; @@ -724,6 +1005,10 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, min_stripe_size = 1 * 1024 * 1024; } + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + /* we don't want a chunk larger than 10% of the FS */ percent_max = div_factor(btrfs_super_total_bytes(&info->super_copy), 1); max_chunk_size = min(percent_max, max_chunk_size); @@ -759,11 +1044,19 @@ again: avail = device->total_bytes - device->bytes_used; cur = cur->next; + if (avail >= min_free) { - list_move_tail(&device->dev_alloc_list, &private_devs); - index++; - if (type & BTRFS_BLOCK_GROUP_DUP) + u64 ignored_start = 0; + ret = find_free_dev_extent(trans, device, path, + min_free, + &ignored_start); + if (ret == 0) { + list_move_tail(&device->dev_alloc_list, + &private_devs); index++; + if (type & BTRFS_BLOCK_GROUP_DUP) + index++; + } } else if (avail > max_avail) max_avail = avail; if (cur == dev_list) @@ -785,30 +1078,37 @@ again: calc_size = max_avail; goto again; } + btrfs_free_path(path); return -ENOSPC; } key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; key.type = BTRFS_CHUNK_ITEM_KEY; ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID, &key.offset); - if (ret) + if (ret) { + btrfs_free_path(path); return ret; + } chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); - if (!chunk) + if (!chunk) { + btrfs_free_path(path); return -ENOMEM; + } map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); if (!map) { kfree(chunk); + btrfs_free_path(path); return -ENOMEM; } + btrfs_free_path(path); + path = NULL; stripes = &chunk->stripe; *num_bytes = chunk_bytes_by_type(type, calc_size, num_stripes, sub_stripes); - index = 0; printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes); while(index < num_stripes) { @@ -874,6 +1174,11 @@ printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.offset, cal em->len = *num_bytes; em->block_start = 0; + if (type & BTRFS_BLOCK_GROUP_SYSTEM) { + ret = btrfs_add_system_chunk(trans, chunk_root, &key, + chunk, btrfs_chunk_item_size(num_stripes)); + BUG_ON(ret); + } kfree(chunk); em_tree = &extent_root->fs_info->mapping_tree.map_tree; @@ -1376,11 +1681,6 @@ int btrfs_read_sys_array(struct btrfs_root *root) array_size = btrfs_super_sys_array_size(super_copy); - /* - * we do this loop twice, once for the device items and - * once for all of the chunks. This way there are device - * structs filled in for every chunk - */ ptr = super_copy->sys_chunk_array; sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); cur = 0; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index a2660d2b6b3..6fe8440b37a 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -128,4 +128,9 @@ int btrfs_cleanup_fs_uuids(void); int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len); int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, u64 logical, struct page *page); +int btrfs_grow_device(struct btrfs_trans_handle *trans, + struct btrfs_device *device, u64 new_size); +struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, + u8 *uuid); +int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); #endif -- cgit v1.2.3-70-g09d2 From 8e7bf94fd5f44fa585e29fbe6a1bfabc04aea7cf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 28 Apr 2008 09:02:36 -0400 Subject: Btrfs: Do more optimal file RA during shrinking and defrag Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 16 +++++++++++++++- fs/btrfs/inode.c | 15 +++++++-------- 2 files changed, 22 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a589912fdd5..fe4fe709c31 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2263,6 +2263,12 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) return 0; } +static unsigned long calc_ra(unsigned long start, unsigned long last, + unsigned long nr) +{ + return min(last, start + nr - 1); +} + static int noinline relocate_inode_pages(struct inode *inode, u64 start, u64 len) { @@ -2275,6 +2281,8 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, struct page *page; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct file_ra_state *ra; + unsigned long total_read = 0; + unsigned long ra_pages; ra = kzalloc(sizeof(*ra), GFP_NOFS); @@ -2282,11 +2290,17 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, i = start >> PAGE_CACHE_SHIFT; last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; + ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages; + file_ra_state_init(ra, inode->i_mapping); - btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index); kfree(ra); for (; i <= last_index; i++) { + if (total_read % ra_pages == 0) { + btrfs_force_ra(inode->i_mapping, ra, NULL, i, + calc_ra(i, last_index, ra_pages)); + } + total_read++; page = grab_cache_page(inode->i_mapping, i); if (!page) goto out_unlock; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4d12aa532c5..ab707c0930d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2814,14 +2814,12 @@ unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, pgoff_t offset, pgoff_t last_index) { - pgoff_t req_size; + pgoff_t req_size = last_index - offset + 1; #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) - req_size = last_index - offset + 1; offset = page_cache_readahead(mapping, ra, file, offset, req_size); return offset; #else - req_size = min(last_index - offset + 1, (pgoff_t)128); page_cache_sync_readahead(mapping, ra, file, offset, req_size); return offset + req_size; #endif @@ -2833,7 +2831,8 @@ int btrfs_defrag_file(struct file *file) { struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct page *page; unsigned long last_index; - unsigned long ra_index = 0; + unsigned long ra_pages = root->fs_info->bdi.ra_pages; + unsigned long total_read = 0; u64 page_start; u64 page_end; unsigned long i; @@ -2848,11 +2847,11 @@ int btrfs_defrag_file(struct file *file) { mutex_lock(&inode->i_mutex); last_index = inode->i_size >> PAGE_CACHE_SHIFT; for (i = 0; i <= last_index; i++) { - if (i == ra_index) { - ra_index = btrfs_force_ra(inode->i_mapping, - &file->f_ra, - file, ra_index, last_index); + if (total_read % ra_pages == 0) { + btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, + min(last_index, i + ra_pages - 1)); } + total_read++; page = grab_cache_page(inode->i_mapping, i); if (!page) goto out_unlock; -- cgit v1.2.3-70-g09d2 From 788f20eb5affef584e75ea84bb80a4c3352a2c0e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 28 Apr 2008 15:29:42 -0400 Subject: Btrfs: Add new ioctl to add devices Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 ++ fs/btrfs/inode.c | 23 +++++++++++++++++ fs/btrfs/ioctl.h | 6 +++++ fs/btrfs/super.c | 1 + fs/btrfs/volumes.c | 75 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/volumes.h | 1 + 6 files changed, 108 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ac7106ec535..b9257b37bb9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -551,6 +551,8 @@ struct btrfs_fs_info { u64 data_alloc_profile; u64 metadata_alloc_profile; u64 system_alloc_profile; + + void *bdev_holder; }; /* diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index ab707c0930d..1a74b501869 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3070,6 +3070,27 @@ static int btrfs_ioctl_defrag(struct file *file) return 0; } +long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) +{ + struct btrfs_ioctl_vol_args *vol_args; + int ret; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + ret = btrfs_init_new_device(root, vol_args->name); + +out: + kfree(vol_args); + return ret; +} + long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { @@ -3082,6 +3103,8 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_defrag(file); case BTRFS_IOC_RESIZE: return btrfs_ioctl_resize(root, (void __user *)arg); + case BTRFS_IOC_ADD_DEV: + return btrfs_ioctl_add_dev(root, (void __user *)arg); } return -ENOTTY; diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 4551e82013c..8ad35fc4ba5 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -36,4 +36,10 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \ struct btrfs_ioctl_vol_args) +#define BTRFS_IOC_ADD_DEV _IOW(BTRFS_IOCTL_MAGIC, 10, \ + struct btrfs_ioctl_vol_args) +#define BTRFS_IOC_RM_DEV _IOW(BTRFS_IOCTL_MAGIC, 11, \ + struct btrfs_ioctl_vol_args) +#define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \ + struct btrfs_ioctl_vol_args) #endif diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 89286490688..7153dfaa340 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -388,6 +388,7 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type, goto error; } + btrfs_sb(s)->fs_info->bdev_holder = fs_type; s->s_flags |= MS_ACTIVE; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index a2c56de1548..b93c15aa17d 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -19,6 +19,7 @@ #include #include #include +#include #include #include "ctree.h" #include "extent_map.h" @@ -592,6 +593,80 @@ out: return ret; } +int btrfs_init_new_device(struct btrfs_root *root, char *device_path) +{ + struct btrfs_trans_handle *trans; + struct btrfs_device *device; + struct block_device *bdev; + struct list_head *cur; + struct list_head *devices; + u64 total_bytes; + int ret = 0; + + + bdev = open_bdev_excl(device_path, 0, root->fs_info->bdev_holder); + if (!bdev) { + return -EIO; + } + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 1); + devices = &root->fs_info->fs_devices->devices; + list_for_each(cur, devices) { + device = list_entry(cur, struct btrfs_device, dev_list); + if (device->bdev == bdev) { + ret = -EEXIST; + goto out; + } + } + + device = kzalloc(sizeof(*device), GFP_NOFS); + if (!device) { + /* we can safely leave the fs_devices entry around */ + ret = -ENOMEM; + goto out_close_bdev; + } + + device->barriers = 1; + generate_random_uuid(device->uuid); + spin_lock_init(&device->io_lock); + device->name = kstrdup(device_path, GFP_NOFS); + if (!device->name) { + kfree(device); + goto out_close_bdev; + } + device->io_width = root->sectorsize; + device->io_align = root->sectorsize; + device->sector_size = root->sectorsize; + device->total_bytes = i_size_read(bdev->bd_inode); + device->dev_root = root->fs_info->dev_root; + device->bdev = bdev; + + ret = btrfs_add_device(trans, root, device); + if (ret) + goto out_close_bdev; + + total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); + btrfs_set_super_total_bytes(&root->fs_info->super_copy, + total_bytes + device->total_bytes); + + total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy); + btrfs_set_super_num_devices(&root->fs_info->super_copy, + total_bytes + 1); + + list_add(&device->dev_list, &root->fs_info->fs_devices->devices); + list_add(&device->dev_alloc_list, + &root->fs_info->fs_devices->alloc_list); + root->fs_info->fs_devices->num_devices++; +out: + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; + +out_close_bdev: + close_bdev_excl(bdev); + goto out; +} + int btrfs_update_device(struct btrfs_trans_handle *trans, struct btrfs_device *device) { diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 6fe8440b37a..6f173450378 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -133,4 +133,5 @@ int btrfs_grow_device(struct btrfs_trans_handle *trans, struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, u8 *uuid); int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); +int btrfs_init_new_device(struct btrfs_root *root, char *path); #endif -- cgit v1.2.3-70-g09d2 From ec44a35cbeb26ab2da84cb280d778260f2312feb Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 28 Apr 2008 15:29:52 -0400 Subject: Btrfs: Add balance ioctl to restripe the chunks Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 +- fs/btrfs/extent-tree.c | 106 +++++++++++++++++++++++++++++++++++++-------- fs/btrfs/inode.c | 11 +++++ fs/btrfs/volumes.c | 115 +++++++++++++++++++++++++++++++++++++++++++++---- fs/btrfs/volumes.h | 1 + 5 files changed, 208 insertions(+), 27 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b9257b37bb9..73b92dd150f 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1364,7 +1364,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, - u64 search_end, struct btrfs_key *ins, int data); + u64 search_end, struct btrfs_key *ins, u64 data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index fe4fe709c31..95aee5a2937 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -17,6 +17,7 @@ */ #include #include +#include #include "hash.h" #include "crc32c.h" #include "ctree.h" @@ -1058,6 +1059,26 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) } } +static u64 reduce_alloc_profile(u64 flags) +{ + if ((flags & BTRFS_BLOCK_GROUP_DUP) && + (flags & (BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_RAID10))) + flags &= ~BTRFS_BLOCK_GROUP_DUP; + + if ((flags & BTRFS_BLOCK_GROUP_RAID1) && + (flags & BTRFS_BLOCK_GROUP_RAID10)) + flags &= ~BTRFS_BLOCK_GROUP_RAID1; + + if ((flags & BTRFS_BLOCK_GROUP_RAID0) && + ((flags & BTRFS_BLOCK_GROUP_RAID1) | + (flags & BTRFS_BLOCK_GROUP_RAID10) | + (flags & BTRFS_BLOCK_GROUP_DUP))) + flags &= ~BTRFS_BLOCK_GROUP_RAID0; + return flags; +} + + static int do_chunk_alloc(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 alloc_bytes, u64 flags) @@ -1068,6 +1089,8 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, u64 num_bytes; int ret; + flags = reduce_alloc_profile(flags); + space_info = __find_space_info(extent_root->fs_info, flags); if (!space_info) { ret = update_space_info(extent_root->fs_info, flags, @@ -1684,6 +1707,7 @@ enospc: error: return ret; } + /* * finds a free extent and does all the dirty work required for allocation * returns the key for the extent through ins, and a tree buffer for @@ -1697,7 +1721,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, - u64 search_end, struct btrfs_key *ins, int data) + u64 search_end, struct btrfs_key *ins, u64 data) { int ret; int pending_ret; @@ -1727,6 +1751,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } again: + data = reduce_alloc_profile(data); if (root->ref_cows) { if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, @@ -1752,6 +1777,9 @@ again: num_bytes = max(num_bytes, min_alloc_size); goto again; } + if (ret) { + printk("allocation failed flags %Lu\n", data); + } BUG_ON(ret); if (ret) return ret; @@ -2274,8 +2302,6 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, { u64 page_start; u64 page_end; - u64 delalloc_start; - u64 existing_delalloc; unsigned long last_index; unsigned long i; struct page *page; @@ -2293,7 +2319,6 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages; file_ra_state_init(ra, inode->i_mapping); - kfree(ra); for (; i <= last_index; i++) { if (total_read % ra_pages == 0) { @@ -2313,26 +2338,30 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, goto out_unlock; } } +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + ClearPageDirty(page); +#else + cancel_dirty_page(page, PAGE_CACHE_SIZE); +#endif + wait_on_page_writeback(page); + set_page_extent_mapped(page); page_start = (u64)page->index << PAGE_CACHE_SHIFT; page_end = page_start + PAGE_CACHE_SIZE - 1; lock_extent(io_tree, page_start, page_end, GFP_NOFS); - delalloc_start = page_start; - existing_delalloc = count_range_bits(io_tree, - &delalloc_start, page_end, - PAGE_CACHE_SIZE, EXTENT_DELALLOC); - + set_page_dirty(page); set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); - set_page_dirty(page); unlock_page(page); page_cache_release(page); + balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); } out_unlock: + kfree(ra); mutex_unlock(&inode->i_mutex); return 0; } @@ -2397,8 +2426,6 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, goto out; } relocate_inode_pages(inode, ref_offset, extent_key->offset); - /* FIXME, data=ordered will help get rid of this */ - filemap_fdatawrite(inode->i_mapping); iput(inode); mutex_lock(&extent_root->fs_info->fs_mutex); } else { @@ -2486,6 +2513,47 @@ out: return ret; } +static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) +{ + u64 num_devices; + u64 stripped = BTRFS_BLOCK_GROUP_RAID0 | + BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10; + + num_devices = btrfs_super_num_devices(&root->fs_info->super_copy); + if (num_devices == 1) { + stripped |= BTRFS_BLOCK_GROUP_DUP; + stripped = flags & ~stripped; + + /* turn raid0 into single device chunks */ + if (flags & BTRFS_BLOCK_GROUP_RAID0) + return stripped; + + /* turn mirroring into duplication */ + if (flags & (BTRFS_BLOCK_GROUP_RAID1 | + BTRFS_BLOCK_GROUP_RAID10)) + return stripped | BTRFS_BLOCK_GROUP_DUP; + return flags; + } else { + /* they already had raid on here, just return */ + if ((flags & BTRFS_BLOCK_GROUP_DUP) && + (flags & BTRFS_BLOCK_GROUP_RAID1)) { + } + if (flags & stripped) + return flags; + + stripped |= BTRFS_BLOCK_GROUP_DUP; + stripped = flags & ~stripped; + + /* switch duplicated blocks with raid1 */ + if (flags & BTRFS_BLOCK_GROUP_DUP) + return stripped | BTRFS_BLOCK_GROUP_RAID1; + + /* turn single device chunks into raid0 */ + return stripped | BTRFS_BLOCK_GROUP_RAID0; + } + return flags; +} + int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) { struct btrfs_trans_handle *trans; @@ -2494,6 +2562,7 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) u64 cur_byte; u64 total_found; u64 shrink_last_byte; + u64 new_alloc_flags; struct btrfs_block_group_cache *shrink_block_group; struct btrfs_fs_info *info = root->fs_info; struct btrfs_key key; @@ -2511,17 +2580,20 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) shrink_block_group->space_info->total_bytes -= shrink_block_group->key.offset; -printk("shrink_extent_tree %Lu -> %Lu type %Lu\n", shrink_start, shrink_last_byte, shrink_block_group->flags); path = btrfs_alloc_path(); root = root->fs_info->extent_root; path->reada = 2; again: - trans = btrfs_start_transaction(root, 1); - do_chunk_alloc(trans, root->fs_info->extent_root, + if (btrfs_block_group_used(&shrink_block_group->item) > 0) { + trans = btrfs_start_transaction(root, 1); + new_alloc_flags = update_block_group_flags(root, + shrink_block_group->flags); + do_chunk_alloc(trans, root->fs_info->extent_root, btrfs_block_group_used(&shrink_block_group->item) + - 2 * 1024 * 1024, shrink_block_group->flags); - btrfs_end_transaction(trans, root); + 2 * 1024 * 1024, new_alloc_flags); + btrfs_end_transaction(trans, root); + } shrink_block_group->ro = 1; total_found = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1a74b501869..99483447459 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2864,6 +2864,15 @@ int btrfs_defrag_file(struct file *file) { goto out_unlock; } } + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + ClearPageDirty(page); +#else + cancel_dirty_page(page, PAGE_CACHE_SIZE); +#endif + wait_on_page_writeback(page); + set_page_extent_mapped(page); + page_start = (u64)page->index << PAGE_CACHE_SHIFT; page_end = page_start + PAGE_CACHE_SIZE - 1; @@ -3105,6 +3114,8 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_resize(root, (void __user *)arg); case BTRFS_IOC_ADD_DEV: return btrfs_ioctl_add_dev(root, (void __user *)arg); + case BTRFS_IOC_BALANCE: + return btrfs_balance(root->fs_info->dev_root); } return -ENOTTY; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index b93c15aa17d..6476ecbf132 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -869,6 +869,107 @@ out: return 0; } +static u64 div_factor(u64 num, int factor) +{ + if (factor == 10) + return num; + num *= factor; + do_div(num, 10); + return num; +} + + +int btrfs_balance(struct btrfs_root *dev_root) +{ + int ret; + struct list_head *cur; + struct list_head *devices = &dev_root->fs_info->fs_devices->devices; + struct btrfs_device *device; + u64 old_size; + u64 size_to_free; + struct btrfs_path *path; + struct btrfs_key key; + struct btrfs_chunk *chunk; + struct btrfs_root *chunk_root = dev_root->fs_info->chunk_root; + struct btrfs_trans_handle *trans; + struct btrfs_key found_key; + + + dev_root = dev_root->fs_info->dev_root; + + mutex_lock(&dev_root->fs_info->fs_mutex); + /* step one make some room on all the devices */ + list_for_each(cur, devices) { + device = list_entry(cur, struct btrfs_device, dev_list); + old_size = device->total_bytes; + size_to_free = div_factor(old_size, 1); + size_to_free = min(size_to_free, (u64)1 * 1024 * 1024); + if (device->total_bytes - device->bytes_used > size_to_free) + continue; + + ret = btrfs_shrink_device(device, old_size - size_to_free); + BUG_ON(ret); + + trans = btrfs_start_transaction(dev_root, 1); + BUG_ON(!trans); + + ret = btrfs_grow_device(trans, device, old_size); + BUG_ON(ret); + + btrfs_end_transaction(trans, dev_root); + } + + /* step two, relocate all the chunks */ + path = btrfs_alloc_path(); + BUG_ON(!path); + + key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; + key.offset = (u64)-1; + key.type = BTRFS_CHUNK_ITEM_KEY; + + while(1) { + ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); + if (ret < 0) + goto error; + + /* + * this shouldn't happen, it means the last relocate + * failed + */ + if (ret == 0) + break; + + ret = btrfs_previous_item(chunk_root, path, 0, + BTRFS_CHUNK_ITEM_KEY); + if (ret) { + break; + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + if (found_key.objectid != key.objectid) + break; + chunk = btrfs_item_ptr(path->nodes[0], + path->slots[0], + struct btrfs_chunk); + key.offset = found_key.offset; + /* chunk zero is special */ + if (key.offset == 0) + break; + + ret = btrfs_relocate_chunk(chunk_root, + chunk_root->root_key.objectid, + found_key.objectid, + found_key.offset); + BUG_ON(ret); + btrfs_release_path(chunk_root, path); + } + ret = 0; +error: + btrfs_free_path(path); + mutex_unlock(&dev_root->fs_info->fs_mutex); + return ret; +} + /* * shrinking a device means finding all of the device extents past * the new size, and then following the back refs to the chunks. @@ -985,15 +1086,6 @@ int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, return 0; } -static u64 div_factor(u64 num, int factor) -{ - if (factor == 10) - return num; - num *= factor; - do_div(num, 10); - return num; -} - static u64 chunk_bytes_by_type(u64 type, u64 calc_size, int num_stripes, int sub_stripes) { @@ -1040,6 +1132,11 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, int stripe_len = 64 * 1024; struct btrfs_key key; + if ((type & BTRFS_BLOCK_GROUP_RAID1) && + (type & BTRFS_BLOCK_GROUP_DUP)) { + WARN_ON(1); + type &= ~BTRFS_BLOCK_GROUP_DUP; + } dev_list = &extent_root->fs_info->fs_devices->alloc_list; if (list_empty(dev_list)) return -ENOSPC; diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 6f173450378..3f9a17f2e41 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -134,4 +134,5 @@ struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, u8 *uuid); int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); int btrfs_init_new_device(struct btrfs_root *root, char *path); +int btrfs_balance(struct btrfs_root *dev_root); #endif -- cgit v1.2.3-70-g09d2 From f2eb0a241f0e5c135d93243b0236cb1f14c305e0 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Fri, 2 May 2008 14:43:14 -0400 Subject: Btrfs: Clone file data ioctl Add a new ioctl to clone file data Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 4 +- fs/btrfs/file-item.c | 12 ++-- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 171 ++++++++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/ioctl.h | 1 + 5 files changed, 179 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 73b92dd150f..3b6f8524a4a 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1516,9 +1516,9 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root /* file-item.c */ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 objectid, u64 pos, u64 offset, + u64 objectid, u64 pos, u64 disk_offset, u64 disk_num_bytes, - u64 num_bytes); + u64 num_bytes, u64 offset); int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid, diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 9259aece6ed..f537eb43c2c 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -28,10 +28,10 @@ sizeof(struct btrfs_item) * 2) / \ BTRFS_CRC32_SIZE) - 1)) int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 objectid, u64 pos, - u64 offset, u64 disk_num_bytes, - u64 num_bytes) + struct btrfs_root *root, + u64 objectid, u64 pos, + u64 disk_offset, u64 disk_num_bytes, + u64 num_bytes, u64 offset) { int ret = 0; struct btrfs_file_extent_item *item; @@ -53,9 +53,9 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, leaf = path->nodes[0]; item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - btrfs_set_file_extent_disk_bytenr(leaf, item, offset); + btrfs_set_file_extent_disk_bytenr(leaf, item, disk_offset); btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes); - btrfs_set_file_extent_offset(leaf, item, 0); + btrfs_set_file_extent_offset(leaf, item, offset); btrfs_set_file_extent_num_bytes(leaf, item, num_bytes); btrfs_set_file_extent_generation(leaf, item, trans->transid); btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 8effdf4f5d6..a50507f3056 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -285,7 +285,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, err = btrfs_insert_file_extent(trans, root, inode->i_ino, last_pos_in_file, - 0, 0, hole_size); + 0, 0, hole_size, 0); btrfs_drop_extent_cache(inode, last_pos_in_file, last_pos_in_file + hole_size -1); btrfs_check_file(root, inode); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 99483447459..c6fae29c0b9 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -19,6 +19,7 @@ #include #include #include +#include #include #include #include @@ -141,7 +142,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) cur_alloc_size = ins.offset; ret = btrfs_insert_file_extent(trans, root, inode->i_ino, start, ins.objectid, ins.offset, - ins.offset); + ins.offset, 0); inode->i_blocks += ins.offset >> 9; btrfs_check_file(root, inode); if (num_bytes < cur_alloc_size) { @@ -1227,7 +1228,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) err = btrfs_insert_file_extent(trans, root, inode->i_ino, hole_start, 0, 0, - hole_size); + hole_size, 0); btrfs_drop_extent_cache(inode, hole_start, (u64)-1); btrfs_check_file(root, inode); @@ -3100,6 +3101,170 @@ out: return ret; } +void dup_item_to_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *leaf, + int slot, + struct btrfs_key *key, + u64 destino) +{ + struct btrfs_path *cpath = btrfs_alloc_path(); + int len = btrfs_item_size_nr(leaf, slot); + int dstoff; + struct btrfs_key ckey = *key; + int ret; + + ckey.objectid = destino; + ret = btrfs_insert_empty_item(trans, root, cpath, &ckey, len); + dstoff = btrfs_item_ptr_offset(cpath->nodes[0], cpath->slots[0]); + copy_extent_buffer(cpath->nodes[0], leaf, dstoff, + btrfs_item_ptr_offset(leaf, slot), + len); + btrfs_release_path(root, cpath); +} + +long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct file *src_file; + struct inode *src; + struct btrfs_trans_handle *trans; + int ret; + u64 pos; + struct btrfs_path *path; + struct btrfs_key key; + struct extent_buffer *leaf; + u32 nritems; + int nextret; + int slot; + + src_file = fget(src_fd); + if (!src_file) + return -EBADF; + src = src_file->f_dentry->d_inode; + + ret = -EXDEV; + if (src->i_sb != inode->i_sb) + goto out_fput; + + if (inode < src) { + mutex_lock(&inode->i_mutex); + mutex_lock(&src->i_mutex); + } else { + mutex_lock(&src->i_mutex); + mutex_lock(&inode->i_mutex); + } + + ret = -ENOTEMPTY; + if (inode->i_size) + goto out_unlock; + + /* do any pending delalloc/csum calc on src, one way or + another, and lock file content */ + while (1) { + filemap_write_and_wait(src->i_mapping); + lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + if (BTRFS_I(src)->delalloc_bytes == 0) + break; + unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + } + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 0); + path = btrfs_alloc_path(); + pos = 0; + while (1) { + ret = btrfs_lookup_file_extent(trans, root, path, src->i_ino, + pos, 0); + if (ret < 0) + goto out; + if (ret > 0) { + if (path->slots[0] == 0) { + ret = 0; + goto out; + } + path->slots[0]--; + } +next_slot: + leaf = path->nodes[0]; + slot = path->slots[0]; + btrfs_item_key_to_cpu(leaf, &key, slot); + nritems = btrfs_header_nritems(leaf); + + if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || + key.objectid != src->i_ino) + goto out; + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { + struct btrfs_file_extent_item *extent; + int found_type; + pos = key.offset; + extent = btrfs_item_ptr(leaf, slot, + struct btrfs_file_extent_item); + found_type = btrfs_file_extent_type(leaf, extent); + if (found_type == BTRFS_FILE_EXTENT_REG) { + u64 len = btrfs_file_extent_num_bytes(leaf, + extent); + u64 ds = btrfs_file_extent_disk_bytenr(leaf, + extent); + u64 dl = btrfs_file_extent_disk_num_bytes(leaf, + extent); + u64 off = btrfs_file_extent_offset(leaf, + extent); + btrfs_insert_file_extent(trans, root, + inode->i_ino, pos, + ds, dl, len, off); + /* ds == 0 means there's a hole */ + if (ds != 0) { + btrfs_inc_extent_ref(trans, root, + ds, dl, + root->root_key.objectid, + trans->transid, + inode->i_ino, pos); + } + pos = key.offset + len; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + dup_item_to_inode(trans, root, path, leaf, slot, + &key, inode->i_ino); + pos = key.offset + btrfs_item_size_nr(leaf, + slot); + } + } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) + dup_item_to_inode(trans, root, path, leaf, slot, &key, + inode->i_ino); + + if (slot >= nritems - 1) { + nextret = btrfs_next_leaf(root, path); + if (nextret) + goto out; + } else { + path->slots[0]++; + } + goto next_slot; + } + +out: + btrfs_free_path(path); + ret = 0; + + inode->i_blocks = src->i_blocks; + i_size_write(inode, src->i_size); + btrfs_update_inode(trans, root, inode); + + unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + +out_unlock: + mutex_unlock(&src->i_mutex); + mutex_unlock(&inode->i_mutex); +out_fput: + fput(src_file); + return ret; +} + long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { @@ -3116,6 +3281,8 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_add_dev(root, (void __user *)arg); case BTRFS_IOC_BALANCE: return btrfs_balance(root->fs_info->dev_root); + case BTRFS_IOC_CLONE: + return btrfs_ioctl_clone(file, arg); } return -ENOTTY; diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 8ad35fc4ba5..b0e73f51d63 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -36,6 +36,7 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \ struct btrfs_ioctl_vol_args) +#define BTRFS_IOC_CLONE _IOW(BTRFS_IOCTL_MAGIC, 9, int) #define BTRFS_IOC_ADD_DEV _IOW(BTRFS_IOCTL_MAGIC, 10, \ struct btrfs_ioctl_vol_args) #define BTRFS_IOC_RM_DEV _IOW(BTRFS_IOCTL_MAGIC, 11, \ -- cgit v1.2.3-70-g09d2 From b9d86667c94e5fe4bf9f6aa500e7ff1138e717ff Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 2 May 2008 16:13:49 -0400 Subject: Btrfs: Silence bogus inode.c compiler warnings Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c6fae29c0b9..5204599e430 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2035,12 +2035,12 @@ fail: static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) { - struct inode *inode; + struct inode *inode = NULL; struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; int err = 0; int drop_on_err = 0; - u64 objectid; + u64 objectid = 0; unsigned long nr = 1; mutex_lock(&root->fs_info->fs_mutex); -- cgit v1.2.3-70-g09d2 From 5d9cd9ecbf40b8bd5045a3c2f1feb35db6a12266 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 5 May 2008 06:26:21 -0400 Subject: Btrfs: Fix clone ioctl to not hold the path over inserts Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 79 +++++++++++++++++++++++++++++++++----------------------- 1 file changed, 47 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5204599e430..f7beb9b0d37 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3101,7 +3101,7 @@ out: return ret; } -void dup_item_to_inode(struct btrfs_trans_handle *trans, +int dup_item_to_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct extent_buffer *leaf, @@ -3109,19 +3109,22 @@ void dup_item_to_inode(struct btrfs_trans_handle *trans, struct btrfs_key *key, u64 destino) { - struct btrfs_path *cpath = btrfs_alloc_path(); + char *dup; int len = btrfs_item_size_nr(leaf, slot); - int dstoff; struct btrfs_key ckey = *key; - int ret; + int ret = 0; + + dup = kmalloc(len, GFP_NOFS); + if (!dup) + return -ENOMEM; + + read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len); + btrfs_release_path(root, path); ckey.objectid = destino; - ret = btrfs_insert_empty_item(trans, root, cpath, &ckey, len); - dstoff = btrfs_item_ptr_offset(cpath->nodes[0], cpath->slots[0]); - copy_extent_buffer(cpath->nodes[0], leaf, dstoff, - btrfs_item_ptr_offset(leaf, slot), - len); - btrfs_release_path(root, cpath); + ret = btrfs_insert_item(trans, root, &ckey, dup, len); + kfree(dup); + return ret; } long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) @@ -3137,7 +3140,6 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) struct btrfs_key key; struct extent_buffer *leaf; u32 nritems; - int nextret; int slot; src_file = fget(src_fd); @@ -3174,20 +3176,32 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 0); path = btrfs_alloc_path(); + if (!path) { + ret = -ENOMEM; + goto out; + } + key.offset = 0; + key.type = BTRFS_EXTENT_DATA_KEY; + key.objectid = src->i_ino; pos = 0; + path->reada = 2; + while (1) { - ret = btrfs_lookup_file_extent(trans, root, path, src->i_ino, - pos, 0); + /* + * note the key will change type as we walk through the + * tree. + */ + ret = btrfs_search_slot(trans, root, &key, path, 0, 0); if (ret < 0) goto out; - if (ret > 0) { - if (path->slots[0] == 0) { - ret = 0; + + if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) goto out; - } - path->slots[0]--; + if (ret > 0) + break; } -next_slot: leaf = path->nodes[0]; slot = path->slots[0]; btrfs_item_key_to_cpu(leaf, &key, slot); @@ -3195,7 +3209,8 @@ next_slot: if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || key.objectid != src->i_ino) - goto out; + break; + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { struct btrfs_file_extent_item *extent; int found_type; @@ -3225,28 +3240,28 @@ next_slot: } pos = key.offset + len; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - dup_item_to_inode(trans, root, path, leaf, slot, - &key, inode->i_ino); + ret = dup_item_to_inode(trans, root, path, + leaf, slot, &key, + inode->i_ino); + if (ret) + goto out; pos = key.offset + btrfs_item_size_nr(leaf, slot); } - } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) - dup_item_to_inode(trans, root, path, leaf, slot, &key, - inode->i_ino); + } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { + ret = dup_item_to_inode(trans, root, path, leaf, + slot, &key, inode->i_ino); - if (slot >= nritems - 1) { - nextret = btrfs_next_leaf(root, path); - if (nextret) + if (ret) goto out; - } else { - path->slots[0]++; } - goto next_slot; + key.offset++; + btrfs_release_path(root, path); } + ret = 0; out: btrfs_free_path(path); - ret = 0; inode->i_blocks = src->i_blocks; i_size_write(inode, src->i_size); -- cgit v1.2.3-70-g09d2 From a061fc8da7b990faa41ca503e66faef3ecdeead0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 7 May 2008 11:43:44 -0400 Subject: Btrfs: Add support for online device removal This required a few structural changes to the code that manages bdev pointers: The VFS super block now gets an anon-bdev instead of a pointer to the lowest bdev. This allows us to avoid swapping the super block bdev pointer around at run time. The code to read in the super block no longer goes through the extent buffer interface. Things got ugly keeping the mapping constant. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 +- fs/btrfs/disk-io.c | 101 ++++++++++------------ fs/btrfs/extent-tree.c | 139 +++++++++++++++++++++++++------ fs/btrfs/inode.c | 27 +++++- fs/btrfs/super.c | 35 ++------ fs/btrfs/transaction.c | 5 +- fs/btrfs/volumes.c | 221 +++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/volumes.h | 3 + 8 files changed, 412 insertions(+), 122 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 3b6f8524a4a..33ab165591c 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -505,7 +505,7 @@ struct btrfs_fs_info { u64 alloc_start; struct btrfs_transaction *running_transaction; struct btrfs_super_block super_copy; - struct extent_buffer *sb_buffer; + struct btrfs_super_block super_for_commit; struct block_device *__bdev; struct super_block *sb; struct inode *btree_inode; @@ -1208,6 +1208,7 @@ BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64); /* struct btrfs_super_block */ BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64); +BTRFS_SETGET_STACK_FUNCS(super_flags, struct btrfs_super_block, flags, 64); BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block, generation, 64); BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index fabc31b334b..9d5424ad01a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -78,9 +78,13 @@ struct extent_map *btree_get_extent(struct inode *inode, struct page *page, spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, len); - spin_unlock(&em_tree->lock); - if (em) + if (em) { + em->bdev = + BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; + spin_unlock(&em_tree->lock); goto out; + } + spin_unlock(&em_tree->lock); em = alloc_extent_map(GFP_NOFS); if (!em) { @@ -90,7 +94,7 @@ struct extent_map *btree_get_extent(struct inode *inode, struct page *page, em->start = 0; em->len = (u64)-1; em->block_start = 0; - em->bdev = inode->i_sb->s_bdev; + em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); @@ -435,11 +439,6 @@ static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); BUG_ON(ret); - if (offset == BTRFS_SUPER_INFO_OFFSET) { - bio->bi_bdev = root->fs_info->fs_devices->latest_bdev; - submit_bio(rw, bio); - return 0; - } return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); } @@ -587,8 +586,7 @@ static int close_all_devices(struct btrfs_fs_info *fs_info) list = &fs_info->fs_devices->devices; list_for_each(next, list) { device = list_entry(next, struct btrfs_device, dev_list); - if (device->bdev && device->bdev != fs_info->sb->s_bdev) - close_bdev_excl(device->bdev); + close_bdev_excl(device->bdev); device->bdev = NULL; } return 0; @@ -1118,6 +1116,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, u32 leafsize; u32 blocksize; u32 stripesize; + struct buffer_head *bh; struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), GFP_NOFS); struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), @@ -1153,7 +1152,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, spin_lock_init(&fs_info->new_trans_lock); init_completion(&fs_info->kobj_unregister); - sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE); fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; fs_info->chunk_root = chunk_root; @@ -1170,6 +1168,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; + sb->s_blocksize = 4096; + sb->s_blocksize_bits = blksize_bits(4096); + /* * we set the i_size on the btree inode to the max possible int. * the real end of the address space is determined by all of @@ -1229,19 +1230,16 @@ struct btrfs_root *open_ctree(struct super_block *sb, __setup_root(4096, 4096, 4096, 4096, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); - fs_info->sb_buffer = read_tree_block(tree_root, - BTRFS_SUPER_INFO_OFFSET, - 4096); - if (!fs_info->sb_buffer) + bh = __bread(fs_devices->latest_bdev, + BTRFS_SUPER_INFO_OFFSET / 4096, 4096); + if (!bh) goto fail_iput; - read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0, - sizeof(fs_info->super_copy)); + memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); + brelse(bh); - read_extent_buffer(fs_info->sb_buffer, fs_info->fsid, - (unsigned long)btrfs_super_fsid(fs_info->sb_buffer), - BTRFS_FSID_SIZE); + memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); disk_super = &fs_info->super_copy; if (!btrfs_super_root(disk_super)) @@ -1263,7 +1261,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, tree_root->leafsize = leafsize; tree_root->sectorsize = sectorsize; tree_root->stripesize = stripesize; - sb_set_blocksize(sb, sectorsize); + + sb->s_blocksize = sectorsize; + sb->s_blocksize_bits = blksize_bits(sectorsize); if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, sizeof(disk_super->magic))) { @@ -1339,7 +1339,6 @@ fail_tree_root: fail_sys_array: mutex_unlock(&fs_info->fs_mutex); fail_sb_buffer: - free_extent_buffer(fs_info->sb_buffer); extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); fail_iput: iput(fs_info->btree_inode); @@ -1380,41 +1379,44 @@ int write_all_supers(struct btrfs_root *root) struct list_head *cur; struct list_head *head = &root->fs_info->fs_devices->devices; struct btrfs_device *dev; - struct extent_buffer *sb; + struct btrfs_super_block *sb; struct btrfs_dev_item *dev_item; struct buffer_head *bh; int ret; int do_barriers; int max_errors; int total_errors = 0; + u32 crc; + u64 flags; max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; do_barriers = !btrfs_test_opt(root, NOBARRIER); - sb = root->fs_info->sb_buffer; - dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, - dev_item); + sb = &root->fs_info->super_for_commit; + dev_item = &sb->dev_item; list_for_each(cur, head) { dev = list_entry(cur, struct btrfs_device, dev_list); - btrfs_set_device_type(sb, dev_item, dev->type); - btrfs_set_device_id(sb, dev_item, dev->devid); - btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes); - btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used); - btrfs_set_device_io_align(sb, dev_item, dev->io_align); - btrfs_set_device_io_width(sb, dev_item, dev->io_width); - btrfs_set_device_sector_size(sb, dev_item, dev->sector_size); - write_extent_buffer(sb, dev->uuid, - (unsigned long)btrfs_device_uuid(dev_item), - BTRFS_UUID_SIZE); - - btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN); - csum_tree_block(root, sb, 0); - - bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / - root->fs_info->sb->s_blocksize, + btrfs_set_stack_device_type(dev_item, dev->type); + btrfs_set_stack_device_id(dev_item, dev->devid); + btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); + btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); + btrfs_set_stack_device_io_align(dev_item, dev->io_align); + btrfs_set_stack_device_io_width(dev_item, dev->io_width); + btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); + memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); + flags = btrfs_super_flags(sb); + btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); + + + crc = ~(u32)0; + crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, + BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); + btrfs_csum_final(crc, sb->csum); + + bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, BTRFS_SUPER_INFO_SIZE); - read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE); + memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); dev->pending_io = bh; get_bh(bh); @@ -1483,15 +1485,6 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root int ret; ret = write_all_supers(root); -#if 0 - if (!btrfs_test_opt(root, NOBARRIER)) - blkdev_issue_flush(sb->s_bdev, NULL); - set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super); - ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping, - super->start, super->len); - if (!btrfs_test_opt(root, NOBARRIER)) - blkdev_issue_flush(sb->s_bdev, NULL); -#endif return ret; } @@ -1570,8 +1563,6 @@ int close_ctree(struct btrfs_root *root) if (root->fs_info->dev_root->node); free_extent_buffer(root->fs_info->dev_root->node); - free_extent_buffer(fs_info->sb_buffer); - btrfs_free_block_groups(root->fs_info); del_fs_roots(fs_info); @@ -1652,7 +1643,7 @@ void btrfs_throttle(struct btrfs_root *root) { struct backing_dev_info *bdi; - bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info; + bdi = &root->fs_info->bdi; if (root->fs_info->throttles && bdi_write_congested(bdi)) { #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) congestion_wait(WRITE, HZ/20); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index f94794a9932..c0e67bde842 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -147,6 +147,8 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct u64 end; int ret; + bytenr = max_t(u64, bytenr, + BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); block_group_cache = &info->block_group_cache; ret = find_first_extent_bit(block_group_cache, bytenr, &start, &end, @@ -1059,16 +1061,25 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) } } -static u64 reduce_alloc_profile(u64 flags) +static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags) { + u64 num_devices = root->fs_info->fs_devices->num_devices; + + if (num_devices == 1) + flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0); + if (num_devices < 4) + flags &= ~BTRFS_BLOCK_GROUP_RAID10; + if ((flags & BTRFS_BLOCK_GROUP_DUP) && (flags & (BTRFS_BLOCK_GROUP_RAID1 | - BTRFS_BLOCK_GROUP_RAID10))) + BTRFS_BLOCK_GROUP_RAID10))) { flags &= ~BTRFS_BLOCK_GROUP_DUP; + } if ((flags & BTRFS_BLOCK_GROUP_RAID1) && - (flags & BTRFS_BLOCK_GROUP_RAID10)) + (flags & BTRFS_BLOCK_GROUP_RAID10)) { flags &= ~BTRFS_BLOCK_GROUP_RAID1; + } if ((flags & BTRFS_BLOCK_GROUP_RAID0) && ((flags & BTRFS_BLOCK_GROUP_RAID1) | @@ -1078,7 +1089,6 @@ static u64 reduce_alloc_profile(u64 flags) return flags; } - static int do_chunk_alloc(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, u64 alloc_bytes, u64 flags) @@ -1089,7 +1099,7 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, u64 num_bytes; int ret; - flags = reduce_alloc_profile(flags); + flags = reduce_alloc_profile(extent_root, flags); space_info = __find_space_info(extent_root->fs_info, flags); if (!space_info) { @@ -1169,6 +1179,21 @@ static int update_block_group(struct btrfs_trans_handle *trans, return 0; } +static u64 first_logical_byte(struct btrfs_root *root, u64 search_start) +{ + u64 start; + u64 end; + int ret; + ret = find_first_extent_bit(&root->fs_info->block_group_cache, + search_start, &start, &end, + BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | + BLOCK_GROUP_SYSTEM); + if (ret) + return 0; + return start; +} + + static int update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin) { @@ -1185,16 +1210,25 @@ static int update_pinned_extents(struct btrfs_root *root, } while (num > 0) { cache = btrfs_lookup_block_group(fs_info, bytenr); - WARN_ON(!cache); - len = min(num, cache->key.offset - - (bytenr - cache->key.objectid)); + if (!cache) { + u64 first = first_logical_byte(root, bytenr); + WARN_ON(first < bytenr); + len = min(first - bytenr, num); + } else { + len = min(num, cache->key.offset - + (bytenr - cache->key.objectid)); + } if (pin) { - cache->pinned += len; - cache->space_info->bytes_pinned += len; + if (cache) { + cache->pinned += len; + cache->space_info->bytes_pinned += len; + } fs_info->total_pinned += len; } else { - cache->pinned -= len; - cache->space_info->bytes_pinned -= len; + if (cache) { + cache->pinned -= len; + cache->space_info->bytes_pinned -= len; + } fs_info->total_pinned -= len; } bytenr += len; @@ -1547,7 +1581,7 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, int data) { int ret; - u64 orig_search_start = search_start; + u64 orig_search_start; struct btrfs_root * root = orig_root->fs_info->extent_root; struct btrfs_fs_info *info = root->fs_info; u64 total_needed = num_bytes; @@ -1577,6 +1611,9 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, } } + search_start = max(search_start, first_logical_byte(root, 0)); + orig_search_start = search_start; + if (search_end == (u64)-1) search_end = btrfs_super_total_bytes(&info->super_copy); @@ -1751,7 +1788,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } again: - data = reduce_alloc_profile(data); + data = reduce_alloc_profile(root, data); if (root->ref_cows) { if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { ret = do_chunk_alloc(trans, root->fs_info->extent_root, @@ -2309,6 +2346,7 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, struct file_ra_state *ra; unsigned long total_read = 0; unsigned long ra_pages; + struct btrfs_trans_handle *trans; ra = kzalloc(sizeof(*ra), GFP_NOFS); @@ -2326,9 +2364,13 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, calc_ra(i, last_index, ra_pages)); } total_read++; + if (((u64)i << PAGE_CACHE_SHIFT) > inode->i_size) + goto truncate_racing; + page = grab_cache_page(inode->i_mapping, i); - if (!page) + if (!page) { goto out_unlock; + } if (!PageUptodate(page)) { btrfs_readpage(NULL, page); lock_page(page); @@ -2350,20 +2392,33 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, lock_extent(io_tree, page_start, page_end, GFP_NOFS); - set_page_dirty(page); set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); + set_page_dirty(page); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); unlock_page(page); page_cache_release(page); - balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); } + balance_dirty_pages_ratelimited_nr(inode->i_mapping, + total_read); out_unlock: kfree(ra); + trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1); + if (trans) { + btrfs_add_ordered_inode(inode); + btrfs_end_transaction(trans, BTRFS_I(inode)->root); + mark_inode_dirty(inode); + } mutex_unlock(&inode->i_mutex); return 0; + +truncate_racing: + vmtruncate(inode, inode->i_size); + balance_dirty_pages_ratelimited_nr(inode->i_mapping, + total_read); + goto out_unlock; } /* @@ -2466,6 +2521,27 @@ out: return 0; } +static int noinline del_extent_zero(struct btrfs_root *extent_root, + struct btrfs_path *path, + struct btrfs_key *extent_key) +{ + int ret; + struct btrfs_trans_handle *trans; + + trans = btrfs_start_transaction(extent_root, 1); + ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1); + if (ret > 0) { + ret = -EIO; + goto out; + } + if (ret < 0) + goto out; + ret = btrfs_del_item(trans, extent_root, path); +out: + btrfs_end_transaction(trans, extent_root); + return ret; +} + static int noinline relocate_one_extent(struct btrfs_root *extent_root, struct btrfs_path *path, struct btrfs_key *extent_key) @@ -2477,6 +2553,10 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, u32 item_size; int ret = 0; + if (extent_key->objectid == 0) { + ret = del_extent_zero(extent_root, path, extent_key); + goto out; + } key.objectid = extent_key->objectid; key.type = BTRFS_EXTENT_REF_KEY; key.offset = 0; @@ -2490,15 +2570,24 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, ret = 0; leaf = path->nodes[0]; nritems = btrfs_header_nritems(leaf); - if (path->slots[0] == nritems) - goto out; + if (path->slots[0] == nritems) { + ret = btrfs_next_leaf(extent_root, path); + if (ret > 0) { + ret = 0; + goto out; + } + if (ret < 0) + goto out; + } btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (found_key.objectid != extent_key->objectid) + if (found_key.objectid != extent_key->objectid) { break; + } - if (found_key.type != BTRFS_EXTENT_REF_KEY) + if (found_key.type != BTRFS_EXTENT_REF_KEY) { break; + } key.offset = found_key.offset + 1; item_size = btrfs_item_size_nr(leaf, path->slots[0]); @@ -2519,7 +2608,7 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) u64 stripped = BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10; - num_devices = btrfs_super_num_devices(&root->fs_info->super_copy); + num_devices = root->fs_info->fs_devices->num_devices; if (num_devices == 1) { stripped |= BTRFS_BLOCK_GROUP_DUP; stripped = flags & ~stripped; @@ -2535,9 +2624,6 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) return flags; } else { /* they already had raid on here, just return */ - if ((flags & BTRFS_BLOCK_GROUP_DUP) && - (flags & BTRFS_BLOCK_GROUP_RAID1)) { - } if (flags & stripped) return flags; @@ -2570,7 +2656,7 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) struct extent_buffer *leaf; u32 nritems; int ret; - int progress = 0; + int progress; shrink_block_group = btrfs_lookup_block_group(root->fs_info, shrink_start); @@ -2597,6 +2683,7 @@ again: shrink_block_group->ro = 1; total_found = 0; + progress = 0; key.objectid = shrink_start; key.offset = 0; key.type = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f7beb9b0d37..b437d3bdf95 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2194,6 +2194,8 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, again: spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, len); + if (em) + em->bdev = root->fs_info->fs_devices->latest_bdev; spin_unlock(&em_tree->lock); if (em) { @@ -2212,7 +2214,7 @@ again: em->start = EXTENT_MAP_HOLE; em->len = (u64)-1; - em->bdev = inode->i_sb->s_bdev; + em->bdev = root->fs_info->fs_devices->latest_bdev; ret = btrfs_lookup_file_extent(trans, root, path, objectid, start, trans != NULL); if (ret < 0) { @@ -3101,6 +3103,27 @@ out: return ret; } +long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) +{ + struct btrfs_ioctl_vol_args *vol_args; + int ret; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + ret = btrfs_rm_device(root, vol_args->name); + +out: + kfree(vol_args); + return ret; +} + int dup_item_to_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, @@ -3294,6 +3317,8 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_resize(root, (void __user *)arg); case BTRFS_IOC_ADD_DEV: return btrfs_ioctl_add_dev(root, (void __user *)arg); + case BTRFS_IOC_RM_DEV: + return btrfs_ioctl_rm_dev(root, (void __user *)arg); case BTRFS_IOC_BALANCE: return btrfs_balance(root->fs_info->dev_root); case BTRFS_IOC_CLONE: diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 7153dfaa340..020e5a83e31 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -315,24 +315,12 @@ static void btrfs_write_super(struct super_block *sb) sb->s_dirt = 0; } -/* - * This is almost a copy of get_sb_bdev in fs/super.c. - * We need the local copy to allow direct mounting of - * subvolumes, but this could be easily integrated back - * into the generic version. --hch - */ - -/* start copy & paste */ -static int set_bdev_super(struct super_block *s, void *data) +static int btrfs_test_super(struct super_block *s, void *data) { - s->s_bdev = data; - s->s_dev = s->s_bdev->bd_dev; - return 0; -} + struct btrfs_fs_devices *test_fs_devices = data; + struct btrfs_root *root = btrfs_sb(s); -static int test_bdev_super(struct super_block *s, void *data) -{ - return (void *)s->s_bdev == data; + return root->fs_info->fs_devices == test_fs_devices; } int btrfs_get_sb_bdev(struct file_system_type *fs_type, @@ -354,14 +342,9 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type, return error; bdev = fs_devices->lowest_bdev; - /* - * once the super is inserted into the list by sget, s_umount - * will protect the lockfs code from trying to start a snapshot - * while we are mounting - */ - down(&bdev->bd_mount_sem); - s = sget(fs_type, test_bdev_super, set_bdev_super, bdev); - up(&bdev->bd_mount_sem); + btrfs_lock_volumes(); + s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices); + btrfs_unlock_volumes(); if (IS_ERR(s)) goto error_s; @@ -373,13 +356,11 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type, goto error_bdev; } - close_bdev_excl(bdev); } else { char b[BDEVNAME_SIZE]; s->s_flags = flags; strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); - sb_set_blocksize(s, block_size(bdev)); error = btrfs_fill_super(s, fs_devices, data, flags & MS_SILENT ? 1 : 0); if (error) { @@ -458,7 +439,7 @@ static struct file_system_type btrfs_fs_type = { .owner = THIS_MODULE, .name = "btrfs", .get_sb = btrfs_get_sb, - .kill_sb = kill_block_super, + .kill_sb = kill_anon_super, .fs_flags = FS_REQUIRES_DEV, }; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 9826942fa18..57746c11eae 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -738,9 +738,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, chunk_root->node->start); btrfs_set_super_chunk_root_level(&root->fs_info->super_copy, btrfs_header_level(chunk_root->node)); - write_extent_buffer(root->fs_info->sb_buffer, - &root->fs_info->super_copy, 0, - sizeof(root->fs_info->super_copy)); + memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy, + sizeof(root->fs_info->super_copy)); btrfs_copy_pinned(root, pinned_copy); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index b3818757310..55da5f0c56e 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -45,6 +45,16 @@ struct map_lookup { static DEFINE_MUTEX(uuid_mutex); static LIST_HEAD(fs_uuids); +void btrfs_lock_volumes(void) +{ + mutex_lock(&uuid_mutex); +} + +void btrfs_unlock_volumes(void) +{ + mutex_unlock(&uuid_mutex); +} + int btrfs_cleanup_fs_uuids(void) { struct btrfs_fs_devices *fs_devices; @@ -193,12 +203,14 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, ret = PTR_ERR(bdev); goto fail; } + set_blocksize(bdev, 4096); if (device->devid == fs_devices->latest_devid) fs_devices->latest_bdev = bdev; if (device->devid == fs_devices->lowest_devid) { fs_devices->lowest_bdev = bdev; } device->bdev = bdev; + } mutex_unlock(&uuid_mutex); return 0; @@ -393,6 +405,9 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; struct btrfs_root *root = device->dev_root; struct btrfs_key key; + struct btrfs_key found_key; + struct extent_buffer *leaf = NULL; + struct btrfs_dev_extent *extent = NULL; path = btrfs_alloc_path(); if (!path) @@ -403,8 +418,25 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, key.type = BTRFS_DEV_EXTENT_KEY; ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret > 0) { + ret = btrfs_previous_item(root, path, key.objectid, + BTRFS_DEV_EXTENT_KEY); + BUG_ON(ret); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + extent = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_dev_extent); + BUG_ON(found_key.offset > start || found_key.offset + + btrfs_dev_extent_length(leaf, extent) < start); + ret = 0; + } else if (ret == 0) { + leaf = path->nodes[0]; + extent = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_dev_extent); + } BUG_ON(ret); + device->bytes_used -= btrfs_dev_extent_length(leaf, extent); ret = btrfs_del_item(trans, root, path); BUG_ON(ret); @@ -593,6 +625,170 @@ out: return ret; } +static int btrfs_rm_dev_item(struct btrfs_root *root, + struct btrfs_device *device) +{ + int ret; + struct btrfs_path *path; + struct block_device *bdev = device->bdev; + struct btrfs_device *next_dev; + struct btrfs_key key; + u64 total_bytes; + struct btrfs_fs_devices *fs_devices; + struct btrfs_trans_handle *trans; + + root = root->fs_info->chunk_root; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + trans = btrfs_start_transaction(root, 1); + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.type = BTRFS_DEV_ITEM_KEY; + key.offset = device->devid; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) + goto out; + + if (ret > 0) { + ret = -ENOENT; + goto out; + } + + ret = btrfs_del_item(trans, root, path); + if (ret) + goto out; + + /* + * at this point, the device is zero sized. We want to + * remove it from the devices list and zero out the old super + */ + list_del_init(&device->dev_list); + list_del_init(&device->dev_alloc_list); + fs_devices = root->fs_info->fs_devices; + + next_dev = list_entry(fs_devices->devices.next, struct btrfs_device, + dev_list); + if (bdev == fs_devices->lowest_bdev) + fs_devices->lowest_bdev = next_dev->bdev; + if (bdev == root->fs_info->sb->s_bdev) + root->fs_info->sb->s_bdev = next_dev->bdev; + if (bdev == fs_devices->latest_bdev) + fs_devices->latest_bdev = next_dev->bdev; + + total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); + btrfs_set_super_total_bytes(&root->fs_info->super_copy, + total_bytes - device->total_bytes); + + total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy); + btrfs_set_super_num_devices(&root->fs_info->super_copy, + total_bytes - 1); +out: + btrfs_free_path(path); + btrfs_commit_transaction(trans, root); + return ret; +} + +int btrfs_rm_device(struct btrfs_root *root, char *device_path) +{ + struct btrfs_device *device; + struct block_device *bdev; + struct buffer_head *bh; + struct btrfs_super_block *disk_super; + u64 all_avail; + u64 devid; + int ret = 0; + + mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&uuid_mutex); + + all_avail = root->fs_info->avail_data_alloc_bits | + root->fs_info->avail_system_alloc_bits | + root->fs_info->avail_metadata_alloc_bits; + + if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) && + root->fs_info->fs_devices->num_devices <= 4) { + printk("btrfs: unable to go below four devices on raid10\n"); + ret = -EINVAL; + goto out; + } + + if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) && + root->fs_info->fs_devices->num_devices <= 2) { + printk("btrfs: unable to go below two devices on raid1\n"); + ret = -EINVAL; + goto out; + } + + bdev = open_bdev_excl(device_path, 0, root->fs_info->bdev_holder); + if (IS_ERR(bdev)) { + ret = PTR_ERR(bdev); + goto out; + } + + bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); + if (!bh) { + ret = -EIO; + goto error_close; + } + disk_super = (struct btrfs_super_block *)bh->b_data; + if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, + sizeof(disk_super->magic))) { + ret = -ENOENT; + goto error_brelse; + } + if (memcmp(disk_super->fsid, root->fs_info->fsid, BTRFS_FSID_SIZE)) { + ret = -ENOENT; + goto error_brelse; + } + devid = le64_to_cpu(disk_super->dev_item.devid); + device = btrfs_find_device(root, devid, NULL); + if (!device) { + ret = -ENOENT; + goto error_brelse; + } + + root->fs_info->fs_devices->num_devices--; + + ret = btrfs_shrink_device(device, 0); + if (ret) + goto error_brelse; + + + ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device); + if (ret) + goto error_brelse; + + /* make sure this device isn't detected as part of the FS anymore */ + memset(&disk_super->magic, 0, sizeof(disk_super->magic)); + set_buffer_dirty(bh); + sync_dirty_buffer(bh); + + brelse(bh); + + /* one close for the device struct or super_block */ + close_bdev_excl(device->bdev); + + /* one close for us */ + close_bdev_excl(device->bdev); + + kfree(device->name); + kfree(device); + ret = 0; + goto out; + +error_brelse: + brelse(bh); +error_close: + close_bdev_excl(bdev); +out: + mutex_unlock(&uuid_mutex); + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + int btrfs_init_new_device(struct btrfs_root *root, char *device_path) { struct btrfs_trans_handle *trans; @@ -831,13 +1027,17 @@ int btrfs_relocate_chunk(struct btrfs_root *root, em = lookup_extent_mapping(em_tree, chunk_offset, 1); spin_unlock(&em_tree->lock); - BUG_ON(em->start > chunk_offset || em->start + em->len < chunk_offset); + BUG_ON(em->start > chunk_offset || + em->start + em->len < chunk_offset); map = (struct map_lookup *)em->bdev; for (i = 0; i < map->num_stripes; i++) { ret = btrfs_free_dev_extent(trans, map->stripes[i].dev, map->stripes[i].physical); BUG_ON(ret); + + ret = btrfs_update_device(trans, map->stripes[i].dev); + BUG_ON(ret); } ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid, chunk_offset); @@ -847,11 +1047,8 @@ int btrfs_relocate_chunk(struct btrfs_root *root, if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) { ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset); BUG_ON(ret); - goto out; } - - spin_lock(&em_tree->lock); remove_extent_mapping(em_tree, em); kfree(map); @@ -861,7 +1058,6 @@ int btrfs_relocate_chunk(struct btrfs_root *root, free_extent_map(em); spin_unlock(&em_tree->lock); -out: /* once for us */ free_extent_map(em); @@ -1449,7 +1645,7 @@ again: return 0; if (!em) { - printk("unable to find logical %Lu\n", logical); + printk("unable to find logical %Lu len %Lu\n", logical, *length); BUG(); } @@ -1712,6 +1908,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, logical = key->offset; length = btrfs_chunk_length(leaf, chunk); + spin_lock(&map_tree->map_tree.lock); em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); spin_unlock(&map_tree->map_tree.lock); @@ -1845,7 +2042,7 @@ int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) int btrfs_read_sys_array(struct btrfs_root *root) { struct btrfs_super_block *super_copy = &root->fs_info->super_copy; - struct extent_buffer *sb = root->fs_info->sb_buffer; + struct extent_buffer *sb; struct btrfs_disk_key *disk_key; struct btrfs_chunk *chunk; u8 *ptr; @@ -1857,6 +2054,12 @@ int btrfs_read_sys_array(struct btrfs_root *root) u32 cur; struct btrfs_key key; + sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET, + BTRFS_SUPER_INFO_SIZE); + if (!sb) + return -ENOMEM; + btrfs_set_buffer_uptodate(sb); + write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE); array_size = btrfs_super_sys_array_size(super_copy); ptr = super_copy->sys_chunk_array; @@ -1867,8 +2070,7 @@ int btrfs_read_sys_array(struct btrfs_root *root) disk_key = (struct btrfs_disk_key *)ptr; btrfs_disk_key_to_cpu(&key, disk_key); - len = sizeof(*disk_key); - ptr += len; + len = sizeof(*disk_key); ptr += len; sb_ptr += len; cur += len; @@ -1887,6 +2089,7 @@ int btrfs_read_sys_array(struct btrfs_root *root) sb_ptr += len; cur += len; } + free_extent_buffer(sb); return ret; } diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index a9663e92bb1..0f94a69e6eb 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -125,6 +125,7 @@ int btrfs_close_devices(struct btrfs_fs_devices *fs_devices); int btrfs_add_device(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_device *device); +int btrfs_rm_device(struct btrfs_root *root, char *device_path); int btrfs_cleanup_fs_uuids(void); int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len); int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, @@ -136,4 +137,6 @@ struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, int btrfs_shrink_device(struct btrfs_device *device, u64 new_size); int btrfs_init_new_device(struct btrfs_root *root, char *path); int btrfs_balance(struct btrfs_root *dev_root); +void btrfs_unlock_volumes(void); +void btrfs_lock_volumes(void); #endif -- cgit v1.2.3-70-g09d2 From a68d5933a0e409592860229b35230c8e87155472 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 8 May 2008 14:11:56 -0400 Subject: Btrfs: Update nodatacow mode to support cloned single files and resizing Before, nodatacow only checked to make sure multiple roots didn't have references on a single extent. This check makes sure that multiple inodes don't have references. nodatacow needed an extra check to see if the block group was currently readonly. This way cows forced by the chunk relocation code are honored. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 +- fs/btrfs/extent-tree.c | 9 +++++++++ fs/btrfs/inode.c | 13 +++++++++---- 3 files changed, 19 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f4b4677bec7..6c2c2c4e4d2 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1325,7 +1325,7 @@ static inline struct dentry *fdentry(struct file *file) { /* extent-tree.c */ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, struct btrfs_path *count_path, - u64 first_extent); + u64 expected_owner, u64 first_extent); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 8e69b5acfef..30a5094fffa 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -721,12 +721,14 @@ out: u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, struct btrfs_path *count_path, + u64 expected_owner, u64 first_extent) { struct btrfs_root *extent_root = root->fs_info->extent_root; struct btrfs_path *path; u64 bytenr; u64 found_objectid; + u64 found_owner; u64 root_objectid = root->root_key.objectid; u32 total_count = 0; u32 cur_count; @@ -792,6 +794,13 @@ again: total_count = 2; goto out; } + if (level == -1) { + found_owner = btrfs_ref_objectid(l, ref_item); + if (found_owner != expected_owner) { + total_count = 2; + goto out; + } + } total_count = 1; path->slots[0]++; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b437d3bdf95..1bf37d15b17 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -172,6 +172,7 @@ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) u64 loops = 0; u64 total_fs_bytes; struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_block_group_cache *block_group; struct extent_buffer *leaf; int found_type; struct btrfs_path *path; @@ -230,16 +231,20 @@ again: if (bytenr == 0) goto not_found; + if (btrfs_count_snapshots_in_path(root, path, inode->i_ino, + bytenr) != 1) { + goto not_found; + } + /* * we may be called by the resizer, make sure we're inside * the limits of the FS */ - if (bytenr + extent_num_bytes > total_fs_bytes) + block_group = btrfs_lookup_block_group(root->fs_info, + bytenr); + if (!block_group || block_group->ro) goto not_found; - if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) { - goto not_found; - } start = extent_end; } else { -- cgit v1.2.3-70-g09d2 From bbaf549e0c3d28399fc5abd68020d4025ae5c3a7 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 8 May 2008 16:31:21 -0400 Subject: Btrfs: A number of nodatacow fixes Once part of a delalloc request fails the cow checks, just cow the entire range It is possible for the back references to all be from the same root, but still have snapshots against an extent. The checks are now more strict, forcing cow any time there are multiple refs against the data extent. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 20 ++++++++++++++++++++ fs/btrfs/inode.c | 8 +++----- 2 files changed, 23 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 30a5094fffa..db996f0edf0 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -731,6 +731,7 @@ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, u64 found_owner; u64 root_objectid = root->root_key.objectid; u32 total_count = 0; + u32 extent_refs; u32 cur_count; u32 nritems; int ret; @@ -767,6 +768,7 @@ again: } item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); + extent_refs = btrfs_extent_refs(l, item); while (1) { l = path->nodes[0]; nritems = btrfs_header_nritems(l); @@ -800,10 +802,28 @@ again: total_count = 2; goto out; } + /* + * nasty. we don't count a reference held by + * the running transaction. This allows nodatacow + * to avoid cow most of the time + */ + if (found_owner >= BTRFS_FIRST_FREE_OBJECTID && + btrfs_ref_generation(l, ref_item) == + root->fs_info->generation) { + extent_refs--; + } } total_count = 1; path->slots[0]++; } + /* + * if there is more than one reference against a data extent, + * we have to assume the other ref is another snapshot + */ + if (level == -1 && extent_refs > 1) { + total_count = 2; + goto out; + } if (cur_count == 0) { total_count = 0; goto out; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1bf37d15b17..a492fd238c8 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -207,9 +207,8 @@ again: btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); found_type = btrfs_key_type(&found_key); if (found_key.objectid != inode->i_ino || - found_type != BTRFS_EXTENT_DATA_KEY) { + found_type != BTRFS_EXTENT_DATA_KEY) goto not_found; - } found_type = btrfs_file_extent_type(leaf, item); extent_start = found_key.offset; @@ -245,7 +244,6 @@ again: if (!block_group || block_group->ro) goto not_found; - start = extent_end; } else { goto not_found; @@ -260,8 +258,8 @@ loop: goto again; not_found: - cow_file_range(inode, start, cow_end); - start = cow_end + 1; + cow_file_range(inode, start, end); + start = end + 1; goto loop; } -- cgit v1.2.3-70-g09d2 From 1259ab75c62462b8ffad90067b5e1f6312786a18 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 12 May 2008 13:39:03 -0400 Subject: Btrfs: Handle write errors on raid1 and raid10 When duplicate copies exist, writes are allowed to fail to one of those copies. This changeset includes a few changes that allow the FS to continue even when some IOs fail. It also adds verification of the parent generation number for btree blocks. This generation is stored in the pointer to a block, and it ensures that missed writes to are detected. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 2 +- fs/btrfs/disk-io.c | 56 ++++++++++++++++++++++++++++++++++++++---- fs/btrfs/disk-io.h | 2 +- fs/btrfs/extent-tree.c | 4 +-- fs/btrfs/extent_io.c | 49 +++++++++++++++++++++++++++++++------ fs/btrfs/extent_io.h | 7 +++++- fs/btrfs/inode.c | 66 +++++++++++++++++++++++++++++++------------------- fs/btrfs/tree-defrag.c | 7 +++--- fs/btrfs/volumes.c | 11 ++++++--- 9 files changed, 155 insertions(+), 49 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 289d71d8653..02e571e6ee6 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -379,7 +379,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, cur = btrfs_find_tree_block(root, blocknr, blocksize); if (cur) - uptodate = btrfs_buffer_uptodate(cur); + uptodate = btrfs_buffer_uptodate(cur, gen); else uptodate = 0; if (!cur || !uptodate) { diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index edee7a44f86..574b1245964 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -205,6 +205,33 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, return 0; } +static int verify_parent_transid(struct extent_io_tree *io_tree, + struct extent_buffer *eb, u64 parent_transid) +{ + int ret; + + if (!parent_transid || btrfs_header_generation(eb) == parent_transid) + return 0; + + lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); + if (extent_buffer_uptodate(io_tree, eb) && + btrfs_header_generation(eb) == parent_transid) { + ret = 0; + goto out; + } + printk("parent transid verify failed on %llu wanted %llu found %llu\n", + (unsigned long long)eb->start, + (unsigned long long)parent_transid, + (unsigned long long)btrfs_header_generation(eb)); + ret = 1; +out: + clear_extent_buffer_uptodate(io_tree, eb); + unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, + GFP_NOFS); + return ret; + +} + static int btree_read_extent_buffer_pages(struct btrfs_root *root, struct extent_buffer *eb, u64 start, u64 parent_transid) @@ -218,7 +245,8 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root, while (1) { ret = read_extent_buffer_pages(io_tree, eb, start, 1, btree_get_extent, mirror_num); - if (!ret) + if (!ret && + !verify_parent_transid(io_tree, eb, parent_transid)) return ret; num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, @@ -330,6 +358,13 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, ret = -EIO; goto err; } + if (memcmp_extent_buffer(eb, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(eb), + BTRFS_FSID_SIZE)) { + printk("bad fsid on block %Lu\n", eb->start); + ret = -EIO; + goto err; + } found_level = btrfs_header_level(eb); ret = csum_tree_block(root, eb, 1); @@ -1363,7 +1398,9 @@ static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) "I/O error on %s\n", bdevname(bh->b_bdev, b)); } - set_buffer_write_io_error(bh); + /* note, we dont' set_buffer_write_io_error because we have + * our own ways of dealing with the IO errors + */ clear_buffer_uptodate(bh); } unlock_buffer(bh); @@ -1459,7 +1496,8 @@ int write_all_supers(struct btrfs_root *root) ret = submit_bh(WRITE, bh); BUG_ON(ret); wait_on_buffer(bh); - BUG_ON(!buffer_uptodate(bh)); + if (!buffer_uptodate(bh)) + total_errors++; } else { total_errors++; } @@ -1607,10 +1645,18 @@ int close_ctree(struct btrfs_root *root) return 0; } -int btrfs_buffer_uptodate(struct extent_buffer *buf) +int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) { + int ret; struct inode *btree_inode = buf->first_page->mapping->host; - return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); + + ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); + if (!ret) + return ret; + + ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, + parent_transid); + return !ret; } int btrfs_set_buffer_uptodate(struct extent_buffer *buf) diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index e29c895d523..30d1ed293c2 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -56,7 +56,7 @@ int btrfs_insert_dev_radix(struct btrfs_root *root, void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr); int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root); void btrfs_mark_buffer_dirty(struct extent_buffer *buf); -int btrfs_buffer_uptodate(struct extent_buffer *buf); +int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid); int btrfs_set_buffer_uptodate(struct extent_buffer *buf); int wait_on_tree_block_writeback(struct btrfs_root *root, struct extent_buffer *buf); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index db07dde4a87..605018c6045 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1366,7 +1366,7 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, if (!pending) { buf = btrfs_find_tree_block(root, bytenr, num_bytes); if (buf) { - if (btrfs_buffer_uptodate(buf)) { + if (btrfs_buffer_uptodate(buf, 0)) { u64 transid = root->fs_info->running_transaction->transid; u64 header_transid = @@ -2151,7 +2151,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, continue; } next = btrfs_find_tree_block(root, bytenr, blocksize); - if (!next || !btrfs_buffer_uptodate(next)) { + if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) { free_extent_buffer(next); reada_walk_down(root, cur, path->slots[*level]); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index dd403b426ff..2a3624adc0c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1366,7 +1366,7 @@ static int end_bio_extent_writepage(struct bio *bio, unsigned int bytes_done, int err) #endif { - const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + int uptodate = err == 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct extent_state *state = bio->bi_private; struct extent_io_tree *tree = state->tree; @@ -1375,6 +1375,7 @@ static int end_bio_extent_writepage(struct bio *bio, u64 end; u64 cur; int whole_page; + int ret; unsigned long flags; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) @@ -1395,17 +1396,30 @@ static int end_bio_extent_writepage(struct bio *bio, if (--bvec >= bio->bi_io_vec) prefetchw(&bvec->bv_page->flags); + if (tree->ops && tree->ops->writepage_end_io_hook) { + ret = tree->ops->writepage_end_io_hook(page, start, + end, state); + if (ret) + uptodate = 0; + } + + if (!uptodate && tree->ops && + tree->ops->writepage_io_failed_hook) { + ret = tree->ops->writepage_io_failed_hook(bio, page, + start, end, state); + if (ret == 0) { + state = NULL; + uptodate = (err == 0); + continue; + } + } + if (!uptodate) { clear_extent_uptodate(tree, start, end, GFP_ATOMIC); ClearPageUptodate(page); SetPageError(page); } - if (tree->ops && tree->ops->writepage_end_io_hook) { - tree->ops->writepage_end_io_hook(page, start, end, - state); - } - /* * bios can get merged in funny ways, and so we need to * be careful with the state variable. We know the @@ -2073,9 +2087,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, } else { ret = 0; } - if (ret) + if (ret) { SetPageError(page); - else { + } else { unsigned long max_nr = end_index + 1; set_range_writeback(tree, cur, cur + iosize - 1); if (!PageWriteback(page)) { @@ -2948,6 +2962,25 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, } EXPORT_SYMBOL(set_extent_buffer_dirty); +int clear_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb) +{ + unsigned long i; + struct page *page; + unsigned long num_pages; + + num_pages = num_extent_pages(eb->start, eb->len); + eb->flags &= ~EXTENT_UPTODATE; + + clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, + GFP_NOFS); + for (i = 0; i < num_pages; i++) { + page = extent_buffer_page(eb, i); + ClearPageUptodate(page); + } + return 0; +} + int set_extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb) { diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index e4834614751..f1960dafaa1 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -36,9 +36,12 @@ struct extent_io_ops { int (*readpage_io_failed_hook)(struct bio *bio, struct page *page, u64 start, u64 end, struct extent_state *state); + int (*writepage_io_failed_hook)(struct bio *bio, struct page *page, + u64 start, u64 end, + struct extent_state *state); int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); - void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, + int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); int (*set_bit_hook)(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits); @@ -212,6 +215,8 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, struct extent_buffer *eb); int set_extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb); +int clear_extent_buffer_uptodate(struct extent_io_tree *tree, + struct extent_buffer *eb); int extent_buffer_uptodate(struct extent_io_tree *tree, struct extent_buffer *eb); int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a492fd238c8..08760ff9bab 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -430,9 +430,9 @@ struct io_failure_record { int last_mirror; }; -int btrfs_readpage_io_failed_hook(struct bio *failed_bio, - struct page *page, u64 start, u64 end, - struct extent_state *state) +int btrfs_io_failed_hook(struct bio *failed_bio, + struct page *page, u64 start, u64 end, + struct extent_state *state) { struct io_failure_record *failrec = NULL; u64 private; @@ -443,6 +443,7 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, struct bio *bio; int num_copies; int ret; + int rw; u64 logical; ret = get_state_private(failure_tree, start, &private); @@ -505,7 +506,41 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio, bio->bi_bdev = failed_bio->bi_bdev; bio->bi_size = 0; bio_add_page(bio, page, failrec->len, start - page_offset(page)); - btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror); + if (failed_bio->bi_rw & (1 << BIO_RW)) + rw = WRITE; + else + rw = READ; + + BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio, + failrec->last_mirror); + return 0; +} + +int btrfs_clean_io_failures(struct inode *inode, u64 start) +{ + u64 private; + u64 private_failure; + struct io_failure_record *failure; + int ret; + + private = 0; + if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private, + (u64)-1, 1, EXTENT_DIRTY)) { + ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, + start, &private_failure); + if (ret == 0) { + failure = (struct io_failure_record *)(unsigned long) + private_failure; + set_state_private(&BTRFS_I(inode)->io_failure_tree, + failure->start, 0); + clear_extent_bits(&BTRFS_I(inode)->io_failure_tree, + failure->start, + failure->start + failure->len - 1, + EXTENT_DIRTY | EXTENT_LOCKED, + GFP_NOFS); + kfree(failure); + } + } return 0; } @@ -547,26 +582,7 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, /* if the io failure tree for this inode is non-empty, * check to see if we've recovered from a failed IO */ - private = 0; - if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private, - (u64)-1, 1, EXTENT_DIRTY)) { - u64 private_failure; - struct io_failure_record *failure; - ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, - start, &private_failure); - if (ret == 0) { - failure = (struct io_failure_record *)(unsigned long) - private_failure; - set_state_private(&BTRFS_I(inode)->io_failure_tree, - failure->start, 0); - clear_extent_bits(&BTRFS_I(inode)->io_failure_tree, - failure->start, - failure->start + failure->len - 1, - EXTENT_DIRTY | EXTENT_LOCKED, - GFP_NOFS); - kfree(failure); - } - } + btrfs_clean_io_failures(inode, start); return 0; zeroit: @@ -3657,7 +3673,7 @@ static struct extent_io_ops btrfs_extent_io_ops = { .merge_bio_hook = btrfs_merge_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, - .readpage_io_failed_hook = btrfs_readpage_io_failed_hook, + .readpage_io_failed_hook = btrfs_io_failed_hook, .set_bit_hook = btrfs_set_bit_hook, .clear_bit_hook = btrfs_clear_bit_hook, }; diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 5085e9e693b..c02e2bf2f02 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -51,6 +51,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, struct extent_buffer *next; struct extent_buffer *cur; u64 bytenr; + u64 ptr_gen; int ret = 0; int is_extent = 0; @@ -93,11 +94,12 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, break; } bytenr = btrfs_node_blockptr(cur, path->slots[*level]); + ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]); if (cache_only) { next = btrfs_find_tree_block(root, bytenr, btrfs_level_size(root, *level - 1)); - if (!next || !btrfs_buffer_uptodate(next) || + if (!next || !btrfs_buffer_uptodate(next, ptr_gen) || !btrfs_buffer_defrag(next)) { free_extent_buffer(next); path->slots[*level]++; @@ -106,8 +108,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans, } else { next = read_tree_block(root, bytenr, btrfs_level_size(root, *level - 1), - btrfs_node_ptr_generation(cur, - path->slots[*level])); + ptr_gen); } ret = btrfs_cow_block(trans, root, next, path->nodes[*level], path->slots[*level], &next); diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index b5d7bd1915b..5fc7fb48147 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1807,14 +1807,19 @@ static int end_bio_multi_stripe(struct bio *bio, if (atomic_dec_and_test(&multi->stripes_pending)) { bio->bi_private = multi->private; bio->bi_end_io = multi->end_io; - /* only send an error to the higher layers if it is * beyond the tolerance of the multi-bio */ - if (atomic_read(&multi->error) > multi->max_errors) + if (atomic_read(&multi->error) > multi->max_errors) { err = -EIO; - else + } else if (err) { + /* + * this bio is actually up to date, we didn't + * go over the max number of errors + */ + set_bit(BIO_UPTODATE, &bio->bi_flags); err = 0; + } kfree(multi); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) -- cgit v1.2.3-70-g09d2 From 211c17f51f46dc6c308c742098273dd46b5ca59c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 15 May 2008 09:13:45 -0400 Subject: Fix corners in writepage and btrfs_truncate_page The extent_io writepage calls needed an extra check for discarding pages that started on th last byte in the file. btrfs_truncate_page needed checks to make sure the page was still part of the file after reading it, and most importantly, needed to wait for all IO to the page to finish before freeing the corresponding extents on disk. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 15 +++++++++------ fs/btrfs/file.c | 8 -------- fs/btrfs/inode.c | 9 ++++++++- 3 files changed, 17 insertions(+), 15 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 2a3624adc0c..17c508a941e 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1988,9 +1988,12 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 nr_delalloc; u64 delalloc_end; + WARN_ON(!PageLocked(page)); - if (page->index > end_index) { - clear_extent_dirty(tree, start, page_end, GFP_NOFS); + page_offset = i_size & (PAGE_CACHE_SIZE - 1); + if (page->index > end_index || + (page->index == end_index && !page_offset)) { + page->mapping->a_ops->invalidatepage(page, 0); unlock_page(page); return 0; } @@ -1998,13 +2001,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, if (page->index == end_index) { char *userpage; - size_t offset = i_size & (PAGE_CACHE_SIZE - 1); - userpage = kmap_atomic(page, KM_USER0); - memset(userpage + offset, 0, PAGE_CACHE_SIZE - offset); - flush_dcache_page(page); + memset(userpage + page_offset, 0, + PAGE_CACHE_SIZE - page_offset); kunmap_atomic(userpage, KM_USER0); + flush_dcache_page(page); } + page_offset = 0; set_page_extent_mapped(page); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index df97d470812..d968900c697 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -305,7 +305,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, (inline_size & (root->sectorsize -1)) == 0 || inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { u64 last_end; - u64 existing_delalloc = 0; for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; @@ -315,13 +314,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, last_end = (u64)(pages[num_pages -1]->index) << PAGE_CACHE_SHIFT; last_end += PAGE_CACHE_SIZE - 1; - if (start_pos < isize) { - u64 delalloc_start = start_pos; - existing_delalloc = count_range_bits(io_tree, - &delalloc_start, - end_of_last_block, (u64)-1, - EXTENT_DELALLOC); - } set_extent_delalloc(io_tree, start_pos, end_of_last_block, GFP_NOFS); btrfs_add_ordered_inode(inode); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 08760ff9bab..40b4a8ec17f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1180,19 +1180,26 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) goto out; ret = -ENOMEM; +again: page = grab_cache_page(mapping, index); if (!page) goto out; if (!PageUptodate(page)) { ret = btrfs_readpage(NULL, page); lock_page(page); + if (page->mapping != mapping) { + unlock_page(page); + page_cache_release(page); + goto again; + } if (!PageUptodate(page)) { ret = -EIO; goto out; } } - page_start = (u64)page->index << PAGE_CACHE_SHIFT; + page_start = (u64)page->index << PAGE_CACHE_SHIFT; + wait_on_page_writeback(page); ret = btrfs_cow_one_page(inode, page, offset); unlock_page(page); -- cgit v1.2.3-70-g09d2 From e1b81e6761bd8419146d4bbe0aadd8b2d348a01c Mon Sep 17 00:00:00 2001 From: Mingming Date: Tue, 27 May 2008 10:55:43 -0400 Subject: btrfs delete ordered inode handling fix Use btrfs_release_file instead of a put_inode call Signed-off-by: Chris Mason --- fs/btrfs/file.c | 7 +++++++ fs/btrfs/inode.c | 24 +----------------------- fs/btrfs/ordered-data.c | 21 ++++++++++++++------- fs/btrfs/ordered-data.h | 2 +- fs/btrfs/super.c | 1 - 5 files changed, 23 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index d968900c697..c4fa4664a45 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -978,6 +978,12 @@ out_nolock: return num_written ? num_written : err; } +static int btrfs_release_file (struct inode * inode, struct file * filp) +{ + btrfs_del_ordered_inode(inode); + return 0; +} + static int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) { @@ -1044,6 +1050,7 @@ struct file_operations btrfs_file_operations = { .write = btrfs_file_write, .mmap = btrfs_file_mmap, .open = generic_file_open, + .release = btrfs_release_file, .fsync = btrfs_sync_file, .unlocked_ioctl = btrfs_ioctl, #ifdef CONFIG_COMPAT diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 40b4a8ec17f..1569fb86451 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -857,15 +857,11 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) nr = trans->blocks_used; if (inode->i_nlink == 0) { - int found; /* if the inode isn't linked anywhere, * we don't need to worry about * data=ordered */ - found = btrfs_del_ordered_inode(inode); - if (found == 1) { - atomic_dec(&inode->i_count); - } + btrfs_del_ordered_inode(inode); } btrfs_end_transaction(trans, root); @@ -1271,24 +1267,6 @@ fail: return err; } -void btrfs_put_inode(struct inode *inode) -{ - int ret; - - if (!BTRFS_I(inode)->ordered_trans) { - return; - } - - if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || - mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK)) - return; - - ret = btrfs_del_ordered_inode(inode); - if (ret == 1) { - atomic_dec(&inode->i_count); - } -} - void btrfs_delete_inode(struct inode *inode) { struct btrfs_trans_handle *trans; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 3814cb0246b..8dd8180183f 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -231,7 +231,7 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, return 1; } -static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, +static void __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, struct inode *inode, u64 root_objectid, u64 objectid) { @@ -243,31 +243,38 @@ static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, node = __tree_search(&tree->tree, root_objectid, objectid, &prev); if (!node) { write_unlock(&tree->lock); - return 0; + return; } rb_erase(node, &tree->tree); BTRFS_I(inode)->ordered_trans = 0; write_unlock(&tree->lock); + atomic_dec(&inode->i_count); entry = rb_entry(node, struct tree_entry, rb_node); kfree(entry); - return 1; + return; } -int btrfs_del_ordered_inode(struct inode *inode) +void btrfs_del_ordered_inode(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 root_objectid = root->root_key.objectid; - int ret = 0; + + if (!BTRFS_I(inode)->ordered_trans) { + return; + } + + if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || + mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK)) + return; spin_lock(&root->fs_info->new_trans_lock); if (root->fs_info->running_transaction) { struct btrfs_ordered_inode_tree *tree; tree = &root->fs_info->running_transaction->ordered_inode_tree; - ret = __btrfs_del_ordered_inode(tree, inode, root_objectid, + __btrfs_del_ordered_inode(tree, inode, root_objectid, inode->i_ino); } spin_unlock(&root->fs_info->new_trans_lock); - return ret; } int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode) diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 29047e0abaa..c515c4b3999 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -38,6 +38,6 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, u64 *root_objectid, u64 *objectid, struct inode **inode); -int btrfs_del_ordered_inode(struct inode *inode); +void btrfs_del_ordered_inode(struct inode *inode); int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode); #endif diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 273a5b511f5..05029656e42 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -487,7 +487,6 @@ static void btrfs_unlockfs(struct super_block *sb) static struct super_operations btrfs_super_ops = { .delete_inode = btrfs_delete_inode, - .put_inode = btrfs_put_inode, .put_super = btrfs_put_super, .write_super = btrfs_write_super, .sync_fs = btrfs_sync_fs, -- cgit v1.2.3-70-g09d2 From 3b96362cc8d314c935c335d5c3c42eb93c23166b Mon Sep 17 00:00:00 2001 From: Sven Wegener Date: Mon, 9 Jun 2008 21:57:42 -0400 Subject: Btrfs: Invalidate dcache entry after creating snapshot and We need to invalidate an existing dcache entry after creating a new snapshot or subvolume, because a negative dache entry will stop us from accessing the new snapshot or subvolume. --- ctree.h | 23 +++++++++++++++++++++++ inode.c | 4 ++++ transaction.c | 4 ++++ 3 files changed, 31 insertions(+) Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 +++ fs/btrfs/inode.c | 29 +++++++++++++++++++++++++++++ fs/btrfs/transaction.c | 8 +++++++- 3 files changed, 39 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 7b73a9c3d86..ad4eacca7f5 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1544,6 +1544,9 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, + int namelen); + int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1569fb86451..31aa4ba06fc 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2642,6 +2642,31 @@ static void btrfs_truncate(struct inode *inode) btrfs_throttle(root); } +/* + * Invalidate a single dcache entry at the root of the filesystem. + * Needed after creation of snapshot or subvolume. + */ +void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, + int namelen) +{ + struct dentry *alias, *entry; + struct qstr qstr; + + alias = d_find_alias(root->fs_info->sb->s_root->d_inode); + if (alias) { + qstr.name = name; + qstr.len = namelen; + /* change me if btrfs ever gets a d_hash operation */ + qstr.hash = full_name_hash(qstr.name, qstr.len); + entry = d_lookup(alias, &qstr); + dput(alias); + if (entry) { + d_invalidate(entry); + dput(entry); + } + } +} + static int noinline create_subvol(struct btrfs_root *root, char *name, int namelen) { @@ -2761,6 +2786,10 @@ static int noinline create_subvol(struct btrfs_root *root, char *name, ret = btrfs_update_inode(trans, new_root, inode); if (ret) goto fail; + + /* Invalidate existing dcache entry for new subvolume. */ + btrfs_invalidate_dcache_root(root, name, namelen); + fail: nr = trans->blocks_used; err = btrfs_commit_transaction(trans, new_root); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 1cb084efd6e..f04684f7fea 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -560,6 +560,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root *root = pending->root; struct extent_buffer *tmp; int ret; + int namelen; u64 objectid; new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); @@ -595,8 +596,9 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, * insert the directory item */ key.offset = (u64)-1; + namelen = strlen(pending->name); ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, - pending->name, strlen(pending->name), + pending->name, namelen, root->fs_info->sb->s_root->d_inode->i_ino, &key, BTRFS_FT_DIR); @@ -606,6 +608,10 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, pending->name, strlen(pending->name), objectid, root->fs_info->sb->s_root->d_inode->i_ino); + + /* Invalidate existing dcache entry for new snapshot. */ + btrfs_invalidate_dcache_root(root, pending->name, namelen); + fail: kfree(new_root_item); return ret; -- cgit v1.2.3-70-g09d2 From 6bf13c0cc833bf5ba013d6aa60379484bf48c4e6 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Tue, 10 Jun 2008 10:07:39 -0400 Subject: Btrfs: transaction ioctls These ioctls let a user application hold a transaction open while it performs a series of operations. A final ioctl does a sync on the fs (closing the current transaction). This is the main requirement for Ceph's OSD to be able to keep the data it's storing in a btrfs volume consistent, and AFAICS it works just fine. The application would do something like fd = ::open("some/file", O_RDONLY); ::ioctl(fd, BTRFS_IOC_TRANS_START); /* do a bunch of stuff */ ::ioctl(fd, BTRFS_IOC_TRANS_END); or just ::close(fd); And to ensure it commits to disk, ::ioctl(fd, BTRFS_IOC_SYNC); When a transaction is held open, the trans_handle is attached to the struct file (via private_data) so that it will get cleaned up if the process dies unexpectedly. A held transaction is also ended on fsync() to avoid a deadlock. A misbehaving application could also deliberately hold a transaction open, effectively locking up the FS, so it may make sense to restrict something like this to root or something. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 4 ++++ fs/btrfs/file.c | 7 ++++++- fs/btrfs/inode.c | 63 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/ioctl.h | 9 ++++++++ fs/btrfs/super.c | 2 +- 5 files changed, 83 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ad4eacca7f5..1dcf4fb5b68 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1575,6 +1575,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb); void btrfs_destroy_inode(struct inode *inode); int btrfs_init_cachep(void); void btrfs_destroy_cachep(void); +long btrfs_ioctl_trans_end(struct file *file); long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); @@ -1595,6 +1596,8 @@ extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 start, u64 end, u64 inline_limit, u64 *hint_block); +int btrfs_release_file(struct inode *inode, struct file *file); + /* tree-defrag.c */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only); @@ -1615,4 +1618,5 @@ int btrfs_delete_xattrs(struct btrfs_trans_handle *trans, u64 btrfs_parse_size(char *str); int btrfs_parse_options(char *options, struct btrfs_root *root, char **subvol_name); +int btrfs_sync_fs(struct super_block *sb, int wait); #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index c4fa4664a45..73c6d085bd9 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -978,9 +978,11 @@ out_nolock: return num_written ? num_written : err; } -static int btrfs_release_file (struct inode * inode, struct file * filp) +int btrfs_release_file(struct inode * inode, struct file * filp) { btrfs_del_ordered_inode(inode); + if (filp->private_data) + btrfs_ioctl_trans_end(filp); return 0; } @@ -1011,6 +1013,9 @@ static int btrfs_sync_file(struct file *file, /* * ok we haven't committed the transaction yet, lets do a commit */ + if (file->private_data) + btrfs_ioctl_trans_end(file); + trans = btrfs_start_transaction(root, 1); if (!trans) { ret = -ENOMEM; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 31aa4ba06fc..0f14697bece 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3336,6 +3336,61 @@ out_fput: return ret; } +/* + * there are many ways the trans_start and trans_end ioctls can lead + * to deadlocks. They should only be used by applications that + * basically own the machine, and have a very in depth understanding + * of all the possible deadlocks and enospc problems. + */ +long btrfs_ioctl_trans_start(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + mutex_lock(&root->fs_info->fs_mutex); + if (file->private_data) { + ret = -EINPROGRESS; + goto out; + } + trans = btrfs_start_transaction(root, 0); + if (trans) + file->private_data = trans; + else + ret = -ENOMEM; + /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ +out: + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +/* + * there are many ways the trans_start and trans_end ioctls can lead + * to deadlocks. They should only be used by applications that + * basically own the machine, and have a very in depth understanding + * of all the possible deadlocks and enospc problems. + */ +long btrfs_ioctl_trans_end(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + mutex_lock(&root->fs_info->fs_mutex); + trans = file->private_data; + if (!trans) { + ret = -EINVAL; + goto out; + } + btrfs_end_transaction(trans, root); + file->private_data = 0; +out: + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { @@ -3356,6 +3411,13 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_balance(root->fs_info->dev_root); case BTRFS_IOC_CLONE: return btrfs_ioctl_clone(file, arg); + case BTRFS_IOC_TRANS_START: + return btrfs_ioctl_trans_start(file); + case BTRFS_IOC_TRANS_END: + return btrfs_ioctl_trans_end(file); + case BTRFS_IOC_SYNC: + btrfs_sync_fs(file->f_dentry->d_sb, 1); + return 0; } return -ENOTTY; @@ -3679,6 +3741,7 @@ static struct file_operations btrfs_dir_file_operations = { #ifdef CONFIG_COMPAT .compat_ioctl = btrfs_ioctl, #endif + .release = btrfs_release_file, }; static struct extent_io_ops btrfs_extent_io_ops = { diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index b0e73f51d63..85ed35a775b 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -36,6 +36,14 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \ struct btrfs_ioctl_vol_args) +/* trans start and trans end are dangerous, and only for + * use by applications that know how to avoid the + * resulting deadlocks + */ +#define BTRFS_IOC_TRANS_START _IO(BTRFS_IOCTL_MAGIC, 6) +#define BTRFS_IOC_TRANS_END _IO(BTRFS_IOCTL_MAGIC, 7) +#define BTRFS_IOC_SYNC _IO(BTRFS_IOCTL_MAGIC, 8) + #define BTRFS_IOC_CLONE _IOW(BTRFS_IOCTL_MAGIC, 9, int) #define BTRFS_IOC_ADD_DEV _IOW(BTRFS_IOCTL_MAGIC, 10, \ struct btrfs_ioctl_vol_args) @@ -43,4 +51,5 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \ struct btrfs_ioctl_vol_args) + #endif diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 77f44494e22..39bb86945ed 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -293,7 +293,7 @@ fail_close: return err; } -static int btrfs_sync_fs(struct super_block *sb, int wait) +int btrfs_sync_fs(struct super_block *sb, int wait) { struct btrfs_trans_handle *trans; struct btrfs_root *root; -- cgit v1.2.3-70-g09d2 From 8b7128429235d9bd72cfd5ed20c77c4f3118f744 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 11 Jun 2008 16:50:36 -0400 Subject: Btrfs: Add async worker threads for pre and post IO checksumming Btrfs has been using workqueues to spread the checksumming load across other CPUs in the system. But, workqueues only schedule work on the same CPU that queued the work, giving them a limited benefit for systems with higher CPU counts. This code adds a generic facility to schedule work with pools of kthreads, and changes the bio submission code to queue bios up. The queueing is important to make sure large numbers of procs on the system don't turn streaming workloads into random workloads by sending IO down concurrently. The end result of all of this is much higher performance (and CPU usage) when doing checksumming on large machines. Two worker pools are created, one for writes and one for endio processing. The two could deadlock if we tried to service both from a single pool. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/async-thread.c | 288 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/async-thread.h | 78 +++++++++++++ fs/btrfs/ctree.h | 14 ++- fs/btrfs/disk-io.c | 200 ++++++++++++++------------------- fs/btrfs/inode.c | 4 +- fs/btrfs/volumes.c | 162 ++++++++++++++++++++++++++- fs/btrfs/volumes.h | 10 +- 8 files changed, 626 insertions(+), 132 deletions(-) create mode 100644 fs/btrfs/async-thread.c create mode 100644 fs/btrfs/async-thread.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index e5fc3cfea0a..9dcfc2fe333 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ - extent_io.o volumes.o + extent_io.o volumes.o async-thread.o btrfs-$(CONFIG_FS_POSIX_ACL) += acl.o else diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c new file mode 100644 index 00000000000..2911b67bd6f --- /dev/null +++ b/fs/btrfs/async-thread.c @@ -0,0 +1,288 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include +#include +#include "async-thread.h" + +/* + * container for the kthread task pointer and the list of pending work + * One of these is allocated per thread. + */ +struct btrfs_worker_thread { + /* list of struct btrfs_work that are waiting for service */ + struct list_head pending; + + /* list of worker threads from struct btrfs_workers */ + struct list_head worker_list; + + /* kthread */ + struct task_struct *task; + + /* number of things on the pending list */ + atomic_t num_pending; + + /* protects the pending list. */ + spinlock_t lock; + + /* set to non-zero when this thread is already awake and kicking */ + int working; +}; + +/* + * main loop for servicing work items + */ +static int worker_loop(void *arg) +{ + struct btrfs_worker_thread *worker = arg; + struct list_head *cur; + struct btrfs_work *work; + do { + spin_lock_irq(&worker->lock); + while(!list_empty(&worker->pending)) { + cur = worker->pending.next; + work = list_entry(cur, struct btrfs_work, list); + list_del(&work->list); + clear_bit(0, &work->flags); + + work->worker = worker; + spin_unlock_irq(&worker->lock); + + work->func(work); + + atomic_dec(&worker->num_pending); + spin_lock_irq(&worker->lock); + } + worker->working = 0; + if (freezing(current)) { + refrigerator(); + } else { + set_current_state(TASK_INTERRUPTIBLE); + spin_unlock_irq(&worker->lock); + schedule(); + __set_current_state(TASK_RUNNING); + } + } while (!kthread_should_stop()); + return 0; +} + +/* + * this will wait for all the worker threads to shutdown + */ +int btrfs_stop_workers(struct btrfs_workers *workers) +{ + struct list_head *cur; + struct btrfs_worker_thread *worker; + + while(!list_empty(&workers->worker_list)) { + cur = workers->worker_list.next; + worker = list_entry(cur, struct btrfs_worker_thread, + worker_list); + kthread_stop(worker->task); + list_del(&worker->worker_list); + kfree(worker); + } + return 0; +} + +/* + * simple init on struct btrfs_workers + */ +void btrfs_init_workers(struct btrfs_workers *workers, int max) +{ + workers->num_workers = 0; + INIT_LIST_HEAD(&workers->worker_list); + workers->last = NULL; + spin_lock_init(&workers->lock); + workers->max_workers = max; +} + +/* + * starts new worker threads. This does not enforce the max worker + * count in case you need to temporarily go past it. + */ +int btrfs_start_workers(struct btrfs_workers *workers, int num_workers) +{ + struct btrfs_worker_thread *worker; + int ret = 0; + int i; + + for (i = 0; i < num_workers; i++) { + worker = kzalloc(sizeof(*worker), GFP_NOFS); + if (!worker) { + ret = -ENOMEM; + goto fail; + } + + INIT_LIST_HEAD(&worker->pending); + INIT_LIST_HEAD(&worker->worker_list); + spin_lock_init(&worker->lock); + atomic_set(&worker->num_pending, 0); + worker->task = kthread_run(worker_loop, worker, "btrfs"); + if (IS_ERR(worker->task)) { + ret = PTR_ERR(worker->task); + goto fail; + } + + spin_lock_irq(&workers->lock); + list_add_tail(&worker->worker_list, &workers->worker_list); + workers->last = worker; + workers->num_workers++; + spin_unlock_irq(&workers->lock); + } + return 0; +fail: + btrfs_stop_workers(workers); + return ret; +} + +/* + * run through the list and find a worker thread that doesn't have a lot + * to do right now. This can return null if we aren't yet at the thread + * count limit and all of the threads are busy. + */ +static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers) +{ + struct btrfs_worker_thread *worker; + struct list_head *next; + struct list_head *start; + int enforce_min = workers->num_workers < workers->max_workers; + + /* start with the last thread if it isn't busy */ + worker = workers->last; + if (atomic_read(&worker->num_pending) < 64) + goto done; + + next = worker->worker_list.next; + start = &worker->worker_list; + + /* + * check all the workers for someone that is bored. FIXME, do + * something smart here + */ + while(next != start) { + if (next == &workers->worker_list) { + next = workers->worker_list.next; + continue; + } + worker = list_entry(next, struct btrfs_worker_thread, + worker_list); + if (atomic_read(&worker->num_pending) < 64 || !enforce_min) + goto done; + next = next->next; + } + /* + * nobody was bored, if we're already at the max thread count, + * use the last thread + */ + if (!enforce_min || atomic_read(&workers->last->num_pending) < 64) { + return workers->last; + } + return NULL; +done: + workers->last = worker; + return worker; +} + +static struct btrfs_worker_thread *find_worker(struct btrfs_workers *workers) +{ + struct btrfs_worker_thread *worker; + unsigned long flags; + +again: + spin_lock_irqsave(&workers->lock, flags); + worker = next_worker(workers); + spin_unlock_irqrestore(&workers->lock, flags); + + if (!worker) { + spin_lock_irqsave(&workers->lock, flags); + if (workers->num_workers >= workers->max_workers) { + /* + * we have failed to find any workers, just + * return the force one + */ + worker = list_entry(workers->worker_list.next, + struct btrfs_worker_thread, worker_list); + spin_unlock_irqrestore(&workers->lock, flags); + } else { + spin_unlock_irqrestore(&workers->lock, flags); + /* we're below the limit, start another worker */ + btrfs_start_workers(workers, 1); + goto again; + } + } + return worker; +} + +/* + * btrfs_requeue_work just puts the work item back on the tail of the list + * it was taken from. It is intended for use with long running work functions + * that make some progress and want to give the cpu up for others. + */ +int btrfs_requeue_work(struct btrfs_work *work) +{ + struct btrfs_worker_thread *worker = work->worker; + unsigned long flags; + + if (test_and_set_bit(0, &work->flags)) + goto out; + + spin_lock_irqsave(&worker->lock, flags); + atomic_inc(&worker->num_pending); + list_add_tail(&work->list, &worker->pending); + spin_unlock_irqrestore(&worker->lock, flags); +out: + return 0; +} + +/* + * places a struct btrfs_work into the pending queue of one of the kthreads + */ +int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work) +{ + struct btrfs_worker_thread *worker; + unsigned long flags; + int wake = 0; + + /* don't requeue something already on a list */ + if (test_and_set_bit(0, &work->flags)) + goto out; + + worker = find_worker(workers); + + spin_lock_irqsave(&worker->lock, flags); + atomic_inc(&worker->num_pending); + list_add_tail(&work->list, &worker->pending); + + /* + * avoid calling into wake_up_process if this thread has already + * been kicked + */ + if (!worker->working) + wake = 1; + worker->working = 1; + + spin_unlock_irqrestore(&worker->lock, flags); + + if (wake) + wake_up_process(worker->task); +out: + return 0; +} diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h new file mode 100644 index 00000000000..52fc9da0f9e --- /dev/null +++ b/fs/btrfs/async-thread.h @@ -0,0 +1,78 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_ASYNC_THREAD_ +#define __BTRFS_ASYNC_THREAD_ + +struct btrfs_worker_thread; + +/* + * This is similar to a workqueue, but it is meant to spread the operations + * across all available cpus instead of just the CPU that was used to + * queue the work. There is also some batching introduced to try and + * cut down on context switches. + * + * By default threads are added on demand up to 2 * the number of cpus. + * Changing struct btrfs_workers->max_workers is one way to prevent + * demand creation of kthreads. + * + * the basic model of these worker threads is to embed a btrfs_work + * structure in your own data struct, and use container_of in a + * work function to get back to your data struct. + */ +struct btrfs_work { + /* + * only func should be set to the function you want called + * your work struct is passed as the only arg + */ + void (*func)(struct btrfs_work *work); + + /* + * flags should be set to zero. It is used to make sure the + * struct is only inserted once into the list. + */ + unsigned long flags; + + /* don't touch these */ + struct btrfs_worker_thread *worker; + struct list_head list; +}; + +struct btrfs_workers { + /* current number of running workers */ + int num_workers; + + /* max number of workers allowed. changed by btrfs_start_workers */ + int max_workers; + + /* list with all the work threads */ + struct list_head worker_list; + + /* the last worker thread to have something queued */ + struct btrfs_worker_thread *last; + + /* lock for finding the next worker thread to queue on */ + spinlock_t lock; +}; + +int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work); +int btrfs_start_workers(struct btrfs_workers *workers, int num_workers); +int btrfs_stop_workers(struct btrfs_workers *workers); +void btrfs_init_workers(struct btrfs_workers *workers, int max); +int btrfs_requeue_work(struct btrfs_work *work); +#endif diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 49cbc62b42f..6c91a510c96 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -30,6 +30,7 @@ #include "bit-radix.h" #include "extent_io.h" #include "extent_map.h" +#include "async-thread.h" struct btrfs_trans_handle; struct btrfs_transaction; @@ -518,13 +519,20 @@ struct btrfs_fs_info { struct list_head hashers; struct list_head dead_roots; struct list_head end_io_work_list; - struct list_head async_submit_work_list; struct work_struct end_io_work; - struct work_struct async_submit_work; spinlock_t end_io_work_lock; - spinlock_t async_submit_work_lock; atomic_t nr_async_submits; + /* + * there is a pool of worker threads for checksumming during writes + * and a pool for checksumming after reads. This is because readers + * can run with FS locks held, and the writers may be waiting for + * those locks. We don't want ordering in the pending list to cause + * deadlocks, and so the two are serviced separately. + */ + struct btrfs_workers workers; + struct btrfs_workers endio_workers; + #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) struct work_struct trans_work; #else diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index b9a53646ceb..98ff4fbcb38 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -31,6 +31,7 @@ #include "btrfs_inode.h" #include "volumes.h" #include "print-tree.h" +#include "async-thread.h" #if 0 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) @@ -46,8 +47,7 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) #endif static struct extent_io_ops btree_extent_io_ops; -static struct workqueue_struct *end_io_workqueue; -static struct workqueue_struct *async_submit_workqueue; +static void end_workqueue_fn(struct btrfs_work *work); struct end_io_wq { struct bio *bio; @@ -57,6 +57,7 @@ struct end_io_wq { int error; int metadata; struct list_head list; + struct btrfs_work work; }; struct async_submit_bio { @@ -66,6 +67,7 @@ struct async_submit_bio { extent_submit_bio_hook_t *submit_bio_hook; int rw; int mirror_num; + struct btrfs_work work; }; struct extent_map *btree_get_extent(struct inode *inode, struct page *page, @@ -389,7 +391,6 @@ static int end_workqueue_bio(struct bio *bio, { struct end_io_wq *end_io_wq = bio->bi_private; struct btrfs_fs_info *fs_info; - unsigned long flags; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) if (bio->bi_size) @@ -397,11 +398,10 @@ static int end_workqueue_bio(struct bio *bio, #endif fs_info = end_io_wq->info; - spin_lock_irqsave(&fs_info->end_io_work_lock, flags); end_io_wq->error = err; - list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list); - spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); - queue_work(end_io_workqueue, &fs_info->end_io_work); + end_io_wq->work.func = end_workqueue_fn; + end_io_wq->work.flags = 0; + btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) return 0; @@ -428,6 +428,19 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, return 0; } +static void run_one_async_submit(struct btrfs_work *work) +{ + struct btrfs_fs_info *fs_info; + struct async_submit_bio *async; + + async = container_of(work, struct async_submit_bio, work); + fs_info = BTRFS_I(async->inode)->root->fs_info; + atomic_dec(&fs_info->nr_async_submits); + async->submit_bio_hook(async->inode, async->rw, async->bio, + async->mirror_num); + kfree(async); +} + int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, int rw, struct bio *bio, int mirror_num, extent_submit_bio_hook_t *submit_bio_hook) @@ -443,13 +456,10 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, async->bio = bio; async->mirror_num = mirror_num; async->submit_bio_hook = submit_bio_hook; - - spin_lock(&fs_info->async_submit_work_lock); - list_add_tail(&async->list, &fs_info->async_submit_work_list); + async->work.func = run_one_async_submit; + async->work.flags = 0; atomic_inc(&fs_info->nr_async_submits); - spin_unlock(&fs_info->async_submit_work_lock); - - queue_work(async_submit_workqueue, &fs_info->async_submit_work); + btrfs_queue_worker(&fs_info->workers, &async->work); return 0; } @@ -462,19 +472,32 @@ static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, offset = bio->bi_sector << 9; + /* + * when we're called for a write, we're already in the async + * submission context. Just jump ingo btrfs_map_bio + */ if (rw & (1 << BIO_RW)) { - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, + mirror_num, 0); } + /* + * called for a read, do the setup so that checksum validation + * can happen in the async kernel threads + */ ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); BUG_ON(ret); - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); } static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { + /* + * kthread helpers are used to submit writes so that checksumming + * can happen in parallel across all CPUs + */ if (!(rw & (1 << BIO_RW))) { return __btree_submit_bio_hook(inode, rw, bio, mirror_num); } @@ -1036,95 +1059,40 @@ static int bio_ready_for_csum(struct bio *bio) return ret; } -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) -static void btrfs_end_io_csum(void *p) -#else -static void btrfs_end_io_csum(struct work_struct *work) -#endif +/* + * called by the kthread helper functions to finally call the bio end_io + * functions. This is where read checksum verification actually happens + */ +static void end_workqueue_fn(struct btrfs_work *work) { -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - struct btrfs_fs_info *fs_info = p; -#else - struct btrfs_fs_info *fs_info = container_of(work, - struct btrfs_fs_info, - end_io_work); -#endif - unsigned long flags; - struct end_io_wq *end_io_wq; struct bio *bio; - struct list_head *next; + struct end_io_wq *end_io_wq; + struct btrfs_fs_info *fs_info; int error; - int was_empty; - while(1) { - spin_lock_irqsave(&fs_info->end_io_work_lock, flags); - if (list_empty(&fs_info->end_io_work_list)) { - spin_unlock_irqrestore(&fs_info->end_io_work_lock, - flags); - return; - } - next = fs_info->end_io_work_list.next; - list_del(next); - spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); - - end_io_wq = list_entry(next, struct end_io_wq, list); - - bio = end_io_wq->bio; - if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { - spin_lock_irqsave(&fs_info->end_io_work_lock, flags); - was_empty = list_empty(&fs_info->end_io_work_list); - list_add_tail(&end_io_wq->list, - &fs_info->end_io_work_list); - spin_unlock_irqrestore(&fs_info->end_io_work_lock, - flags); - if (was_empty) - return; - continue; - } - error = end_io_wq->error; - bio->bi_private = end_io_wq->private; - bio->bi_end_io = end_io_wq->end_io; - kfree(end_io_wq); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - bio_endio(bio, bio->bi_size, error); -#else - bio_endio(bio, error); -#endif - } -} + end_io_wq = container_of(work, struct end_io_wq, work); + bio = end_io_wq->bio; + fs_info = end_io_wq->info; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) -static void btrfs_async_submit_work(void *p) -#else -static void btrfs_async_submit_work(struct work_struct *work) -#endif -{ -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - struct btrfs_fs_info *fs_info = p; + /* metadata bios are special because the whole tree block must + * be checksummed at once. This makes sure the entire block is in + * ram and up to date before trying to verify things. For + * blocksize <= pagesize, it is basically a noop + */ + if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { + btrfs_queue_worker(&fs_info->endio_workers, + &end_io_wq->work); + return; + } + error = end_io_wq->error; + bio->bi_private = end_io_wq->private; + bio->bi_end_io = end_io_wq->end_io; + kfree(end_io_wq); +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) + bio_endio(bio, bio->bi_size, error); #else - struct btrfs_fs_info *fs_info = container_of(work, - struct btrfs_fs_info, - async_submit_work); + bio_endio(bio, error); #endif - struct async_submit_bio *async; - struct list_head *next; - - while(1) { - spin_lock(&fs_info->async_submit_work_lock); - if (list_empty(&fs_info->async_submit_work_list)) { - spin_unlock(&fs_info->async_submit_work_lock); - return; - } - next = fs_info->async_submit_work_list.next; - list_del(next); - atomic_dec(&fs_info->nr_async_submits); - spin_unlock(&fs_info->async_submit_work_lock); - - async = list_entry(next, struct async_submit_bio, list); - async->submit_bio_hook(async->inode, async->rw, async->bio, - async->mirror_num); - kfree(async); - } } struct btrfs_root *open_ctree(struct super_block *sb, @@ -1155,19 +1123,11 @@ struct btrfs_root *open_ctree(struct super_block *sb, err = -ENOMEM; goto fail; } - end_io_workqueue = create_workqueue("btrfs-end-io"); - BUG_ON(!end_io_workqueue); - async_submit_workqueue = create_workqueue("btrfs-async-submit"); - INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); INIT_LIST_HEAD(&fs_info->trans_list); INIT_LIST_HEAD(&fs_info->dead_roots); INIT_LIST_HEAD(&fs_info->hashers); - INIT_LIST_HEAD(&fs_info->end_io_work_list); - INIT_LIST_HEAD(&fs_info->async_submit_work_list); spin_lock_init(&fs_info->hash_lock); - spin_lock_init(&fs_info->end_io_work_lock); - spin_lock_init(&fs_info->async_submit_work_lock); spin_lock_init(&fs_info->delalloc_lock); spin_lock_init(&fs_info->new_trans_lock); @@ -1222,13 +1182,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->do_barriers = 1; #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info); - INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work, - fs_info); INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); #else - INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum); - INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work); INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); #endif BTRFS_I(fs_info->btree_inode)->root = tree_root; @@ -1240,6 +1195,19 @@ struct btrfs_root *open_ctree(struct super_block *sb, mutex_init(&fs_info->trans_mutex); mutex_init(&fs_info->fs_mutex); + /* we need to start all the end_io workers up front because the + * queue work function gets called at interrupt time. The endio + * workers don't normally start IO, so some number of them <= the + * number of cpus is fine. They handle checksumming after a read. + * + * The other worker threads do start IO, so the max is larger than + * the number of CPUs. FIXME, tune this for huge machines + */ + btrfs_init_workers(&fs_info->workers, num_online_cpus() * 2); + btrfs_init_workers(&fs_info->endio_workers, num_online_cpus()); + btrfs_start_workers(&fs_info->workers, 1); + btrfs_start_workers(&fs_info->endio_workers, num_online_cpus()); + #if 0 ret = add_hasher(fs_info, "crc32c"); if (ret) { @@ -1375,6 +1343,8 @@ fail_sb_buffer: extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); fail_iput: iput(fs_info->btree_inode); + btrfs_stop_workers(&fs_info->workers); + btrfs_stop_workers(&fs_info->endio_workers); fail: btrfs_close_devices(fs_info->fs_devices); btrfs_mapping_tree_free(&fs_info->mapping_tree); @@ -1623,16 +1593,10 @@ int close_ctree(struct btrfs_root *root) extent_io_tree_empty_lru(&fs_info->extent_ins); extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); - flush_workqueue(async_submit_workqueue); - flush_workqueue(end_io_workqueue); - truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); - flush_workqueue(async_submit_workqueue); - destroy_workqueue(async_submit_workqueue); - - flush_workqueue(end_io_workqueue); - destroy_workqueue(end_io_workqueue); + btrfs_stop_workers(&fs_info->workers); + btrfs_stop_workers(&fs_info->endio_workers); iput(fs_info->btree_inode); #if 0 diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0f14697bece..7daef8d3700 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -359,7 +359,7 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, kfree(sums); - return btrfs_map_bio(root, rw, bio, mirror_num); + return btrfs_map_bio(root, rw, bio, mirror_num, 1); } int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, @@ -383,7 +383,7 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, inode, rw, bio, mirror_num, __btrfs_submit_bio_hook); mapit: - return btrfs_map_bio(root, rw, bio, mirror_num); + return btrfs_map_bio(root, rw, bio, mirror_num, 0); } int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 722eb455015..c57458ce633 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -27,6 +27,7 @@ #include "transaction.h" #include "print-tree.h" #include "volumes.h" +#include "async-thread.h" struct map_lookup { u64 type; @@ -110,6 +111,101 @@ static struct btrfs_fs_devices *find_fsid(u8 *fsid) return NULL; } +/* + * we try to collect pending bios for a device so we don't get a large + * number of procs sending bios down to the same device. This greatly + * improves the schedulers ability to collect and merge the bios. + * + * But, it also turns into a long list of bios to process and that is sure + * to eventually make the worker thread block. The solution here is to + * make some progress and then put this work struct back at the end of + * the list if the block device is congested. This way, multiple devices + * can make progress from a single worker thread. + */ +int run_scheduled_bios(struct btrfs_device *device) +{ + struct bio *pending; + struct backing_dev_info *bdi; + struct bio *tail; + struct bio *cur; + int again = 0; + unsigned long num_run = 0; + + bdi = device->bdev->bd_inode->i_mapping->backing_dev_info; +loop: + spin_lock(&device->io_lock); + + /* take all the bios off the list at once and process them + * later on (without the lock held). But, remember the + * tail and other pointers so the bios can be properly reinserted + * into the list if we hit congestion + */ + pending = device->pending_bios; + tail = device->pending_bio_tail; + WARN_ON(pending && !tail); + device->pending_bios = NULL; + device->pending_bio_tail = NULL; + + /* + * if pending was null this time around, no bios need processing + * at all and we can stop. Otherwise it'll loop back up again + * and do an additional check so no bios are missed. + * + * device->running_pending is used to synchronize with the + * schedule_bio code. + */ + if (pending) { + again = 1; + device->running_pending = 1; + } else { + again = 0; + device->running_pending = 0; + } + spin_unlock(&device->io_lock); + + while(pending) { + cur = pending; + pending = pending->bi_next; + cur->bi_next = NULL; + atomic_dec(&device->dev_root->fs_info->nr_async_submits); + submit_bio(cur->bi_rw, cur); + num_run++; + + /* + * we made progress, there is more work to do and the bdi + * is now congested. Back off and let other work structs + * run instead + */ + if (pending && num_run && bdi_write_congested(bdi)) { + struct bio *old_head; + + spin_lock(&device->io_lock); + old_head = device->pending_bios; + device->pending_bios = pending; + if (device->pending_bio_tail) + tail->bi_next = old_head; + else + device->pending_bio_tail = tail; + + spin_unlock(&device->io_lock); + btrfs_requeue_work(&device->work); + goto done; + } + } + if (again) + goto loop; +done: + return 0; +} + +void pending_bios_fn(struct btrfs_work *work) +{ + struct btrfs_device *device; + + device = container_of(work, struct btrfs_device, work); + run_scheduled_bios(device); +} + static int device_list_add(const char *path, struct btrfs_super_block *disk_super, u64 devid, struct btrfs_fs_devices **fs_devices_ret) @@ -141,6 +237,7 @@ static int device_list_add(const char *path, return -ENOMEM; } device->devid = devid; + device->work.func = pending_bios_fn; memcpy(device->uuid, disk_super->dev_item.uuid, BTRFS_UUID_SIZE); device->barriers = 1; @@ -925,6 +1022,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) } device->barriers = 1; + device->work.func = pending_bios_fn; generate_random_uuid(device->uuid); spin_lock_init(&device->io_lock); device->name = kstrdup(device_path, GFP_NOFS); @@ -1965,8 +2063,61 @@ static int end_bio_multi_stripe(struct bio *bio, #endif } +struct async_sched { + struct bio *bio; + int rw; + struct btrfs_fs_info *info; + struct btrfs_work work; +}; + +/* + * see run_scheduled_bios for a description of why bios are collected for + * async submit. + * + * This will add one bio to the pending list for a device and make sure + * the work struct is scheduled. + */ +int schedule_bio(struct btrfs_root *root, struct btrfs_device *device, + int rw, struct bio *bio) +{ + int should_queue = 1; + + /* don't bother with additional async steps for reads, right now */ + if (!(rw & (1 << BIO_RW))) { + submit_bio(rw, bio); + return 0; + } + + /* + * nr_async_sumbits allows us to reliably return congestion to the + * higher layers. Otherwise, the async bio makes it appear we have + * made progress against dirty pages when we've really just put it + * on a queue for later + */ + atomic_inc(&root->fs_info->nr_async_submits); + bio->bi_next = NULL; + bio->bi_rw |= rw; + + spin_lock(&device->io_lock); + + if (device->pending_bio_tail) + device->pending_bio_tail->bi_next = bio; + + device->pending_bio_tail = bio; + if (!device->pending_bios) + device->pending_bios = bio; + if (device->running_pending) + should_queue = 0; + + spin_unlock(&device->io_lock); + + if (should_queue) + btrfs_queue_worker(&root->fs_info->workers, &device->work); + return 0; +} + int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, - int mirror_num) + int mirror_num, int async_submit) { struct btrfs_mapping_tree *map_tree; struct btrfs_device *dev; @@ -2012,10 +2163,10 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, dev = multi->stripes[dev_nr].dev; if (dev && dev->bdev) { bio->bi_bdev = dev->bdev; - spin_lock(&dev->io_lock); - dev->total_ios++; - spin_unlock(&dev->io_lock); - submit_bio(rw, bio); + if (async_submit) + schedule_bio(root, dev, rw, bio); + else + submit_bio(rw, bio); } else { bio->bi_bdev = root->fs_info->fs_devices->latest_bdev; bio->bi_sector = logical >> 9; @@ -2054,6 +2205,7 @@ static struct btrfs_device *add_missing_dev(struct btrfs_root *root, device->barriers = 1; device->dev_root = root->fs_info->dev_root; device->devid = devid; + device->work.func = pending_bios_fn; fs_devices->num_devices++; spin_lock_init(&device->io_lock); memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE); diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 4df6b1608f9..48a44f7a938 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -20,6 +20,7 @@ #define __BTRFS_VOLUMES_ #include +#include "async-thread.h" struct buffer_head; struct btrfs_device { @@ -27,6 +28,9 @@ struct btrfs_device { struct list_head dev_alloc_list; struct btrfs_root *dev_root; struct buffer_head *pending_io; + struct bio *pending_bios; + struct bio *pending_bio_tail; + int running_pending; u64 generation; int barriers; @@ -36,8 +40,6 @@ struct btrfs_device { struct block_device *bdev; - u64 total_ios; - char *name; /* the internal btrfs device id */ @@ -63,6 +65,8 @@ struct btrfs_device { /* physical drive uuid (or lvm uuid) */ u8 uuid[BTRFS_UUID_SIZE]; + + struct btrfs_work work; }; struct btrfs_fs_devices { @@ -117,7 +121,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, void btrfs_mapping_init(struct btrfs_mapping_tree *tree); void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree); int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, - int mirror_num); + int mirror_num, int async_submit); int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf); int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, int flags, void *holder); -- cgit v1.2.3-70-g09d2 From f46b5a66b3316ef2f8febfe4c56e2d555e2c3979 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 11 Jun 2008 21:53:53 -0400 Subject: Btrfs: split out ioctl.c Split the ioctl handling out of inode.c into a file of it's own. Also fix up checkpatch.pl warnings for the moved code. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.h | 9 +- fs/btrfs/inode.c | 733 +------------------------------------------------- fs/btrfs/ioctl.c | 781 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 796 insertions(+), 729 deletions(-) create mode 100644 fs/btrfs/ioctl.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 9dcfc2fe333..7ed6b39e42d 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ - extent_io.o volumes.o async-thread.o + extent_io.o volumes.o async-thread.o ioctl.o btrfs-$(CONFIG_FS_POSIX_ACL) += acl.o else diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 7ae4666103c..181c81d2989 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1553,6 +1553,10 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +int btrfs_create_subvol_root(struct btrfs_root *new_root, + struct btrfs_trans_handle *trans, u64 new_dirid, + struct btrfs_block_group_cache *block_group); + void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, int namelen); @@ -1585,7 +1589,6 @@ void btrfs_destroy_inode(struct inode *inode); int btrfs_init_cachep(void); void btrfs_destroy_cachep(void); long btrfs_ioctl_trans_end(struct file *file); -long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, @@ -1598,6 +1601,10 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, int btrfs_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode); + +/* ioctl.c */ +long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); + /* file.c */ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); int btrfs_check_file(struct btrfs_root *root, struct inode *inode); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7daef8d3700..0c79346fd2c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2667,114 +2667,17 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, } } -static int noinline create_subvol(struct btrfs_root *root, char *name, - int namelen) +int btrfs_create_subvol_root(struct btrfs_root *new_root, + struct btrfs_trans_handle *trans, u64 new_dirid, + struct btrfs_block_group_cache *block_group) { - struct btrfs_trans_handle *trans; - struct btrfs_key key; - struct btrfs_root_item root_item; - struct btrfs_inode_item *inode_item; - struct extent_buffer *leaf; - struct btrfs_root *new_root = root; struct inode *inode; - struct inode *dir; int ret; - int err; - u64 objectid; - u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; - unsigned long nr = 1; - - mutex_lock(&root->fs_info->fs_mutex); - ret = btrfs_check_free_space(root, 1, 0); - if (ret) - goto fail_commit; - - trans = btrfs_start_transaction(root, 1); - BUG_ON(!trans); - - ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, - 0, &objectid); - if (ret) - goto fail; - - leaf = __btrfs_alloc_free_block(trans, root, root->leafsize, - objectid, trans->transid, 0, 0, - 0, 0); - if (IS_ERR(leaf)) - return PTR_ERR(leaf); - - btrfs_set_header_nritems(leaf, 0); - btrfs_set_header_level(leaf, 0); - btrfs_set_header_bytenr(leaf, leaf->start); - btrfs_set_header_generation(leaf, trans->transid); - btrfs_set_header_owner(leaf, objectid); - - write_extent_buffer(leaf, root->fs_info->fsid, - (unsigned long)btrfs_header_fsid(leaf), - BTRFS_FSID_SIZE); - btrfs_mark_buffer_dirty(leaf); - - inode_item = &root_item.inode; - memset(inode_item, 0, sizeof(*inode_item)); - inode_item->generation = cpu_to_le64(1); - inode_item->size = cpu_to_le64(3); - inode_item->nlink = cpu_to_le32(1); - inode_item->nblocks = cpu_to_le64(1); - inode_item->mode = cpu_to_le32(S_IFDIR | 0755); - - btrfs_set_root_bytenr(&root_item, leaf->start); - btrfs_set_root_level(&root_item, 0); - btrfs_set_root_refs(&root_item, 1); - btrfs_set_root_used(&root_item, 0); - - memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); - root_item.drop_level = 0; - - free_extent_buffer(leaf); - leaf = NULL; - - btrfs_set_root_dirid(&root_item, new_dirid); - - key.objectid = objectid; - key.offset = 1; - btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); - ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, - &root_item); - if (ret) - goto fail; - - /* - * insert the directory item - */ - key.offset = (u64)-1; - dir = root->fs_info->sb->s_root->d_inode; - ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, - name, namelen, dir->i_ino, &key, - BTRFS_FT_DIR); - if (ret) - goto fail; - - ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, - name, namelen, objectid, - root->fs_info->sb->s_root->d_inode->i_ino); - if (ret) - goto fail; - - ret = btrfs_commit_transaction(trans, root); - if (ret) - goto fail_commit; - - new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); - BUG_ON(!new_root); - - trans = btrfs_start_transaction(new_root, 1); - BUG_ON(!trans); inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid, - new_dirid, - BTRFS_I(dir)->block_group, S_IFDIR | 0700); + new_dirid, block_group, S_IFDIR | 0700); if (IS_ERR(inode)) - goto fail; + return PTR_ERR(inode); inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; new_root->inode = inode; @@ -2783,67 +2686,8 @@ static int noinline create_subvol(struct btrfs_root *root, char *name, new_dirid); inode->i_nlink = 1; inode->i_size = 0; - ret = btrfs_update_inode(trans, new_root, inode); - if (ret) - goto fail; - - /* Invalidate existing dcache entry for new subvolume. */ - btrfs_invalidate_dcache_root(root, name, namelen); -fail: - nr = trans->blocks_used; - err = btrfs_commit_transaction(trans, new_root); - if (err && !ret) - ret = err; -fail_commit: - mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); - return ret; -} - -static int create_snapshot(struct btrfs_root *root, char *name, int namelen) -{ - struct btrfs_pending_snapshot *pending_snapshot; - struct btrfs_trans_handle *trans; - int ret; - int err; - unsigned long nr = 0; - - if (!root->ref_cows) - return -EINVAL; - - mutex_lock(&root->fs_info->fs_mutex); - ret = btrfs_check_free_space(root, 1, 0); - if (ret) - goto fail_unlock; - - pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); - if (!pending_snapshot) { - ret = -ENOMEM; - goto fail_unlock; - } - pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); - if (!pending_snapshot->name) { - ret = -ENOMEM; - kfree(pending_snapshot); - goto fail_unlock; - } - memcpy(pending_snapshot->name, name, namelen); - pending_snapshot->name[namelen] = '\0'; - trans = btrfs_start_transaction(root, 1); - BUG_ON(!trans); - pending_snapshot->root = root; - list_add(&pending_snapshot->list, - &trans->transaction->pending_snapshots); - ret = btrfs_update_inode(trans, root, root->inode); - err = btrfs_commit_transaction(trans, root); - -fail_unlock: - mutex_unlock(&root->fs_info->fs_mutex); - btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); - return ret; + return btrfs_update_inode(trans, new_root, inode); } unsigned long btrfs_force_ra(struct address_space *mapping, @@ -2861,571 +2705,6 @@ unsigned long btrfs_force_ra(struct address_space *mapping, #endif } -int btrfs_defrag_file(struct file *file) { - struct inode *inode = fdentry(file)->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct page *page; - unsigned long last_index; - unsigned long ra_pages = root->fs_info->bdi.ra_pages; - unsigned long total_read = 0; - u64 page_start; - u64 page_end; - unsigned long i; - int ret; - - mutex_lock(&root->fs_info->fs_mutex); - ret = btrfs_check_free_space(root, inode->i_size, 0); - mutex_unlock(&root->fs_info->fs_mutex); - if (ret) - return -ENOSPC; - - mutex_lock(&inode->i_mutex); - last_index = inode->i_size >> PAGE_CACHE_SHIFT; - for (i = 0; i <= last_index; i++) { - if (total_read % ra_pages == 0) { - btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, - min(last_index, i + ra_pages - 1)); - } - total_read++; - page = grab_cache_page(inode->i_mapping, i); - if (!page) - goto out_unlock; - if (!PageUptodate(page)) { - btrfs_readpage(NULL, page); - lock_page(page); - if (!PageUptodate(page)) { - unlock_page(page); - page_cache_release(page); - goto out_unlock; - } - } - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - ClearPageDirty(page); -#else - cancel_dirty_page(page, PAGE_CACHE_SIZE); -#endif - wait_on_page_writeback(page); - set_page_extent_mapped(page); - - page_start = (u64)page->index << PAGE_CACHE_SHIFT; - page_end = page_start + PAGE_CACHE_SIZE - 1; - - lock_extent(io_tree, page_start, page_end, GFP_NOFS); - set_extent_delalloc(io_tree, page_start, - page_end, GFP_NOFS); - - unlock_extent(io_tree, page_start, page_end, GFP_NOFS); - set_page_dirty(page); - unlock_page(page); - page_cache_release(page); - balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); - } - -out_unlock: - mutex_unlock(&inode->i_mutex); - return 0; -} - -static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) -{ - u64 new_size; - u64 old_size; - u64 devid = 1; - struct btrfs_ioctl_vol_args *vol_args; - struct btrfs_trans_handle *trans; - struct btrfs_device *device = NULL; - char *sizestr; - char *devstr = NULL; - int ret = 0; - int namelen; - int mod = 0; - - vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); - - if (!vol_args) - return -ENOMEM; - - if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { - ret = -EFAULT; - goto out; - } - namelen = strlen(vol_args->name); - if (namelen > BTRFS_VOL_NAME_MAX) { - ret = -EINVAL; - goto out; - } - - mutex_lock(&root->fs_info->fs_mutex); - sizestr = vol_args->name; - devstr = strchr(sizestr, ':'); - if (devstr) { - char *end; - sizestr = devstr + 1; - *devstr = '\0'; - devstr = vol_args->name; - devid = simple_strtoull(devstr, &end, 10); -printk("resizing devid %Lu\n", devid); - } - device = btrfs_find_device(root, devid, NULL); - if (!device) { - printk("resizer unable to find device %Lu\n", devid); - ret = -EINVAL; - goto out_unlock; - } - if (!strcmp(sizestr, "max")) - new_size = device->bdev->bd_inode->i_size; - else { - if (sizestr[0] == '-') { - mod = -1; - sizestr++; - } else if (sizestr[0] == '+') { - mod = 1; - sizestr++; - } - new_size = btrfs_parse_size(sizestr); - if (new_size == 0) { - ret = -EINVAL; - goto out_unlock; - } - } - - old_size = device->total_bytes; - - if (mod < 0) { - if (new_size > old_size) { - ret = -EINVAL; - goto out_unlock; - } - new_size = old_size - new_size; - } else if (mod > 0) { - new_size = old_size + new_size; - } - - if (new_size < 256 * 1024 * 1024) { - ret = -EINVAL; - goto out_unlock; - } - if (new_size > device->bdev->bd_inode->i_size) { - ret = -EFBIG; - goto out_unlock; - } - - do_div(new_size, root->sectorsize); - new_size *= root->sectorsize; - -printk("new size for %s is %llu\n", device->name, (unsigned long long)new_size); - - if (new_size > old_size) { - trans = btrfs_start_transaction(root, 1); - ret = btrfs_grow_device(trans, device, new_size); - btrfs_commit_transaction(trans, root); - } else { - ret = btrfs_shrink_device(device, new_size); - } - -out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); -out: - kfree(vol_args); - return ret; -} - -static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root, - void __user *arg) -{ - struct btrfs_ioctl_vol_args *vol_args; - struct btrfs_dir_item *di; - struct btrfs_path *path; - u64 root_dirid; - int namelen; - int ret; - - vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); - - if (!vol_args) - return -ENOMEM; - - if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { - ret = -EFAULT; - goto out; - } - - namelen = strlen(vol_args->name); - if (namelen > BTRFS_VOL_NAME_MAX) { - ret = -EINVAL; - goto out; - } - if (strchr(vol_args->name, '/')) { - ret = -EINVAL; - goto out; - } - - path = btrfs_alloc_path(); - if (!path) { - ret = -ENOMEM; - goto out; - } - - root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, - mutex_lock(&root->fs_info->fs_mutex); - di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, - path, root_dirid, - vol_args->name, namelen, 0); - mutex_unlock(&root->fs_info->fs_mutex); - btrfs_free_path(path); - - if (di && !IS_ERR(di)) { - ret = -EEXIST; - goto out; - } - - if (IS_ERR(di)) { - ret = PTR_ERR(di); - goto out; - } - - if (root == root->fs_info->tree_root) - ret = create_subvol(root, vol_args->name, namelen); - else - ret = create_snapshot(root, vol_args->name, namelen); -out: - kfree(vol_args); - return ret; -} - -static int btrfs_ioctl_defrag(struct file *file) -{ - struct inode *inode = fdentry(file)->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - - switch (inode->i_mode & S_IFMT) { - case S_IFDIR: - mutex_lock(&root->fs_info->fs_mutex); - btrfs_defrag_root(root, 0); - btrfs_defrag_root(root->fs_info->extent_root, 0); - mutex_unlock(&root->fs_info->fs_mutex); - break; - case S_IFREG: - btrfs_defrag_file(file); - break; - } - - return 0; -} - -long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) -{ - struct btrfs_ioctl_vol_args *vol_args; - int ret; - - vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); - - if (!vol_args) - return -ENOMEM; - - if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { - ret = -EFAULT; - goto out; - } - ret = btrfs_init_new_device(root, vol_args->name); - -out: - kfree(vol_args); - return ret; -} - -long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) -{ - struct btrfs_ioctl_vol_args *vol_args; - int ret; - - vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); - - if (!vol_args) - return -ENOMEM; - - if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { - ret = -EFAULT; - goto out; - } - ret = btrfs_rm_device(root, vol_args->name); - -out: - kfree(vol_args); - return ret; -} - -int dup_item_to_inode(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, - struct extent_buffer *leaf, - int slot, - struct btrfs_key *key, - u64 destino) -{ - char *dup; - int len = btrfs_item_size_nr(leaf, slot); - struct btrfs_key ckey = *key; - int ret = 0; - - dup = kmalloc(len, GFP_NOFS); - if (!dup) - return -ENOMEM; - - read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len); - btrfs_release_path(root, path); - - ckey.objectid = destino; - ret = btrfs_insert_item(trans, root, &ckey, dup, len); - kfree(dup); - return ret; -} - -long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) -{ - struct inode *inode = fdentry(file)->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct file *src_file; - struct inode *src; - struct btrfs_trans_handle *trans; - int ret; - u64 pos; - struct btrfs_path *path; - struct btrfs_key key; - struct extent_buffer *leaf; - u32 nritems; - int slot; - - src_file = fget(src_fd); - if (!src_file) - return -EBADF; - src = src_file->f_dentry->d_inode; - - ret = -EXDEV; - if (src->i_sb != inode->i_sb) - goto out_fput; - - if (inode < src) { - mutex_lock(&inode->i_mutex); - mutex_lock(&src->i_mutex); - } else { - mutex_lock(&src->i_mutex); - mutex_lock(&inode->i_mutex); - } - - ret = -ENOTEMPTY; - if (inode->i_size) - goto out_unlock; - - /* do any pending delalloc/csum calc on src, one way or - another, and lock file content */ - while (1) { - filemap_write_and_wait(src->i_mapping); - lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); - if (BTRFS_I(src)->delalloc_bytes == 0) - break; - unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); - } - - mutex_lock(&root->fs_info->fs_mutex); - trans = btrfs_start_transaction(root, 0); - path = btrfs_alloc_path(); - if (!path) { - ret = -ENOMEM; - goto out; - } - key.offset = 0; - key.type = BTRFS_EXTENT_DATA_KEY; - key.objectid = src->i_ino; - pos = 0; - path->reada = 2; - - while (1) { - /* - * note the key will change type as we walk through the - * tree. - */ - ret = btrfs_search_slot(trans, root, &key, path, 0, 0); - if (ret < 0) - goto out; - - if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { - ret = btrfs_next_leaf(root, path); - if (ret < 0) - goto out; - if (ret > 0) - break; - } - leaf = path->nodes[0]; - slot = path->slots[0]; - btrfs_item_key_to_cpu(leaf, &key, slot); - nritems = btrfs_header_nritems(leaf); - - if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || - key.objectid != src->i_ino) - break; - - if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { - struct btrfs_file_extent_item *extent; - int found_type; - pos = key.offset; - extent = btrfs_item_ptr(leaf, slot, - struct btrfs_file_extent_item); - found_type = btrfs_file_extent_type(leaf, extent); - if (found_type == BTRFS_FILE_EXTENT_REG) { - u64 len = btrfs_file_extent_num_bytes(leaf, - extent); - u64 ds = btrfs_file_extent_disk_bytenr(leaf, - extent); - u64 dl = btrfs_file_extent_disk_num_bytes(leaf, - extent); - u64 off = btrfs_file_extent_offset(leaf, - extent); - btrfs_insert_file_extent(trans, root, - inode->i_ino, pos, - ds, dl, len, off); - /* ds == 0 means there's a hole */ - if (ds != 0) { - btrfs_inc_extent_ref(trans, root, - ds, dl, - root->root_key.objectid, - trans->transid, - inode->i_ino, pos); - } - pos = key.offset + len; - } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - ret = dup_item_to_inode(trans, root, path, - leaf, slot, &key, - inode->i_ino); - if (ret) - goto out; - pos = key.offset + btrfs_item_size_nr(leaf, - slot); - } - } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { - ret = dup_item_to_inode(trans, root, path, leaf, - slot, &key, inode->i_ino); - - if (ret) - goto out; - } - key.offset++; - btrfs_release_path(root, path); - } - - ret = 0; -out: - btrfs_free_path(path); - - inode->i_blocks = src->i_blocks; - i_size_write(inode, src->i_size); - btrfs_update_inode(trans, root, inode); - - unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); - - btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); - -out_unlock: - mutex_unlock(&src->i_mutex); - mutex_unlock(&inode->i_mutex); -out_fput: - fput(src_file); - return ret; -} - -/* - * there are many ways the trans_start and trans_end ioctls can lead - * to deadlocks. They should only be used by applications that - * basically own the machine, and have a very in depth understanding - * of all the possible deadlocks and enospc problems. - */ -long btrfs_ioctl_trans_start(struct file *file) -{ - struct inode *inode = fdentry(file)->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; - int ret = 0; - - mutex_lock(&root->fs_info->fs_mutex); - if (file->private_data) { - ret = -EINPROGRESS; - goto out; - } - trans = btrfs_start_transaction(root, 0); - if (trans) - file->private_data = trans; - else - ret = -ENOMEM; - /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ -out: - mutex_unlock(&root->fs_info->fs_mutex); - return ret; -} - -/* - * there are many ways the trans_start and trans_end ioctls can lead - * to deadlocks. They should only be used by applications that - * basically own the machine, and have a very in depth understanding - * of all the possible deadlocks and enospc problems. - */ -long btrfs_ioctl_trans_end(struct file *file) -{ - struct inode *inode = fdentry(file)->d_inode; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; - int ret = 0; - - mutex_lock(&root->fs_info->fs_mutex); - trans = file->private_data; - if (!trans) { - ret = -EINVAL; - goto out; - } - btrfs_end_transaction(trans, root); - file->private_data = 0; -out: - mutex_unlock(&root->fs_info->fs_mutex); - return ret; -} - -long btrfs_ioctl(struct file *file, unsigned int - cmd, unsigned long arg) -{ - struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; - - switch (cmd) { - case BTRFS_IOC_SNAP_CREATE: - return btrfs_ioctl_snap_create(root, (void __user *)arg); - case BTRFS_IOC_DEFRAG: - return btrfs_ioctl_defrag(file); - case BTRFS_IOC_RESIZE: - return btrfs_ioctl_resize(root, (void __user *)arg); - case BTRFS_IOC_ADD_DEV: - return btrfs_ioctl_add_dev(root, (void __user *)arg); - case BTRFS_IOC_RM_DEV: - return btrfs_ioctl_rm_dev(root, (void __user *)arg); - case BTRFS_IOC_BALANCE: - return btrfs_balance(root->fs_info->dev_root); - case BTRFS_IOC_CLONE: - return btrfs_ioctl_clone(file, arg); - case BTRFS_IOC_TRANS_START: - return btrfs_ioctl_trans_start(file); - case BTRFS_IOC_TRANS_END: - return btrfs_ioctl_trans_end(file); - case BTRFS_IOC_SYNC: - btrfs_sync_fs(file->f_dentry->d_sb, 1); - return 0; - } - - return -ENOTTY; -} - -/* - * Called inside transaction, so use GFP_NOFS - */ struct inode *btrfs_alloc_inode(struct super_block *sb) { struct btrfs_inode *ei; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c new file mode 100644 index 00000000000..da8de6cfdb5 --- /dev/null +++ b/fs/btrfs/ioctl.c @@ -0,0 +1,781 @@ +/* + * Copyright (C) 2007 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "ctree.h" +#include "disk-io.h" +#include "transaction.h" +#include "btrfs_inode.h" +#include "ioctl.h" +#include "print-tree.h" +#include "volumes.h" + + + +static noinline int create_subvol(struct btrfs_root *root, char *name, + int namelen) +{ + struct btrfs_trans_handle *trans; + struct btrfs_key key; + struct btrfs_root_item root_item; + struct btrfs_inode_item *inode_item; + struct extent_buffer *leaf; + struct btrfs_root *new_root = root; + struct inode *dir; + int ret; + int err; + u64 objectid; + u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; + unsigned long nr = 1; + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 0); + if (ret) + goto fail_commit; + + trans = btrfs_start_transaction(root, 1); + BUG_ON(!trans); + + ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root, + 0, &objectid); + if (ret) + goto fail; + + leaf = __btrfs_alloc_free_block(trans, root, root->leafsize, + objectid, trans->transid, 0, 0, + 0, 0); + if (IS_ERR(leaf)) + return PTR_ERR(leaf); + + btrfs_set_header_nritems(leaf, 0); + btrfs_set_header_level(leaf, 0); + btrfs_set_header_bytenr(leaf, leaf->start); + btrfs_set_header_generation(leaf, trans->transid); + btrfs_set_header_owner(leaf, objectid); + + write_extent_buffer(leaf, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(leaf), + BTRFS_FSID_SIZE); + btrfs_mark_buffer_dirty(leaf); + + inode_item = &root_item.inode; + memset(inode_item, 0, sizeof(*inode_item)); + inode_item->generation = cpu_to_le64(1); + inode_item->size = cpu_to_le64(3); + inode_item->nlink = cpu_to_le32(1); + inode_item->nblocks = cpu_to_le64(1); + inode_item->mode = cpu_to_le32(S_IFDIR | 0755); + + btrfs_set_root_bytenr(&root_item, leaf->start); + btrfs_set_root_level(&root_item, 0); + btrfs_set_root_refs(&root_item, 1); + btrfs_set_root_used(&root_item, 0); + + memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); + root_item.drop_level = 0; + + free_extent_buffer(leaf); + leaf = NULL; + + btrfs_set_root_dirid(&root_item, new_dirid); + + key.objectid = objectid; + key.offset = 1; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, + &root_item); + if (ret) + goto fail; + + /* + * insert the directory item + */ + key.offset = (u64)-1; + dir = root->fs_info->sb->s_root->d_inode; + ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, + name, namelen, dir->i_ino, &key, + BTRFS_FT_DIR); + if (ret) + goto fail; + + ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, + name, namelen, objectid, + root->fs_info->sb->s_root->d_inode->i_ino); + if (ret) + goto fail; + + ret = btrfs_commit_transaction(trans, root); + if (ret) + goto fail_commit; + + new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); + BUG_ON(!new_root); + + trans = btrfs_start_transaction(new_root, 1); + BUG_ON(!trans); + + ret = btrfs_create_subvol_root(new_root, trans, new_dirid, + BTRFS_I(dir)->block_group); + if (ret) + goto fail; + + /* Invalidate existing dcache entry for new subvolume. */ + btrfs_invalidate_dcache_root(root, name, namelen); + +fail: + nr = trans->blocks_used; + err = btrfs_commit_transaction(trans, new_root); + if (err && !ret) + ret = err; +fail_commit: + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); + return ret; +} + +static int create_snapshot(struct btrfs_root *root, char *name, int namelen) +{ + struct btrfs_pending_snapshot *pending_snapshot; + struct btrfs_trans_handle *trans; + int ret; + int err; + unsigned long nr = 0; + + if (!root->ref_cows) + return -EINVAL; + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, 1, 0); + if (ret) + goto fail_unlock; + + pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); + if (!pending_snapshot) { + ret = -ENOMEM; + goto fail_unlock; + } + pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS); + if (!pending_snapshot->name) { + ret = -ENOMEM; + kfree(pending_snapshot); + goto fail_unlock; + } + memcpy(pending_snapshot->name, name, namelen); + pending_snapshot->name[namelen] = '\0'; + trans = btrfs_start_transaction(root, 1); + BUG_ON(!trans); + pending_snapshot->root = root; + list_add(&pending_snapshot->list, + &trans->transaction->pending_snapshots); + ret = btrfs_update_inode(trans, root, root->inode); + err = btrfs_commit_transaction(trans, root); + +fail_unlock: + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_btree_balance_dirty(root, nr); + btrfs_throttle(root); + return ret; +} + +int btrfs_defrag_file(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct page *page; + unsigned long last_index; + unsigned long ra_pages = root->fs_info->bdi.ra_pages; + unsigned long total_read = 0; + u64 page_start; + u64 page_end; + unsigned long i; + int ret; + + mutex_lock(&root->fs_info->fs_mutex); + ret = btrfs_check_free_space(root, inode->i_size, 0); + mutex_unlock(&root->fs_info->fs_mutex); + if (ret) + return -ENOSPC; + + mutex_lock(&inode->i_mutex); + last_index = inode->i_size >> PAGE_CACHE_SHIFT; + for (i = 0; i <= last_index; i++) { + if (total_read % ra_pages == 0) { + btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, + min(last_index, i + ra_pages - 1)); + } + total_read++; + page = grab_cache_page(inode->i_mapping, i); + if (!page) + goto out_unlock; + if (!PageUptodate(page)) { + btrfs_readpage(NULL, page); + lock_page(page); + if (!PageUptodate(page)) { + unlock_page(page); + page_cache_release(page); + goto out_unlock; + } + } + +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + ClearPageDirty(page); +#else + cancel_dirty_page(page, PAGE_CACHE_SIZE); +#endif + wait_on_page_writeback(page); + set_page_extent_mapped(page); + + page_start = (u64)page->index << PAGE_CACHE_SHIFT; + page_end = page_start + PAGE_CACHE_SIZE - 1; + + lock_extent(io_tree, page_start, page_end, GFP_NOFS); + set_extent_delalloc(io_tree, page_start, + page_end, GFP_NOFS); + + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); + set_page_dirty(page); + unlock_page(page); + page_cache_release(page); + balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); + } + +out_unlock: + mutex_unlock(&inode->i_mutex); + return 0; +} + +/* + * Called inside transaction, so use GFP_NOFS + */ + +static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) +{ + u64 new_size; + u64 old_size; + u64 devid = 1; + struct btrfs_ioctl_vol_args *vol_args; + struct btrfs_trans_handle *trans; + struct btrfs_device *device = NULL; + char *sizestr; + char *devstr = NULL; + int ret = 0; + int namelen; + int mod = 0; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + namelen = strlen(vol_args->name); + if (namelen > BTRFS_VOL_NAME_MAX) { + ret = -EINVAL; + goto out; + } + + mutex_lock(&root->fs_info->fs_mutex); + sizestr = vol_args->name; + devstr = strchr(sizestr, ':'); + if (devstr) { + char *end; + sizestr = devstr + 1; + *devstr = '\0'; + devstr = vol_args->name; + devid = simple_strtoull(devstr, &end, 10); + printk(KERN_INFO "resizing devid %llu\n", devid); + } + device = btrfs_find_device(root, devid, NULL); + if (!device) { + printk(KERN_INFO "resizer unable to find device %llu\n", devid); + ret = -EINVAL; + goto out_unlock; + } + if (!strcmp(sizestr, "max")) + new_size = device->bdev->bd_inode->i_size; + else { + if (sizestr[0] == '-') { + mod = -1; + sizestr++; + } else if (sizestr[0] == '+') { + mod = 1; + sizestr++; + } + new_size = btrfs_parse_size(sizestr); + if (new_size == 0) { + ret = -EINVAL; + goto out_unlock; + } + } + + old_size = device->total_bytes; + + if (mod < 0) { + if (new_size > old_size) { + ret = -EINVAL; + goto out_unlock; + } + new_size = old_size - new_size; + } else if (mod > 0) { + new_size = old_size + new_size; + } + + if (new_size < 256 * 1024 * 1024) { + ret = -EINVAL; + goto out_unlock; + } + if (new_size > device->bdev->bd_inode->i_size) { + ret = -EFBIG; + goto out_unlock; + } + + do_div(new_size, root->sectorsize); + new_size *= root->sectorsize; + + printk(KERN_INFO "new size for %s is %llu\n", + device->name, (unsigned long long)new_size); + + if (new_size > old_size) { + trans = btrfs_start_transaction(root, 1); + ret = btrfs_grow_device(trans, device, new_size); + btrfs_commit_transaction(trans, root); + } else { + ret = btrfs_shrink_device(device, new_size); + } + +out_unlock: + mutex_unlock(&root->fs_info->fs_mutex); +out: + kfree(vol_args); + return ret; +} + +static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, + void __user *arg) +{ + struct btrfs_ioctl_vol_args *vol_args; + struct btrfs_dir_item *di; + struct btrfs_path *path; + u64 root_dirid; + int namelen; + int ret; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + + namelen = strlen(vol_args->name); + if (namelen > BTRFS_VOL_NAME_MAX) { + ret = -EINVAL; + goto out; + } + if (strchr(vol_args->name, '/')) { + ret = -EINVAL; + goto out; + } + + path = btrfs_alloc_path(); + if (!path) { + ret = -ENOMEM; + goto out; + } + + root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, + mutex_lock(&root->fs_info->fs_mutex); + di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, + path, root_dirid, + vol_args->name, namelen, 0); + mutex_unlock(&root->fs_info->fs_mutex); + btrfs_free_path(path); + + if (di && !IS_ERR(di)) { + ret = -EEXIST; + goto out; + } + + if (IS_ERR(di)) { + ret = PTR_ERR(di); + goto out; + } + + if (root == root->fs_info->tree_root) + ret = create_subvol(root, vol_args->name, namelen); + else + ret = create_snapshot(root, vol_args->name, namelen); +out: + kfree(vol_args); + return ret; +} + +static int btrfs_ioctl_defrag(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + + switch (inode->i_mode & S_IFMT) { + case S_IFDIR: + mutex_lock(&root->fs_info->fs_mutex); + btrfs_defrag_root(root, 0); + btrfs_defrag_root(root->fs_info->extent_root, 0); + mutex_unlock(&root->fs_info->fs_mutex); + break; + case S_IFREG: + btrfs_defrag_file(file); + break; + } + + return 0; +} + +long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) +{ + struct btrfs_ioctl_vol_args *vol_args; + int ret; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + ret = btrfs_init_new_device(root, vol_args->name); + +out: + kfree(vol_args); + return ret; +} + +long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) +{ + struct btrfs_ioctl_vol_args *vol_args; + int ret; + + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); + + if (!vol_args) + return -ENOMEM; + + if (copy_from_user(vol_args, arg, sizeof(*vol_args))) { + ret = -EFAULT; + goto out; + } + ret = btrfs_rm_device(root, vol_args->name); + +out: + kfree(vol_args); + return ret; +} + +int dup_item_to_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *leaf, + int slot, + struct btrfs_key *key, + u64 destino) +{ + char *dup; + int len = btrfs_item_size_nr(leaf, slot); + struct btrfs_key ckey = *key; + int ret = 0; + + dup = kmalloc(len, GFP_NOFS); + if (!dup) + return -ENOMEM; + + read_extent_buffer(leaf, dup, btrfs_item_ptr_offset(leaf, slot), len); + btrfs_release_path(root, path); + + ckey.objectid = destino; + ret = btrfs_insert_item(trans, root, &ckey, dup, len); + kfree(dup); + return ret; +} + +long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct file *src_file; + struct inode *src; + struct btrfs_trans_handle *trans; + int ret; + u64 pos; + struct btrfs_path *path; + struct btrfs_key key; + struct extent_buffer *leaf; + u32 nritems; + int slot; + + src_file = fget(src_fd); + if (!src_file) + return -EBADF; + src = src_file->f_dentry->d_inode; + + ret = -EXDEV; + if (src->i_sb != inode->i_sb) + goto out_fput; + + if (inode < src) { + mutex_lock(&inode->i_mutex); + mutex_lock(&src->i_mutex); + } else { + mutex_lock(&src->i_mutex); + mutex_lock(&inode->i_mutex); + } + + ret = -ENOTEMPTY; + if (inode->i_size) + goto out_unlock; + + /* do any pending delalloc/csum calc on src, one way or + another, and lock file content */ + while (1) { + filemap_write_and_wait(src->i_mapping); + lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + if (BTRFS_I(src)->delalloc_bytes == 0) + break; + unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + } + + mutex_lock(&root->fs_info->fs_mutex); + trans = btrfs_start_transaction(root, 0); + path = btrfs_alloc_path(); + if (!path) { + ret = -ENOMEM; + goto out; + } + key.offset = 0; + key.type = BTRFS_EXTENT_DATA_KEY; + key.objectid = src->i_ino; + pos = 0; + path->reada = 2; + + while (1) { + /* + * note the key will change type as we walk through the + * tree. + */ + ret = btrfs_search_slot(trans, root, &key, path, 0, 0); + if (ret < 0) + goto out; + + if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) + goto out; + if (ret > 0) + break; + } + leaf = path->nodes[0]; + slot = path->slots[0]; + btrfs_item_key_to_cpu(leaf, &key, slot); + nritems = btrfs_header_nritems(leaf); + + if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || + key.objectid != src->i_ino) + break; + + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { + struct btrfs_file_extent_item *extent; + int found_type; + pos = key.offset; + extent = btrfs_item_ptr(leaf, slot, + struct btrfs_file_extent_item); + found_type = btrfs_file_extent_type(leaf, extent); + if (found_type == BTRFS_FILE_EXTENT_REG) { + u64 len = btrfs_file_extent_num_bytes(leaf, + extent); + u64 ds = btrfs_file_extent_disk_bytenr(leaf, + extent); + u64 dl = btrfs_file_extent_disk_num_bytes(leaf, + extent); + u64 off = btrfs_file_extent_offset(leaf, + extent); + btrfs_insert_file_extent(trans, root, + inode->i_ino, pos, + ds, dl, len, off); + /* ds == 0 means there's a hole */ + if (ds != 0) { + btrfs_inc_extent_ref(trans, root, + ds, dl, + root->root_key.objectid, + trans->transid, + inode->i_ino, pos); + } + pos = key.offset + len; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + ret = dup_item_to_inode(trans, root, path, + leaf, slot, &key, + inode->i_ino); + if (ret) + goto out; + pos = key.offset + btrfs_item_size_nr(leaf, + slot); + } + } else if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { + ret = dup_item_to_inode(trans, root, path, leaf, + slot, &key, inode->i_ino); + + if (ret) + goto out; + } + key.offset++; + btrfs_release_path(root, path); + } + + ret = 0; +out: + btrfs_free_path(path); + + inode->i_blocks = src->i_blocks; + i_size_write(inode, src->i_size); + btrfs_update_inode(trans, root, inode); + + unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); + + btrfs_end_transaction(trans, root); + mutex_unlock(&root->fs_info->fs_mutex); + +out_unlock: + mutex_unlock(&src->i_mutex); + mutex_unlock(&inode->i_mutex); +out_fput: + fput(src_file); + return ret; +} + +/* + * there are many ways the trans_start and trans_end ioctls can lead + * to deadlocks. They should only be used by applications that + * basically own the machine, and have a very in depth understanding + * of all the possible deadlocks and enospc problems. + */ +long btrfs_ioctl_trans_start(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + mutex_lock(&root->fs_info->fs_mutex); + if (file->private_data) { + ret = -EINPROGRESS; + goto out; + } + trans = btrfs_start_transaction(root, 0); + if (trans) + file->private_data = trans; + else + ret = -ENOMEM; + /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ +out: + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +/* + * there are many ways the trans_start and trans_end ioctls can lead + * to deadlocks. They should only be used by applications that + * basically own the machine, and have a very in depth understanding + * of all the possible deadlocks and enospc problems. + */ +long btrfs_ioctl_trans_end(struct file *file) +{ + struct inode *inode = fdentry(file)->d_inode; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + int ret = 0; + + mutex_lock(&root->fs_info->fs_mutex); + trans = file->private_data; + if (!trans) { + ret = -EINVAL; + goto out; + } + btrfs_end_transaction(trans, root); + file->private_data = 0; +out: + mutex_unlock(&root->fs_info->fs_mutex); + return ret; +} + +long btrfs_ioctl(struct file *file, unsigned int + cmd, unsigned long arg) +{ + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; + + switch (cmd) { + case BTRFS_IOC_SNAP_CREATE: + return btrfs_ioctl_snap_create(root, (void __user *)arg); + case BTRFS_IOC_DEFRAG: + return btrfs_ioctl_defrag(file); + case BTRFS_IOC_RESIZE: + return btrfs_ioctl_resize(root, (void __user *)arg); + case BTRFS_IOC_ADD_DEV: + return btrfs_ioctl_add_dev(root, (void __user *)arg); + case BTRFS_IOC_RM_DEV: + return btrfs_ioctl_rm_dev(root, (void __user *)arg); + case BTRFS_IOC_BALANCE: + return btrfs_balance(root->fs_info->dev_root); + case BTRFS_IOC_CLONE: + return btrfs_ioctl_clone(file, arg); + case BTRFS_IOC_TRANS_START: + return btrfs_ioctl_trans_start(file); + case BTRFS_IOC_TRANS_END: + return btrfs_ioctl_trans_end(file); + case BTRFS_IOC_SYNC: + btrfs_sync_fs(file->f_dentry->d_sb, 1); + return 0; + } + + return -ENOTTY; +} -- cgit v1.2.3-70-g09d2 From 925baeddc5b0764a53f2214a1253251bab0e0324 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:30 -0400 Subject: Btrfs: Start btree concurrency work. The allocation trees and the chunk trees are serialized via their own dedicated mutexes. This means allocation location is still not very fine grained. The main FS btree is protected by locks on each block in the btree. Locks are taken top / down, and as processing finishes on a given level of the tree, the lock is released after locking the lower level. The end result of a search is now a path where only the lowest level is locked. Releasing or freeing the path drops any locks held. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.c | 386 ++++++++++++++++++++++++++++++++++++------------- fs/btrfs/ctree.h | 23 ++- fs/btrfs/disk-io.c | 13 +- fs/btrfs/extent-tree.c | 252 +++++++++++++++++++++----------- fs/btrfs/extent_io.c | 8 - fs/btrfs/inode.c | 8 +- fs/btrfs/ioctl.c | 8 +- fs/btrfs/locking.c | 50 +++++++ fs/btrfs/locking.h | 26 ++++ fs/btrfs/transaction.c | 15 +- fs/btrfs/tree-defrag.c | 2 + 12 files changed, 579 insertions(+), 214 deletions(-) create mode 100644 fs/btrfs/locking.c create mode 100644 fs/btrfs/locking.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 7ed6b39e42d..0e2dcc718de 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ - extent_io.o volumes.o async-thread.o ioctl.o + extent_io.o volumes.o async-thread.o ioctl.o locking.o btrfs-$(CONFIG_FS_POSIX_ACL) += acl.o else diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 1d404bde3cf..75625c68fdd 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -21,6 +21,7 @@ #include "disk-io.h" #include "transaction.h" #include "print-tree.h" +#include "locking.h" static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level); @@ -64,12 +65,47 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) int i; for (i = 0; i < BTRFS_MAX_LEVEL; i++) { if (!p->nodes[i]) - break; + continue; + if (p->locks[i]) { + btrfs_tree_unlock(p->nodes[i]); + p->locks[i] = 0; + } free_extent_buffer(p->nodes[i]); } memset(p, 0, sizeof(*p)); } +struct extent_buffer *btrfs_root_node(struct btrfs_root *root) +{ + struct extent_buffer *eb; + spin_lock(&root->node_lock); + eb = root->node; + extent_buffer_get(eb); + spin_unlock(&root->node_lock); + return eb; +} + +struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root) +{ + struct extent_buffer *eb; + + while(1) { + eb = btrfs_root_node(root); + btrfs_tree_lock(eb); + + spin_lock(&root->node_lock); + if (eb == root->node) { + spin_unlock(&root->node_lock); + break; + } + spin_unlock(&root->node_lock); + + btrfs_tree_unlock(eb); + free_extent_buffer(eb); + } + return eb; +} + static void add_root_to_dirty_list(struct btrfs_root *root) { if (root->track_dirty && list_empty(&root->dirty_list)) { @@ -111,7 +147,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, } else { first_key.objectid = 0; } - cow = __btrfs_alloc_free_block(trans, new_root, buf->len, + cow = btrfs_alloc_free_block(trans, new_root, buf->len, new_root_objectid, trans->transid, first_key.objectid, level, buf->start, 0); @@ -151,8 +187,14 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, int ret = 0; int different_trans = 0; int level; + int unlock_orig = 0; struct btrfs_key first_key; + if (*cow_ret == buf) + unlock_orig = 1; + + WARN_ON(!btrfs_tree_locked(buf)); + if (root->ref_cows) { root_gen = trans->transid; } else { @@ -172,7 +214,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, } else { first_key.objectid = 0; } - cow = __btrfs_alloc_free_block(trans, root, buf->len, + cow = btrfs_alloc_free_block(trans, root, buf->len, root->root_key.objectid, root_gen, first_key.objectid, level, search_start, empty_size); @@ -196,9 +238,14 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, } if (buf == root->node) { + WARN_ON(parent && parent != buf); root_gen = btrfs_header_generation(buf); + + spin_lock(&root->node_lock); root->node = cow; extent_buffer_get(cow); + spin_unlock(&root->node_lock); + if (buf != root->commit_root) { btrfs_free_extent(trans, root, buf->start, buf->len, root->root_key.objectid, @@ -219,6 +266,8 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, btrfs_header_owner(parent), root_gen, 0, 0, 1); } + if (unlock_orig) + btrfs_tree_unlock(buf); free_extent_buffer(buf); btrfs_mark_buffer_dirty(cow); *cow_ret = cow; @@ -316,6 +365,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, int progress_passed = 0; struct btrfs_disk_key disk_key; + /* FIXME this code needs locking */ + return 0; + parent_level = btrfs_header_level(parent); if (cache_only && parent_level != 1) return 0; @@ -729,6 +781,7 @@ static int balance_level(struct btrfs_trans_handle *trans, return 0; mid = path->nodes[level]; + WARN_ON(!path->locks[level]); WARN_ON(btrfs_header_generation(mid) != trans->transid); orig_ptr = btrfs_node_blockptr(mid, orig_slot); @@ -749,14 +802,21 @@ static int balance_level(struct btrfs_trans_handle *trans, /* promote the child to a root */ child = read_node_slot(root, mid, 0); + btrfs_tree_lock(child); BUG_ON(!child); ret = btrfs_cow_block(trans, root, child, mid, 0, &child); BUG_ON(ret); + spin_lock(&root->node_lock); root->node = child; + spin_unlock(&root->node_lock); + add_root_to_dirty_list(root); + btrfs_tree_unlock(child); + path->locks[level] = 0; path->nodes[level] = NULL; clean_tree_block(trans, root, mid); + btrfs_tree_unlock(mid); /* once for the path */ free_extent_buffer(mid); ret = btrfs_free_extent(trans, root, mid->start, mid->len, @@ -775,6 +835,7 @@ static int balance_level(struct btrfs_trans_handle *trans, left = read_node_slot(root, parent, pslot - 1); if (left) { + btrfs_tree_lock(left); wret = btrfs_cow_block(trans, root, left, parent, pslot - 1, &left); if (wret) { @@ -784,6 +845,7 @@ static int balance_level(struct btrfs_trans_handle *trans, } right = read_node_slot(root, parent, pslot + 1); if (right) { + btrfs_tree_lock(right); wret = btrfs_cow_block(trans, root, right, parent, pslot + 1, &right); if (wret) { @@ -815,6 +877,7 @@ static int balance_level(struct btrfs_trans_handle *trans, u32 blocksize = right->len; clean_tree_block(trans, root, right); + btrfs_tree_unlock(right); free_extent_buffer(right); right = NULL; wret = del_ptr(trans, root, path, level + 1, pslot + @@ -862,7 +925,9 @@ static int balance_level(struct btrfs_trans_handle *trans, u64 root_gen = btrfs_header_generation(parent); u64 bytenr = mid->start; u32 blocksize = mid->len; + clean_tree_block(trans, root, mid); + btrfs_tree_unlock(mid); free_extent_buffer(mid); mid = NULL; wret = del_ptr(trans, root, path, level + 1, pslot); @@ -885,11 +950,14 @@ static int balance_level(struct btrfs_trans_handle *trans, if (left) { if (btrfs_header_nritems(left) > orig_slot) { extent_buffer_get(left); + /* left was locked after cow */ path->nodes[level] = left; path->slots[level + 1] -= 1; path->slots[level] = orig_slot; - if (mid) + if (mid) { + btrfs_tree_unlock(mid); free_extent_buffer(mid); + } } else { orig_slot -= btrfs_header_nritems(left); path->slots[level] = orig_slot; @@ -901,10 +969,15 @@ static int balance_level(struct btrfs_trans_handle *trans, btrfs_node_blockptr(path->nodes[level], path->slots[level])) BUG(); enospc: - if (right) + if (right) { + btrfs_tree_unlock(right); free_extent_buffer(right); - if (left) + } + if (left) { + if (path->nodes[level] != left) + btrfs_tree_unlock(left); free_extent_buffer(left); + } return ret; } @@ -942,6 +1015,8 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, /* first, try to make some room in the middle buffer */ if (left) { u32 left_nr; + + btrfs_tree_lock(left); left_nr = btrfs_header_nritems(left); if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; @@ -967,24 +1042,28 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, path->nodes[level] = left; path->slots[level + 1] -= 1; path->slots[level] = orig_slot; + btrfs_tree_unlock(mid); free_extent_buffer(mid); } else { orig_slot -= btrfs_header_nritems(left); path->slots[level] = orig_slot; + btrfs_tree_unlock(left); free_extent_buffer(left); } return 0; } + btrfs_tree_unlock(left); free_extent_buffer(left); } - right= read_node_slot(root, parent, pslot + 1); + right = read_node_slot(root, parent, pslot + 1); /* * then try to empty the right most buffer into the middle */ if (right) { u32 right_nr; + btrfs_tree_lock(right); right_nr = btrfs_header_nritems(right); if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) { wret = 1; @@ -1013,12 +1092,15 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, path->slots[level + 1] += 1; path->slots[level] = orig_slot - btrfs_header_nritems(mid); + btrfs_tree_unlock(mid); free_extent_buffer(mid); } else { + btrfs_tree_unlock(right); free_extent_buffer(right); } return 0; } + btrfs_tree_unlock(right); free_extent_buffer(right); } return 1; @@ -1050,6 +1132,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, return; node = path->nodes[level]; + WARN_ON(!path->skip_locking && !btrfs_tree_locked(node)); + search = btrfs_node_blockptr(node, slot); blocksize = btrfs_level_size(root, level - 1); eb = btrfs_find_tree_block(root, search, blocksize); @@ -1098,6 +1182,39 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, highest_read = search; } } + +static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock) +{ + int i; + int skip_level = level; + struct extent_buffer *t; + + for (i = level; i < BTRFS_MAX_LEVEL; i++) { + if (!path->nodes[i]) + break; + if (!path->locks[i]) + break; + if (path->slots[i] == 0) { + skip_level = i + 1; + continue; + } + if (path->keep_locks) { + u32 nritems; + t = path->nodes[i]; + nritems = btrfs_header_nritems(t); + if (path->slots[i] >= nritems - 1) { + skip_level = i + 1; + continue; + } + } + t = path->nodes[i]; + if (i >= lowest_unlock && i > skip_level && path->locks[i]) { + btrfs_tree_unlock(t); + path->locks[i] = 0; + } + } +} + /* * look for key in the tree. path is filled in with nodes along the way * if key is found, we return zero and you can find the item in the leaf @@ -1120,15 +1237,27 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root int ret; int level; int should_reada = p->reada; + int lowest_unlock = 1; u8 lowest_level = 0; lowest_level = p->lowest_level; WARN_ON(lowest_level && ins_len); WARN_ON(p->nodes[0] != NULL); - WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex)); + // WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex)); + WARN_ON(root == root->fs_info->extent_root && + !mutex_is_locked(&root->fs_info->alloc_mutex)); + WARN_ON(root == root->fs_info->chunk_root && + !mutex_is_locked(&root->fs_info->chunk_mutex)); + WARN_ON(root == root->fs_info->dev_root && + !mutex_is_locked(&root->fs_info->chunk_mutex)); + if (ins_len < 0) + lowest_unlock = 2; again: - b = root->node; - extent_buffer_get(b); + if (!p->skip_locking) + b = btrfs_lock_root_node(root); + else + b = btrfs_root_node(root); + while (b) { level = btrfs_header_level(b); if (cow) { @@ -1147,9 +1276,12 @@ again: WARN_ON(1); level = btrfs_header_level(b); p->nodes[level] = b; + if (!p->skip_locking) + p->locks[level] = 1; ret = check_block(root, p, level); if (ret) return -1; + ret = bin_search(b, key, level, &slot); if (level != 0) { if (ret && slot > 0) @@ -1177,14 +1309,19 @@ again: BUG_ON(btrfs_header_nritems(b) == 1); } /* this is only true while dropping a snapshot */ - if (level == lowest_level) + if (level == lowest_level) { + unlock_up(p, level, lowest_unlock); break; + } if (should_reada) reada_for_search(root, p, level, slot, key->objectid); b = read_node_slot(root, b, slot); + if (!p->skip_locking) + btrfs_tree_lock(b); + unlock_up(p, level, lowest_unlock); } else { p->slots[level] = slot; if (ins_len > 0 && btrfs_leaf_free_space(root, b) < @@ -1195,6 +1332,7 @@ again: if (sret) return sret; } + unlock_up(p, level, lowest_unlock); return ret; } } @@ -1225,6 +1363,13 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, break; t = path->nodes[i]; btrfs_set_node_key(t, key, tslot); + if (!btrfs_tree_locked(path->nodes[i])) { + int ii; +printk("fixup without lock on level %d\n", btrfs_header_level(path->nodes[i])); + for (ii = 0; ii < BTRFS_MAX_LEVEL; ii++) { +printk("level %d slot %d\n", ii, path->slots[ii]); + } + } btrfs_mark_buffer_dirty(path->nodes[i]); if (tslot != 0) break; @@ -1370,6 +1515,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, u64 lower_gen; struct extent_buffer *lower; struct extent_buffer *c; + struct extent_buffer *old; struct btrfs_disk_key lower_key; BUG_ON(path->nodes[level]); @@ -1386,12 +1532,13 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, else btrfs_node_key(lower, &lower_key, 0); - c = __btrfs_alloc_free_block(trans, root, root->nodesize, + c = btrfs_alloc_free_block(trans, root, root->nodesize, root->root_key.objectid, root_gen, lower_key.objectid, level, root->node->start, 0); if (IS_ERR(c)) return PTR_ERR(c); + memset_extent_buffer(c, 0, 0, root->nodesize); btrfs_set_header_nritems(c, 1); btrfs_set_header_level(c, level); @@ -1416,23 +1563,31 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, btrfs_mark_buffer_dirty(c); - /* the super has an extra ref to root->node */ - free_extent_buffer(root->node); + spin_lock(&root->node_lock); + old = root->node; root->node = c; + spin_unlock(&root->node_lock); + + /* the super has an extra ref to root->node */ + free_extent_buffer(old); + add_root_to_dirty_list(root); extent_buffer_get(c); path->nodes[level] = c; + path->locks[level] = 1; path->slots[level] = 0; if (root->ref_cows && lower_gen != trans->transid) { struct btrfs_path *back_path = btrfs_alloc_path(); int ret; + mutex_lock(&root->fs_info->alloc_mutex); ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, path, lower->start, root->root_key.objectid, trans->transid, 0, 0); BUG_ON(ret); + mutex_unlock(&root->fs_info->alloc_mutex); btrfs_free_path(back_path); } return 0; @@ -1521,7 +1676,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root root_gen = 0; btrfs_node_key(c, &disk_key, 0); - split = __btrfs_alloc_free_block(trans, root, root->nodesize, + split = btrfs_alloc_free_block(trans, root, root->nodesize, root->root_key.objectid, root_gen, btrfs_disk_key_objectid(&disk_key), @@ -1564,10 +1719,12 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root if (path->slots[level] >= mid) { path->slots[level] -= mid; + btrfs_tree_unlock(c); free_extent_buffer(c); path->nodes[level] = split; path->slots[level + 1] += 1; } else { + btrfs_tree_unlock(split); free_extent_buffer(split); } return ret; @@ -1648,30 +1805,24 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root return 1; right = read_node_slot(root, upper, slot + 1); + btrfs_tree_lock(right); free_space = btrfs_leaf_free_space(root, right); - if (free_space < data_size + sizeof(struct btrfs_item)) { - free_extent_buffer(right); - return 1; - } + if (free_space < data_size + sizeof(struct btrfs_item)) + goto out_unlock; /* cow and double check */ ret = btrfs_cow_block(trans, root, right, upper, slot + 1, &right); - if (ret) { - free_extent_buffer(right); - return 1; - } + if (ret) + goto out_unlock; + free_space = btrfs_leaf_free_space(root, right); - if (free_space < data_size + sizeof(struct btrfs_item)) { - free_extent_buffer(right); - return 1; - } + if (free_space < data_size + sizeof(struct btrfs_item)) + goto out_unlock; left_nritems = btrfs_header_nritems(left); - if (left_nritems == 0) { - free_extent_buffer(right); - return 1; - } + if (left_nritems == 0) + goto out_unlock; if (empty) nr = 0; @@ -1707,10 +1858,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root left->map_token = NULL; } - if (push_items == 0) { - free_extent_buffer(right); - return 1; - } + if (push_items == 0) + goto out_unlock; if (!empty && push_items == left_nritems) WARN_ON(1); @@ -1778,14 +1927,24 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root /* then fixup the leaf pointer in the path */ if (path->slots[0] >= left_nritems) { path->slots[0] -= left_nritems; + if (btrfs_header_nritems(path->nodes[0]) == 0) + clean_tree_block(trans, root, path->nodes[0]); + btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); path->nodes[0] = right; path->slots[1] += 1; } else { + btrfs_tree_unlock(right); free_extent_buffer(right); } return 0; + +out_unlock: + btrfs_tree_unlock(right); + free_extent_buffer(right); + return 1; } + /* * push some data in the path leaf to the left, trying to free up at * least data_size bytes. returns zero if the push worked, nonzero otherwise @@ -1823,10 +1982,11 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root } left = read_node_slot(root, path->nodes[1], slot - 1); + btrfs_tree_lock(left); free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { - free_extent_buffer(left); - return 1; + ret = 1; + goto out; } /* cow and double check */ @@ -1834,14 +1994,14 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root path->nodes[1], slot - 1, &left); if (ret) { /* we hit -ENOSPC, but it isn't fatal here */ - free_extent_buffer(left); - return 1; + ret = 1; + goto out; } free_space = btrfs_leaf_free_space(root, left); if (free_space < data_size + sizeof(struct btrfs_item)) { - free_extent_buffer(left); - return 1; + ret = 1; + goto out; } if (empty) @@ -1876,8 +2036,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root } if (push_items == 0) { - free_extent_buffer(left); - return 1; + ret = 1; + goto out; } if (!empty && push_items == btrfs_header_nritems(right)) WARN_ON(1); @@ -1975,15 +2135,23 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root /* then fixup the leaf pointer in the path */ if (path->slots[0] < push_items) { path->slots[0] += old_left_nritems; + if (btrfs_header_nritems(path->nodes[0]) == 0) + clean_tree_block(trans, root, path->nodes[0]); + btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); path->nodes[0] = left; path->slots[1] -= 1; } else { + btrfs_tree_unlock(left); free_extent_buffer(left); path->slots[0] -= push_items; } BUG_ON(path->slots[0] < 0); return ret; +out: + btrfs_tree_unlock(left); + free_extent_buffer(left); + return ret; } /* @@ -2052,7 +2220,7 @@ again: btrfs_item_key(l, &disk_key, 0); - right = __btrfs_alloc_free_block(trans, root, root->leafsize, + right = btrfs_alloc_free_block(trans, root, root->leafsize, root->root_key.objectid, root_gen, disk_key.objectid, 0, l->start, 0); @@ -2085,6 +2253,8 @@ again: path->slots[1] + 1, 1); if (wret) ret = wret; + + btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); path->nodes[0] = right; path->slots[0] = 0; @@ -2111,6 +2281,7 @@ again: path->slots[1], 1); if (wret) ret = wret; + btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); path->nodes[0] = right; path->slots[0] = 0; @@ -2184,12 +2355,15 @@ again: BUG_ON(path->slots[0] != slot); if (mid <= slot) { + btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); path->nodes[0] = right; path->slots[0] -= mid; path->slots[1] += 1; - } else + } else { + btrfs_tree_unlock(right); free_extent_buffer(right); + } BUG_ON(path->slots[0] < 0); @@ -2418,10 +2592,6 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, total_data += data_size[i]; } - /* create a root if there isn't one */ - if (!root->node) - BUG(); - total_size = total_data + (nr - 1) * sizeof(struct btrfs_item); ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); if (ret == 0) { @@ -2516,7 +2686,6 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, btrfs_print_leaf(root, leaf); BUG(); } - out: return ret; } @@ -2655,7 +2824,6 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, btrfs_set_header_level(leaf, 0); } else { u64 root_gen = btrfs_header_generation(path->nodes[1]); - clean_tree_block(trans, root, leaf); wret = del_ptr(trans, root, path, 1, path->slots[1]); if (wret) ret = wret; @@ -2706,8 +2874,6 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, root_gen = btrfs_header_generation( path->nodes[1]); - clean_tree_block(trans, root, leaf); - wret = del_ptr(trans, root, path, 1, slot); if (wret) ret = wret; @@ -2720,7 +2886,13 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (wret) ret = wret; } else { - btrfs_mark_buffer_dirty(leaf); + /* if we're still in the path, make sure + * we're dirty. Otherwise, one of the + * push_leaf functions must have already + * dirtied this buffer + */ + if (path->nodes[0] == leaf) + btrfs_mark_buffer_dirty(leaf); free_extent_buffer(leaf); } } else { @@ -2731,56 +2903,40 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, } /* - * walk up the tree as far as required to find the previous leaf. + * search the tree again to find a leaf with lesser keys * returns 0 if it found something or 1 if there are no lesser leaves. * returns < 0 on io errors. */ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) { - int slot; - int level = 1; - struct extent_buffer *c; - struct extent_buffer *next = NULL; + struct btrfs_key key; + struct btrfs_disk_key found_key; + int ret; - while(level < BTRFS_MAX_LEVEL) { - if (!path->nodes[level]) - return 1; + btrfs_item_key_to_cpu(path->nodes[0], &key, 0); - slot = path->slots[level]; - c = path->nodes[level]; - if (slot == 0) { - level++; - if (level == BTRFS_MAX_LEVEL) - return 1; - continue; - } - slot--; - - if (next) - free_extent_buffer(next); + if (key.offset > 0) + key.offset--; + else if (key.type > 0) + key.type--; + else if (key.objectid > 0) + key.objectid--; + else + return 1; - next = read_node_slot(root, c, slot); - break; - } - path->slots[level] = slot; - while(1) { - level--; - c = path->nodes[level]; - free_extent_buffer(c); - slot = btrfs_header_nritems(next); - if (slot != 0) - slot--; - path->nodes[level] = next; - path->slots[level] = slot; - if (!level) - break; - next = read_node_slot(root, next, slot); - } - return 0; + btrfs_release_path(root, path); + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + return ret; + btrfs_item_key(path->nodes[0], &found_key, 0); + ret = comp_keys(&found_key, &key); + if (ret < 0) + return 0; + return 1; } /* - * walk up the tree as far as required to find the next leaf. + * search the tree again to find a leaf with greater keys * returns 0 if it found something or 1 if there are no greater leaves. * returns < 0 on io errors. */ @@ -2790,6 +2946,28 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) int level = 1; struct extent_buffer *c; struct extent_buffer *next = NULL; + struct btrfs_key key; + u32 nritems; + int ret; + + nritems = btrfs_header_nritems(path->nodes[0]); + if (nritems == 0) { + return 1; + } + + btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1); + + path->keep_locks = 1; + btrfs_release_path(root, path); + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + path->keep_locks = 0; + + if (ret < 0) + return ret; + + if (path->slots[0] < nritems - 1) { + goto done; + } while(level < BTRFS_MAX_LEVEL) { if (!path->nodes[level]) @@ -2799,33 +2977,45 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) c = path->nodes[level]; if (slot >= btrfs_header_nritems(c)) { level++; - if (level == BTRFS_MAX_LEVEL) + if (level == BTRFS_MAX_LEVEL) { return 1; + } continue; } - if (next) + if (next) { + btrfs_tree_unlock(next); free_extent_buffer(next); + } - if (path->reada) + if (level == 1 && path->locks[1] && path->reada) reada_for_search(root, path, level, slot, 0); next = read_node_slot(root, c, slot); + if (!path->skip_locking) + btrfs_tree_lock(next); break; } path->slots[level] = slot; while(1) { level--; c = path->nodes[level]; + if (path->locks[level]) + btrfs_tree_unlock(c); free_extent_buffer(c); path->nodes[level] = next; path->slots[level] = 0; + path->locks[level] = 1; if (!level) break; - if (path->reada) - reada_for_search(root, path, level, 0, 0); + if (level == 1 && path->locks[1] && path->reada) + reada_for_search(root, path, level, slot, 0); next = read_node_slot(root, next, 0); + if (!path->skip_locking) + btrfs_tree_lock(next); } +done: + unlock_up(path, 0, 1); return 0; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index dcea9d706d9..50891b39f36 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -330,8 +330,13 @@ struct btrfs_node { struct btrfs_path { struct extent_buffer *nodes[BTRFS_MAX_LEVEL]; int slots[BTRFS_MAX_LEVEL]; + /* if there is real range locking, this locks field will change */ + int locks[BTRFS_MAX_LEVEL]; int reada; + /* keep some upper locks as we walk down */ + int keep_locks; int lowest_level; + int skip_locking; }; /* @@ -515,6 +520,8 @@ struct btrfs_fs_info { spinlock_t hash_lock; struct mutex trans_mutex; struct mutex fs_mutex; + struct mutex alloc_mutex; + struct mutex chunk_mutex; struct list_head trans_list; struct list_head hashers; struct list_head dead_roots; @@ -576,6 +583,10 @@ struct btrfs_fs_info { */ struct btrfs_root { struct extent_buffer *node; + + /* the node lock is held while changing the node pointer */ + spinlock_t node_lock; + struct extent_buffer *commit_root; struct btrfs_root_item root_item; struct btrfs_key root_key; @@ -1353,13 +1364,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *hint, u64 search_start, int data, int owner); -int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 owner_objectid); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u32 size, - u64 root_objectid, - u64 hint, u64 empty_size); -struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u32 blocksize, u64 root_objectid, @@ -1368,8 +1373,6 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, int level, u64 hint, u64 empty_size); -int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 new_size); int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size); int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -1409,6 +1412,10 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, int btrfs_previous_item(struct btrfs_root *root, struct btrfs_path *path, u64 min_objectid, int type); + +struct extent_buffer *btrfs_root_node(struct btrfs_root *root); +struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root); + int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e5c758e306d..fe40bdd984f 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -32,6 +32,7 @@ #include "volumes.h" #include "print-tree.h" #include "async-thread.h" +#include "locking.h" #if 0 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) @@ -681,9 +682,11 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, { struct inode *btree_inode = root->fs_info->btree_inode; if (btrfs_header_generation(buf) == - root->fs_info->running_transaction->transid) + root->fs_info->running_transaction->transid) { + WARN_ON(!btrfs_tree_locked(buf)); clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); + } return 0; } @@ -720,6 +723,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->in_sysfs = 0; INIT_LIST_HEAD(&root->dirty_list); + spin_lock_init(&root->node_lock); memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); @@ -1196,6 +1200,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, mutex_init(&fs_info->trans_mutex); mutex_init(&fs_info->fs_mutex); + mutex_init(&fs_info->alloc_mutex); + mutex_init(&fs_info->chunk_mutex); #if 0 ret = add_hasher(fs_info, "crc32c"); @@ -1274,7 +1280,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, mutex_lock(&fs_info->fs_mutex); + mutex_lock(&fs_info->chunk_mutex); ret = btrfs_read_sys_array(tree_root); + mutex_unlock(&fs_info->chunk_mutex); if (ret) { printk("btrfs: failed to read the system array on %s\n", sb->s_id); @@ -1296,7 +1304,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); + mutex_lock(&fs_info->chunk_mutex); ret = btrfs_read_chunk_tree(chunk_root); + mutex_unlock(&fs_info->chunk_mutex); BUG_ON(ret); btrfs_close_extra_devices(fs_devices); @@ -1654,6 +1664,7 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) u64 transid = btrfs_header_generation(buf); struct inode *btree_inode = root->fs_info->btree_inode; + WARN_ON(!btrfs_tree_locked(buf)); if (transid != root->fs_info->generation) { printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", (unsigned long long)buf->start, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 41a63462d3e..7e40c516fe6 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -25,6 +25,7 @@ #include "print-tree.h" #include "transaction.h" #include "volumes.h" +#include "locking.h" #define BLOCK_GROUP_DATA EXTENT_WRITEBACK #define BLOCK_GROUP_METADATA EXTENT_UPTODATE @@ -36,7 +37,28 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); static int del_pending_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); +static struct btrfs_block_group_cache * +__btrfs_find_block_group(struct btrfs_root *root, + struct btrfs_block_group_cache *hint, + u64 search_start, int data, int owner); +void maybe_lock_mutex(struct btrfs_root *root) +{ + if (root != root->fs_info->extent_root && + root != root->fs_info->chunk_root && + root != root->fs_info->dev_root) { + mutex_lock(&root->fs_info->alloc_mutex); + } +} + +void maybe_unlock_mutex(struct btrfs_root *root) +{ + if (root != root->fs_info->extent_root && + root != root->fs_info->chunk_root && + root != root->fs_info->dev_root) { + mutex_unlock(&root->fs_info->alloc_mutex); + } +} static int cache_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *block_group) @@ -66,6 +88,7 @@ static int cache_block_group(struct btrfs_root *root, return -ENOMEM; path->reada = 2; + path->skip_locking = 1; first_free = block_group->key.objectid; key.objectid = block_group->key.objectid; key.offset = 0; @@ -290,7 +313,7 @@ no_cache: cache = btrfs_lookup_first_block_group(root->fs_info, last); } cache_miss = 0; - cache = btrfs_find_block_group(root, cache, last, data, 0); + cache = __btrfs_find_block_group(root, cache, last, data, 0); if (!cache) goto no_cache; *cache_ret = cache; @@ -318,10 +341,10 @@ static int block_group_state_bits(u64 flags) return bits; } -struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, - struct btrfs_block_group_cache - *hint, u64 search_start, - int data, int owner) +static struct btrfs_block_group_cache * +__btrfs_find_block_group(struct btrfs_root *root, + struct btrfs_block_group_cache *hint, + u64 search_start, int data, int owner) { struct btrfs_block_group_cache *cache; struct extent_io_tree *block_group_cache; @@ -411,6 +434,18 @@ found: return found_group; } +struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, + struct btrfs_block_group_cache + *hint, u64 search_start, + int data, int owner) +{ + + struct btrfs_block_group_cache *ret; + mutex_lock(&root->fs_info->alloc_mutex); + ret = __btrfs_find_block_group(root, hint, search_start, data, owner); + mutex_unlock(&root->fs_info->alloc_mutex); + return ret; +} static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset) { @@ -646,7 +681,7 @@ out: return ret; } -int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, +static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 root_objectid, u64 ref_generation, @@ -696,6 +731,22 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, return 0; } +int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset) +{ + int ret; + + mutex_lock(&root->fs_info->alloc_mutex); + ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, + root_objectid, ref_generation, + owner, owner_offset); + mutex_unlock(&root->fs_info->alloc_mutex); + return ret; +} + int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root) { @@ -760,6 +811,10 @@ u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, struct btrfs_extent_ref *ref_item; int level = -1; + /* FIXME, needs locking */ + BUG(); + + mutex_lock(&root->fs_info->alloc_mutex); path = btrfs_alloc_path(); again: if (level == -1) @@ -854,33 +909,9 @@ again: out: btrfs_free_path(path); + mutex_unlock(&root->fs_info->alloc_mutex); return total_count; } -int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 owner_objectid) -{ - u64 generation; - u64 key_objectid; - u64 level; - u32 nritems; - struct btrfs_disk_key disk_key; - - level = btrfs_header_level(root->node); - generation = trans->transid; - nritems = btrfs_header_nritems(root->node); - if (nritems > 0) { - if (level == 0) - btrfs_item_key(root->node, &disk_key, 0); - else - btrfs_node_key(root->node, &disk_key, 0); - key_objectid = btrfs_disk_key_objectid(&disk_key); - } else { - key_objectid = 0; - } - return btrfs_inc_extent_ref(trans, root, root->node->start, - root->node->len, owner_objectid, - generation, level, key_objectid); -} int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf) @@ -897,6 +928,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (!root->ref_cows) return 0; + mutex_lock(&root->fs_info->alloc_mutex); level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); for (i = 0; i < nritems; i++) { @@ -913,7 +945,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); if (disk_bytenr == 0) continue; - ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, + ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(buf, fi), root->root_key.objectid, trans->transid, key.objectid, key.offset); @@ -924,7 +956,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, } else { bytenr = btrfs_node_blockptr(buf, i); btrfs_node_key_to_cpu(buf, &key, i); - ret = btrfs_inc_extent_ref(trans, root, bytenr, + ret = __btrfs_inc_extent_ref(trans, root, bytenr, btrfs_level_size(root, level - 1), root->root_key.objectid, trans->transid, @@ -935,6 +967,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, } } } + mutex_unlock(&root->fs_info->alloc_mutex); return 0; fail: WARN_ON(1); @@ -965,6 +998,7 @@ fail: } } #endif + mutex_unlock(&root->fs_info->alloc_mutex); return ret; } @@ -1019,6 +1053,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, if (!path) return -ENOMEM; + mutex_lock(&root->fs_info->alloc_mutex); while(1) { ret = find_first_extent_bit(block_group_cache, last, &start, &end, BLOCK_GROUP_DIRTY); @@ -1045,6 +1080,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, BLOCK_GROUP_DIRTY, GFP_NOFS); } btrfs_free_path(path); + mutex_unlock(&root->fs_info->alloc_mutex); return werr; } @@ -1162,26 +1198,28 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, space_info->force_alloc = 0; } if (space_info->full) - return 0; + goto out; thresh = div_factor(space_info->total_bytes, 6); if (!force && (space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < thresh) - return 0; + goto out; + mutex_lock(&extent_root->fs_info->chunk_mutex); ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); if (ret == -ENOSPC) { printk("space info full %Lu\n", flags); space_info->full = 1; - return 0; + goto out; } BUG_ON(ret); ret = btrfs_make_block_group(trans, extent_root, 0, flags, BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes); BUG_ON(ret); - + mutex_unlock(&extent_root->fs_info->chunk_mutex); +out: return 0; } @@ -1318,6 +1356,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct extent_io_tree *free_space_cache; free_space_cache = &root->fs_info->free_space_cache; + mutex_lock(&root->fs_info->alloc_mutex); while(1) { ret = find_first_extent_bit(unpin, 0, &start, &end, EXTENT_DIRTY); @@ -1327,6 +1366,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, clear_extent_dirty(unpin, start, end, GFP_NOFS); set_extent_dirty(free_space_cache, start, end, GFP_NOFS); } + mutex_unlock(&root->fs_info->alloc_mutex); return 0; } @@ -1363,18 +1403,24 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, GFP_NOFS); eb = read_tree_block(extent_root, ins.objectid, ins.offset, trans->transid); + btrfs_tree_lock(eb); level = btrfs_header_level(eb); if (level == 0) { btrfs_item_key(eb, &first, 0); } else { btrfs_node_key(eb, &first, 0); } + btrfs_tree_unlock(eb); + free_extent_buffer(eb); + /* + * the first key is just a hint, so the race we've created + * against reading it is fine + */ err = btrfs_insert_extent_backref(trans, extent_root, path, start, extent_root->root_key.objectid, 0, level, btrfs_disk_key_objectid(&first)); BUG_ON(err); - free_extent_buffer(eb); } btrfs_free_path(path); return 0; @@ -1384,12 +1430,14 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, int pending) { int err = 0; - struct extent_buffer *buf; if (!pending) { +#if 0 + struct extent_buffer *buf; buf = btrfs_find_tree_block(root, bytenr, num_bytes); if (buf) { - if (btrfs_buffer_uptodate(buf, 0)) { + if (!btrfs_try_tree_lock(buf) && + btrfs_buffer_uptodate(buf, 0)) { u64 transid = root->fs_info->running_transaction->transid; u64 header_transid = @@ -1398,12 +1446,15 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { clean_tree_block(NULL, root, buf); + btrfs_tree_unlock(buf); free_extent_buffer(buf); return 1; } + btrfs_tree_unlock(buf); } free_extent_buffer(buf); } +#endif update_pinned_extents(root, bytenr, num_bytes, 1); } else { set_extent_bits(&root->fs_info->pending_del, @@ -1586,10 +1637,11 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct /* * remove an extent from the root, returns 0 on success */ -int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, - u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, int pin) +static int __btrfs_free_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 num_bytes, u64 root_objectid, + u64 ref_generation, u64 owner_objectid, + u64 owner_offset, int pin) { struct btrfs_root *extent_root = root->fs_info->extent_root; int pending_ret; @@ -1610,6 +1662,22 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root return ret ? ret : pending_ret; } +int btrfs_free_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 num_bytes, u64 root_objectid, + u64 ref_generation, u64 owner_objectid, + u64 owner_offset, int pin) +{ + int ret; + + maybe_lock_mutex(root); + ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, + root_objectid, ref_generation, + owner_objectid, owner_offset, pin); + maybe_unlock_mutex(root); + return ret; +} + static u64 stripe_align(struct btrfs_root *root, u64 val) { u64 mask = ((u64)root->stripesize - 1); @@ -1679,12 +1747,12 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, block_group = btrfs_lookup_first_block_group(info, hint_byte); if (!block_group) hint_byte = search_start; - block_group = btrfs_find_block_group(root, block_group, + block_group = __btrfs_find_block_group(root, block_group, hint_byte, data, 1); if (last_ptr && *last_ptr == 0 && block_group) hint_byte = block_group->key.objectid; } else { - block_group = btrfs_find_block_group(root, + block_group = __btrfs_find_block_group(root, trans->block_group, search_start, data, 1); } @@ -1806,7 +1874,7 @@ enospc: } block_group = btrfs_lookup_first_block_group(info, search_start); cond_resched(); - block_group = btrfs_find_block_group(root, block_group, + block_group = __btrfs_find_block_group(root, block_group, search_start, data, 0); goto check_failed; @@ -1843,6 +1911,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; struct btrfs_key keys[2]; + maybe_lock_mutex(root); + if (data) { alloc_profile = info->avail_data_alloc_bits & info->data_alloc_profile; @@ -1892,9 +1962,10 @@ again: if (ret) { printk("allocation failed flags %Lu\n", data); } - BUG_ON(ret); - if (ret) - return ret; + if (ret) { + BUG(); + goto out; + } /* block accounting for super block */ super_used = btrfs_super_bytes_used(&info->super_copy); @@ -1953,11 +2024,11 @@ again: finish_current_insert(trans, extent_root); pending_ret = del_pending_extents(trans, extent_root); - if (ret) { - return ret; - } + if (ret) + goto out; if (pending_ret) { - return pending_ret; + ret = pending_ret; + goto out; } update_block: @@ -1967,36 +2038,15 @@ update_block: ins->objectid, ins->offset); BUG(); } - return 0; +out: + maybe_unlock_mutex(root); + return ret; } - /* * helper function to allocate a block for a given tree * returns the tree buffer or NULL. */ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u32 blocksize, - u64 root_objectid, u64 hint, - u64 empty_size) -{ - u64 ref_generation; - - if (root->ref_cows) - ref_generation = trans->transid; - else - ref_generation = 0; - - - return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, - ref_generation, 0, 0, hint, empty_size); -} - -/* - * helper function to allocate a block for a given tree - * returns the tree buffer or NULL. - */ -struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u32 blocksize, u64 root_objectid, @@ -2026,6 +2076,7 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, return ERR_PTR(-ENOMEM); } btrfs_set_header_generation(buf, trans->transid); + btrfs_tree_lock(buf); clean_tree_block(trans, root, buf); btrfs_set_buffer_uptodate(buf); @@ -2076,7 +2127,7 @@ static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); if (disk_bytenr == 0) continue; - ret = btrfs_free_extent(trans, root, disk_bytenr, + ret = __btrfs_free_extent(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(leaf, fi), leaf_owner, leaf_generation, key.objectid, key.offset, 0); @@ -2151,6 +2202,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, int ret; u32 refs; + mutex_lock(&root->fs_info->alloc_mutex); + WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); ret = lookup_extent_ref(trans, root, @@ -2182,6 +2235,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, bytenr = btrfs_node_blockptr(cur, path->slots[*level]); ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]); blocksize = btrfs_level_size(root, *level - 1); + ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); BUG_ON(ret); if (refs != 1) { @@ -2189,7 +2243,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, root_owner = btrfs_header_owner(parent); root_gen = btrfs_header_generation(parent); path->slots[*level]++; - ret = btrfs_free_extent(trans, root, bytenr, + ret = __btrfs_free_extent(trans, root, bytenr, blocksize, root_owner, root_gen, 0, 0, 1); BUG_ON(ret); @@ -2201,9 +2255,11 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, reada_walk_down(root, cur, path->slots[*level]); mutex_unlock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->alloc_mutex); next = read_tree_block(root, bytenr, blocksize, ptr_gen); mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->alloc_mutex); /* we've dropped the lock, double check */ ret = lookup_extent_ref(trans, root, bytenr, @@ -2216,7 +2272,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, path->slots[*level]++; free_extent_buffer(next); - ret = btrfs_free_extent(trans, root, bytenr, + ret = __btrfs_free_extent(trans, root, bytenr, blocksize, root_owner, root_gen, 0, 0, 1); @@ -2244,13 +2300,14 @@ out: } root_gen = btrfs_header_generation(parent); - ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, + ret = __btrfs_free_extent(trans, root, path->nodes[*level]->start, path->nodes[*level]->len, root_owner, root_gen, 0, 0, 1); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; BUG_ON(ret); + mutex_unlock(&root->fs_info->alloc_mutex); return 0; } @@ -2350,6 +2407,12 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_node_key(node, &found_key, path->slots[level]); WARN_ON(memcmp(&found_key, &root_item->drop_progress, sizeof(found_key))); + for (i = 0; i < BTRFS_MAX_LEVEL; i++) { + if (path->nodes[i] && path->locks[i]) { + path->locks[i] = 0; + btrfs_tree_unlock(path->nodes[i]); + } + } } while(1) { wret = walk_down_tree(trans, root, path, &level); @@ -2383,6 +2446,8 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) u64 end; u64 ptr; int ret; + + mutex_lock(&info->alloc_mutex); while(1) { ret = find_first_extent_bit(&info->block_group_cache, 0, &start, &end, (unsigned int)-1); @@ -2402,6 +2467,7 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) clear_extent_dirty(&info->free_space_cache, start, end, GFP_NOFS); } + mutex_unlock(&info->alloc_mutex); return 0; } @@ -2678,6 +2744,7 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, eb = read_tree_block(found_root, extent_key->objectid, extent_key->offset, 0); + btrfs_tree_lock(eb); level = btrfs_header_level(eb); if (level == 0) @@ -2685,6 +2752,7 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, else btrfs_node_key_to_cpu(eb, &found_key, 0); + btrfs_tree_unlock(eb); free_extent_buffer(eb); ret = find_root_for_ref(extent_root, path, &found_key, @@ -2888,6 +2956,7 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) int ret; int progress; + mutex_lock(&root->fs_info->alloc_mutex); shrink_block_group = btrfs_lookup_block_group(root->fs_info, shrink_start); BUG_ON(!shrink_block_group); @@ -3044,20 +3113,22 @@ next: (unsigned int)-1, GFP_NOFS); out: btrfs_free_path(path); + mutex_unlock(&root->fs_info->alloc_mutex); return ret; } int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *key) { - int ret; + int ret = 0; struct btrfs_key found_key; struct extent_buffer *leaf; int slot; ret = btrfs_search_slot(NULL, root, key, path, 0, 0); if (ret < 0) - return ret; + goto out; + while(1) { slot = path->slots[0]; leaf = path->nodes[0]; @@ -3066,18 +3137,20 @@ int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, if (ret == 0) continue; if (ret < 0) - goto error; + goto out; break; } btrfs_item_key_to_cpu(leaf, &found_key, slot); if (found_key.objectid >= key->objectid && - found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) - return 0; + found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) { + ret = 0; + goto out; + } path->slots[0]++; } ret = -ENOENT; -error: +out: return ret; } @@ -3103,6 +3176,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) if (!path) return -ENOMEM; + mutex_lock(&root->fs_info->alloc_mutex); while(1) { ret = find_first_block_group(root, path, &key); if (ret > 0) { @@ -3158,6 +3232,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) ret = 0; error: btrfs_free_path(path); + mutex_unlock(&root->fs_info->alloc_mutex); return ret; } @@ -3205,5 +3280,6 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, ret = del_pending_extents(trans, extent_root); BUG_ON(ret); set_avail_alloc_bits(extent_root->fs_info, type); + return 0; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 17c508a941e..bd15cdcaba9 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2889,7 +2889,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); - lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); else @@ -2907,7 +2906,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, end = start + PAGE_CACHE_SIZE - 1; if (test_range_bit(tree, start, end, EXTENT_DIRTY, 0)) { - unlock_page(page); continue; } } @@ -2919,7 +2917,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, PAGECACHE_TAG_DIRTY); } read_unlock_irq(&page->mapping->tree_lock); - unlock_page(page); } return 0; } @@ -2948,17 +2945,12 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, * on us if the page isn't already dirty. */ if (i == 0) { - lock_page(page); set_page_extent_head(page, eb->len); } else if (PagePrivate(page) && page->private != EXTENT_PAGE_PRIVATE) { - lock_page(page); set_page_extent_mapped(page); - unlock_page(page); } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); - if (i == 0) - unlock_page(page); } return set_extent_dirty(tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0c79346fd2c..61bd8953a68 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -115,6 +115,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); btrfs_set_trans_block_group(trans, inode); + mutex_unlock(&root->fs_info->fs_mutex); num_bytes = (end - start + blocksize) & ~(blocksize - 1); num_bytes = max(blocksize, num_bytes); @@ -159,6 +160,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) btrfs_add_ordered_inode(inode); btrfs_update_inode(trans, root, inode); out: + mutex_lock(&root->fs_info->fs_mutex); btrfs_end_transaction(trans, root); return ret; } @@ -349,10 +351,12 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); + mutex_unlock(&root->fs_info->fs_mutex); btrfs_set_trans_block_group(trans, inode); btrfs_csum_file_blocks(trans, root, inode, bio, sums); + mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_end_transaction(trans, root); BUG_ON(ret); mutex_unlock(&root->fs_info->fs_mutex); @@ -807,6 +811,7 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, goto err; } ret = btrfs_delete_one_dir_name(trans, root, path, di); + btrfs_release_path(root, path); dentry->d_inode->i_ctime = dir->i_ctime; ret = btrfs_del_inode_ref(trans, root, name, name_len, @@ -881,8 +886,9 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) struct btrfs_trans_handle *trans; unsigned long nr = 0; - if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) + if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) { return -ENOTEMPTY; + } mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 1); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 6fb45580275..3fbf74e93db 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -43,6 +43,7 @@ #include "ioctl.h" #include "print-tree.h" #include "volumes.h" +#include "locking.h" @@ -75,9 +76,9 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, if (ret) goto fail; - leaf = __btrfs_alloc_free_block(trans, root, root->leafsize, - objectid, trans->transid, 0, 0, - 0, 0); + leaf = btrfs_alloc_free_block(trans, root, root->leafsize, + objectid, trans->transid, 0, 0, + 0, 0); if (IS_ERR(leaf)) return PTR_ERR(leaf); @@ -108,6 +109,7 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); root_item.drop_level = 0; + btrfs_tree_unlock(leaf); free_extent_buffer(leaf); leaf = NULL; diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c new file mode 100644 index 00000000000..80813a307b4 --- /dev/null +++ b/fs/btrfs/locking.c @@ -0,0 +1,50 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ +#include +#include +#include +#include +#include +#include +#include "ctree.h" +#include "extent_io.h" +#include "locking.h" + +int btrfs_tree_lock(struct extent_buffer *eb) +{ + lock_page(eb->first_page); + return 0; +} + +int btrfs_try_tree_lock(struct extent_buffer *eb) +{ + return TestSetPageLocked(eb->first_page); +} + +int btrfs_tree_unlock(struct extent_buffer *eb) +{ + WARN_ON(!PageLocked(eb->first_page)); + unlock_page(eb->first_page); + return 0; +} + +int btrfs_tree_locked(struct extent_buffer *eb) +{ + return PageLocked(eb->first_page); +} + diff --git a/fs/btrfs/locking.h b/fs/btrfs/locking.h new file mode 100644 index 00000000000..2dab96d8280 --- /dev/null +++ b/fs/btrfs/locking.h @@ -0,0 +1,26 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_LOCKING_ +#define __BTRFS_LOCKING_ + +int btrfs_tree_lock(struct extent_buffer *eb); +int btrfs_tree_unlock(struct extent_buffer *eb); +int btrfs_tree_locked(struct extent_buffer *eb); +int btrfs_try_tree_lock(struct extent_buffer *eb); +#endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index f04684f7fea..1ed433a7149 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -23,6 +23,7 @@ #include "ctree.h" #include "disk-io.h" #include "transaction.h" +#include "locking.h" static int total_trans = 0; extern struct kmem_cache *btrfs_trans_handle_cachep; @@ -96,8 +97,7 @@ static noinline int record_root_in_trans(struct btrfs_root *root) radix_tree_tag_set(&root->fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, BTRFS_ROOT_DEFRAG_TAG); - root->commit_root = root->node; - extent_buffer_get(root->node); + root->commit_root = btrfs_root_node(root); } else { WARN_ON(1); } @@ -559,6 +559,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root *tree_root = fs_info->tree_root; struct btrfs_root *root = pending->root; struct extent_buffer *tmp; + struct extent_buffer *old; int ret; int namelen; u64 objectid; @@ -578,16 +579,18 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, key.offset = 1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); - extent_buffer_get(root->node); - btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); - free_extent_buffer(tmp); + old = btrfs_lock_root_node(root); + btrfs_cow_block(trans, root, old, NULL, 0, &old); - btrfs_copy_root(trans, root, root->node, &tmp, objectid); + btrfs_copy_root(trans, root, old, &tmp, objectid); + btrfs_tree_unlock(old); + free_extent_buffer(old); btrfs_set_root_bytenr(new_root_item, tmp->start); btrfs_set_root_level(new_root_item, btrfs_header_level(tmp)); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, new_root_item); + btrfs_tree_unlock(tmp); free_extent_buffer(tmp); if (ret) goto fail; diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 155961c7b4d..fab851d8538 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -181,6 +181,8 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, if (root->fs_info->extent_root == root) is_extent = 1; + goto out; + if (root->ref_cows == 0 && !is_extent) goto out; -- cgit v1.2.3-70-g09d2 From a213501153fd66e2359e091b1612841305ba6551 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:30 -0400 Subject: Btrfs: Replace the big fs_mutex with a collection of other locks Extent alloctions are still protected by a large alloc_mutex. Objectid allocations are covered by a objectid mutex Other btree operations are protected by a lock on individual btree nodes Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 40 +++++++++++++++++++++----- fs/btrfs/ctree.h | 5 ++-- fs/btrfs/disk-io.c | 15 +++++----- fs/btrfs/extent-tree.c | 18 +++++------- fs/btrfs/file.c | 7 +---- fs/btrfs/inode-map.c | 8 ++++++ fs/btrfs/inode.c | 76 ++++---------------------------------------------- fs/btrfs/ioctl.c | 24 ++++------------ fs/btrfs/super.c | 2 -- fs/btrfs/transaction.c | 42 +++++++++++----------------- fs/btrfs/volumes.c | 19 +++++++++---- fs/btrfs/xattr.c | 10 +------ 12 files changed, 101 insertions(+), 165 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 75625c68fdd..dff4da082d0 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -63,6 +63,9 @@ void btrfs_free_path(struct btrfs_path *p) void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) { int i; + int skip = p->skip_locking; + int keep = p->keep_locks; + for (i = 0; i < BTRFS_MAX_LEVEL; i++) { if (!p->nodes[i]) continue; @@ -73,6 +76,8 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) free_extent_buffer(p->nodes[i]); } memset(p, 0, sizeof(*p)); + p->skip_locking = skip; + p->keep_locks = keep; } struct extent_buffer *btrfs_root_node(struct btrfs_root *root) @@ -1202,13 +1207,19 @@ static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock) u32 nritems; t = path->nodes[i]; nritems = btrfs_header_nritems(t); - if (path->slots[i] >= nritems - 1) { + if (nritems < 2 || path->slots[i] >= nritems - 2) { +if (path->keep_locks) { +//printk("path %p skip level now %d\n", path, skip_level); +} skip_level = i + 1; continue; } } t = path->nodes[i]; if (i >= lowest_unlock && i > skip_level && path->locks[i]) { +if (path->keep_locks) { +//printk("path %p unlocking level %d slot %d nritems %d skip_level %d\n", path, i, path->slots[i], btrfs_header_nritems(t), skip_level); +} btrfs_tree_unlock(t); path->locks[i] = 0; } @@ -1243,7 +1254,6 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root lowest_level = p->lowest_level; WARN_ON(lowest_level && ins_len); WARN_ON(p->nodes[0] != NULL); - // WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex)); WARN_ON(root == root->fs_info->extent_root && !mutex_is_locked(&root->fs_info->alloc_mutex)); WARN_ON(root == root->fs_info->chunk_root && @@ -1321,7 +1331,7 @@ again: b = read_node_slot(root, b, slot); if (!p->skip_locking) btrfs_tree_lock(b); - unlock_up(p, level, lowest_unlock); + unlock_up(p, level + 1, lowest_unlock); } else { p->slots[level] = slot; if (ins_len > 0 && btrfs_leaf_free_space(root, b) < @@ -1804,6 +1814,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root if (slot >= btrfs_header_nritems(upper) - 1) return 1; + WARN_ON(!btrfs_tree_locked(path->nodes[1])); + right = read_node_slot(root, upper, slot + 1); btrfs_tree_lock(right); free_space = btrfs_leaf_free_space(root, right); @@ -1981,6 +1993,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root return 1; } + WARN_ON(!btrfs_tree_locked(path->nodes[1])); + left = read_node_slot(root, path->nodes[1], slot - 1); btrfs_tree_lock(left); free_space = btrfs_leaf_free_space(root, left); @@ -2957,15 +2971,16 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1); - path->keep_locks = 1; btrfs_release_path(root, path); + path->keep_locks = 1; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); path->keep_locks = 0; if (ret < 0) return ret; - if (path->slots[0] < nritems - 1) { + nritems = btrfs_header_nritems(path->nodes[0]); + if (nritems > 0 && path->slots[0] < nritems - 1) { goto done; } @@ -2992,8 +3007,17 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) reada_for_search(root, path, level, slot, 0); next = read_node_slot(root, c, slot); - if (!path->skip_locking) + if (!path->skip_locking) { + if (!btrfs_tree_locked(c)) { + int i; + WARN_ON(1); +printk("path %p no lock on level %d\n", path, level); +for (i = 0; i < BTRFS_MAX_LEVEL; i++) { +printk("path %p level %d slot %d nritems %d\n", path, i, path->slots[i], btrfs_header_nritems(path->nodes[i])); +} + } btrfs_tree_lock(next); + } break; } path->slots[level] = slot; @@ -3011,8 +3035,10 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) if (level == 1 && path->locks[1] && path->reada) reada_for_search(root, path, level, slot, 0); next = read_node_slot(root, next, 0); - if (!path->skip_locking) + if (!path->skip_locking) { + WARN_ON(!btrfs_tree_locked(path->nodes[level])); btrfs_tree_lock(next); + } } done: unlock_up(path, 0, 1); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 50891b39f36..692b8ea42de 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -519,9 +519,9 @@ struct btrfs_fs_info { struct backing_dev_info bdi; spinlock_t hash_lock; struct mutex trans_mutex; - struct mutex fs_mutex; struct mutex alloc_mutex; struct mutex chunk_mutex; + struct mutex drop_mutex; struct list_head trans_list; struct list_head hashers; struct list_head dead_roots; @@ -554,7 +554,7 @@ struct btrfs_fs_info { struct completion kobj_unregister; int do_barriers; int closing; - unsigned long throttles; + atomic_t throttles; u64 total_pinned; struct list_head dirty_cowonly_roots; @@ -594,6 +594,7 @@ struct btrfs_root { struct inode *inode; struct kobject root_kobj; struct completion kobj_unregister; + struct mutex objectid_mutex; u64 objectid; u64 last_trans; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index fe40bdd984f..f638803549e 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -724,6 +724,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, INIT_LIST_HEAD(&root->dirty_list); spin_lock_init(&root->node_lock); + mutex_init(&root->objectid_mutex); memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); @@ -1146,6 +1147,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->space_info); btrfs_mapping_init(&fs_info->mapping_tree); atomic_set(&fs_info->nr_async_submits, 0); + atomic_set(&fs_info->throttles, 0); fs_info->sb = sb; fs_info->max_extent = (u64)-1; fs_info->max_inline = 8192 * 1024; @@ -1199,7 +1201,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); mutex_init(&fs_info->trans_mutex); - mutex_init(&fs_info->fs_mutex); + mutex_init(&fs_info->drop_mutex); mutex_init(&fs_info->alloc_mutex); mutex_init(&fs_info->chunk_mutex); @@ -1278,8 +1280,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, goto fail_sb_buffer; } - mutex_lock(&fs_info->fs_mutex); - mutex_lock(&fs_info->chunk_mutex); ret = btrfs_read_sys_array(tree_root); mutex_unlock(&fs_info->chunk_mutex); @@ -1342,7 +1342,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->metadata_alloc_profile = (u64)-1; fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; - mutex_unlock(&fs_info->fs_mutex); return tree_root; fail_extent_root: @@ -1350,7 +1349,6 @@ fail_extent_root: fail_tree_root: free_extent_buffer(tree_root->node); fail_sys_array: - mutex_unlock(&fs_info->fs_mutex); fail_sb_buffer: extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); btrfs_stop_workers(&fs_info->workers); @@ -1562,8 +1560,9 @@ int close_ctree(struct btrfs_root *root) struct btrfs_fs_info *fs_info = root->fs_info; fs_info->closing = 1; + smp_mb(); + btrfs_transaction_flush_work(root); - mutex_lock(&fs_info->fs_mutex); btrfs_defrag_dirty_roots(root->fs_info); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); @@ -1574,7 +1573,6 @@ int close_ctree(struct btrfs_root *root) BUG_ON(ret); write_ctree_super(NULL, root); - mutex_unlock(&fs_info->fs_mutex); btrfs_transaction_flush_work(root); @@ -1679,7 +1677,8 @@ void btrfs_throttle(struct btrfs_root *root) struct backing_dev_info *bdi; bdi = &root->fs_info->bdi; - if (root->fs_info->throttles && bdi_write_congested(bdi)) { + if (atomic_read(&root->fs_info->throttles) && + bdi_write_congested(bdi)) { #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) congestion_wait(WRITE, HZ/20); #else diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 7e40c516fe6..890b9e9d8e2 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1577,9 +1577,11 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root } /* block accounting for super block */ + spin_lock_irq(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); btrfs_set_super_bytes_used(&info->super_copy, super_used - num_bytes); + spin_unlock_irq(&info->delalloc_lock); /* block accounting for root item */ root_used = btrfs_root_used(&root->root_item); @@ -1968,8 +1970,10 @@ again: } /* block accounting for super block */ + spin_lock_irq(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); + spin_unlock_irq(&info->delalloc_lock); /* block accounting for root item */ root_used = btrfs_root_used(&root->root_item); @@ -2172,12 +2176,12 @@ static void noinline reada_walk_down(struct btrfs_root *root, continue; } } - mutex_unlock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->alloc_mutex); ret = readahead_tree_block(root, bytenr, blocksize, btrfs_node_ptr_generation(node, i)); last = bytenr + blocksize; cond_resched(); - mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->alloc_mutex); if (ret) break; } @@ -2254,11 +2258,9 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, free_extent_buffer(next); reada_walk_down(root, cur, path->slots[*level]); - mutex_unlock(&root->fs_info->fs_mutex); mutex_unlock(&root->fs_info->alloc_mutex); next = read_tree_block(root, bytenr, blocksize, ptr_gen); - mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->alloc_mutex); /* we've dropped the lock, double check */ @@ -2381,6 +2383,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root int orig_level; struct btrfs_root_item *root_item = &root->root_item; + WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex)); path = btrfs_alloc_path(); BUG_ON(!path); @@ -2710,7 +2713,6 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, *last_file_root == ref_root) goto out; - mutex_unlock(&extent_root->fs_info->fs_mutex); inode = btrfs_iget_locked(extent_root->fs_info->sb, ref_objectid, found_root); if (inode->i_state & I_NEW) { @@ -2727,7 +2729,6 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, * the latest version of the tree root */ if (is_bad_inode(inode)) { - mutex_lock(&extent_root->fs_info->fs_mutex); goto out; } *last_file_objectid = inode->i_ino; @@ -2736,7 +2737,6 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root, relocate_inode_pages(inode, ref_offset, extent_key->offset); iput(inode); - mutex_lock(&extent_root->fs_info->fs_mutex); } else { struct btrfs_trans_handle *trans; struct extent_buffer *eb; @@ -3033,9 +3033,7 @@ next: if (progress && need_resched()) { memcpy(&key, &found_key, sizeof(key)); - mutex_unlock(&root->fs_info->fs_mutex); cond_resched(); - mutex_lock(&root->fs_info->fs_mutex); btrfs_release_path(root, path); btrfs_search_slot(NULL, root, &key, path, 0, 0); progress = 0; @@ -3068,9 +3066,7 @@ next: trans = btrfs_start_transaction(tree_root, 1); btrfs_commit_transaction(trans, tree_root); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_clean_old_snapshots(tree_root); - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(tree_root, 1); btrfs_commit_transaction(trans, tree_root); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 73c6d085bd9..18bbe108a0e 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -252,7 +252,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, end_of_last_block = start_pos + num_bytes - 1; lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); if (!trans) { err = -ENOMEM; @@ -341,7 +340,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, failed: err = btrfs_end_transaction(trans, root); out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); return err; } @@ -905,9 +903,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, WARN_ON(num_pages > nrptrs); memset(pages, 0, sizeof(pages)); - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, write_bytes, 0); - mutex_unlock(&root->fs_info->fs_mutex); if (ret) goto out; @@ -998,9 +994,9 @@ static int btrfs_sync_file(struct file *file, * check the transaction that last modified this inode * and see if its already been committed */ - mutex_lock(&root->fs_info->fs_mutex); if (!BTRFS_I(inode)->last_trans) goto out; + mutex_lock(&root->fs_info->trans_mutex); if (BTRFS_I(inode)->last_trans <= root->fs_info->last_trans_committed) { @@ -1023,7 +1019,6 @@ static int btrfs_sync_file(struct file *file, } ret = btrfs_commit_transaction(trans, root); out: - mutex_unlock(&root->fs_info->fs_mutex); return ret > 0 ? EIO : ret; } diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index a0925eabdaa..298346ae148 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -69,6 +69,12 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, struct btrfs_key search_key; u64 search_start = dirid; + mutex_lock(&root->objectid_mutex); + if (root->last_inode_alloc) { + *objectid = ++root->last_inode_alloc; + mutex_unlock(&root->objectid_mutex); + return 0; + } path = btrfs_alloc_path(); BUG_ON(!path); search_start = root->last_inode_alloc; @@ -124,9 +130,11 @@ found: btrfs_release_path(root, path); btrfs_free_path(path); BUG_ON(*objectid < search_start); + mutex_unlock(&root->objectid_mutex); return 0; error: btrfs_release_path(root, path); btrfs_free_path(path); + mutex_unlock(&root->objectid_mutex); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 61bd8953a68..b2251e27ac8 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -79,12 +79,15 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int for_del) { - u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy); - u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy); + u64 total; + u64 used; u64 thresh; unsigned long flags; int ret = 0; + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + total = btrfs_super_total_bytes(&root->fs_info->super_copy); + used = btrfs_super_bytes_used(&root->fs_info->super_copy); if (for_del) thresh = total * 90; else @@ -92,7 +95,6 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, do_div(thresh, 100); - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); if (used + root->fs_info->delalloc_bytes + num_required > thresh) ret = -ENOSPC; spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); @@ -115,7 +117,6 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); btrfs_set_trans_block_group(trans, inode); - mutex_unlock(&root->fs_info->fs_mutex); num_bytes = (end - start + blocksize) & ~(blocksize - 1); num_bytes = max(blocksize, num_bytes); @@ -160,7 +161,6 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) btrfs_add_ordered_inode(inode); btrfs_update_inode(trans, root, inode); out: - mutex_lock(&root->fs_info->fs_mutex); btrfs_end_transaction(trans, root); return ret; } @@ -269,14 +269,13 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret; - mutex_lock(&root->fs_info->fs_mutex); + if (btrfs_test_opt(root, NODATACOW) || btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, start, end); else ret = cow_file_range(inode, start, end); - mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -349,17 +348,13 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, ret = btrfs_csum_one_bio(root, bio, &sums); BUG_ON(ret); - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_set_trans_block_group(trans, inode); btrfs_csum_file_blocks(trans, root, inode, bio, sums); - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_end_transaction(trans, root); BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); kfree(sums); @@ -404,7 +399,6 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) btrfs_test_flag(inode, NODATASUM)) return 0; - mutex_lock(&root->fs_info->fs_mutex); path = btrfs_alloc_path(); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); if (IS_ERR(item)) { @@ -422,7 +416,6 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) out: if (path) btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -616,7 +609,6 @@ void btrfs_read_locked_inode(struct inode *inode) path = btrfs_alloc_path(); BUG_ON(!path); - mutex_lock(&root->fs_info->fs_mutex); memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); ret = btrfs_lookup_inode(NULL, root, path, &location, 0); @@ -662,8 +654,6 @@ void btrfs_read_locked_inode(struct inode *inode) btrfs_free_path(path); inode_item = NULL; - mutex_unlock(&root->fs_info->fs_mutex); - switch (inode->i_mode & S_IFMT) { case S_IFREG: inode->i_mapping->a_ops = &btrfs_aops; @@ -691,9 +681,7 @@ void btrfs_read_locked_inode(struct inode *inode) return; make_bad: - btrfs_release_path(root, path); btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); make_bad_inode(inode); } @@ -758,7 +746,6 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, btrfs_set_inode_last_trans(trans, inode); ret = 0; failed: - btrfs_release_path(root, path); btrfs_free_path(path); return ret; } @@ -849,7 +836,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) unsigned long nr = 0; root = BTRFS_I(dir)->root; - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 1); if (ret) @@ -871,7 +857,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) btrfs_end_transaction(trans, root); fail: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); return ret; @@ -890,7 +875,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) return -ENOTEMPTY; } - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 1); if (ret) goto fail; @@ -907,7 +891,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) nr = trans->blocks_used; ret = btrfs_end_transaction(trans, root); fail: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); @@ -1129,7 +1112,6 @@ error: ret = btrfs_del_items(trans, root, path, pending_del_slot, pending_del_nr); } - btrfs_release_path(root, path); btrfs_free_path(path); inode->i_sb->s_dirt = 1; return ret; @@ -1234,9 +1216,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) if (attr->ia_size <= hole_start) goto out; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); - mutex_unlock(&root->fs_info->fs_mutex); if (err) goto fail; @@ -1245,7 +1225,6 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); hole_size = block_end - hole_start; - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); err = btrfs_drop_extents(trans, root, inode, @@ -1262,7 +1241,6 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_check_file(root, inode); } btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); if (err) return err; @@ -1286,7 +1264,6 @@ void btrfs_delete_inode(struct inode *inode) } inode->i_size = 0; - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -1298,7 +1275,6 @@ void btrfs_delete_inode(struct inode *inode) clear_inode(inode); btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); return; @@ -1306,7 +1282,6 @@ void btrfs_delete_inode(struct inode *inode) no_delete_lock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); no_delete: @@ -1402,7 +1377,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, path = btrfs_alloc_path(); BUG_ON(!path); - mutex_lock(&root->fs_info->fs_mutex); *sub_root = btrfs_read_fs_root(root->fs_info, location, dentry->d_name.name, @@ -1416,7 +1390,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, location->offset = 0; btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); return 0; } @@ -1482,9 +1455,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, if (dentry->d_name.len > BTRFS_NAME_LEN) return ERR_PTR(-ENAMETOOLONG); - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_inode_by_name(dir, dentry, &location); - mutex_unlock(&root->fs_info->fs_mutex); if (ret < 0) return ERR_PTR(ret); @@ -1559,7 +1530,6 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) filp->f_pos = 1; } - mutex_lock(&root->fs_info->fs_mutex); key.objectid = inode->i_ino; path = btrfs_alloc_path(); path->reada = 2; @@ -1668,9 +1638,7 @@ read_dir_items: nopos: ret = 0; err: - btrfs_release_path(root, path); btrfs_free_path(path); - mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -1681,11 +1649,9 @@ int btrfs_write_inode(struct inode *inode, int wait) int ret = 0; if (wait) { - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); ret = btrfs_commit_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); } return ret; } @@ -1701,12 +1667,10 @@ void btrfs_dirty_inode(struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); btrfs_update_inode(trans, root, inode); btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); } static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, @@ -1874,7 +1838,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, if (!new_valid_dev(rdev)) return -EINVAL; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; @@ -1912,8 +1875,6 @@ out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); fail: - mutex_unlock(&root->fs_info->fs_mutex); - if (drop_inode) { inode_dec_link_count(inode); iput(inode); @@ -1934,7 +1895,6 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, unsigned long nr = 0; u64 objectid; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; @@ -1980,8 +1940,6 @@ out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); fail: - mutex_unlock(&root->fs_info->fs_mutex); - if (drop_inode) { inode_dec_link_count(inode); iput(inode); @@ -2009,7 +1967,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, #else inc_nlink(inode); #endif - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; @@ -2032,8 +1989,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, nr = trans->blocks_used; btrfs_end_transaction(trans, root); fail: - mutex_unlock(&root->fs_info->fs_mutex); - if (drop_inode) { inode_dec_link_count(inode); iput(inode); @@ -2053,7 +2008,6 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) u64 objectid = 0; unsigned long nr = 1; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); if (err) goto out_unlock; @@ -2106,7 +2060,6 @@ out_fail: btrfs_end_transaction(trans, root); out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); if (drop_on_err) iput(inode); btrfs_btree_balance_dirty(root, nr); @@ -2199,7 +2152,6 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, path = btrfs_alloc_path(); BUG_ON(!path); - mutex_lock(&root->fs_info->fs_mutex); again: spin_lock(&em_tree->lock); @@ -2402,7 +2354,6 @@ out: if (!err) err = ret; } - mutex_unlock(&root->fs_info->fs_mutex); if (err) { free_extent_map(em); WARN_ON(1); @@ -2584,9 +2535,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) int ret; u64 page_start; - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); - mutex_unlock(&root->fs_info->fs_mutex); if (ret) goto out; @@ -2631,7 +2580,6 @@ static void btrfs_truncate(struct inode *inode) btrfs_truncate_page(inode->i_mapping, inode->i_size); - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -2643,7 +2591,6 @@ static void btrfs_truncate(struct inode *inode) ret = btrfs_end_transaction(trans, root); BUG_ON(ret); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); } @@ -2827,7 +2774,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, struct inode *new_inode = new_dentry->d_inode; struct inode *old_inode = old_dentry->d_inode; struct timespec ctime = CURRENT_TIME; - struct btrfs_path *path; int ret; if (S_ISDIR(old_inode->i_mode) && new_inode && @@ -2835,7 +2781,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, return -ENOTEMPTY; } - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 0); if (ret) goto out_unlock; @@ -2843,11 +2788,6 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, new_dir); - path = btrfs_alloc_path(); - if (!path) { - ret = -ENOMEM; - goto out_fail; - } old_dentry->d_inode->i_nlink++; old_dir->i_ctime = old_dir->i_mtime = ctime; @@ -2869,10 +2809,8 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, goto out_fail; out_fail: - btrfs_free_path(path); btrfs_end_transaction(trans, root); out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -2898,7 +2836,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) return -ENAMETOOLONG; - mutex_lock(&root->fs_info->fs_mutex); err = btrfs_check_free_space(root, 1, 0); if (err) goto out_fail; @@ -2979,7 +2916,6 @@ out_unlock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); out_fail: - mutex_unlock(&root->fs_info->fs_mutex); if (drop_inode) { inode_dec_link_count(inode); iput(inode); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 3fbf74e93db..6002eb64daf 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -63,7 +63,6 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; unsigned long nr = 1; - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 0); if (ret) goto fail_commit; @@ -164,7 +163,6 @@ fail: if (err && !ret) ret = err; fail_commit: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); return ret; @@ -181,7 +179,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (!root->ref_cows) return -EINVAL; - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, 1, 0); if (ret) goto fail_unlock; @@ -208,7 +205,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) err = btrfs_commit_transaction(trans, root); fail_unlock: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_btree_balance_dirty(root, nr); btrfs_throttle(root); return ret; @@ -228,9 +224,7 @@ int btrfs_defrag_file(struct file *file) unsigned long i; int ret; - mutex_lock(&root->fs_info->fs_mutex); ret = btrfs_check_free_space(root, inode->i_size, 0); - mutex_unlock(&root->fs_info->fs_mutex); if (ret) return -ENOSPC; @@ -315,7 +309,8 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) goto out; } - mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->alloc_mutex); + mutex_lock(&root->fs_info->chunk_mutex); sizestr = vol_args->name; devstr = strchr(sizestr, ':'); if (devstr) { @@ -385,7 +380,8 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) } out_unlock: - mutex_unlock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->alloc_mutex); + mutex_lock(&root->fs_info->chunk_mutex); out: kfree(vol_args); return ret; @@ -428,11 +424,9 @@ static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, } root_dirid = root->fs_info->sb->s_root->d_inode->i_ino, - mutex_lock(&root->fs_info->fs_mutex); di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path, root_dirid, vol_args->name, namelen, 0); - mutex_unlock(&root->fs_info->fs_mutex); btrfs_free_path(path); if (di && !IS_ERR(di)) { @@ -445,10 +439,12 @@ static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, goto out; } + mutex_lock(&root->fs_info->drop_mutex); if (root == root->fs_info->tree_root) ret = create_subvol(root, vol_args->name, namelen); else ret = create_snapshot(root, vol_args->name, namelen); + mutex_unlock(&root->fs_info->drop_mutex); out: kfree(vol_args); return ret; @@ -461,10 +457,8 @@ static int btrfs_ioctl_defrag(struct file *file) switch (inode->i_mode & S_IFMT) { case S_IFDIR: - mutex_lock(&root->fs_info->fs_mutex); btrfs_defrag_root(root, 0); btrfs_defrag_root(root->fs_info->extent_root, 0); - mutex_unlock(&root->fs_info->fs_mutex); break; case S_IFREG: btrfs_defrag_file(file); @@ -588,7 +582,6 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); } - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 0); path = btrfs_alloc_path(); if (!path) { @@ -685,7 +678,6 @@ out: unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); out_unlock: mutex_unlock(&src->i_mutex); @@ -711,7 +703,6 @@ long btrfs_ioctl_trans_start(struct file *file) if (!capable(CAP_SYS_ADMIN)) return -EPERM; - mutex_lock(&root->fs_info->fs_mutex); if (file->private_data) { ret = -EINPROGRESS; goto out; @@ -723,7 +714,6 @@ long btrfs_ioctl_trans_start(struct file *file) ret = -ENOMEM; /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/ out: - mutex_unlock(&root->fs_info->fs_mutex); return ret; } @@ -740,7 +730,6 @@ long btrfs_ioctl_trans_end(struct file *file) struct btrfs_trans_handle *trans; int ret = 0; - mutex_lock(&root->fs_info->fs_mutex); trans = file->private_data; if (!trans) { ret = -EINVAL; @@ -749,7 +738,6 @@ long btrfs_ioctl_trans_end(struct file *file) btrfs_end_transaction(trans, root); file->private_data = 0; out: - mutex_unlock(&root->fs_info->fs_mutex); return ret; } diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 196d0e280b1..b61ded7a20c 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -366,12 +366,10 @@ int btrfs_sync_fs(struct super_block *sb, int wait) return 0; } btrfs_clean_old_snapshots(root); - mutex_lock(&root->fs_info->fs_mutex); btrfs_defrag_dirty_roots(root->fs_info); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); sb->s_dirt = 0; - mutex_unlock(&root->fs_info->fs_mutex); return ret; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 1ed433a7149..5a1ee0665ae 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -370,6 +370,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) struct btrfs_trans_handle *trans; unsigned long nr; + smp_mb(); if (root->defrag_running) return 0; trans = btrfs_start_transaction(root, 1); @@ -378,16 +379,15 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) ret = btrfs_defrag_leaves(trans, root, cacheonly); nr = trans->blocks_used; btrfs_end_transaction(trans, root); - mutex_unlock(&info->fs_mutex); btrfs_btree_balance_dirty(info->tree_root, nr); cond_resched(); - mutex_lock(&info->fs_mutex); trans = btrfs_start_transaction(root, 1); if (ret != -EAGAIN) break; } root->defrag_running = 0; + smp_mb(); radix_tree_tag_clear(&info->fs_roots_radix, (unsigned long)root->root_key.objectid, BTRFS_ROOT_DEFRAG_TAG); @@ -435,14 +435,14 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, while(!list_empty(list)) { struct btrfs_root *root; - mutex_lock(&tree_root->fs_info->fs_mutex); dirty = list_entry(list->next, struct dirty_root, list); list_del_init(&dirty->list); num_bytes = btrfs_root_used(&dirty->root->root_item); root = dirty->latest_root; - root->fs_info->throttles++; + atomic_inc(&root->fs_info->throttles); + mutex_lock(&root->fs_info->drop_mutex); while(1) { trans = btrfs_start_transaction(tree_root, 1); ret = btrfs_drop_snapshot(trans, dirty->root); @@ -459,14 +459,16 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, nr = trans->blocks_used; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); - mutex_unlock(&tree_root->fs_info->fs_mutex); + + mutex_unlock(&root->fs_info->drop_mutex); btrfs_btree_balance_dirty(tree_root, nr); cond_resched(); - mutex_lock(&tree_root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->drop_mutex); } BUG_ON(ret); - root->fs_info->throttles--; + atomic_dec(&root->fs_info->throttles); + mutex_lock(&root->fs_info->alloc_mutex); num_bytes -= btrfs_root_used(&dirty->root->root_item); bytes_used = btrfs_root_used(&root->root_item); if (num_bytes) { @@ -474,11 +476,15 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, btrfs_set_root_used(&root->root_item, bytes_used - num_bytes); } + mutex_unlock(&root->fs_info->alloc_mutex); + ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); if (ret) { BUG(); break; } + mutex_unlock(&root->fs_info->drop_mutex); + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); @@ -486,7 +492,6 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, free_extent_buffer(dirty->root->node); kfree(dirty->root); kfree(dirty); - mutex_unlock(&tree_root->fs_info->fs_mutex); btrfs_btree_balance_dirty(tree_root, nr); cond_resched(); @@ -503,7 +508,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, u64 objectid = 0; int ret; - root->fs_info->throttles++; + atomic_inc(&root->fs_info->throttles); while(1) { ret = btrfs_find_first_ordered_inode( &cur_trans->ordered_inode_tree, @@ -512,7 +517,6 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, break; mutex_unlock(&root->fs_info->trans_mutex); - mutex_unlock(&root->fs_info->fs_mutex); if (S_ISREG(inode->i_mode)) { atomic_inc(&BTRFS_I(inode)->ordered_writeback); @@ -521,7 +525,6 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, } iput(inode); - mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); } while(1) { @@ -533,7 +536,6 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, if (!ret) break; mutex_unlock(&root->fs_info->trans_mutex); - mutex_unlock(&root->fs_info->fs_mutex); if (S_ISREG(inode->i_mode)) { atomic_inc(&BTRFS_I(inode)->ordered_writeback); @@ -543,10 +545,9 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, atomic_dec(&inode->i_count); iput(inode); - mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); } - root->fs_info->throttles--; + atomic_dec(&root->fs_info->throttles); return 0; } @@ -661,7 +662,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->trans_mutex); btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); ret = wait_for_commit(root, cur_trans); BUG_ON(ret); @@ -669,7 +669,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, put_transaction(cur_trans); mutex_unlock(&root->fs_info->trans_mutex); - mutex_lock(&root->fs_info->fs_mutex); return 0; } @@ -687,12 +686,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction, list); if (!prev_trans->commit_done) { prev_trans->use_count++; - mutex_unlock(&root->fs_info->fs_mutex); mutex_unlock(&root->fs_info->trans_mutex); wait_for_commit(root, prev_trans); - mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); put_transaction(prev_trans); } @@ -709,12 +706,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, else timeout = 1; - mutex_unlock(&root->fs_info->fs_mutex); mutex_unlock(&root->fs_info->trans_mutex); schedule_timeout(timeout); - mutex_lock(&root->fs_info->fs_mutex); mutex_lock(&root->fs_info->trans_mutex); finish_wait(&cur_trans->writer_wait, &wait); ret = btrfs_write_ordered_inodes(trans, root); @@ -755,12 +750,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_copy_pinned(root, pinned_copy); mutex_unlock(&root->fs_info->trans_mutex); - mutex_unlock(&root->fs_info->fs_mutex); ret = btrfs_write_and_wait_transaction(trans, root); BUG_ON(ret); write_ctree_super(trans, root); - mutex_lock(&root->fs_info->fs_mutex); btrfs_finish_extent_commit(trans, root, pinned_copy); mutex_lock(&root->fs_info->trans_mutex); @@ -781,9 +774,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, kmem_cache_free(btrfs_trans_handle_cachep, trans); if (root->fs_info->closing) { - mutex_unlock(&root->fs_info->fs_mutex); drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots); - mutex_lock(&root->fs_info->fs_mutex); } return ret; } @@ -823,7 +814,7 @@ void btrfs_transaction_cleaner(struct work_struct *work) unsigned long delay = HZ * 30; int ret; - mutex_lock(&root->fs_info->fs_mutex); + smp_mb(); if (root->fs_info->closing) goto out; @@ -844,7 +835,6 @@ void btrfs_transaction_cleaner(struct work_struct *work) trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); out: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_clean_old_snapshots(root); btrfs_transaction_queue_work(root, delay); } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index ba396857102..869864ddcc2 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -866,7 +866,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) u64 devid; int ret = 0; - mutex_lock(&root->fs_info->fs_mutex); + mutex_lock(&root->fs_info->alloc_mutex); + mutex_lock(&root->fs_info->chunk_mutex); mutex_lock(&uuid_mutex); all_avail = root->fs_info->avail_data_alloc_bits | @@ -984,7 +985,8 @@ error_close: close_bdev_excl(bdev); out: mutex_unlock(&uuid_mutex); - mutex_unlock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->chunk_mutex); + mutex_unlock(&root->fs_info->alloc_mutex); return ret; } @@ -1003,7 +1005,10 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) if (!bdev) { return -EIO; } - mutex_lock(&root->fs_info->fs_mutex); + + mutex_lock(&root->fs_info->alloc_mutex); + mutex_lock(&root->fs_info->chunk_mutex); + trans = btrfs_start_transaction(root, 1); devices = &root->fs_info->fs_devices->devices; list_for_each(cur, devices) { @@ -1057,7 +1062,9 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) root->fs_info->fs_devices->open_devices++; out: btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); + mutex_unlock(&root->fs_info->chunk_mutex); + mutex_unlock(&root->fs_info->alloc_mutex); + return ret; out_close_bdev: @@ -1297,9 +1304,10 @@ int btrfs_balance(struct btrfs_root *dev_root) struct btrfs_key found_key; + BUG(); /* FIXME, needs locking */ + dev_root = dev_root->fs_info->dev_root; - mutex_lock(&dev_root->fs_info->fs_mutex); /* step one make some room on all the devices */ list_for_each(cur, devices) { device = list_entry(cur, struct btrfs_device, dev_list); @@ -1368,7 +1376,6 @@ int btrfs_balance(struct btrfs_root *dev_root) ret = 0; error: btrfs_free_path(path); - mutex_unlock(&dev_root->fs_info->fs_mutex); return ret; } diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c index f63488dc2f1..3e9eb91c3c8 100644 --- a/fs/btrfs/xattr.c +++ b/fs/btrfs/xattr.c @@ -153,7 +153,6 @@ ssize_t btrfs_xattr_get(struct inode *inode, int name_index, return -ENOMEM; } - mutex_lock(&root->fs_info->fs_mutex); /* lookup the xattr by name */ di = btrfs_lookup_xattr(NULL, root, path, inode->i_ino, name, strlen(name), 0); @@ -181,7 +180,6 @@ ssize_t btrfs_xattr_get(struct inode *inode, int name_index, ret = btrfs_dir_data_len(leaf, di); out: - mutex_unlock(&root->fs_info->fs_mutex); kfree(name); btrfs_free_path(path); return ret; @@ -210,7 +208,6 @@ int btrfs_xattr_set(struct inode *inode, int name_index, return -ENOMEM; } - mutex_lock(&root->fs_info->fs_mutex); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -260,7 +257,6 @@ out: } btrfs_end_transaction(trans, root); - mutex_unlock(&root->fs_info->fs_mutex); kfree(name); btrfs_free_path(path); @@ -297,8 +293,6 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size) return -ENOMEM; path->reada = 2; - mutex_lock(&root->fs_info->fs_mutex); - /* search for our xattrs */ ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) @@ -379,15 +373,13 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size) ret = total_size; err: - mutex_unlock(&root->fs_info->fs_mutex); btrfs_free_path(path); return ret; } /* - * delete all the xattrs associated with the inode. fs_mutex should be - * held when we come into here + * delete all the xattrs associated with the inode. */ int btrfs_delete_xattrs(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) -- cgit v1.2.3-70-g09d2 From 594a24eb0e7fa8413f8b443863be4b7c72bfde9f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:30 -0400 Subject: Fix btrfs_del_ordered_inode to allow forcing the drop during unlinks This allows us to delete an unlinked inode with dirty pages from the list instead of forcing commit to write these out before deleting the inode. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 18 +++++++++++++----- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 4 +--- fs/btrfs/ordered-data.c | 6 +++--- fs/btrfs/ordered-data.h | 2 +- 5 files changed, 19 insertions(+), 13 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 1b756fae279..9601241e552 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1245,7 +1245,10 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root int level; int should_reada = p->reada; int lowest_unlock = 1; + int blocksize; u8 lowest_level = 0; + u64 blocknr; + u64 gen; lowest_level = p->lowest_level; WARN_ON(lowest_level && ins_len); @@ -1320,11 +1323,12 @@ again: reada_for_search(root, p, level, slot, key->objectid); - tmp = btrfs_find_tree_block(root, - btrfs_node_blockptr(b, slot), - btrfs_level_size(root, level - 1)); - if (tmp && btrfs_buffer_uptodate(tmp, - btrfs_node_ptr_generation(b, slot))) { + blocknr = btrfs_node_blockptr(b, slot); + gen = btrfs_node_ptr_generation(b, slot); + blocksize = btrfs_level_size(root, level - 1); + + tmp = btrfs_find_tree_block(root, blocknr, blocksize); + if (tmp && btrfs_buffer_uptodate(tmp, gen)) { b = tmp; } else { /* @@ -1334,6 +1338,10 @@ again: */ if (level > 1) { btrfs_release_path(NULL, p); + if (tmp) + free_extent_buffer(tmp); + tmp = read_tree_block(root, blocknr, + blocksize, gen); if (tmp) free_extent_buffer(tmp); goto again; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 18bbe108a0e..b7f8f92daf8 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -976,7 +976,7 @@ out_nolock: int btrfs_release_file(struct inode * inode, struct file * filp) { - btrfs_del_ordered_inode(inode); + btrfs_del_ordered_inode(inode, 0); if (filp->private_data) btrfs_ioctl_trans_end(filp); return 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b2251e27ac8..cf27b598462 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -852,7 +852,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) * we don't need to worry about * data=ordered */ - btrfs_del_ordered_inode(inode); + btrfs_del_ordered_inode(inode, 1); } btrfs_end_transaction(trans, root); @@ -1276,14 +1276,12 @@ void btrfs_delete_inode(struct inode *inode) btrfs_end_transaction(trans, root); btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return; no_delete_lock: nr = trans->blocks_used; btrfs_end_transaction(trans, root); btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); no_delete: clear_inode(inode); } diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 8dd8180183f..5e4c0d95ce4 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -254,7 +254,7 @@ static void __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, return; } -void btrfs_del_ordered_inode(struct inode *inode) +void btrfs_del_ordered_inode(struct inode *inode, int force) { struct btrfs_root *root = BTRFS_I(inode)->root; u64 root_objectid = root->root_key.objectid; @@ -263,8 +263,8 @@ void btrfs_del_ordered_inode(struct inode *inode) return; } - if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || - mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK)) + if (!force && (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || + mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))) return; spin_lock(&root->fs_info->new_trans_lock); diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index c515c4b3999..4fa78736423 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -38,6 +38,6 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, u64 *root_objectid, u64 *objectid, struct inode **inode); -void btrfs_del_ordered_inode(struct inode *inode); +void btrfs_del_ordered_inode(struct inode *inode, int force); int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode); #endif -- cgit v1.2.3-70-g09d2 From 89ce8a63d0c761fbb02089850605360f389477d8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:31 -0400 Subject: Add btrfs_end_transaction_throttle to force writers to wait for pending commits The existing throttle mechanism was often not sufficient to prevent new writers from coming in and making a given transaction run forever. This adds an explicit wait at the end of most operations so they will allow the current transaction to close. There is no wait inside file_write, inode updates, or cow filling, all which have different deadlock possibilities. This is a temporary measure until better asynchronous commit support is added. This code leads to stalls as it waits for data=ordered writeback, and it really needs to be fixed. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 18 -------------- fs/btrfs/disk-io.h | 1 - fs/btrfs/file.c | 1 - fs/btrfs/inode.c | 24 ++++++------------ fs/btrfs/ioctl.c | 2 -- fs/btrfs/transaction.c | 66 ++++++++++++++++++++++++++++++++++---------------- fs/btrfs/transaction.h | 2 ++ 7 files changed, 55 insertions(+), 59 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3cc480b8381..52569b57692 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1672,24 +1672,6 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); } -void btrfs_throttle(struct btrfs_root *root) -{ -#if 0 - struct backing_dev_info *bdi; - - bdi = &root->fs_info->bdi; - if (atomic_read(&root->fs_info->throttles) && - bdi_write_congested(bdi)) { -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) - congestion_wait(WRITE, HZ/20); -#else - blk_congestion_wait(WRITE, HZ/20); -#endif - - } -#endif -} - void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) { /* diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 2bc64fefe6e..deff6b4815a 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -70,7 +70,6 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf); int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid); u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len); void btrfs_csum_final(u32 crc, char *result); -void btrfs_throttle(struct btrfs_root *root); int btrfs_open_device(struct btrfs_device *dev); int btrfs_verify_block_csum(struct btrfs_root *root, struct extent_buffer *buf); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index b7f8f92daf8..ece221cba90 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -934,7 +934,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1) btrfs_btree_balance_dirty(root, 1); - btrfs_throttle(root); cond_resched(); } out: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index cf27b598462..bbba3350d02 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -855,10 +855,9 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) btrfs_del_ordered_inode(inode, 1); } - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return ret; } @@ -889,10 +888,9 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) } nr = trans->blocks_used; - ret = btrfs_end_transaction(trans, root); + ret = btrfs_end_transaction_throttle(trans, root); fail: btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); if (ret && !err) err = ret; @@ -1871,14 +1869,13 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); iput(inode); } btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return err; } @@ -1936,14 +1933,13 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); iput(inode); } btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return err; } @@ -1985,14 +1981,13 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); iput(inode); } btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return err; } @@ -2055,13 +2050,12 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) out_fail: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); out_unlock: if (drop_on_err) iput(inode); btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return err; } @@ -2587,10 +2581,9 @@ static void btrfs_truncate(struct inode *inode) btrfs_update_inode(trans, root, inode); nr = trans->blocks_used; - ret = btrfs_end_transaction(trans, root); + ret = btrfs_end_transaction_throttle(trans, root); BUG_ON(ret); btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); } /* @@ -2912,14 +2905,13 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); out_fail: if (drop_inode) { inode_dec_link_count(inode); iput(inode); } btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return err; } diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 6002eb64daf..026039a2ac5 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -164,7 +164,6 @@ fail: ret = err; fail_commit: btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return ret; } @@ -206,7 +205,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) fail_unlock: btrfs_btree_balance_dirty(root, nr); - btrfs_throttle(root); return ret; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 5a1ee0665ae..69ed5f85a38 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -130,8 +130,27 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, return h; } -int btrfs_end_transaction(struct btrfs_trans_handle *trans, - struct btrfs_root *root) +static noinline int wait_for_commit(struct btrfs_root *root, + struct btrfs_transaction *commit) +{ + DEFINE_WAIT(wait); + mutex_lock(&root->fs_info->trans_mutex); + while(!commit->commit_done) { + prepare_to_wait(&commit->commit_wait, &wait, + TASK_UNINTERRUPTIBLE); + if (commit->commit_done) + break; + mutex_unlock(&root->fs_info->trans_mutex); + schedule(); + mutex_lock(&root->fs_info->trans_mutex); + } + mutex_unlock(&root->fs_info->trans_mutex); + finish_wait(&commit->commit_wait, &wait); + return 0; +} + +static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root, int throttle) { struct btrfs_transaction *cur_trans; @@ -140,8 +159,18 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans, WARN_ON(cur_trans != trans->transaction); WARN_ON(cur_trans->num_writers < 1); cur_trans->num_writers--; + if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); + + if (cur_trans->in_commit && throttle) { + int ret; + mutex_unlock(&root->fs_info->trans_mutex); + ret = wait_for_commit(root, cur_trans); + BUG_ON(ret); + mutex_lock(&root->fs_info->trans_mutex); + } + put_transaction(cur_trans); mutex_unlock(&root->fs_info->trans_mutex); memset(trans, 0, sizeof(*trans)); @@ -149,6 +178,18 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans, return 0; } +int btrfs_end_transaction(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + return __btrfs_end_transaction(trans, root, 0); +} + +int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + return __btrfs_end_transaction(trans, root, 1); +} + int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root) @@ -240,25 +281,6 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, return 0; } -static noinline int wait_for_commit(struct btrfs_root *root, - struct btrfs_transaction *commit) -{ - DEFINE_WAIT(wait); - mutex_lock(&root->fs_info->trans_mutex); - while(!commit->commit_done) { - prepare_to_wait(&commit->commit_wait, &wait, - TASK_UNINTERRUPTIBLE); - if (commit->commit_done) - break; - mutex_unlock(&root->fs_info->trans_mutex); - schedule(); - mutex_lock(&root->fs_info->trans_mutex); - } - mutex_unlock(&root->fs_info->trans_mutex); - finish_wait(&commit->commit_wait, &wait); - return 0; -} - struct dirty_root { struct list_head list; struct btrfs_root *root; @@ -680,6 +702,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, root->fs_info->btree_inode->i_mapping, GFP_NOFS); trans->transaction->in_commit = 1; +printk("trans %Lu in commit\n", trans->transid); cur_trans = trans->transaction; if (cur_trans->list.prev != &root->fs_info->trans_list) { prev_trans = list_entry(cur_trans->list.prev, @@ -760,6 +783,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, kfree(pinned_copy); cur_trans->commit_done = 1; +printk("trans %Lu done in commit\n", cur_trans->transid); root->fs_info->last_trans_committed = cur_trans->transid; wake_up(&cur_trans->commit_wait); put_transaction(cur_trans); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index c3172ddb332..52559b51b18 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -101,4 +101,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, struct btrfs_root *root); +int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, + struct btrfs_root *root); #endif -- cgit v1.2.3-70-g09d2 From 1b1e2135dc1e4efbcf25ac9ac9979316d4e1193e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 25 Jun 2008 16:01:31 -0400 Subject: Btrfs: Add a per-inode csum mutex to avoid races creating csum items Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/file.c | 6 +++--- fs/btrfs/inode.c | 6 ++++++ fs/btrfs/ordered-data.c | 5 +++-- fs/btrfs/transaction.c | 2 +- fs/btrfs/tree-defrag.c | 7 +++++++ 6 files changed, 21 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 5ba83894c8b..40b4e0c9cd0 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -30,6 +30,7 @@ struct btrfs_inode { struct extent_map_tree extent_tree; struct extent_io_tree io_tree; struct extent_io_tree io_failure_tree; + struct mutex csum_mutex; struct inode vfs_inode; atomic_t ordered_writeback; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index ece221cba90..8037792f878 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -267,13 +267,13 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, /* FIXME...EIEIO, ENOSPC and more */ /* insert any holes we need to create */ - if (isize < end_pos) { + if (isize < start_pos) { u64 last_pos_in_file; u64 hole_size; u64 mask = root->sectorsize - 1; last_pos_in_file = (isize + mask) & ~mask; - hole_size = (end_pos - last_pos_in_file + mask) & ~mask; - if (last_pos_in_file < end_pos) { + hole_size = (start_pos - last_pos_in_file + mask) & ~mask; + if (last_pos_in_file < start_pos) { err = btrfs_drop_extents(trans, root, inode, last_pos_in_file, last_pos_in_file + hole_size, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bbba3350d02..d39433dfb2c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -351,7 +351,9 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); + mutex_lock(&BTRFS_I(inode)->csum_mutex); btrfs_csum_file_blocks(trans, root, inode, bio, sums); + mutex_unlock(&BTRFS_I(inode)->csum_mutex); ret = btrfs_end_transaction(trans, root); BUG_ON(ret); @@ -1400,6 +1402,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + mutex_init(&BTRFS_I(inode)->csum_mutex); atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); return 0; } @@ -1701,6 +1704,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + mutex_init(&BTRFS_I(inode)->csum_mutex); atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->root = root; @@ -1924,6 +1928,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; @@ -2862,6 +2867,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 5e4c0d95ce4..254da822566 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -152,12 +152,13 @@ int btrfs_add_ordered_inode(struct inode *inode) inode->i_ino, &entry->rb_node); BTRFS_I(inode)->ordered_trans = transid; + if (!node) + igrab(inode); write_unlock(&tree->lock); + if (node) kfree(entry); - else - igrab(inode); return 0; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 0c53ff775b9..8e909cb97c6 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -477,7 +477,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, if (err) ret = err; nr = trans->blocks_used; - ret = btrfs_end_transaction(trans, tree_root); + ret = btrfs_end_transaction_throttle(trans, tree_root); BUG_ON(ret); mutex_unlock(&root->fs_info->drop_mutex); diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index 1677e4edaf6..b17693f61fb 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -38,8 +38,15 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, u64 last_ret = 0; if (root->fs_info->extent_root == root) { + /* + * there's recursion here right now in the tree locking, + * we can't defrag the extent root without deadlock + */ + goto out; +#if 0 mutex_lock(&root->fs_info->alloc_mutex); is_extent = 1; +#endif } if (root->ref_cows == 0 && !is_extent) -- cgit v1.2.3-70-g09d2 From e6dcd2dc9c489108648e2ed543315dd134d50a9a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:53:50 -0400 Subject: Btrfs: New data=ordered implementation The old data=ordered code would force commit to wait until all the data extents from the transaction were fully on disk. This introduced large latencies into the commit and stalled new writers in the transaction for a long time. The new code changes the way data allocations and extents work: * When delayed allocation is filled, data extents are reserved, and the extent bit EXTENT_ORDERED is set on the entire range of the extent. A struct btrfs_ordered_extent is allocated an inserted into a per-inode rbtree to track the pending extents. * As each page is written EXTENT_ORDERED is cleared on the bytes corresponding to that page. * When all of the bytes corresponding to a single struct btrfs_ordered_extent are written, The previously reserved extent is inserted into the FS btree and into the extent allocation trees. The checksums for the file data are also updated. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 4 +- fs/btrfs/ctree.h | 19 +- fs/btrfs/disk-io.c | 13 +- fs/btrfs/extent-tree.c | 132 +++++++++----- fs/btrfs/extent_io.c | 52 +++++- fs/btrfs/extent_io.h | 14 +- fs/btrfs/extent_map.c | 5 +- fs/btrfs/file-item.c | 62 ++++--- fs/btrfs/file.c | 67 ++++--- fs/btrfs/inode.c | 447 ++++++++++++++++++++++++++++++----------------- fs/btrfs/ordered-data.c | 455 ++++++++++++++++++++++++++++++------------------ fs/btrfs/ordered-data.h | 71 ++++++-- fs/btrfs/transaction.c | 67 +------ fs/btrfs/transaction.h | 4 - 14 files changed, 910 insertions(+), 502 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 40b4e0c9cd0..8d03687510e 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -21,6 +21,7 @@ #include "extent_map.h" #include "extent_io.h" +#include "ordered-data.h" /* in memory btrfs inode */ struct btrfs_inode { @@ -32,9 +33,8 @@ struct btrfs_inode { struct extent_io_tree io_failure_tree; struct mutex csum_mutex; struct inode vfs_inode; - atomic_t ordered_writeback; + struct btrfs_ordered_inode_tree ordered_tree; - u64 ordered_trans; /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f3783dbd9b6..ceebc052ddc 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -25,6 +25,7 @@ #include #include #include +#include #include #include "bit-radix.h" #include "extent_io.h" @@ -37,6 +38,7 @@ extern struct kmem_cache *btrfs_trans_handle_cachep; extern struct kmem_cache *btrfs_transaction_cachep; extern struct kmem_cache *btrfs_bit_radix_cachep; extern struct kmem_cache *btrfs_path_cachep; +struct btrfs_ordered_sum; #define BTRFS_MAGIC "_B5RfS_M" @@ -510,6 +512,7 @@ struct btrfs_fs_info { u64 max_inline; u64 alloc_start; struct btrfs_transaction *running_transaction; + wait_queue_head_t transaction_throttle; struct btrfs_super_block super_copy; struct btrfs_super_block super_for_commit; struct block_device *__bdev; @@ -541,6 +544,7 @@ struct btrfs_fs_info { */ struct btrfs_workers workers; struct btrfs_workers endio_workers; + struct btrfs_workers endio_write_workers; struct btrfs_workers submit_workers; struct task_struct *transaction_kthread; struct task_struct *cleaner_kthread; @@ -1384,6 +1388,17 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, u64 data); +int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + struct btrfs_key *ins); +int btrfs_reserve_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 num_bytes, u64 min_alloc_size, + u64 empty_size, u64 hint_byte, + u64 search_end, struct btrfs_key *ins, + u64 data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf); int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root @@ -1556,9 +1571,9 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, u64 bytenr, int mod); int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - struct bio *bio, char *sums); + struct btrfs_ordered_sum *sums); int btrfs_csum_one_bio(struct btrfs_root *root, - struct bio *bio, char **sums_ret); + struct bio *bio, struct btrfs_ordered_sum **sums_ret); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index b01b3f4f92a..4a5ebafb935 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -407,7 +407,11 @@ static int end_workqueue_bio(struct bio *bio, end_io_wq->error = err; end_io_wq->work.func = end_workqueue_fn; end_io_wq->work.flags = 0; - btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); + if (bio->bi_rw & (1 << BIO_RW)) + btrfs_queue_worker(&fs_info->endio_write_workers, + &end_io_wq->work); + else + btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) return 0; @@ -1286,6 +1290,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, mutex_init(&fs_info->transaction_kthread_mutex); mutex_init(&fs_info->cleaner_mutex); mutex_init(&fs_info->volume_mutex); + init_waitqueue_head(&fs_info->transaction_throttle); #if 0 ret = add_hasher(fs_info, "crc32c"); @@ -1325,9 +1330,13 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->endio_write_workers, + fs_info->thread_pool_size); btrfs_start_workers(&fs_info->workers, 1); btrfs_start_workers(&fs_info->submit_workers, 1); btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); + btrfs_start_workers(&fs_info->endio_write_workers, + fs_info->thread_pool_size); err = -EINVAL; if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { @@ -1447,6 +1456,7 @@ fail_sb_buffer: extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); + btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); fail_iput: iput(fs_info->btree_inode); @@ -1702,6 +1712,7 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); + btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); iput(fs_info->btree_inode); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 8ebfa6be079..343d1101c31 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1895,36 +1895,17 @@ error: return ret; } -/* - * finds a free extent and does all the dirty work required for allocation - * returns the key for the extent through ins, and a tree buffer for - * the first block of the extent through buf. - * - * returns 0 if everything worked, non-zero otherwise. - */ -int btrfs_alloc_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 num_bytes, u64 min_alloc_size, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - u64 empty_size, u64 hint_byte, - u64 search_end, struct btrfs_key *ins, u64 data) +static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 num_bytes, u64 min_alloc_size, + u64 empty_size, u64 hint_byte, + u64 search_end, struct btrfs_key *ins, + u64 data) { int ret; - int pending_ret; - u64 super_used; - u64 root_used; u64 search_start = 0; u64 alloc_profile; - u32 sizes[2]; struct btrfs_fs_info *info = root->fs_info; - struct btrfs_root *extent_root = info->extent_root; - struct btrfs_extent_item *extent_item; - struct btrfs_extent_ref *ref; - struct btrfs_path *path; - struct btrfs_key keys[2]; - - maybe_lock_mutex(root); if (data) { alloc_profile = info->avail_data_alloc_bits & @@ -1974,11 +1955,48 @@ again: } if (ret) { printk("allocation failed flags %Lu\n", data); - } - if (ret) { BUG(); - goto out; } + clear_extent_dirty(&root->fs_info->free_space_cache, + ins->objectid, ins->objectid + ins->offset - 1, + GFP_NOFS); + return 0; +} + +int btrfs_reserve_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 num_bytes, u64 min_alloc_size, + u64 empty_size, u64 hint_byte, + u64 search_end, struct btrfs_key *ins, + u64 data) +{ + int ret; + maybe_lock_mutex(root); + ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size, + empty_size, hint_byte, search_end, ins, + data); + maybe_unlock_mutex(root); + return ret; +} + +static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + struct btrfs_key *ins) +{ + int ret; + int pending_ret; + u64 super_used; + u64 root_used; + u64 num_bytes = ins->offset; + u32 sizes[2]; + struct btrfs_fs_info *info = root->fs_info; + struct btrfs_root *extent_root = info->extent_root; + struct btrfs_extent_item *extent_item; + struct btrfs_extent_ref *ref; + struct btrfs_path *path; + struct btrfs_key keys[2]; /* block accounting for super block */ spin_lock_irq(&info->delalloc_lock); @@ -1990,10 +2008,6 @@ again: root_used = btrfs_root_used(&root->root_item); btrfs_set_root_used(&root->root_item, root_used + num_bytes); - clear_extent_dirty(&root->fs_info->free_space_cache, - ins->objectid, ins->objectid + ins->offset - 1, - GFP_NOFS); - if (root == extent_root) { set_extent_bits(&root->fs_info->extent_ins, ins->objectid, ins->objectid + ins->offset - 1, @@ -2001,10 +2015,6 @@ again: goto update_block; } - WARN_ON(trans->alloc_exclude_nr); - trans->alloc_exclude_start = ins->objectid; - trans->alloc_exclude_nr = ins->offset; - memcpy(&keys[0], ins, sizeof(*ins)); keys[1].offset = hash_extent_ref(root_objectid, ref_generation, owner, owner_offset); @@ -2054,6 +2064,51 @@ update_block: BUG(); } out: + return ret; +} + +int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + struct btrfs_key *ins) +{ + int ret; + maybe_lock_mutex(root); + ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, + ref_generation, owner, + owner_offset, ins); + maybe_unlock_mutex(root); + return ret; +} +/* + * finds a free extent and does all the dirty work required for allocation + * returns the key for the extent through ins, and a tree buffer for + * the first block of the extent through buf. + * + * returns 0 if everything worked, non-zero otherwise. + */ +int btrfs_alloc_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 num_bytes, u64 min_alloc_size, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + u64 empty_size, u64 hint_byte, + u64 search_end, struct btrfs_key *ins, u64 data) +{ + int ret; + + maybe_lock_mutex(root); + + ret = __btrfs_reserve_extent(trans, root, num_bytes, + min_alloc_size, empty_size, hint_byte, + search_end, ins, data); + BUG_ON(ret); + ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, + ref_generation, owner, + owner_offset, ins); + BUG_ON(ret); + maybe_unlock_mutex(root); return ret; } @@ -2288,8 +2343,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->alloc_mutex); /* we've dropped the lock, double check */ - ret = drop_snap_lookup_refcount(root, bytenr, - blocksize, &refs); + ret = lookup_extent_ref(NULL, root, bytenr, blocksize, + &refs); BUG_ON(ret); if (refs != 1) { parent = path->nodes[*level]; @@ -2584,7 +2639,6 @@ out_unlock: kfree(ra); trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1); if (trans) { - btrfs_add_ordered_inode(inode); btrfs_end_transaction(trans, BTRFS_I(inode)->root); mark_inode_dirty(inode); } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 40a5f53cb04..3f82a6e9ca4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -793,6 +793,13 @@ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_dirty); +int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return set_extent_bit(tree, start, end, EXTENT_ORDERED, 0, NULL, mask); +} +EXPORT_SYMBOL(set_extent_ordered); + int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask) { @@ -812,8 +819,8 @@ int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, - EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL, - mask); + EXTENT_DELALLOC | EXTENT_DIRTY, + 0, NULL, mask); } EXPORT_SYMBOL(set_extent_delalloc); @@ -825,6 +832,13 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(clear_extent_dirty); +int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask) +{ + return clear_extent_bit(tree, start, end, EXTENT_ORDERED, 1, 0, mask); +} +EXPORT_SYMBOL(clear_extent_ordered); + int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { @@ -1395,10 +1409,9 @@ static int end_bio_extent_writepage(struct bio *bio, if (--bvec >= bio->bi_io_vec) prefetchw(&bvec->bv_page->flags); - if (tree->ops && tree->ops->writepage_end_io_hook) { ret = tree->ops->writepage_end_io_hook(page, start, - end, state); + end, state, uptodate); if (ret) uptodate = 0; } @@ -1868,9 +1881,14 @@ static int __extent_read_full_page(struct extent_io_tree *tree, unlock_extent(tree, cur, end, GFP_NOFS); break; } - extent_offset = cur - em->start; + if (extent_map_end(em) <= cur) { +printk("bad mapping em [%Lu %Lu] cur %Lu\n", em->start, extent_map_end(em), cur); + } BUG_ON(extent_map_end(em) <= cur); + if (end < cur) { +printk("2bad mapping end %Lu cur %Lu\n", end, cur); + } BUG_ON(end < cur); iosize = min(extent_map_end(em) - cur, end - cur + 1); @@ -1976,6 +1994,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 last_byte = i_size_read(inode); u64 block_start; u64 iosize; + u64 unlock_start; sector_t sector; struct extent_map *em; struct block_device *bdev; @@ -1988,7 +2007,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 nr_delalloc; u64 delalloc_end; - WARN_ON(!PageLocked(page)); page_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || @@ -2030,6 +2048,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = delalloc_end + 1; } lock_extent(tree, start, page_end, GFP_NOFS); + unlock_start = start; end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { @@ -2038,6 +2057,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, if (last_byte <= start) { clear_extent_dirty(tree, start, page_end, GFP_NOFS); + unlock_extent(tree, start, page_end, GFP_NOFS); + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, + page_end, NULL, 1); + unlock_start = page_end + 1; goto done; } @@ -2047,6 +2071,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, while (cur <= end) { if (cur >= last_byte) { clear_extent_dirty(tree, cur, page_end, GFP_NOFS); + unlock_extent(tree, unlock_start, page_end, GFP_NOFS); + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, cur, + page_end, NULL, 1); + unlock_start = page_end + 1; break; } em = epd->get_extent(inode, page, page_offset, cur, @@ -2071,8 +2100,16 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, block_start == EXTENT_MAP_INLINE) { clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); + + unlock_extent(tree, unlock_start, cur + iosize -1, + GFP_NOFS); + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, cur, + cur + iosize - 1, + NULL, 1); cur = cur + iosize; page_offset += iosize; + unlock_start = cur; continue; } @@ -2119,7 +2156,8 @@ done: set_page_writeback(page); end_page_writeback(page); } - unlock_extent(tree, start, page_end, GFP_NOFS); + if (unlock_start <= page_end) + unlock_extent(tree, unlock_start, page_end, GFP_NOFS); unlock_page(page); return 0; } diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index f1960dafaa1..2268a799589 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -13,6 +13,8 @@ #define EXTENT_DEFRAG (1 << 6) #define EXTENT_DEFRAG_DONE (1 << 7) #define EXTENT_BUFFER_FILLED (1 << 8) +#define EXTENT_ORDERED (1 << 9) +#define EXTENT_ORDERED_METADATA (1 << 10) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) /* @@ -42,7 +44,7 @@ struct extent_io_ops { int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end, struct extent_state *state); int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end, - struct extent_state *state); + struct extent_state *state, int uptodate); int (*set_bit_hook)(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits); int (*clear_bit_hook)(struct inode *inode, u64 start, u64 end, @@ -131,6 +133,8 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, int filled); int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask); +int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int wake, int delete, gfp_t mask); int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, int bits, gfp_t mask); int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, @@ -141,8 +145,14 @@ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); +int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); +int clear_extent_ordered_metadata(struct extent_io_tree *tree, u64 start, + u64 end, gfp_t mask); int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask); +int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end, + gfp_t mask); int find_first_extent_bit(struct extent_io_tree *tree, u64 start, u64 *start_ret, u64 *end_ret, int bits); struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, @@ -209,6 +219,8 @@ void memset_extent_buffer(struct extent_buffer *eb, char c, unsigned long start, unsigned long len); int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, struct extent_buffer *eb); +int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end); +int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits); int clear_extent_buffer_dirty(struct extent_io_tree *tree, struct extent_buffer *eb); int set_extent_buffer_dirty(struct extent_io_tree *tree, diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index f5a04eb9a2a..81123277c2b 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -206,10 +206,11 @@ int add_extent_mapping(struct extent_map_tree *tree, struct extent_map *merge = NULL; struct rb_node *rb; + BUG_ON(spin_trylock(&tree->lock)); rb = tree_insert(&tree->map, em->start, &em->rb_node); if (rb) { - merge = rb_entry(rb, struct extent_map, rb_node); ret = -EEXIST; + free_extent_map(merge); goto out; } atomic_inc(&em->refs); @@ -268,6 +269,7 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, struct rb_node *next = NULL; u64 end = range_end(start, len); + BUG_ON(spin_trylock(&tree->lock)); em = tree->last; if (em && end > em->start && start < extent_map_end(em)) goto found; @@ -318,6 +320,7 @@ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) { int ret = 0; + BUG_ON(spin_trylock(&tree->lock)); rb_erase(&em->rb_node, &tree->map); em->in_tree = 0; if (tree->last == em) diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index f537eb43c2c..345caf8ff51 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -135,26 +135,37 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, } int btrfs_csum_one_bio(struct btrfs_root *root, - struct bio *bio, char **sums_ret) + struct bio *bio, struct btrfs_ordered_sum **sums_ret) { - u32 *sums; + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; char *data; struct bio_vec *bvec = bio->bi_io_vec; int bio_index = 0; - sums = kmalloc(bio->bi_vcnt * BTRFS_CRC32_SIZE, GFP_NOFS); + WARN_ON(bio->bi_vcnt <= 0); + sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS); if (!sums) return -ENOMEM; - *sums_ret = (char *)sums; + *sums_ret = sums; + sector_sum = &sums->sums; + sums->file_offset = page_offset(bvec->bv_page); + sums->len = bio->bi_size; + INIT_LIST_HEAD(&sums->list); while(bio_index < bio->bi_vcnt) { data = kmap_atomic(bvec->bv_page, KM_USER0); - *sums = ~(u32)0; - *sums = btrfs_csum_data(root, data + bvec->bv_offset, - *sums, bvec->bv_len); + sector_sum->sum = ~(u32)0; + sector_sum->sum = btrfs_csum_data(root, + data + bvec->bv_offset, + sector_sum->sum, + bvec->bv_len); kunmap_atomic(data, KM_USER0); - btrfs_csum_final(*sums, (char *)sums); - sums++; + btrfs_csum_final(sector_sum->sum, + (char *)§or_sum->sum); + sector_sum->offset = page_offset(bvec->bv_page) + + bvec->bv_offset; + sector_sum++; bio_index++; bvec++; } @@ -163,7 +174,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, - struct bio *bio, char *sums) + struct btrfs_ordered_sum *sums) { u64 objectid = inode->i_ino; u64 offset; @@ -171,17 +182,16 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_key file_key; struct btrfs_key found_key; u64 next_offset; + u64 total_bytes = 0; int found_next; struct btrfs_path *path; struct btrfs_csum_item *item; struct btrfs_csum_item *item_end; struct extent_buffer *leaf = NULL; u64 csum_offset; - u32 *sums32 = (u32 *)sums; + struct btrfs_sector_sum *sector_sum; u32 nritems; u32 ins_size; - int bio_index = 0; - struct bio_vec *bvec = bio->bi_io_vec; char *eb_map; char *eb_token; unsigned long map_len; @@ -189,10 +199,11 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); + sector_sum = &sums->sums; again: next_offset = (u64)-1; found_next = 0; - offset = page_offset(bvec->bv_page) + bvec->bv_offset; + offset = sector_sum->offset; file_key.objectid = objectid; file_key.offset = offset; btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); @@ -303,7 +314,7 @@ found: item_end = (struct btrfs_csum_item *)((unsigned char *)item_end + btrfs_item_size_nr(leaf, path->slots[0])); eb_token = NULL; -next_bvec: +next_sector: if (!eb_token || (unsigned long)item + BTRFS_CRC32_SIZE >= map_start + map_len) { @@ -321,21 +332,20 @@ next_bvec: } if (eb_token) { memcpy(eb_token + ((unsigned long)item & (PAGE_CACHE_SIZE - 1)), - sums32, BTRFS_CRC32_SIZE); + §or_sum->sum, BTRFS_CRC32_SIZE); } else { - write_extent_buffer(leaf, sums32, (unsigned long)item, - BTRFS_CRC32_SIZE); + write_extent_buffer(leaf, §or_sum->sum, + (unsigned long)item, BTRFS_CRC32_SIZE); } - bio_index++; - bvec++; - sums32++; - if (bio_index < bio->bi_vcnt) { + total_bytes += root->sectorsize; + sector_sum++; + if (total_bytes < sums->len) { item = (struct btrfs_csum_item *)((char *)item + BTRFS_CRC32_SIZE); if (item < item_end && offset + PAGE_CACHE_SIZE == - page_offset(bvec->bv_page)) { - offset = page_offset(bvec->bv_page); - goto next_bvec; + sector_sum->offset) { + offset = sector_sum->offset; + goto next_sector; } } if (eb_token) { @@ -343,7 +353,7 @@ next_bvec: eb_token = NULL; } btrfs_mark_buffer_dirty(path->nodes[0]); - if (bio_index < bio->bi_vcnt) { + if (total_bytes < sums->len) { btrfs_release_path(root, path); goto again; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 8037792f878..12e765f7e0d 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -34,7 +34,6 @@ #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" -#include "ordered-data.h" #include "ioctl.h" #include "print-tree.h" #include "compat.h" @@ -273,7 +272,9 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 mask = root->sectorsize - 1; last_pos_in_file = (isize + mask) & ~mask; hole_size = (start_pos - last_pos_in_file + mask) & ~mask; - if (last_pos_in_file < start_pos) { + if (hole_size > 0) { + btrfs_wait_ordered_range(inode, last_pos_in_file, + last_pos_in_file + hole_size); err = btrfs_drop_extents(trans, root, inode, last_pos_in_file, last_pos_in_file + hole_size, @@ -303,19 +304,17 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inline_size > root->fs_info->max_inline || (inline_size & (root->sectorsize -1)) == 0 || inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { - u64 last_end; - + /* check for reserved extents on each page, we don't want + * to reset the delalloc bit on things that already have + * extents reserved. + */ + set_extent_delalloc(io_tree, start_pos, + end_of_last_block, GFP_NOFS); for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; SetPageUptodate(p); set_page_dirty(p); } - last_end = (u64)(pages[num_pages -1]->index) << - PAGE_CACHE_SHIFT; - last_end += PAGE_CACHE_SIZE - 1; - set_extent_delalloc(io_tree, start_pos, end_of_last_block, - GFP_NOFS); - btrfs_add_ordered_inode(inode); } else { u64 aligned_end; /* step one, delete the existing extents in this range */ @@ -350,10 +349,13 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) struct extent_map *split = NULL; struct extent_map *split2 = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_map *tmp; u64 len = end - start + 1; + u64 next_start; int ret; int testend = 1; + WARN_ON(end < start); if (end == (u64)-1) { len = (u64)-1; testend = 0; @@ -370,6 +372,8 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) spin_unlock(&em_tree->lock); break; } + tmp = rb_entry(&em->rb_node, struct extent_map, rb_node); + next_start = tmp->start; remove_extent_mapping(em_tree, em); if (em->block_start < EXTENT_MAP_LAST_BYTE && @@ -778,37 +782,58 @@ static int prepare_pages(struct btrfs_root *root, struct file *file, struct inode *inode = fdentry(file)->d_inode; int err = 0; u64 start_pos; + u64 last_pos; start_pos = pos & ~((u64)root->sectorsize - 1); + last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT; memset(pages, 0, num_pages * sizeof(struct page *)); - +again: for (i = 0; i < num_pages; i++) { pages[i] = grab_cache_page(inode->i_mapping, index + i); if (!pages[i]) { err = -ENOMEM; BUG_ON(1); } -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - ClearPageDirty(pages[i]); -#else - cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); -#endif wait_on_page_writeback(pages[i]); - set_page_extent_mapped(pages[i]); - WARN_ON(!PageLocked(pages[i])); } if (start_pos < inode->i_size) { - u64 last_pos; - last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT; + struct btrfs_ordered_extent *ordered; lock_extent(&BTRFS_I(inode)->io_tree, start_pos, last_pos - 1, GFP_NOFS); + ordered = btrfs_lookup_first_ordered_extent(inode, last_pos -1); + if (ordered && + ordered->file_offset + ordered->len > start_pos && + ordered->file_offset < last_pos) { + btrfs_put_ordered_extent(ordered); + unlock_extent(&BTRFS_I(inode)->io_tree, + start_pos, last_pos - 1, GFP_NOFS); + for (i = 0; i < num_pages; i++) { + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + btrfs_wait_ordered_range(inode, start_pos, + last_pos - start_pos); + goto again; + } + if (ordered) + btrfs_put_ordered_extent(ordered); + clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos, last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC, GFP_NOFS); unlock_extent(&BTRFS_I(inode)->io_tree, start_pos, last_pos - 1, GFP_NOFS); } + for (i = 0; i < num_pages; i++) { +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) + ClearPageDirty(pages[i]); +#else + cancel_dirty_page(pages[i], PAGE_CACHE_SIZE); +#endif + set_page_extent_mapped(pages[i]); + WARN_ON(!PageLocked(pages[i])); + } return 0; } @@ -969,13 +994,11 @@ out_nolock: (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); } current->backing_dev_info = NULL; - btrfs_ordered_throttle(root, inode); return num_written ? num_written : err; } int btrfs_release_file(struct inode * inode, struct file * filp) { - btrfs_del_ordered_inode(inode, 0); if (filp->private_data) btrfs_ioctl_trans_end(filp); return 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d39433dfb2c..c5a62f0b959 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -43,6 +43,7 @@ #include "ioctl.h" #include "print-tree.h" #include "volumes.h" +#include "ordered-data.h" struct btrfs_iget_args { u64 ino; @@ -109,10 +110,11 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) u64 num_bytes; u64 cur_alloc_size; u64 blocksize = root->sectorsize; - u64 orig_start = start; u64 orig_num_bytes; struct btrfs_key ins; - int ret; + struct extent_map *em; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + int ret = 0; trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); @@ -120,33 +122,44 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) num_bytes = (end - start + blocksize) & ~(blocksize - 1); num_bytes = max(blocksize, num_bytes); - ret = btrfs_drop_extents(trans, root, inode, - start, start + num_bytes, start, &alloc_hint); orig_num_bytes = num_bytes; if (alloc_hint == EXTENT_MAP_INLINE) goto out; BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); + btrfs_drop_extent_cache(inode, start, start + num_bytes - 1); while(num_bytes > 0) { cur_alloc_size = min(num_bytes, root->fs_info->max_extent); - ret = btrfs_alloc_extent(trans, root, cur_alloc_size, - root->sectorsize, - root->root_key.objectid, - trans->transid, - inode->i_ino, start, 0, - alloc_hint, (u64)-1, &ins, 1); + ret = btrfs_reserve_extent(trans, root, cur_alloc_size, + root->sectorsize, 0, 0, + (u64)-1, &ins, 1); if (ret) { WARN_ON(1); goto out; } + em = alloc_extent_map(GFP_NOFS); + em->start = start; + em->len = ins.offset; + em->block_start = ins.objectid; + em->bdev = root->fs_info->fs_devices->latest_bdev; + while(1) { + spin_lock(&em_tree->lock); + ret = add_extent_mapping(em_tree, em); + spin_unlock(&em_tree->lock); + if (ret != -EEXIST) { + free_extent_map(em); + break; + } + btrfs_drop_extent_cache(inode, start, + start + ins.offset - 1); + } + cur_alloc_size = ins.offset; - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - start, ins.objectid, ins.offset, - ins.offset, 0); - inode->i_blocks += ins.offset >> 9; - btrfs_check_file(root, inode); + ret = btrfs_add_ordered_extent(inode, start, ins.objectid, + ins.offset); + BUG_ON(ret); if (num_bytes < cur_alloc_size) { printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes, cur_alloc_size); @@ -156,10 +169,6 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; } - btrfs_drop_extent_cache(inode, orig_start, - orig_start + orig_num_bytes - 1); - btrfs_add_ordered_inode(inode); - btrfs_update_inode(trans, root, inode); out: btrfs_end_transaction(trans, root); return ret; @@ -341,25 +350,15 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_trans_handle *trans; int ret = 0; - char *sums = NULL; + struct btrfs_ordered_sum *sums; ret = btrfs_csum_one_bio(root, bio, &sums); BUG_ON(ret); - trans = btrfs_start_transaction(root, 1); - - btrfs_set_trans_block_group(trans, inode); - mutex_lock(&BTRFS_I(inode)->csum_mutex); - btrfs_csum_file_blocks(trans, root, inode, bio, sums); - mutex_unlock(&BTRFS_I(inode)->csum_mutex); - - ret = btrfs_end_transaction(trans, root); + ret = btrfs_add_ordered_sum(inode, sums); BUG_ON(ret); - kfree(sums); - return btrfs_map_bio(root, rw, bio, mirror_num, 1); } @@ -369,14 +368,10 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; - if (!(rw & (1 << BIO_RW))) { - ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); - BUG_ON(ret); - goto mapit; - } + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); + BUG_ON(ret); - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) { + if (!(rw & (1 << BIO_RW))) { goto mapit; } @@ -387,6 +382,96 @@ mapit: return btrfs_map_bio(root, rw, bio, mirror_num, 0); } +static int add_pending_csums(struct btrfs_trans_handle *trans, + struct inode *inode, u64 file_offset, + struct list_head *list) +{ + struct list_head *cur; + struct btrfs_ordered_sum *sum; + + btrfs_set_trans_block_group(trans, inode); + while(!list_empty(list)) { + cur = list->next; + sum = list_entry(cur, struct btrfs_ordered_sum, list); + mutex_lock(&BTRFS_I(inode)->csum_mutex); + btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, + inode, sum); + mutex_unlock(&BTRFS_I(inode)->csum_mutex); + list_del(&sum->list); + kfree(sum); + } + return 0; +} + +int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, + struct extent_state *state, int uptodate) +{ + struct inode *inode = page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + struct btrfs_ordered_extent *ordered_extent; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + u64 alloc_hint = 0; + struct list_head list; + struct btrfs_key ins; + int ret; + + ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); + if (!ret) { + return 0; + } + + trans = btrfs_start_transaction(root, 1); + + ordered_extent = btrfs_lookup_ordered_extent(inode, start); + BUG_ON(!ordered_extent); + + lock_extent(io_tree, ordered_extent->file_offset, + ordered_extent->file_offset + ordered_extent->len - 1, + GFP_NOFS); + + INIT_LIST_HEAD(&list); + + ins.objectid = ordered_extent->start; + ins.offset = ordered_extent->len; + ins.type = BTRFS_EXTENT_ITEM_KEY; + ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid, + trans->transid, inode->i_ino, + ordered_extent->file_offset, &ins); + BUG_ON(ret); + ret = btrfs_drop_extents(trans, root, inode, + ordered_extent->file_offset, + ordered_extent->file_offset + + ordered_extent->len, + ordered_extent->file_offset, &alloc_hint); + BUG_ON(ret); + ret = btrfs_insert_file_extent(trans, root, inode->i_ino, + ordered_extent->file_offset, + ordered_extent->start, + ordered_extent->len, + ordered_extent->len, 0); + BUG_ON(ret); + btrfs_drop_extent_cache(inode, ordered_extent->file_offset, + ordered_extent->file_offset + + ordered_extent->len - 1); + inode->i_blocks += ordered_extent->len >> 9; + unlock_extent(io_tree, ordered_extent->file_offset, + ordered_extent->file_offset + ordered_extent->len - 1, + GFP_NOFS); + add_pending_csums(trans, inode, ordered_extent->file_offset, + &ordered_extent->list); + + btrfs_remove_ordered_extent(inode, ordered_extent); + /* once for us */ + btrfs_put_ordered_extent(ordered_extent); + /* once for the tree */ + btrfs_put_ordered_extent(ordered_extent); + + btrfs_update_inode(trans, root, inode); + btrfs_end_transaction(trans, root); + return 0; +} + int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) { int ret = 0; @@ -409,7 +494,8 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) if (ret == -ENOENT || ret == -EFBIG) ret = 0; csum = 0; - printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start); + printk("no csum found for inode %lu start %Lu\n", inode->i_ino, + start); goto out; } read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, @@ -833,7 +919,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) { struct btrfs_root *root; struct btrfs_trans_handle *trans; - struct inode *inode = dentry->d_inode; int ret; unsigned long nr = 0; @@ -849,14 +934,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) ret = btrfs_unlink_trans(trans, root, dir, dentry); nr = trans->blocks_used; - if (inode->i_nlink == 0) { - /* if the inode isn't linked anywhere, - * we don't need to worry about - * data=ordered - */ - btrfs_del_ordered_inode(inode, 1); - } - btrfs_end_transaction_throttle(trans, root); fail: btrfs_btree_balance_dirty(root, nr); @@ -931,6 +1008,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int extent_type = -1; u64 mask = root->sectorsize - 1; + btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1); path = btrfs_alloc_path(); path->reada = -1; @@ -1117,34 +1195,6 @@ error: return ret; } -static int btrfs_cow_one_page(struct inode *inode, struct page *page, - size_t zero_start) -{ - char *kaddr; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; - u64 page_end = page_start + PAGE_CACHE_SIZE - 1; - int ret = 0; - - WARN_ON(!PageLocked(page)); - set_page_extent_mapped(page); - - lock_extent(io_tree, page_start, page_end, GFP_NOFS); - set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, - page_end, GFP_NOFS); - - if (zero_start != PAGE_CACHE_SIZE) { - kaddr = kmap(page); - memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); - flush_dcache_page(page); - kunmap(page); - } - set_page_dirty(page); - unlock_extent(io_tree, page_start, page_end, GFP_NOFS); - - return ret; -} - /* * taken from block_truncate_page, but does cow as it zeros out * any bytes left in the last page in the file. @@ -1153,12 +1203,16 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from) { struct inode *inode = mapping->host; struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct btrfs_ordered_extent *ordered; + char *kaddr; u32 blocksize = root->sectorsize; pgoff_t index = from >> PAGE_CACHE_SHIFT; unsigned offset = from & (PAGE_CACHE_SIZE-1); struct page *page; int ret = 0; u64 page_start; + u64 page_end; if ((offset & (blocksize - 1)) == 0) goto out; @@ -1168,6 +1222,10 @@ again: page = grab_cache_page(mapping, index); if (!page) goto out; + + page_start = page_offset(page); + page_end = page_start + PAGE_CACHE_SIZE - 1; + if (!PageUptodate(page)) { ret = btrfs_readpage(NULL, page); lock_page(page); @@ -1181,10 +1239,32 @@ again: goto out; } } - - page_start = (u64)page->index << PAGE_CACHE_SHIFT; wait_on_page_writeback(page); - ret = btrfs_cow_one_page(inode, page, offset); + + lock_extent(io_tree, page_start, page_end, GFP_NOFS); + set_page_extent_mapped(page); + + ordered = btrfs_lookup_ordered_extent(inode, page_start); + if (ordered) { + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); + unlock_page(page); + page_cache_release(page); + btrfs_wait_ordered_extent(inode, ordered); + btrfs_put_ordered_extent(ordered); + goto again; + } + + set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, + page_end, GFP_NOFS); + ret = 0; + if (offset != PAGE_CACHE_SIZE) { + kaddr = kmap(page); + memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); + flush_dcache_page(page); + kunmap(page); + } + set_page_dirty(page); + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); unlock_page(page); page_cache_release(page); @@ -1222,8 +1302,9 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); - lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); hole_size = block_end - hole_start; + btrfs_wait_ordered_range(inode, hole_start, hole_size); + lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -1258,6 +1339,7 @@ void btrfs_delete_inode(struct inode *inode) unsigned long nr; int ret; + btrfs_wait_ordered_range(inode, 0, (u64)-1); truncate_inode_pages(&inode->i_data, 0); if (is_bad_inode(inode)) { goto no_delete; @@ -1403,7 +1485,6 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); - atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); return 0; } @@ -1705,7 +1786,6 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); - atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->root = root; @@ -1930,7 +2010,6 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; - atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -2066,64 +2145,18 @@ out_unlock: static int merge_extent_mapping(struct extent_map_tree *em_tree, struct extent_map *existing, - struct extent_map *em) + struct extent_map *em, + u64 map_start, u64 map_len) { u64 start_diff; - u64 new_end; - int ret = 0; - int real_blocks = existing->block_start < EXTENT_MAP_LAST_BYTE; - - if (real_blocks && em->block_start >= EXTENT_MAP_LAST_BYTE) - goto invalid; - - if (!real_blocks && em->block_start != existing->block_start) - goto invalid; - - new_end = max(existing->start + existing->len, em->start + em->len); - - if (existing->start >= em->start) { - if (em->start + em->len < existing->start) - goto invalid; - start_diff = existing->start - em->start; - if (real_blocks && em->block_start + start_diff != - existing->block_start) - goto invalid; - - em->len = new_end - em->start; - - remove_extent_mapping(em_tree, existing); - /* free for the tree */ - free_extent_map(existing); - ret = add_extent_mapping(em_tree, em); - - } else if (em->start > existing->start) { - - if (existing->start + existing->len < em->start) - goto invalid; - - start_diff = em->start - existing->start; - if (real_blocks && existing->block_start + start_diff != - em->block_start) - goto invalid; - - remove_extent_mapping(em_tree, existing); - em->block_start = existing->block_start; - em->start = existing->start; - em->len = new_end - existing->start; - free_extent_map(existing); - - ret = add_extent_mapping(em_tree, em); - } else { - goto invalid; - } - return ret; - -invalid: - printk("invalid extent map merge [%Lu %Lu %Lu] [%Lu %Lu %Lu]\n", - existing->start, existing->len, existing->block_start, - em->start, em->len, em->block_start); - return -EIO; + BUG_ON(map_start < em->start || map_start >= extent_map_end(em)); + start_diff = map_start - em->start; + em->start = map_start; + em->len = map_len; + if (em->block_start < EXTENT_MAP_LAST_BYTE) + em->block_start += start_diff; + return add_extent_mapping(em_tree, em); } struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, @@ -2170,10 +2203,9 @@ again: err = -ENOMEM; goto out; } - + em->bdev = root->fs_info->fs_devices->latest_bdev; em->start = EXTENT_MAP_HOLE; em->len = (u64)-1; - em->bdev = root->fs_info->fs_devices->latest_bdev; ret = btrfs_lookup_file_extent(trans, root, path, objectid, start, trans != NULL); if (ret < 0) { @@ -2314,6 +2346,9 @@ insert: */ if (ret == -EEXIST) { struct extent_map *existing; + + ret = 0; + existing = lookup_extent_mapping(em_tree, start, len); if (existing && (existing->start > start || existing->start + existing->len <= start)) { @@ -2325,7 +2360,8 @@ insert: em->len); if (existing) { err = merge_extent_mapping(em_tree, existing, - em); + em, start, + root->sectorsize); free_extent_map(existing); if (err) { free_extent_map(em); @@ -2341,6 +2377,7 @@ insert: } else { free_extent_map(em); em = existing; + err = 0; } } spin_unlock(&em_tree->lock); @@ -2348,8 +2385,9 @@ out: btrfs_free_path(path); if (trans) { ret = btrfs_end_transaction(trans, root); - if (!err) + if (!err) { err = ret; + } } if (err) { free_extent_map(em); @@ -2474,8 +2512,7 @@ btrfs_readpages(struct file *file, struct address_space *mapping, return extent_readpages(tree, mapping, pages, nr_pages, btrfs_get_extent); } - -static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) +static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags) { struct extent_io_tree *tree; struct extent_map_tree *map; @@ -2493,15 +2530,54 @@ static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) return ret; } +static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) +{ + struct btrfs_ordered_extent *ordered; + + ordered = btrfs_lookup_ordered_extent(page->mapping->host, + page_offset(page)); + if (ordered) { + btrfs_put_ordered_extent(ordered); + return 0; + } + return __btrfs_releasepage(page, gfp_flags); +} + static void btrfs_invalidatepage(struct page *page, unsigned long offset) { struct extent_io_tree *tree; + struct btrfs_ordered_extent *ordered; + u64 page_start = page_offset(page); + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + wait_on_page_writeback(page); tree = &BTRFS_I(page->mapping->host)->io_tree; - extent_invalidatepage(tree, page, offset); - btrfs_releasepage(page, GFP_NOFS); + if (offset) { + btrfs_releasepage(page, GFP_NOFS); + return; + } + + lock_extent(tree, page_start, page_end, GFP_NOFS); + ordered = btrfs_lookup_ordered_extent(page->mapping->host, + page_offset(page)); + if (ordered) { + clear_extent_bit(tree, page_start, page_end, + EXTENT_DIRTY | EXTENT_DELALLOC | + EXTENT_LOCKED, 1, 0, GFP_NOFS); + btrfs_writepage_end_io_hook(page, page_start, + page_end, NULL, 1); + btrfs_put_ordered_extent(ordered); + lock_extent(tree, page_start, page_end, GFP_NOFS); + } + clear_extent_bit(tree, page_start, page_end, + EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC | + EXTENT_ORDERED, + 1, 1, GFP_NOFS); + __btrfs_releasepage(page, GFP_NOFS); + if (PagePrivate(page)) { - invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); + invalidate_extent_lru(tree, page_offset(page), + PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); @@ -2527,35 +2603,63 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) { struct inode *inode = fdentry(vma->vm_file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; - unsigned long end; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct btrfs_ordered_extent *ordered; + char *kaddr; + unsigned long zero_start; loff_t size; int ret; u64 page_start; + u64 page_end; ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); if (ret) goto out; ret = -EINVAL; - +again: lock_page(page); - wait_on_page_writeback(page); size = i_size_read(inode); - page_start = (u64)page->index << PAGE_CACHE_SHIFT; + page_start = page_offset(page); + page_end = page_start + PAGE_CACHE_SIZE - 1; if ((page->mapping != inode->i_mapping) || - (page_start > size)) { + (page_start >= size)) { /* page got truncated out from underneath us */ goto out_unlock; } + wait_on_page_writeback(page); + + lock_extent(io_tree, page_start, page_end, GFP_NOFS); + set_page_extent_mapped(page); + + ordered = btrfs_lookup_ordered_extent(inode, page_start); + if (ordered) { + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); + unlock_page(page); + btrfs_wait_ordered_extent(inode, ordered); + btrfs_put_ordered_extent(ordered); + goto again; + } + + set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, + page_end, GFP_NOFS); + ret = 0; /* page is wholly or partially inside EOF */ if (page_start + PAGE_CACHE_SIZE > size) - end = size & ~PAGE_CACHE_MASK; + zero_start = size & ~PAGE_CACHE_MASK; else - end = PAGE_CACHE_SIZE; + zero_start = PAGE_CACHE_SIZE; - ret = btrfs_cow_one_page(inode, page, end); + if (zero_start != PAGE_CACHE_SIZE) { + kaddr = kmap(page); + memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); + flush_dcache_page(page); + kunmap(page); + } + set_page_dirty(page); + unlock_extent(io_tree, page_start, page_end, GFP_NOFS); out_unlock: unlock_page(page); @@ -2662,15 +2766,28 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) if (!ei) return NULL; ei->last_trans = 0; - ei->ordered_trans = 0; + btrfs_ordered_inode_tree_init(&ei->ordered_tree); return &ei->vfs_inode; } void btrfs_destroy_inode(struct inode *inode) { + struct btrfs_ordered_extent *ordered; WARN_ON(!list_empty(&inode->i_dentry)); WARN_ON(inode->i_data.nrpages); + while(1) { + ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); + if (!ordered) + break; + else { + printk("found ordered extent %Lu %Lu\n", + ordered->file_offset, ordered->len); + btrfs_remove_ordered_extent(inode, ordered); + btrfs_put_ordered_extent(ordered); + btrfs_put_ordered_extent(ordered); + } + } btrfs_drop_extent_cache(inode, 0, (u64)-1); kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } @@ -2869,7 +2986,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; - atomic_set(&BTRFS_I(inode)->ordered_writeback, 0); BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -2921,6 +3037,20 @@ out_fail: return err; } +static int btrfs_set_page_dirty(struct page *page) +{ + struct inode *inode = page->mapping->host; + u64 page_start = page_offset(page); + u64 page_end = page_start + PAGE_CACHE_SIZE - 1; + + if (!test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end, + EXTENT_DELALLOC, 0)) { +printk("inode %lu page %Lu not delalloc\n", inode->i_ino, page_offset(page)); +WARN_ON(1); + } + return __set_page_dirty_nobuffers(page); +} + static int btrfs_permission(struct inode *inode, int mask, struct nameidata *nd) { @@ -2967,6 +3097,7 @@ static struct extent_io_ops btrfs_extent_io_ops = { .merge_bio_hook = btrfs_merge_bio_hook, .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, + .writepage_end_io_hook = btrfs_writepage_end_io_hook, .readpage_io_failed_hook = btrfs_io_failed_hook, .set_bit_hook = btrfs_set_bit_hook, .clear_bit_hook = btrfs_clear_bit_hook, @@ -2982,7 +3113,7 @@ static struct address_space_operations btrfs_aops = { .direct_IO = btrfs_direct_IO, .invalidatepage = btrfs_invalidatepage, .releasepage = btrfs_releasepage, - .set_page_dirty = __set_page_dirty_nobuffers, + .set_page_dirty = btrfs_set_page_dirty, }; static struct address_space_operations btrfs_symlink_aops = { diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 254da822566..6513270f054 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -22,48 +22,30 @@ #include "ctree.h" #include "transaction.h" #include "btrfs_inode.h" +#include "extent_io.h" -struct tree_entry { - u64 root_objectid; - u64 objectid; - struct inode *inode; - struct rb_node rb_node; -}; -/* - * returns > 0 if entry passed (root, objectid) is > entry, - * < 0 if (root, objectid) < entry and zero if they are equal - */ -static int comp_entry(struct tree_entry *entry, u64 root_objectid, - u64 objectid) +static u64 entry_end(struct btrfs_ordered_extent *entry) { - if (root_objectid < entry->root_objectid) - return -1; - if (root_objectid > entry->root_objectid) - return 1; - if (objectid < entry->objectid) - return -1; - if (objectid > entry->objectid) - return 1; - return 0; + if (entry->file_offset + entry->len < entry->file_offset) + return (u64)-1; + return entry->file_offset + entry->len; } -static struct rb_node *tree_insert(struct rb_root *root, u64 root_objectid, - u64 objectid, struct rb_node *node) +static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, + struct rb_node *node) { struct rb_node ** p = &root->rb_node; struct rb_node * parent = NULL; - struct tree_entry *entry; - int comp; + struct btrfs_ordered_extent *entry; while(*p) { parent = *p; - entry = rb_entry(parent, struct tree_entry, rb_node); + entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node); - comp = comp_entry(entry, root_objectid, objectid); - if (comp < 0) + if (file_offset < entry->file_offset) p = &(*p)->rb_left; - else if (comp > 0) + else if (file_offset >= entry_end(entry)) p = &(*p)->rb_right; else return parent; @@ -74,24 +56,23 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 root_objectid, return NULL; } -static struct rb_node *__tree_search(struct rb_root *root, u64 root_objectid, - u64 objectid, struct rb_node **prev_ret) +static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, + struct rb_node **prev_ret) { struct rb_node * n = root->rb_node; struct rb_node *prev = NULL; - struct tree_entry *entry; - struct tree_entry *prev_entry = NULL; - int comp; + struct rb_node *test; + struct btrfs_ordered_extent *entry; + struct btrfs_ordered_extent *prev_entry = NULL; while(n) { - entry = rb_entry(n, struct tree_entry, rb_node); + entry = rb_entry(n, struct btrfs_ordered_extent, rb_node); prev = n; prev_entry = entry; - comp = comp_entry(entry, root_objectid, objectid); - if (comp < 0) + if (file_offset < entry->file_offset) n = n->rb_left; - else if (comp > 0) + else if (file_offset >= entry_end(entry)) n = n->rb_right; else return n; @@ -99,195 +80,329 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 root_objectid, if (!prev_ret) return NULL; - while(prev && comp_entry(prev_entry, root_objectid, objectid) >= 0) { - prev = rb_next(prev); - prev_entry = rb_entry(prev, struct tree_entry, rb_node); + while(prev && file_offset >= entry_end(prev_entry)) { + test = rb_next(prev); + if (!test) + break; + prev_entry = rb_entry(test, struct btrfs_ordered_extent, + rb_node); + if (file_offset < entry_end(prev_entry)) + break; + + prev = test; + } + if (prev) + prev_entry = rb_entry(prev, struct btrfs_ordered_extent, + rb_node); + while(prev && file_offset < entry_end(prev_entry)) { + test = rb_prev(prev); + if (!test) + break; + prev_entry = rb_entry(test, struct btrfs_ordered_extent, + rb_node); + prev = test; } *prev_ret = prev; return NULL; } -static inline struct rb_node *tree_search(struct rb_root *root, - u64 root_objectid, u64 objectid) +static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset) +{ + if (file_offset < entry->file_offset || + entry->file_offset + entry->len <= file_offset) + return 0; + return 1; +} + +static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, + u64 file_offset) { + struct rb_root *root = &tree->tree; struct rb_node *prev; struct rb_node *ret; - ret = __tree_search(root, root_objectid, objectid, &prev); + struct btrfs_ordered_extent *entry; + + if (tree->last) { + entry = rb_entry(tree->last, struct btrfs_ordered_extent, + rb_node); + if (offset_in_entry(entry, file_offset)) + return tree->last; + } + ret = __tree_search(root, file_offset, &prev); if (!ret) - return prev; + ret = prev; + if (ret) + tree->last = ret; return ret; } -int btrfs_add_ordered_inode(struct inode *inode) +int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, + u64 start, u64 len) { - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 root_objectid = root->root_key.objectid; - u64 transid = root->fs_info->running_transaction->transid; - struct tree_entry *entry; - struct rb_node *node; struct btrfs_ordered_inode_tree *tree; + struct rb_node *node; + struct btrfs_ordered_extent *entry; - if (transid <= BTRFS_I(inode)->ordered_trans) - return 0; - - tree = &root->fs_info->running_transaction->ordered_inode_tree; - - read_lock(&tree->lock); - node = __tree_search(&tree->tree, root_objectid, inode->i_ino, NULL); - read_unlock(&tree->lock); - if (node) { - return 0; - } - - entry = kmalloc(sizeof(*entry), GFP_NOFS); + tree = &BTRFS_I(inode)->ordered_tree; + entry = kzalloc(sizeof(*entry), GFP_NOFS); if (!entry) return -ENOMEM; - write_lock(&tree->lock); - entry->objectid = inode->i_ino; - entry->root_objectid = root_objectid; + mutex_lock(&tree->mutex); + entry->file_offset = file_offset; + entry->start = start; + entry->len = len; entry->inode = inode; + /* one ref for the tree */ + atomic_set(&entry->refs, 1); + init_waitqueue_head(&entry->wait); + INIT_LIST_HEAD(&entry->list); - node = tree_insert(&tree->tree, root_objectid, - inode->i_ino, &entry->rb_node); - - BTRFS_I(inode)->ordered_trans = transid; - if (!node) - igrab(inode); - - write_unlock(&tree->lock); + node = tree_insert(&tree->tree, file_offset, + &entry->rb_node); + if (node) { + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + atomic_inc(&entry->refs); + } + set_extent_ordered(&BTRFS_I(inode)->io_tree, file_offset, + entry_end(entry) - 1, GFP_NOFS); - if (node) - kfree(entry); + set_bit(BTRFS_ORDERED_START, &entry->flags); + mutex_unlock(&tree->mutex); + BUG_ON(node); return 0; } -int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, - u64 *root_objectid, u64 *objectid, - struct inode **inode) +int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum) { - struct tree_entry *entry; + struct btrfs_ordered_inode_tree *tree; struct rb_node *node; + struct btrfs_ordered_extent *entry; - write_lock(&tree->lock); - node = tree_search(&tree->tree, *root_objectid, *objectid); + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + node = tree_search(tree, sum->file_offset); if (!node) { - write_unlock(&tree->lock); - return 0; +search_fail: +printk("add ordered sum failed to find a node for inode %lu offset %Lu\n", inode->i_ino, sum->file_offset); + node = rb_first(&tree->tree); + while(node) { + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + printk("entry %Lu %Lu %Lu\n", entry->file_offset, entry->file_offset + entry->len, entry->start); + node = rb_next(node); + } + BUG(); } - entry = rb_entry(node, struct tree_entry, rb_node); + BUG_ON(!node); - while(comp_entry(entry, *root_objectid, *objectid) >= 0) { - node = rb_next(node); - if (!node) - break; - entry = rb_entry(node, struct tree_entry, rb_node); - } - if (!node) { - write_unlock(&tree->lock); - return 0; + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + if (!offset_in_entry(entry, sum->file_offset)) { + goto search_fail; } - *root_objectid = entry->root_objectid; - *inode = entry->inode; - atomic_inc(&entry->inode->i_count); - *objectid = entry->objectid; - write_unlock(&tree->lock); - return 1; + list_add_tail(&sum->list, &entry->list); + mutex_unlock(&tree->mutex); + return 0; } -int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, - u64 *root_objectid, u64 *objectid, - struct inode **inode) +int btrfs_dec_test_ordered_pending(struct inode *inode, + u64 file_offset, u64 io_size) { - struct tree_entry *entry; + struct btrfs_ordered_inode_tree *tree; struct rb_node *node; - - write_lock(&tree->lock); - node = tree_search(&tree->tree, *root_objectid, *objectid); + struct btrfs_ordered_extent *entry; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + int ret; + + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + clear_extent_ordered(io_tree, file_offset, file_offset + io_size - 1, + GFP_NOFS); + node = tree_search(tree, file_offset); if (!node) { - write_unlock(&tree->lock); - return 0; + ret = 1; + goto out; } - entry = rb_entry(node, struct tree_entry, rb_node); - while(comp_entry(entry, *root_objectid, *objectid) >= 0) { - node = rb_next(node); - if (!node) - break; - entry = rb_entry(node, struct tree_entry, rb_node); + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + if (!offset_in_entry(entry, file_offset)) { + ret = 1; + goto out; } - if (!node) { - write_unlock(&tree->lock); - return 0; + + ret = test_range_bit(io_tree, entry->file_offset, + entry->file_offset + entry->len - 1, + EXTENT_ORDERED, 0); + if (!test_bit(BTRFS_ORDERED_START, &entry->flags)) { +printk("inode %lu not ready yet for extent %Lu %Lu\n", inode->i_ino, entry->file_offset, entry_end(entry)); } + if (ret == 0) + ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags); +out: + mutex_unlock(&tree->mutex); + return ret == 0; +} - *root_objectid = entry->root_objectid; - *objectid = entry->objectid; - *inode = entry->inode; - atomic_inc(&entry->inode->i_count); - rb_erase(node, &tree->tree); - write_unlock(&tree->lock); - kfree(entry); - return 1; +int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) +{ + if (atomic_dec_and_test(&entry->refs)) + kfree(entry); + return 0; } -static void __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree, - struct inode *inode, - u64 root_objectid, u64 objectid) +int btrfs_remove_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry) { - struct tree_entry *entry; + struct btrfs_ordered_inode_tree *tree; struct rb_node *node; - struct rb_node *prev; - write_lock(&tree->lock); - node = __tree_search(&tree->tree, root_objectid, objectid, &prev); - if (!node) { - write_unlock(&tree->lock); - return; - } + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + node = &entry->rb_node; rb_erase(node, &tree->tree); - BTRFS_I(inode)->ordered_trans = 0; - write_unlock(&tree->lock); - atomic_dec(&inode->i_count); - entry = rb_entry(node, struct tree_entry, rb_node); - kfree(entry); - return; + tree->last = NULL; + set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags); + mutex_unlock(&tree->mutex); + wake_up(&entry->wait); + return 0; } -void btrfs_del_ordered_inode(struct inode *inode, int force) +void btrfs_wait_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry) { - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 root_objectid = root->root_key.objectid; + u64 start = entry->file_offset; + u64 end = start + entry->len - 1; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) + do_sync_file_range(file, start, end, SYNC_FILE_RANGE_WRITE); +#else + do_sync_mapping_range(inode->i_mapping, start, end, + SYNC_FILE_RANGE_WRITE); +#endif + wait_event(entry->wait, + test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); +} - if (!BTRFS_I(inode)->ordered_trans) { - return; - } +static void btrfs_start_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry, int wait) +{ + u64 start = entry->file_offset; + u64 end = start + entry->len - 1; - if (!force && (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) || - mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))) - return; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) + do_sync_file_range(file, start, end, SYNC_FILE_RANGE_WRITE); +#else + do_sync_mapping_range(inode->i_mapping, start, end, + SYNC_FILE_RANGE_WRITE); +#endif + if (wait) + wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, + &entry->flags)); +} - spin_lock(&root->fs_info->new_trans_lock); - if (root->fs_info->running_transaction) { - struct btrfs_ordered_inode_tree *tree; - tree = &root->fs_info->running_transaction->ordered_inode_tree; - __btrfs_del_ordered_inode(tree, inode, root_objectid, - inode->i_ino); +void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) +{ + u64 end; + struct btrfs_ordered_extent *ordered; + int found; + int should_wait = 0; + +again: + if (start + len < start) + end = (u64)-1; + else + end = start + len - 1; + found = 0; + while(1) { + ordered = btrfs_lookup_first_ordered_extent(inode, end); + if (!ordered) { + break; + } + if (ordered->file_offset >= start + len) { + btrfs_put_ordered_extent(ordered); + break; + } + if (ordered->file_offset + ordered->len < start) { + btrfs_put_ordered_extent(ordered); + break; + } + btrfs_start_ordered_extent(inode, ordered, should_wait); + found++; + end = ordered->file_offset; + btrfs_put_ordered_extent(ordered); + if (end == 0) + break; + end--; + } + if (should_wait && found) { + should_wait = 0; + goto again; } - spin_unlock(&root->fs_info->new_trans_lock); } -int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode) +int btrfs_add_ordered_pending(struct inode *inode, + struct btrfs_ordered_extent *ordered, + u64 start, u64 len) { - struct btrfs_transaction *cur = root->fs_info->running_transaction; - while(cur == root->fs_info->running_transaction && - atomic_read(&BTRFS_I(inode)->ordered_writeback)) { -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) - congestion_wait(WRITE, HZ/20); -#else - blk_congestion_wait(WRITE, HZ/20); -#endif - } + WARN_ON(1); return 0; +#if 0 + int ret; + struct btrfs_ordered_inode_tree *tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + if (test_bit(BTRFS_ORDERED_IO_DONE, &ordered->flags)) { + ret = -EAGAIN; + goto out; + } + set_extent_ordered(io_tree, start, start + len - 1, GFP_NOFS); + ret = 0; +out: + mutex_unlock(&tree->mutex); + return ret; +#endif +} + +struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, + u64 file_offset) +{ + struct btrfs_ordered_inode_tree *tree; + struct rb_node *node; + struct btrfs_ordered_extent *entry = NULL; + + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + node = tree_search(tree, file_offset); + if (!node) + goto out; + + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + if (!offset_in_entry(entry, file_offset)) + entry = NULL; + if (entry) + atomic_inc(&entry->refs); +out: + mutex_unlock(&tree->mutex); + return entry; +} + +struct btrfs_ordered_extent * +btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset) +{ + struct btrfs_ordered_inode_tree *tree; + struct rb_node *node; + struct btrfs_ordered_extent *entry = NULL; + + tree = &BTRFS_I(inode)->ordered_tree; + mutex_lock(&tree->mutex); + node = tree_search(tree, file_offset); + if (!node) + goto out; + + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + atomic_inc(&entry->refs); +out: + mutex_unlock(&tree->mutex); + return entry; } diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 4fa78736423..33292c5fe90 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -20,24 +20,73 @@ #define __BTRFS_ORDERED_DATA__ struct btrfs_ordered_inode_tree { - rwlock_t lock; + struct mutex mutex; struct rb_root tree; + struct rb_node *last; }; +struct btrfs_sector_sum { + u64 offset; + u32 sum; +}; + +struct btrfs_ordered_sum { + u64 file_offset; + u64 len; + struct list_head list; + struct btrfs_sector_sum sums; +}; + +/* bits for the flags field */ +#define BTRFS_ORDERED_IO_DONE 0 /* set when all the pages are written */ +#define BTRFS_ORDERED_COMPLETE 1 /* set when removed from the tree */ +#define BTRFS_ORDERED_START 2 /* set when tree setup */ + +struct btrfs_ordered_extent { + u64 file_offset; + u64 start; + u64 len; + unsigned long flags; + atomic_t refs; + struct list_head list; + struct inode *inode; + wait_queue_head_t wait; + struct rb_node rb_node; +}; + + +static inline int btrfs_ordered_sum_size(struct btrfs_root *root, u64 bytes) +{ + unsigned long num_sectors = (bytes + root->sectorsize - 1) / + root->sectorsize; + return sizeof(struct btrfs_ordered_sum) + + num_sectors * sizeof(struct btrfs_sector_sum); +} + static inline void btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t) { - rwlock_init(&t->lock); + mutex_init(&t->mutex); t->tree.rb_node = NULL; + t->last = NULL; } -int btrfs_add_ordered_inode(struct inode *inode); -int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, - u64 *root_objectid, u64 *objectid, - struct inode **inode); -int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, - u64 *root_objectid, u64 *objectid, - struct inode **inode); -void btrfs_del_ordered_inode(struct inode *inode, int force); -int btrfs_ordered_throttle(struct btrfs_root *root, struct inode *inode); +int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry); +int btrfs_remove_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry); +int btrfs_dec_test_ordered_pending(struct inode *inode, + u64 file_offset, u64 io_size); +int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, + u64 start, u64 len); +int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum); +struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, + u64 file_offset); +void btrfs_wait_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry); +void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len); +struct btrfs_ordered_extent * +btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); +int btrfs_add_ordered_pending(struct inode *inode, + struct btrfs_ordered_extent *ordered, + u64 start, u64 len); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index a8a3cb03de5..86a5acc19ce 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -67,7 +67,6 @@ static noinline int join_transaction(struct btrfs_root *root) cur_trans->start_time = get_seconds(); INIT_LIST_HEAD(&cur_trans->pending_snapshots); list_add_tail(&cur_trans->list, &root->fs_info->trans_list); - btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree); extent_io_tree_init(&cur_trans->dirty_pages, root->fs_info->btree_inode->i_mapping, GFP_NOFS); @@ -158,10 +157,12 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, wake_up(&cur_trans->writer_wait); if (cur_trans->in_commit && throttle) { - int ret; + DEFINE_WAIT(wait); mutex_unlock(&root->fs_info->trans_mutex); - ret = wait_for_commit(root, cur_trans); - BUG_ON(ret); + prepare_to_wait(&root->fs_info->transaction_throttle, &wait, + TASK_UNINTERRUPTIBLE); + schedule(); + finish_wait(&root->fs_info->transaction_throttle, &wait); mutex_lock(&root->fs_info->trans_mutex); } @@ -486,58 +487,6 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, return ret; } -int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, - struct btrfs_root *root) -{ - struct btrfs_transaction *cur_trans = trans->transaction; - struct inode *inode; - u64 root_objectid = 0; - u64 objectid = 0; - int ret; - - atomic_inc(&root->fs_info->throttles); - while(1) { - ret = btrfs_find_first_ordered_inode( - &cur_trans->ordered_inode_tree, - &root_objectid, &objectid, &inode); - if (!ret) - break; - - mutex_unlock(&root->fs_info->trans_mutex); - - if (S_ISREG(inode->i_mode)) { - atomic_inc(&BTRFS_I(inode)->ordered_writeback); - filemap_fdatawrite(inode->i_mapping); - atomic_dec(&BTRFS_I(inode)->ordered_writeback); - } - iput(inode); - - mutex_lock(&root->fs_info->trans_mutex); - } - while(1) { - root_objectid = 0; - objectid = 0; - ret = btrfs_find_del_first_ordered_inode( - &cur_trans->ordered_inode_tree, - &root_objectid, &objectid, &inode); - if (!ret) - break; - mutex_unlock(&root->fs_info->trans_mutex); - - if (S_ISREG(inode->i_mode)) { - atomic_inc(&BTRFS_I(inode)->ordered_writeback); - filemap_write_and_wait(inode->i_mapping); - atomic_dec(&BTRFS_I(inode)->ordered_writeback); - } - atomic_dec(&inode->i_count); - iput(inode); - - mutex_lock(&root->fs_info->trans_mutex); - } - atomic_dec(&root->fs_info->throttles); - return 0; -} - static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, struct btrfs_pending_snapshot *pending) @@ -666,6 +615,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, extent_io_tree_init(pinned_copy, root->fs_info->btree_inode->i_mapping, GFP_NOFS); +printk("commit trans %Lu\n", trans->transid); trans->transaction->in_commit = 1; cur_trans = trans->transaction; if (cur_trans->list.prev != &root->fs_info->trans_list) { @@ -699,8 +649,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->trans_mutex); finish_wait(&cur_trans->writer_wait, &wait); - ret = btrfs_write_ordered_inodes(trans, root); - } while (cur_trans->num_writers > 1 || (cur_trans->num_joined != joined)); @@ -736,6 +684,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_copy_pinned(root, pinned_copy); + wake_up(&root->fs_info->transaction_throttle); + mutex_unlock(&root->fs_info->trans_mutex); ret = btrfs_write_and_wait_transaction(trans, root); BUG_ON(ret); @@ -758,6 +708,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); mutex_unlock(&root->fs_info->trans_mutex); +printk("done commit trans %Lu\n", trans->transid); kmem_cache_free(btrfs_trans_handle_cachep, trans); if (root->fs_info->closing) { diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 9ccd5a5b170..910350cd4cf 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -19,7 +19,6 @@ #ifndef __BTRFS_TRANSACTION__ #define __BTRFS_TRANSACTION__ #include "btrfs_inode.h" -#include "ordered-data.h" struct btrfs_transaction { u64 transid; @@ -31,7 +30,6 @@ struct btrfs_transaction { struct list_head list; struct extent_io_tree dirty_pages; unsigned long start_time; - struct btrfs_ordered_inode_tree ordered_inode_tree; wait_queue_head_t writer_wait; wait_queue_head_t commit_wait; struct list_head pending_snapshots; @@ -88,8 +86,6 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly); int btrfs_clean_old_snapshots(struct btrfs_root *root); int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); -int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, - struct btrfs_root *root); int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, struct btrfs_root *root); #endif -- cgit v1.2.3-70-g09d2 From 247e743cbe6e655768c3679f84821e03c1577902 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:53:51 -0400 Subject: Btrfs: Use async helpers to deal with pages that have been improperly dirtied Higher layers sometimes call set_page_dirty without asking the filesystem to help. This causes many problems for the data=ordered and cow code. This commit detects pages that haven't been properly setup for IO and kicks off an async helper to deal with them. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 6 ++++ fs/btrfs/disk-io.c | 4 +++ fs/btrfs/extent_io.c | 10 ++++++ fs/btrfs/extent_io.h | 1 + fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 93 +++++++++++++++++++++++++++++++++++++++++++++++----- 6 files changed, 106 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ceebc052ddc..4ddc8a8f82c 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -546,6 +546,12 @@ struct btrfs_fs_info { struct btrfs_workers endio_workers; struct btrfs_workers endio_write_workers; struct btrfs_workers submit_workers; + /* + * fixup workers take dirty pages that didn't properly go through + * the cow mechanism and make them safe to write. It happens + * for the sys_munmap function call path + */ + struct btrfs_workers fixup_workers; struct task_struct *transaction_kthread; struct task_struct *cleaner_kthread; int thread_pool_size; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 4a5ebafb935..66466d125c0 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1329,11 +1329,13 @@ struct btrfs_root *open_ctree(struct super_block *sb, */ btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->fixup_workers, 1); btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_write_workers, fs_info->thread_pool_size); btrfs_start_workers(&fs_info->workers, 1); btrfs_start_workers(&fs_info->submit_workers, 1); + btrfs_start_workers(&fs_info->fixup_workers, 1); btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); btrfs_start_workers(&fs_info->endio_write_workers, fs_info->thread_pool_size); @@ -1454,6 +1456,7 @@ fail_tree_root: fail_sys_array: fail_sb_buffer: extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); + btrfs_stop_workers(&fs_info->fixup_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_write_workers); @@ -1710,6 +1713,7 @@ int close_ctree(struct btrfs_root *root) truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); + btrfs_stop_workers(&fs_info->fixup_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_write_workers); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3f82a6e9ca4..feff16cb9b4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2050,6 +2050,16 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, lock_extent(tree, start, page_end, GFP_NOFS); unlock_start = start; + if (tree->ops && tree->ops->writepage_start_hook) { + ret = tree->ops->writepage_start_hook(page, start, page_end); + if (ret == -EAGAIN) { + unlock_extent(tree, start, page_end, GFP_NOFS); + redirty_page_for_writepage(wbc, page); + unlock_page(page); + return 0; + } + } + end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { printk("found delalloc bits after lock_extent\n"); diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 2268a799589..23affd27af5 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -30,6 +30,7 @@ typedef int (extent_submit_bio_hook_t)(struct inode *inode, int rw, struct bio *bio, int mirror_num); struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); + int (*writepage_start_hook)(struct page *page, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); extent_submit_bio_hook_t *submit_bio_hook; int (*merge_bio_hook)(struct page *page, unsigned long offset, diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 12e765f7e0d..20928639d17 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -313,6 +313,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; SetPageUptodate(p); + ClearPageChecked(p); set_page_dirty(p); } } else { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c5a62f0b959..47a008c1930 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -403,6 +403,87 @@ static int add_pending_csums(struct btrfs_trans_handle *trans, return 0; } +struct btrfs_writepage_fixup { + struct page *page; + struct btrfs_work work; +}; + +/* see btrfs_writepage_start_hook for details on why this is required */ +void btrfs_writepage_fixup_worker(struct btrfs_work *work) +{ + struct btrfs_writepage_fixup *fixup; + struct btrfs_ordered_extent *ordered; + struct page *page; + struct inode *inode; + u64 page_start; + u64 page_end; + + fixup = container_of(work, struct btrfs_writepage_fixup, work); + page = fixup->page; + + lock_page(page); + if (!page->mapping || !PageDirty(page) || !PageChecked(page)) { + ClearPageChecked(page); + goto out_page; + } + + inode = page->mapping->host; + page_start = page_offset(page); + page_end = page_offset(page) + PAGE_CACHE_SIZE - 1; + + lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); + ordered = btrfs_lookup_ordered_extent(inode, page_start); + if (ordered) + goto out; + + set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end, + GFP_NOFS); + ClearPageChecked(page); +out: + unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); +out_page: + unlock_page(page); + page_cache_release(page); +} + +/* + * There are a few paths in the higher layers of the kernel that directly + * set the page dirty bit without asking the filesystem if it is a + * good idea. This causes problems because we want to make sure COW + * properly happens and the data=ordered rules are followed. + * + * In our case any range that doesn't have the EXTENT_ORDERED bit set + * hasn't been properly setup for IO. We kick off an async process + * to fix it up. The async helper will wait for ordered extents, set + * the delalloc bit and make it safe to write the page. + */ +int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) +{ + struct inode *inode = page->mapping->host; + struct btrfs_writepage_fixup *fixup; + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret; + + ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end, + EXTENT_ORDERED, 0); + if (ret) + return 0; + + if (PageChecked(page)) + return -EAGAIN; + + fixup = kzalloc(sizeof(*fixup), GFP_NOFS); + if (!fixup) + return -EAGAIN; +printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page)); + SetPageChecked(page); + page_cache_get(page); + fixup->work.func = btrfs_writepage_fixup_worker; + fixup->page = page; + btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work); + return -EAGAIN; +} + int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state, int uptodate) { @@ -1263,6 +1344,7 @@ again: flush_dcache_page(page); kunmap(page); } + ClearPageChecked(page); set_page_dirty(page); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); @@ -2658,6 +2740,7 @@ again: flush_dcache_page(page); kunmap(page); } + ClearPageChecked(page); set_page_dirty(page); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); @@ -3039,15 +3122,6 @@ out_fail: static int btrfs_set_page_dirty(struct page *page) { - struct inode *inode = page->mapping->host; - u64 page_start = page_offset(page); - u64 page_end = page_start + PAGE_CACHE_SIZE - 1; - - if (!test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end, - EXTENT_DELALLOC, 0)) { -printk("inode %lu page %Lu not delalloc\n", inode->i_ino, page_offset(page)); -WARN_ON(1); - } return __set_page_dirty_nobuffers(page); } @@ -3098,6 +3172,7 @@ static struct extent_io_ops btrfs_extent_io_ops = { .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .writepage_end_io_hook = btrfs_writepage_end_io_hook, + .writepage_start_hook = btrfs_writepage_start_hook, .readpage_io_failed_hook = btrfs_io_failed_hook, .set_bit_hook = btrfs_set_bit_hook, .clear_bit_hook = btrfs_clear_bit_hook, -- cgit v1.2.3-70-g09d2 From dbe674a99c8af088faa4c95eddaeb271a3140ab6 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:54:05 -0400 Subject: Btrfs: Update on disk i_size only after pending ordered extents are done This changes the ordered data code to update i_size after the extent is on disk. An on disk i_size is maintained in the in-memory btrfs inode structures, and this is updated as extents finish. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 9 +++++ fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 28 ++++++++++------ fs/btrfs/ordered-data.c | 89 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/ordered-data.h | 2 ++ 5 files changed, 119 insertions(+), 11 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 8d03687510e..81c0444f37b 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -40,11 +40,20 @@ struct btrfs_inode { */ u64 last_trans; u64 delalloc_bytes; + u64 disk_i_size; u32 flags; }; + static inline struct btrfs_inode *BTRFS_I(struct inode *inode) { return container_of(inode, struct btrfs_inode, vfs_inode); } +static inline void btrfs_i_size_write(struct inode *inode, u64 size) +{ + inode->i_size = size; + BTRFS_I(inode)->disk_i_size = size; +} + + #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 20928639d17..3e4e5c227c0 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -338,7 +338,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, btrfs_update_inode(trans, root, inode); } failed: - err = btrfs_end_transaction(trans, root); + err = btrfs_end_transaction_throttle(trans, root); out_unlock: unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); return err; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 47a008c1930..baf46017d0d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -542,6 +542,7 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, add_pending_csums(trans, inode, ordered_extent->file_offset, &ordered_extent->list); + btrfs_ordered_update_i_size(inode, ordered_extent); btrfs_remove_ordered_extent(inode, ordered_extent); /* once for us */ btrfs_put_ordered_extent(ordered_extent); @@ -792,7 +793,7 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_nlink = btrfs_inode_nlink(leaf, inode_item); inode->i_uid = btrfs_inode_uid(leaf, inode_item); inode->i_gid = btrfs_inode_gid(leaf, inode_item); - inode->i_size = btrfs_inode_size(leaf, inode_item); + btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item)); tspec = btrfs_inode_atime(inode_item); inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec); @@ -860,7 +861,7 @@ static void fill_inode_item(struct extent_buffer *leaf, { btrfs_set_inode_uid(leaf, item, inode->i_uid); btrfs_set_inode_gid(leaf, item, inode->i_gid); - btrfs_set_inode_size(leaf, item, inode->i_size); + btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size); btrfs_set_inode_mode(leaf, item, inode->i_mode); btrfs_set_inode_nlink(leaf, item, inode->i_nlink); @@ -982,7 +983,7 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, err: btrfs_free_path(path); if (!ret) { - dir->i_size -= name_len * 2; + btrfs_i_size_write(dir, dir->i_size - name_len * 2); dir->i_mtime = dir->i_ctime = CURRENT_TIME; btrfs_update_inode(trans, root, dir); #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) @@ -1044,7 +1045,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) /* now the directory is empty */ err = btrfs_unlink_trans(trans, root, dir, dentry); if (!err) { - inode->i_size = 0; + btrfs_i_size_write(inode, 0); } nr = trans->blocks_used; @@ -1089,7 +1090,6 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int extent_type = -1; u64 mask = root->sectorsize - 1; - btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1); path = btrfs_alloc_path(); path->reada = -1; @@ -1427,7 +1427,7 @@ void btrfs_delete_inode(struct inode *inode) goto no_delete; } - inode->i_size = 0; + btrfs_i_size_write(inode, 0); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -1561,6 +1561,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_ino = args->ino; BTRFS_I(inode)->root = args->root; BTRFS_I(inode)->delalloc_bytes = 0; + BTRFS_I(inode)->disk_i_size = 0; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); @@ -1869,6 +1870,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; + BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->root = root; if (mode & S_IFDIR) @@ -1964,7 +1966,8 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_parent->d_inode->i_ino); } parent_inode = dentry->d_parent->d_inode; - parent_inode->i_size += dentry->d_name.len * 2; + btrfs_i_size_write(parent_inode, parent_inode->i_size + + dentry->d_name.len * 2); parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; ret = btrfs_update_inode(trans, root, dentry->d_parent->d_inode); @@ -2092,6 +2095,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; + BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -2199,7 +2203,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) inode->i_fop = &btrfs_dir_file_operations; btrfs_set_trans_block_group(trans, inode); - inode->i_size = 0; + btrfs_i_size_write(inode, 0); err = btrfs_update_inode(trans, root, inode); if (err) goto out_fail; @@ -2756,6 +2760,7 @@ static void btrfs_truncate(struct inode *inode) int ret; struct btrfs_trans_handle *trans; unsigned long nr; + u64 mask = root->sectorsize - 1; if (!S_ISREG(inode->i_mode)) return; @@ -2766,6 +2771,8 @@ static void btrfs_truncate(struct inode *inode) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); + btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); + btrfs_i_size_write(inode, inode->i_size); /* FIXME, add redo link to tree so we don't leak on crash */ ret = btrfs_truncate_in_trans(trans, root, inode, @@ -2821,7 +2828,7 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid, new_dirid); inode->i_nlink = 1; - inode->i_size = 0; + btrfs_i_size_write(inode, 0); return btrfs_update_inode(trans, new_root, inode); } @@ -3069,6 +3076,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; + BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } dir->i_sb->s_dirt = 1; @@ -3103,7 +3111,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; inode->i_mapping->backing_dev_info = &root->fs_info->bdi; - inode->i_size = name_len - 1; + btrfs_i_size_write(inode, name_len - 1); err = btrfs_update_inode(trans, root, inode); if (err) drop_inode = 1; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 6513270f054..d86a953ae51 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -406,3 +406,92 @@ out: mutex_unlock(&tree->mutex); return entry; } + +int btrfs_ordered_update_i_size(struct inode *inode, + struct btrfs_ordered_extent *ordered) +{ + struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + u64 disk_i_size; + u64 new_i_size; + u64 i_size_test; + struct rb_node *node; + struct btrfs_ordered_extent *test; + + mutex_lock(&tree->mutex); + disk_i_size = BTRFS_I(inode)->disk_i_size; + + /* + * if the disk i_size is already at the inode->i_size, or + * this ordered extent is inside the disk i_size, we're done + */ + if (disk_i_size >= inode->i_size || + ordered->file_offset + ordered->len <= disk_i_size) { + goto out; + } + + /* + * we can't update the disk_isize if there are delalloc bytes + * between disk_i_size and this ordered extent + */ + if (test_range_bit(io_tree, disk_i_size, + ordered->file_offset + ordered->len - 1, + EXTENT_DELALLOC, 0)) { + goto out; + } + /* + * walk backward from this ordered extent to disk_i_size. + * if we find an ordered extent then we can't update disk i_size + * yet + */ + while(1) { + node = rb_prev(&ordered->rb_node); + if (!node) + break; + test = rb_entry(node, struct btrfs_ordered_extent, rb_node); + if (test->file_offset + test->len <= disk_i_size) + break; + if (test->file_offset >= inode->i_size) + break; + if (test->file_offset >= disk_i_size) + goto out; + } + new_i_size = min_t(u64, entry_end(ordered), i_size_read(inode)); + + /* + * at this point, we know we can safely update i_size to at least + * the offset from this ordered extent. But, we need to + * walk forward and see if ios from higher up in the file have + * finished. + */ + node = rb_next(&ordered->rb_node); + i_size_test = 0; + if (node) { + /* + * do we have an area where IO might have finished + * between our ordered extent and the next one. + */ + test = rb_entry(node, struct btrfs_ordered_extent, rb_node); + if (test->file_offset > entry_end(ordered)) { + i_size_test = test->file_offset - 1; + } + } else { + i_size_test = i_size_read(inode); + } + + /* + * i_size_test is the end of a region after this ordered + * extent where there are no ordered extents. As long as there + * are no delalloc bytes in this area, it is safe to update + * disk_i_size to the end of the region. + */ + if (i_size_test > entry_end(ordered) && + !test_range_bit(io_tree, entry_end(ordered), i_size_test, + EXTENT_DELALLOC, 0)) { + new_i_size = min_t(u64, i_size_test, i_size_read(inode)); + } + BTRFS_I(inode)->disk_i_size = new_i_size; +out: + mutex_unlock(&tree->mutex); + return 0; +} diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 33292c5fe90..40e9126ad95 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -89,4 +89,6 @@ btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); int btrfs_add_ordered_pending(struct inode *inode, struct btrfs_ordered_extent *ordered, u64 start, u64 len); +int btrfs_ordered_update_i_size(struct inode *inode, + struct btrfs_ordered_extent *ordered); #endif -- cgit v1.2.3-70-g09d2 From f9295749388f82c8d2f485e99c72cd7c7876a99b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:54:14 -0400 Subject: btrfs_start_transaction: wait for commits in progress to finish btrfs_commit_transaction has to loop waiting for any writers in the transaction to finish before it can proceed. btrfs_start_transaction should be polite and not join a transaction that is in the process of being finished off. There are a few places that can't wait, basically the ones doing IO that might be needed to finish the transaction. For them, btrfs_join_transaction is added. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 1 + fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 10 +++++----- fs/btrfs/transaction.c | 43 ++++++++++++++++++++++++++++++++++++++++--- fs/btrfs/transaction.h | 3 +++ 6 files changed, 51 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 4ddc8a8f82c..acbce542d29 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -513,6 +513,7 @@ struct btrfs_fs_info { u64 alloc_start; struct btrfs_transaction *running_transaction; wait_queue_head_t transaction_throttle; + wait_queue_head_t transaction_wait; struct btrfs_super_block super_copy; struct btrfs_super_block super_for_commit; struct block_device *__bdev; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 66466d125c0..99bb385c298 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1291,6 +1291,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, mutex_init(&fs_info->cleaner_mutex); mutex_init(&fs_info->volume_mutex); init_waitqueue_head(&fs_info->transaction_throttle); + init_waitqueue_head(&fs_info->transaction_wait); #if 0 ret = add_hasher(fs_info, "crc32c"); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3e4e5c227c0..d6505892cd5 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -251,7 +251,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, end_of_last_block = start_pos + num_bytes - 1; lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); if (!trans) { err = -ENOMEM; goto out_unlock; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index baf46017d0d..0a687326c0b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -116,7 +116,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; int ret = 0; - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); BUG_ON(!trans); btrfs_set_trans_block_group(trans, inode); @@ -502,7 +502,7 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return 0; } - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); ordered_extent = btrfs_lookup_ordered_extent(inode, start); BUG_ON(!ordered_extent); @@ -1812,7 +1812,7 @@ int btrfs_write_inode(struct inode *inode, int wait) int ret = 0; if (wait) { - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); ret = btrfs_commit_transaction(trans, root); } @@ -1830,7 +1830,7 @@ void btrfs_dirty_inode(struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); btrfs_update_inode(trans, root, inode); btrfs_end_transaction(trans, root); @@ -2395,7 +2395,7 @@ again: free_extent_map(em); em = NULL; btrfs_release_path(root, path); - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); goto again; } write_extent_buffer(leaf, map + pg_offset, ptr, diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 86a5acc19ce..05823904ecb 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -62,6 +62,7 @@ static noinline int join_transaction(struct btrfs_root *root) init_waitqueue_head(&cur_trans->writer_wait); init_waitqueue_head(&cur_trans->commit_wait); cur_trans->in_commit = 0; + cur_trans->blocked = 0; cur_trans->use_count = 1; cur_trans->commit_done = 0; cur_trans->start_time = get_seconds(); @@ -99,14 +100,36 @@ static noinline int record_root_in_trans(struct btrfs_root *root) return 0; } -struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, - int num_blocks) +struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, + int num_blocks, int join) { struct btrfs_trans_handle *h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); + struct btrfs_transaction *cur_trans; int ret; mutex_lock(&root->fs_info->trans_mutex); + cur_trans = root->fs_info->running_transaction; + if (cur_trans && cur_trans->blocked && !join) { + DEFINE_WAIT(wait); + cur_trans->use_count++; + while(1) { + prepare_to_wait(&root->fs_info->transaction_wait, &wait, + TASK_UNINTERRUPTIBLE); + if (cur_trans->blocked) { + mutex_unlock(&root->fs_info->trans_mutex); + schedule(); + mutex_lock(&root->fs_info->trans_mutex); + finish_wait(&root->fs_info->transaction_wait, + &wait); + } else { + finish_wait(&root->fs_info->transaction_wait, + &wait); + break; + } + } + put_transaction(cur_trans); + } ret = join_transaction(root); BUG_ON(ret); @@ -123,6 +146,17 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, return h; } +struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, + int num_blocks) +{ + return start_transaction(root, num_blocks, 0); +} +struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root, + int num_blocks) +{ + return start_transaction(root, num_blocks, 1); +} + static noinline int wait_for_commit(struct btrfs_root *root, struct btrfs_transaction *commit) { @@ -156,7 +190,7 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); - if (cur_trans->in_commit && throttle) { + if (0 && cur_trans->in_commit && throttle) { DEFINE_WAIT(wait); mutex_unlock(&root->fs_info->trans_mutex); prepare_to_wait(&root->fs_info->transaction_throttle, &wait, @@ -617,6 +651,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, printk("commit trans %Lu\n", trans->transid); trans->transaction->in_commit = 1; + trans->transaction->blocked = 1; cur_trans = trans->transaction; if (cur_trans->list.prev != &root->fs_info->trans_list) { prev_trans = list_entry(cur_trans->list.prev, @@ -684,7 +719,9 @@ printk("commit trans %Lu\n", trans->transid); btrfs_copy_pinned(root, pinned_copy); + trans->transaction->blocked = 0; wake_up(&root->fs_info->transaction_throttle); + wake_up(&root->fs_info->transaction_wait); mutex_unlock(&root->fs_info->trans_mutex); ret = btrfs_write_and_wait_transaction(trans, root); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 910350cd4cf..11fbdeceb26 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -27,6 +27,7 @@ struct btrfs_transaction { int in_commit; int use_count; int commit_done; + int blocked; struct list_head list; struct extent_io_tree dirty_pages; unsigned long start_time; @@ -75,6 +76,8 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, int num_blocks); +struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root, + int num_blocks); int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, -- cgit v1.2.3-70-g09d2 From ba1da2f442ec91a1534afa893f9bef7e33056ace Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:54:15 -0400 Subject: Btrfs: Don't pin pages in ram until the entire ordered extent is on disk. Checksum items are not inserted until the entire ordered extent is on disk, but individual pages might be clean and available for reclaim long before the whole extent is on disk. In order to allow those pages to be freed, we need to be able to search the list of ordered extents to find the checksum that is going to be inserted in the tree. This way if the page needs to be read back in before the checksums are in the btree, we'll be able to verify the checksum on the page. This commit adds the ability to search the pending ordered extents for a given offset in the file, and changes btrfs_releasepage to allow ordered pages to be freed. Signed-off-by: Chris Mason --- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 37 +++++++++++++++++++++---------------- fs/btrfs/ordered-data.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/ordered-data.h | 1 + 4 files changed, 69 insertions(+), 19 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index d6505892cd5..3e4e5c227c0 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -251,7 +251,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, end_of_last_block = start_pos + num_bytes - 1; lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_start_transaction(root, 1); if (!trans) { err = -ENOMEM; goto out_unlock; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0a687326c0b..293355c92a4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -382,7 +382,7 @@ mapit: return btrfs_map_bio(root, rw, bio, mirror_num, 0); } -static int add_pending_csums(struct btrfs_trans_handle *trans, +static noinline int add_pending_csums(struct btrfs_trans_handle *trans, struct inode *inode, u64 file_offset, struct list_head *list) { @@ -390,15 +390,12 @@ static int add_pending_csums(struct btrfs_trans_handle *trans, struct btrfs_ordered_sum *sum; btrfs_set_trans_block_group(trans, inode); - while(!list_empty(list)) { - cur = list->next; + list_for_each(cur, list) { sum = list_entry(cur, struct btrfs_ordered_sum, list); mutex_lock(&BTRFS_I(inode)->csum_mutex); btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, inode, sum); mutex_unlock(&BTRFS_I(inode)->csum_mutex); - list_del(&sum->list); - kfree(sum); } return 0; } @@ -498,9 +495,8 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, int ret; ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); - if (!ret) { + if (!ret) return 0; - } trans = btrfs_join_transaction(root, 1); @@ -571,6 +567,18 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) path = btrfs_alloc_path(); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); if (IS_ERR(item)) { + /* + * It is possible there is an ordered extent that has + * not yet finished for this range in the file. If so, + * that extent will have a csum cached, and it will insert + * the sum after all the blocks in the extent are fully + * on disk. So, look for an ordered extent and use the + * sum if found. + */ + ret = btrfs_find_ordered_sum(inode, start, &csum); + if (ret == 0) + goto found; + ret = PTR_ERR(item); /* a csum that isn't present is a preallocated region. */ if (ret == -ENOENT || ret == -EFBIG) @@ -582,6 +590,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) } read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, BTRFS_CRC32_SIZE); +found: set_state_private(io_tree, start, csum); out: if (path) @@ -888,7 +897,7 @@ static void fill_inode_item(struct extent_buffer *leaf, BTRFS_I(inode)->block_group->key.objectid); } -int btrfs_update_inode(struct btrfs_trans_handle *trans, +int noinline btrfs_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) { @@ -1567,6 +1576,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); return 0; } @@ -1868,6 +1878,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; @@ -2097,6 +2108,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; + btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2618,14 +2630,6 @@ static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags) static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) { - struct btrfs_ordered_extent *ordered; - - ordered = btrfs_lookup_ordered_extent(page->mapping->host, - page_offset(page)); - if (ordered) { - btrfs_put_ordered_extent(ordered); - return 0; - } return __btrfs_releasepage(page, gfp_flags); } @@ -3078,6 +3082,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; + btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index d86a953ae51..b739e3abebb 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -245,8 +245,18 @@ out: int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) { - if (atomic_dec_and_test(&entry->refs)) + struct list_head *cur; + struct btrfs_ordered_sum *sum; + + if (atomic_dec_and_test(&entry->refs)) { + while(!list_empty(&entry->list)) { + cur = entry->list.next; + sum = list_entry(cur, struct btrfs_ordered_sum, list); + list_del(&sum->list); + kfree(sum); + } kfree(entry); + } return 0; } @@ -444,8 +454,9 @@ int btrfs_ordered_update_i_size(struct inode *inode, * if we find an ordered extent then we can't update disk i_size * yet */ + node = &ordered->rb_node; while(1) { - node = rb_prev(&ordered->rb_node); + node = rb_prev(node); if (!node) break; test = rb_entry(node, struct btrfs_ordered_extent, rb_node); @@ -495,3 +506,36 @@ out: mutex_unlock(&tree->mutex); return 0; } + +int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) +{ + struct btrfs_ordered_sum *ordered_sum; + struct btrfs_sector_sum *sector_sums; + struct btrfs_ordered_extent *ordered; + struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree; + struct list_head *cur; + int ret = 1; + int index; + + ordered = btrfs_lookup_ordered_extent(inode, offset); + if (!ordered) + return 1; + + mutex_lock(&tree->mutex); + list_for_each_prev(cur, &ordered->list) { + ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); + if (offset >= ordered_sum->file_offset && + offset < ordered_sum->file_offset + ordered_sum->len) { + index = (offset - ordered_sum->file_offset) / + BTRFS_I(inode)->root->sectorsize;; + sector_sums = &ordered_sum->sums; + *sum = sector_sums[index].sum; + ret = 0; + goto out; + } + } +out: + mutex_unlock(&tree->mutex); + return ret; +} + diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 40e9126ad95..33f0d9e91b1 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -91,4 +91,5 @@ int btrfs_add_ordered_pending(struct inode *inode, u64 start, u64 len); int btrfs_ordered_update_i_size(struct inode *inode, struct btrfs_ordered_extent *ordered); +int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum); #endif -- cgit v1.2.3-70-g09d2 From ee6e6504e147a59a9f4d582662c105e9d72ae638 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 12:54:40 -0400 Subject: Add a per-inode lock around btrfs_drop_extents btrfs_drop_extents is always called with a range lock held on the inode. But, it may operate on extents outside that range as it drops and splits them. This patch adds a per-inode mutex that is held while calling btrfs_drop_extents and while inserting new extents into the tree. It prevents races from two procs working against adjacent ranges in the tree. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 1 + fs/btrfs/extent-tree.c | 2 ++ fs/btrfs/file.c | 8 ++++++++ fs/btrfs/inode.c | 10 ++++++++++ 4 files changed, 21 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 81c0444f37b..3bf40591742 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -32,6 +32,7 @@ struct btrfs_inode { struct extent_io_tree io_tree; struct extent_io_tree io_failure_tree; struct mutex csum_mutex; + struct mutex extent_mutex; struct inode vfs_inode; struct btrfs_ordered_inode_tree ordered_tree; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 343d1101c31..4036c62b667 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1671,6 +1671,8 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, ref_generation, owner_objectid, owner_offset, pin, pin == 0); + + finish_current_insert(trans, root->fs_info->extent_root); pending_ret = del_pending_extents(trans, root->fs_info->extent_root); return ret ? ret : pending_ret; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3e4e5c227c0..40ad1b2958c 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -242,6 +242,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 end_of_last_block; u64 end_pos = pos + write_bytes; u64 inline_size; + int did_inline = 0; loff_t isize = i_size_read(inode); start_pos = pos & ~((u64)root->sectorsize - 1); @@ -275,6 +276,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, if (hole_size > 0) { btrfs_wait_ordered_range(inode, last_pos_in_file, last_pos_in_file + hole_size); + mutex_lock(&BTRFS_I(inode)->extent_mutex); err = btrfs_drop_extents(trans, root, inode, last_pos_in_file, last_pos_in_file + hole_size, @@ -289,6 +291,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, 0, 0, hole_size, 0); btrfs_drop_extent_cache(inode, last_pos_in_file, last_pos_in_file + hole_size -1); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); btrfs_check_file(root, inode); } if (err) @@ -321,6 +324,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, /* step one, delete the existing extents in this range */ aligned_end = (pos + write_bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); + mutex_lock(&BTRFS_I(inode)->extent_mutex); err = btrfs_drop_extents(trans, root, inode, start_pos, aligned_end, aligned_end, &hint_byte); if (err) @@ -332,9 +336,13 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inline_size, pages, 0, num_pages); btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1); BUG_ON(err); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); + did_inline = 1; } if (end_pos > isize) { i_size_write(inode, end_pos); + if (did_inline) + BTRFS_I(inode)->disk_i_size = end_pos; btrfs_update_inode(trans, root, inode); } failed: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 293355c92a4..0d1e2ba47a9 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -516,6 +516,8 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, trans->transid, inode->i_ino, ordered_extent->file_offset, &ins); BUG_ON(ret); + + mutex_lock(&BTRFS_I(inode)->extent_mutex); ret = btrfs_drop_extents(trans, root, inode, ordered_extent->file_offset, ordered_extent->file_offset + @@ -531,6 +533,8 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); + inode->i_blocks += ordered_extent->len >> 9; unlock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, @@ -1399,6 +1403,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); + mutex_lock(&BTRFS_I(inode)->extent_mutex); err = btrfs_drop_extents(trans, root, inode, hole_start, block_end, hole_start, &alloc_hint); @@ -1412,6 +1417,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) (u64)-1); btrfs_check_file(root, inode); } + mutex_unlock(&BTRFS_I(inode)->extent_mutex); btrfs_end_transaction(trans, root); unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); if (err) @@ -1578,6 +1584,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_mapping, GFP_NOFS); btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); + mutex_init(&BTRFS_I(inode)->extent_mutex); return 0; } @@ -1880,6 +1887,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mapping, GFP_NOFS); btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); + mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->root = root; @@ -2105,6 +2113,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); + mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; @@ -3079,6 +3088,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); mutex_init(&BTRFS_I(inode)->csum_mutex); + mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; -- cgit v1.2.3-70-g09d2 From eb84ae039e10f1f80443d846ba1350122bbdc753 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 17 Jul 2008 13:53:27 -0400 Subject: Btrfs: Cleanup and comment ordered-data.c Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 12 ++++- fs/btrfs/ordered-data.c | 130 ++++++++++++++++++++++++++---------------------- fs/btrfs/ordered-data.h | 49 +++++++++++++++--- 3 files changed, 121 insertions(+), 70 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0d1e2ba47a9..f37e09e724f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1343,7 +1343,7 @@ again: unlock_extent(io_tree, page_start, page_end, GFP_NOFS); unlock_page(page); page_cache_release(page); - btrfs_wait_ordered_extent(inode, ordered); + btrfs_start_ordered_extent(inode, ordered, 1); btrfs_put_ordered_extent(ordered); goto again; } @@ -2660,6 +2660,10 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) ordered = btrfs_lookup_ordered_extent(page->mapping->host, page_offset(page)); if (ordered) { + /* + * IO on this page will never be started, so we need + * to account for any ordered extents now + */ clear_extent_bit(tree, page_start, page_end, EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_LOCKED, 1, 0, GFP_NOFS); @@ -2732,11 +2736,15 @@ again: lock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_extent_mapped(page); + /* + * we can't set the delalloc bits if there are pending ordered + * extents. Drop our locks and wait for them to finish + */ ordered = btrfs_lookup_ordered_extent(inode, page_start); if (ordered) { unlock_extent(io_tree, page_start, page_end, GFP_NOFS); unlock_page(page); - btrfs_wait_ordered_extent(inode, ordered); + btrfs_start_ordered_extent(inode, ordered, 1); btrfs_put_ordered_extent(ordered); goto again; } diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index b739e3abebb..230fd3ca6b2 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -136,6 +136,19 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, return ret; } +/* allocate and add a new ordered_extent into the per-inode tree. + * file_offset is the logical offset in the file + * + * start is the disk block number of an extent already reserved in the + * extent allocation tree + * + * len is the length of the extent + * + * This also sets the EXTENT_ORDERED bit on the range in the inode. + * + * The tree is given a single reference on the ordered extent that was + * inserted. + */ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, u64 start, u64 len) { @@ -152,7 +165,6 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, entry->file_offset = file_offset; entry->start = start; entry->len = len; - entry->inode = inode; /* one ref for the tree */ atomic_set(&entry->refs, 1); init_waitqueue_head(&entry->wait); @@ -167,12 +179,15 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, set_extent_ordered(&BTRFS_I(inode)->io_tree, file_offset, entry_end(entry) - 1, GFP_NOFS); - set_bit(BTRFS_ORDERED_START, &entry->flags); mutex_unlock(&tree->mutex); BUG_ON(node); return 0; } +/* + * Add a struct btrfs_ordered_sum into the list of checksums to be inserted + * when an ordered extent is finished. + */ int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum) { struct btrfs_ordered_inode_tree *tree; @@ -182,29 +197,25 @@ int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum) tree = &BTRFS_I(inode)->ordered_tree; mutex_lock(&tree->mutex); node = tree_search(tree, sum->file_offset); - if (!node) { -search_fail: -printk("add ordered sum failed to find a node for inode %lu offset %Lu\n", inode->i_ino, sum->file_offset); - node = rb_first(&tree->tree); - while(node) { - entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - printk("entry %Lu %Lu %Lu\n", entry->file_offset, entry->file_offset + entry->len, entry->start); - node = rb_next(node); - } - BUG(); - } BUG_ON(!node); entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - if (!offset_in_entry(entry, sum->file_offset)) { - goto search_fail; - } + BUG_ON(!offset_in_entry(entry, sum->file_offset)); list_add_tail(&sum->list, &entry->list); mutex_unlock(&tree->mutex); return 0; } +/* + * this is used to account for finished IO across a given range + * of the file. The IO should not span ordered extents. If + * a given ordered_extent is completely done, 1 is returned, otherwise + * 0. + * + * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used + * to make sure this function only returns 1 once for a given ordered extent. + */ int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size) { @@ -233,9 +244,6 @@ int btrfs_dec_test_ordered_pending(struct inode *inode, ret = test_range_bit(io_tree, entry->file_offset, entry->file_offset + entry->len - 1, EXTENT_ORDERED, 0); - if (!test_bit(BTRFS_ORDERED_START, &entry->flags)) { -printk("inode %lu not ready yet for extent %Lu %Lu\n", inode->i_ino, entry->file_offset, entry_end(entry)); - } if (ret == 0) ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags); out: @@ -243,6 +251,10 @@ out: return ret == 0; } +/* + * used to drop a reference on an ordered extent. This will free + * the extent if the last reference is dropped + */ int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) { struct list_head *cur; @@ -260,6 +272,10 @@ int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) return 0; } +/* + * remove an ordered extent from the tree. No references are dropped + * but, anyone waiting on this extent is woken up. + */ int btrfs_remove_ordered_extent(struct inode *inode, struct btrfs_ordered_extent *entry) { @@ -277,27 +293,25 @@ int btrfs_remove_ordered_extent(struct inode *inode, return 0; } -void btrfs_wait_ordered_extent(struct inode *inode, - struct btrfs_ordered_extent *entry) -{ - u64 start = entry->file_offset; - u64 end = start + entry->len - 1; -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) - do_sync_file_range(file, start, end, SYNC_FILE_RANGE_WRITE); -#else - do_sync_mapping_range(inode->i_mapping, start, end, - SYNC_FILE_RANGE_WRITE); -#endif - wait_event(entry->wait, - test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); -} - -static void btrfs_start_ordered_extent(struct inode *inode, - struct btrfs_ordered_extent *entry, int wait) +/* + * Used to start IO or wait for a given ordered extent to finish. + * + * If wait is one, this effectively waits on page writeback for all the pages + * in the extent, and it waits on the io completion code to insert + * metadata into the btree corresponding to the extent + */ +void btrfs_start_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry, + int wait) { u64 start = entry->file_offset; u64 end = start + entry->len - 1; + /* + * pages in the range can be dirty, clean or writeback. We + * start IO on any dirty ones so the wait doesn't stall waiting + * for pdflush to find them + */ #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) do_sync_file_range(file, start, end, SYNC_FILE_RANGE_WRITE); #else @@ -309,6 +323,9 @@ static void btrfs_start_ordered_extent(struct inode *inode, &entry->flags)); } +/* + * Used to wait on ordered extents across a large range of bytes. + */ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) { u64 end; @@ -349,31 +366,11 @@ again: } } -int btrfs_add_ordered_pending(struct inode *inode, - struct btrfs_ordered_extent *ordered, - u64 start, u64 len) -{ - WARN_ON(1); - return 0; -#if 0 - int ret; - struct btrfs_ordered_inode_tree *tree; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - - tree = &BTRFS_I(inode)->ordered_tree; - mutex_lock(&tree->mutex); - if (test_bit(BTRFS_ORDERED_IO_DONE, &ordered->flags)) { - ret = -EAGAIN; - goto out; - } - set_extent_ordered(io_tree, start, start + len - 1, GFP_NOFS); - ret = 0; -out: - mutex_unlock(&tree->mutex); - return ret; -#endif -} +/* + * find an ordered extent corresponding to file_offset. return NULL if + * nothing is found, otherwise take a reference on the extent and return it + */ struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, u64 file_offset) { @@ -397,6 +394,10 @@ out: return entry; } +/* + * lookup and return any extent before 'file_offset'. NULL is returned + * if none is found + */ struct btrfs_ordered_extent * btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset) { @@ -417,6 +418,10 @@ out: return entry; } +/* + * After an extent is done, call this to conditionally update the on disk + * i_size. i_size is updated to cover any fully written part of the file. + */ int btrfs_ordered_update_i_size(struct inode *inode, struct btrfs_ordered_extent *ordered) { @@ -507,6 +512,11 @@ out: return 0; } +/* + * search the ordered extents for one corresponding to 'offset' and + * try to find a checksum. This is used because we allow pages to + * be reclaimed before their checksum is actually put into the btree + */ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) { struct btrfs_ordered_sum *ordered_sum; diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 33f0d9e91b1..98f491d1022 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -19,12 +19,19 @@ #ifndef __BTRFS_ORDERED_DATA__ #define __BTRFS_ORDERED_DATA__ +/* one of these per inode */ struct btrfs_ordered_inode_tree { struct mutex mutex; struct rb_root tree; struct rb_node *last; }; +/* + * these are used to collect checksums done just before bios submission. + * They are attached via a list into the ordered extent, and + * checksum items are inserted into the tree after all the blocks in + * the ordered extent are on disk + */ struct btrfs_sector_sum { u64 offset; u32 sum; @@ -34,27 +41,56 @@ struct btrfs_ordered_sum { u64 file_offset; u64 len; struct list_head list; + /* last field is a variable length array of btrfs_sector_sums */ struct btrfs_sector_sum sums; }; -/* bits for the flags field */ +/* + * bits for the flags field: + * + * BTRFS_ORDERED_IO_DONE is set when all of the blocks are written. + * It is used to make sure metadata is inserted into the tree only once + * per extent. + * + * BTRFS_ORDERED_COMPLETE is set when the extent is removed from the + * rbtree, just before waking any waiters. It is used to indicate the + * IO is done and any metadata is inserted into the tree. + */ #define BTRFS_ORDERED_IO_DONE 0 /* set when all the pages are written */ + #define BTRFS_ORDERED_COMPLETE 1 /* set when removed from the tree */ -#define BTRFS_ORDERED_START 2 /* set when tree setup */ struct btrfs_ordered_extent { + /* logical offset in the file */ u64 file_offset; + + /* disk byte number */ u64 start; + + /* length of the extent in bytes */ u64 len; + + /* flags (described above) */ unsigned long flags; + + /* reference count */ atomic_t refs; + + /* list of checksums for insertion when the extent io is done */ struct list_head list; - struct inode *inode; + + /* used to wait for the BTRFS_ORDERED_COMPLETE bit */ wait_queue_head_t wait; + + /* our friendly rbtree entry */ struct rb_node rb_node; }; +/* + * calculates the total size you need to allocate for an ordered sum + * structure spanning 'bytes' in the file + */ static inline int btrfs_ordered_sum_size(struct btrfs_root *root, u64 bytes) { unsigned long num_sectors = (bytes + root->sectorsize - 1) / @@ -81,14 +117,11 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum); struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, u64 file_offset); -void btrfs_wait_ordered_extent(struct inode *inode, - struct btrfs_ordered_extent *entry); +void btrfs_start_ordered_extent(struct inode *inode, + struct btrfs_ordered_extent *entry, int wait); void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len); struct btrfs_ordered_extent * btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); -int btrfs_add_ordered_pending(struct inode *inode, - struct btrfs_ordered_extent *ordered, - u64 start, u64 len); int btrfs_ordered_update_i_size(struct inode *inode, struct btrfs_ordered_extent *ordered); int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum); -- cgit v1.2.3-70-g09d2 From 3edf7d33f4edb1e4a9bb0a4c0a84d95fb4d22a09 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 06:17:13 -0400 Subject: Btrfs: Handle data checksumming on bios that span multiple ordered extents Data checksumming is done right before the bio is sent down the IO stack, which means a single bio might span more than one ordered extent. In this case, the checksumming data is split between two ordered extents. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 4 ++-- fs/btrfs/file-item.c | 43 +++++++++++++++++++++++++++++++++++++++---- fs/btrfs/inode.c | 6 +----- fs/btrfs/ordered-data.c | 36 ++++++++++++++++++------------------ fs/btrfs/ordered-data.h | 11 +++++++++-- 5 files changed, 69 insertions(+), 31 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index acbce542d29..96ab2797c09 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1579,8 +1579,8 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, struct btrfs_ordered_sum *sums); -int btrfs_csum_one_bio(struct btrfs_root *root, - struct bio *bio, struct btrfs_ordered_sum **sums_ret); +int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, + struct bio *bio); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 345caf8ff51..e02f1e5acb0 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -134,26 +134,53 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } -int btrfs_csum_one_bio(struct btrfs_root *root, - struct bio *bio, struct btrfs_ordered_sum **sums_ret) +int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, + struct bio *bio) { struct btrfs_ordered_sum *sums; struct btrfs_sector_sum *sector_sum; + struct btrfs_ordered_extent *ordered; char *data; struct bio_vec *bvec = bio->bi_io_vec; int bio_index = 0; + unsigned long total_bytes = 0; + unsigned long this_sum_bytes = 0; + u64 offset; WARN_ON(bio->bi_vcnt <= 0); sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS); if (!sums) return -ENOMEM; - *sums_ret = sums; + sector_sum = &sums->sums; - sums->file_offset = page_offset(bvec->bv_page); + sums->file_offset = page_offset(bvec->bv_page) + bvec->bv_offset; sums->len = bio->bi_size; INIT_LIST_HEAD(&sums->list); + ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset); + BUG_ON(!ordered); while(bio_index < bio->bi_vcnt) { + offset = page_offset(bvec->bv_page) + bvec->bv_offset; + if (offset >= ordered->file_offset + ordered->len) { + unsigned long bytes_left; + sums->len = this_sum_bytes; + this_sum_bytes = 0; + btrfs_add_ordered_sum(inode, ordered, sums); + btrfs_put_ordered_extent(ordered); + + bytes_left = bio->bi_size - total_bytes; + + sums = kzalloc(btrfs_ordered_sum_size(root, bytes_left), + GFP_NOFS); + BUG_ON(!sums); + sector_sum = &sums->sums; + sums->len = bytes_left; + sums->file_offset = offset; + ordered = btrfs_lookup_ordered_extent(inode, + sums->file_offset); + BUG_ON(!ordered); + } + data = kmap_atomic(bvec->bv_page, KM_USER0); sector_sum->sum = ~(u32)0; sector_sum->sum = btrfs_csum_data(root, @@ -165,10 +192,18 @@ int btrfs_csum_one_bio(struct btrfs_root *root, (char *)§or_sum->sum); sector_sum->offset = page_offset(bvec->bv_page) + bvec->bv_offset; + sector_sum++; bio_index++; + total_bytes += bvec->bv_len; + this_sum_bytes += bvec->bv_len; bvec++; } + btrfs_add_ordered_sum(inode, ordered, sums); + btrfs_put_ordered_extent(ordered); + if (total_bytes != bio->bi_size) { +printk("warning, total bytes %lu bio size %u\n", total_bytes, bio->bi_size); + } return 0; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f37e09e724f..4d729d90d2b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -351,12 +351,8 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, { struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; - struct btrfs_ordered_sum *sums; - ret = btrfs_csum_one_bio(root, bio, &sums); - BUG_ON(ret); - - ret = btrfs_add_ordered_sum(inode, sums); + ret = btrfs_csum_one_bio(root, inode, bio); BUG_ON(ret); return btrfs_map_bio(root, rw, bio, mirror_num, 1); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 230fd3ca6b2..1ddb7bceea9 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -186,22 +186,17 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, /* * Add a struct btrfs_ordered_sum into the list of checksums to be inserted - * when an ordered extent is finished. + * when an ordered extent is finished. If the list covers more than one + * ordered extent, it is split across multiples. */ -int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum) +int btrfs_add_ordered_sum(struct inode *inode, + struct btrfs_ordered_extent *entry, + struct btrfs_ordered_sum *sum) { struct btrfs_ordered_inode_tree *tree; - struct rb_node *node; - struct btrfs_ordered_extent *entry; tree = &BTRFS_I(inode)->ordered_tree; mutex_lock(&tree->mutex); - node = tree_search(tree, sum->file_offset); - BUG_ON(!node); - - entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - BUG_ON(!offset_in_entry(entry, sum->file_offset)); - list_add_tail(&sum->list, &entry->list); mutex_unlock(&tree->mutex); return 0; @@ -524,8 +519,10 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) struct btrfs_ordered_extent *ordered; struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree; struct list_head *cur; + unsigned long num_sectors; + unsigned long i; + u32 sectorsize = BTRFS_I(inode)->root->sectorsize; int ret = 1; - int index; ordered = btrfs_lookup_ordered_extent(inode, offset); if (!ordered) @@ -534,14 +531,17 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) mutex_lock(&tree->mutex); list_for_each_prev(cur, &ordered->list) { ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); - if (offset >= ordered_sum->file_offset && - offset < ordered_sum->file_offset + ordered_sum->len) { - index = (offset - ordered_sum->file_offset) / - BTRFS_I(inode)->root->sectorsize;; + if (offset >= ordered_sum->file_offset) { + num_sectors = ordered_sum->len / sectorsize; sector_sums = &ordered_sum->sums; - *sum = sector_sums[index].sum; - ret = 0; - goto out; + for (i = 0; i < num_sectors; i++) { + if (sector_sums[i].offset == offset) { +printk("find ordered sum inode %lu offset %Lu\n", inode->i_ino, offset); + *sum = sector_sums[i].sum; + ret = 0; + goto out; + } + } } } out: diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 98f491d1022..1794efd13ca 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -39,7 +39,11 @@ struct btrfs_sector_sum { struct btrfs_ordered_sum { u64 file_offset; - u64 len; + /* + * this is the length in bytes covered by the sums array below. + * But, the sums array may not be contiguous in the file. + */ + unsigned long len; struct list_head list; /* last field is a variable length array of btrfs_sector_sums */ struct btrfs_sector_sum sums; @@ -95,6 +99,7 @@ static inline int btrfs_ordered_sum_size(struct btrfs_root *root, u64 bytes) { unsigned long num_sectors = (bytes + root->sectorsize - 1) / root->sectorsize; + num_sectors++; return sizeof(struct btrfs_ordered_sum) + num_sectors * sizeof(struct btrfs_sector_sum); } @@ -114,7 +119,9 @@ int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size); int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, u64 start, u64 len); -int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum); +int btrfs_add_ordered_sum(struct inode *inode, + struct btrfs_ordered_extent *entry, + struct btrfs_ordered_sum *sum); struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, u64 file_offset); void btrfs_start_ordered_extent(struct inode *inode, -- cgit v1.2.3-70-g09d2 From 211f90e68b679d27fe23c5505f86d6ce62c98bae Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 11:56:15 -0400 Subject: Btrfs: Don't allow releasepage to succeed if EXTENT_ORDERED is set Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 3 ++- fs/btrfs/inode.c | 14 +++++++++----- 2 files changed, 11 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index feff16cb9b4..3153b4fdd0b 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2542,7 +2542,8 @@ int try_release_extent_state(struct extent_map_tree *map, u64 end = start + PAGE_CACHE_SIZE - 1; int ret = 1; - if (test_range_bit(tree, start, end, EXTENT_IOBITS, 0)) + if (test_range_bit(tree, start, end, + EXTENT_IOBITS | EXTENT_ORDERED, 0)) ret = 0; else { if ((mask & GFP_NOFS) == GFP_NOFS) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4d729d90d2b..8803abc89bb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -477,10 +477,8 @@ printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page return -EAGAIN; } -int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, - struct extent_state *state, int uptodate) +static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) { - struct inode *inode = page->mapping->host; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; struct btrfs_ordered_extent *ordered_extent; @@ -550,6 +548,12 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return 0; } +int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, + struct extent_state *state, int uptodate) +{ + return btrfs_finish_ordered_io(page->mapping->host, start, end); +} + int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) { int ret = 0; @@ -2663,8 +2667,8 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) clear_extent_bit(tree, page_start, page_end, EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_LOCKED, 1, 0, GFP_NOFS); - btrfs_writepage_end_io_hook(page, page_start, - page_end, NULL, 1); + btrfs_finish_ordered_io(page->mapping->host, + page_start, page_end); btrfs_put_ordered_extent(ordered); lock_extent(tree, page_start, page_end, GFP_NOFS); } -- cgit v1.2.3-70-g09d2 From 7f3c74fb831fa19bafe087e817c0a5ff3883f1ea Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 12:01:11 -0400 Subject: Btrfs: Keep extent mappings in ram until pending ordered extents are done It was possible for stale mappings from disk to be used instead of the new pending ordered extent. This adds a flag to the extent map struct to keep it pinned until the pending ordered extent is actually on disk. Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 27 +++++++++++++++------------ fs/btrfs/extent_map.c | 4 ++++ fs/btrfs/extent_map.h | 3 +++ fs/btrfs/file-item.c | 5 +---- fs/btrfs/file.c | 14 ++++++++++---- fs/btrfs/inode.c | 15 +++++++++++++++ 6 files changed, 48 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 3153b4fdd0b..d4a63ae7ed1 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2000,7 +2000,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, struct block_device *bdev; int ret; int nr = 0; - size_t page_offset = 0; + size_t pg_offset = 0; size_t blocksize; loff_t i_size = i_size_read(inode); unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; @@ -2008,9 +2008,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 delalloc_end; WARN_ON(!PageLocked(page)); - page_offset = i_size & (PAGE_CACHE_SIZE - 1); + pg_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || - (page->index == end_index && !page_offset)) { + (page->index == end_index && !pg_offset)) { page->mapping->a_ops->invalidatepage(page, 0); unlock_page(page); return 0; @@ -2020,12 +2020,12 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, char *userpage; userpage = kmap_atomic(page, KM_USER0); - memset(userpage + page_offset, 0, - PAGE_CACHE_SIZE - page_offset); + memset(userpage + pg_offset, 0, + PAGE_CACHE_SIZE - pg_offset); kunmap_atomic(userpage, KM_USER0); flush_dcache_page(page); } - page_offset = 0; + pg_offset = 0; set_page_extent_mapped(page); @@ -2088,7 +2088,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_start = page_end + 1; break; } - em = epd->get_extent(inode, page, page_offset, cur, + em = epd->get_extent(inode, page, pg_offset, cur, end - cur + 1, 1); if (IS_ERR(em) || !em) { SetPageError(page); @@ -2113,12 +2113,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_extent(tree, unlock_start, cur + iosize -1, GFP_NOFS); + if (tree->ops && tree->ops->writepage_end_io_hook) tree->ops->writepage_end_io_hook(page, cur, cur + iosize - 1, NULL, 1); cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; unlock_start = cur; continue; } @@ -2127,7 +2128,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, if (0 && !test_range_bit(tree, cur, cur + iosize - 1, EXTENT_DIRTY, 0)) { cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; continue; } clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -2141,6 +2142,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, SetPageError(page); } else { unsigned long max_nr = end_index + 1; + set_range_writeback(tree, cur, cur + iosize - 1); if (!PageWriteback(page)) { printk("warning page %lu not writeback, " @@ -2150,14 +2152,14 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, } ret = submit_extent_page(WRITE, tree, page, sector, - iosize, page_offset, bdev, + iosize, pg_offset, bdev, &epd->bio, max_nr, end_bio_extent_writepage, 0); if (ret) SetPageError(page); } cur = cur + iosize; - page_offset += iosize; + pg_offset += iosize; nr++; } done: @@ -2579,7 +2581,8 @@ int try_release_extent_mapping(struct extent_map_tree *map, spin_unlock(&map->lock); break; } - if (em->start != start) { + if (test_bit(EXTENT_FLAG_PINNED, &em->flags) || + em->start != start) { spin_unlock(&map->lock); free_extent_map(em); break; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 81123277c2b..71b1ac15535 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -173,6 +173,9 @@ static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) static int mergable_maps(struct extent_map *prev, struct extent_map *next) { + if (test_bit(EXTENT_FLAG_PINNED, &prev->flags)) + return 0; + if (extent_map_end(prev) == next->start && prev->flags == next->flags && prev->bdev == next->bdev && @@ -320,6 +323,7 @@ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) { int ret = 0; + WARN_ON(test_bit(EXTENT_FLAG_PINNED, &em->flags)); BUG_ON(spin_trylock(&tree->lock)); rb_erase(&em->rb_node, &tree->map); em->in_tree = 0; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 56314217cfc..a3978ec2784 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -8,6 +8,9 @@ #define EXTENT_MAP_INLINE (u64)-2 #define EXTENT_MAP_DELALLOC (u64)-1 +/* bits for the flags field */ +#define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */ + struct extent_map { struct rb_node rb_node; diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index e02f1e5acb0..d9c69e16d36 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -192,7 +192,6 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, (char *)§or_sum->sum); sector_sum->offset = page_offset(bvec->bv_page) + bvec->bv_offset; - sector_sum++; bio_index++; total_bytes += bvec->bv_len; @@ -201,9 +200,6 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, } btrfs_add_ordered_sum(inode, ordered, sums); btrfs_put_ordered_extent(ordered); - if (total_bytes != bio->bi_size) { -printk("warning, total bytes %lu bio size %u\n", total_bytes, bio->bi_size); - } return 0; } @@ -372,6 +368,7 @@ next_sector: write_extent_buffer(leaf, §or_sum->sum, (unsigned long)item, BTRFS_CRC32_SIZE); } + total_bytes += root->sectorsize; sector_sum++; if (total_bytes < sums->len) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 40ad1b2958c..eccdb9562ba 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -358,9 +358,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) struct extent_map *split = NULL; struct extent_map *split2 = NULL; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - struct extent_map *tmp; u64 len = end - start + 1; - u64 next_start; int ret; int testend = 1; @@ -381,8 +379,16 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) spin_unlock(&em_tree->lock); break; } - tmp = rb_entry(&em->rb_node, struct extent_map, rb_node); - next_start = tmp->start; + if (test_bit(EXTENT_FLAG_PINNED, &em->flags)) { + start = em->start + em->len; + free_extent_map(em); + spin_unlock(&em_tree->lock); + if (start < end) { + len = end - start + 1; + continue; + } + break; + } remove_extent_mapping(em_tree, em); if (em->block_start < EXTENT_MAP_LAST_BYTE && diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8803abc89bb..08dbe738b51 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -144,6 +144,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) em->len = ins.offset; em->block_start = ins.objectid; em->bdev = root->fs_info->fs_devices->latest_bdev; + set_bit(EXTENT_FLAG_PINNED, &em->flags); while(1) { spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); @@ -483,6 +484,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; struct btrfs_ordered_extent *ordered_extent; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_map *em; u64 alloc_hint = 0; struct list_head list; struct btrfs_key ins; @@ -524,6 +527,17 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ordered_extent->len, ordered_extent->len, 0); BUG_ON(ret); + + + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, ordered_extent->file_offset, + ordered_extent->len); + if (em) { + clear_bit(EXTENT_FLAG_PINNED, &em->flags); + free_extent_map(em); + } + spin_unlock(&em_tree->lock); + btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1); @@ -538,6 +552,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) btrfs_ordered_update_i_size(inode, ordered_extent); btrfs_remove_ordered_extent(inode, ordered_extent); + /* once for us */ btrfs_put_ordered_extent(ordered_extent); /* once for the tree */ -- cgit v1.2.3-70-g09d2 From e5a2217ef6ff088d08a27208929a6f9c635d672c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 18 Jul 2008 20:42:20 -0400 Subject: Fix btrfs_wait_ordered_extent_range to properly wait Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 3 ++- fs/btrfs/inode.c | 26 ++++++++++++++++++++------ fs/btrfs/ordered-data.c | 45 +++++++++++++++++++++++++++------------------ 3 files changed, 49 insertions(+), 25 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index d9c69e16d36..45127e4797c 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -161,7 +161,8 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, while(bio_index < bio->bi_vcnt) { offset = page_offset(bvec->bv_page) + bvec->bv_offset; - if (offset >= ordered->file_offset + ordered->len) { + if (offset >= ordered->file_offset + ordered->len || + offset < ordered->file_offset) { unsigned long bytes_left; sums->len = this_sum_bytes; this_sum_bytes = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 08dbe738b51..50ee4befac8 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -128,7 +128,9 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) goto out; BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); + mutex_lock(&BTRFS_I(inode)->extent_mutex); btrfs_drop_extent_cache(inode, start, start + num_bytes - 1); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); while(num_bytes > 0) { cur_alloc_size = min(num_bytes, root->fs_info->max_extent); @@ -144,6 +146,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) em->len = ins.offset; em->block_start = ins.objectid; em->bdev = root->fs_info->fs_devices->latest_bdev; + mutex_lock(&BTRFS_I(inode)->extent_mutex); set_bit(EXTENT_FLAG_PINNED, &em->flags); while(1) { spin_lock(&em_tree->lock); @@ -156,6 +159,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) btrfs_drop_extent_cache(inode, start, start + ins.offset - 1); } + mutex_unlock(&BTRFS_I(inode)->extent_mutex); cur_alloc_size = ins.offset; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, @@ -487,6 +491,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct extent_map *em; u64 alloc_hint = 0; + u64 clear_start; + u64 clear_end; struct list_head list; struct btrfs_key ins; int ret; @@ -509,12 +515,14 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ins.objectid = ordered_extent->start; ins.offset = ordered_extent->len; ins.type = BTRFS_EXTENT_ITEM_KEY; + ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid, trans->transid, inode->i_ino, ordered_extent->file_offset, &ins); BUG_ON(ret); mutex_lock(&BTRFS_I(inode)->extent_mutex); + ret = btrfs_drop_extents(trans, root, inode, ordered_extent->file_offset, ordered_extent->file_offset + @@ -528,13 +536,19 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ordered_extent->len, 0); BUG_ON(ret); - spin_lock(&em_tree->lock); - em = lookup_extent_mapping(em_tree, ordered_extent->file_offset, - ordered_extent->len); - if (em) { - clear_bit(EXTENT_FLAG_PINNED, &em->flags); - free_extent_map(em); + clear_start = ordered_extent->file_offset; + clear_end = ordered_extent->file_offset + ordered_extent->len; + while(clear_start < clear_end) { + em = lookup_extent_mapping(em_tree, clear_start, + clear_end - clear_start); + if (em) { + clear_bit(EXTENT_FLAG_PINNED, &em->flags); + clear_start = em->start + em->len; + free_extent_map(em); + } else { + break; + } } spin_unlock(&em_tree->lock); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 1ddb7bceea9..c2b4a9c4ddb 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -324,22 +324,37 @@ void btrfs_start_ordered_extent(struct inode *inode, void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) { u64 end; + u64 orig_end; + u64 wait_end; struct btrfs_ordered_extent *ordered; - int found; - int should_wait = 0; - -again: - if (start + len < start) - end = (u64)-1; - else - end = start + len - 1; - found = 0; + u64 mask = BTRFS_I(inode)->root->sectorsize - 1; + + if (start + len < start) { + wait_end = (inode->i_size + mask) & ~mask; + orig_end = (u64)-1; + } else { + orig_end = start + len - 1; + wait_end = orig_end; + } + + /* start IO across the range first to instantiate any delalloc + * extents + */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) + do_sync_file_range(file, start, wait_end, SYNC_FILE_RANGE_WRITE); +#else + do_sync_mapping_range(inode->i_mapping, start, wait_end, + SYNC_FILE_RANGE_WRITE); +#endif + end = orig_end; + wait_on_extent_writeback(&BTRFS_I(inode)->io_tree, start, orig_end); + while(1) { ordered = btrfs_lookup_first_ordered_extent(inode, end); if (!ordered) { break; } - if (ordered->file_offset >= start + len) { + if (ordered->file_offset > orig_end) { btrfs_put_ordered_extent(ordered); break; } @@ -347,21 +362,15 @@ again: btrfs_put_ordered_extent(ordered); break; } - btrfs_start_ordered_extent(inode, ordered, should_wait); - found++; + btrfs_start_ordered_extent(inode, ordered, 1); end = ordered->file_offset; btrfs_put_ordered_extent(ordered); - if (end == 0) + if (end == 0 || end == start) break; end--; } - if (should_wait && found) { - should_wait = 0; - goto again; - } } - /* * find an ordered extent corresponding to file_offset. return NULL if * nothing is found, otherwise take a reference on the extent and return it -- cgit v1.2.3-70-g09d2 From 4a09675279674041862d2210635b0cc1f60be28e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 21 Jul 2008 10:29:44 -0400 Subject: Btrfs: Data ordered fixes * In btrfs_delete_inode, wait for ordered extents after calling truncate_inode_pages. This is much faster, and more correct * Properly clear our the PageChecked bit everywhere we redirty the page. * Change the writepage fixup handler to lock the page range and check to see if an ordered extent had been inserted since the improperly dirtied page was discovered * Wait for ordered extents outside the transaction. This isn't required for locking rules but does improve transaction latencies * Reduce contention on the alloc_mutex by dropping it while incrementing refs on a node/leaf and while dropping refs on a leaf. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 18 +++++++++++++++--- fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 23 ++++++++++++++++++----- fs/btrfs/ordered-data.c | 11 +++++++++-- 4 files changed, 43 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index be2aef1cb7a..ccd49322f79 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -934,7 +934,6 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (!root->ref_cows) return 0; - mutex_lock(&root->fs_info->alloc_mutex); level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); for (i = 0; i < nritems; i++) { @@ -951,29 +950,36 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); if (disk_bytenr == 0) continue; + + mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(buf, fi), root->root_key.objectid, trans->transid, key.objectid, key.offset); + mutex_unlock(&root->fs_info->alloc_mutex); if (ret) { faili = i; + WARN_ON(1); goto fail; } } else { bytenr = btrfs_node_blockptr(buf, i); btrfs_node_key_to_cpu(buf, &key, i); + + mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_inc_extent_ref(trans, root, bytenr, btrfs_level_size(root, level - 1), root->root_key.objectid, trans->transid, level - 1, key.objectid); + mutex_unlock(&root->fs_info->alloc_mutex); if (ret) { faili = i; + WARN_ON(1); goto fail; } } } - mutex_unlock(&root->fs_info->alloc_mutex); return 0; fail: WARN_ON(1); @@ -1004,7 +1010,6 @@ fail: } } #endif - mutex_unlock(&root->fs_info->alloc_mutex); return ret; } @@ -2180,6 +2185,8 @@ static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, leaf_owner = btrfs_header_owner(leaf); leaf_generation = btrfs_header_generation(leaf); + mutex_unlock(&root->fs_info->alloc_mutex); + for (i = 0; i < nritems; i++) { u64 disk_bytenr; @@ -2197,12 +2204,17 @@ static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); if (disk_bytenr == 0) continue; + + mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(leaf, fi), leaf_owner, leaf_generation, key.objectid, key.offset, 0); + mutex_unlock(&root->fs_info->alloc_mutex); BUG_ON(ret); } + + mutex_lock(&root->fs_info->alloc_mutex); return 0; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index eccdb9562ba..591a30208ac 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -75,6 +75,7 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages) for (i = 0; i < num_pages; i++) { if (!pages[i]) break; + ClearPageChecked(pages[i]); unlock_page(pages[i]); mark_page_accessed(pages[i]); page_cache_release(pages[i]); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 50ee4befac8..8fb6dc25e7a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -418,7 +418,7 @@ void btrfs_writepage_fixup_worker(struct btrfs_work *work) fixup = container_of(work, struct btrfs_writepage_fixup, work); page = fixup->page; - +again: lock_page(page); if (!page->mapping || !PageDirty(page) || !PageChecked(page)) { ClearPageChecked(page); @@ -430,9 +430,21 @@ void btrfs_writepage_fixup_worker(struct btrfs_work *work) page_end = page_offset(page) + PAGE_CACHE_SIZE - 1; lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); - ordered = btrfs_lookup_ordered_extent(inode, page_start); - if (ordered) + + /* already ordered? We're done */ + if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end, + EXTENT_ORDERED, 0)) { goto out; + } + + ordered = btrfs_lookup_ordered_extent(inode, page_start); + if (ordered) { + unlock_extent(&BTRFS_I(inode)->io_tree, page_start, + page_end, GFP_NOFS); + unlock_page(page); + btrfs_start_ordered_extent(inode, ordered, 1); + goto again; + } set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); @@ -1465,11 +1477,11 @@ void btrfs_delete_inode(struct inode *inode) unsigned long nr; int ret; - btrfs_wait_ordered_range(inode, 0, (u64)-1); truncate_inode_pages(&inode->i_data, 0); if (is_bad_inode(inode)) { goto no_delete; } + btrfs_wait_ordered_range(inode, 0, (u64)-1); btrfs_i_size_write(inode, 0); trans = btrfs_start_transaction(root, 1); @@ -2707,6 +2719,7 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) 1, 1, GFP_NOFS); __btrfs_releasepage(page, GFP_NOFS); + ClearPageChecked(page); if (PagePrivate(page)) { invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); @@ -2818,10 +2831,10 @@ static void btrfs_truncate(struct inode *inode) return; btrfs_truncate_page(inode->i_mapping, inode->i_size); + btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); - btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); btrfs_i_size_write(inode, inode->i_size); /* FIXME, add redo link to tree so we don't leak on crash */ diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index c2b4a9c4ddb..0d87795fdd8 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -336,7 +336,7 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) orig_end = start + len - 1; wait_end = orig_end; } - +again: /* start IO across the range first to instantiate any delalloc * extents */ @@ -369,6 +369,14 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) break; end--; } + if (test_range_bit(&BTRFS_I(inode)->io_tree, start, orig_end, + EXTENT_ORDERED | EXTENT_DELALLOC, 0)) { + printk("inode %lu still ordered or delalloc after wait " + "%llu %llu\n", inode->i_ino, + (unsigned long long)start, + (unsigned long long)orig_end); + goto again; + } } /* @@ -545,7 +553,6 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) sector_sums = &ordered_sum->sums; for (i = 0; i < num_sectors; i++) { if (sector_sums[i].offset == offset) { -printk("find ordered sum inode %lu offset %Lu\n", inode->i_ino, offset); *sum = sector_sums[i].sum; ret = 0; goto out; -- cgit v1.2.3-70-g09d2 From 6af118ce51b52ceda357c671550c79628b9c4a65 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 11:18:07 -0400 Subject: Btrfs: Index extent buffers in an rbtree Before, extent buffers were a temporary object, meant to map a number of pages at once and collect operations on them. But, a few extra fields have crept in, and they are also the best place to store a per-tree block lock field as well. This commit puts the extent buffers into an rbtree, and ensures a single extent buffer for each tree block. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 26 ++--- fs/btrfs/extent_io.c | 309 +++++++++++++++++++-------------------------------- fs/btrfs/extent_io.h | 11 +- fs/btrfs/inode.c | 3 - 4 files changed, 129 insertions(+), 220 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 99bb385c298..86e84a8579e 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -381,7 +381,6 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, end = min_t(u64, eb->len, PAGE_CACHE_SIZE); end = eb->start + end - 1; - release_extent_buffer_tail_pages(eb); err: free_extent_buffer(eb); out: @@ -563,21 +562,21 @@ static int btree_releasepage(struct page *page, gfp_t gfp_flags) struct extent_map_tree *map; int ret; - if (page_count(page) > 3) { - /* once for page->private, once for the caller, once - * once for the page cache - */ - return 0; - } tree = &BTRFS_I(page->mapping->host)->io_tree; map = &BTRFS_I(page->mapping->host)->extent_tree; + ret = try_release_extent_state(map, tree, page, gfp_flags); + if (!ret) { + return 0; + } + + ret = try_release_extent_buffer(tree, page); if (ret == 1) { - invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); } + return ret; } @@ -588,7 +587,8 @@ static void btree_invalidatepage(struct page *page, unsigned long offset) extent_invalidatepage(tree, page, offset); btree_releasepage(page, GFP_NOFS); if (PagePrivate(page)) { - invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); + printk("warning page private not zero on page %Lu\n", + page_offset(page)); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); @@ -1456,7 +1456,6 @@ fail_tree_root: free_extent_buffer(tree_root->node); fail_sys_array: fail_sb_buffer: - extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); btrfs_stop_workers(&fs_info->fixup_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); @@ -1705,13 +1704,6 @@ int close_ctree(struct btrfs_root *root) filemap_write_and_wait(fs_info->btree_inode->i_mapping); - extent_io_tree_empty_lru(&fs_info->free_space_cache); - extent_io_tree_empty_lru(&fs_info->block_group_cache); - extent_io_tree_empty_lru(&fs_info->pinned_extents); - extent_io_tree_empty_lru(&fs_info->pending_del); - extent_io_tree_empty_lru(&fs_info->extent_ins); - extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); - truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); btrfs_stop_workers(&fs_info->fixup_workers); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index d4a63ae7ed1..32bb4ed3723 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -91,29 +91,16 @@ void extent_io_tree_init(struct extent_io_tree *tree, struct address_space *mapping, gfp_t mask) { tree->state.rb_node = NULL; + tree->buffer.rb_node = NULL; tree->ops = NULL; tree->dirty_bytes = 0; spin_lock_init(&tree->lock); - spin_lock_init(&tree->lru_lock); + spin_lock_init(&tree->buffer_lock); tree->mapping = mapping; - INIT_LIST_HEAD(&tree->buffer_lru); - tree->lru_size = 0; tree->last = NULL; } EXPORT_SYMBOL(extent_io_tree_init); -void extent_io_tree_empty_lru(struct extent_io_tree *tree) -{ - struct extent_buffer *eb; - while(!list_empty(&tree->buffer_lru)) { - eb = list_entry(tree->buffer_lru.next, struct extent_buffer, - lru); - list_del_init(&eb->lru); - free_extent_buffer(eb); - } -} -EXPORT_SYMBOL(extent_io_tree_empty_lru); - struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; @@ -245,6 +232,50 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, return ret; } +static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree, + u64 offset, struct rb_node *node) +{ + struct rb_root *root = &tree->buffer; + struct rb_node ** p = &root->rb_node; + struct rb_node * parent = NULL; + struct extent_buffer *eb; + + while(*p) { + parent = *p; + eb = rb_entry(parent, struct extent_buffer, rb_node); + + if (offset < eb->start) + p = &(*p)->rb_left; + else if (offset > eb->start) + p = &(*p)->rb_right; + else + return eb; + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); + return NULL; +} + +static struct extent_buffer *buffer_search(struct extent_io_tree *tree, + u64 offset) +{ + struct rb_root *root = &tree->buffer; + struct rb_node * n = root->rb_node; + struct extent_buffer *eb; + + while(n) { + eb = rb_entry(n, struct extent_buffer, rb_node); + if (offset < eb->start) + n = n->rb_left; + else if (offset > eb->start) + n = n->rb_right; + else + return eb; + } + return NULL; +} + /* * utility function to look for merge candidates inside a given range. * Any extents with matching state are merged together into a single @@ -1817,9 +1848,8 @@ void set_page_extent_mapped(struct page *page) { if (!PagePrivate(page)) { SetPagePrivate(page); - WARN_ON(!page->mapping->a_ops->invalidatepage); - set_page_private(page, EXTENT_PAGE_PRIVATE); page_cache_get(page); + set_page_private(page, EXTENT_PAGE_PRIVATE); } } @@ -2627,51 +2657,6 @@ out: return sector; } -static int add_lru(struct extent_io_tree *tree, struct extent_buffer *eb) -{ - if (list_empty(&eb->lru)) { - extent_buffer_get(eb); - list_add(&eb->lru, &tree->buffer_lru); - tree->lru_size++; - if (tree->lru_size >= BUFFER_LRU_MAX) { - struct extent_buffer *rm; - rm = list_entry(tree->buffer_lru.prev, - struct extent_buffer, lru); - tree->lru_size--; - list_del_init(&rm->lru); - free_extent_buffer(rm); - } - } else - list_move(&eb->lru, &tree->buffer_lru); - return 0; -} -static struct extent_buffer *find_lru(struct extent_io_tree *tree, - u64 start, unsigned long len) -{ - struct list_head *lru = &tree->buffer_lru; - struct list_head *cur = lru->next; - struct extent_buffer *eb; - - if (list_empty(lru)) - return NULL; - - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start == start && eb->len == len) { - extent_buffer_get(eb); - return eb; - } - cur = cur->next; - } while (cur != lru); - return NULL; -} - -static inline unsigned long num_extent_pages(u64 start, u64 len) -{ - return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - - (start >> PAGE_CACHE_SHIFT); -} - static inline struct page *extent_buffer_page(struct extent_buffer *eb, unsigned long i) { @@ -2688,44 +2673,10 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, return p; } -int release_extent_buffer_tail_pages(struct extent_buffer *eb) -{ - unsigned long num_pages = num_extent_pages(eb->start, eb->len); - struct page *page; - unsigned long i; - - if (num_pages == 1) - return 0; - for (i = 1; i < num_pages; i++) { - page = extent_buffer_page(eb, i); - page_cache_release(page); - } - return 0; -} - - -int invalidate_extent_lru(struct extent_io_tree *tree, u64 start, - unsigned long len) +static inline unsigned long num_extent_pages(u64 start, u64 len) { - struct list_head *lru = &tree->buffer_lru; - struct list_head *cur = lru->next; - struct extent_buffer *eb; - int found = 0; - - spin_lock(&tree->lru_lock); - if (list_empty(lru)) - goto out; - - do { - eb = list_entry(cur, struct extent_buffer, lru); - if (eb->start <= start && eb->start + eb->len > start) { - eb->flags &= ~EXTENT_UPTODATE; - } - cur = cur->next; - } while (cur != lru); -out: - spin_unlock(&tree->lru_lock); - return found; + return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - + (start >> PAGE_CACHE_SHIFT); } static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, @@ -2736,15 +2687,7 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, struct extent_buffer *eb = NULL; unsigned long flags; - spin_lock(&tree->lru_lock); - eb = find_lru(tree, start, len); - spin_unlock(&tree->lru_lock); - if (eb) { - return eb; - } - eb = kmem_cache_zalloc(extent_buffer_cache, mask); - INIT_LIST_HEAD(&eb->lru); eb->start = start; eb->len = len; spin_lock_irqsave(&leak_lock, flags); @@ -2773,17 +2716,24 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; + struct extent_buffer *exists = NULL; struct page *p; struct address_space *mapping = tree->mapping; int uptodate = 1; + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (eb) { + atomic_inc(&eb->refs); + spin_unlock(&tree->buffer_lock); + return eb; + } + spin_unlock(&tree->buffer_lock); + eb = __alloc_extent_buffer(tree, start, len, mask); if (!eb) return NULL; - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - if (page0) { eb->first_page = page0; i = 1; @@ -2800,7 +2750,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); if (!p) { WARN_ON(1); - goto fail; + goto free_eb; } set_page_extent_mapped(p); mark_page_accessed(p); @@ -2818,25 +2768,28 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, eb->flags |= EXTENT_UPTODATE; eb->flags |= EXTENT_BUFFER_FILLED; -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); + spin_lock(&tree->buffer_lock); + exists = buffer_tree_insert(tree, start, &eb->rb_node); + if (exists) { + /* add one reference for the caller */ + atomic_inc(&exists->refs); + spin_unlock(&tree->buffer_lock); + goto free_eb; + } + spin_unlock(&tree->buffer_lock); + + /* add one reference for the tree */ + atomic_inc(&eb->refs); return eb; -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); +free_eb: if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { + return exists; + for (index = 1; index < i; index++) page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); + page_cache_release(extent_buffer_page(eb, 0)); __free_extent_buffer(eb); - return NULL; + return exists; } EXPORT_SYMBOL(alloc_extent_buffer); @@ -2844,89 +2797,27 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, u64 start, unsigned long len, gfp_t mask) { - unsigned long num_pages = num_extent_pages(start, len); - unsigned long i; - unsigned long index = start >> PAGE_CACHE_SHIFT; struct extent_buffer *eb; - struct page *p; - struct address_space *mapping = tree->mapping; - int uptodate = 1; - eb = __alloc_extent_buffer(tree, start, len, mask); - if (!eb) - return NULL; - - if (eb->flags & EXTENT_BUFFER_FILLED) - goto lru_add; - - for (i = 0; i < num_pages; i++, index++) { - p = find_get_page(mapping, index); - if (!p) { - goto fail; - } - if (TestSetPageLocked(p)) { - page_cache_release(p); - goto fail; - } - - set_page_extent_mapped(p); - mark_page_accessed(p); - - if (i == 0) { - eb->first_page = p; - set_page_extent_head(p, len); - } else { - set_page_private(p, EXTENT_PAGE_PRIVATE); - } - - if (!PageUptodate(p)) - uptodate = 0; - unlock_page(p); - } - if (uptodate) - eb->flags |= EXTENT_UPTODATE; - eb->flags |= EXTENT_BUFFER_FILLED; + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (eb) + atomic_inc(&eb->refs); + spin_unlock(&tree->buffer_lock); -lru_add: - spin_lock(&tree->lru_lock); - add_lru(tree, eb); - spin_unlock(&tree->lru_lock); return eb; -fail: - spin_lock(&tree->lru_lock); - list_del_init(&eb->lru); - spin_unlock(&tree->lru_lock); - if (!atomic_dec_and_test(&eb->refs)) - return NULL; - for (index = 1; index < i; index++) { - page_cache_release(extent_buffer_page(eb, index)); - } - if (i > 0) - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); - return NULL; } EXPORT_SYMBOL(find_extent_buffer); void free_extent_buffer(struct extent_buffer *eb) { - unsigned long i; - unsigned long num_pages; - if (!eb) return; if (!atomic_dec_and_test(&eb->refs)) return; - WARN_ON(!list_empty(&eb->lru)); - num_pages = num_extent_pages(eb->start, eb->len); - - for (i = 1; i < num_pages; i++) { - page_cache_release(extent_buffer_page(eb, i)); - } - page_cache_release(extent_buffer_page(eb, 0)); - __free_extent_buffer(eb); + WARN_ON(1); } EXPORT_SYMBOL(free_extent_buffer); @@ -3583,3 +3474,35 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, } } EXPORT_SYMBOL(memmove_extent_buffer); + +int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page) +{ + u64 start = page_offset(page); + struct extent_buffer *eb; + int ret = 1; + unsigned long i; + unsigned long num_pages; + + spin_lock(&tree->buffer_lock); + eb = buffer_search(tree, start); + if (!eb) + goto out; + + if (atomic_read(&eb->refs) > 1) { + ret = 0; + goto out; + } + /* at this point we can safely release the extent buffer */ + num_pages = num_extent_pages(eb->start, eb->len); + for (i = 0; i < num_pages; i++) { + struct page *page = extent_buffer_page(eb, i); + page_cache_release(page); + } + rb_erase(&eb->rb_node, &tree->buffer); + __free_extent_buffer(eb); +out: + spin_unlock(&tree->buffer_lock); + return ret; +} +EXPORT_SYMBOL(try_release_extent_buffer); + diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 23affd27af5..dd367617d78 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -54,13 +54,12 @@ struct extent_io_ops { struct extent_io_tree { struct rb_root state; + struct rb_root buffer; struct address_space *mapping; u64 dirty_bytes; spinlock_t lock; + spinlock_t buffer_lock; struct extent_io_ops *ops; - spinlock_t lru_lock; - struct list_head buffer_lru; - int lru_size; struct extent_state *last; }; @@ -87,10 +86,10 @@ struct extent_buffer { unsigned long map_start; unsigned long map_len; struct page *first_page; - struct list_head lru; atomic_t refs; int flags; struct list_head leak_list; + struct rb_node rb_node; }; struct extent_map_tree; @@ -112,10 +111,10 @@ typedef struct extent_map *(get_extent_t)(struct inode *inode, void extent_io_tree_init(struct extent_io_tree *tree, struct address_space *mapping, gfp_t mask); -void extent_io_tree_empty_lru(struct extent_io_tree *tree); int try_release_extent_mapping(struct extent_map_tree *map, struct extent_io_tree *tree, struct page *page, gfp_t mask); +int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page); int try_release_extent_state(struct extent_map_tree *map, struct extent_io_tree *tree, struct page *page, gfp_t mask); @@ -241,8 +240,6 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset, unsigned long *map_start, unsigned long *map_len, int km); void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); -int invalidate_extent_lru(struct extent_io_tree *tree, u64 start, - unsigned long len); int release_extent_buffer_tail_pages(struct extent_buffer *eb); int extent_range_uptodate(struct extent_io_tree *tree, u64 start, u64 end); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8fb6dc25e7a..60852ada658 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2670,7 +2670,6 @@ static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags) map = &BTRFS_I(page->mapping->host)->extent_tree; ret = try_release_extent_mapping(map, tree, page, gfp_flags); if (ret == 1) { - invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); @@ -2721,8 +2720,6 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset) ClearPageChecked(page); if (PagePrivate(page)) { - invalidate_extent_lru(tree, page_offset(page), - PAGE_CACHE_SIZE); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); -- cgit v1.2.3-70-g09d2 From f421950f86bf96a11fef932e167ab2e70d4c43a0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 11:18:09 -0400 Subject: Btrfs: Fix some data=ordered related data corruptions Stress testing was showing data checksum errors, most of which were caused by a lookup bug in the extent_map tree. The tree was caching the last pointer returned, and searches would check the last pointer first. But, search callers also expect the search to return the very first matching extent in the range, which wasn't always true with the last pointer usage. For now, the code to cache the last return value is just removed. It is easy to fix, but I think lookups are rare enough that it isn't required anymore. This commit also replaces do_sync_mapping_range with a local copy of the related functions. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 + fs/btrfs/extent_io.c | 20 --------- fs/btrfs/extent_io.h | 1 - fs/btrfs/extent_map.c | 9 ---- fs/btrfs/extent_map.h | 1 - fs/btrfs/file.c | 15 +++---- fs/btrfs/inode.c | 54 +++++++++++++---------- fs/btrfs/ordered-data.c | 115 ++++++++++++++++++++++++++++++++++++++++-------- fs/btrfs/ordered-data.h | 4 ++ fs/btrfs/transaction.c | 2 - 10 files changed, 140 insertions(+), 83 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 96ab2797c09..f8fccdac305 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1590,6 +1590,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); /* inode.c */ +int btrfs_writepages(struct address_space *mapping, + struct writeback_control *wbc); int btrfs_create_subvol_root(struct btrfs_root *new_root, struct btrfs_trans_handle *trans, u64 new_dirid, struct btrfs_block_group_cache *block_group); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7380449cb5b..9965993748d 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -97,7 +97,6 @@ void extent_io_tree_init(struct extent_io_tree *tree, spin_lock_init(&tree->lock); spin_lock_init(&tree->buffer_lock); tree->mapping = mapping; - tree->last = NULL; } EXPORT_SYMBOL(extent_io_tree_init); @@ -173,12 +172,6 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, struct tree_entry *entry; struct tree_entry *prev_entry = NULL; - if (tree->last) { - struct extent_state *state; - state = tree->last; - if (state->start <= offset && offset <= state->end) - return &tree->last->rb_node; - } while(n) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; @@ -189,7 +182,6 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, else if (offset > entry->end) n = n->rb_right; else { - tree->last = rb_entry(n, struct extent_state, rb_node); return n; } } @@ -223,10 +215,6 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, ret = __etree_search(tree, offset, &prev, NULL); if (!ret) { - if (prev) { - tree->last = rb_entry(prev, struct extent_state, - rb_node); - } return prev; } return ret; @@ -301,8 +289,6 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { state->start = other->start; other->tree = NULL; - if (tree->last == other) - tree->last = state; rb_erase(&other->rb_node, &tree->state); free_extent_state(other); } @@ -314,8 +300,6 @@ static int merge_state(struct extent_io_tree *tree, other->state == state->state) { other->start = state->start; state->tree = NULL; - if (tree->last == state) - tree->last = other; rb_erase(&state->rb_node, &tree->state); free_extent_state(state); } @@ -378,7 +362,6 @@ static int insert_state(struct extent_io_tree *tree, return -EEXIST; } state->tree = tree; - tree->last = state; merge_state(tree, state); return 0; } @@ -444,9 +427,6 @@ static int clear_state_bit(struct extent_io_tree *tree, if (delete || state->state == 0) { if (state->tree) { clear_state_cb(tree, state, state->state); - if (tree->last == state) { - tree->last = extent_state_next(state); - } rb_erase(&state->rb_node, &tree->state); state->tree = NULL; free_extent_state(state); diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 6c03e6a1993..315cfceae31 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -60,7 +60,6 @@ struct extent_io_tree { spinlock_t lock; spinlock_t buffer_lock; struct extent_io_ops *ops; - struct extent_state *last; }; struct extent_state { diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 71b1ac15535..8a502ee2f23 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -42,7 +42,6 @@ void extent_map_exit(void) void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask) { tree->map.rb_node = NULL; - tree->last = NULL; spin_lock_init(&tree->lock); } EXPORT_SYMBOL(extent_map_tree_init); @@ -239,7 +238,6 @@ int add_extent_mapping(struct extent_map_tree *tree, merge->in_tree = 0; free_extent_map(merge); } - tree->last = em; out: return ret; } @@ -273,10 +271,6 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, u64 end = range_end(start, len); BUG_ON(spin_trylock(&tree->lock)); - em = tree->last; - if (em && end > em->start && start < extent_map_end(em)) - goto found; - rb_node = __tree_search(&tree->map, start, &prev, &next); if (!rb_node && prev) { em = rb_entry(prev, struct extent_map, rb_node); @@ -305,7 +299,6 @@ struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree, found: atomic_inc(&em->refs); - tree->last = em; out: return em; } @@ -327,8 +320,6 @@ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em) BUG_ON(spin_trylock(&tree->lock)); rb_erase(&em->rb_node, &tree->map); em->in_tree = 0; - if (tree->last == em) - tree->last = NULL; return ret; } EXPORT_SYMBOL(remove_extent_mapping); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index a3978ec2784..26ac6fe0b26 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -26,7 +26,6 @@ struct extent_map { struct extent_map_tree { struct rb_root map; - struct extent_map *last; spinlock_t lock; }; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 591a30208ac..e5ffb66ad32 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -381,14 +381,13 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) break; } if (test_bit(EXTENT_FLAG_PINNED, &em->flags)) { - start = em->start + em->len; - free_extent_map(em); - spin_unlock(&em_tree->lock); - if (start < end) { - len = end - start + 1; - continue; - } - break; + printk(KERN_CRIT "inode %lu trying to drop pinned " + "extent start %llu end %llu, em [%llu %llu]\n", + inode->i_ino, + (unsigned long long)start, + (unsigned long long)end, + (unsigned long long)em->start, + (unsigned long long)em->len); } remove_extent_mapping(em_tree, em); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 60852ada658..3da12a4d913 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -485,7 +485,7 @@ int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) fixup = kzalloc(sizeof(*fixup), GFP_NOFS); if (!fixup) return -EAGAIN; -printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page)); + SetPageChecked(page); page_cache_get(page); fixup->work.func = btrfs_writepage_fixup_worker; @@ -502,11 +502,13 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct extent_map *em; + struct extent_map *em_orig; u64 alloc_hint = 0; u64 clear_start; u64 clear_end; struct list_head list; struct btrfs_key ins; + struct rb_node *rb; int ret; ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); @@ -535,6 +537,22 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) mutex_lock(&BTRFS_I(inode)->extent_mutex); + spin_lock(&em_tree->lock); + clear_start = ordered_extent->file_offset; + clear_end = ordered_extent->file_offset + ordered_extent->len; + em = lookup_extent_mapping(em_tree, clear_start, + ordered_extent->len); + em_orig = em; + while(em && clear_start < extent_map_end(em) && clear_end > em->start) { + clear_bit(EXTENT_FLAG_PINNED, &em->flags); + rb = rb_next(&em->rb_node); + if (!rb) + break; + em = rb_entry(rb, struct extent_map, rb_node); + } + free_extent_map(em_orig); + spin_unlock(&em_tree->lock); + ret = btrfs_drop_extents(trans, root, inode, ordered_extent->file_offset, ordered_extent->file_offset + @@ -548,22 +566,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ordered_extent->len, 0); BUG_ON(ret); - spin_lock(&em_tree->lock); - clear_start = ordered_extent->file_offset; - clear_end = ordered_extent->file_offset + ordered_extent->len; - while(clear_start < clear_end) { - em = lookup_extent_mapping(em_tree, clear_start, - clear_end - clear_start); - if (em) { - clear_bit(EXTENT_FLAG_PINNED, &em->flags); - clear_start = em->start + em->len; - free_extent_map(em); - } else { - break; - } - } - spin_unlock(&em_tree->lock); - btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1); @@ -2318,7 +2320,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, u64 extent_end = 0; u64 objectid = inode->i_ino; u32 found_type; - struct btrfs_path *path; + struct btrfs_path *path = NULL; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_file_extent_item *item; struct extent_buffer *leaf; @@ -2328,9 +2330,6 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct btrfs_trans_handle *trans = NULL; - path = btrfs_alloc_path(); - BUG_ON(!path); - again: spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, len); @@ -2354,6 +2353,12 @@ again: em->bdev = root->fs_info->fs_devices->latest_bdev; em->start = EXTENT_MAP_HOLE; em->len = (u64)-1; + + if (!path) { + path = btrfs_alloc_path(); + BUG_ON(!path); + } + ret = btrfs_lookup_file_extent(trans, root, path, objectid, start, trans != NULL); if (ret < 0) { @@ -2530,7 +2535,8 @@ insert: } spin_unlock(&em_tree->lock); out: - btrfs_free_path(path); + if (path) + btrfs_free_path(path); if (trans) { ret = btrfs_end_transaction(trans, root); if (!err) { @@ -2643,8 +2649,8 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc) return extent_write_full_page(tree, page, btrfs_get_extent, wbc); } -static int btrfs_writepages(struct address_space *mapping, - struct writeback_control *wbc) +int btrfs_writepages(struct address_space *mapping, + struct writeback_control *wbc) { struct extent_io_tree *tree; tree = &BTRFS_I(mapping->host)->io_tree; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 0d87795fdd8..830dbaea685 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -19,6 +19,8 @@ #include #include #include +#include +#include #include "ctree.h" #include "transaction.h" #include "btrfs_inode.h" @@ -307,12 +309,7 @@ void btrfs_start_ordered_extent(struct inode *inode, * start IO on any dirty ones so the wait doesn't stall waiting * for pdflush to find them */ -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) - do_sync_file_range(file, start, end, SYNC_FILE_RANGE_WRITE); -#else - do_sync_mapping_range(inode->i_mapping, start, end, - SYNC_FILE_RANGE_WRITE); -#endif + btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_NONE); if (wait) wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); @@ -327,28 +324,26 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) u64 orig_end; u64 wait_end; struct btrfs_ordered_extent *ordered; - u64 mask = BTRFS_I(inode)->root->sectorsize - 1; if (start + len < start) { - wait_end = (inode->i_size + mask) & ~mask; - orig_end = (u64)-1; + orig_end = INT_LIMIT(loff_t); } else { orig_end = start + len - 1; - wait_end = orig_end; + if (orig_end > INT_LIMIT(loff_t)) + orig_end = INT_LIMIT(loff_t); } + wait_end = orig_end; again: /* start IO across the range first to instantiate any delalloc * extents */ -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) - do_sync_file_range(file, start, wait_end, SYNC_FILE_RANGE_WRITE); -#else - do_sync_mapping_range(inode->i_mapping, start, wait_end, - SYNC_FILE_RANGE_WRITE); -#endif - end = orig_end; - wait_on_extent_writeback(&BTRFS_I(inode)->io_tree, start, orig_end); + btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_NONE); + + btrfs_wait_on_page_writeback_range(inode->i_mapping, + start >> PAGE_CACHE_SHIFT, + orig_end >> PAGE_CACHE_SHIFT); + end = orig_end; while(1) { ordered = btrfs_lookup_first_ordered_extent(inode, end); if (!ordered) { @@ -565,3 +560,87 @@ out: return ret; } + +/** + * taken from mm/filemap.c because it isn't exported + * + * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range + * @mapping: address space structure to write + * @start: offset in bytes where the range starts + * @end: offset in bytes where the range ends (inclusive) + * @sync_mode: enable synchronous operation + * + * Start writeback against all of a mapping's dirty pages that lie + * within the byte offsets inclusive. + * + * If sync_mode is WB_SYNC_ALL then this is a "data integrity" operation, as + * opposed to a regular memory cleansing writeback. The difference between + * these two operations is that if a dirty page/buffer is encountered, it must + * be waited upon, and not just skipped over. + */ +int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, + loff_t end, int sync_mode) +{ + struct writeback_control wbc = { + .sync_mode = sync_mode, + .nr_to_write = mapping->nrpages * 2, + .range_start = start, + .range_end = end, + .for_writepages = 1, + }; + return btrfs_writepages(mapping, &wbc); +} + +/** + * taken from mm/filemap.c because it isn't exported + * + * wait_on_page_writeback_range - wait for writeback to complete + * @mapping: target address_space + * @start: beginning page index + * @end: ending page index + * + * Wait for writeback to complete against pages indexed by start->end + * inclusive + */ +int btrfs_wait_on_page_writeback_range(struct address_space *mapping, + pgoff_t start, pgoff_t end) +{ + struct pagevec pvec; + int nr_pages; + int ret = 0; + pgoff_t index; + + if (end < start) + return 0; + + pagevec_init(&pvec, 0); + index = start; + while ((index <= end) && + (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, + PAGECACHE_TAG_WRITEBACK, + min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) { + unsigned i; + + for (i = 0; i < nr_pages; i++) { + struct page *page = pvec.pages[i]; + + /* until radix tree lookup accepts end_index */ + if (page->index > end) + continue; + + wait_on_page_writeback(page); + if (PageError(page)) + ret = -EIO; + } + pagevec_release(&pvec); + cond_resched(); + } + + /* Check for outstanding write errors */ + if (test_and_clear_bit(AS_ENOSPC, &mapping->flags)) + ret = -ENOSPC; + if (test_and_clear_bit(AS_EIO, &mapping->flags)) + ret = -EIO; + + return ret; +} diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 1794efd13ca..8e8e3c0404f 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -132,4 +132,8 @@ btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); int btrfs_ordered_update_i_size(struct inode *inode, struct btrfs_ordered_extent *ordered); int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum); +int btrfs_wait_on_page_writeback_range(struct address_space *mapping, + pgoff_t start, pgoff_t end); +int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, + loff_t end, int sync_mode); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 05823904ecb..38c75a0256c 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -649,7 +649,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, extent_io_tree_init(pinned_copy, root->fs_info->btree_inode->i_mapping, GFP_NOFS); -printk("commit trans %Lu\n", trans->transid); trans->transaction->in_commit = 1; trans->transaction->blocked = 1; cur_trans = trans->transaction; @@ -745,7 +744,6 @@ printk("commit trans %Lu\n", trans->transid); list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); mutex_unlock(&root->fs_info->trans_mutex); -printk("done commit trans %Lu\n", trans->transid); kmem_cache_free(btrfs_trans_handle_cachep, trans); if (root->fs_info->closing) { -- cgit v1.2.3-70-g09d2 From ed98b56a6393c5e150fd5095b9eb7fd7d3cfb041 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 22 Jul 2008 23:06:42 -0400 Subject: Btrfs: Take the csum mutex while reading checksums Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 8 +++++--- fs/btrfs/inode.c | 2 ++ fs/btrfs/ordered-data.c | 2 +- fs/btrfs/ordered-data.h | 2 +- fs/btrfs/transaction.c | 3 +++ 5 files changed, 12 insertions(+), 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 45127e4797c..afe42d00b5a 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -152,7 +152,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, if (!sums) return -ENOMEM; - sector_sum = &sums->sums; + sector_sum = sums->sums; sums->file_offset = page_offset(bvec->bv_page) + bvec->bv_offset; sums->len = bio->bi_size; INIT_LIST_HEAD(&sums->list); @@ -174,7 +174,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, sums = kzalloc(btrfs_ordered_sum_size(root, bytes_left), GFP_NOFS); BUG_ON(!sums); - sector_sum = &sums->sums; + sector_sum = sums->sums; sums->len = bytes_left; sums->file_offset = offset; ordered = btrfs_lookup_ordered_extent(inode, @@ -193,12 +193,14 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, (char *)§or_sum->sum); sector_sum->offset = page_offset(bvec->bv_page) + bvec->bv_offset; + sector_sum++; bio_index++; total_bytes += bvec->bv_len; this_sum_bytes += bvec->bv_len; bvec++; } + this_sum_bytes = 0; btrfs_add_ordered_sum(inode, ordered, sums); btrfs_put_ordered_extent(ordered); return 0; @@ -231,7 +233,7 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); BUG_ON(!path); - sector_sum = &sums->sums; + sector_sum = sums->sums; again: next_offset = (u64)-1; found_next = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3da12a4d913..28e667052ec 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -612,6 +612,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) return 0; path = btrfs_alloc_path(); + mutex_lock(&BTRFS_I(inode)->csum_mutex); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); if (IS_ERR(item)) { /* @@ -640,6 +641,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) found: set_state_private(io_tree, start, csum); out: + mutex_unlock(&BTRFS_I(inode)->csum_mutex); if (path) btrfs_free_path(path); return ret; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 830dbaea685..b695f5b29c4 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -545,7 +545,7 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); if (offset >= ordered_sum->file_offset) { num_sectors = ordered_sum->len / sectorsize; - sector_sums = &ordered_sum->sums; + sector_sums = ordered_sum->sums; for (i = 0; i < num_sectors; i++) { if (sector_sums[i].offset == offset) { *sum = sector_sums[i].sum; diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 8e8e3c0404f..36e63f1f79b 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -46,7 +46,7 @@ struct btrfs_ordered_sum { unsigned long len; struct list_head list; /* last field is a variable length array of btrfs_sector_sums */ - struct btrfs_sector_sum sums; + struct btrfs_sector_sum sums[]; }; /* diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 38c75a0256c..0f756e0175c 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -382,6 +382,9 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, memcpy(dirty->root, root, sizeof(*root)); dirty->root->node = root->commit_root; dirty->latest_root = root; + spin_lock_init(&dirty->root->node_lock); + mutex_init(&dirty->root->objectid_mutex); + root->commit_root = NULL; root->root_key.offset = root->fs_info->generation; -- cgit v1.2.3-70-g09d2 From 89642229a582a5c2b6d2ed8ec16986387d9a9047 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 24 Jul 2008 09:41:53 -0400 Subject: Btrfs: Search data ordered extents first for checksums on read Checksum items are not inserted into the tree until all of the io from a given extent is complete. This means one dirty page from an extent may be written, freed, and then read again before the entire extent is on disk and the checksum item is inserted. The checksums themselves are stored in the ordered extent so they can be inserted in bulk when IO is complete. On read, if a checksum item isn't found, the ordered extents were being searched for a checksum record. This all worked most of the time, but the checksum insertion code tries to reduce the number of tree operations by pre-inserting checksum items based on i_size and a few other factors. This means the read code might find a checksum item that hasn't yet really been filled in. This commit changes things to check the ordered extents first and only dive into the btree if nothing was found. This removes the need for extra locking and is more reliable. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 9 ++++++++- fs/btrfs/extent_io.c | 8 ++++---- fs/btrfs/inode.c | 33 ++++++++++++++++++--------------- fs/btrfs/ordered-data.c | 1 + 4 files changed, 31 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 86e84a8579e..7ce3f83c5dd 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1011,9 +1011,16 @@ void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); spin_unlock(&em_tree->lock); - if (!em) + if (!em) { + __unplug_io_fn(bdi, page); return; + } + if (em->block_start >= EXTENT_MAP_LAST_BYTE) { + free_extent_map(em); + __unplug_io_fn(bdi, page); + return; + } offset = offset - em->start; btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, em->block_start + offset, page); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 9965993748d..e3547a992d5 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1949,18 +1949,18 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); cur + iosize - 1); } if (!ret) { - unsigned long nr = (last_byte >> PAGE_CACHE_SHIFT) + 1; - nr -= page->index; + unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; + pnr -= page->index; ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, - bdev, bio, nr, + bdev, bio, pnr, end_bio_extent_readpage, mirror_num); + nr++; } if (ret) SetPageError(page); cur = cur + iosize; page_offset += iosize; - nr++; } if (!nr) { if (!PageError(page)) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 28e667052ec..0e90315ea80 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -611,22 +611,25 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) btrfs_test_flag(inode, NODATASUM)) return 0; + /* + * It is possible there is an ordered extent that has + * not yet finished for this range in the file. If so, + * that extent will have a csum cached, and it will insert + * the sum after all the blocks in the extent are fully + * on disk. So, look for an ordered extent and use the + * sum if found. We have to do this before looking in the + * btree because csum items are pre-inserted based on + * the file size. btrfs_lookup_csum might find an item + * that still hasn't been fully filled. + */ + ret = btrfs_find_ordered_sum(inode, start, &csum); + if (ret == 0) + goto found; + + ret = 0; path = btrfs_alloc_path(); - mutex_lock(&BTRFS_I(inode)->csum_mutex); item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); if (IS_ERR(item)) { - /* - * It is possible there is an ordered extent that has - * not yet finished for this range in the file. If so, - * that extent will have a csum cached, and it will insert - * the sum after all the blocks in the extent are fully - * on disk. So, look for an ordered extent and use the - * sum if found. - */ - ret = btrfs_find_ordered_sum(inode, start, &csum); - if (ret == 0) - goto found; - ret = PTR_ERR(item); /* a csum that isn't present is a preallocated region. */ if (ret == -ENOENT || ret == -EFBIG) @@ -641,7 +644,6 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) found: set_state_private(io_tree, start, csum); out: - mutex_unlock(&BTRFS_I(inode)->csum_mutex); if (path) btrfs_free_path(path); return ret; @@ -1375,7 +1377,7 @@ again: } if (!PageUptodate(page)) { ret = -EIO; - goto out; + goto out_unlock; } } wait_on_page_writeback(page); @@ -1406,6 +1408,7 @@ again: set_page_dirty(page); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); +out_unlock: unlock_page(page); page_cache_release(page); out: diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index b695f5b29c4..e42fd233e04 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -557,6 +557,7 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) } out: mutex_unlock(&tree->mutex); + btrfs_put_ordered_extent(ordered); return ret; } -- cgit v1.2.3-70-g09d2 From aec7477b3b0e8ec93f6d274f25ba40b0665134d4 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Thu, 24 Jul 2008 12:12:38 -0400 Subject: Btrfs: Implement new dir index format Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 15 ++++++ fs/btrfs/ctree.h | 8 +-- fs/btrfs/dir-item.c | 4 +- fs/btrfs/inode-item.c | 10 +++- fs/btrfs/inode.c | 143 +++++++++++++++++++++++++++++++++++++++++-------- fs/btrfs/ioctl.c | 4 +- fs/btrfs/transaction.c | 4 +- 7 files changed, 155 insertions(+), 33 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 3bf40591742..b7bd60e4fdd 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -43,6 +43,21 @@ struct btrfs_inode { u64 delalloc_bytes; u64 disk_i_size; u32 flags; + + /* + * if this is a directory then index_cnt is the counter for the index + * number for new files that are created + */ + u64 index_cnt; + + /* + * index holds the directory index for this inode on creation, so + * add_link can do what its supposed to. This isn't populated when the + * inode is read because there isn't really a reason to know this unless + * we are creating the directory index or deleting it, and deletion + * reads the index off of the inode reference at unlink time. + */ + u64 index; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 6675e916ebc..beb05b1de54 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -372,6 +372,7 @@ struct btrfs_dev_extent { } __attribute__ ((__packed__)); struct btrfs_inode_ref { + __le64 index; __le16 name_len; /* name goes here */ } __attribute__ ((__packed__)); @@ -902,6 +903,7 @@ BTRFS_SETGET_STACK_FUNCS(block_group_flags, /* struct btrfs_inode_ref */ BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16); +BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64); /* struct btrfs_inode_item */ BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); @@ -1528,7 +1530,7 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, /* dir-item.c */ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, u64 dir, - struct btrfs_key *location, u8 type); + struct btrfs_key *location, u8 type, u64 index); struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 dir, @@ -1566,11 +1568,11 @@ int btrfs_find_highest_inode(struct btrfs_root *fs_root, u64 *objectid); int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, - u64 inode_objectid, u64 ref_objectid); + u64 inode_objectid, u64 ref_objectid, u64 index); int btrfs_del_inode_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, - u64 inode_objectid, u64 ref_objectid); + u64 inode_objectid, u64 ref_objectid, u64 *index); int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid); diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 7a73dc59dc4..eb4dd3d75cf 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -110,7 +110,7 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, u64 dir, - struct btrfs_key *location, u8 type) + struct btrfs_key *location, u8 type, u64 index) { int ret = 0; int ret2 = 0; @@ -156,7 +156,7 @@ second_insert: btrfs_release_path(root, path); btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY); - key.offset = location->objectid; + key.offset = index; dir_item = insert_with_overflow(trans, root, path, &key, data_size, name, name_len); if (IS_ERR(dir_item)) { diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index cba30b6cc6f..d93451c66ba 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -52,7 +52,7 @@ int find_name_in_backref(struct btrfs_path *path, const char * name, int btrfs_del_inode_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, - u64 inode_objectid, u64 ref_objectid) + u64 inode_objectid, u64 ref_objectid, u64 *index) { struct btrfs_path *path; struct btrfs_key key; @@ -86,6 +86,10 @@ int btrfs_del_inode_ref(struct btrfs_trans_handle *trans, } leaf = path->nodes[0]; item_size = btrfs_item_size_nr(leaf, path->slots[0]); + + if (index) + *index = btrfs_inode_ref_index(leaf, ref); + if (del_len == item_size) { ret = btrfs_del_item(trans, root, path); goto out; @@ -106,7 +110,7 @@ out: int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, - u64 inode_objectid, u64 ref_objectid) + u64 inode_objectid, u64 ref_objectid, u64 index) { struct btrfs_path *path; struct btrfs_key key; @@ -138,6 +142,7 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, struct btrfs_inode_ref); ref = (struct btrfs_inode_ref *)((unsigned long)ref + old_size); btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + btrfs_set_inode_ref_index(path->nodes[0], ref, index); ptr = (unsigned long)(ref + 1); ret = 0; } else if (ret < 0) { @@ -146,6 +151,7 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans, ref = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_inode_ref); btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + btrfs_set_inode_ref_index(path->nodes[0], ref, index); ptr = (unsigned long)(ref + 1); } write_extent_buffer(path->nodes[0], name, ptr, name_len); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0e90315ea80..8d371d6fe55 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -872,6 +872,8 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_rdev = 0; rdev = btrfs_inode_rdev(leaf, inode_item); + BTRFS_I(inode)->index_cnt = (u64)-1; + alloc_group_block = btrfs_inode_block_group(leaf, inode_item); BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, alloc_group_block); @@ -993,6 +995,7 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, struct extent_buffer *leaf; struct btrfs_dir_item *di; struct btrfs_key key; + u64 index; path = btrfs_alloc_path(); if (!path) { @@ -1017,8 +1020,19 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, goto err; btrfs_release_path(root, path); + ret = btrfs_del_inode_ref(trans, root, name, name_len, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino, &index); + if (ret) { + printk("failed to delete reference to %.*s, " + "inode %lu parent %lu\n", name_len, name, + dentry->d_inode->i_ino, + dentry->d_parent->d_inode->i_ino); + goto err; + } + di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, - key.objectid, name, name_len, -1); + index, name, name_len, -1); if (IS_ERR(di)) { ret = PTR_ERR(di); goto err; @@ -1031,15 +1045,6 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, btrfs_release_path(root, path); dentry->d_inode->i_ctime = dir->i_ctime; - ret = btrfs_del_inode_ref(trans, root, name, name_len, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino); - if (ret) { - printk("failed to delete reference to %.*s, " - "inode %lu parent %lu\n", name_len, name, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino); - } err: btrfs_free_path(path); if (!ret) { @@ -1625,6 +1630,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) BTRFS_I(inode)->root = args->root; BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; + BTRFS_I(inode)->index_cnt = (u64)-1; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); @@ -1901,8 +1907,77 @@ void btrfs_dirty_inode(struct inode *inode) btrfs_end_transaction(trans, root); } +static int btrfs_set_inode_index_count(struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_key key, found_key; + struct btrfs_path *path; + struct extent_buffer *leaf; + int ret; + + key.objectid = inode->i_ino; + btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY); + key.offset = (u64)-1; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto out; + /* FIXME: we should be able to handle this */ + if (ret == 0) + goto out; + ret = 0; + + /* + * MAGIC NUMBER EXPLANATION: + * since we search a directory based on f_pos we have to start at 2 + * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody + * else has to start at 2 + */ + if (path->slots[0] == 0) { + BTRFS_I(inode)->index_cnt = 2; + goto out; + } + + path->slots[0]--; + + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + if (found_key.objectid != inode->i_ino || + btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) { + BTRFS_I(inode)->index_cnt = 2; + goto out; + } + + BTRFS_I(inode)->index_cnt = found_key.offset + 1; +out: + btrfs_free_path(path); + return ret; +} + +static int btrfs_set_inode_index(struct inode *dir, struct inode *inode) +{ + int ret = 0; + + if (BTRFS_I(dir)->index_cnt == (u64)-1) { + ret = btrfs_set_inode_index_count(dir); + if (ret) + return ret; + } + + BTRFS_I(inode)->index = BTRFS_I(dir)->index_cnt; + BTRFS_I(dir)->index_cnt++; + + return ret; +} + static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct inode *dir, const char *name, int name_len, u64 ref_objectid, u64 objectid, @@ -1928,6 +2003,20 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, if (!inode) return ERR_PTR(-ENOMEM); + if (dir) { + ret = btrfs_set_inode_index(dir, inode); + if (ret) + return ERR_PTR(ret); + } else { + BTRFS_I(inode)->index = 0; + } + /* + * index_cnt is ignored for everything but a dir, + * btrfs_get_inode_index_count has an explanation for the magic + * number + */ + BTRFS_I(inode)->index_cnt = 2; + extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); @@ -1984,6 +2073,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, struct btrfs_inode_ref); btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); + btrfs_set_inode_ref_index(path->nodes[0], ref, BTRFS_I(inode)->index); ptr = (unsigned long)(ref + 1); write_extent_buffer(path->nodes[0], name, ptr, name_len); @@ -1998,6 +2088,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, insert_inode_hash(inode); return inode; fail: + if (dir) + BTRFS_I(dir)->index_cnt--; btrfs_free_path(path); return ERR_PTR(ret); } @@ -2014,7 +2106,7 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, int ret; struct btrfs_key key; struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root; - struct inode *parent_inode; + struct inode *parent_inode = dentry->d_parent->d_inode; key.objectid = inode->i_ino; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); @@ -2023,16 +2115,17 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, ret = btrfs_insert_dir_item(trans, root, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, - &key, btrfs_inode_type(inode)); + &key, btrfs_inode_type(inode), + BTRFS_I(inode)->index); if (ret == 0) { if (add_backref) { ret = btrfs_insert_inode_ref(trans, root, dentry->d_name.name, dentry->d_name.len, inode->i_ino, - dentry->d_parent->d_inode->i_ino); + parent_inode->i_ino, + BTRFS_I(inode)->index); } - parent_inode = dentry->d_parent->d_inode; btrfs_i_size_write(parent_inode, parent_inode->i_size + dentry->d_name.len * 2); parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; @@ -2083,7 +2176,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, dentry->d_name.name, + inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, mode); @@ -2138,7 +2231,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, dentry->d_name.name, + inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, mode); @@ -2203,10 +2296,15 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; + err = btrfs_set_inode_index(dir, inode); + if (err) + goto fail; + trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); atomic_inc(&inode->i_count); + err = btrfs_add_nondir(trans, dentry, inode, 1); if (err) @@ -2258,7 +2356,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) goto out_unlock; } - inode = btrfs_new_inode(trans, root, dentry->d_name.name, + inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, S_IFDIR | mode); @@ -2886,9 +2984,8 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct btrfs_block_group_cache *block_group) { struct inode *inode; - int ret; - inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid, + inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, new_dirid, block_group, S_IFDIR | 0700); if (IS_ERR(inode)) return PTR_ERR(inode); @@ -2896,8 +2993,6 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, inode->i_fop = &btrfs_dir_file_operations; new_root->inode = inode; - ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid, - new_dirid); inode->i_nlink = 1; btrfs_i_size_write(inode, 0); @@ -3078,6 +3173,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, if (ret) goto out_fail; } + ret = btrfs_set_inode_index(new_dir, old_inode); + if (ret) + goto out_fail; + ret = btrfs_add_link(trans, new_dentry, old_inode, 1); if (ret) goto out_fail; @@ -3123,7 +3222,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, goto out_unlock; } - inode = btrfs_new_inode(trans, root, dentry->d_name.name, + inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index a61f2e7e2db..faf081302d0 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -129,13 +129,13 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, dir = root->fs_info->sb->s_root->d_inode; ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, name, namelen, dir->i_ino, &key, - BTRFS_FT_DIR); + BTRFS_FT_DIR, 0); if (ret) goto fail; ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, name, namelen, objectid, - root->fs_info->sb->s_root->d_inode->i_ino); + root->fs_info->sb->s_root->d_inode->i_ino, 0); if (ret) goto fail; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 0f756e0175c..768b0d223e6 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -577,14 +577,14 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, pending->name, namelen, root->fs_info->sb->s_root->d_inode->i_ino, - &key, BTRFS_FT_DIR); + &key, BTRFS_FT_DIR, 0); if (ret) goto fail; ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, pending->name, strlen(pending->name), objectid, - root->fs_info->sb->s_root->d_inode->i_ino); + root->fs_info->sb->s_root->d_inode->i_ino, 0); /* Invalidate existing dcache entry for new snapshot. */ btrfs_invalidate_dcache_root(root, pending->name, namelen); -- cgit v1.2.3-70-g09d2 From 33268eaf0b3db5e2bd12c0ada81a8e8f87a46d68 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Thu, 24 Jul 2008 12:16:36 -0400 Subject: Btrfs: Add ACL support Signed-off-by: Chris Mason --- fs/btrfs/acl.c | 296 +++++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/btrfs_inode.h | 3 + fs/btrfs/ctree.h | 7 ++ fs/btrfs/inode.c | 42 ++++++- fs/btrfs/super.c | 9 +- fs/btrfs/xattr.c | 26 +++-- 6 files changed, 335 insertions(+), 48 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c index ff0fcc72848..b192659b4f5 100644 --- a/fs/btrfs/acl.c +++ b/fs/btrfs/acl.c @@ -20,76 +20,302 @@ #include #include #include +#include #include + #include "ctree.h" +#include "btrfs_inode.h" #include "xattr.h" -#ifndef is_owner_or_cap -#define is_owner_or_cap(inode) \ - ((current->fsuid == (inode)->i_uid) || capable(CAP_FOWNER)) -#endif + +static void btrfs_update_cached_acl(struct inode *inode, + struct posix_acl **p_acl, + struct posix_acl *acl) +{ + spin_lock(&inode->i_lock); + if (*p_acl && *p_acl != BTRFS_ACL_NOT_CACHED) + posix_acl_release(*p_acl); + *p_acl = posix_acl_dup(acl); + spin_unlock(&inode->i_lock); +} + +static struct posix_acl *btrfs_get_acl(struct inode *inode, int type) +{ + int size, name_index; + char *value = NULL; + struct posix_acl *acl = NULL, **p_acl; + + switch (type) { + case ACL_TYPE_ACCESS: + name_index = BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS; + p_acl = &BTRFS_I(inode)->i_acl; + break; + case ACL_TYPE_DEFAULT: + name_index = BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT; + p_acl = &BTRFS_I(inode)->i_default_acl; + break; + default: + return ERR_PTR(-EINVAL); + } + + spin_lock(&inode->i_lock); + if (*p_acl != BTRFS_ACL_NOT_CACHED) + acl = posix_acl_dup(*p_acl); + spin_unlock(&inode->i_lock); + + if (acl) + return acl; + + + size = btrfs_xattr_get(inode, name_index, "", NULL, 0); + if (size > 0) { + value = kzalloc(size, GFP_NOFS); + if (!value) + return ERR_PTR(-ENOMEM); + size = btrfs_xattr_get(inode, name_index, "", value, size); + if (size > 0) { + acl = posix_acl_from_xattr(value, size); + btrfs_update_cached_acl(inode, p_acl, acl); + } + kfree(value); + } else if (size == -ENOENT) { + acl = NULL; + btrfs_update_cached_acl(inode, p_acl, acl); + } + + return acl; +} + +static int btrfs_xattr_get_acl(struct inode *inode, int type, + void *value, size_t size) +{ + struct posix_acl *acl; + int ret = 0; + + acl = btrfs_get_acl(inode, type); + + if (IS_ERR(acl)) + return PTR_ERR(acl); + if (acl == NULL) + return -ENODATA; + ret = posix_acl_to_xattr(acl, value, size); + posix_acl_release(acl); + + return ret; +} + +/* + * Needs to be called with fs_mutex held + */ +static int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type) +{ + int ret, name_index = 0, size = 0; + struct posix_acl **p_acl; + char *value = NULL; + mode_t mode; + + if (acl) { + ret = posix_acl_valid(acl); + if (ret < 0) + return ret; + ret = 0; + } + + switch (type) { + case ACL_TYPE_ACCESS: + mode = inode->i_mode; + ret = posix_acl_equiv_mode(acl, &mode); + if (ret < 0) + return ret; + ret = 0; + inode->i_mode = mode; + name_index = BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS; + p_acl = &BTRFS_I(inode)->i_acl; + break; + case ACL_TYPE_DEFAULT: + if (!S_ISDIR(inode->i_mode)) + return acl ? -EINVAL : 0; + name_index = BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT; + p_acl = &BTRFS_I(inode)->i_default_acl; + break; + default: + return -EINVAL; + } + + if (acl) { + size = posix_acl_xattr_size(acl->a_count); + value = kmalloc(size, GFP_NOFS); + if (!value) { + ret = -ENOMEM; + goto out; + } + + ret = posix_acl_to_xattr(acl, value, size); + if (ret < 0) + goto out; + } + + ret = btrfs_xattr_set(inode, name_index, "", value, size, 0); + +out: + if (value) + kfree(value); + + if (!ret) + btrfs_update_cached_acl(inode, p_acl, acl); + + return ret; +} static int btrfs_xattr_set_acl(struct inode *inode, int type, const void *value, size_t size) { int ret = 0; - struct posix_acl *acl; + struct posix_acl *acl = NULL; - if (!is_owner_or_cap(inode)) - return -EPERM; if (value) { acl = posix_acl_from_xattr(value, size); if (acl == NULL) { value = NULL; size = 0; } else if (IS_ERR(acl)) { - ret = PTR_ERR(acl); - } else { - ret = posix_acl_valid(acl); - posix_acl_release(acl); + return PTR_ERR(acl); } - if (ret) - return ret; } - return btrfs_xattr_set(inode, type, "", value, size, 0); -} -static int btrfs_xattr_get_acl(struct inode *inode, int type, - void *value, size_t size) -{ - return btrfs_xattr_get(inode, type, "", value, size); + ret = btrfs_set_acl(inode, acl, type); + + posix_acl_release(acl); + + return ret; } + + static int btrfs_xattr_acl_access_get(struct inode *inode, const char *name, void *value, size_t size) { - if (*name != '\0') - return -EINVAL; - return btrfs_xattr_get_acl(inode, BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, - value, size); + return btrfs_xattr_get_acl(inode, ACL_TYPE_ACCESS, value, size); } + static int btrfs_xattr_acl_access_set(struct inode *inode, const char *name, const void *value, size_t size, int flags) { - if (*name != '\0') - return -EINVAL; - return btrfs_xattr_set_acl(inode, BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, - value, size); + return btrfs_xattr_set_acl(inode, ACL_TYPE_ACCESS, value, size); } + static int btrfs_xattr_acl_default_get(struct inode *inode, const char *name, void *value, size_t size) { - if (*name != '\0') - return -EINVAL; - return btrfs_xattr_get_acl(inode, BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, - value, size); + return btrfs_xattr_get_acl(inode, ACL_TYPE_DEFAULT, value, size); } + static int btrfs_xattr_acl_default_set(struct inode *inode, const char *name, const void *value, size_t size, int flags) { - if (*name != '\0') - return -EINVAL; - return btrfs_xattr_set_acl(inode, BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, - value, size); + return btrfs_xattr_set_acl(inode, ACL_TYPE_DEFAULT, value, size); +} + +int btrfs_check_acl(struct inode *inode, int mask) +{ + struct posix_acl *acl; + int error = -EAGAIN; + + acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS); + + if (IS_ERR(acl)) + return PTR_ERR(acl); + if (acl) { + error = posix_acl_permission(inode, acl, mask); + posix_acl_release(acl); + } + + return error; +} + +/* + * btrfs_init_acl is already generally called under fs_mutex, so the locking + * stuff has been fixed to work with that. If the locking stuff changes, we + * need to re-evaluate the acl locking stuff. + */ +int btrfs_init_acl(struct inode *inode, struct inode *dir) +{ + struct posix_acl *acl = NULL; + int ret = 0; + + /* this happens with subvols */ + if (!dir) + return 0; + + if (!S_ISLNK(inode->i_mode)) { + if (IS_POSIXACL(dir)) { + acl = btrfs_get_acl(dir, ACL_TYPE_DEFAULT); + if (IS_ERR(acl)) + return PTR_ERR(acl); + } + + if (!acl) + inode->i_mode &= ~current->fs->umask; + } + + if (IS_POSIXACL(dir) && acl) { + struct posix_acl *clone; + mode_t mode; + + if (S_ISDIR(inode->i_mode)) { + ret = btrfs_set_acl(inode, acl, ACL_TYPE_DEFAULT); + if (ret) + goto failed; + } + clone = posix_acl_clone(acl, GFP_NOFS); + ret = -ENOMEM; + if (!clone) + goto failed; + + mode = inode->i_mode; + ret = posix_acl_create_masq(clone, &mode); + if (ret >= 0) { + inode->i_mode = mode; + if (ret > 0) { + /* we need an acl */ + ret = btrfs_set_acl(inode, clone, + ACL_TYPE_ACCESS); + } + } + } +failed: + posix_acl_release(acl); + + return ret; +} + +int btrfs_acl_chmod(struct inode *inode) +{ + struct posix_acl *acl, *clone; + int ret = 0; + + if (S_ISLNK(inode->i_mode)) + return -EOPNOTSUPP; + + if (!IS_POSIXACL(inode)) + return 0; + + acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS); + if (IS_ERR(acl) || !acl) + return PTR_ERR(acl); + + clone = posix_acl_clone(acl, GFP_KERNEL); + posix_acl_release(acl); + if (!clone) + return -ENOMEM; + + ret = posix_acl_chmod_masq(clone, inode->i_mode); + if (!ret) + ret = btrfs_set_acl(inode, clone, ACL_TYPE_ACCESS); + + posix_acl_release(clone); + + return ret; } + struct xattr_handler btrfs_xattr_acl_default_handler = { .prefix = POSIX_ACL_XATTR_DEFAULT, .list = btrfs_xattr_generic_list, diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index b7bd60e4fdd..9f2a4ef944a 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -36,6 +36,9 @@ struct btrfs_inode { struct inode vfs_inode; struct btrfs_ordered_inode_tree ordered_tree; + struct posix_acl *i_acl; + struct posix_acl *i_default_acl; + /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 53d315bdd16..f87d7263f2d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -42,6 +42,8 @@ struct btrfs_ordered_sum; #define BTRFS_MAGIC "_B5RfS_M" +#define BTRFS_ACL_NOT_CACHED ((void *)-1) + #ifdef CONFIG_LOCKDEP # define BTRFS_MAX_LEVEL 7 #else @@ -1694,4 +1696,9 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size); u64 btrfs_parse_size(char *str); int btrfs_parse_options(struct btrfs_root *root, char *options); int btrfs_sync_fs(struct super_block *sb, int wait); + +/* acl.c */ +int btrfs_check_acl(struct inode *inode, int mask); +int btrfs_init_acl(struct inode *inode, struct inode *dir); +int btrfs_acl_chmod(struct inode *inode); #endif diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8d371d6fe55..2d8853543a7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -36,6 +36,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -1478,6 +1479,9 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) } out: err = inode_setattr(inode, attr); + + if (!err && ((attr->ia_valid & ATTR_MODE))) + err = btrfs_acl_chmod(inode); fail: return err; } @@ -2184,6 +2188,12 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, if (IS_ERR(inode)) goto out_unlock; + err = btrfs_init_acl(inode, dir); + if (err) { + drop_inode = 1; + goto out_unlock; + } + btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) @@ -2239,6 +2249,12 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, if (IS_ERR(inode)) goto out_unlock; + err = btrfs_init_acl(inode, dir); + if (err) { + drop_inode = 1; + goto out_unlock; + } + btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) @@ -2366,6 +2382,11 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) } drop_on_err = 1; + + err = btrfs_init_acl(inode, dir); + if (err) + goto out_fail; + inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; btrfs_set_trans_block_group(trans, inode); @@ -3023,6 +3044,8 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) return NULL; ei->last_trans = 0; btrfs_ordered_inode_tree_init(&ei->ordered_tree); + ei->i_acl = BTRFS_ACL_NOT_CACHED; + ei->i_default_acl = BTRFS_ACL_NOT_CACHED; return &ei->vfs_inode; } @@ -3032,6 +3055,13 @@ void btrfs_destroy_inode(struct inode *inode) WARN_ON(!list_empty(&inode->i_dentry)); WARN_ON(inode->i_data.nrpages); + if (BTRFS_I(inode)->i_acl && + BTRFS_I(inode)->i_acl != BTRFS_ACL_NOT_CACHED) + posix_acl_release(BTRFS_I(inode)->i_acl); + if (BTRFS_I(inode)->i_default_acl && + BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED) + posix_acl_release(BTRFS_I(inode)->i_default_acl); + while(1) { ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); if (!ordered) @@ -3230,6 +3260,12 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, if (IS_ERR(inode)) goto out_unlock; + err = btrfs_init_acl(inode, dir); + if (err) { + drop_inode = 1; + goto out_unlock; + } + btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dentry, inode, 0); if (err) @@ -3310,7 +3346,7 @@ static int btrfs_permission(struct inode *inode, int mask, { if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE)) return -EACCES; - return generic_permission(inode, mask, NULL); + return generic_permission(inode, mask, btrfs_check_acl); } static struct inode_operations btrfs_dir_inode_operations = { @@ -3392,6 +3428,10 @@ static struct inode_operations btrfs_special_inode_operations = { .getattr = btrfs_getattr, .setattr = btrfs_setattr, .permission = btrfs_permission, + .setxattr = generic_setxattr, + .getxattr = generic_getxattr, + .listxattr = btrfs_listxattr, + .removexattr = generic_removexattr, }; static struct inode_operations btrfs_symlink_inode_operations = { .readlink = generic_readlink, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 4cb6aac5122..a6a418b6894 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -68,7 +68,7 @@ static void btrfs_put_super (struct super_block * sb) enum { Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, - Opt_ssd, Opt_thread_pool, Opt_err, + Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_err, }; static match_table_t tokens = { @@ -83,7 +83,8 @@ static match_table_t tokens = { {Opt_alloc_start, "alloc_start=%s"}, {Opt_thread_pool, "thread_pool=%d"}, {Opt_ssd, "ssd"}, - {Opt_err, NULL} + {Opt_noacl, "noacl"}, + {Opt_err, NULL}, }; u64 btrfs_parse_size(char *str) @@ -215,6 +216,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) info->alloc_start); } break; + case Opt_noacl: + root->fs_info->sb->s_flags &= ~MS_POSIXACL; + break; default: break; } @@ -301,6 +305,7 @@ static int btrfs_fill_super(struct super_block * sb, sb->s_op = &btrfs_super_ops; sb->s_xattr = btrfs_xattr_handlers; sb->s_time_gran = 1; + sb->s_flags |= MS_POSIXACL; tree_root = open_ctree(sb, fs_devices, (char *)data); diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c index 6730b595884..121c9550314 100644 --- a/fs/btrfs/xattr.c +++ b/fs/btrfs/xattr.c @@ -26,25 +26,27 @@ #include "transaction.h" #include "xattr.h" #include "disk-io.h" + static struct xattr_handler *btrfs_xattr_handler_map[] = { [BTRFS_XATTR_INDEX_USER] = &btrfs_xattr_user_handler, #ifdef CONFIG_FS_POSIX_ACL -// [BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS] = &btrfs_xattr_acl_access_handler, -// [BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT] = &btrfs_xattr_acl_default_handler, + [BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS] = &btrfs_xattr_acl_access_handler, + [BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT] = &btrfs_xattr_acl_default_handler, #endif [BTRFS_XATTR_INDEX_TRUSTED] = &btrfs_xattr_trusted_handler, [BTRFS_XATTR_INDEX_SECURITY] = &btrfs_xattr_security_handler, -// [BTRFS_XATTR_INDEX_SYSTEM] = &btrfs_xattr_system_handler, + [BTRFS_XATTR_INDEX_SYSTEM] = &btrfs_xattr_system_handler, }; + struct xattr_handler *btrfs_xattr_handlers[] = { &btrfs_xattr_user_handler, #ifdef CONFIG_FS_POSIX_ACL -// &btrfs_xattr_acl_access_handler, -// &btrfs_xattr_acl_default_handler, + &btrfs_xattr_acl_access_handler, + &btrfs_xattr_acl_default_handler, #endif &btrfs_xattr_trusted_handler, &btrfs_xattr_security_handler, -// &btrfs_xattr_system_handler, + &btrfs_xattr_system_handler, NULL, }; @@ -237,10 +239,14 @@ int btrfs_xattr_set(struct inode *inode, int name_index, mod = 1; goto out; } - } else if (flags & XATTR_REPLACE) { - /* we couldn't find the attr to replace, so error out */ - ret = -ENODATA; - goto out; + } else { + btrfs_release_path(root, path); + + if (flags & XATTR_REPLACE) { + /* we couldn't find the attr to replace */ + ret = -ENODATA; + goto out; + } } /* ok we have to create a completely new xattr */ -- cgit v1.2.3-70-g09d2 From 7b1287662304c3cb05cb38f5e3e2d69f386e8f10 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Thu, 24 Jul 2008 12:17:14 -0400 Subject: Btrfs: Create orphan inode records to prevent lost files after a crash Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/btrfs_inode.h | 3 + fs/btrfs/ctree.c | 2 +- fs/btrfs/ctree.h | 14 +++ fs/btrfs/disk-io.c | 2 + fs/btrfs/inode.c | 237 ++++++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/orphan.c | 67 ++++++++++++++ 7 files changed, 322 insertions(+), 5 deletions(-) create mode 100644 fs/btrfs/orphan.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index f8a38b0da7f..5a0fd7b0e3e 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ hash.o file-item.o inode-item.o inode-map.o disk-io.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ - extent_io.o volumes.o async-thread.o ioctl.o locking.o + extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o btrfs-$(CONFIG_FS_POSIX_ACL) += acl.o else diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 9f2a4ef944a..df624fd735c 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -39,6 +39,9 @@ struct btrfs_inode { struct posix_acl *i_acl; struct posix_acl *i_default_acl; + /* for keeping track of orphaned inodes */ + struct list_head i_orphan; + /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index cdc713062b0..f2a94999c37 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2622,7 +2622,7 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, total_data += data_size[i]; } - total_size = total_data + (nr - 1) * sizeof(struct btrfs_item); + total_size = total_data + (nr * sizeof(struct btrfs_item)); ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); if (ret == 0) { return -EEXIST; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f87d7263f2d..012ad529cb1 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -74,6 +74,9 @@ struct btrfs_ordered_sum; /* directory objectid inside the root tree */ #define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL +/* orhpan objectid for tracking unlinked/truncated files */ +#define BTRFS_ORPHAN_OBJECTID -5ULL + /* * All files have objectids higher than this. */ @@ -646,6 +649,9 @@ struct btrfs_root { /* the dirty list is only used by non-reference counted roots */ struct list_head dirty_list; + + spinlock_t orphan_lock; + struct list_head orphan_list; }; /* @@ -657,6 +663,7 @@ struct btrfs_root { #define BTRFS_INODE_ITEM_KEY 1 #define BTRFS_INODE_REF_KEY 2 #define BTRFS_XATTR_ITEM_KEY 8 +#define BTRFS_ORPHAN_ITEM_KEY 9 /* reserve 2-15 close to the inode for later flexibility */ /* @@ -1560,6 +1567,13 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans, struct btrfs_path *path, u64 dir, const char *name, u16 name_len, int mod); + +/* orphan.c */ +int btrfs_insert_orphan_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 offset); +int btrfs_del_orphan_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 offset); + /* inode-map.c */ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, struct btrfs_root *fs_root, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index ec01062eb41..d6092396734 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -732,7 +732,9 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->in_sysfs = 0; INIT_LIST_HEAD(&root->dirty_list); + INIT_LIST_HEAD(&root->orphan_list); spin_lock_init(&root->node_lock); + spin_lock_init(&root->orphan_lock); mutex_init(&root->objectid_mutex); memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2d8853543a7..0c9ec8aa304 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -78,6 +78,8 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, }; +static void btrfs_truncate(struct inode *inode); + int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int for_del) { @@ -826,6 +828,190 @@ zeroit: return -EIO; } +/* + * This creates an orphan entry for the given inode in case something goes + * wrong in the middle of an unlink/truncate. + */ +int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret = 0; + + spin_lock(&root->orphan_lock); + + /* already on the orphan list, we're good */ + if (!list_empty(&BTRFS_I(inode)->i_orphan)) { + spin_unlock(&root->orphan_lock); + return 0; + } + + list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); + + spin_unlock(&root->orphan_lock); + + /* + * insert an orphan item to track this unlinked/truncated file + */ + ret = btrfs_insert_orphan_item(trans, root, inode->i_ino); + + return ret; +} + +/* + * We have done the truncate/delete so we can go ahead and remove the orphan + * item for this particular inode. + */ +int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + int ret = 0; + + spin_lock(&root->orphan_lock); + + if (list_empty(&BTRFS_I(inode)->i_orphan)) { + spin_unlock(&root->orphan_lock); + return 0; + } + + list_del_init(&BTRFS_I(inode)->i_orphan); + if (!trans) { + spin_unlock(&root->orphan_lock); + return 0; + } + + spin_unlock(&root->orphan_lock); + + ret = btrfs_del_orphan_item(trans, root, inode->i_ino); + + return ret; +} + +/* + * this cleans up any orphans that may be left on the list from the last use + * of this root. + */ +void btrfs_orphan_cleanup(struct btrfs_root *root) +{ + struct btrfs_path *path; + struct extent_buffer *leaf; + struct btrfs_item *item; + struct btrfs_key key, found_key; + struct btrfs_trans_handle *trans; + struct inode *inode; + int ret = 0, nr_unlink = 0, nr_truncate = 0; + + /* don't do orphan cleanup if the fs is readonly. */ + if (root->inode->i_sb->s_flags & MS_RDONLY) + return; + + path = btrfs_alloc_path(); + if (!path) + return; + path->reada = -1; + + key.objectid = BTRFS_ORPHAN_OBJECTID; + btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); + key.offset = (u64)-1; + + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, root->inode); + + while (1) { + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) { + printk(KERN_ERR "Error searching slot for orphan: %d" + "\n", ret); + break; + } + + /* + * if ret == 0 means we found what we were searching for, which + * is weird, but possible, so only screw with path if we didnt + * find the key and see if we have stuff that matches + */ + if (ret > 0) { + if (path->slots[0] == 0) + break; + path->slots[0]--; + } + + /* pull out the item */ + leaf = path->nodes[0]; + item = btrfs_item_nr(leaf, path->slots[0]); + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + + /* make sure the item matches what we want */ + if (found_key.objectid != BTRFS_ORPHAN_OBJECTID) + break; + if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY) + break; + + /* release the path since we're done with it */ + btrfs_release_path(root, path); + + /* + * this is where we are basically btrfs_lookup, without the + * crossing root thing. we store the inode number in the + * offset of the orphan item. + */ + inode = btrfs_iget_locked(root->inode->i_sb, + found_key.offset, root); + if (!inode) + break; + + if (inode->i_state & I_NEW) { + BTRFS_I(inode)->root = root; + + /* have to set the location manually */ + BTRFS_I(inode)->location.objectid = inode->i_ino; + BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; + BTRFS_I(inode)->location.offset = 0; + + btrfs_read_locked_inode(inode); + unlock_new_inode(inode); + } + + /* + * add this inode to the orphan list so btrfs_orphan_del does + * the proper thing when we hit it + */ + spin_lock(&root->orphan_lock); + list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); + spin_unlock(&root->orphan_lock); + + /* + * if this is a bad inode, means we actually succeeded in + * removing the inode, but not the orphan record, which means + * we need to manually delete the orphan since iput will just + * do a destroy_inode + */ + if (is_bad_inode(inode)) { + btrfs_orphan_del(trans, inode); + iput(inode); + continue; + } + + /* if we have links, this was a truncate, lets do that */ + if (inode->i_nlink) { + nr_truncate++; + btrfs_truncate(inode); + } else { + nr_unlink++; + } + + /* this will do delete_inode and everything for us */ + iput(inode); + } + + if (nr_unlink) + printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink); + if (nr_truncate) + printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate); + + btrfs_free_path(path); + btrfs_end_transaction(trans, root); +} + void btrfs_read_locked_inode(struct inode *inode) { struct btrfs_path *path; @@ -1067,6 +1253,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) { struct btrfs_root *root; struct btrfs_trans_handle *trans; + struct inode *inode = dentry->d_inode; int ret; unsigned long nr = 0; @@ -1080,6 +1267,10 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) btrfs_set_trans_block_group(trans, dir); ret = btrfs_unlink_trans(trans, root, dir, dentry); + + if (inode->i_nlink == 0) + ret = btrfs_orphan_add(trans, inode); + nr = trans->blocks_used; btrfs_end_transaction_throttle(trans, root); @@ -1108,12 +1299,17 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); + err = btrfs_orphan_add(trans, inode); + if (err) + goto fail_trans; + /* now the directory is empty */ err = btrfs_unlink_trans(trans, root, dir, dentry); if (!err) { btrfs_i_size_write(inode, 0); } +fail_trans: nr = trans->blocks_used; ret = btrfs_end_transaction_throttle(trans, root); fail: @@ -1131,6 +1327,9 @@ fail: * * csum items that cross the new i_size are truncated to the new size * as well. + * + * min_type is the minimum key type to truncate down to. If set to 0, this + * will kill all the items on this inode, including the INODE_ITEM_KEY. */ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -1495,6 +1694,7 @@ void btrfs_delete_inode(struct inode *inode) truncate_inode_pages(&inode->i_data, 0); if (is_bad_inode(inode)) { + btrfs_orphan_del(NULL, inode); goto no_delete; } btrfs_wait_ordered_range(inode, 0, (u64)-1); @@ -1504,8 +1704,12 @@ void btrfs_delete_inode(struct inode *inode) btrfs_set_trans_block_group(trans, inode); ret = btrfs_truncate_in_trans(trans, root, inode, 0); - if (ret) + if (ret) { + btrfs_orphan_del(NULL, inode); goto no_delete_lock; + } + + btrfs_orphan_del(trans, inode); nr = trans->blocks_used; clear_inode(inode); @@ -1688,7 +1892,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = bi->root; struct btrfs_root *sub_root = root; struct btrfs_key location; - int ret; + int ret, do_orphan = 0; if (dentry->d_name.len > BTRFS_NAME_LEN) return ERR_PTR(-ENAMETOOLONG); @@ -1706,6 +1910,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, return ERR_PTR(ret); if (ret > 0) return ERR_PTR(-ENOENT); + inode = btrfs_iget_locked(dir->i_sb, location.objectid, sub_root); if (!inode) @@ -1715,6 +1920,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, if (sub_root != root) { igrab(inode); sub_root->inode = inode; + do_orphan = 1; } BTRFS_I(inode)->root = sub_root; memcpy(&BTRFS_I(inode)->location, &location, @@ -1723,6 +1929,10 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, unlock_new_inode(inode); } } + + if (unlikely(do_orphan)) + btrfs_orphan_cleanup(sub_root); + return d_splice_alias(inode, dentry); } @@ -2964,12 +3174,19 @@ static void btrfs_truncate(struct inode *inode) btrfs_set_trans_block_group(trans, inode); btrfs_i_size_write(inode, inode->i_size); + ret = btrfs_orphan_add(trans, inode); + if (ret) + goto out; /* FIXME, add redo link to tree so we don't leak on crash */ ret = btrfs_truncate_in_trans(trans, root, inode, BTRFS_EXTENT_DATA_KEY); btrfs_update_inode(trans, root, inode); - nr = trans->blocks_used; + ret = btrfs_orphan_del(trans, inode); + BUG_ON(ret); + +out: + nr = trans->blocks_used; ret = btrfs_end_transaction_throttle(trans, root); BUG_ON(ret); btrfs_btree_balance_dirty(root, nr); @@ -3046,6 +3263,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) btrfs_ordered_inode_tree_init(&ei->ordered_tree); ei->i_acl = BTRFS_ACL_NOT_CACHED; ei->i_default_acl = BTRFS_ACL_NOT_CACHED; + INIT_LIST_HEAD(&ei->i_orphan); return &ei->vfs_inode; } @@ -3062,6 +3280,14 @@ void btrfs_destroy_inode(struct inode *inode) BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED) posix_acl_release(BTRFS_I(inode)->i_default_acl); + spin_lock(&BTRFS_I(inode)->root->orphan_lock); + if (!list_empty(&BTRFS_I(inode)->i_orphan)) { + printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan" + " list\n", inode->i_ino); + dump_stack(); + } + spin_unlock(&BTRFS_I(inode)->root->orphan_lock); + while(1) { ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); if (!ordered) @@ -3202,6 +3428,11 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry); if (ret) goto out_fail; + if (new_inode->i_nlink == 0) { + ret = btrfs_orphan_add(trans, new_inode); + if (ret) + goto out_fail; + } } ret = btrfs_set_inode_index(new_dir, old_inode); if (ret) diff --git a/fs/btrfs/orphan.c b/fs/btrfs/orphan.c new file mode 100644 index 00000000000..3c0d52af4f8 --- /dev/null +++ b/fs/btrfs/orphan.c @@ -0,0 +1,67 @@ +/* + * Copyright (C) 2008 Red Hat. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include "ctree.h" +#include "disk-io.h" + +int btrfs_insert_orphan_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 offset) +{ + struct btrfs_path *path; + struct btrfs_key key; + int ret = 0; + + key.objectid = BTRFS_ORPHAN_OBJECTID; + btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); + key.offset = offset; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = btrfs_insert_empty_item(trans, root, path, &key, 0); + + btrfs_free_path(path); + return ret; +} + +int btrfs_del_orphan_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 offset) +{ + struct btrfs_path *path; + struct btrfs_key key; + int ret = 0; + + key.objectid = BTRFS_ORPHAN_OBJECTID; + btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); + key.offset = offset; + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret) + goto out; + + ret = btrfs_del_item(trans, root, path); + +out: + btrfs_free_path(path); + return ret; +} -- cgit v1.2.3-70-g09d2 From 45467261edb590fd5a2d676d9686edb82762fb3f Mon Sep 17 00:00:00 2001 From: Balaji Rao Date: Thu, 24 Jul 2008 12:17:15 -0400 Subject: Btrfs: Remove unused variable in fixup_tree_root_location Remove a unused variable 'path' in fixup_tree_root_location. Signed-off-by: Balaji Rao Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0c9ec8aa304..8bc981c4f5e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1805,7 +1805,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, struct btrfs_root **sub_root, struct dentry *dentry) { - struct btrfs_path *path; struct btrfs_root_item *ri; if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY) @@ -1813,9 +1812,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) return 0; - path = btrfs_alloc_path(); - BUG_ON(!path); - *sub_root = btrfs_read_fs_root(root->fs_info, location, dentry->d_name.name, dentry->d_name.len); @@ -1827,7 +1823,6 @@ static int fixup_tree_root_location(struct btrfs_root *root, btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); location->offset = 0; - btrfs_free_path(path); return 0; } -- cgit v1.2.3-70-g09d2 From 445dceb78f3445b9bcade90e93ca35cae6120172 Mon Sep 17 00:00:00 2001 From: Yan Date: Thu, 24 Jul 2008 12:19:32 -0400 Subject: Btrfs: Fix .. lookup corner case Inode ref item can be in the next leaf when we find "path->slots[0] == btrfs_header_nritems(...)". Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8bc981c4f5e..cf9534b79ab 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1752,21 +1752,18 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, if (namelen == 2 && strcmp(name, "..") == 0) { struct btrfs_key key; struct extent_buffer *leaf; - u32 nritems; int slot; key.objectid = dir->i_ino; + key.offset = (u64)-1; btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); - key.offset = 0; + if (ret < 0 || path->slots[0] == 0) + goto out_err; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); BUG_ON(ret == 0); ret = 0; - leaf = path->nodes[0]; - slot = path->slots[0]; - nritems = btrfs_header_nritems(leaf); - if (slot >= nritems) - goto out_err; + slot = path->slots[0] - 1; btrfs_item_key_to_cpu(leaf, &key, slot); if (key.objectid != dir->i_ino || @@ -1980,16 +1977,15 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) /* special case for .., just use the back ref */ if (filp->f_pos == 1) { btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); - key.offset = 0; + key.offset = (u64)-1; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); - BUG_ON(ret == 0); - leaf = path->nodes[0]; - slot = path->slots[0]; - nritems = btrfs_header_nritems(leaf); - if (slot >= nritems) { + if (ret < 0 || path->slots[0] == 0) { btrfs_release_path(root, path); goto read_dir_items; } + BUG_ON(ret == 0); + leaf = path->nodes[0]; + slot = path->slots[0] - 1; btrfs_item_key_to_cpu(leaf, &found_key, slot); btrfs_release_path(root, path); if (found_key.objectid != key.objectid || -- cgit v1.2.3-70-g09d2 From 017e5369eb353559d68a11d4a718faa634533821 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 28 Jul 2008 15:32:51 -0400 Subject: Btrfs: Leaf reference cache update This changes the reference cache to make a single cache per root instead of one cache per transaction, and to key by the byte number of the disk block instead of the keys inside. This makes it much less likely to have cache misses if a snapshot or something has an extra reference on a higher node or a leaf while the first transaction that added the leaf into the cache is dropping. Some throttling is added to functions that free blocks heavily so they wait for old transactions to drop. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 11 +++++++- fs/btrfs/disk-io.c | 8 +++--- fs/btrfs/extent-tree.c | 18 +++++-------- fs/btrfs/file.c | 2 +- fs/btrfs/inode.c | 8 +++--- fs/btrfs/ref-cache.c | 71 +++++++++++--------------------------------------- fs/btrfs/ref-cache.h | 18 ++++++------- fs/btrfs/transaction.c | 34 +++++++++--------------- 8 files changed, 61 insertions(+), 109 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 34ed23d64eb..4eca0aa1ce7 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -594,7 +594,6 @@ struct btrfs_fs_info { spinlock_t ref_cache_lock; u64 total_ref_cache_size; - u64 running_ref_cache_size; u64 avail_data_alloc_bits; u64 avail_metadata_alloc_bits; @@ -606,10 +605,18 @@ struct btrfs_fs_info { void *bdev_holder; }; +struct btrfs_leaf_ref_tree { + struct rb_root root; + struct btrfs_leaf_ref *last; + struct list_head list; + spinlock_t lock; +}; + /* * in ram representation of the tree. extent_root is used for all allocations * and for the extent tree extent_root root. */ +struct dirty_root; struct btrfs_root { struct extent_buffer *node; @@ -618,6 +625,8 @@ struct btrfs_root { struct extent_buffer *commit_root; struct btrfs_leaf_ref_tree *ref_tree; + struct btrfs_leaf_ref_tree ref_tree_struct; + struct dirty_root *dirty_root; struct btrfs_root_item root_item; struct btrfs_key root_key; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 4f0e1d06c38..eccdf13a95a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -40,6 +40,7 @@ #include "print-tree.h" #include "async-thread.h" #include "locking.h" +#include "ref-cache.h" #if 0 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) @@ -737,6 +738,10 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, spin_lock_init(&root->node_lock); spin_lock_init(&root->orphan_lock); mutex_init(&root->objectid_mutex); + + btrfs_leaf_ref_tree_init(&root->ref_tree_struct); + root->ref_tree = &root->ref_tree_struct; + memset(&root->root_key, 0, sizeof(root->root_key)); memset(&root->root_item, 0, sizeof(root->root_item)); memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); @@ -1176,9 +1181,6 @@ static int transaction_kthread(void *arg) goto sleep; } - printk("btrfs: running reference cache size %Lu\n", - root->fs_info->running_ref_cache_size); - now = get_seconds(); if (now < cur->start_time || now - cur->start_time < 30) { mutex_unlock(&root->fs_info->trans_mutex); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 7b24f151165..0e294cfaa60 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1004,8 +1004,6 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, goto out; } - btrfs_item_key_to_cpu(buf, &ref->key, 0); - ref->bytenr = buf->start; ref->owner = btrfs_header_owner(buf); ref->generation = btrfs_header_generation(buf); @@ -2387,19 +2385,15 @@ static void noinline reada_walk_down(struct btrfs_root *root, } } -/* - * we want to avoid as much random IO as we can with the alloc mutex - * held, so drop the lock and do the lookup, then do it again with the - * lock held. - */ int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len, u32 *refs) { + int ret; mutex_unlock(&root->fs_info->alloc_mutex); - lookup_extent_ref(NULL, root, start, len, refs); + ret = lookup_extent_ref(NULL, root, start, len, refs); cond_resched(); mutex_lock(&root->fs_info->alloc_mutex); - return lookup_extent_ref(NULL, root, start, len, refs); + return ret; } /* @@ -2468,11 +2462,11 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, BUG_ON(ret); continue; } - + if (*level == 1) { struct btrfs_key key; btrfs_node_key_to_cpu(cur, &key, path->slots[*level]); - ref = btrfs_lookup_leaf_ref(root, &key); + ref = btrfs_lookup_leaf_ref(root, bytenr); if (ref) { ret = drop_leaf_ref(trans, root, ref); BUG_ON(ret); @@ -2482,7 +2476,6 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, break; } } - next = btrfs_find_tree_block(root, bytenr, blocksize); if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) { free_extent_buffer(next); @@ -2672,6 +2665,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root ret = -EAGAIN; break; } + wake_up(&root->fs_info->transaction_throttle); } for (i = 0; i <= orig_level; i++) { if (path->nodes[i]) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index e5ffb66ad32..3efec25e34b 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -347,7 +347,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, btrfs_update_inode(trans, root, inode); } failed: - err = btrfs_end_transaction_throttle(trans, root); + err = btrfs_end_transaction(trans, root); out_unlock: unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); return err; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index cf9534b79ab..4f977ea5497 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2482,7 +2482,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; - btrfs_end_transaction_throttle(trans, root); + btrfs_end_transaction(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); @@ -2535,7 +2535,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; nr = trans->blocks_used; - btrfs_end_transaction_throttle(trans, root); + btrfs_end_transaction(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); @@ -2609,7 +2609,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) out_fail: nr = trans->blocks_used; - btrfs_end_transaction_throttle(trans, root); + btrfs_end_transaction(trans, root); out_unlock: if (drop_on_err) @@ -3548,7 +3548,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; - btrfs_end_transaction_throttle(trans, root); + btrfs_end_transaction(trans, root); out_fail: if (drop_inode) { inode_dec_link_count(inode); diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c index 95a9faeb9dc..ec9587784a3 100644 --- a/fs/btrfs/ref-cache.c +++ b/fs/btrfs/ref-cache.c @@ -29,6 +29,7 @@ struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents) if (ref) { memset(ref, 0, sizeof(*ref)); atomic_set(&ref->usage, 1); + INIT_LIST_HEAD(&ref->list); } return ref; } @@ -44,40 +45,21 @@ void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref) } } -static int comp_keys(struct btrfs_key *k1, struct btrfs_key *k2) -{ - if (k1->objectid > k2->objectid) - return 1; - if (k1->objectid < k2->objectid) - return -1; - if (k1->type > k2->type) - return 1; - if (k1->type < k2->type) - return -1; - if (k1->offset > k2->offset) - return 1; - if (k1->offset < k2->offset) - return -1; - return 0; -} - -static struct rb_node *tree_insert(struct rb_root *root, struct btrfs_key *key, +static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr, struct rb_node *node) { struct rb_node ** p = &root->rb_node; struct rb_node * parent = NULL; struct btrfs_leaf_ref *entry; - int ret; while(*p) { parent = *p; entry = rb_entry(parent, struct btrfs_leaf_ref, rb_node); WARN_ON(!entry->in_tree); - ret = comp_keys(key, &entry->key); - if (ret < 0) + if (bytenr < entry->bytenr) p = &(*p)->rb_left; - else if (ret > 0) + else if (bytenr > entry->bytenr) p = &(*p)->rb_right; else return parent; @@ -90,20 +72,18 @@ static struct rb_node *tree_insert(struct rb_root *root, struct btrfs_key *key, return NULL; } -static struct rb_node *tree_search(struct rb_root *root, struct btrfs_key *key) +static struct rb_node *tree_search(struct rb_root *root, u64 bytenr) { struct rb_node * n = root->rb_node; struct btrfs_leaf_ref *entry; - int ret; while(n) { entry = rb_entry(n, struct btrfs_leaf_ref, rb_node); WARN_ON(!entry->in_tree); - ret = comp_keys(key, &entry->key); - if (ret < 0) + if (bytenr < entry->bytenr) n = n->rb_left; - else if (ret > 0) + else if (bytenr > entry->bytenr) n = n->rb_right; else return n; @@ -122,11 +102,11 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root) spin_lock(&tree->lock); while(!btrfs_leaf_ref_tree_empty(tree)) { - tree->last = NULL; rb = rb_first(&tree->root); ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node); rb_erase(&ref->rb_node, &tree->root); ref->in_tree = 0; + list_del_init(&ref->list); spin_unlock(&tree->lock); @@ -140,7 +120,7 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root) } struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root, - struct btrfs_key *key) + u64 bytenr) { struct rb_node *rb; struct btrfs_leaf_ref *ref = NULL; @@ -150,15 +130,9 @@ struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root, return NULL; spin_lock(&tree->lock); - if (tree->last && comp_keys(key, &tree->last->key) == 0) { - ref = tree->last; - } else { - rb = tree_search(&tree->root, key); - if (rb) { - ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node); - tree->last = ref; - } - } + rb = tree_search(&tree->root, bytenr); + if (rb) + ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node); if (ref) atomic_inc(&ref->usage); spin_unlock(&tree->lock); @@ -171,21 +145,17 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) struct rb_node *rb; size_t size = btrfs_leaf_ref_size(ref->nritems); struct btrfs_leaf_ref_tree *tree = root->ref_tree; - struct btrfs_transaction *trans = root->fs_info->running_transaction; spin_lock(&tree->lock); - rb = tree_insert(&tree->root, &ref->key, &ref->rb_node); + rb = tree_insert(&tree->root, ref->bytenr, &ref->rb_node); if (rb) { ret = -EEXIST; } else { spin_lock(&root->fs_info->ref_cache_lock); root->fs_info->total_ref_cache_size += size; - if (trans && tree->generation == trans->transid) - root->fs_info->running_ref_cache_size += size; spin_unlock(&root->fs_info->ref_cache_lock); - - tree->last = ref; atomic_inc(&ref->usage); + list_add_tail(&ref->list, &tree->list); } spin_unlock(&tree->lock); return ret; @@ -195,28 +165,17 @@ int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { size_t size = btrfs_leaf_ref_size(ref->nritems); struct btrfs_leaf_ref_tree *tree = root->ref_tree; - struct btrfs_transaction *trans = root->fs_info->running_transaction; BUG_ON(!ref->in_tree); spin_lock(&tree->lock); spin_lock(&root->fs_info->ref_cache_lock); root->fs_info->total_ref_cache_size -= size; - if (trans && tree->generation == trans->transid) - root->fs_info->running_ref_cache_size -= size; spin_unlock(&root->fs_info->ref_cache_lock); - if (tree->last == ref) { - struct rb_node *next = rb_next(&ref->rb_node); - if (next) { - tree->last = rb_entry(next, struct btrfs_leaf_ref, - rb_node); - } else - tree->last = NULL; - } - rb_erase(&ref->rb_node, &tree->root); ref->in_tree = 0; + list_del_init(&ref->list); spin_unlock(&tree->lock); diff --git a/fs/btrfs/ref-cache.h b/fs/btrfs/ref-cache.h index 79ecc47110f..823c049f72f 100644 --- a/fs/btrfs/ref-cache.h +++ b/fs/btrfs/ref-cache.h @@ -15,6 +15,8 @@ * Free Software Foundation, Inc., 59 Temple Place - Suite 330, * Boston, MA 021110-1307, USA. */ +#ifndef __REFCACHE__ +#define __REFCACHE__ struct btrfs_extent_info { u64 bytenr; @@ -25,7 +27,6 @@ struct btrfs_extent_info { struct btrfs_leaf_ref { struct rb_node rb_node; - struct btrfs_key key; int in_tree; atomic_t usage; @@ -33,14 +34,9 @@ struct btrfs_leaf_ref { u64 owner; u64 generation; int nritems; - struct btrfs_extent_info extents[]; -}; -struct btrfs_leaf_ref_tree { - struct rb_root root; - struct btrfs_leaf_ref *last; - u64 generation; - spinlock_t lock; + struct list_head list; + struct btrfs_extent_info extents[]; }; static inline size_t btrfs_leaf_ref_size(int nr_extents) @@ -53,7 +49,7 @@ static inline void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree) { tree->root.rb_node = NULL; tree->last = NULL; - tree->generation = 0; + INIT_LIST_HEAD(&tree->list); spin_lock_init(&tree->lock); } @@ -66,7 +62,9 @@ void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree); struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents); void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref); struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root, - struct btrfs_key *key); + u64 bytenr); int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref); int btrfs_remove_leaf_refs(struct btrfs_root *root); int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref); + +#endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 543e5ee4033..fcef3cae0c9 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -36,7 +36,6 @@ struct dirty_root { struct list_head list; struct btrfs_root *root; struct btrfs_root *latest_root; - struct btrfs_leaf_ref_tree ref_tree; }; static noinline void put_transaction(struct btrfs_transaction *transaction) @@ -108,13 +107,13 @@ static noinline int record_root_in_trans(struct btrfs_root *root) dirty->latest_root = root; INIT_LIST_HEAD(&dirty->list); - btrfs_leaf_ref_tree_init(&dirty->ref_tree); - dirty->ref_tree.generation = running_trans_id; root->commit_root = btrfs_root_node(root); - root->ref_tree = &dirty->ref_tree; + root->dirty_root = dirty; memcpy(dirty->root, root, sizeof(*root)); + dirty->root->ref_tree = &root->ref_tree_struct; + spin_lock_init(&dirty->root->node_lock); mutex_init(&dirty->root->objectid_mutex); dirty->root->node = root->commit_root; @@ -217,12 +216,13 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); - if (0 && cur_trans->in_commit && throttle) { + if (throttle && atomic_read(&root->fs_info->throttles)) { DEFINE_WAIT(wait); mutex_unlock(&root->fs_info->trans_mutex); prepare_to_wait(&root->fs_info->transaction_throttle, &wait, TASK_UNINTERRUPTIBLE); - schedule(); + if (atomic_read(&root->fs_info->throttles)) + schedule(); finish_wait(&root->fs_info->transaction_throttle, &wait); mutex_lock(&root->fs_info->trans_mutex); } @@ -333,6 +333,8 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, list_del_init(next); root = list_entry(next, struct btrfs_root, dirty_list); update_cowonly_root(trans, root); + if (root->fs_info->closing) + btrfs_remove_leaf_refs(root); } return 0; } @@ -346,10 +348,8 @@ int btrfs_add_dead_root(struct btrfs_root *root, dirty = kmalloc(sizeof(*dirty), GFP_NOFS); if (!dirty) return -ENOMEM; - btrfs_leaf_ref_tree_init(&dirty->ref_tree); dirty->root = root; dirty->latest_root = latest; - root->ref_tree = NULL; list_add(&dirty->list, dead_list); return 0; } @@ -379,18 +379,14 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, BTRFS_ROOT_TRANS_TAG); BUG_ON(!root->ref_tree); - dirty = container_of(root->ref_tree, struct dirty_root, - ref_tree); + dirty = root->dirty_root; if (root->commit_root == root->node) { WARN_ON(root->node->start != btrfs_root_bytenr(&root->root_item)); - BUG_ON(!btrfs_leaf_ref_tree_empty( - root->ref_tree)); free_extent_buffer(root->commit_root); root->commit_root = NULL; - root->ref_tree = NULL; kfree(dirty->root); kfree(dirty); @@ -410,7 +406,6 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, sizeof(struct btrfs_disk_key)); root->root_item.drop_level = 0; root->commit_root = NULL; - root->ref_tree = NULL; root->root_key.offset = root->fs_info->generation; btrfs_set_root_bytenr(&root->root_item, root->node->start); @@ -485,7 +480,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, while(!list_empty(list)) { struct btrfs_root *root; - dirty = list_entry(list->next, struct dirty_root, list); + dirty = list_entry(list->prev, struct dirty_root, list); list_del_init(&dirty->list); num_bytes = btrfs_root_used(&dirty->root->root_item); @@ -507,7 +502,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, if (err) ret = err; nr = trans->blocks_used; - ret = btrfs_end_transaction_throttle(trans, tree_root); + ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); mutex_unlock(&root->fs_info->drop_mutex); @@ -517,6 +512,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, } BUG_ON(ret); atomic_dec(&root->fs_info->throttles); + wake_up(&root->fs_info->transaction_throttle); mutex_lock(&root->fs_info->alloc_mutex); num_bytes -= btrfs_root_used(&dirty->root->root_item); @@ -539,8 +535,6 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); - btrfs_remove_leaf_refs(dirty->root); - free_extent_buffer(dirty->root->node); kfree(dirty->root); kfree(dirty); @@ -725,10 +719,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, &dirty_fs_roots); BUG_ON(ret); - spin_lock(&root->fs_info->ref_cache_lock); - root->fs_info->running_ref_cache_size = 0; - spin_unlock(&root->fs_info->ref_cache_lock); - ret = btrfs_commit_tree_roots(trans, root); BUG_ON(ret); -- cgit v1.2.3-70-g09d2 From ab78c84de1ce4db1b2a2cef361625ad80abbab3f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 29 Jul 2008 16:15:18 -0400 Subject: Btrfs: Throttle operations if the reference cache gets too large A large reference cache is directly related to a lot of work pending for the cleaner thread. This throttles back new operations based on the size of the reference cache so the cleaner thread will be able to keep up. Overall, this actually makes the FS faster because the cleaner thread will be more likely to find things in cache. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 7 ++++-- fs/btrfs/extent-tree.c | 1 + fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 10 ++++----- fs/btrfs/transaction.c | 59 +++++++++++++++++++++++++++++++++++++------------- fs/btrfs/transaction.h | 1 + 7 files changed, 58 insertions(+), 22 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 4eca0aa1ce7..5517dfc6f71 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -580,6 +580,7 @@ struct btrfs_fs_info { int do_barriers; int closing; atomic_t throttles; + atomic_t throttle_gen; u64 total_pinned; struct list_head dirty_cowonly_roots; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index eccdf13a95a..27ffa9b7ddc 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1171,8 +1171,10 @@ static int transaction_kthread(void *arg) vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); mutex_lock(&root->fs_info->transaction_kthread_mutex); - printk("btrfs: total reference cache size %Lu\n", - root->fs_info->total_ref_cache_size); + if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { + printk("btrfs: total reference cache size %Lu\n", + root->fs_info->total_ref_cache_size); + } mutex_lock(&root->fs_info->trans_mutex); cur = root->fs_info->running_transaction; @@ -1256,6 +1258,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_mapping_init(&fs_info->mapping_tree); atomic_set(&fs_info->nr_async_submits, 0); atomic_set(&fs_info->throttles, 0); + atomic_set(&fs_info->throttle_gen, 0); fs_info->sb = sb; fs_info->max_extent = (u64)-1; fs_info->max_inline = 8192 * 1024; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 0e294cfaa60..6290cf41d64 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2650,6 +2650,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root } } while(1) { + atomic_inc(&root->fs_info->throttle_gen); wret = walk_down_tree(trans, root, path, &level); if (wret > 0) break; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3efec25e34b..ded5281f846 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -974,6 +974,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages); if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1) btrfs_btree_balance_dirty(root, 1); + btrfs_throttle(root); cond_resched(); } out: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4f977ea5497..7c87f863d6f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2482,7 +2482,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); @@ -2535,7 +2535,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); fail: if (drop_inode) { inode_dec_link_count(inode); @@ -2609,7 +2609,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) out_fail: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); out_unlock: if (drop_on_err) @@ -3434,7 +3434,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, goto out_fail; out_fail: - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); out_unlock: return ret; } @@ -3548,7 +3548,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, out_unlock: nr = trans->blocks_used; - btrfs_end_transaction(trans, root); + btrfs_end_transaction_throttle(trans, root); out_fail: if (drop_inode) { inode_dec_link_count(inode); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index fcef3cae0c9..b8be6703189 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -202,35 +202,64 @@ static noinline int wait_for_commit(struct btrfs_root *root, return 0; } +void btrfs_throttle(struct btrfs_root *root) +{ + struct btrfs_fs_info *info = root->fs_info; + +harder: + if (atomic_read(&info->throttles)) { + DEFINE_WAIT(wait); + int thr; + int harder_count = 0; + thr = atomic_read(&info->throttle_gen); + + do { + prepare_to_wait(&info->transaction_throttle, + &wait, TASK_UNINTERRUPTIBLE); + if (!atomic_read(&info->throttles)) { + finish_wait(&info->transaction_throttle, &wait); + break; + } + schedule(); + finish_wait(&info->transaction_throttle, &wait); + } while (thr == atomic_read(&info->throttle_gen)); + + if (harder_count < 5 && + info->total_ref_cache_size > 5 * 1024 * 1024) { + harder_count++; + goto harder; + } + + if (harder_count < 10 && + info->total_ref_cache_size > 10 * 1024 * 1024) { + harder_count++; + goto harder; + } + } +} + static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root, int throttle) { struct btrfs_transaction *cur_trans; + struct btrfs_fs_info *info = root->fs_info; - mutex_lock(&root->fs_info->trans_mutex); - cur_trans = root->fs_info->running_transaction; + mutex_lock(&info->trans_mutex); + cur_trans = info->running_transaction; WARN_ON(cur_trans != trans->transaction); WARN_ON(cur_trans->num_writers < 1); cur_trans->num_writers--; if (waitqueue_active(&cur_trans->writer_wait)) wake_up(&cur_trans->writer_wait); - - if (throttle && atomic_read(&root->fs_info->throttles)) { - DEFINE_WAIT(wait); - mutex_unlock(&root->fs_info->trans_mutex); - prepare_to_wait(&root->fs_info->transaction_throttle, &wait, - TASK_UNINTERRUPTIBLE); - if (atomic_read(&root->fs_info->throttles)) - schedule(); - finish_wait(&root->fs_info->transaction_throttle, &wait); - mutex_lock(&root->fs_info->trans_mutex); - } - put_transaction(cur_trans); - mutex_unlock(&root->fs_info->trans_mutex); + mutex_unlock(&info->trans_mutex); memset(trans, 0, sizeof(*trans)); kmem_cache_free(btrfs_trans_handle_cachep, trans); + + if (throttle) + btrfs_throttle(root); + return 0; } diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 11fbdeceb26..df2ca2aad1c 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -91,4 +91,5 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, struct btrfs_root *root); +void btrfs_throttle(struct btrfs_root *root); #endif -- cgit v1.2.3-70-g09d2 From f321e4910398cf7922265d269fb17fd26f312571 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Wed, 30 Jul 2008 09:26:11 -0400 Subject: Btrfs: Update and fix mount -o nodatacow To check whether a given file extent is referenced by multiple snapshots, the checker walks down the fs tree through dead root and checks all tree blocks in the path. We can easily detect whether a given tree block is directly referenced by other snapshot. We can also detect any indirect reference from other snapshot by checking reference's generation. The checker can always detect multiple references, but can't reliably detect cases of single reference. So btrfs may do file data cow even there is only one reference. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 9 +-- fs/btrfs/extent-tree.c | 202 ++++++++++++++++++++++++++++++------------------- fs/btrfs/inode.c | 6 +- fs/btrfs/transaction.c | 16 ++-- fs/btrfs/transaction.h | 5 ++ 5 files changed, 142 insertions(+), 96 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5517dfc6f71..83422088c62 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -617,7 +617,7 @@ struct btrfs_leaf_ref_tree { * in ram representation of the tree. extent_root is used for all allocations * and for the extent tree extent_root root. */ -struct dirty_root; +struct btrfs_dirty_root; struct btrfs_root { struct extent_buffer *node; @@ -627,7 +627,7 @@ struct btrfs_root { struct extent_buffer *commit_root; struct btrfs_leaf_ref_tree *ref_tree; struct btrfs_leaf_ref_tree ref_tree_struct; - struct dirty_root *dirty_root; + struct btrfs_dirty_root *dirty_root; struct btrfs_root_item root_item; struct btrfs_key root_key; @@ -1399,9 +1399,8 @@ static inline struct dentry *fdentry(struct file *file) { } /* extent-tree.c */ -u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, - struct btrfs_path *count_path, - u64 expected_owner, u64 first_extent); +int btrfs_cross_ref_exists(struct btrfs_root *root, + struct btrfs_key *key, u64 bytenr); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 6290cf41d64..fe1ddbd2bfd 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -802,70 +802,57 @@ out: return 0; } -u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, - struct btrfs_path *count_path, - u64 expected_owner, - u64 first_extent) + +static int get_reference_status(struct btrfs_root *root, u64 bytenr, + u64 parent_gen, u64 ref_objectid, + u64 *min_generation, u32 *ref_count) { struct btrfs_root *extent_root = root->fs_info->extent_root; struct btrfs_path *path; - u64 bytenr; - u64 found_objectid; - u64 found_owner; + struct extent_buffer *leaf; + struct btrfs_extent_ref *ref_item; + struct btrfs_key key; + struct btrfs_key found_key; u64 root_objectid = root->root_key.objectid; - u32 total_count = 0; - u32 extent_refs; - u32 cur_count; + u64 ref_generation; u32 nritems; int ret; - struct btrfs_key key; - struct btrfs_key found_key; - struct extent_buffer *l; - struct btrfs_extent_item *item; - struct btrfs_extent_ref *ref_item; - int level = -1; - /* FIXME, needs locking */ - BUG(); - - mutex_lock(&root->fs_info->alloc_mutex); - path = btrfs_alloc_path(); -again: - if (level == -1) - bytenr = first_extent; - else - bytenr = count_path->nodes[level]->start; - - cur_count = 0; key.objectid = bytenr; key.offset = 0; + key.type = BTRFS_EXTENT_ITEM_KEY; - btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); + path = btrfs_alloc_path(); + mutex_lock(&root->fs_info->alloc_mutex); ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); if (ret < 0) goto out; BUG_ON(ret == 0); - l = path->nodes[0]; - btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (found_key.objectid != bytenr || found_key.type != BTRFS_EXTENT_ITEM_KEY) { + ret = 1; goto out; } - item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); - extent_refs = btrfs_extent_refs(l, item); + *ref_count = 0; + *min_generation = (u64)-1; + while (1) { - l = path->nodes[0]; - nritems = btrfs_header_nritems(l); + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); if (path->slots[0] >= nritems) { ret = btrfs_next_leaf(extent_root, path); + if (ret < 0) + goto out; if (ret == 0) continue; break; } - btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (found_key.objectid != bytenr) break; @@ -874,57 +861,120 @@ again: continue; } - cur_count++; - ref_item = btrfs_item_ptr(l, path->slots[0], + ref_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); - found_objectid = btrfs_ref_root(l, ref_item); - - if (found_objectid != root_objectid) { - total_count = 2; - goto out; - } - if (level == -1) { - found_owner = btrfs_ref_objectid(l, ref_item); - if (found_owner != expected_owner) { - total_count = 2; - goto out; - } - /* - * nasty. we don't count a reference held by - * the running transaction. This allows nodatacow - * to avoid cow most of the time - */ - if (found_owner >= BTRFS_FIRST_FREE_OBJECTID && - btrfs_ref_generation(l, ref_item) == - root->fs_info->generation) { - extent_refs--; - } + ref_generation = btrfs_ref_generation(leaf, ref_item); + /* + * For (parent_gen > 0 && parent_gen > ref_gen): + * + * we reach here through the oldest root, therefore + * all other reference from same snapshot should have + * a larger generation. + */ + if ((root_objectid != btrfs_ref_root(leaf, ref_item)) || + (parent_gen > 0 && parent_gen > ref_generation) || + (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID && + ref_objectid != btrfs_ref_objectid(leaf, ref_item))) { + if (ref_count) + *ref_count = 2; + break; } - total_count = 1; + + *ref_count = 1; + if (*min_generation > ref_generation) + *min_generation = ref_generation; + path->slots[0]++; } - /* - * if there is more than one reference against a data extent, - * we have to assume the other ref is another snapshot - */ - if (level == -1 && extent_refs > 1) { - total_count = 2; + ret = 0; +out: + mutex_unlock(&root->fs_info->alloc_mutex); + btrfs_free_path(path); + return ret; +} + +int btrfs_cross_ref_exists(struct btrfs_root *root, + struct btrfs_key *key, u64 bytenr) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *old_root; + struct btrfs_path *path = NULL; + struct extent_buffer *eb; + struct btrfs_file_extent_item *item; + u64 ref_generation; + u64 min_generation; + u64 extent_start; + u32 ref_count; + int level; + int ret; + + BUG_ON(key->type != BTRFS_EXTENT_DATA_KEY); + ret = get_reference_status(root, bytenr, 0, key->objectid, + &min_generation, &ref_count); + if (ret) + return ret; + + if (ref_count != 1) + return 1; + + trans = btrfs_start_transaction(root, 0); + old_root = root->dirty_root->root; + ref_generation = old_root->root_key.offset; + + /* all references are created in running transaction */ + if (min_generation > ref_generation) { + ret = 0; goto out; } - if (cur_count == 0) { - total_count = 0; + + path = btrfs_alloc_path(); + if (!path) { + ret = -ENOMEM; goto out; } - if (level >= 0 && root->node == count_path->nodes[level]) + + path->skip_locking = 1; + /* if no item found, the extent is referenced by other snapshot */ + ret = btrfs_search_slot(NULL, old_root, key, path, 0, 0); + if (ret) goto out; - level++; - btrfs_release_path(root, path); - goto again; + eb = path->nodes[0]; + item = btrfs_item_ptr(eb, path->slots[0], + struct btrfs_file_extent_item); + if (btrfs_file_extent_type(eb, item) != BTRFS_FILE_EXTENT_REG || + btrfs_file_extent_disk_bytenr(eb, item) != bytenr) { + ret = 1; + goto out; + } + + for (level = BTRFS_MAX_LEVEL - 1; level >= -1; level--) { + if (level >= 0) { + eb = path->nodes[level]; + if (!eb) + continue; + extent_start = eb->start; + } else + extent_start = bytenr; + + ret = get_reference_status(root, extent_start, ref_generation, + 0, &min_generation, &ref_count); + if (ret) + goto out; + + if (ref_count != 1) { + ret = 1; + goto out; + } + if (level >= 0) + ref_generation = btrfs_header_generation(eb); + } + ret = 0; out: - btrfs_free_path(path); - mutex_unlock(&root->fs_info->alloc_mutex); - return total_count; + if (path) + btrfs_free_path(path); + btrfs_end_transaction(trans, root); + return ret; } int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7c87f863d6f..3aa82cec6bf 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -249,11 +249,8 @@ again: if (bytenr == 0) goto not_found; - if (btrfs_count_snapshots_in_path(root, path, inode->i_ino, - bytenr) != 1) { + if (btrfs_cross_ref_exists(root, &found_key, bytenr)) goto not_found; - } - /* * we may be called by the resizer, make sure we're inside * the limits of the FS @@ -277,6 +274,7 @@ loop: goto again; not_found: + btrfs_release_path(root, path); cow_file_range(inode, start, end); start = end + 1; goto loop; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index b8be6703189..216f3157162 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -32,12 +32,6 @@ extern struct kmem_cache *btrfs_transaction_cachep; #define BTRFS_ROOT_TRANS_TAG 0 -struct dirty_root { - struct list_head list; - struct btrfs_root *root; - struct btrfs_root *latest_root; -}; - static noinline void put_transaction(struct btrfs_transaction *transaction) { WARN_ON(transaction->use_count == 0); @@ -91,7 +85,7 @@ static noinline int join_transaction(struct btrfs_root *root) static noinline int record_root_in_trans(struct btrfs_root *root) { - struct dirty_root *dirty; + struct btrfs_dirty_root *dirty; u64 running_trans_id = root->fs_info->running_transaction->transid; if (root->ref_cows && root->last_trans < running_trans_id) { WARN_ON(root == root->fs_info->extent_root); @@ -372,7 +366,7 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest, struct list_head *dead_list) { - struct dirty_root *dirty; + struct btrfs_dirty_root *dirty; dirty = kmalloc(sizeof(*dirty), GFP_NOFS); if (!dirty) @@ -387,7 +381,7 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, struct radix_tree_root *radix, struct list_head *list) { - struct dirty_root *dirty; + struct btrfs_dirty_root *dirty; struct btrfs_root *gang[8]; struct btrfs_root *root; int i; @@ -498,7 +492,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) static noinline int drop_dirty_roots(struct btrfs_root *tree_root, struct list_head *list) { - struct dirty_root *dirty; + struct btrfs_dirty_root *dirty; struct btrfs_trans_handle *trans; unsigned long nr; u64 num_bytes; @@ -509,7 +503,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, while(!list_empty(list)) { struct btrfs_root *root; - dirty = list_entry(list->prev, struct dirty_root, list); + dirty = list_entry(list->prev, struct btrfs_dirty_root, list); list_del_init(&dirty->list); num_bytes = btrfs_root_used(&dirty->root->root_item); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index df2ca2aad1c..f5adb23151f 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -52,6 +52,11 @@ struct btrfs_pending_snapshot { struct list_head list; }; +struct btrfs_dirty_root { + struct list_head list; + struct btrfs_root *root; + struct btrfs_root *latest_root; +}; static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans, struct inode *inode) -- cgit v1.2.3-70-g09d2 From bcc63abbf3e9bf948a1b0129b3e6120ec7d7f698 Mon Sep 17 00:00:00 2001 From: Yan Date: Wed, 30 Jul 2008 16:29:20 -0400 Subject: Btrfs: implement memory reclaim for leaf reference cache The memory reclaiming issue happens when snapshot exists. In that case, some cache entries may not be used during old snapshot dropping, so they will remain in the cache until umount. The patch adds a field to struct btrfs_leaf_ref to record create time. Besides, the patch makes all dead roots of a given snapshot linked together in order of create time. After a old snapshot was completely dropped, we check the dead root list and remove all cache entries created before the oldest dead root in the list. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 1 - fs/btrfs/ctree.h | 3 ++- fs/btrfs/dir-item.c | 1 - fs/btrfs/disk-io.c | 5 +++-- fs/btrfs/extent-tree.c | 18 +++++++++--------- fs/btrfs/extent_io.c | 1 - fs/btrfs/file-item.c | 1 - fs/btrfs/file.c | 1 - fs/btrfs/inode.c | 22 +++++++++++----------- fs/btrfs/locking.c | 1 - fs/btrfs/print-tree.c | 1 - fs/btrfs/ref-cache.c | 48 +++++++++++++++++++++++++----------------------- fs/btrfs/ref-cache.h | 11 ++++++----- fs/btrfs/transaction.c | 40 ++++++++++++++++++++++++++++++---------- fs/btrfs/volumes.c | 1 - 15 files changed, 86 insertions(+), 69 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 245eb00435d..c4792062dd5 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -3275,4 +3275,3 @@ int btrfs_previous_item(struct btrfs_root *root, } return 1; } - diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 83422088c62..be16cd49ef6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -666,7 +666,8 @@ struct btrfs_root { /* the dirty list is only used by non-reference counted roots */ struct list_head dirty_list; - spinlock_t orphan_lock; + spinlock_t list_lock; + struct list_head dead_list; struct list_head orphan_list; }; diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index eb4dd3d75cf..125094617fe 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -340,4 +340,3 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, } return 0; } - diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index ec1ba8ddb35..e826730d750 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -735,8 +735,9 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, INIT_LIST_HEAD(&root->dirty_list); INIT_LIST_HEAD(&root->orphan_list); + INIT_LIST_HEAD(&root->dead_list); spin_lock_init(&root->node_lock); - spin_lock_init(&root->orphan_lock); + spin_lock_init(&root->list_lock); mutex_init(&root->objectid_mutex); btrfs_leaf_ref_tree_init(&root->ref_tree_struct); @@ -1717,7 +1718,7 @@ int close_ctree(struct btrfs_root *root) printk("btrfs: at umount reference cache size %Lu\n", fs_info->total_ref_cache_size); } - + if (fs_info->extent_root->node) free_extent_buffer(fs_info->extent_root->node); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index fe1ddbd2bfd..37ca8df30c3 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -867,8 +867,8 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, /* * For (parent_gen > 0 && parent_gen > ref_gen): * - * we reach here through the oldest root, therefore - * all other reference from same snapshot should have + * we reach here through the oldest root, therefore + * all other reference from same snapshot should have * a larger generation. */ if ((root_objectid != btrfs_ref_root(leaf, ref_item)) || @@ -954,7 +954,7 @@ int btrfs_cross_ref_exists(struct btrfs_root *root, if (!eb) continue; extent_start = eb->start; - } else + } else extent_start = bytenr; ret = get_reference_status(root, extent_start, ref_generation, @@ -1048,7 +1048,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_leaf_ref *ref; struct btrfs_extent_info *info; - ref = btrfs_alloc_leaf_ref(nr_file_extents); + ref = btrfs_alloc_leaf_ref(root, nr_file_extents); if (!ref) { WARN_ON(1); goto out; @@ -1059,7 +1059,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, ref->generation = btrfs_header_generation(buf); ref->nritems = nr_file_extents; info = ref->extents; - + for (i = 0; nr_file_extents > 0 && i < nritems; i++) { u64 disk_bytenr; btrfs_item_key_to_cpu(buf, &key, i); @@ -1085,7 +1085,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, BUG_ON(!root->ref_tree); ret = btrfs_add_leaf_ref(root, ref); WARN_ON(ret); - btrfs_free_leaf_ref(ref); + btrfs_free_leaf_ref(root, ref); } out: return 0; @@ -2316,7 +2316,7 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, } static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, struct extent_buffer *leaf) { u64 leaf_owner; @@ -2367,7 +2367,7 @@ static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans, } static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, struct btrfs_leaf_ref *ref) { int i; @@ -2521,7 +2521,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, ret = drop_leaf_ref(trans, root, ref); BUG_ON(ret); btrfs_remove_leaf_ref(root, ref); - btrfs_free_leaf_ref(ref); + btrfs_free_leaf_ref(root, ref); *level = 0; break; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 964ec1622d6..5368e3b6eb9 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3497,4 +3497,3 @@ out: return ret; } EXPORT_SYMBOL(try_release_extent_buffer); - diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index afe42d00b5a..2311061f070 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -422,4 +422,3 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, BUG_ON(ret); return ret; } - diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index ded5281f846..412ab4a2638 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1095,4 +1095,3 @@ struct file_operations btrfs_file_operations = { .compat_ioctl = btrfs_ioctl, #endif }; - diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3aa82cec6bf..7af8be076ee 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -835,17 +835,17 @@ int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; - spin_lock(&root->orphan_lock); + spin_lock(&root->list_lock); /* already on the orphan list, we're good */ if (!list_empty(&BTRFS_I(inode)->i_orphan)) { - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); return 0; } list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); /* * insert an orphan item to track this unlinked/truncated file @@ -864,20 +864,20 @@ int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; - spin_lock(&root->orphan_lock); + spin_lock(&root->list_lock); if (list_empty(&BTRFS_I(inode)->i_orphan)) { - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); return 0; } list_del_init(&BTRFS_I(inode)->i_orphan); if (!trans) { - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); return 0; } - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); ret = btrfs_del_orphan_item(trans, root, inode->i_ino); @@ -973,9 +973,9 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) * add this inode to the orphan list so btrfs_orphan_del does * the proper thing when we hit it */ - spin_lock(&root->orphan_lock); + spin_lock(&root->list_lock); list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); - spin_unlock(&root->orphan_lock); + spin_unlock(&root->list_lock); /* * if this is a bad inode, means we actually succeeded in @@ -3269,13 +3269,13 @@ void btrfs_destroy_inode(struct inode *inode) BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED) posix_acl_release(BTRFS_I(inode)->i_default_acl); - spin_lock(&BTRFS_I(inode)->root->orphan_lock); + spin_lock(&BTRFS_I(inode)->root->list_lock); if (!list_empty(&BTRFS_I(inode)->i_orphan)) { printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan" " list\n", inode->i_ino); dump_stack(); } - spin_unlock(&BTRFS_I(inode)->root->orphan_lock); + spin_unlock(&BTRFS_I(inode)->root->list_lock); while(1) { ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c index d617c29787f..d43e14c7471 100644 --- a/fs/btrfs/locking.c +++ b/fs/btrfs/locking.c @@ -56,4 +56,3 @@ int btrfs_tree_locked(struct extent_buffer *eb) { return mutex_is_locked(&eb->mutex); } - diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 14d86372030..f1374d597a1 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -198,4 +198,3 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c) free_extent_buffer(next); } } - diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c index ec9587784a3..272b9890c98 100644 --- a/fs/btrfs/ref-cache.c +++ b/fs/btrfs/ref-cache.c @@ -21,12 +21,18 @@ #include "ref-cache.h" #include "transaction.h" -struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents) +struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root, + int nr_extents) { struct btrfs_leaf_ref *ref; + size_t size = btrfs_leaf_ref_size(nr_extents); - ref = kmalloc(btrfs_leaf_ref_size(nr_extents), GFP_NOFS); + ref = kmalloc(size, GFP_NOFS); if (ref) { + spin_lock(&root->fs_info->ref_cache_lock); + root->fs_info->total_ref_cache_size += size; + spin_unlock(&root->fs_info->ref_cache_lock); + memset(ref, 0, sizeof(*ref)); atomic_set(&ref->usage, 1); INIT_LIST_HEAD(&ref->list); @@ -34,14 +40,20 @@ struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents) return ref; } -void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref) +void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { if (!ref) return; WARN_ON(atomic_read(&ref->usage) == 0); if (atomic_dec_and_test(&ref->usage)) { + size_t size = btrfs_leaf_ref_size(ref->nritems); + BUG_ON(ref->in_tree); kfree(ref); + + spin_lock(&root->fs_info->ref_cache_lock); + root->fs_info->total_ref_cache_size -= size; + spin_unlock(&root->fs_info->ref_cache_lock); } } @@ -64,7 +76,7 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr, else return parent; } - + entry = rb_entry(node, struct btrfs_leaf_ref, rb_node); entry->in_tree = 1; rb_link_node(node, parent, p); @@ -91,9 +103,8 @@ static struct rb_node *tree_search(struct rb_root *root, u64 bytenr) return NULL; } -int btrfs_remove_leaf_refs(struct btrfs_root *root) +int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen) { - struct rb_node *rb; struct btrfs_leaf_ref *ref = NULL; struct btrfs_leaf_ref_tree *tree = root->ref_tree; @@ -101,17 +112,18 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root) return 0; spin_lock(&tree->lock); - while(!btrfs_leaf_ref_tree_empty(tree)) { - rb = rb_first(&tree->root); - ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node); + while(!list_empty(&tree->list)) { + ref = list_entry(tree->list.next, struct btrfs_leaf_ref, list); + BUG_ON(!ref->in_tree); + if (ref->root_gen > max_root_gen) + break; + rb_erase(&ref->rb_node, &tree->root); ref->in_tree = 0; list_del_init(&ref->list); spin_unlock(&tree->lock); - - btrfs_free_leaf_ref(ref); - + btrfs_free_leaf_ref(root, ref); cond_resched(); spin_lock(&tree->lock); } @@ -143,7 +155,6 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { int ret = 0; struct rb_node *rb; - size_t size = btrfs_leaf_ref_size(ref->nritems); struct btrfs_leaf_ref_tree *tree = root->ref_tree; spin_lock(&tree->lock); @@ -151,9 +162,6 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) if (rb) { ret = -EEXIST; } else { - spin_lock(&root->fs_info->ref_cache_lock); - root->fs_info->total_ref_cache_size += size; - spin_unlock(&root->fs_info->ref_cache_lock); atomic_inc(&ref->usage); list_add_tail(&ref->list, &tree->list); } @@ -163,15 +171,10 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { - size_t size = btrfs_leaf_ref_size(ref->nritems); struct btrfs_leaf_ref_tree *tree = root->ref_tree; BUG_ON(!ref->in_tree); spin_lock(&tree->lock); - - spin_lock(&root->fs_info->ref_cache_lock); - root->fs_info->total_ref_cache_size -= size; - spin_unlock(&root->fs_info->ref_cache_lock); rb_erase(&ref->rb_node, &tree->root); ref->in_tree = 0; @@ -179,7 +182,6 @@ int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) spin_unlock(&tree->lock); - btrfs_free_leaf_ref(ref); + btrfs_free_leaf_ref(root, ref); return 0; } - diff --git a/fs/btrfs/ref-cache.h b/fs/btrfs/ref-cache.h index 823c049f72f..c361b321c0c 100644 --- a/fs/btrfs/ref-cache.h +++ b/fs/btrfs/ref-cache.h @@ -30,6 +30,7 @@ struct btrfs_leaf_ref { int in_tree; atomic_t usage; + u64 root_gen; u64 bytenr; u64 owner; u64 generation; @@ -41,14 +42,13 @@ struct btrfs_leaf_ref { static inline size_t btrfs_leaf_ref_size(int nr_extents) { - return sizeof(struct btrfs_leaf_ref) + + return sizeof(struct btrfs_leaf_ref) + sizeof(struct btrfs_extent_info) * nr_extents; } static inline void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree) { tree->root.rb_node = NULL; - tree->last = NULL; INIT_LIST_HEAD(&tree->list); spin_lock_init(&tree->lock); } @@ -59,12 +59,13 @@ static inline int btrfs_leaf_ref_tree_empty(struct btrfs_leaf_ref_tree *tree) } void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree); -struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(int nr_extents); -void btrfs_free_leaf_ref(struct btrfs_leaf_ref *ref); +struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root, + int nr_extents); +void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref); struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root, u64 bytenr); int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref); -int btrfs_remove_leaf_refs(struct btrfs_root *root); +int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen); int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 216f3157162..52c5524896a 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -98,20 +98,24 @@ static noinline int record_root_in_trans(struct btrfs_root *root) BUG_ON(!dirty); dirty->root = kmalloc(sizeof(*dirty->root), GFP_NOFS); BUG_ON(!dirty->root); - dirty->latest_root = root; INIT_LIST_HEAD(&dirty->list); root->commit_root = btrfs_root_node(root); - root->dirty_root = dirty; memcpy(dirty->root, root, sizeof(*root)); - dirty->root->ref_tree = &root->ref_tree_struct; - spin_lock_init(&dirty->root->node_lock); + spin_lock_init(&dirty->root->list_lock); mutex_init(&dirty->root->objectid_mutex); + INIT_LIST_HEAD(&dirty->root->dead_list); dirty->root->node = root->commit_root; dirty->root->commit_root = NULL; + + spin_lock(&root->list_lock); + list_add(&dirty->root->dead_list, &root->dead_list); + spin_unlock(&root->list_lock); + + root->dirty_root = dirty; } else { WARN_ON(1); } @@ -356,8 +360,6 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, list_del_init(next); root = list_entry(next, struct btrfs_root, dirty_list); update_cowonly_root(trans, root); - if (root->fs_info->closing) - btrfs_remove_leaf_refs(root); } return 0; } @@ -410,7 +412,11 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, free_extent_buffer(root->commit_root); root->commit_root = NULL; - + + spin_lock(&root->list_lock); + list_del_init(&dirty->root->dead_list); + spin_unlock(&root->list_lock); + kfree(dirty->root); kfree(dirty); @@ -497,6 +503,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, unsigned long nr; u64 num_bytes; u64 bytes_used; + u64 max_useless; int ret = 0; int err; @@ -554,10 +561,25 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, } mutex_unlock(&root->fs_info->drop_mutex); + spin_lock(&root->list_lock); + list_del_init(&dirty->root->dead_list); + if (!list_empty(&root->dead_list)) { + struct btrfs_root *oldest; + oldest = list_entry(root->dead_list.prev, + struct btrfs_root, dead_list); + max_useless = oldest->root_key.offset - 1; + } else { + max_useless = root->root_key.offset - 1; + } + spin_unlock(&root->list_lock); + nr = trans->blocks_used; ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); + ret = btrfs_remove_leaf_refs(root, max_useless); + BUG_ON(ret); + free_extent_buffer(dirty->root->node); kfree(dirty->root); kfree(dirty); @@ -785,10 +807,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, put_transaction(cur_trans); put_transaction(cur_trans); + list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); if (root->fs_info->closing) list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); - else - list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); mutex_unlock(&root->fs_info->trans_mutex); kmem_cache_free(btrfs_trans_handle_cachep, trans); @@ -814,4 +835,3 @@ again: } return 0; } - diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 5e6ee7a6f73..18db4cbe279 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2527,4 +2527,3 @@ again: error: return ret; } - -- cgit v1.2.3-70-g09d2 From 0ee0fda06b943d7ef65f3cec50ab9d427a2baf97 Mon Sep 17 00:00:00 2001 From: Sven Wegener Date: Wed, 30 Jul 2008 16:54:26 -0400 Subject: Btrfs: Add compatibility for kernels >= 2.6.27-rc1 Add a couple of #if's to follow API changes. Signed-off-by: Sven Wegener Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 23 +++++++++++++++++++++++ fs/btrfs/file.c | 4 ++++ fs/btrfs/inode.c | 12 ++++++++++-- 3 files changed, 37 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5368e3b6eb9..1cf4baba434 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2649,9 +2649,24 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, mapping = eb->first_page->mapping; if (!mapping) return NULL; + + /* + * extent_buffer_page is only called after pinning the page + * by increasing the reference count. So we know the page must + * be in the radix tree. + */ +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + rcu_read_lock(); +#else read_lock_irq(&mapping->tree_lock); +#endif p = radix_tree_lookup(&mapping->page_tree, i); + +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + rcu_read_unlock(); +#else read_unlock_irq(&mapping->tree_lock); +#endif return p; } @@ -2843,13 +2858,21 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + spin_lock_irq(&page->mapping->tree_lock); +#else read_lock_irq(&page->mapping->tree_lock); +#endif if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + spin_unlock_irq(&page->mapping->tree_lock); +#else read_unlock_irq(&page->mapping->tree_lock); +#endif unlock_page(page); } return 0; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 412ab4a2638..388ac397c2c 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -886,7 +886,11 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, #ifdef REMOVE_SUID_PATH err = remove_suid(&file->f_path); #else +# if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + err = file_remove_suid(file); +# else err = remove_suid(fdentry(file)); +# endif #endif if (err) goto out_nolock; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7af8be076ee..c4afa9d78da 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3293,7 +3293,9 @@ void btrfs_destroy_inode(struct inode *inode) kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) +static void init_once(void *foo) +#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void init_once(struct kmem_cache * cachep, void *foo) #else static void init_once(void * foo, struct kmem_cache * cachep, @@ -3321,7 +3323,9 @@ void btrfs_destroy_cachep(void) struct kmem_cache *btrfs_cache_create(const char *name, size_t size, unsigned long extra_flags, -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + void (*ctor)(void *) +#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) void (*ctor)(struct kmem_cache *, void *) #else void (*ctor)(void *, struct kmem_cache *, @@ -3561,8 +3565,12 @@ static int btrfs_set_page_dirty(struct page *page) return __set_page_dirty_nobuffers(page); } +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) +static int btrfs_permission(struct inode *inode, int mask) +#else static int btrfs_permission(struct inode *inode, int mask, struct nameidata *nd) +#endif { if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE)) return -EACCES; -- cgit v1.2.3-70-g09d2 From 61b4944018449003ac5f9757f4d125dce519cf51 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 31 Jul 2008 15:42:53 -0400 Subject: Btrfs: Fix streaming read performance with checksumming on Large streaming reads make for large bios, which means each entry on the list async work queues represents a large amount of data. IO congestion throttling on the device was kicking in before the async worker threads decided a single thread was busy and needed some help. The end result was that a streaming read would result in a single CPU running at 100% instead of balancing the work off to other CPUs. This patch also changes the pre-IO checksum lookup done by reads to work on a per-bio basis instead of a per-page. This results in many extra btree lookups on large streaming reads. Doing the checksum lookup right before bio submit allows us to reuse searches while processing adjacent offsets. Signed-off-by: Chris Mason --- fs/btrfs/async-thread.c | 2 +- fs/btrfs/ctree.h | 2 ++ fs/btrfs/disk-io.c | 15 ++++++++++ fs/btrfs/file-item.c | 77 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/inode.c | 57 +++--------------------------------- 5 files changed, 99 insertions(+), 54 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index 5fe6a0d532e..bc2980c433e 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -160,7 +160,7 @@ void btrfs_init_workers(struct btrfs_workers *workers, int max) INIT_LIST_HEAD(&workers->idle_list); spin_lock_init(&workers->lock); workers->max_workers = max; - workers->idle_thresh = 64; + workers->idle_thresh = 32; } /* diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index be16cd49ef6..d788ab0dcd9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1613,6 +1613,8 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_key *location, int mod); /* file-item.c */ +int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, + struct bio *bio); int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, u64 disk_offset, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e826730d750..d2d1cc87e8a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1357,10 +1357,25 @@ struct btrfs_root *open_ctree(struct super_block *sb, */ btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size); + + /* a higher idle thresh on the submit workers makes it much more + * likely that bios will be send down in a sane order to the + * devices + */ + fs_info->submit_workers.idle_thresh = 64; + btrfs_init_workers(&fs_info->fixup_workers, 1); btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_write_workers, fs_info->thread_pool_size); + + /* + * endios are largely parallel and should have a very + * low idle thresh + */ + fs_info->endio_workers.idle_thresh = 4; + fs_info->endio_write_workers.idle_thresh = 4; + btrfs_start_workers(&fs_info->workers, 1); btrfs_start_workers(&fs_info->submit_workers, 1); btrfs_start_workers(&fs_info->fixup_workers, 1); diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 2311061f070..a5ff19b34b2 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -134,6 +134,83 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } +int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, + struct bio *bio) +{ + u32 sum; + struct bio_vec *bvec = bio->bi_io_vec; + int bio_index = 0; + u64 offset; + u64 item_start_offset = 0; + u64 item_last_offset = 0; + u32 diff; + int ret; + struct btrfs_path *path; + struct btrfs_csum_item *item = NULL; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + + path = btrfs_alloc_path(); + path->reada = 2; + + WARN_ON(bio->bi_vcnt <= 0); + + while(bio_index < bio->bi_vcnt) { + offset = page_offset(bvec->bv_page) + bvec->bv_offset; + ret = btrfs_find_ordered_sum(inode, offset, &sum); + if (ret == 0) + goto found; + + if (!item || offset < item_start_offset || + offset >= item_last_offset) { + struct btrfs_key found_key; + u32 item_size; + + if (item) + btrfs_release_path(root, path); + item = btrfs_lookup_csum(NULL, root, path, + inode->i_ino, offset, 0); + if (IS_ERR(item)) { + ret = PTR_ERR(item); + if (ret == -ENOENT || ret == -EFBIG) + ret = 0; + sum = 0; + printk("no csum found for inode %lu start " + "%llu\n", inode->i_ino, + (unsigned long long)offset); + goto found; + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + + item_start_offset = found_key.offset; + item_size = btrfs_item_size_nr(path->nodes[0], + path->slots[0]); + item_last_offset = item_start_offset + + (item_size / BTRFS_CRC32_SIZE) * + root->sectorsize; + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_csum_item); + } + /* + * this byte range must be able to fit inside + * a single leaf so it will also fit inside a u32 + */ + diff = offset - item_start_offset; + diff = diff / root->sectorsize; + diff = diff * BTRFS_CRC32_SIZE; + + read_extent_buffer(path->nodes[0], &sum, + (unsigned long)item + diff, + BTRFS_CRC32_SIZE); +found: + set_state_private(io_tree, offset, sum); + bio_index++; + bvec++; + } + btrfs_free_path(path); + return 0; +} + int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c4afa9d78da..31d52c51acc 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -374,6 +374,10 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, BUG_ON(ret); if (!(rw & (1 << BIO_RW))) { + if (!btrfs_test_opt(root, NODATASUM) && + !btrfs_test_flag(inode, NODATASUM)) { + btrfs_lookup_bio_sums(root, inode, bio); + } goto mapit; } @@ -598,58 +602,6 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return btrfs_finish_ordered_io(page->mapping->host, start, end); } -int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) -{ - int ret = 0; - struct inode *inode = page->mapping->host; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct btrfs_csum_item *item; - struct btrfs_path *path = NULL; - u32 csum; - - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) - return 0; - - /* - * It is possible there is an ordered extent that has - * not yet finished for this range in the file. If so, - * that extent will have a csum cached, and it will insert - * the sum after all the blocks in the extent are fully - * on disk. So, look for an ordered extent and use the - * sum if found. We have to do this before looking in the - * btree because csum items are pre-inserted based on - * the file size. btrfs_lookup_csum might find an item - * that still hasn't been fully filled. - */ - ret = btrfs_find_ordered_sum(inode, start, &csum); - if (ret == 0) - goto found; - - ret = 0; - path = btrfs_alloc_path(); - item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); - if (IS_ERR(item)) { - ret = PTR_ERR(item); - /* a csum that isn't present is a preallocated region. */ - if (ret == -ENOENT || ret == -EFBIG) - ret = 0; - csum = 0; - printk("no csum found for inode %lu start %Lu\n", inode->i_ino, - start); - goto out; - } - read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, - BTRFS_CRC32_SIZE); -found: - set_state_private(io_tree, start, csum); -out: - if (path) - btrfs_free_path(path); - return ret; -} - struct io_failure_record { struct page *page; u64 start; @@ -3613,7 +3565,6 @@ static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, .submit_bio_hook = btrfs_submit_bio_hook, .merge_bio_hook = btrfs_merge_bio_hook, - .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .writepage_end_io_hook = btrfs_writepage_end_io_hook, .writepage_start_hook = btrfs_writepage_start_hook, -- cgit v1.2.3-70-g09d2 From 3ce7e67a069b919be774a341b82fc20978b7f69d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 31 Jul 2008 15:42:54 -0400 Subject: Btrfs: Drop some debugging around the extent_map pinned flag Signed-off-by: Chris Mason --- fs/btrfs/file.c | 10 +--------- fs/btrfs/inode.c | 22 ---------------------- 2 files changed, 1 insertion(+), 31 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index d3f2fe0b7c6..c78f184ee5c 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -380,15 +380,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) spin_unlock(&em_tree->lock); break; } - if (test_bit(EXTENT_FLAG_PINNED, &em->flags)) { - printk(KERN_CRIT "inode %lu trying to drop pinned " - "extent start %llu end %llu, em [%llu %llu]\n", - inode->i_ino, - (unsigned long long)start, - (unsigned long long)end, - (unsigned long long)em->start, - (unsigned long long)em->len); - } + clear_bit(EXTENT_FLAG_PINNED, &em->flags); remove_extent_mapping(em_tree, em); if (em->block_start < EXTENT_MAP_LAST_BYTE && diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 31d52c51acc..76365ad9bfb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -505,15 +505,9 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; struct btrfs_ordered_extent *ordered_extent; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; - struct extent_map *em; - struct extent_map *em_orig; u64 alloc_hint = 0; - u64 clear_start; - u64 clear_end; struct list_head list; struct btrfs_key ins; - struct rb_node *rb; int ret; ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); @@ -542,22 +536,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) mutex_lock(&BTRFS_I(inode)->extent_mutex); - spin_lock(&em_tree->lock); - clear_start = ordered_extent->file_offset; - clear_end = ordered_extent->file_offset + ordered_extent->len; - em = lookup_extent_mapping(em_tree, clear_start, - ordered_extent->len); - em_orig = em; - while(em && clear_start < extent_map_end(em) && clear_end > em->start) { - clear_bit(EXTENT_FLAG_PINNED, &em->flags); - rb = rb_next(&em->rb_node); - if (!rb) - break; - em = rb_entry(rb, struct extent_map, rb_node); - } - free_extent_map(em_orig); - spin_unlock(&em_tree->lock); - ret = btrfs_drop_extents(trans, root, inode, ordered_extent->file_offset, ordered_extent->file_offset + -- cgit v1.2.3-70-g09d2 From 6dab81574346c831ded96ae3ab0e8f9ca72c37ae Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 4 Aug 2008 08:35:53 -0400 Subject: Btrfs: Hold csum mutex while reading in sums during readpages Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 2 +- fs/btrfs/inode.c | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index a5ff19b34b2..70af24aa582 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -150,7 +150,6 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; path = btrfs_alloc_path(); - path->reada = 2; WARN_ON(bio->bi_vcnt <= 0); @@ -177,6 +176,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, printk("no csum found for inode %lu start " "%llu\n", inode->i_ino, (unsigned long long)offset); + item = NULL; goto found; } btrfs_item_key_to_cpu(path->nodes[0], &found_key, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 76365ad9bfb..640648c66b2 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -376,7 +376,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, if (!(rw & (1 << BIO_RW))) { if (!btrfs_test_opt(root, NODATASUM) && !btrfs_test_flag(inode, NODATASUM)) { + mutex_lock(&BTRFS_I(inode)->csum_mutex); btrfs_lookup_bio_sums(root, inode, bio); + mutex_unlock(&BTRFS_I(inode)->csum_mutex); } goto mapit; } -- cgit v1.2.3-70-g09d2 From ea8c281947950fac5f78818b767821d696c9512a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 4 Aug 2008 23:17:27 -0400 Subject: Btrfs: Maintain a list of inodes that are delalloc and a way to wait on them Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 2 ++ fs/btrfs/ctree.h | 4 ++++ fs/btrfs/disk-io.c | 1 + fs/btrfs/extent-tree.c | 12 +++++++++--- fs/btrfs/file.c | 3 +-- fs/btrfs/inode.c | 47 +++++++++++++++++++++++++++++++++++++++++------ fs/btrfs/ioctl.c | 4 ++-- 7 files changed, 60 insertions(+), 13 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index df624fd735c..534ac4153dd 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -42,6 +42,8 @@ struct btrfs_inode { /* for keeping track of orphaned inodes */ struct list_head i_orphan; + struct list_head delalloc_inodes; + /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 62499dd761b..116aee21bf7 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -551,6 +551,7 @@ struct btrfs_fs_info { */ spinlock_t ordered_extent_lock; struct list_head ordered_extents; + struct list_head delalloc_inodes; /* * there is a pool of worker threads for checksumming during writes @@ -637,6 +638,7 @@ struct btrfs_root { struct kobject root_kobj; struct completion kobj_unregister; struct mutex objectid_mutex; + u64 objectid; u64 last_trans; @@ -1651,6 +1653,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, #define PageChecked PageFsMisc #endif +int btrfs_start_delalloc_inodes(struct btrfs_root *root); +int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end); int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc); int btrfs_create_subvol_root(struct btrfs_root *new_root, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index da9dda4338a..76543683f3b 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1234,6 +1234,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->trans_list); INIT_LIST_HEAD(&fs_info->dead_roots); INIT_LIST_HEAD(&fs_info->hashers); + INIT_LIST_HEAD(&fs_info->delalloc_inodes); spin_lock_init(&fs_info->hash_lock); spin_lock_init(&fs_info->delalloc_lock); spin_lock_init(&fs_info->new_trans_lock); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index dbde1288124..33cb2ac4cb2 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1230,7 +1230,6 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, found->total_bytes += total_bytes; found->bytes_used += bytes_used; found->full = 0; - WARN_ON(found->total_bytes < found->bytes_used); *space_info = found; return 0; } @@ -2841,8 +2840,7 @@ again: */ clear_page_dirty_for_io(page); - set_extent_delalloc(io_tree, page_start, - page_end, GFP_NOFS); + btrfs_set_extent_delalloc(inode, page_start, page_end); set_page_dirty(page); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); @@ -3319,6 +3317,13 @@ again: key.type = 0; cur_byte = key.objectid; + mutex_unlock(&root->fs_info->alloc_mutex); + + btrfs_start_delalloc_inodes(root); + btrfs_wait_ordered_extents(tree_root); + + mutex_lock(&root->fs_info->alloc_mutex); + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto out; @@ -3401,6 +3406,7 @@ next: btrfs_clean_old_snapshots(tree_root); + btrfs_start_delalloc_inodes(root); btrfs_wait_ordered_extents(tree_root); trans = btrfs_start_transaction(tree_root, 1); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 8915f2dc1bc..eb8e4556fa7 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -312,8 +312,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, * to reset the delalloc bit on things that already have * extents reserved. */ - set_extent_delalloc(io_tree, start_pos, - end_of_last_block, GFP_NOFS); + btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block); for (i = 0; i < num_pages; i++) { struct page *p = pages[i]; SetPageUptodate(p); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 640648c66b2..8a405a5fa6a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -303,6 +303,10 @@ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); BTRFS_I(inode)->delalloc_bytes += end - start + 1; root->fs_info->delalloc_bytes += end - start + 1; + if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) { + list_add_tail(&BTRFS_I(inode)->delalloc_inodes, + &root->fs_info->delalloc_inodes); + } spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); } return 0; @@ -325,6 +329,10 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, root->fs_info->delalloc_bytes -= end - start + 1; BTRFS_I(inode)->delalloc_bytes -= end - start + 1; } + if (BTRFS_I(inode)->delalloc_bytes == 0 && + !list_empty(&BTRFS_I(inode)->delalloc_inodes)) { + list_del_init(&BTRFS_I(inode)->delalloc_inodes); + } spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); } return 0; @@ -408,6 +416,12 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, return 0; } +int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end) +{ + return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, + GFP_NOFS); +} + struct btrfs_writepage_fixup { struct page *page; struct btrfs_work work; @@ -453,8 +467,7 @@ again: goto again; } - set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end, - GFP_NOFS); + btrfs_set_extent_delalloc(inode, page_start, page_end); ClearPageChecked(page); out: unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); @@ -1530,8 +1543,7 @@ again: goto again; } - set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, - page_end, GFP_NOFS); + btrfs_set_extent_delalloc(inode, page_start, page_end); ret = 0; if (offset != PAGE_CACHE_SIZE) { kaddr = kmap(page); @@ -1766,6 +1778,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); @@ -2158,6 +2171,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); + INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; @@ -2400,6 +2414,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; @@ -3049,8 +3064,7 @@ again: goto again; } - set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, - page_end, GFP_NOFS); + btrfs_set_extent_delalloc(inode, page_start, page_end); ret = 0; /* page is wholly or partially inside EOF */ @@ -3373,6 +3387,26 @@ out_unlock: return ret; } +int btrfs_start_delalloc_inodes(struct btrfs_root *root) +{ + struct list_head *head = &root->fs_info->delalloc_inodes; + struct btrfs_inode *binode; + unsigned long flags; + + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + while(!list_empty(head)) { + binode = list_entry(head->next, struct btrfs_inode, + delalloc_inodes); + atomic_inc(&binode->vfs_inode.i_count); + spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + filemap_write_and_wait(binode->vfs_inode.i_mapping); + iput(&binode->vfs_inode); + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + } + spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + return 0; +} + static int btrfs_symlink(struct inode *dir, struct dentry *dentry, const char *symname) { @@ -3436,6 +3470,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, inode->i_mapping, GFP_NOFS); + INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 0b63c3c77cf..e1046a54b1c 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -274,8 +274,7 @@ again: */ clear_page_dirty_for_io(page); - set_extent_delalloc(io_tree, page_start, - page_end, GFP_NOFS); + btrfs_set_extent_delalloc(inode, page_start, page_end); unlock_extent(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); @@ -784,6 +783,7 @@ long btrfs_ioctl(struct file *file, unsigned int case BTRFS_IOC_TRANS_END: return btrfs_ioctl_trans_end(file); case BTRFS_IOC_SYNC: + btrfs_start_delalloc_inodes(root); btrfs_sync_fs(file->f_dentry->d_sb, 1); return 0; } -- cgit v1.2.3-70-g09d2 From 3de9d6b649b4cc60687be92e71cef36d7d4e8f2f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 4 Aug 2008 23:17:27 -0400 Subject: btrfs_lookup_bio_sums seems broken, go back to the readpage_io_hook for now Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 4 +++- fs/btrfs/inode.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++------ 2 files changed, 56 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 70af24aa582..51aba8cee7c 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -134,6 +134,7 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } +#if 0 /* broken */ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct bio *bio) { @@ -200,7 +201,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, diff = diff * BTRFS_CRC32_SIZE; read_extent_buffer(path->nodes[0], &sum, - (unsigned long)item + diff, + ((unsigned long)item) + diff, BTRFS_CRC32_SIZE); found: set_state_private(io_tree, offset, sum); @@ -210,6 +211,7 @@ found: btrfs_free_path(path); return 0; } +#endif int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8a405a5fa6a..99121a55ffb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -382,12 +382,6 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, BUG_ON(ret); if (!(rw & (1 << BIO_RW))) { - if (!btrfs_test_opt(root, NODATASUM) && - !btrfs_test_flag(inode, NODATASUM)) { - mutex_lock(&BTRFS_I(inode)->csum_mutex); - btrfs_lookup_bio_sums(root, inode, bio); - mutex_unlock(&BTRFS_I(inode)->csum_mutex); - } goto mapit; } @@ -595,6 +589,58 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return btrfs_finish_ordered_io(page->mapping->host, start, end); } +int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) +{ + int ret = 0; + struct inode *inode = page->mapping->host; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct btrfs_csum_item *item; + struct btrfs_path *path = NULL; + u32 csum; + + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) + return 0; + + /* + * It is possible there is an ordered extent that has + * not yet finished for this range in the file. If so, + * that extent will have a csum cached, and it will insert + * the sum after all the blocks in the extent are fully + * on disk. So, look for an ordered extent and use the + * sum if found. We have to do this before looking in the + * btree because csum items are pre-inserted based on + * the file size. btrfs_lookup_csum might find an item + * that still hasn't been fully filled. + */ + ret = btrfs_find_ordered_sum(inode, start, &csum); + if (ret == 0) + goto found; + + ret = 0; + path = btrfs_alloc_path(); + item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); + if (IS_ERR(item)) { + ret = PTR_ERR(item); + /* a csum that isn't present is a preallocated region. */ + if (ret == -ENOENT || ret == -EFBIG) + ret = 0; + csum = 0; + printk("no csum found for inode %lu start %Lu\n", inode->i_ino, + start); + goto out; + } + read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, + BTRFS_CRC32_SIZE); +found: + set_state_private(io_tree, start, csum); +out: + if (path) + btrfs_free_path(path); + return ret; +} + struct io_failure_record { struct page *page; u64 start; @@ -3580,6 +3626,7 @@ static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, .submit_bio_hook = btrfs_submit_bio_hook, .merge_bio_hook = btrfs_merge_bio_hook, + .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .writepage_end_io_hook = btrfs_writepage_end_io_hook, .writepage_start_hook = btrfs_writepage_start_hook, -- cgit v1.2.3-70-g09d2 From 00e4e6b33a0f78aab4b788d6d31c884fd8bf88da Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 5 Aug 2008 11:18:09 -0400 Subject: Get rid of BTRFS_I(inode)->index and use local vars instead rename and link don't always have a lock on the source inode, and our use of a per-inode index variable was racy. This changes things to store the index in a local variable instead. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 9 -------- fs/btrfs/inode.c | 59 +++++++++++++++++++++++++++++--------------------- 2 files changed, 34 insertions(+), 34 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 534ac4153dd..111f90524ae 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -57,15 +57,6 @@ struct btrfs_inode { * number for new files that are created */ u64 index_cnt; - - /* - * index holds the directory index for this inode on creation, so - * add_link can do what its supposed to. This isn't populated when the - * inode is read because there isn't really a reason to know this unless - * we are creating the directory index or deleting it, and deletion - * reads the index off of the inode reference at unlink time. - */ - u64 index; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 99121a55ffb..4d8ffc01931 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2153,7 +2153,8 @@ out: return ret; } -static int btrfs_set_inode_index(struct inode *dir, struct inode *inode) +static int btrfs_set_inode_index(struct inode *dir, struct inode *inode, + u64 *index) { int ret = 0; @@ -2163,7 +2164,7 @@ static int btrfs_set_inode_index(struct inode *dir, struct inode *inode) return ret; } - BTRFS_I(inode)->index = BTRFS_I(dir)->index_cnt; + *index = BTRFS_I(dir)->index_cnt; BTRFS_I(dir)->index_cnt++; return ret; @@ -2176,7 +2177,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, u64 ref_objectid, u64 objectid, struct btrfs_block_group_cache *group, - int mode) + int mode, u64 *index) { struct inode *inode; struct btrfs_inode_item *inode_item; @@ -2198,11 +2199,9 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, return ERR_PTR(-ENOMEM); if (dir) { - ret = btrfs_set_inode_index(dir, inode); + ret = btrfs_set_inode_index(dir, inode, index); if (ret) return ERR_PTR(ret); - } else { - BTRFS_I(inode)->index = 0; } /* * index_cnt is ignored for everything but a dir, @@ -2268,7 +2267,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, struct btrfs_inode_ref); btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); - btrfs_set_inode_ref_index(path->nodes[0], ref, BTRFS_I(inode)->index); + btrfs_set_inode_ref_index(path->nodes[0], ref, *index); ptr = (unsigned long)(ref + 1); write_extent_buffer(path->nodes[0], name, ptr, name_len); @@ -2296,7 +2295,7 @@ static inline u8 btrfs_inode_type(struct inode *inode) static int btrfs_add_link(struct btrfs_trans_handle *trans, struct dentry *dentry, struct inode *inode, - int add_backref) + int add_backref, u64 index) { int ret; struct btrfs_key key; @@ -2311,7 +2310,7 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, &key, btrfs_inode_type(inode), - BTRFS_I(inode)->index); + index); if (ret == 0) { if (add_backref) { ret = btrfs_insert_inode_ref(trans, root, @@ -2319,7 +2318,7 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, dentry->d_name.len, inode->i_ino, parent_inode->i_ino, - BTRFS_I(inode)->index); + index); } btrfs_i_size_write(parent_inode, parent_inode->i_size + dentry->d_name.len * 2); @@ -2332,9 +2331,9 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans, static int btrfs_add_nondir(struct btrfs_trans_handle *trans, struct dentry *dentry, struct inode *inode, - int backref) + int backref, u64 index) { - int err = btrfs_add_link(trans, dentry, inode, backref); + int err = btrfs_add_link(trans, dentry, inode, backref, index); if (!err) { d_instantiate(dentry, inode); return 0; @@ -2354,6 +2353,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, int drop_inode = 0; u64 objectid; unsigned long nr = 0; + u64 index = 0; if (!new_valid_dev(rdev)) return -EINVAL; @@ -2374,7 +2374,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, - BTRFS_I(dir)->block_group, mode); + BTRFS_I(dir)->block_group, mode, &index); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; @@ -2386,7 +2386,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, } btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode, 0); + err = btrfs_add_nondir(trans, dentry, inode, 0, index); if (err) drop_inode = 1; else { @@ -2419,6 +2419,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, int drop_inode = 0; unsigned long nr = 0; u64 objectid; + u64 index = 0; err = btrfs_check_free_space(root, 1, 0); if (err) @@ -2435,7 +2436,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, - objectid, BTRFS_I(dir)->block_group, mode); + objectid, BTRFS_I(dir)->block_group, mode, + &index); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; @@ -2447,7 +2449,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, } btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode, 0); + err = btrfs_add_nondir(trans, dentry, inode, 0, index); if (err) drop_inode = 1; else { @@ -2489,6 +2491,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; struct inode *inode = old_dentry->d_inode; + u64 index; unsigned long nr = 0; int err; int drop_inode = 0; @@ -2504,7 +2507,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; - err = btrfs_set_inode_index(dir, inode); + err = btrfs_set_inode_index(dir, inode, &index); if (err) goto fail; @@ -2513,7 +2516,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, btrfs_set_trans_block_group(trans, dir); atomic_inc(&inode->i_count); - err = btrfs_add_nondir(trans, dentry, inode, 1); + err = btrfs_add_nondir(trans, dentry, inode, 1, index); if (err) drop_inode = 1; @@ -2544,6 +2547,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) int err = 0; int drop_on_err = 0; u64 objectid = 0; + u64 index = 0; unsigned long nr = 1; err = btrfs_check_free_space(root, 1, 0); @@ -2567,7 +2571,8 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, - BTRFS_I(dir)->block_group, S_IFDIR | mode); + BTRFS_I(dir)->block_group, S_IFDIR | mode, + &index); if (IS_ERR(inode)) { err = PTR_ERR(inode); goto out_fail; @@ -2588,7 +2593,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) if (err) goto out_fail; - err = btrfs_add_link(trans, dentry, inode, 0); + err = btrfs_add_link(trans, dentry, inode, 0, index); if (err) goto out_fail; @@ -3203,9 +3208,10 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct btrfs_block_group_cache *block_group) { struct inode *inode; + u64 index = 0; inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, - new_dirid, block_group, S_IFDIR | 0700); + new_dirid, block_group, S_IFDIR | 0700, &index); if (IS_ERR(inode)) return PTR_ERR(inode); inode->i_op = &btrfs_dir_inode_operations; @@ -3384,6 +3390,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, struct inode *new_inode = new_dentry->d_inode; struct inode *old_inode = old_dentry->d_inode; struct timespec ctime = CURRENT_TIME; + u64 index = 0; int ret; if (S_ISDIR(old_inode->i_mode) && new_inode && @@ -3419,11 +3426,11 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, goto out_fail; } } - ret = btrfs_set_inode_index(new_dir, old_inode); + ret = btrfs_set_inode_index(new_dir, old_inode, &index); if (ret) goto out_fail; - ret = btrfs_add_link(trans, new_dentry, old_inode, 1); + ret = btrfs_add_link(trans, new_dentry, old_inode, 1, index); if (ret) goto out_fail; @@ -3464,6 +3471,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, int err; int drop_inode = 0; u64 objectid; + u64 index = 0 ; int name_len; int datasize; unsigned long ptr; @@ -3491,7 +3499,8 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dentry->d_parent->d_inode->i_ino, objectid, - BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO); + BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO, + &index); err = PTR_ERR(inode); if (IS_ERR(inode)) goto out_unlock; @@ -3503,7 +3512,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, } btrfs_set_trans_block_group(trans, inode); - err = btrfs_add_nondir(trans, dentry, inode, 0); + err = btrfs_add_nondir(trans, dentry, inode, 0, index); if (err) drop_inode = 1; else { -- cgit v1.2.3-70-g09d2 From 7ea394f1192bee1af67ea4762c88ef4b7b0487a8 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Tue, 5 Aug 2008 13:05:02 -0400 Subject: Btrfs: Fix nodatacow for the new data=ordered mode Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 3 ++- fs/btrfs/extent-tree.c | 11 +++++---- fs/btrfs/inode.c | 60 +++++++++++++++++++++++++++++++------------------ fs/btrfs/ioctl.c | 1 + fs/btrfs/ordered-data.c | 16 ++++++++++--- fs/btrfs/ordered-data.h | 6 +++-- fs/btrfs/transaction.c | 11 +++++++++ 7 files changed, 74 insertions(+), 34 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 116aee21bf7..f90e5a7ac16 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1403,7 +1403,8 @@ static inline struct dentry *fdentry(struct file *file) { } /* extent-tree.c */ -int btrfs_cross_ref_exists(struct btrfs_root *root, +int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_key *key, u64 bytenr); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 33cb2ac4cb2..fff219ed61d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -893,10 +893,10 @@ out: return ret; } -int btrfs_cross_ref_exists(struct btrfs_root *root, +int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_key *key, u64 bytenr) { - struct btrfs_trans_handle *trans; struct btrfs_root *old_root; struct btrfs_path *path = NULL; struct extent_buffer *eb; @@ -908,6 +908,7 @@ int btrfs_cross_ref_exists(struct btrfs_root *root, int level; int ret; + BUG_ON(trans == NULL); BUG_ON(key->type != BTRFS_EXTENT_DATA_KEY); ret = get_reference_status(root, bytenr, 0, key->objectid, &min_generation, &ref_count); @@ -917,7 +918,6 @@ int btrfs_cross_ref_exists(struct btrfs_root *root, if (ref_count != 1) return 1; - trans = btrfs_start_transaction(root, 0); old_root = root->dirty_root->root; ref_generation = old_root->root_key.offset; @@ -973,7 +973,6 @@ int btrfs_cross_ref_exists(struct btrfs_root *root, out: if (path) btrfs_free_path(path); - btrfs_end_transaction(trans, root); return ret; } @@ -3320,7 +3319,7 @@ again: mutex_unlock(&root->fs_info->alloc_mutex); btrfs_start_delalloc_inodes(root); - btrfs_wait_ordered_extents(tree_root); + btrfs_wait_ordered_extents(tree_root, 0); mutex_lock(&root->fs_info->alloc_mutex); @@ -3407,7 +3406,7 @@ next: btrfs_clean_old_snapshots(tree_root); btrfs_start_delalloc_inodes(root); - btrfs_wait_ordered_extents(tree_root); + btrfs_wait_ordered_extents(tree_root, 0); trans = btrfs_start_transaction(tree_root, 1); btrfs_commit_transaction(trans, tree_root); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4d8ffc01931..c33053ba381 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -166,7 +166,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) cur_alloc_size = ins.offset; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, - ins.offset); + ins.offset, 0); BUG_ON(ret); if (num_bytes < cur_alloc_size) { printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes, @@ -187,31 +187,32 @@ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) u64 extent_start; u64 extent_end; u64 bytenr; - u64 cow_end; u64 loops = 0; u64 total_fs_bytes; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_block_group_cache *block_group; + struct btrfs_trans_handle *trans; struct extent_buffer *leaf; int found_type; struct btrfs_path *path; struct btrfs_file_extent_item *item; int ret; - int err; + int err = 0; struct btrfs_key found_key; total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); path = btrfs_alloc_path(); BUG_ON(!path); + trans = btrfs_join_transaction(root, 1); + BUG_ON(!trans); again: ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, start, 0); if (ret < 0) { - btrfs_free_path(path); - return ret; + err = ret; + goto out; } - cow_end = end; if (ret != 0) { if (path->slots[0] == 0) goto not_found; @@ -244,12 +245,11 @@ again: if (start < extent_start || start >= extent_end) goto not_found; - cow_end = min(end, extent_end - 1); bytenr = btrfs_file_extent_disk_bytenr(leaf, item); if (bytenr == 0) goto not_found; - if (btrfs_cross_ref_exists(root, &found_key, bytenr)) + if (btrfs_cross_ref_exists(trans, root, &found_key, bytenr)) goto not_found; /* * we may be called by the resizer, make sure we're inside @@ -260,24 +260,32 @@ again: if (!block_group || block_group->ro) goto not_found; + bytenr += btrfs_file_extent_offset(leaf, item); + extent_num_bytes = min(end + 1, extent_end) - start; + ret = btrfs_add_ordered_extent(inode, start, bytenr, + extent_num_bytes, 1); + if (ret) { + err = ret; + goto out; + } + + btrfs_release_path(root, path); start = extent_end; + if (start <= end) { + loops++; + goto again; + } } else { - goto not_found; - } -loop: - if (start > end) { +not_found: + btrfs_end_transaction(trans, root); btrfs_free_path(path); - return 0; + return cow_file_range(inode, start, end); } - btrfs_release_path(root, path); - loops++; - goto again; - -not_found: - btrfs_release_path(root, path); - cow_file_range(inode, start, end); - start = end + 1; - goto loop; +out: + WARN_ON(err); + btrfs_end_transaction(trans, root); + btrfs_free_path(path); + return err; } static int run_delalloc_range(struct inode *inode, u64 start, u64 end) @@ -385,6 +393,11 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, goto mapit; } + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) { + goto mapit; + } + return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, __btrfs_submit_bio_hook); @@ -527,6 +540,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ordered_extent = btrfs_lookup_ordered_extent(inode, start); BUG_ON(!ordered_extent); + if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) + goto nocow; lock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, @@ -567,6 +582,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) unlock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, GFP_NOFS); +nocow: add_pending_csums(trans, inode, ordered_extent->file_offset, &ordered_extent->list); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 3932c7cd0fa..59b64c738fd 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -36,6 +36,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index db200e6baf7..da6d43eb41d 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -152,7 +152,7 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, * inserted. */ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len) + u64 start, u64 len, int nocow) { struct btrfs_ordered_inode_tree *tree; struct rb_node *node; @@ -168,6 +168,8 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, entry->start = start; entry->len = len; entry->inode = inode; + if (nocow) + set_bit(BTRFS_ORDERED_NOCOW, &entry->flags); /* one ref for the tree */ atomic_set(&entry->refs, 1); @@ -303,10 +305,11 @@ int btrfs_remove_ordered_extent(struct inode *inode, return 0; } -int btrfs_wait_ordered_extents(struct btrfs_root *root) +int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only) { struct list_head splice; struct list_head *cur; + struct list_head *tmp; struct btrfs_ordered_extent *ordered; struct inode *inode; @@ -314,10 +317,16 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root) spin_lock(&root->fs_info->ordered_extent_lock); list_splice_init(&root->fs_info->ordered_extents, &splice); - while(!list_empty(&splice)) { + list_for_each_safe(cur, tmp, &splice) { cur = splice.next; ordered = list_entry(cur, struct btrfs_ordered_extent, root_extent_list); + if (nocow_only && + !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) { + cond_resched_lock(&root->fs_info->ordered_extent_lock); + continue; + } + list_del_init(&ordered->root_extent_list); atomic_inc(&ordered->refs); inode = ordered->inode; @@ -338,6 +347,7 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root) spin_lock(&root->fs_info->ordered_extent_lock); } + list_splice_init(&splice, &root->fs_info->ordered_extents); spin_unlock(&root->fs_info->ordered_extent_lock); return 0; } diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 5efe6b63c74..fd45519f30a 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -64,6 +64,8 @@ struct btrfs_ordered_sum { #define BTRFS_ORDERED_COMPLETE 1 /* set when removed from the tree */ +#define BTRFS_ORDERED_NOCOW 2 /* set when we want to write in place */ + struct btrfs_ordered_extent { /* logical offset in the file */ u64 file_offset; @@ -125,7 +127,7 @@ int btrfs_remove_ordered_extent(struct inode *inode, int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size); int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len); + u64 start, u64 len, int nocow); int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_extent *entry, struct btrfs_ordered_sum *sum); @@ -143,5 +145,5 @@ int btrfs_wait_on_page_writeback_range(struct address_space *mapping, pgoff_t start, pgoff_t end); int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, loff_t end, int sync_mode); -int btrfs_wait_ordered_extents(struct btrfs_root *root); +int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index ebf5362da1d..9d3d08e9f8d 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -438,6 +438,7 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, free_extent_buffer(root->commit_root); root->commit_root = NULL; + root->dirty_root = NULL; spin_lock(&root->list_lock); list_del_init(&dirty->root->dead_list); @@ -461,6 +462,7 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, sizeof(struct btrfs_disk_key)); root->root_item.drop_level = 0; root->commit_root = NULL; + root->dirty_root = NULL; root->root_key.offset = root->fs_info->generation; btrfs_set_root_bytenr(&root->root_item, root->node->start); @@ -762,7 +764,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, } do { + int snap_pending = 0; joined = cur_trans->num_joined; + if (!list_empty(&trans->transaction->pending_snapshots)) + snap_pending = 1; + WARN_ON(cur_trans != trans->transaction); prepare_to_wait(&cur_trans->writer_wait, &wait, TASK_UNINTERRUPTIBLE); @@ -774,6 +780,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, mutex_unlock(&root->fs_info->trans_mutex); + if (snap_pending) { + ret = btrfs_wait_ordered_extents(root, 1); + BUG_ON(ret); + } + schedule_timeout(timeout); mutex_lock(&root->fs_info->trans_mutex); -- cgit v1.2.3-70-g09d2 From 4ca8b41e3fe76cc5774fbcc6cba64259d69e0f96 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 5 Aug 2008 13:30:48 -0400 Subject: Btrfs: Avoid calling into the FS for the final iput on fake root inodes Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 1 + fs/btrfs/inode.c | 3 +++ 2 files changed, 4 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 76543683f3b..56c54a41dbb 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1739,6 +1739,7 @@ int close_ctree(struct btrfs_root *root) free_extent_buffer(root->fs_info->dev_root->node); btrfs_free_block_groups(root->fs_info); + fs_info->closing = 2; del_fs_roots(fs_info); filemap_write_and_wait(fs_info->btree_inode->i_mapping); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c33053ba381..a26d365111e 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2092,6 +2092,9 @@ int btrfs_write_inode(struct inode *inode, int wait) struct btrfs_trans_handle *trans; int ret = 0; + if (root->fs_info->closing > 1) + return 0; + if (wait) { trans = btrfs_join_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); -- cgit v1.2.3-70-g09d2 From db69e0ebae944690de89851315404f483e6464e0 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 15 Aug 2008 15:34:14 -0400 Subject: Btrfs: Init address_space->writeback_index properly The writeback_index field is used by write_cache_pages to pick up where writeback on a given inode left off. But, it is never set to a sane value, so writeback can often start at a random offset in the file. Kernels 2.6.28 and higher will have this fixed, but for everyone else, we also fill in the value in btrfs. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a26d365111e..33b990878d8 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1833,6 +1833,7 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) inode->i_ino = args->ino; BTRFS_I(inode)->root = args->root; BTRFS_I(inode)->delalloc_bytes = 0; + inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->index_cnt = (u64)-1; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); @@ -2239,6 +2240,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; + inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->root = root; @@ -2486,6 +2488,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; + inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } @@ -3549,6 +3552,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, mutex_init(&BTRFS_I(inode)->extent_mutex); BTRFS_I(inode)->delalloc_bytes = 0; BTRFS_I(inode)->disk_i_size = 0; + inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } -- cgit v1.2.3-70-g09d2 From 53863232ef961778aa414b700ed88a48e8e871e6 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 15 Aug 2008 15:34:18 -0400 Subject: Btrfs: Lower contention on the csum mutex This takes the csum mutex deeper in the call chain and releases it more often. Signed-off-by: Chris Mason --- fs/btrfs/async-thread.c | 3 ++- fs/btrfs/disk-io.c | 9 ++++++++- fs/btrfs/file-item.c | 16 ++++++++++++---- fs/btrfs/inode.c | 2 -- 4 files changed, 22 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index 958cd8b5f0d..2ee30174019 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -48,6 +48,7 @@ struct btrfs_worker_thread { /* number of things on the pending list */ atomic_t num_pending; + unsigned long sequence; /* protects the pending list. */ @@ -242,7 +243,7 @@ static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers) worker = list_entry(next, struct btrfs_worker_thread, worker_list); atomic_inc(&worker->num_pending); worker->sequence++; - if (worker->sequence % 4 == 0) + if (worker->sequence % workers->idle_thresh == 0) list_move_tail(next, &workers->worker_list); return worker; } diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 1aed1f4616b..92e14dd9bdd 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1386,7 +1386,14 @@ struct btrfs_root *open_ctree(struct super_block *sb, * devices */ fs_info->submit_workers.idle_thresh = 64; - fs_info->workers.idle_thresh = 32; + + /* fs_info->workers is responsible for checksumming file data + * blocks and metadata. Using a larger idle thresh allows each + * worker thread to operate on things in roughly the order they + * were sent by the writeback daemons, improving overall locality + * of the IO going down the pipe. + */ + fs_info->workers.idle_thresh = 128; btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); btrfs_init_workers(&fs_info->endio_workers, "endio", diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 51aba8cee7c..9454e0a07c8 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -321,6 +321,7 @@ again: file_key.offset = offset; btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); + mutex_lock(&BTRFS_I(inode)->csum_mutex); item = btrfs_lookup_csum(trans, root, path, objectid, offset, 1); if (!IS_ERR(item)) { leaf = path->nodes[0]; @@ -367,7 +368,7 @@ again: ret = btrfs_search_slot(trans, root, &file_key, path, BTRFS_CRC32_SIZE, 1); if (ret < 0) - goto fail; + goto fail_unlock; if (ret == 0) { BUG(); } @@ -411,10 +412,10 @@ insert: ret = btrfs_insert_empty_item(trans, root, path, &file_key, ins_size); if (ret < 0) - goto fail; + goto fail_unlock; if (ret != 0) { WARN_ON(1); - goto fail; + goto fail_unlock; } csum: leaf = path->nodes[0]; @@ -427,6 +428,8 @@ found: item_end = (struct btrfs_csum_item *)((unsigned char *)item_end + btrfs_item_size_nr(leaf, path->slots[0])); eb_token = NULL; + mutex_unlock(&BTRFS_I(inode)->csum_mutex); + cond_resched(); next_sector: if (!eb_token || @@ -467,13 +470,18 @@ next_sector: eb_token = NULL; } btrfs_mark_buffer_dirty(path->nodes[0]); + cond_resched(); if (total_bytes < sums->len) { btrfs_release_path(root, path); goto again; } -fail: +out: btrfs_free_path(path); return ret; + +fail_unlock: + mutex_unlock(&BTRFS_I(inode)->csum_mutex); + goto out; } int btrfs_csum_truncate(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 33b990878d8..65107894a5b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -415,10 +415,8 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, btrfs_set_trans_block_group(trans, inode); list_for_each(cur, list) { sum = list_entry(cur, struct btrfs_ordered_sum, list); - mutex_lock(&BTRFS_I(inode)->csum_mutex); btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, inode, sum); - mutex_unlock(&BTRFS_I(inode)->csum_mutex); } return 0; } -- cgit v1.2.3-70-g09d2 From 7c2fe32a238eb12422beca5cbd5194a594baa559 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Aug 2008 08:51:50 -0400 Subject: Btrfs: Fix add_extent_mapping to check for duplicates across the whole range add_extent_mapping was allowing the insertion of overlapping extents. This never used to happen because it only inserted the extents from disk and those were never overlapping. But, with the data=ordered code, the disk and memory representations of the file are not the same. add_extent_mapping needs to ensure a new extent does not overlap before it inserts. Signed-off-by: Chris Mason --- fs/btrfs/extent_map.c | 7 +++++++ fs/btrfs/inode.c | 21 +++++++++++++++++---- 2 files changed, 24 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 954b047639a..78ced11d18c 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -207,7 +207,14 @@ int add_extent_mapping(struct extent_map_tree *tree, int ret = 0; struct extent_map *merge = NULL; struct rb_node *rb; + struct extent_map *exist; + exist = lookup_extent_mapping(tree, em->start, em->len); + if (exist) { + free_extent_map(exist); + ret = -EEXIST; + goto out; + } assert_spin_locked(&tree->lock); rb = tree_insert(&tree->map, em->start, &em->rb_node); if (rb) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 65107894a5b..6c778043207 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -641,8 +641,9 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) if (ret == -ENOENT || ret == -EFBIG) ret = 0; csum = 0; - printk("no csum found for inode %lu start %Lu\n", inode->i_ino, - start); + if (printk_ratelimit()) + printk("no csum found for inode %lu start %Lu\n", inode->i_ino, + start); goto out; } read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, @@ -1653,8 +1654,20 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) btrfs_truncate_page(inode->i_mapping, inode->i_size); hole_size = block_end - hole_start; - btrfs_wait_ordered_range(inode, hole_start, hole_size); - lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + while(1) { + struct btrfs_ordered_extent *ordered; + btrfs_wait_ordered_range(inode, hole_start, hole_size); + + lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + ordered = btrfs_lookup_ordered_extent(inode, hole_start); + if (ordered) { + unlock_extent(io_tree, hole_start, + block_end - 1, GFP_NOFS); + btrfs_put_ordered_extent(ordered); + } else { + break; + } + } trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); -- cgit v1.2.3-70-g09d2 From 4d1b5fb4d7075f862848dbff8873e22382abd482 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 20 Aug 2008 09:44:52 -0400 Subject: Btrfs: Lookup readpage checksums on bio submission again This optimization had been removed because I thought it was triggering csum errors. The real cause of the errors was elsewhere, and so this optimization is back. Signed-off-by: Chris Mason --- fs/btrfs/file-item.c | 4 ++-- fs/btrfs/inode.c | 62 ++++------------------------------------------------ 2 files changed, 6 insertions(+), 60 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 9454e0a07c8..f861bd50692 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -134,7 +134,6 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } -#if 0 /* broken */ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct bio *bio) { @@ -151,6 +150,8 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; path = btrfs_alloc_path(); + if (bio->bi_size > PAGE_CACHE_SIZE * 8) + path->reada = 2; WARN_ON(bio->bi_vcnt <= 0); @@ -211,7 +212,6 @@ found: btrfs_free_path(path); return 0; } -#endif int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6c778043207..38d7a81763c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -389,15 +389,15 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); - if (!(rw & (1 << BIO_RW))) { - goto mapit; - } - if (btrfs_test_opt(root, NODATASUM) || btrfs_test_flag(inode, NODATASUM)) { goto mapit; } + if (!(rw & (1 << BIO_RW))) { + btrfs_lookup_bio_sums(root, inode, bio); + goto mapit; + } return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, __btrfs_submit_bio_hook); @@ -603,59 +603,6 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return btrfs_finish_ordered_io(page->mapping->host, start, end); } -int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end) -{ - int ret = 0; - struct inode *inode = page->mapping->host; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct btrfs_csum_item *item; - struct btrfs_path *path = NULL; - u32 csum; - - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) - return 0; - - /* - * It is possible there is an ordered extent that has - * not yet finished for this range in the file. If so, - * that extent will have a csum cached, and it will insert - * the sum after all the blocks in the extent are fully - * on disk. So, look for an ordered extent and use the - * sum if found. We have to do this before looking in the - * btree because csum items are pre-inserted based on - * the file size. btrfs_lookup_csum might find an item - * that still hasn't been fully filled. - */ - ret = btrfs_find_ordered_sum(inode, start, &csum); - if (ret == 0) - goto found; - - ret = 0; - path = btrfs_alloc_path(); - item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0); - if (IS_ERR(item)) { - ret = PTR_ERR(item); - /* a csum that isn't present is a preallocated region. */ - if (ret == -ENOENT || ret == -EFBIG) - ret = 0; - csum = 0; - if (printk_ratelimit()) - printk("no csum found for inode %lu start %Lu\n", inode->i_ino, - start); - goto out; - } - read_extent_buffer(path->nodes[0], &csum, (unsigned long)item, - BTRFS_CRC32_SIZE); -found: - set_state_private(io_tree, start, csum); -out: - if (path) - btrfs_free_path(path); - return ret; -} - struct io_failure_record { struct page *page; u64 start; @@ -3669,7 +3616,6 @@ static struct extent_io_ops btrfs_extent_io_ops = { .fill_delalloc = run_delalloc_range, .submit_bio_hook = btrfs_submit_bio_hook, .merge_bio_hook = btrfs_merge_bio_hook, - .readpage_io_hook = btrfs_readpage_io_hook, .readpage_end_io_hook = btrfs_readpage_end_io_hook, .writepage_end_io_hook = btrfs_writepage_end_io_hook, .writepage_start_hook = btrfs_writepage_start_hook, -- cgit v1.2.3-70-g09d2 From 1a54ef8c11a0eaef59ff418089f109c27f8bd48d Mon Sep 17 00:00:00 2001 From: Balaji Rao Date: Mon, 21 Jul 2008 02:01:04 +0530 Subject: Introduce btrfs_iget helper Date: Mon, 21 Jul 2008 02:01:04 +0530 This patch introduces a btrfs_iget helper to be used in NFS support. Signed-off-by: Balaji Rao Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 ++ fs/btrfs/inode.c | 55 +++++++++++++++++++++++++++++++++++++------------------ 2 files changed, 39 insertions(+), 18 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 04021335939..cecf29e0363 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1700,6 +1700,8 @@ struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, u64 root_objectid); +struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, + struct btrfs_root *root, int *is_new); int btrfs_commit_write(struct file *file, struct page *page, unsigned from, unsigned to); struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 38d7a81763c..266ca5514b3 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1840,6 +1840,33 @@ struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, return inode; } +/* Get an inode object given its location and corresponding root. + * Returns in *is_new if the inode was read from disk + */ +struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, + struct btrfs_root *root, int *is_new) +{ + struct inode *inode; + + inode = btrfs_iget_locked(s, location->objectid, root); + if (!inode) + return ERR_PTR(-EACCES); + + if (inode->i_state & I_NEW) { + BTRFS_I(inode)->root = root; + memcpy(&BTRFS_I(inode)->location, location, sizeof(*location)); + btrfs_read_locked_inode(inode); + unlock_new_inode(inode); + if (is_new) + *is_new = 1; + } else { + if (is_new) + *is_new = 0; + } + + return inode; +} + static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) { @@ -1848,7 +1875,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = bi->root; struct btrfs_root *sub_root = root; struct btrfs_key location; - int ret, do_orphan = 0; + int ret, new, do_orphan = 0; if (dentry->d_name.len > BTRFS_NAME_LEN) return ERR_PTR(-ENAMETOOLONG); @@ -1866,23 +1893,15 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, return ERR_PTR(ret); if (ret > 0) return ERR_PTR(-ENOENT); - - inode = btrfs_iget_locked(dir->i_sb, location.objectid, - sub_root); - if (!inode) - return ERR_PTR(-EACCES); - if (inode->i_state & I_NEW) { - /* the inode and parent dir are two different roots */ - if (sub_root != root) { - igrab(inode); - sub_root->inode = inode; - do_orphan = 1; - } - BTRFS_I(inode)->root = sub_root; - memcpy(&BTRFS_I(inode)->location, &location, - sizeof(location)); - btrfs_read_locked_inode(inode); - unlock_new_inode(inode); + inode = btrfs_iget(dir->i_sb, &location, sub_root, &new); + if (IS_ERR(inode)) + return ERR_CAST(inode); + + /* the inode and parent dir are two different roots */ + if (new && root != sub_root) { + igrab(inode); + sub_root->inode = inode; + do_orphan = 1; } } -- cgit v1.2.3-70-g09d2 From cbdf5a2442330102c08f5a2ad3058e29e90a43a9 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Wed, 6 Aug 2008 19:42:33 +0100 Subject: Btrfs: Implement our own copy of the nfsd readdir hack, for older kernels Date: Wed, 6 Aug 2008 19:42:33 +0100 Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 102 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 266ca5514b3..0da1ae4ee40 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1915,7 +1915,8 @@ static unsigned char btrfs_filetype_table[] = { DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK }; -static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir) +static int btrfs_real_readdir(struct file *filp, void *dirent, + filldir_t filldir) { struct inode *inode = filp->f_dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -2064,6 +2065,101 @@ err: return ret; } +/* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd + will call the file system's ->lookup() method from within its + filldir callback, which in turn was called from the file system's + ->readdir() method. And will deadlock for many file systems. */ +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) + +struct nfshack_dirent { + u64 ino; + loff_t offset; + int namlen; + unsigned int d_type; + char name[]; +}; + +struct nfshack_readdir { + char *dirent; + size_t used; +}; + + + +static int btrfs_nfshack_filldir(void *__buf, const char *name, int namlen, + loff_t offset, u64 ino, unsigned int d_type) +{ + struct nfshack_readdir *buf = __buf; + struct nfshack_dirent *de = (void *)(buf->dirent + buf->used); + unsigned int reclen; + + reclen = ALIGN(sizeof(struct nfshack_dirent) + namlen, sizeof(u64)); + if (buf->used + reclen > PAGE_SIZE) + return -EINVAL; + + de->namlen = namlen; + de->offset = offset; + de->ino = ino; + de->d_type = d_type; + memcpy(de->name, name, namlen); + buf->used += reclen; + + return 0; +} + +static int btrfs_nfshack_readdir(struct file *file, void *dirent, + filldir_t filldir) +{ + struct nfshack_readdir buf; + struct nfshack_dirent *de; + int err; + int size; + loff_t offset; + + buf.dirent = (void *)__get_free_page(GFP_KERNEL); + if (!buf.dirent) + return -ENOMEM; + + offset = file->f_pos; + + while (1) { + unsigned int reclen; + + buf.used = 0; + + err = btrfs_real_readdir(file, &buf, btrfs_nfshack_filldir); + if (err) + break; + + size = buf.used; + + if (!size) + break; + + de = (struct nfshack_dirent *)buf.dirent; + while (size > 0) { + offset = de->offset; + + if (filldir(dirent, de->name, de->namlen, de->offset, + de->ino, de->d_type)) + goto done; + offset = file->f_pos; + + reclen = ALIGN(sizeof(*de) + de->namlen, + sizeof(u64)); + size -= reclen; + de = (struct nfshack_dirent *)((char *)de + reclen); + } + } + + done: + free_page((unsigned long)buf.dirent); + file->f_pos = offset; + + return err; +} +#endif + int btrfs_write_inode(struct inode *inode, int wait) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -3623,7 +3719,11 @@ static struct inode_operations btrfs_dir_ro_inode_operations = { static struct file_operations btrfs_dir_file_operations = { .llseek = generic_file_llseek, .read = generic_read_dir, - .readdir = btrfs_readdir, +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) + .readdir = btrfs_nfshack_readdir, +#else /* NFSd readdir/lookup deadlock is fixed */ + .readdir = btrfs_real_readdir, +#endif .unlocked_ioctl = btrfs_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = btrfs_ioctl, -- cgit v1.2.3-70-g09d2 From 5ecc7e5d1d58731b3a6a69303d69aca7fb0fe1f5 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sun, 17 Aug 2008 15:14:48 +0100 Subject: Btrfs: Remove special cases for "." and ".." Date: Sun, 17 Aug 2008 15:14:48 +0100 We never get asked by the VFS to lookup either of them, and we can handle the readdir() case a lot more simply, too. Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 52 ++-------------------------------------------------- 1 file changed, 2 insertions(+), 50 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0da1ae4ee40..4520a0e86e7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1702,42 +1702,9 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = BTRFS_I(dir)->root; int ret = 0; - if (namelen == 1 && strcmp(name, ".") == 0) { - location->objectid = dir->i_ino; - location->type = BTRFS_INODE_ITEM_KEY; - location->offset = 0; - return 0; - } path = btrfs_alloc_path(); BUG_ON(!path); - if (namelen == 2 && strcmp(name, "..") == 0) { - struct btrfs_key key; - struct extent_buffer *leaf; - int slot; - - key.objectid = dir->i_ino; - key.offset = (u64)-1; - btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); - if (ret < 0 || path->slots[0] == 0) - goto out_err; - ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); - BUG_ON(ret == 0); - ret = 0; - leaf = path->nodes[0]; - slot = path->slots[0] - 1; - - btrfs_item_key_to_cpu(leaf, &key, slot); - if (key.objectid != dir->i_ino || - key.type != BTRFS_INODE_REF_KEY) { - goto out_err; - } - location->objectid = key.offset; - location->type = BTRFS_INODE_ITEM_KEY; - location->offset = 0; - goto out; - } - di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, namelen, 0); if (IS_ERR(di)) @@ -1960,29 +1927,14 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, /* special case for .., just use the back ref */ if (filp->f_pos == 1) { - btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY); - key.offset = (u64)-1; - ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); - if (ret < 0 || path->slots[0] == 0) { - btrfs_release_path(root, path); - goto read_dir_items; - } - BUG_ON(ret == 0); - leaf = path->nodes[0]; - slot = path->slots[0] - 1; - btrfs_item_key_to_cpu(leaf, &found_key, slot); - btrfs_release_path(root, path); - if (found_key.objectid != key.objectid || - found_key.type != BTRFS_INODE_REF_KEY) - goto read_dir_items; + u64 pino = parent_ino(filp->f_path.dentry); over = filldir(dirent, "..", 2, - 2, found_key.offset, DT_DIR); + 2, pino, DT_DIR); if (over) goto nopos; filp->f_pos = 2; } -read_dir_items: btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; -- cgit v1.2.3-70-g09d2 From 49593bfa575b7e3fda073b6d1033ee273bdaf97c Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sun, 17 Aug 2008 17:08:36 +0100 Subject: Minor cleanup of btrfs_real_readdir() Date: Sun, 17 Aug 2008 17:08:36 +0100 Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 36 +++++++++++++++++++++--------------- 1 file changed, 21 insertions(+), 15 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4520a0e86e7..c7b7095634d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1920,34 +1920,34 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, return 0; filp->f_pos = 1; } - - key.objectid = inode->i_ino; - path = btrfs_alloc_path(); - path->reada = 2; - /* special case for .., just use the back ref */ if (filp->f_pos == 1) { u64 pino = parent_ino(filp->f_path.dentry); over = filldir(dirent, "..", 2, 2, pino, DT_DIR); if (over) - goto nopos; + return 0; filp->f_pos = 2; } + path = btrfs_alloc_path(); + path->reada = 2; + btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; + key.objectid = inode->i_ino; ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; advance = 0; - while(1) { + + while (1) { leaf = path->nodes[0]; nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; if (advance || slot >= nritems) { - if (slot >= nritems -1) { + if (slot >= nritems - 1) { ret = btrfs_next_leaf(root, path); if (ret) break; @@ -1971,19 +1971,23 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, continue; filp->f_pos = found_key.offset; - advance = 1; + di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); di_cur = 0; di_total = btrfs_item_size(leaf, item); - while(di_cur < di_total) { + + while (di_cur < di_total) { struct btrfs_key location; name_len = btrfs_dir_name_len(leaf, di); - if (name_len < 32) { + if (name_len <= sizeof(tmp_name)) { name_ptr = tmp_name; } else { name_ptr = kmalloc(name_len, GFP_NOFS); - BUG_ON(!name_ptr); + if (!name_ptr) { + ret = -ENOMEM; + goto err; + } } read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1), name_len); @@ -1991,8 +1995,7 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)]; btrfs_dir_item_key_to_cpu(leaf, di, &location); over = filldir(dirent, name_ptr, name_len, - found_key.offset, - location.objectid, + found_key.offset, location.objectid, d_type); if (name_ptr != tmp_name) @@ -2000,12 +2003,15 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, if (over) goto nopos; + di_len = btrfs_dir_name_len(leaf, di) + - btrfs_dir_data_len(leaf, di) +sizeof(*di); + btrfs_dir_data_len(leaf, di) + sizeof(*di); di_cur += di_len; di = (struct btrfs_dir_item *)((char *)di + di_len); } } + + /* Reached end of directory/root. Bump pos past the last item. */ if (key_type == BTRFS_DIR_INDEX_KEY) filp->f_pos = INT_LIMIT(typeof(filp->f_pos)); else -- cgit v1.2.3-70-g09d2 From f2322b1c652add8bcd64b10843d76b0211ab1fc6 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Sun, 17 Aug 2008 17:12:56 +0100 Subject: Btrfs: Optimise NFS readdir hack slightly; don't call readdir() again when done Date: Sun, 17 Aug 2008 17:12:56 +0100 Signed-off-by: David Woodhouse Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c7b7095634d..10f26f44532 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2040,6 +2040,7 @@ struct nfshack_dirent { struct nfshack_readdir { char *dirent; size_t used; + int full; }; @@ -2052,8 +2053,10 @@ static int btrfs_nfshack_filldir(void *__buf, const char *name, int namlen, unsigned int reclen; reclen = ALIGN(sizeof(struct nfshack_dirent) + namlen, sizeof(u64)); - if (buf->used + reclen > PAGE_SIZE) + if (buf->used + reclen > PAGE_SIZE) { + buf->full = 1; return -EINVAL; + } de->namlen = namlen; de->offset = offset; @@ -2080,11 +2083,11 @@ static int btrfs_nfshack_readdir(struct file *file, void *dirent, offset = file->f_pos; - while (1) { + do { unsigned int reclen; buf.used = 0; - + buf.full = 0; err = btrfs_real_readdir(file, &buf, btrfs_nfshack_filldir); if (err) break; @@ -2108,7 +2111,7 @@ static int btrfs_nfshack_readdir(struct file *file, void *dirent, size -= reclen; de = (struct nfshack_dirent *)((char *)de + reclen); } - } + } while (buf.full); done: free_page((unsigned long)buf.dirent); -- cgit v1.2.3-70-g09d2 From 95819c05732c511338b43c115ffbcee978c02888 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 28 Aug 2008 06:21:17 -0400 Subject: Btrfs: optimize btrget/set/removexattr btrfs actually stores the whole xattr name, including the prefix ondisk, so using the generic resolver that strips off the prefix is not very helpful. Instead do the real ondisk xattrs manually and only use the generic resolver for synthetic xattrs like ACLs. (Sorry Josef for guiding you towards the wrong direction here intially) Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/acl.c | 20 +++--- fs/btrfs/inode.c | 19 +++--- fs/btrfs/xattr.c | 204 ++++++++++++++++++++----------------------------------- fs/btrfs/xattr.h | 31 +++------ 4 files changed, 104 insertions(+), 170 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c index 2f865311460..867eaf1f8ef 100644 --- a/fs/btrfs/acl.c +++ b/fs/btrfs/acl.c @@ -42,17 +42,18 @@ static void btrfs_update_cached_acl(struct inode *inode, static struct posix_acl *btrfs_get_acl(struct inode *inode, int type) { - int size, name_index; + int size; + const char *name; char *value = NULL; struct posix_acl *acl = NULL, **p_acl; switch (type) { case ACL_TYPE_ACCESS: - name_index = BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS; + name = POSIX_ACL_XATTR_ACCESS; p_acl = &BTRFS_I(inode)->i_acl; break; case ACL_TYPE_DEFAULT: - name_index = BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT; + name = POSIX_ACL_XATTR_DEFAULT; p_acl = &BTRFS_I(inode)->i_default_acl; break; default: @@ -68,12 +69,12 @@ static struct posix_acl *btrfs_get_acl(struct inode *inode, int type) return acl; - size = btrfs_xattr_get(inode, name_index, "", NULL, 0); + size = __btrfs_getxattr(inode, name, "", 0); if (size > 0) { value = kzalloc(size, GFP_NOFS); if (!value) return ERR_PTR(-ENOMEM); - size = btrfs_xattr_get(inode, name_index, "", value, size); + size = __btrfs_getxattr(inode, name, value, size); if (size > 0) { acl = posix_acl_from_xattr(value, size); btrfs_update_cached_acl(inode, p_acl, acl); @@ -110,7 +111,8 @@ static int btrfs_xattr_get_acl(struct inode *inode, int type, */ static int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type) { - int ret, name_index = 0, size = 0; + int ret, size = 0; + const char *name; struct posix_acl **p_acl; char *value = NULL; mode_t mode; @@ -130,13 +132,13 @@ static int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type) return ret; ret = 0; inode->i_mode = mode; - name_index = BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS; + name = POSIX_ACL_XATTR_ACCESS; p_acl = &BTRFS_I(inode)->i_acl; break; case ACL_TYPE_DEFAULT: if (!S_ISDIR(inode->i_mode)) return acl ? -EINVAL : 0; - name_index = BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT; + name = POSIX_ACL_XATTR_DEFAULT; p_acl = &BTRFS_I(inode)->i_default_acl; break; default: @@ -156,7 +158,7 @@ static int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type) goto out; } - ret = btrfs_xattr_set(inode, name_index, "", value, size, 0); + ret = __btrfs_setxattr(inode, name, value, size, 0); out: if (value) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 10f26f44532..43d3f2649ca 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -45,6 +45,7 @@ #include "print-tree.h" #include "volumes.h" #include "ordered-data.h" +#include "xattr.h" struct btrfs_iget_args { u64 ino; @@ -3667,10 +3668,10 @@ static struct inode_operations btrfs_dir_inode_operations = { .symlink = btrfs_symlink, .setattr = btrfs_setattr, .mknod = btrfs_mknod, - .setxattr = generic_setxattr, - .getxattr = generic_getxattr, + .setxattr = btrfs_setxattr, + .getxattr = btrfs_getxattr, .listxattr = btrfs_listxattr, - .removexattr = generic_removexattr, + .removexattr = btrfs_removexattr, .permission = btrfs_permission, }; static struct inode_operations btrfs_dir_ro_inode_operations = { @@ -3728,20 +3729,20 @@ static struct inode_operations btrfs_file_inode_operations = { .truncate = btrfs_truncate, .getattr = btrfs_getattr, .setattr = btrfs_setattr, - .setxattr = generic_setxattr, - .getxattr = generic_getxattr, + .setxattr = btrfs_setxattr, + .getxattr = btrfs_getxattr, .listxattr = btrfs_listxattr, - .removexattr = generic_removexattr, + .removexattr = btrfs_removexattr, .permission = btrfs_permission, }; static struct inode_operations btrfs_special_inode_operations = { .getattr = btrfs_getattr, .setattr = btrfs_setattr, .permission = btrfs_permission, - .setxattr = generic_setxattr, - .getxattr = generic_getxattr, + .setxattr = btrfs_setxattr, + .getxattr = btrfs_getxattr, .listxattr = btrfs_listxattr, - .removexattr = generic_removexattr, + .removexattr = btrfs_removexattr, }; static struct inode_operations btrfs_symlink_inode_operations = { .readlink = generic_readlink, diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c index fdfece41dd1..adb4b32a9d5 100644 --- a/fs/btrfs/xattr.c +++ b/fs/btrfs/xattr.c @@ -27,91 +27,20 @@ #include "xattr.h" #include "disk-io.h" -static struct xattr_handler *btrfs_xattr_handler_map[] = { - [BTRFS_XATTR_INDEX_USER] = &btrfs_xattr_user_handler, -#ifdef CONFIG_FS_POSIX_ACL - [BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS] = &btrfs_xattr_acl_access_handler, - [BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT] = &btrfs_xattr_acl_default_handler, -#endif - [BTRFS_XATTR_INDEX_TRUSTED] = &btrfs_xattr_trusted_handler, - [BTRFS_XATTR_INDEX_SECURITY] = &btrfs_xattr_security_handler, - [BTRFS_XATTR_INDEX_SYSTEM] = &btrfs_xattr_system_handler, -}; - -struct xattr_handler *btrfs_xattr_handlers[] = { - &btrfs_xattr_user_handler, -#ifdef CONFIG_FS_POSIX_ACL - &btrfs_xattr_acl_access_handler, - &btrfs_xattr_acl_default_handler, -#endif - &btrfs_xattr_trusted_handler, - &btrfs_xattr_security_handler, - &btrfs_xattr_system_handler, - NULL, -}; - -/* - * @param name_index - the index for the xattr handler - * @return the xattr_handler if we found it, NULL otherwise - * - * use this if we know the type of the xattr already - */ -static struct xattr_handler *btrfs_xattr_handler(int name_index) -{ - struct xattr_handler *handler = NULL; - - if (name_index >= 0 && - name_index < ARRAY_SIZE(btrfs_xattr_handler_map)) - handler = btrfs_xattr_handler_map[name_index]; - - return handler; -} - -static inline char *get_name(const char *name, int name_index) -{ - char *ret = NULL; - struct xattr_handler *handler = btrfs_xattr_handler(name_index); - int prefix_len; - - if (!handler) - return ret; - - prefix_len = strlen(handler->prefix); - - ret = kmalloc(strlen(name) + prefix_len + 1, GFP_KERNEL); - if (!ret) - return ret; - - memcpy(ret, handler->prefix, prefix_len); - memcpy(ret+prefix_len, name, strlen(name)); - ret[prefix_len + strlen(name)] = '\0'; - - return ret; -} -ssize_t btrfs_xattr_get(struct inode *inode, int name_index, - const char *attr_name, void *buffer, size_t size) +ssize_t __btrfs_getxattr(struct inode *inode, const char *name, + void *buffer, size_t size) { struct btrfs_dir_item *di; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_path *path; struct extent_buffer *leaf; - struct xattr_handler *handler = btrfs_xattr_handler(name_index); int ret = 0; unsigned long data_ptr; - char *name; - - if (!handler) - return -EOPNOTSUPP; - name = get_name(attr_name, name_index); - if (!name) - return -ENOMEM; path = btrfs_alloc_path(); - if (!path) { - kfree(name); + if (!path) return -ENOMEM; - } /* lookup the xattr by name */ di = btrfs_lookup_xattr(NULL, root, path, inode->i_ino, name, @@ -140,33 +69,22 @@ ssize_t btrfs_xattr_get(struct inode *inode, int name_index, ret = btrfs_dir_data_len(leaf, di); out: - kfree(name); btrfs_free_path(path); return ret; } -int btrfs_xattr_set(struct inode *inode, int name_index, - const char *attr_name, const void *value, size_t size, - int flags) +int __btrfs_setxattr(struct inode *inode, const char *name, + const void *value, size_t size, int flags) { struct btrfs_dir_item *di; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; struct btrfs_path *path; - struct xattr_handler *handler = btrfs_xattr_handler(name_index); - char *name; int ret = 0, mod = 0; - if (!handler) - return -EOPNOTSUPP; - name = get_name(attr_name, name_index); - if (!name) - return -ENOMEM; path = btrfs_alloc_path(); - if (!path) { - kfree(name); + if (!path) return -ENOMEM; - } trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); @@ -221,9 +139,7 @@ out: } btrfs_end_transaction(trans, root); - kfree(name); btrfs_free_path(path); - return ret; } @@ -329,51 +245,77 @@ err: } /* - * Handler functions + * List of handlers for synthetic system.* attributes. All real ondisk + * attributes are handled directly. + */ +struct xattr_handler *btrfs_xattr_handlers[] = { +#ifdef CONFIG_FS_POSIX_ACL + &btrfs_xattr_acl_access_handler, + &btrfs_xattr_acl_default_handler, +#endif + NULL, +}; + +/* + * Check if the attribute is in a supported namespace. + * + * This applied after the check for the synthetic attributes in the system + * namespace. */ -#define BTRFS_XATTR_SETGET_FUNCS(name, index) \ -static int btrfs_xattr_##name##_get(struct inode *inode, \ - const char *name, void *value, \ - size_t size) \ -{ \ - if (*name == '\0') \ - return -EINVAL; \ - return btrfs_xattr_get(inode, index, name, value, size); \ -} \ -static int btrfs_xattr_##name##_set(struct inode *inode, \ - const char *name, const void *value,\ - size_t size, int flags) \ -{ \ - if (*name == '\0') \ - return -EINVAL; \ - return btrfs_xattr_set(inode, index, name, value, size, flags); \ +static bool btrfs_is_valid_xattr(const char *name) +{ + return !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) || + !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN) || + !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) || + !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN); } -BTRFS_XATTR_SETGET_FUNCS(security, BTRFS_XATTR_INDEX_SECURITY); -BTRFS_XATTR_SETGET_FUNCS(system, BTRFS_XATTR_INDEX_SYSTEM); -BTRFS_XATTR_SETGET_FUNCS(user, BTRFS_XATTR_INDEX_USER); -BTRFS_XATTR_SETGET_FUNCS(trusted, BTRFS_XATTR_INDEX_TRUSTED); +ssize_t btrfs_getxattr(struct dentry *dentry, const char *name, + void *buffer, size_t size) +{ + /* + * If this is a request for a synthetic attribute in the system.* + * namespace use the generic infrastructure to resolve a handler + * for it via sb->s_xattr. + */ + if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN)) + return generic_getxattr(dentry, name, buffer, size); -struct xattr_handler btrfs_xattr_security_handler = { - .prefix = XATTR_SECURITY_PREFIX, - .get = btrfs_xattr_security_get, - .set = btrfs_xattr_security_set, -}; + if (!btrfs_is_valid_xattr(name)) + return -EOPNOTSUPP; + return __btrfs_getxattr(dentry->d_inode, name, buffer, size); +} -struct xattr_handler btrfs_xattr_system_handler = { - .prefix = XATTR_SYSTEM_PREFIX, - .get = btrfs_xattr_system_get, - .set = btrfs_xattr_system_set, -}; +int btrfs_setxattr(struct dentry *dentry, const char *name, const void *value, + size_t size, int flags) +{ + /* + * If this is a request for a synthetic attribute in the system.* + * namespace use the generic infrastructure to resolve a handler + * for it via sb->s_xattr. + */ + if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN)) + return generic_setxattr(dentry, name, value, size, flags); -struct xattr_handler btrfs_xattr_user_handler = { - .prefix = XATTR_USER_PREFIX, - .get = btrfs_xattr_user_get, - .set = btrfs_xattr_user_set, -}; + if (!btrfs_is_valid_xattr(name)) + return -EOPNOTSUPP; -struct xattr_handler btrfs_xattr_trusted_handler = { - .prefix = XATTR_TRUSTED_PREFIX, - .get = btrfs_xattr_trusted_get, - .set = btrfs_xattr_trusted_set, -}; + if (size == 0) + value = ""; /* empty EA, do not remove */ + return __btrfs_setxattr(dentry->d_inode, name, value, size, flags); +} + +int btrfs_removexattr(struct dentry *dentry, const char *name) +{ + /* + * If this is a request for a synthetic attribute in the system.* + * namespace use the generic infrastructure to resolve a handler + * for it via sb->s_xattr. + */ + if (!strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN)) + return generic_removexattr(dentry, name); + + if (!btrfs_is_valid_xattr(name)) + return -EOPNOTSUPP; + return __btrfs_setxattr(dentry->d_inode, name, NULL, 0, XATTR_REPLACE); +} diff --git a/fs/btrfs/xattr.h b/fs/btrfs/xattr.h index 825e55bd496..5b1d08f8e68 100644 --- a/fs/btrfs/xattr.h +++ b/fs/btrfs/xattr.h @@ -20,31 +20,20 @@ #define __XATTR__ #include -#include "ctree.h" -/* Name indexes */ -enum { - BTRFS_XATTR_INDEX_USER, - BTRFS_XATTR_INDEX_POSIX_ACL_ACCESS, - BTRFS_XATTR_INDEX_POSIX_ACL_DEFAULT, - BTRFS_XATTR_INDEX_TRUSTED, - BTRFS_XATTR_INDEX_SECURITY, - BTRFS_XATTR_INDEX_SYSTEM, - BTRFS_XATTR_INDEX_END, -}; - -extern struct xattr_handler btrfs_xattr_user_handler; -extern struct xattr_handler btrfs_xattr_trusted_handler; extern struct xattr_handler btrfs_xattr_acl_access_handler; extern struct xattr_handler btrfs_xattr_acl_default_handler; -extern struct xattr_handler btrfs_xattr_security_handler; -extern struct xattr_handler btrfs_xattr_system_handler; - extern struct xattr_handler *btrfs_xattr_handlers[]; -ssize_t btrfs_xattr_get(struct inode *inode, int name_index, const char *name, - void *buffer, size_t size); -int btrfs_xattr_set(struct inode *inode, int name_index, const char *name, - const void *value, size_t size, int flags); +extern ssize_t __btrfs_getxattr(struct inode *inode, const char *name, + void *buffer, size_t size); +extern int __btrfs_setxattr(struct inode *inode, const char *name, + const void *value, size_t size, int flags); + +extern ssize_t btrfs_getxattr(struct dentry *dentry, const char *name, + void *buffer, size_t size); +extern int btrfs_setxattr(struct dentry *dentry, const char *name, + const void *value, size_t size, int flags); +extern int btrfs_removexattr(struct dentry *dentry, const char *name); #endif /* __XATTR__ */ -- cgit v1.2.3-70-g09d2 From e02119d5a7b4396c5a872582fddc8bd6d305a70a Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 5 Sep 2008 16:13:11 -0400 Subject: Btrfs: Add a write ahead tree log to optimize synchronous operations File syncs and directory syncs are optimized by copying their items into a special (copy-on-write) log tree. There is one log tree per subvolume and the btrfs super block points to a tree of log tree roots. After a crash, items are copied out of the log tree and back into the subvolume. See tree-log.c for all the details. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 3 +- fs/btrfs/btrfs_inode.h | 8 + fs/btrfs/compat.h | 15 + fs/btrfs/ctree.c | 73 +- fs/btrfs/ctree.h | 72 +- fs/btrfs/dir-item.c | 3 + fs/btrfs/disk-io.c | 138 ++- fs/btrfs/disk-io.h | 8 +- fs/btrfs/extent-tree.c | 93 +- fs/btrfs/file.c | 39 +- fs/btrfs/inode.c | 261 ++--- fs/btrfs/root-tree.c | 5 +- fs/btrfs/transaction.c | 45 +- fs/btrfs/transaction.h | 1 + fs/btrfs/tree-defrag.c | 4 +- fs/btrfs/tree-log.c | 2804 ++++++++++++++++++++++++++++++++++++++++++++++++ fs/btrfs/tree-log.h | 41 + 17 files changed, 3408 insertions(+), 205 deletions(-) create mode 100644 fs/btrfs/tree-log.c create mode 100644 fs/btrfs/tree-log.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 3a01065d4ef..b7addbfd8c2 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -7,8 +7,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \ - ref-cache.o acl.o export.o - + ref-cache.o export.o tree-log.o acl.o else # Normal Makefile diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 111f90524ae..fcc8cf27e90 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -33,6 +33,7 @@ struct btrfs_inode { struct extent_io_tree io_failure_tree; struct mutex csum_mutex; struct mutex extent_mutex; + struct mutex log_mutex; struct inode vfs_inode; struct btrfs_ordered_inode_tree ordered_tree; @@ -44,10 +45,17 @@ struct btrfs_inode { struct list_head delalloc_inodes; + /* full 64 bit generation number */ + u64 generation; + /* * transid of the trans_handle that last modified this inode */ u64 last_trans; + /* + * transid that last logged this inode + */ + u64 logged_trans; u64 delalloc_bytes; u64 disk_i_size; u32 flags; diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h index d45fb37887b..b0ed1887d9b 100644 --- a/fs/btrfs/compat.h +++ b/fs/btrfs/compat.h @@ -22,6 +22,21 @@ static inline struct dentry *d_obtain_alias(struct inode *inode) } #endif +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) +static inline void btrfs_drop_nlink(struct inode *inode) +{ + inode->i_nlink--; +} + +static inline void btrfs_inc_nlink(struct inode *inode) +{ + inode->i_nlink++; +} +#else +# define btrfs_drop_nlink(inode) drop_nlink(inode) +# define btrfs_inc_nlink(inode) inc_nlink(inode) +#endif + /* * Even if AppArmor isn't enabled, it still has different prototypes. * Add more distro/version pairs here to declare which has AppArmor applied. diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 7114faafa9d..579124043d9 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -60,7 +60,7 @@ void btrfs_free_path(struct btrfs_path *p) kmem_cache_free(btrfs_path_cachep, p); } -void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) +void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) { int i; @@ -176,7 +176,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, return 0; } -int __btrfs_cow_block(struct btrfs_trans_handle *trans, +int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, @@ -294,7 +294,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans, return 0; } -int btrfs_cow_block(struct btrfs_trans_handle *trans, +int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, struct extent_buffer **cow_ret, u64 prealloc_dest) @@ -677,9 +677,10 @@ static int noinline check_block(struct btrfs_root *root, * * slot may point to max if the key is bigger than all of the keys */ -static int generic_bin_search(struct extent_buffer *eb, unsigned long p, - int item_size, struct btrfs_key *key, - int max, int *slot) +static noinline int generic_bin_search(struct extent_buffer *eb, + unsigned long p, + int item_size, struct btrfs_key *key, + int max, int *slot) { int low = 0; int high = max; @@ -765,7 +766,7 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key, return -1; } -static struct extent_buffer *read_node_slot(struct btrfs_root *root, +static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root, struct extent_buffer *parent, int slot) { int level = btrfs_header_level(parent); @@ -781,7 +782,7 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root, btrfs_node_ptr_generation(parent, slot)); } -static int balance_level(struct btrfs_trans_handle *trans, +static noinline int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { @@ -1128,8 +1129,9 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, /* * readahead one full node of leaves */ -static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, - int level, int slot, u64 objectid) +static noinline void reada_for_search(struct btrfs_root *root, + struct btrfs_path *path, + int level, int slot, u64 objectid) { struct extent_buffer *node; struct btrfs_disk_key disk_key; @@ -1201,7 +1203,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, } } -static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock) +static noinline void unlock_up(struct btrfs_path *path, int level, + int lowest_unlock) { int i; int skip_level = level; @@ -1759,8 +1762,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root * * returns 0 on success and < 0 on failure */ -static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_path *path, int level) +static noinline int split_node(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int level) { u64 root_gen; struct extent_buffer *c; @@ -1874,7 +1878,8 @@ static int leaf_space_used(struct extent_buffer *l, int start, int nr) * the start of the leaf data. IOW, how much room * the leaf has left for both items and data */ -int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf) +int noinline btrfs_leaf_free_space(struct btrfs_root *root, + struct extent_buffer *leaf) { int nritems = btrfs_header_nritems(leaf); int ret; @@ -2283,9 +2288,11 @@ out: * * returns 0 if all went well and < 0 on failure. */ -static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root - *root, struct btrfs_key *ins_key, - struct btrfs_path *path, int data_size, int extend) +static noinline int split_leaf(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_key *ins_key, + struct btrfs_path *path, int data_size, + int extend) { u64 root_gen; struct extent_buffer *l; @@ -3079,6 +3086,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) * was nothing in the tree that matched the search criteria. */ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, + struct btrfs_key *max_key, struct btrfs_path *path, int cache_only, u64 min_trans) { @@ -3093,6 +3101,7 @@ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, again: cur = btrfs_lock_root_node(root); level = btrfs_header_level(cur); + WARN_ON(path->nodes[level]); path->nodes[level] = cur; path->locks[level] = 1; @@ -3107,6 +3116,8 @@ again: /* at level = 0, we're done, setup the path and exit */ if (level == 0) { + if (slot >= nritems) + goto find_next_key; ret = 0; path->slots[level] = slot; btrfs_item_key_to_cpu(cur, &found_key, slot); @@ -3123,6 +3134,8 @@ again: u64 blockptr; u64 gen; struct extent_buffer *tmp; + struct btrfs_disk_key disk_key; + blockptr = btrfs_node_blockptr(cur, slot); gen = btrfs_node_ptr_generation(cur, slot); if (gen < min_trans) { @@ -3132,6 +3145,14 @@ again: if (!cache_only) break; + if (max_key) { + btrfs_node_key(cur, &disk_key, slot); + if (comp_keys(&disk_key, max_key) >= 0) { + ret = 1; + goto out; + } + } + tmp = btrfs_find_tree_block(root, blockptr, btrfs_level_size(root, level - 1)); @@ -3143,14 +3164,16 @@ again: free_extent_buffer(tmp); slot++; } +find_next_key: /* * we didn't find a candidate key in this node, walk forward * and find another one */ if (slot >= nritems) { - ret = btrfs_find_next_key(root, path, min_key, level, + path->slots[level] = slot; + sret = btrfs_find_next_key(root, path, min_key, level, cache_only, min_trans); - if (ret == 0) { + if (sret == 0) { btrfs_release_path(root, path); goto again; } else { @@ -3351,6 +3374,7 @@ int btrfs_previous_item(struct btrfs_root *root, { struct btrfs_key found_key; struct extent_buffer *leaf; + u32 nritems; int ret; while(1) { @@ -3362,9 +3386,20 @@ int btrfs_previous_item(struct btrfs_root *root, path->slots[0]--; } leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + if (nritems == 0) + return 1; + if (path->slots[0] == nritems) + path->slots[0]--; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (found_key.type == type) return 0; + if (found_key.objectid < min_objectid) + break; + if (found_key.objectid == min_objectid && + found_key.type < type) + break; } return 1; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b305ae7e10b..6532b60683e 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -77,6 +77,10 @@ struct btrfs_ordered_sum; /* orhpan objectid for tracking unlinked/truncated files */ #define BTRFS_ORPHAN_OBJECTID -5ULL +/* does write ahead logging to speed up fsyncs */ +#define BTRFS_TREE_LOG_OBJECTID -6ULL +#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL + /* * All files have objectids higher than this. */ @@ -276,6 +280,7 @@ struct btrfs_super_block { __le64 generation; __le64 root; __le64 chunk_root; + __le64 log_root; __le64 total_bytes; __le64 bytes_used; __le64 root_dir_objectid; @@ -287,6 +292,7 @@ struct btrfs_super_block { __le32 sys_chunk_array_size; u8 root_level; u8 chunk_root_level; + u8 log_root_level; struct btrfs_dev_item dev_item; char label[BTRFS_LABEL_SIZE]; u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE]; @@ -392,7 +398,10 @@ struct btrfs_timespec { * make a new item type */ struct btrfs_inode_item { + /* nfs style generation number */ __le64 generation; + /* transid that last touched this inode */ + __le64 transid; __le64 size; __le64 nblocks; __le64 block_group; @@ -409,8 +418,13 @@ struct btrfs_inode_item { struct btrfs_timespec otime; } __attribute__ ((__packed__)); +struct btrfs_dir_log_item { + __le64 end; +} __attribute__ ((__packed__)); + struct btrfs_dir_item { struct btrfs_disk_key location; + __le64 transid; __le16 data_len; __le16 name_len; u8 type; @@ -505,6 +519,9 @@ struct btrfs_fs_info { struct btrfs_root *tree_root; struct btrfs_root *chunk_root; struct btrfs_root *dev_root; + + /* the log root tree is a directory of all the other log roots */ + struct btrfs_root *log_root_tree; struct radix_tree_root fs_roots_radix; struct extent_io_tree free_space_cache; @@ -518,6 +535,7 @@ struct btrfs_fs_info { u64 generation; u64 last_trans_committed; + u64 last_trans_new_blockgroup; u64 open_ioctl_trans; unsigned long mount_opt; u64 max_extent; @@ -527,6 +545,9 @@ struct btrfs_fs_info { wait_queue_head_t transaction_throttle; wait_queue_head_t transaction_wait; wait_queue_head_t async_submit_wait; + + wait_queue_head_t tree_log_wait; + struct btrfs_super_block super_copy; struct btrfs_super_block super_for_commit; struct block_device *__bdev; @@ -535,6 +556,7 @@ struct btrfs_fs_info { struct backing_dev_info bdi; spinlock_t hash_lock; struct mutex trans_mutex; + struct mutex tree_log_mutex; struct mutex transaction_kthread_mutex; struct mutex cleaner_mutex; struct mutex alloc_mutex; @@ -544,8 +566,13 @@ struct btrfs_fs_info { struct list_head trans_list; struct list_head hashers; struct list_head dead_roots; + atomic_t nr_async_submits; atomic_t nr_async_bios; + atomic_t tree_log_writers; + atomic_t tree_log_commit; + unsigned long tree_log_batch; + u64 tree_log_transid; /* * this is used by the balancing code to wait for all the pending @@ -583,6 +610,7 @@ struct btrfs_fs_info { struct completion kobj_unregister; int do_barriers; int closing; + int log_root_recovering; atomic_t throttles; atomic_t throttle_gen; @@ -596,6 +624,7 @@ struct btrfs_fs_info { u64 delalloc_bytes; u64 last_alloc; u64 last_data_alloc; + u64 last_log_alloc; spinlock_t ref_cache_lock; u64 total_ref_cache_size; @@ -632,6 +661,7 @@ struct btrfs_root { struct btrfs_leaf_ref_tree *ref_tree; struct btrfs_leaf_ref_tree ref_tree_struct; struct btrfs_dirty_root *dirty_root; + struct btrfs_root *log_root; struct btrfs_root_item root_item; struct btrfs_key root_key; @@ -640,6 +670,7 @@ struct btrfs_root { struct kobject root_kobj; struct completion kobj_unregister; struct mutex objectid_mutex; + struct mutex log_mutex; u64 objectid; u64 last_trans; @@ -692,6 +723,8 @@ struct btrfs_root { * dir items are the name -> inode pointers in a directory. There is one * for every name in a directory. */ +#define BTRFS_DIR_LOG_ITEM_KEY 14 +#define BTRFS_DIR_LOG_INDEX_KEY 15 #define BTRFS_DIR_ITEM_KEY 16 #define BTRFS_DIR_INDEX_KEY 17 /* @@ -703,7 +736,8 @@ struct btrfs_root { */ #define BTRFS_CSUM_ITEM_KEY 19 -/* reserve 20-31 for other file stuff */ + +/* reserve 21-31 for other file/dir stuff */ /* * root items point to tree roots. There are typically in the root @@ -938,6 +972,7 @@ BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64); /* struct btrfs_inode_item */ BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); +BTRFS_SETGET_FUNCS(inode_transid, struct btrfs_inode_item, transid, 64); BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64); BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64); BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64); @@ -1126,10 +1161,13 @@ static inline void btrfs_set_item_key(struct extent_buffer *eb, write_eb_member(eb, item, struct btrfs_item, key, disk_key); } +BTRFS_SETGET_FUNCS(dir_log_end, struct btrfs_dir_log_item, end, 64); + /* struct btrfs_dir_item */ BTRFS_SETGET_FUNCS(dir_data_len, struct btrfs_dir_item, data_len, 16); BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8); BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16); +BTRFS_SETGET_FUNCS(dir_transid, struct btrfs_dir_item, transid, 64); static inline void btrfs_dir_item_key(struct extent_buffer *eb, struct btrfs_dir_item *item, @@ -1301,7 +1339,11 @@ BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block, BTRFS_SETGET_STACK_FUNCS(super_chunk_root, struct btrfs_super_block, chunk_root, 64); BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block, - chunk_root_level, 64); + chunk_root_level, 8); +BTRFS_SETGET_STACK_FUNCS(super_log_root, struct btrfs_super_block, + log_root, 64); +BTRFS_SETGET_STACK_FUNCS(super_log_root_level, struct btrfs_super_block, + log_root_level, 8); BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block, total_bytes, 64); BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block, @@ -1405,6 +1447,12 @@ static inline struct dentry *fdentry(struct file *file) { } /* extent-tree.c */ +int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path, + u64 start, u64 len); +int btrfs_update_pinned_extents(struct btrfs_root *root, + u64 bytenr, u64 num, int pin); +int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *leaf); int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, u64 bytenr); @@ -1448,6 +1496,11 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins); +int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + struct btrfs_key *ins); int btrfs_reserve_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 min_alloc_size, @@ -1488,9 +1541,9 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *key, int lowest_level, int cache_only, u64 min_trans); int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, + struct btrfs_key *max_key, struct btrfs_path *path, int cache_only, u64 min_trans); - int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, @@ -1656,6 +1709,18 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, #define PageChecked PageFsMisc #endif +int btrfs_unlink_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *dir, struct inode *inode, + const char *name, int name_len); +int btrfs_add_link(struct btrfs_trans_handle *trans, + struct inode *parent_inode, struct inode *inode, + const char *name, int name_len, int add_backref, u64 index); +int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode, u64 new_size, + u32 min_type); + int btrfs_start_delalloc_inodes(struct btrfs_root *root); int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end); int btrfs_writepages(struct address_space *mapping, @@ -1715,6 +1780,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); /* file.c */ +int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync); int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); int btrfs_check_file(struct btrfs_root *root, struct inode *inode); extern struct file_operations btrfs_file_operations; diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 125094617fe..e4f30090d64 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -96,6 +96,7 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, btrfs_set_dir_item_key(leaf, dir_item, &disk_key); btrfs_set_dir_type(leaf, dir_item, BTRFS_FT_XATTR); btrfs_set_dir_name_len(leaf, dir_item, name_len); + btrfs_set_dir_transid(leaf, dir_item, trans->transid); btrfs_set_dir_data_len(leaf, dir_item, data_len); name_ptr = (unsigned long)(dir_item + 1); data_ptr = (unsigned long)((char *)name_ptr + name_len); @@ -142,6 +143,7 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_set_dir_type(leaf, dir_item, type); btrfs_set_dir_data_len(leaf, dir_item, 0); btrfs_set_dir_name_len(leaf, dir_item, name_len); + btrfs_set_dir_transid(leaf, dir_item, trans->transid); name_ptr = (unsigned long)(dir_item + 1); write_extent_buffer(leaf, name, name_ptr, name_len); @@ -169,6 +171,7 @@ second_insert: btrfs_set_dir_type(leaf, dir_item, type); btrfs_set_dir_data_len(leaf, dir_item, 0); btrfs_set_dir_name_len(leaf, dir_item, name_len); + btrfs_set_dir_transid(leaf, dir_item, trans->transid); name_ptr = (unsigned long)(dir_item + 1); write_extent_buffer(leaf, name, name_ptr, name_len); btrfs_mark_buffer_dirty(leaf); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 8e7a938bfbc..a4373db5967 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -41,6 +41,7 @@ #include "async-thread.h" #include "locking.h" #include "ref-cache.h" +#include "tree-log.h" #if 0 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) @@ -694,6 +695,18 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, } +int btrfs_write_tree_block(struct extent_buffer *buf) +{ + return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, + buf->start + buf->len - 1, WB_SYNC_NONE); +} + +int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) +{ + return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, + buf->start, buf->start + buf->len -1); +} + struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, u64 parent_transid) { @@ -732,15 +745,6 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, return 0; } -int wait_on_tree_block_writeback(struct btrfs_root *root, - struct extent_buffer *buf) -{ - struct inode *btree_inode = root->fs_info->btree_inode; - wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, - buf); - return 0; -} - static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, u32 stripesize, struct btrfs_root *root, struct btrfs_fs_info *fs_info, @@ -771,6 +775,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, spin_lock_init(&root->node_lock); spin_lock_init(&root->list_lock); mutex_init(&root->objectid_mutex); + mutex_init(&root->log_mutex); btrfs_leaf_ref_tree_init(&root->ref_tree_struct); root->ref_tree = &root->ref_tree_struct; @@ -809,11 +814,74 @@ static int find_and_setup_root(struct btrfs_root *tree_root, return 0; } -struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, - struct btrfs_key *location) +int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info) +{ + struct extent_buffer *eb; + int ret; + + if (!fs_info->log_root_tree) + return 0; + + eb = fs_info->log_root_tree->node; + + WARN_ON(btrfs_header_level(eb) != 0); + WARN_ON(btrfs_header_nritems(eb) != 0); + + ret = btrfs_free_extent(trans, fs_info->tree_root, + eb->start, eb->len, + BTRFS_TREE_LOG_OBJECTID, 0, 0, 0, 1); + BUG_ON(ret); + + free_extent_buffer(eb); + kfree(fs_info->log_root_tree); + fs_info->log_root_tree = NULL; + return 0; +} + +int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info) { struct btrfs_root *root; struct btrfs_root *tree_root = fs_info->tree_root; + + root = kzalloc(sizeof(*root), GFP_NOFS); + if (!root) + return -ENOMEM; + + __setup_root(tree_root->nodesize, tree_root->leafsize, + tree_root->sectorsize, tree_root->stripesize, + root, fs_info, BTRFS_TREE_LOG_OBJECTID); + + root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; + root->root_key.type = BTRFS_ROOT_ITEM_KEY; + root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; + root->ref_cows = 0; + + root->node = btrfs_alloc_free_block(trans, root, root->leafsize, + BTRFS_TREE_LOG_OBJECTID, + 0, 0, 0, 0, 0); + + btrfs_set_header_nritems(root->node, 0); + btrfs_set_header_level(root->node, 0); + btrfs_set_header_bytenr(root->node, root->node->start); + btrfs_set_header_generation(root->node, trans->transid); + btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); + + write_extent_buffer(root->node, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(root->node), + BTRFS_FSID_SIZE); + btrfs_mark_buffer_dirty(root->node); + btrfs_tree_unlock(root->node); + fs_info->log_root_tree = root; + return 0; +} + +struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, + struct btrfs_key *location) +{ + struct btrfs_root *root; + struct btrfs_fs_info *fs_info = tree_root->fs_info; struct btrfs_path *path; struct extent_buffer *l; u64 highest_inode; @@ -863,11 +931,13 @@ out: blocksize, 0); BUG_ON(!root->node); insert: - root->ref_cows = 1; - ret = btrfs_find_highest_inode(root, &highest_inode); - if (ret == 0) { - root->highest_inode = highest_inode; - root->last_inode_alloc = highest_inode; + if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { + root->ref_cows = 1; + ret = btrfs_find_highest_inode(root, &highest_inode); + if (ret == 0) { + root->highest_inode = highest_inode; + root->last_inode_alloc = highest_inode; + } } return root; } @@ -907,7 +977,7 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, if (root) return root; - root = btrfs_read_fs_root_no_radix(fs_info, location); + root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); if (IS_ERR(root)) return root; ret = radix_tree_insert(&fs_info->fs_roots_radix, @@ -1250,16 +1320,18 @@ struct btrfs_root *open_ctree(struct super_block *sb, u32 blocksize; u32 stripesize; struct buffer_head *bh; - struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), + struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); - struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), + struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), GFP_NOFS); - struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), + struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); - struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), + struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); + struct btrfs_root *log_tree_root; + int ret; int err = -EINVAL; @@ -1343,6 +1415,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); mutex_init(&fs_info->trans_mutex); + mutex_init(&fs_info->tree_log_mutex); mutex_init(&fs_info->drop_mutex); mutex_init(&fs_info->alloc_mutex); mutex_init(&fs_info->chunk_mutex); @@ -1352,6 +1425,10 @@ struct btrfs_root *open_ctree(struct super_block *sb, init_waitqueue_head(&fs_info->transaction_throttle); init_waitqueue_head(&fs_info->transaction_wait); init_waitqueue_head(&fs_info->async_submit_wait); + init_waitqueue_head(&fs_info->tree_log_wait); + atomic_set(&fs_info->tree_log_commit, 0); + atomic_set(&fs_info->tree_log_writers, 0); + fs_info->tree_log_transid = 0; #if 0 ret = add_hasher(fs_info, "crc32c"); @@ -1532,7 +1609,26 @@ struct btrfs_root *open_ctree(struct super_block *sb, if (!fs_info->transaction_kthread) goto fail_cleaner; + if (btrfs_super_log_root(disk_super) != 0) { + u32 blocksize; + u64 bytenr = btrfs_super_log_root(disk_super); + + blocksize = + btrfs_level_size(tree_root, + btrfs_super_log_root_level(disk_super)); + log_tree_root = kzalloc(sizeof(struct btrfs_root), + GFP_NOFS); + + __setup_root(nodesize, leafsize, sectorsize, stripesize, + log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); + + log_tree_root->node = read_tree_block(tree_root, bytenr, + blocksize, 0); + ret = btrfs_recover_log_trees(log_tree_root); + BUG_ON(ret); + } + fs_info->last_trans_committed = btrfs_super_generation(disk_super); return tree_root; fail_cleaner: diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 2562a273ae1..6b6fdc697f3 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -45,7 +45,7 @@ struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_key *location, const char *name, int namelen); -struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, +struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, struct btrfs_key *location); struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, struct btrfs_key *location); @@ -74,4 +74,10 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, extent_submit_bio_hook_t *submit_bio_hook); int btrfs_congested_async(struct btrfs_fs_info *info, int iodone); unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info); +int btrfs_write_tree_block(struct extent_buffer *buf); +int btrfs_wait_tree_block_writeback(struct extent_buffer *buf); +int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info); +int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info); #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index e63b3b4bed7..646b9148ca2 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -496,6 +496,23 @@ static int match_extent_ref(struct extent_buffer *leaf, return ret == 0; } +/* simple helper to search for an existing extent at a given offset */ +int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path, + u64 start, u64 len) +{ + int ret; + struct btrfs_key key; + + maybe_lock_mutex(root); + key.objectid = start; + key.offset = len; + btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); + ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path, + 0, 0); + maybe_unlock_mutex(root); + return ret; +} + static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 bytenr, @@ -1409,7 +1426,7 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start) } -static int update_pinned_extents(struct btrfs_root *root, +int btrfs_update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin) { u64 len; @@ -1492,7 +1509,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, EXTENT_DIRTY); if (ret) break; - update_pinned_extents(root, start, end + 1 - start, 0); + btrfs_update_pinned_extents(root, start, end + 1 - start, 0); clear_extent_dirty(unpin, start, end, GFP_NOFS); set_extent_dirty(free_space_cache, start, end, GFP_NOFS); if (need_resched()) { @@ -1538,14 +1555,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, GFP_NOFS); - eb = btrfs_find_tree_block(extent_root, ins.objectid, + eb = btrfs_find_create_tree_block(extent_root, ins.objectid, ins.offset); - if (!btrfs_buffer_uptodate(eb, trans->transid)) { - mutex_unlock(&extent_root->fs_info->alloc_mutex); + if (!btrfs_buffer_uptodate(eb, trans->transid)) btrfs_read_buffer(eb, trans->transid); - mutex_lock(&extent_root->fs_info->alloc_mutex); - } btrfs_tree_lock(eb); level = btrfs_header_level(eb); @@ -1585,13 +1599,20 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, struct extent_buffer *buf; buf = btrfs_find_tree_block(root, bytenr, num_bytes); if (buf) { + /* we can reuse a block if it hasn't been written + * and it is from this transaction. We can't + * reuse anything from the tree log root because + * it has tiny sub-transactions. + */ if (btrfs_buffer_uptodate(buf, 0) && btrfs_try_tree_lock(buf)) { u64 transid = root->fs_info->running_transaction->transid; u64 header_transid = btrfs_header_generation(buf); - if (header_transid == transid && + if (btrfs_header_owner(buf) != + BTRFS_TREE_LOG_OBJECTID && + header_transid == transid && !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { clean_tree_block(NULL, root, buf); @@ -1603,7 +1624,7 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, } free_extent_buffer(buf); } - update_pinned_extents(root, bytenr, num_bytes, 1); + btrfs_update_pinned_extents(root, bytenr, num_bytes, 1); } else { set_extent_bits(&root->fs_info->pending_del, bytenr, bytenr + num_bytes - 1, @@ -1801,7 +1822,7 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct GFP_NOFS); if (!test_range_bit(&extent_root->fs_info->extent_ins, start, end, EXTENT_LOCKED, 0)) { - update_pinned_extents(extent_root, start, + btrfs_update_pinned_extents(extent_root, start, end + 1 - start, 1); ret = __free_extent(trans, extent_root, start, end + 1 - start, @@ -1919,6 +1940,12 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { last_ptr = &root->fs_info->last_data_alloc; } + if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) { + last_ptr = &root->fs_info->last_log_alloc; + if (!last_ptr == 0 && root->fs_info->last_alloc) { + *last_ptr = root->fs_info->last_alloc + empty_cluster; + } + } if (last_ptr) { if (*last_ptr) @@ -2268,6 +2295,35 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, maybe_unlock_mutex(root); return ret; } + +/* + * this is used by the tree logging recovery code. It records that + * an extent has been allocated and makes sure to clear the free + * space cache bits as well + */ +int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset, + struct btrfs_key *ins) +{ + int ret; + struct btrfs_block_group_cache *block_group; + + maybe_lock_mutex(root); + block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid); + cache_block_group(root, block_group); + + clear_extent_dirty(&root->fs_info->free_space_cache, + ins->objectid, ins->objectid + ins->offset - 1, + GFP_NOFS); + ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, + ref_generation, owner, + owner_offset, ins); + maybe_unlock_mutex(root); + return ret; +} + /* * finds a free extent and does all the dirty work required for allocation * returns the key for the extent through ins, and a tree buffer for @@ -2350,9 +2406,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, return buf; } -static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct extent_buffer *leaf) +int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *leaf) { u64 leaf_owner; u64 leaf_generation; @@ -2402,9 +2457,9 @@ static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans, return 0; } -static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_leaf_ref *ref) +static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_leaf_ref *ref) { int i; int ret; @@ -2512,7 +2567,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, btrfs_header_nritems(cur)) break; if (*level == 0) { - ret = drop_leaf_ref_no_cache(trans, root, cur); + ret = btrfs_drop_leaf_ref(trans, root, cur); BUG_ON(ret); break; } @@ -2552,7 +2607,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, btrfs_node_key_to_cpu(cur, &key, path->slots[*level]); ref = btrfs_lookup_leaf_ref(root, bytenr); if (ref) { - ret = drop_leaf_ref(trans, root, ref); + ret = cache_drop_leaf_ref(trans, root, ref); BUG_ON(ret); btrfs_remove_leaf_ref(root, ref); btrfs_free_leaf_ref(root, ref); @@ -3628,6 +3683,8 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, extent_root = root->fs_info->extent_root; block_group_cache = &root->fs_info->block_group_cache; + root->fs_info->last_trans_new_blockgroup = trans->transid; + cache = kzalloc(sizeof(*cache), GFP_NOFS); BUG_ON(!cache); cache->key.objectid = chunk_offset; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index e9e86fbaa24..84ecf3ab851 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -36,6 +36,8 @@ #include "btrfs_inode.h" #include "ioctl.h" #include "print-tree.h" +#include "tree-log.h" +#include "locking.h" #include "compat.h" @@ -988,10 +990,27 @@ out_nolock: *ppos = pos; if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) { - err = sync_page_range(inode, inode->i_mapping, - start_pos, num_written); + struct btrfs_trans_handle *trans; + + err = btrfs_fdatawrite_range(inode->i_mapping, start_pos, + start_pos + num_written -1, + WB_SYNC_NONE); + if (err < 0) + num_written = err; + + err = btrfs_wait_on_page_writeback_range(inode->i_mapping, + start_pos, start_pos + num_written - 1); if (err < 0) num_written = err; + + trans = btrfs_start_transaction(root, 1); + ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); + if (ret == 0) { + btrfs_sync_log(trans, root); + btrfs_end_transaction(trans, root); + } else { + btrfs_commit_transaction(trans, root); + } } else if (num_written > 0 && (file->f_flags & O_DIRECT)) { #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) do_sync_file_range(file, start_pos, @@ -1019,8 +1038,7 @@ int btrfs_release_file(struct inode * inode, struct file * filp) return 0; } -static int btrfs_sync_file(struct file *file, - struct dentry *dentry, int datasync) +int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) { struct inode *inode = dentry->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1043,6 +1061,8 @@ static int btrfs_sync_file(struct file *file, } mutex_unlock(&root->fs_info->trans_mutex); + filemap_fdatawait(inode->i_mapping); + /* * ok we haven't committed the transaction yet, lets do a commit */ @@ -1054,7 +1074,16 @@ static int btrfs_sync_file(struct file *file, ret = -ENOMEM; goto out; } - ret = btrfs_commit_transaction(trans, root); + + ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); + if (ret < 0) + goto out; + if (ret > 0) { + ret = btrfs_commit_transaction(trans, root); + } else { + btrfs_sync_log(trans, root); + ret = btrfs_end_transaction(trans, root); + } out: return ret > 0 ? EIO : ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 43d3f2649ca..65df9d83023 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -46,6 +46,8 @@ #include "volumes.h" #include "ordered-data.h" #include "xattr.h" +#include "compat.h" +#include "tree-log.h" struct btrfs_iget_args { u64 ino; @@ -586,6 +588,7 @@ nocow: &ordered_extent->list); btrfs_ordered_update_i_size(inode, ordered_extent); + btrfs_update_inode(trans, root, inode); btrfs_remove_ordered_extent(inode, ordered_extent); /* once for us */ @@ -593,7 +596,6 @@ nocow: /* once for the tree */ btrfs_put_ordered_extent(ordered_extent); - btrfs_update_inode(trans, root, inode); btrfs_end_transaction(trans, root); return 0; } @@ -1007,7 +1009,8 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item); - inode->i_generation = btrfs_inode_generation(leaf, inode_item); + BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item); + inode->i_generation = BTRFS_I(inode)->generation; inode->i_rdev = 0; rdev = btrfs_inode_rdev(leaf, inode_item); @@ -1056,7 +1059,8 @@ make_bad: make_bad_inode(inode); } -static void fill_inode_item(struct extent_buffer *leaf, +static void fill_inode_item(struct btrfs_trans_handle *trans, + struct extent_buffer *leaf, struct btrfs_inode_item *item, struct inode *inode) { @@ -1082,7 +1086,8 @@ static void fill_inode_item(struct extent_buffer *leaf, inode->i_ctime.tv_nsec); btrfs_set_inode_nblocks(leaf, item, inode->i_blocks); - btrfs_set_inode_generation(leaf, item, inode->i_generation); + btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation); + btrfs_set_inode_transid(leaf, item, trans->transid); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); btrfs_set_inode_block_group(leaf, item, @@ -1112,7 +1117,7 @@ int noinline btrfs_update_inode(struct btrfs_trans_handle *trans, inode_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item); - fill_inode_item(leaf, inode_item, inode); + fill_inode_item(trans, leaf, inode_item, inode); btrfs_mark_buffer_dirty(leaf); btrfs_set_inode_last_trans(trans, inode); ret = 0; @@ -1122,14 +1127,12 @@ failed: } -static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *dir, - struct dentry *dentry) +int btrfs_unlink_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *dir, struct inode *inode, + const char *name, int name_len) { struct btrfs_path *path; - const char *name = dentry->d_name.name; - int name_len = dentry->d_name.len; int ret = 0; struct extent_buffer *leaf; struct btrfs_dir_item *di; @@ -1160,13 +1163,12 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, btrfs_release_path(root, path); ret = btrfs_del_inode_ref(trans, root, name, name_len, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino, &index); + inode->i_ino, + dir->i_ino, &index); if (ret) { printk("failed to delete reference to %.*s, " "inode %lu parent %lu\n", name_len, name, - dentry->d_inode->i_ino, - dentry->d_parent->d_inode->i_ino); + inode->i_ino, dir->i_ino); goto err; } @@ -1183,21 +1185,25 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans, ret = btrfs_delete_one_dir_name(trans, root, path, di); btrfs_release_path(root, path); - dentry->d_inode->i_ctime = dir->i_ctime; + ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len, + inode, dir->i_ino); + BUG_ON(ret); + + ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len, + dir, index); + BUG_ON(ret); err: btrfs_free_path(path); - if (!ret) { - btrfs_i_size_write(dir, dir->i_size - name_len * 2); - dir->i_mtime = dir->i_ctime = CURRENT_TIME; - btrfs_update_inode(trans, root, dir); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - dentry->d_inode->i_nlink--; -#else - drop_nlink(dentry->d_inode); -#endif - ret = btrfs_update_inode(trans, root, dentry->d_inode); - dir->i_sb->s_dirt = 1; - } + if (ret) + goto out; + + btrfs_i_size_write(dir, dir->i_size - name_len * 2); + inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME; + btrfs_update_inode(trans, root, dir); + btrfs_drop_nlink(inode); + ret = btrfs_update_inode(trans, root, inode); + dir->i_sb->s_dirt = 1; +out: return ret; } @@ -1218,7 +1224,8 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, dir); - ret = btrfs_unlink_trans(trans, root, dir, dentry); + ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, + dentry->d_name.name, dentry->d_name.len); if (inode->i_nlink == 0) ret = btrfs_orphan_add(trans, inode); @@ -1256,7 +1263,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) goto fail_trans; /* now the directory is empty */ - err = btrfs_unlink_trans(trans, root, dir, dentry); + err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, + dentry->d_name.name, dentry->d_name.len); if (!err) { btrfs_i_size_write(inode, 0); } @@ -1283,10 +1291,10 @@ fail: * min_type is the minimum key type to truncate down to. If set to 0, this * will kill all the items on this inode, including the INODE_ITEM_KEY. */ -static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode, - u32 min_type) +noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode, + u64 new_size, u32 min_type) { int ret; struct btrfs_path *path; @@ -1307,7 +1315,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans, int extent_type = -1; u64 mask = root->sectorsize - 1; - btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1); + if (root->ref_cows) + btrfs_drop_extent_cache(inode, + new_size & (~mask), (u64)-1); path = btrfs_alloc_path(); path->reada = -1; BUG_ON(!path); @@ -1324,7 +1334,13 @@ search_again: goto error; } if (ret > 0) { - BUG_ON(path->slots[0] == 0); + /* there are no items in the tree for us to truncate, we're + * done + */ + if (path->slots[0] == 0) { + ret = 0; + goto error; + } path->slots[0]--; } @@ -1358,10 +1374,10 @@ search_again: } if (found_type == BTRFS_CSUM_ITEM_KEY) { ret = btrfs_csum_truncate(trans, root, path, - inode->i_size); + new_size); BUG_ON(ret); } - if (item_end < inode->i_size) { + if (item_end < new_size) { if (found_type == BTRFS_DIR_ITEM_KEY) { found_type = BTRFS_INODE_ITEM_KEY; } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { @@ -1378,7 +1394,7 @@ search_again: btrfs_set_key_type(&key, found_type); goto next; } - if (found_key.offset >= inode->i_size) + if (found_key.offset >= new_size) del_item = 1; else del_item = 0; @@ -1394,7 +1410,7 @@ search_again: if (!del_item) { u64 orig_num_bytes = btrfs_file_extent_num_bytes(leaf, fi); - extent_num_bytes = inode->i_size - + extent_num_bytes = new_size - found_key.offset + root->sectorsize - 1; extent_num_bytes = extent_num_bytes & ~((u64)root->sectorsize - 1); @@ -1402,7 +1418,7 @@ search_again: extent_num_bytes); num_dec = (orig_num_bytes - extent_num_bytes); - if (extent_start != 0) + if (root->ref_cows && extent_start != 0) dec_i_blocks(inode, num_dec); btrfs_mark_buffer_dirty(leaf); } else { @@ -1413,22 +1429,29 @@ search_again: num_dec = btrfs_file_extent_num_bytes(leaf, fi); if (extent_start != 0) { found_extent = 1; - dec_i_blocks(inode, num_dec); + if (root->ref_cows) + dec_i_blocks(inode, num_dec); + } + if (root->ref_cows) { + root_gen = + btrfs_header_generation(leaf); } - root_gen = btrfs_header_generation(leaf); root_owner = btrfs_header_owner(leaf); } } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { if (!del_item) { - u32 newsize = inode->i_size - found_key.offset; - dec_i_blocks(inode, item_end + 1 - - found_key.offset - newsize); - newsize = - btrfs_file_extent_calc_inline_size(newsize); + u32 size = new_size - found_key.offset; + + if (root->ref_cows) { + dec_i_blocks(inode, item_end + 1 - + found_key.offset - size); + } + size = + btrfs_file_extent_calc_inline_size(size); ret = btrfs_truncate_item(trans, root, path, - newsize, 1); + size, 1); BUG_ON(ret); - } else { + } else if (root->ref_cows) { dec_i_blocks(inode, item_end + 1 - found_key.offset); } @@ -1666,7 +1689,7 @@ void btrfs_delete_inode(struct inode *inode) trans = btrfs_start_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); - ret = btrfs_truncate_in_trans(trans, root, inode, 0); + ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, 0); if (ret) { btrfs_orphan_del(NULL, inode); goto no_delete_lock; @@ -1753,15 +1776,20 @@ static int fixup_tree_root_location(struct btrfs_root *root, return 0; } -static int btrfs_init_locked_inode(struct inode *inode, void *p) +static noinline void init_btrfs_i(struct inode *inode) { - struct btrfs_iget_args *args = p; - inode->i_ino = args->ino; - BTRFS_I(inode)->root = args->root; - BTRFS_I(inode)->delalloc_bytes = 0; - inode->i_mapping->writeback_index = 0; - BTRFS_I(inode)->disk_i_size = 0; - BTRFS_I(inode)->index_cnt = (u64)-1; + struct btrfs_inode *bi = BTRFS_I(inode); + + bi->i_acl = NULL; + bi->i_default_acl = NULL; + + bi->generation = 0; + bi->last_trans = 0; + bi->logged_trans = 0; + bi->delalloc_bytes = 0; + bi->disk_i_size = 0; + bi->flags = 0; + bi->index_cnt = (u64)-1; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); @@ -1771,6 +1799,15 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); + mutex_init(&BTRFS_I(inode)->log_mutex); +} + +static int btrfs_init_locked_inode(struct inode *inode, void *p) +{ + struct btrfs_iget_args *args = p; + inode->i_ino = args->ino; + init_btrfs_i(inode); + BTRFS_I(inode)->root = args->root; return 0; } @@ -2263,21 +2300,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, * btrfs_get_inode_index_count has an explanation for the magic * number */ + init_btrfs_i(inode); BTRFS_I(inode)->index_cnt = 2; - - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_tree, - inode->i_mapping, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, - inode->i_mapping, GFP_NOFS); - btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); - INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); - mutex_init(&BTRFS_I(inode)->csum_mutex); - mutex_init(&BTRFS_I(inode)->extent_mutex); - BTRFS_I(inode)->delalloc_bytes = 0; - inode->i_mapping->writeback_index = 0; - BTRFS_I(inode)->disk_i_size = 0; BTRFS_I(inode)->root = root; + BTRFS_I(inode)->generation = trans->transid; if (mode & S_IFDIR) owner = 0; @@ -2290,7 +2316,6 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, new_inode_group = group; } BTRFS_I(inode)->block_group = new_inode_group; - BTRFS_I(inode)->flags = 0; key[0].objectid = objectid; btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); @@ -2318,7 +2343,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_inode_item); - fill_inode_item(path->nodes[0], inode_item, inode); + fill_inode_item(trans, path->nodes[0], inode_item, inode); ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, struct btrfs_inode_ref); @@ -2349,38 +2374,34 @@ static inline u8 btrfs_inode_type(struct inode *inode) return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT]; } -static int btrfs_add_link(struct btrfs_trans_handle *trans, - struct dentry *dentry, struct inode *inode, - int add_backref, u64 index) +int btrfs_add_link(struct btrfs_trans_handle *trans, + struct inode *parent_inode, struct inode *inode, + const char *name, int name_len, int add_backref, u64 index) { int ret; struct btrfs_key key; - struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root; - struct inode *parent_inode = dentry->d_parent->d_inode; + struct btrfs_root *root = BTRFS_I(parent_inode)->root; key.objectid = inode->i_ino; btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); key.offset = 0; - ret = btrfs_insert_dir_item(trans, root, - dentry->d_name.name, dentry->d_name.len, - dentry->d_parent->d_inode->i_ino, + ret = btrfs_insert_dir_item(trans, root, name, name_len, + parent_inode->i_ino, &key, btrfs_inode_type(inode), index); if (ret == 0) { if (add_backref) { ret = btrfs_insert_inode_ref(trans, root, - dentry->d_name.name, - dentry->d_name.len, - inode->i_ino, - parent_inode->i_ino, - index); + name, name_len, + inode->i_ino, + parent_inode->i_ino, + index); } btrfs_i_size_write(parent_inode, parent_inode->i_size + - dentry->d_name.len * 2); + name_len * 2); parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; - ret = btrfs_update_inode(trans, root, - dentry->d_parent->d_inode); + ret = btrfs_update_inode(trans, root, parent_inode); } return ret; } @@ -2389,7 +2410,9 @@ static int btrfs_add_nondir(struct btrfs_trans_handle *trans, struct dentry *dentry, struct inode *inode, int backref, u64 index) { - int err = btrfs_add_link(trans, dentry, inode, backref, index); + int err = btrfs_add_link(trans, dentry->d_parent->d_inode, + inode, dentry->d_name.name, + dentry->d_name.len, backref, index); if (!err) { d_instantiate(dentry, inode); return 0; @@ -2513,19 +2536,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_mapping->backing_dev_info = &root->fs_info->bdi; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_tree, - inode->i_mapping, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, - inode->i_mapping, GFP_NOFS); - INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); - mutex_init(&BTRFS_I(inode)->csum_mutex); - mutex_init(&BTRFS_I(inode)->extent_mutex); - BTRFS_I(inode)->delalloc_bytes = 0; - BTRFS_I(inode)->disk_i_size = 0; - inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; - btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -2556,11 +2567,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, if (inode->i_nlink == 0) return -ENOENT; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - inode->i_nlink++; -#else - inc_nlink(inode); -#endif + btrfs_inc_nlink(inode); err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; @@ -2650,7 +2657,9 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) if (err) goto out_fail; - err = btrfs_add_link(trans, dentry, inode, 0, index); + err = btrfs_add_link(trans, dentry->d_parent->d_inode, + inode, dentry->d_name.name, + dentry->d_name.len, 0, index); if (err) goto out_fail; @@ -3221,7 +3230,7 @@ static void btrfs_truncate(struct inode *inode) if (ret) goto out; /* FIXME, add redo link to tree so we don't leak on crash */ - ret = btrfs_truncate_in_trans(trans, root, inode, + ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, BTRFS_EXTENT_DATA_KEY); btrfs_update_inode(trans, root, inode); @@ -3304,6 +3313,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) if (!ei) return NULL; ei->last_trans = 0; + ei->logged_trans = 0; btrfs_ordered_inode_tree_init(&ei->ordered_tree); ei->i_acl = BTRFS_ACL_NOT_CACHED; ei->i_default_acl = BTRFS_ACL_NOT_CACHED; @@ -3463,31 +3473,39 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, btrfs_set_trans_block_group(trans, new_dir); - old_dentry->d_inode->i_nlink++; + btrfs_inc_nlink(old_dentry->d_inode); old_dir->i_ctime = old_dir->i_mtime = ctime; new_dir->i_ctime = new_dir->i_mtime = ctime; old_inode->i_ctime = ctime; - ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry); + ret = btrfs_unlink_inode(trans, root, old_dir, old_dentry->d_inode, + old_dentry->d_name.name, + old_dentry->d_name.len); if (ret) goto out_fail; if (new_inode) { new_inode->i_ctime = CURRENT_TIME; - ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry); + ret = btrfs_unlink_inode(trans, root, new_dir, + new_dentry->d_inode, + new_dentry->d_name.name, + new_dentry->d_name.len); if (ret) goto out_fail; if (new_inode->i_nlink == 0) { - ret = btrfs_orphan_add(trans, new_inode); + ret = btrfs_orphan_add(trans, new_dentry->d_inode); if (ret) goto out_fail; } + } ret = btrfs_set_inode_index(new_dir, old_inode, &index); if (ret) goto out_fail; - ret = btrfs_add_link(trans, new_dentry, old_inode, 1, index); + ret = btrfs_add_link(trans, new_dentry->d_parent->d_inode, + old_inode, new_dentry->d_name.name, + new_dentry->d_name.len, 1, index); if (ret) goto out_fail; @@ -3577,19 +3595,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_mapping->backing_dev_info = &root->fs_info->bdi; inode->i_fop = &btrfs_file_operations; inode->i_op = &btrfs_file_inode_operations; - extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_tree, - inode->i_mapping, GFP_NOFS); - extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, - inode->i_mapping, GFP_NOFS); - INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); - mutex_init(&BTRFS_I(inode)->csum_mutex); - mutex_init(&BTRFS_I(inode)->extent_mutex); - BTRFS_I(inode)->delalloc_bytes = 0; - BTRFS_I(inode)->disk_i_size = 0; - inode->i_mapping->writeback_index = 0; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; - btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); } dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); @@ -3691,6 +3697,7 @@ static struct file_operations btrfs_dir_file_operations = { .compat_ioctl = btrfs_ioctl, #endif .release = btrfs_release_file, + .fsync = btrfs_sync_file, }; static struct extent_io_ops btrfs_extent_io_ops = { diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 36726696e58..e3984f902e7 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -202,8 +202,9 @@ again: memcpy(&found_key, &key, sizeof(key)); key.offset++; btrfs_release_path(root, path); - dead_root = btrfs_read_fs_root_no_radix(root->fs_info, - &found_key); + dead_root = + btrfs_read_fs_root_no_radix(root->fs_info->tree_root, + &found_key); if (IS_ERR(dead_root)) { ret = PTR_ERR(dead_root); goto err; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index eff3ad72991..49c4f5b40ed 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -25,6 +25,7 @@ #include "transaction.h" #include "locking.h" #include "ref-cache.h" +#include "tree-log.h" static int total_trans = 0; extern struct kmem_cache *btrfs_trans_handle_cachep; @@ -57,6 +58,7 @@ static noinline int join_transaction(struct btrfs_root *root) root->fs_info->generation++; root->fs_info->last_alloc = 0; root->fs_info->last_data_alloc = 0; + root->fs_info->last_log_alloc = 0; cur_trans->num_writers = 1; cur_trans->num_joined = 0; cur_trans->transid = root->fs_info->generation; @@ -83,7 +85,7 @@ static noinline int join_transaction(struct btrfs_root *root) return 0; } -static noinline int record_root_in_trans(struct btrfs_root *root) +noinline int btrfs_record_root_in_trans(struct btrfs_root *root) { struct btrfs_dirty_root *dirty; u64 running_trans_id = root->fs_info->running_transaction->transid; @@ -151,7 +153,7 @@ static void wait_current_trans(struct btrfs_root *root) } } -struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, +static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, int num_blocks, int wait) { struct btrfs_trans_handle *h = @@ -164,7 +166,7 @@ struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, ret = join_transaction(root); BUG_ON(ret); - record_root_in_trans(root); + btrfs_record_root_in_trans(root); h->transid = root->fs_info->running_transaction->transid; h->transaction = root->fs_info->running_transaction; h->blocks_reserved = num_blocks; @@ -456,6 +458,8 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, BUG_ON(!root->ref_tree); dirty = root->dirty_root; + btrfs_free_log(trans, root); + if (root->commit_root == root->node) { WARN_ON(root->node->start != btrfs_root_bytenr(&root->root_item)); @@ -600,7 +604,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, num_bytes -= btrfs_root_used(&dirty->root->root_item); bytes_used = btrfs_root_used(&root->root_item); if (num_bytes) { - record_root_in_trans(root); + btrfs_record_root_in_trans(root); btrfs_set_root_used(&root->root_item, bytes_used - num_bytes); } @@ -745,7 +749,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, int ret; INIT_LIST_HEAD(&dirty_fs_roots); - mutex_lock(&root->fs_info->trans_mutex); if (trans->transaction->in_commit) { cur_trans = trans->transaction; @@ -821,10 +824,30 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, WARN_ON(cur_trans != trans->transaction); + /* btrfs_commit_tree_roots is responsible for getting the + * various roots consistent with each other. Every pointer + * in the tree of tree roots has to point to the most up to date + * root for every subvolume and other tree. So, we have to keep + * the tree logging code from jumping in and changing any + * of the trees. + * + * At this point in the commit, there can't be any tree-log + * writers, but a little lower down we drop the trans mutex + * and let new people in. By holding the tree_log_mutex + * from now until after the super is written, we avoid races + * with the tree-log code. + */ + mutex_lock(&root->fs_info->tree_log_mutex); + ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, &dirty_fs_roots); BUG_ON(ret); + /* add_dirty_roots gets rid of all the tree log roots, it is now + * safe to free the root of tree log roots + */ + btrfs_free_log_root_tree(trans, root->fs_info); + ret = btrfs_commit_tree_roots(trans, root); BUG_ON(ret); @@ -843,6 +866,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, chunk_root->node->start); btrfs_set_super_chunk_root_level(&root->fs_info->super_copy, btrfs_header_level(chunk_root->node)); + + if (!root->fs_info->log_root_recovering) { + btrfs_set_super_log_root(&root->fs_info->super_copy, 0); + btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0); + } + memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy, sizeof(root->fs_info->super_copy)); @@ -857,6 +886,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, BUG_ON(ret); write_ctree_super(trans, root); + /* + * the super is written, we can safely allow the tree-loggers + * to go about their business + */ + mutex_unlock(&root->fs_info->tree_log_mutex); + btrfs_finish_extent_commit(trans, root, pinned_copy); mutex_lock(&root->fs_info->trans_mutex); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 598baa31241..cc63650d60d 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -98,4 +98,5 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, struct btrfs_root *root); void btrfs_throttle(struct btrfs_root *root); +int btrfs_record_root_in_trans(struct btrfs_root *root); #endif diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index cc2650b0695..b3bb5bbad76 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -81,12 +81,12 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, memcpy(&key, &root->defrag_progress, sizeof(key)); } - path->lowest_level = 1; path->keep_locks = 1; if (cache_only) min_trans = root->defrag_trans_start; - ret = btrfs_search_forward(root, &key, path, cache_only, min_trans); + ret = btrfs_search_forward(root, &key, NULL, path, + cache_only, min_trans); if (ret < 0) goto out; if (ret > 0) { diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c new file mode 100644 index 00000000000..d1ce8314b94 --- /dev/null +++ b/fs/btrfs/tree-log.c @@ -0,0 +1,2804 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include "ctree.h" +#include "transaction.h" +#include "disk-io.h" +#include "locking.h" +#include "print-tree.h" +#include "compat.h" + +/* magic values for the inode_only field in btrfs_log_inode: + * + * LOG_INODE_ALL means to log everything + * LOG_INODE_EXISTS means to log just enough to recreate the inode + * during log replay + */ +#define LOG_INODE_ALL 0 +#define LOG_INODE_EXISTS 1 + +/* + * stages for the tree walking. The first + * stage (0) is to only pin down the blocks we find + * the second stage (1) is to make sure that all the inodes + * we find in the log are created in the subvolume. + * + * The last stage is to deal with directories and links and extents + * and all the other fun semantics + */ +#define LOG_WALK_PIN_ONLY 0 +#define LOG_WALK_REPLAY_INODES 1 +#define LOG_WALK_REPLAY_ALL 2 + +static int __btrfs_log_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + int inode_only); + +/* + * tree logging is a special write ahead log used to make sure that + * fsyncs and O_SYNCs can happen without doing full tree commits. + * + * Full tree commits are expensive because they require commonly + * modified blocks to be recowed, creating many dirty pages in the + * extent tree an 4x-6x higher write load than ext3. + * + * Instead of doing a tree commit on every fsync, we use the + * key ranges and transaction ids to find items for a given file or directory + * that have changed in this transaction. Those items are copied into + * a special tree (one per subvolume root), that tree is written to disk + * and then the fsync is considered complete. + * + * After a crash, items are copied out of the log-tree back into the + * subvolume tree. Any file data extents found are recorded in the extent + * allocation tree, and the log-tree freed. + * + * The log tree is read three times, once to pin down all the extents it is + * using in ram and once, once to create all the inodes logged in the tree + * and once to do all the other items. + */ + +/* + * btrfs_add_log_tree adds a new per-subvolume log tree into the + * tree of log tree roots. This must be called with a tree log transaction + * running (see start_log_trans). + */ +int btrfs_add_log_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + struct btrfs_key key; + struct btrfs_root_item root_item; + struct btrfs_inode_item *inode_item; + struct extent_buffer *leaf; + struct btrfs_root *new_root = root; + int ret; + u64 objectid = root->root_key.objectid; + + leaf = btrfs_alloc_free_block(trans, root, root->leafsize, + BTRFS_TREE_LOG_OBJECTID, + 0, 0, 0, 0, 0); + if (IS_ERR(leaf)) { + ret = PTR_ERR(leaf); + return ret; + } + + btrfs_set_header_nritems(leaf, 0); + btrfs_set_header_level(leaf, 0); + btrfs_set_header_bytenr(leaf, leaf->start); + btrfs_set_header_generation(leaf, trans->transid); + btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); + + write_extent_buffer(leaf, root->fs_info->fsid, + (unsigned long)btrfs_header_fsid(leaf), + BTRFS_FSID_SIZE); + btrfs_mark_buffer_dirty(leaf); + + inode_item = &root_item.inode; + memset(inode_item, 0, sizeof(*inode_item)); + inode_item->generation = cpu_to_le64(1); + inode_item->size = cpu_to_le64(3); + inode_item->nlink = cpu_to_le32(1); + inode_item->nblocks = cpu_to_le64(1); + inode_item->mode = cpu_to_le32(S_IFDIR | 0755); + + btrfs_set_root_bytenr(&root_item, leaf->start); + btrfs_set_root_level(&root_item, 0); + btrfs_set_root_refs(&root_item, 0); + btrfs_set_root_used(&root_item, 0); + + memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); + root_item.drop_level = 0; + + btrfs_tree_unlock(leaf); + free_extent_buffer(leaf); + leaf = NULL; + + btrfs_set_root_dirid(&root_item, 0); + + key.objectid = BTRFS_TREE_LOG_OBJECTID; + key.offset = objectid; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + ret = btrfs_insert_root(trans, root->fs_info->log_root_tree, &key, + &root_item); + if (ret) + goto fail; + + new_root = btrfs_read_fs_root_no_radix(root->fs_info->log_root_tree, + &key); + BUG_ON(!new_root); + + WARN_ON(root->log_root); + root->log_root = new_root; + + /* + * log trees do not get reference counted because they go away + * before a real commit is actually done. They do store pointers + * to file data extents, and those reference counts still get + * updated (along with back refs to the log tree). + */ + new_root->ref_cows = 0; + new_root->last_trans = trans->transid; +fail: + return ret; +} + +/* + * start a sub transaction and setup the log tree + * this increments the log tree writer count to make the people + * syncing the tree wait for us to finish + */ +static int start_log_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + int ret; + mutex_lock(&root->fs_info->tree_log_mutex); + if (!root->fs_info->log_root_tree) { + ret = btrfs_init_log_root_tree(trans, root->fs_info); + BUG_ON(ret); + } + if (!root->log_root) { + ret = btrfs_add_log_tree(trans, root); + BUG_ON(ret); + } + atomic_inc(&root->fs_info->tree_log_writers); + root->fs_info->tree_log_batch++; + mutex_unlock(&root->fs_info->tree_log_mutex); + return 0; +} + +/* + * returns 0 if there was a log transaction running and we were able + * to join, or returns -ENOENT if there were not transactions + * in progress + */ +static int join_running_log_trans(struct btrfs_root *root) +{ + int ret = -ENOENT; + + smp_mb(); + if (!root->log_root) + return -ENOENT; + + mutex_lock(&root->fs_info->tree_log_mutex); + if (root->log_root) { + ret = 0; + atomic_inc(&root->fs_info->tree_log_writers); + root->fs_info->tree_log_batch++; + } + mutex_unlock(&root->fs_info->tree_log_mutex); + return ret; +} + +/* + * indicate we're done making changes to the log tree + * and wake up anyone waiting to do a sync + */ +static int end_log_trans(struct btrfs_root *root) +{ + atomic_dec(&root->fs_info->tree_log_writers); + smp_mb(); + if (waitqueue_active(&root->fs_info->tree_log_wait)) + wake_up(&root->fs_info->tree_log_wait); + return 0; +} + + +/* + * the walk control struct is used to pass state down the chain when + * processing the log tree. The stage field tells us which part + * of the log tree processing we are currently doing. The others + * are state fields used for that specific part + */ +struct walk_control { + /* should we free the extent on disk when done? This is used + * at transaction commit time while freeing a log tree + */ + int free; + + /* should we write out the extent buffer? This is used + * while flushing the log tree to disk during a sync + */ + int write; + + /* should we wait for the extent buffer io to finish? Also used + * while flushing the log tree to disk for a sync + */ + int wait; + + /* pin only walk, we record which extents on disk belong to the + * log trees + */ + int pin; + + /* what stage of the replay code we're currently in */ + int stage; + + /* the root we are currently replaying */ + struct btrfs_root *replay_dest; + + /* the trans handle for the current replay */ + struct btrfs_trans_handle *trans; + + /* the function that gets used to process blocks we find in the + * tree. Note the extent_buffer might not be up to date when it is + * passed in, and it must be checked or read if you need the data + * inside it + */ + int (*process_func)(struct btrfs_root *log, struct extent_buffer *eb, + struct walk_control *wc, u64 gen); +}; + +/* + * process_func used to pin down extents, write them or wait on them + */ +static int process_one_buffer(struct btrfs_root *log, + struct extent_buffer *eb, + struct walk_control *wc, u64 gen) +{ + if (wc->pin) { + mutex_lock(&log->fs_info->alloc_mutex); + btrfs_update_pinned_extents(log->fs_info->extent_root, + eb->start, eb->len, 1); + mutex_unlock(&log->fs_info->alloc_mutex); + } + + if (btrfs_buffer_uptodate(eb, gen)) { + if (wc->write) + btrfs_write_tree_block(eb); + if (wc->wait) + btrfs_wait_tree_block_writeback(eb); + } + return 0; +} + +/* + * Item overwrite used by replay and tree logging. eb, slot and key all refer + * to the src data we are copying out. + * + * root is the tree we are copying into, and path is a scratch + * path for use in this function (it should be released on entry and + * will be released on exit). + * + * If the key is already in the destination tree the existing item is + * overwritten. If the existing item isn't big enough, it is extended. + * If it is too large, it is truncated. + * + * If the key isn't in the destination yet, a new item is inserted. + */ +static noinline int overwrite_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *eb, int slot, + struct btrfs_key *key) +{ + int ret; + u32 item_size; + u64 saved_i_size = 0; + int save_old_i_size = 0; + unsigned long src_ptr; + unsigned long dst_ptr; + int overwrite_root = 0; + + if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) + overwrite_root = 1; + + item_size = btrfs_item_size_nr(eb, slot); + src_ptr = btrfs_item_ptr_offset(eb, slot); + + /* look for the key in the destination tree */ + ret = btrfs_search_slot(NULL, root, key, path, 0, 0); + if (ret == 0) { + char *src_copy; + char *dst_copy; + u32 dst_size = btrfs_item_size_nr(path->nodes[0], + path->slots[0]); + if (dst_size != item_size) + goto insert; + + if (item_size == 0) { + btrfs_release_path(root, path); + return 0; + } + dst_copy = kmalloc(item_size, GFP_NOFS); + src_copy = kmalloc(item_size, GFP_NOFS); + + read_extent_buffer(eb, src_copy, src_ptr, item_size); + + dst_ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]); + read_extent_buffer(path->nodes[0], dst_copy, dst_ptr, + item_size); + ret = memcmp(dst_copy, src_copy, item_size); + + kfree(dst_copy); + kfree(src_copy); + /* + * they have the same contents, just return, this saves + * us from cowing blocks in the destination tree and doing + * extra writes that may not have been done by a previous + * sync + */ + if (ret == 0) { + btrfs_release_path(root, path); + return 0; + } + + } +insert: + btrfs_release_path(root, path); + /* try to insert the key into the destination tree */ + ret = btrfs_insert_empty_item(trans, root, path, + key, item_size); + + /* make sure any existing item is the correct size */ + if (ret == -EEXIST) { + u32 found_size; + found_size = btrfs_item_size_nr(path->nodes[0], + path->slots[0]); + if (found_size > item_size) { + btrfs_truncate_item(trans, root, path, item_size, 1); + } else if (found_size < item_size) { + ret = btrfs_del_item(trans, root, + path); + BUG_ON(ret); + + btrfs_release_path(root, path); + ret = btrfs_insert_empty_item(trans, + root, path, key, item_size); + BUG_ON(ret); + } + } else if (ret) { + BUG(); + } + dst_ptr = btrfs_item_ptr_offset(path->nodes[0], + path->slots[0]); + + /* don't overwrite an existing inode if the generation number + * was logged as zero. This is done when the tree logging code + * is just logging an inode to make sure it exists after recovery. + * + * Also, don't overwrite i_size on directories during replay. + * log replay inserts and removes directory items based on the + * state of the tree found in the subvolume, and i_size is modified + * as it goes + */ + if (key->type == BTRFS_INODE_ITEM_KEY && ret == -EEXIST) { + struct btrfs_inode_item *src_item; + struct btrfs_inode_item *dst_item; + + src_item = (struct btrfs_inode_item *)src_ptr; + dst_item = (struct btrfs_inode_item *)dst_ptr; + + if (btrfs_inode_generation(eb, src_item) == 0) + goto no_copy; + + if (overwrite_root && + S_ISDIR(btrfs_inode_mode(eb, src_item)) && + S_ISDIR(btrfs_inode_mode(path->nodes[0], dst_item))) { + save_old_i_size = 1; + saved_i_size = btrfs_inode_size(path->nodes[0], + dst_item); + } + } + + copy_extent_buffer(path->nodes[0], eb, dst_ptr, + src_ptr, item_size); + + if (save_old_i_size) { + struct btrfs_inode_item *dst_item; + dst_item = (struct btrfs_inode_item *)dst_ptr; + btrfs_set_inode_size(path->nodes[0], dst_item, saved_i_size); + } + + /* make sure the generation is filled in */ + if (key->type == BTRFS_INODE_ITEM_KEY) { + struct btrfs_inode_item *dst_item; + dst_item = (struct btrfs_inode_item *)dst_ptr; + if (btrfs_inode_generation(path->nodes[0], dst_item) == 0) { + btrfs_set_inode_generation(path->nodes[0], dst_item, + trans->transid); + } + } +no_copy: + btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_release_path(root, path); + return 0; +} + +/* + * simple helper to read an inode off the disk from a given root + * This can only be called for subvolume roots and not for the log + */ +static noinline struct inode *read_one_inode(struct btrfs_root *root, + u64 objectid) +{ + struct inode *inode; + inode = btrfs_iget_locked(root->fs_info->sb, objectid, root); + if (inode->i_state & I_NEW) { + BTRFS_I(inode)->root = root; + BTRFS_I(inode)->location.objectid = objectid; + BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; + BTRFS_I(inode)->location.offset = 0; + btrfs_read_locked_inode(inode); + unlock_new_inode(inode); + + } + if (is_bad_inode(inode)) { + iput(inode); + inode = NULL; + } + return inode; +} + +/* replays a single extent in 'eb' at 'slot' with 'key' into the + * subvolume 'root'. path is released on entry and should be released + * on exit. + * + * extents in the log tree have not been allocated out of the extent + * tree yet. So, this completes the allocation, taking a reference + * as required if the extent already exists or creating a new extent + * if it isn't in the extent allocation tree yet. + * + * The extent is inserted into the file, dropping any existing extents + * from the file that overlap the new one. + */ +static noinline int replay_one_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *eb, int slot, + struct btrfs_key *key) +{ + int found_type; + u64 mask = root->sectorsize - 1; + u64 extent_end; + u64 alloc_hint; + u64 start = key->offset; + struct btrfs_file_extent_item *item; + struct inode *inode = NULL; + unsigned long size; + int ret = 0; + + item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); + found_type = btrfs_file_extent_type(eb, item); + + if (found_type == BTRFS_FILE_EXTENT_REG) + extent_end = start + btrfs_file_extent_num_bytes(eb, item); + else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + size = btrfs_file_extent_inline_len(eb, + btrfs_item_nr(eb, slot)); + extent_end = (start + size + mask) & ~mask; + } else { + ret = 0; + goto out; + } + + inode = read_one_inode(root, key->objectid); + if (!inode) { + ret = -EIO; + goto out; + } + + /* + * first check to see if we already have this extent in the + * file. This must be done before the btrfs_drop_extents run + * so we don't try to drop this extent. + */ + ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, + start, 0); + + if (ret == 0 && found_type == BTRFS_FILE_EXTENT_REG) { + struct btrfs_file_extent_item cmp1; + struct btrfs_file_extent_item cmp2; + struct btrfs_file_extent_item *existing; + struct extent_buffer *leaf; + + leaf = path->nodes[0]; + existing = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + + read_extent_buffer(eb, &cmp1, (unsigned long)item, + sizeof(cmp1)); + read_extent_buffer(leaf, &cmp2, (unsigned long)existing, + sizeof(cmp2)); + + /* + * we already have a pointer to this exact extent, + * we don't have to do anything + */ + if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) { + btrfs_release_path(root, path); + goto out; + } + } + btrfs_release_path(root, path); + + /* drop any overlapping extents */ + ret = btrfs_drop_extents(trans, root, inode, + start, extent_end, start, &alloc_hint); + BUG_ON(ret); + + BUG_ON(ret); + if (found_type == BTRFS_FILE_EXTENT_REG) { + struct btrfs_key ins; + + ins.objectid = btrfs_file_extent_disk_bytenr(eb, item); + ins.offset = btrfs_file_extent_disk_num_bytes(eb, item); + ins.type = BTRFS_EXTENT_ITEM_KEY; + + /* insert the extent pointer in the file */ + ret = overwrite_item(trans, root, path, eb, slot, key); + BUG_ON(ret); + + /* + * is this extent already allocated in the extent + * allocation tree? If so, just add a reference + */ + ret = btrfs_lookup_extent(root, path, ins.objectid, ins.offset); + btrfs_release_path(root, path); + if (ret == 0) { + ret = btrfs_inc_extent_ref(trans, root, + ins.objectid, ins.offset, + root->root_key.objectid, + trans->transid, key->objectid, start); + } else { + /* + * insert the extent pointer in the extent + * allocation tree + */ + ret = btrfs_alloc_logged_extent(trans, root, + root->root_key.objectid, + trans->transid, key->objectid, + start, &ins); + BUG_ON(ret); + } + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + /* inline extents are easy, we just overwrite them */ + ret = overwrite_item(trans, root, path, eb, slot, key); + BUG_ON(ret); + } + /* btrfs_drop_extents changes i_blocks, update it here */ + inode->i_blocks += (extent_end - start) >> 9; + btrfs_update_inode(trans, root, inode); +out: + if (inode) + iput(inode); + return ret; +} + +/* + * when cleaning up conflicts between the directory names in the + * subvolume, directory names in the log and directory names in the + * inode back references, we may have to unlink inodes from directories. + * + * This is a helper function to do the unlink of a specific directory + * item + */ +static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct inode *dir, + struct btrfs_dir_item *di) +{ + struct inode *inode; + char *name; + int name_len; + struct extent_buffer *leaf; + struct btrfs_key location; + int ret; + + leaf = path->nodes[0]; + + btrfs_dir_item_key_to_cpu(leaf, di, &location); + name_len = btrfs_dir_name_len(leaf, di); + name = kmalloc(name_len, GFP_NOFS); + read_extent_buffer(leaf, name, (unsigned long)(di + 1), name_len); + btrfs_release_path(root, path); + + inode = read_one_inode(root, location.objectid); + BUG_ON(!inode); + + btrfs_inc_nlink(inode); + ret = btrfs_unlink_inode(trans, root, dir, inode, name, name_len); + kfree(name); + + iput(inode); + return ret; +} + +/* + * helper function to see if a given name and sequence number found + * in an inode back reference are already in a directory and correctly + * point to this inode + */ +static noinline int inode_in_dir(struct btrfs_root *root, + struct btrfs_path *path, + u64 dirid, u64 objectid, u64 index, + const char *name, int name_len) +{ + struct btrfs_dir_item *di; + struct btrfs_key location; + int match = 0; + + di = btrfs_lookup_dir_index_item(NULL, root, path, dirid, + index, name, name_len, 0); + if (di && !IS_ERR(di)) { + btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); + if (location.objectid != objectid) + goto out; + } else + goto out; + btrfs_release_path(root, path); + + di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, name_len, 0); + if (di && !IS_ERR(di)) { + btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); + if (location.objectid != objectid) + goto out; + } else + goto out; + match = 1; +out: + btrfs_release_path(root, path); + return match; +} + +/* + * helper function to check a log tree for a named back reference in + * an inode. This is used to decide if a back reference that is + * found in the subvolume conflicts with what we find in the log. + * + * inode backreferences may have multiple refs in a single item, + * during replay we process one reference at a time, and we don't + * want to delete valid links to a file from the subvolume if that + * link is also in the log. + */ +static noinline int backref_in_log(struct btrfs_root *log, + struct btrfs_key *key, + char *name, int namelen) +{ + struct btrfs_path *path; + struct btrfs_inode_ref *ref; + unsigned long ptr; + unsigned long ptr_end; + unsigned long name_ptr; + int found_name_len; + int item_size; + int ret; + int match = 0; + + path = btrfs_alloc_path(); + ret = btrfs_search_slot(NULL, log, key, path, 0, 0); + if (ret != 0) + goto out; + + item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]); + ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]); + ptr_end = ptr + item_size; + while (ptr < ptr_end) { + ref = (struct btrfs_inode_ref *)ptr; + found_name_len = btrfs_inode_ref_name_len(path->nodes[0], ref); + if (found_name_len == namelen) { + name_ptr = (unsigned long)(ref + 1); + ret = memcmp_extent_buffer(path->nodes[0], name, + name_ptr, namelen); + if (ret == 0) { + match = 1; + goto out; + } + } + ptr = (unsigned long)(ref + 1) + found_name_len; + } +out: + btrfs_free_path(path); + return match; +} + + +/* + * replay one inode back reference item found in the log tree. + * eb, slot and key refer to the buffer and key found in the log tree. + * root is the destination we are replaying into, and path is for temp + * use by this function. (it should be released on return). + */ +static noinline int add_inode_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_root *log, + struct btrfs_path *path, + struct extent_buffer *eb, int slot, + struct btrfs_key *key) +{ + struct inode *dir; + int ret; + struct btrfs_key location; + struct btrfs_inode_ref *ref; + struct btrfs_dir_item *di; + struct inode *inode; + char *name; + int namelen; + unsigned long ref_ptr; + unsigned long ref_end; + + location.objectid = key->objectid; + location.type = BTRFS_INODE_ITEM_KEY; + location.offset = 0; + + /* + * it is possible that we didn't log all the parent directories + * for a given inode. If we don't find the dir, just don't + * copy the back ref in. The link count fixup code will take + * care of the rest + */ + dir = read_one_inode(root, key->offset); + if (!dir) + return -ENOENT; + + inode = read_one_inode(root, key->objectid); + BUG_ON(!dir); + + ref_ptr = btrfs_item_ptr_offset(eb, slot); + ref_end = ref_ptr + btrfs_item_size_nr(eb, slot); + +again: + ref = (struct btrfs_inode_ref *)ref_ptr; + + namelen = btrfs_inode_ref_name_len(eb, ref); + name = kmalloc(namelen, GFP_NOFS); + BUG_ON(!name); + + read_extent_buffer(eb, name, (unsigned long)(ref + 1), namelen); + + /* if we already have a perfect match, we're done */ + if (inode_in_dir(root, path, dir->i_ino, inode->i_ino, + btrfs_inode_ref_index(eb, ref), + name, namelen)) { + goto out; + } + + /* + * look for a conflicting back reference in the metadata. + * if we find one we have to unlink that name of the file + * before we add our new link. Later on, we overwrite any + * existing back reference, and we don't want to create + * dangling pointers in the directory. + */ +conflict_again: + ret = btrfs_search_slot(NULL, root, key, path, 0, 0); + if (ret == 0) { + char *victim_name; + int victim_name_len; + struct btrfs_inode_ref *victim_ref; + unsigned long ptr; + unsigned long ptr_end; + struct extent_buffer *leaf = path->nodes[0]; + + /* are we trying to overwrite a back ref for the root directory + * if so, just jump out, we're done + */ + if (key->objectid == key->offset) + goto out_nowrite; + + /* check all the names in this back reference to see + * if they are in the log. if so, we allow them to stay + * otherwise they must be unlinked as a conflict + */ + ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); + ptr_end = ptr + btrfs_item_size_nr(leaf, path->slots[0]); + while(ptr < ptr_end) { + victim_ref = (struct btrfs_inode_ref *)ptr; + victim_name_len = btrfs_inode_ref_name_len(leaf, + victim_ref); + victim_name = kmalloc(victim_name_len, GFP_NOFS); + BUG_ON(!victim_name); + + read_extent_buffer(leaf, victim_name, + (unsigned long)(victim_ref + 1), + victim_name_len); + + if (!backref_in_log(log, key, victim_name, + victim_name_len)) { + btrfs_inc_nlink(inode); + btrfs_release_path(root, path); + ret = btrfs_unlink_inode(trans, root, dir, + inode, victim_name, + victim_name_len); + kfree(victim_name); + btrfs_release_path(root, path); + goto conflict_again; + } + kfree(victim_name); + ptr = (unsigned long)(victim_ref + 1) + victim_name_len; + } + BUG_ON(ret); + } + btrfs_release_path(root, path); + + /* look for a conflicting sequence number */ + di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, + btrfs_inode_ref_index(eb, ref), + name, namelen, 0); + if (di && !IS_ERR(di)) { + ret = drop_one_dir_item(trans, root, path, dir, di); + BUG_ON(ret); + } + btrfs_release_path(root, path); + + + /* look for a conflicting name */ + di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino, + name, namelen, 0); + if (di && !IS_ERR(di)) { + ret = drop_one_dir_item(trans, root, path, dir, di); + BUG_ON(ret); + } + btrfs_release_path(root, path); + + /* insert our name */ + ret = btrfs_add_link(trans, dir, inode, name, namelen, 0, + btrfs_inode_ref_index(eb, ref)); + BUG_ON(ret); + + btrfs_update_inode(trans, root, inode); + +out: + ref_ptr = (unsigned long)(ref + 1) + namelen; + kfree(name); + if (ref_ptr < ref_end) + goto again; + + /* finally write the back reference in the inode */ + ret = overwrite_item(trans, root, path, eb, slot, key); + BUG_ON(ret); + +out_nowrite: + btrfs_release_path(root, path); + iput(dir); + iput(inode); + return 0; +} + +/* + * replay one csum item from the log tree into the subvolume 'root' + * eb, slot and key all refer to the log tree + * path is for temp use by this function and should be released on return + * + * This copies the checksums out of the log tree and inserts them into + * the subvolume. Any existing checksums for this range in the file + * are overwritten, and new items are added where required. + * + * We keep this simple by reusing the btrfs_ordered_sum code from + * the data=ordered mode. This basically means making a copy + * of all the checksums in ram, which we have to do anyway for kmap + * rules. + * + * The copy is then sent down to btrfs_csum_file_blocks, which + * does all the hard work of finding existing items in the file + * or adding new ones. + */ +static noinline int replay_one_csum(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *eb, int slot, + struct btrfs_key *key) +{ + int ret; + u32 item_size = btrfs_item_size_nr(eb, slot); + u64 cur_offset; + unsigned long file_bytes; + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; + struct inode *inode; + unsigned long ptr; + + file_bytes = (item_size / BTRFS_CRC32_SIZE) * root->sectorsize; + inode = read_one_inode(root, key->objectid); + if (!inode) { + return -EIO; + } + + sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS); + if (!sums) { + iput(inode); + return -ENOMEM; + } + + INIT_LIST_HEAD(&sums->list); + sums->len = file_bytes; + sums->file_offset = key->offset; + + /* + * copy all the sums into the ordered sum struct + */ + sector_sum = sums->sums; + cur_offset = key->offset; + ptr = btrfs_item_ptr_offset(eb, slot); + while(item_size > 0) { + sector_sum->offset = cur_offset; + read_extent_buffer(eb, §or_sum->sum, ptr, BTRFS_CRC32_SIZE); + sector_sum++; + item_size -= BTRFS_CRC32_SIZE; + ptr += BTRFS_CRC32_SIZE; + cur_offset += root->sectorsize; + } + + /* let btrfs_csum_file_blocks add them into the file */ + ret = btrfs_csum_file_blocks(trans, root, inode, sums); + BUG_ON(ret); + kfree(sums); + iput(inode); + + return 0; +} +/* + * There are a few corners where the link count of the file can't + * be properly maintained during replay. So, instead of adding + * lots of complexity to the log code, we just scan the backrefs + * for any file that has been through replay. + * + * The scan will update the link count on the inode to reflect the + * number of back refs found. If it goes down to zero, the iput + * will free the inode. + */ +static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode) +{ + struct btrfs_path *path; + int ret; + struct btrfs_key key; + u64 nlink = 0; + unsigned long ptr; + unsigned long ptr_end; + int name_len; + + key.objectid = inode->i_ino; + key.type = BTRFS_INODE_REF_KEY; + key.offset = (u64)-1; + + path = btrfs_alloc_path(); + + while(1) { + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + break; + if (ret > 0) { + if (path->slots[0] == 0) + break; + path->slots[0]--; + } + btrfs_item_key_to_cpu(path->nodes[0], &key, + path->slots[0]); + if (key.objectid != inode->i_ino || + key.type != BTRFS_INODE_REF_KEY) + break; + ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]); + ptr_end = ptr + btrfs_item_size_nr(path->nodes[0], + path->slots[0]); + while(ptr < ptr_end) { + struct btrfs_inode_ref *ref; + + ref = (struct btrfs_inode_ref *)ptr; + name_len = btrfs_inode_ref_name_len(path->nodes[0], + ref); + ptr = (unsigned long)(ref + 1) + name_len; + nlink++; + } + + if (key.offset == 0) + break; + key.offset--; + btrfs_release_path(root, path); + } + btrfs_free_path(path); + if (nlink != inode->i_nlink) { + inode->i_nlink = nlink; + btrfs_update_inode(trans, root, inode); + } + + return 0; +} + +static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path) +{ + int ret; + struct btrfs_key key; + struct inode *inode; + + key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID; + key.type = BTRFS_ORPHAN_ITEM_KEY; + key.offset = (u64)-1; + while(1) { + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) + break; + + if (ret == 1) { + if (path->slots[0] == 0) + break; + path->slots[0]--; + } + + btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); + if (key.objectid != BTRFS_TREE_LOG_FIXUP_OBJECTID || + key.type != BTRFS_ORPHAN_ITEM_KEY) + break; + + ret = btrfs_del_item(trans, root, path); + BUG_ON(ret); + + btrfs_release_path(root, path); + inode = read_one_inode(root, key.offset); + BUG_ON(!inode); + + ret = fixup_inode_link_count(trans, root, inode); + BUG_ON(ret); + + iput(inode); + + if (key.offset == 0) + break; + key.offset--; + } + btrfs_release_path(root, path); + return 0; +} + + +/* + * record a given inode in the fixup dir so we can check its link + * count when replay is done. The link count is incremented here + * so the inode won't go away until we check it + */ +static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + u64 objectid) +{ + struct btrfs_key key; + int ret = 0; + struct inode *inode; + + inode = read_one_inode(root, objectid); + BUG_ON(!inode); + + key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID; + btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); + key.offset = objectid; + + ret = btrfs_insert_empty_item(trans, root, path, &key, 0); + + btrfs_release_path(root, path); + if (ret == 0) { + btrfs_inc_nlink(inode); + btrfs_update_inode(trans, root, inode); + } else if (ret == -EEXIST) { + ret = 0; + } else { + BUG(); + } + iput(inode); + + return ret; +} + +/* + * when replaying the log for a directory, we only insert names + * for inodes that actually exist. This means an fsync on a directory + * does not implicitly fsync all the new files in it + */ +static noinline int insert_one_name(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + u64 dirid, u64 index, + char *name, int name_len, u8 type, + struct btrfs_key *location) +{ + struct inode *inode; + struct inode *dir; + int ret; + + inode = read_one_inode(root, location->objectid); + if (!inode) + return -ENOENT; + + dir = read_one_inode(root, dirid); + if (!dir) { + iput(inode); + return -EIO; + } + ret = btrfs_add_link(trans, dir, inode, name, name_len, 1, index); + + /* FIXME, put inode into FIXUP list */ + + iput(inode); + iput(dir); + return ret; +} + +/* + * take a single entry in a log directory item and replay it into + * the subvolume. + * + * if a conflicting item exists in the subdirectory already, + * the inode it points to is unlinked and put into the link count + * fix up tree. + * + * If a name from the log points to a file or directory that does + * not exist in the FS, it is skipped. fsyncs on directories + * do not force down inodes inside that directory, just changes to the + * names or unlinks in a directory. + */ +static noinline int replay_one_name(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *eb, + struct btrfs_dir_item *di, + struct btrfs_key *key) +{ + char *name; + int name_len; + struct btrfs_dir_item *dst_di; + struct btrfs_key found_key; + struct btrfs_key log_key; + struct inode *dir; + struct inode *inode; + u8 log_type; + int ret; + + dir = read_one_inode(root, key->objectid); + BUG_ON(!dir); + + name_len = btrfs_dir_name_len(eb, di); + name = kmalloc(name_len, GFP_NOFS); + log_type = btrfs_dir_type(eb, di); + read_extent_buffer(eb, name, (unsigned long)(di + 1), + name_len); + + btrfs_dir_item_key_to_cpu(eb, di, &log_key); + if (key->type == BTRFS_DIR_ITEM_KEY) { + dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid, + name, name_len, 1); + } + else if (key->type == BTRFS_DIR_INDEX_KEY) { + dst_di = btrfs_lookup_dir_index_item(trans, root, path, + key->objectid, + key->offset, name, + name_len, 1); + } else { + BUG(); + } + if (!dst_di || IS_ERR(dst_di)) { + /* we need a sequence number to insert, so we only + * do inserts for the BTRFS_DIR_INDEX_KEY types + */ + if (key->type != BTRFS_DIR_INDEX_KEY) + goto out; + goto insert; + } + + btrfs_dir_item_key_to_cpu(path->nodes[0], dst_di, &found_key); + /* the existing item matches the logged item */ + if (found_key.objectid == log_key.objectid && + found_key.type == log_key.type && + found_key.offset == log_key.offset && + btrfs_dir_type(path->nodes[0], dst_di) == log_type) { + goto out; + } + + /* + * don't drop the conflicting directory entry if the inode + * for the new entry doesn't exist + */ + inode = read_one_inode(root, log_key.objectid); + if (!inode) + goto out; + + iput(inode); + ret = drop_one_dir_item(trans, root, path, dir, dst_di); + BUG_ON(ret); + + if (key->type == BTRFS_DIR_INDEX_KEY) + goto insert; +out: + btrfs_release_path(root, path); + kfree(name); + iput(dir); + return 0; + +insert: + btrfs_release_path(root, path); + ret = insert_one_name(trans, root, path, key->objectid, key->offset, + name, name_len, log_type, &log_key); + + if (ret && ret != -ENOENT) + BUG(); + goto out; +} + +/* + * find all the names in a directory item and reconcile them into + * the subvolume. Only BTRFS_DIR_ITEM_KEY types will have more than + * one name in a directory item, but the same code gets used for + * both directory index types + */ +static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct extent_buffer *eb, int slot, + struct btrfs_key *key) +{ + int ret; + u32 item_size = btrfs_item_size_nr(eb, slot); + struct btrfs_dir_item *di; + int name_len; + unsigned long ptr; + unsigned long ptr_end; + + ptr = btrfs_item_ptr_offset(eb, slot); + ptr_end = ptr + item_size; + while(ptr < ptr_end) { + di = (struct btrfs_dir_item *)ptr; + name_len = btrfs_dir_name_len(eb, di); + ret = replay_one_name(trans, root, path, eb, di, key); + BUG_ON(ret); + ptr = (unsigned long)(di + 1); + ptr += name_len; + } + return 0; +} + +/* + * directory replay has two parts. There are the standard directory + * items in the log copied from the subvolume, and range items + * created in the log while the subvolume was logged. + * + * The range items tell us which parts of the key space the log + * is authoritative for. During replay, if a key in the subvolume + * directory is in a logged range item, but not actually in the log + * that means it was deleted from the directory before the fsync + * and should be removed. + */ +static noinline int find_dir_range(struct btrfs_root *root, + struct btrfs_path *path, + u64 dirid, int key_type, + u64 *start_ret, u64 *end_ret) +{ + struct btrfs_key key; + u64 found_end; + struct btrfs_dir_log_item *item; + int ret; + int nritems; + + if (*start_ret == (u64)-1) + return 1; + + key.objectid = dirid; + key.type = key_type; + key.offset = *start_ret; + + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); + if (ret < 0) + goto out; + if (ret > 0) { + if (path->slots[0] == 0) + goto out; + path->slots[0]--; + } + if (ret != 0) + btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); + + if (key.type != key_type || key.objectid != dirid) { + ret = 1; + goto next; + } + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_dir_log_item); + found_end = btrfs_dir_log_end(path->nodes[0], item); + + if (*start_ret >= key.offset && *start_ret <= found_end) { + ret = 0; + *start_ret = key.offset; + *end_ret = found_end; + goto out; + } + ret = 1; +next: + /* check the next slot in the tree to see if it is a valid item */ + nritems = btrfs_header_nritems(path->nodes[0]); + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret) + goto out; + } else { + path->slots[0]++; + } + + btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); + + if (key.type != key_type || key.objectid != dirid) { + ret = 1; + goto out; + } + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_dir_log_item); + found_end = btrfs_dir_log_end(path->nodes[0], item); + *start_ret = key.offset; + *end_ret = found_end; + ret = 0; +out: + btrfs_release_path(root, path); + return ret; +} + +/* + * this looks for a given directory item in the log. If the directory + * item is not in the log, the item is removed and the inode it points + * to is unlinked + */ +static noinline int check_item_in_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_root *log, + struct btrfs_path *path, + struct btrfs_path *log_path, + struct inode *dir, + struct btrfs_key *dir_key) +{ + int ret; + struct extent_buffer *eb; + int slot; + u32 item_size; + struct btrfs_dir_item *di; + struct btrfs_dir_item *log_di; + int name_len; + unsigned long ptr; + unsigned long ptr_end; + char *name; + struct inode *inode; + struct btrfs_key location; + +again: + eb = path->nodes[0]; + slot = path->slots[0]; + item_size = btrfs_item_size_nr(eb, slot); + ptr = btrfs_item_ptr_offset(eb, slot); + ptr_end = ptr + item_size; + while(ptr < ptr_end) { + di = (struct btrfs_dir_item *)ptr; + name_len = btrfs_dir_name_len(eb, di); + name = kmalloc(name_len, GFP_NOFS); + if (!name) { + ret = -ENOMEM; + goto out; + } + read_extent_buffer(eb, name, (unsigned long)(di + 1), + name_len); + log_di = NULL; + if (dir_key->type == BTRFS_DIR_ITEM_KEY) { + log_di = btrfs_lookup_dir_item(trans, log, log_path, + dir_key->objectid, + name, name_len, 0); + } else if (dir_key->type == BTRFS_DIR_INDEX_KEY) { + log_di = btrfs_lookup_dir_index_item(trans, log, + log_path, + dir_key->objectid, + dir_key->offset, + name, name_len, 0); + } + if (!log_di || IS_ERR(log_di)) { + btrfs_dir_item_key_to_cpu(eb, di, &location); + btrfs_release_path(root, path); + btrfs_release_path(log, log_path); + inode = read_one_inode(root, location.objectid); + BUG_ON(!inode); + + ret = link_to_fixup_dir(trans, root, + path, location.objectid); + BUG_ON(ret); + btrfs_inc_nlink(inode); + ret = btrfs_unlink_inode(trans, root, dir, inode, + name, name_len); + BUG_ON(ret); + kfree(name); + iput(inode); + + /* there might still be more names under this key + * check and repeat if required + */ + ret = btrfs_search_slot(NULL, root, dir_key, path, + 0, 0); + if (ret == 0) + goto again; + ret = 0; + goto out; + } + btrfs_release_path(log, log_path); + kfree(name); + + ptr = (unsigned long)(di + 1); + ptr += name_len; + } + ret = 0; +out: + btrfs_release_path(root, path); + btrfs_release_path(log, log_path); + return ret; +} + +/* + * deletion replay happens before we copy any new directory items + * out of the log or out of backreferences from inodes. It + * scans the log to find ranges of keys that log is authoritative for, + * and then scans the directory to find items in those ranges that are + * not present in the log. + * + * Anything we don't find in the log is unlinked and removed from the + * directory. + */ +static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_root *log, + struct btrfs_path *path, + u64 dirid) +{ + u64 range_start; + u64 range_end; + int key_type = BTRFS_DIR_LOG_ITEM_KEY; + int ret = 0; + struct btrfs_key dir_key; + struct btrfs_key found_key; + struct btrfs_path *log_path; + struct inode *dir; + + dir_key.objectid = dirid; + dir_key.type = BTRFS_DIR_ITEM_KEY; + log_path = btrfs_alloc_path(); + if (!log_path) + return -ENOMEM; + + dir = read_one_inode(root, dirid); + /* it isn't an error if the inode isn't there, that can happen + * because we replay the deletes before we copy in the inode item + * from the log + */ + if (!dir) { + btrfs_free_path(log_path); + return 0; + } +again: + range_start = 0; + range_end = 0; + while(1) { + ret = find_dir_range(log, path, dirid, key_type, + &range_start, &range_end); + if (ret != 0) + break; + + dir_key.offset = range_start; + while(1) { + int nritems; + ret = btrfs_search_slot(NULL, root, &dir_key, path, + 0, 0); + if (ret < 0) + goto out; + + nritems = btrfs_header_nritems(path->nodes[0]); + if (path->slots[0] >= nritems) { + ret = btrfs_next_leaf(root, path); + if (ret) + break; + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + if (found_key.objectid != dirid || + found_key.type != dir_key.type) + goto next_type; + + if (found_key.offset > range_end) + break; + + ret = check_item_in_log(trans, root, log, path, + log_path, dir, &found_key); + BUG_ON(ret); + if (found_key.offset == (u64)-1) + break; + dir_key.offset = found_key.offset + 1; + } + btrfs_release_path(root, path); + if (range_end == (u64)-1) + break; + range_start = range_end + 1; + } + +next_type: + ret = 0; + if (key_type == BTRFS_DIR_LOG_ITEM_KEY) { + key_type = BTRFS_DIR_LOG_INDEX_KEY; + dir_key.type = BTRFS_DIR_INDEX_KEY; + btrfs_release_path(root, path); + goto again; + } +out: + btrfs_release_path(root, path); + btrfs_free_path(log_path); + iput(dir); + return ret; +} + +/* + * the process_func used to replay items from the log tree. This + * gets called in two different stages. The first stage just looks + * for inodes and makes sure they are all copied into the subvolume. + * + * The second stage copies all the other item types from the log into + * the subvolume. The two stage approach is slower, but gets rid of + * lots of complexity around inodes referencing other inodes that exist + * only in the log (references come from either directory items or inode + * back refs). + */ +static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb, + struct walk_control *wc, u64 gen) +{ + int nritems; + struct btrfs_path *path; + struct btrfs_root *root = wc->replay_dest; + struct btrfs_key key; + u32 item_size; + int level; + int i; + int ret; + + btrfs_read_buffer(eb, gen); + + level = btrfs_header_level(eb); + + if (level != 0) + return 0; + + path = btrfs_alloc_path(); + BUG_ON(!path); + + nritems = btrfs_header_nritems(eb); + for (i = 0; i < nritems; i++) { + btrfs_item_key_to_cpu(eb, &key, i); + item_size = btrfs_item_size_nr(eb, i); + + /* inode keys are done during the first stage */ + if (key.type == BTRFS_INODE_ITEM_KEY && + wc->stage == LOG_WALK_REPLAY_INODES) { + struct inode *inode; + struct btrfs_inode_item *inode_item; + u32 mode; + + inode_item = btrfs_item_ptr(eb, i, + struct btrfs_inode_item); + mode = btrfs_inode_mode(eb, inode_item); + if (S_ISDIR(mode)) { + ret = replay_dir_deletes(wc->trans, + root, log, path, key.objectid); + BUG_ON(ret); + } + ret = overwrite_item(wc->trans, root, path, + eb, i, &key); + BUG_ON(ret); + + /* for regular files, truncate away + * extents past the new EOF + */ + if (S_ISREG(mode)) { + inode = read_one_inode(root, + key.objectid); + BUG_ON(!inode); + + ret = btrfs_truncate_inode_items(wc->trans, + root, inode, inode->i_size, + BTRFS_EXTENT_DATA_KEY); + BUG_ON(ret); + iput(inode); + } + ret = link_to_fixup_dir(wc->trans, root, + path, key.objectid); + BUG_ON(ret); + } + if (wc->stage < LOG_WALK_REPLAY_ALL) + continue; + + /* these keys are simply copied */ + if (key.type == BTRFS_XATTR_ITEM_KEY) { + ret = overwrite_item(wc->trans, root, path, + eb, i, &key); + BUG_ON(ret); + } else if (key.type == BTRFS_INODE_REF_KEY) { + ret = add_inode_ref(wc->trans, root, log, path, + eb, i, &key); + BUG_ON(ret && ret != -ENOENT); + } else if (key.type == BTRFS_EXTENT_DATA_KEY) { + ret = replay_one_extent(wc->trans, root, path, + eb, i, &key); + BUG_ON(ret); + } else if (key.type == BTRFS_CSUM_ITEM_KEY) { + ret = replay_one_csum(wc->trans, root, path, + eb, i, &key); + BUG_ON(ret); + } else if (key.type == BTRFS_DIR_ITEM_KEY || + key.type == BTRFS_DIR_INDEX_KEY) { + ret = replay_one_dir_item(wc->trans, root, path, + eb, i, &key); + BUG_ON(ret); + } + } + btrfs_free_path(path); + return 0; +} + +static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int *level, + struct walk_control *wc) +{ + u64 root_owner; + u64 root_gen; + u64 bytenr; + u64 ptr_gen; + struct extent_buffer *next; + struct extent_buffer *cur; + struct extent_buffer *parent; + u32 blocksize; + int ret = 0; + + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + + while(*level > 0) { + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + cur = path->nodes[*level]; + + if (btrfs_header_level(cur) != *level) + WARN_ON(1); + + if (path->slots[*level] >= + btrfs_header_nritems(cur)) + break; + + bytenr = btrfs_node_blockptr(cur, path->slots[*level]); + ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]); + blocksize = btrfs_level_size(root, *level - 1); + + parent = path->nodes[*level]; + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); + + next = btrfs_find_create_tree_block(root, bytenr, blocksize); + + wc->process_func(root, next, wc, ptr_gen); + + if (*level == 1) { + path->slots[*level]++; + if (wc->free) { + btrfs_read_buffer(next, ptr_gen); + + btrfs_tree_lock(next); + clean_tree_block(trans, root, next); + btrfs_wait_tree_block_writeback(next); + btrfs_tree_unlock(next); + + ret = btrfs_drop_leaf_ref(trans, root, next); + BUG_ON(ret); + + WARN_ON(root_owner != + BTRFS_TREE_LOG_OBJECTID); + ret = btrfs_free_extent(trans, root, bytenr, + blocksize, root_owner, + root_gen, 0, 0, 1); + BUG_ON(ret); + } + free_extent_buffer(next); + continue; + } + btrfs_read_buffer(next, ptr_gen); + + WARN_ON(*level <= 0); + if (path->nodes[*level-1]) + free_extent_buffer(path->nodes[*level-1]); + path->nodes[*level-1] = next; + *level = btrfs_header_level(next); + path->slots[*level] = 0; + cond_resched(); + } + WARN_ON(*level < 0); + WARN_ON(*level >= BTRFS_MAX_LEVEL); + + if (path->nodes[*level] == root->node) { + parent = path->nodes[*level]; + } else { + parent = path->nodes[*level + 1]; + } + bytenr = path->nodes[*level]->start; + + blocksize = btrfs_level_size(root, *level); + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); + + wc->process_func(root, path->nodes[*level], wc, + btrfs_header_generation(path->nodes[*level])); + + if (wc->free) { + next = path->nodes[*level]; + btrfs_tree_lock(next); + clean_tree_block(trans, root, next); + btrfs_wait_tree_block_writeback(next); + btrfs_tree_unlock(next); + + if (*level == 0) { + ret = btrfs_drop_leaf_ref(trans, root, next); + BUG_ON(ret); + } + WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID); + ret = btrfs_free_extent(trans, root, bytenr, blocksize, + root_owner, root_gen, 0, 0, 1); + BUG_ON(ret); + } + free_extent_buffer(path->nodes[*level]); + path->nodes[*level] = NULL; + *level += 1; + + cond_resched(); + return 0; +} + +static int noinline walk_up_log_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, int *level, + struct walk_control *wc) +{ + u64 root_owner; + u64 root_gen; + int i; + int slot; + int ret; + + for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { + slot = path->slots[i]; + if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { + struct extent_buffer *node; + node = path->nodes[i]; + path->slots[i]++; + *level = i; + WARN_ON(*level == 0); + return 0; + } else { + if (path->nodes[*level] == root->node) { + root_owner = root->root_key.objectid; + root_gen = + btrfs_header_generation(path->nodes[*level]); + } else { + struct extent_buffer *node; + node = path->nodes[*level + 1]; + root_owner = btrfs_header_owner(node); + root_gen = btrfs_header_generation(node); + } + wc->process_func(root, path->nodes[*level], wc, + btrfs_header_generation(path->nodes[*level])); + if (wc->free) { + struct extent_buffer *next; + + next = path->nodes[*level]; + + btrfs_tree_lock(next); + clean_tree_block(trans, root, next); + btrfs_wait_tree_block_writeback(next); + btrfs_tree_unlock(next); + + if (*level == 0) { + ret = btrfs_drop_leaf_ref(trans, root, + next); + BUG_ON(ret); + } + + WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID); + ret = btrfs_free_extent(trans, root, + path->nodes[*level]->start, + path->nodes[*level]->len, + root_owner, root_gen, 0, 0, 1); + BUG_ON(ret); + } + free_extent_buffer(path->nodes[*level]); + path->nodes[*level] = NULL; + *level = i + 1; + } + } + return 1; +} + +/* + * drop the reference count on the tree rooted at 'snap'. This traverses + * the tree freeing any blocks that have a ref count of zero after being + * decremented. + */ +static int walk_log_tree(struct btrfs_trans_handle *trans, + struct btrfs_root *log, struct walk_control *wc) +{ + int ret = 0; + int wret; + int level; + struct btrfs_path *path; + int i; + int orig_level; + + path = btrfs_alloc_path(); + BUG_ON(!path); + + level = btrfs_header_level(log->node); + orig_level = level; + path->nodes[level] = log->node; + extent_buffer_get(log->node); + path->slots[level] = 0; + + while(1) { + wret = walk_down_log_tree(trans, log, path, &level, wc); + if (wret > 0) + break; + if (wret < 0) + ret = wret; + + wret = walk_up_log_tree(trans, log, path, &level, wc); + if (wret > 0) + break; + if (wret < 0) + ret = wret; + } + + /* was the root node processed? if not, catch it here */ + if (path->nodes[orig_level]) { + wc->process_func(log, path->nodes[orig_level], wc, + btrfs_header_generation(path->nodes[orig_level])); + if (wc->free) { + struct extent_buffer *next; + + next = path->nodes[orig_level]; + + btrfs_tree_lock(next); + clean_tree_block(trans, log, next); + btrfs_wait_tree_block_writeback(next); + btrfs_tree_unlock(next); + + if (orig_level == 0) { + ret = btrfs_drop_leaf_ref(trans, log, + next); + BUG_ON(ret); + } + WARN_ON(log->root_key.objectid != + BTRFS_TREE_LOG_OBJECTID); + ret = btrfs_free_extent(trans, log, + next->start, next->len, + log->root_key.objectid, + btrfs_header_generation(next), + 0, 0, 1); + BUG_ON(ret); + } + } + + for (i = 0; i <= orig_level; i++) { + if (path->nodes[i]) { + free_extent_buffer(path->nodes[i]); + path->nodes[i] = NULL; + } + } + btrfs_free_path(path); + if (wc->free) + free_extent_buffer(log->node); + return ret; +} + +int wait_log_commit(struct btrfs_root *log) +{ + DEFINE_WAIT(wait); + u64 transid = log->fs_info->tree_log_transid; + + do { + prepare_to_wait(&log->fs_info->tree_log_wait, &wait, + TASK_UNINTERRUPTIBLE); + mutex_unlock(&log->fs_info->tree_log_mutex); + if (atomic_read(&log->fs_info->tree_log_commit)) + schedule(); + finish_wait(&log->fs_info->tree_log_wait, &wait); + mutex_lock(&log->fs_info->tree_log_mutex); + } while(transid == log->fs_info->tree_log_transid && + atomic_read(&log->fs_info->tree_log_commit)); + return 0; +} + +/* + * btrfs_sync_log does sends a given tree log down to the disk and + * updates the super blocks to record it. When this call is done, + * you know that any inodes previously logged are safely on disk + */ +int btrfs_sync_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root) +{ + int ret; + unsigned long batch; + struct btrfs_root *log = root->log_root; + struct walk_control wc = { + .write = 1, + .process_func = process_one_buffer + }; + + mutex_lock(&log->fs_info->tree_log_mutex); + if (atomic_read(&log->fs_info->tree_log_commit)) { + wait_log_commit(log); + goto out; + } + atomic_set(&log->fs_info->tree_log_commit, 1); + + while(1) { + mutex_unlock(&log->fs_info->tree_log_mutex); + schedule_timeout_uninterruptible(1); + mutex_lock(&log->fs_info->tree_log_mutex); + batch = log->fs_info->tree_log_batch; + + while(atomic_read(&log->fs_info->tree_log_writers)) { + DEFINE_WAIT(wait); + prepare_to_wait(&log->fs_info->tree_log_wait, &wait, + TASK_UNINTERRUPTIBLE); + batch = log->fs_info->tree_log_batch; + mutex_unlock(&log->fs_info->tree_log_mutex); + if (atomic_read(&log->fs_info->tree_log_writers)) + schedule(); + mutex_lock(&log->fs_info->tree_log_mutex); + finish_wait(&log->fs_info->tree_log_wait, &wait); + } + if (batch == log->fs_info->tree_log_batch) + break; + } + ret = walk_log_tree(trans, log, &wc); + BUG_ON(ret); + + ret = walk_log_tree(trans, log->fs_info->log_root_tree, &wc); + BUG_ON(ret); + + wc.wait = 1; + + ret = walk_log_tree(trans, log, &wc); + BUG_ON(ret); + + ret = walk_log_tree(trans, log->fs_info->log_root_tree, &wc); + BUG_ON(ret); + + btrfs_set_super_log_root(&root->fs_info->super_for_commit, + log->fs_info->log_root_tree->node->start); + btrfs_set_super_log_root_level(&root->fs_info->super_for_commit, + btrfs_header_level(log->fs_info->log_root_tree->node)); + + write_ctree_super(trans, log->fs_info->tree_root); + log->fs_info->tree_log_transid++; + log->fs_info->tree_log_batch = 0; + atomic_set(&log->fs_info->tree_log_commit, 0); + smp_mb(); + if (waitqueue_active(&log->fs_info->tree_log_wait)) + wake_up(&log->fs_info->tree_log_wait); +out: + mutex_unlock(&log->fs_info->tree_log_mutex); + return 0; + +} + +/* + * free all the extents used by the tree log. This should be called + * at commit time of the full transaction + */ +int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root) +{ + int ret; + struct btrfs_root *log; + struct key; + struct walk_control wc = { + .free = 1, + .process_func = process_one_buffer + }; + + if (!root->log_root) + return 0; + + log = root->log_root; + ret = walk_log_tree(trans, log, &wc); + BUG_ON(ret); + + log = root->log_root; + ret = btrfs_del_root(trans, root->fs_info->log_root_tree, + &log->root_key); + BUG_ON(ret); + root->log_root = NULL; + kfree(root->log_root); + return 0; +} + +/* + * helper function to update the item for a given subvolumes log root + * in the tree of log roots + */ +static int update_log_root(struct btrfs_trans_handle *trans, + struct btrfs_root *log) +{ + u64 bytenr = btrfs_root_bytenr(&log->root_item); + int ret; + + if (log->node->start == bytenr) + return 0; + + btrfs_set_root_bytenr(&log->root_item, log->node->start); + btrfs_set_root_level(&log->root_item, btrfs_header_level(log->node)); + ret = btrfs_update_root(trans, log->fs_info->log_root_tree, + &log->root_key, &log->root_item); + BUG_ON(ret); + return ret; +} + +/* + * If both a file and directory are logged, and unlinks or renames are + * mixed in, we have a few interesting corners: + * + * create file X in dir Y + * link file X to X.link in dir Y + * fsync file X + * unlink file X but leave X.link + * fsync dir Y + * + * After a crash we would expect only X.link to exist. But file X + * didn't get fsync'd again so the log has back refs for X and X.link. + * + * We solve this by removing directory entries and inode backrefs from the + * log when a file that was logged in the current transaction is + * unlinked. Any later fsync will include the updated log entries, and + * we'll be able to reconstruct the proper directory items from backrefs. + * + * This optimizations allows us to avoid relogging the entire inode + * or the entire directory. + */ +int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + struct inode *dir, u64 index) +{ + struct btrfs_root *log; + struct btrfs_dir_item *di; + struct btrfs_path *path; + int ret; + int bytes_del = 0; + + ret = join_running_log_trans(root); + if (ret) + return 0; + + mutex_lock(&BTRFS_I(dir)->log_mutex); + + log = root->log_root; + path = btrfs_alloc_path(); + di = btrfs_lookup_dir_item(trans, log, path, dir->i_ino, + name, name_len, -1); + if (di && !IS_ERR(di)) { + ret = btrfs_delete_one_dir_name(trans, log, path, di); + bytes_del += name_len; + BUG_ON(ret); + } + btrfs_release_path(log, path); + di = btrfs_lookup_dir_index_item(trans, log, path, dir->i_ino, + index, name, name_len, -1); + if (di && !IS_ERR(di)) { + ret = btrfs_delete_one_dir_name(trans, log, path, di); + bytes_del += name_len; + BUG_ON(ret); + } + + /* update the directory size in the log to reflect the names + * we have removed + */ + if (bytes_del) { + struct btrfs_key key; + + key.objectid = dir->i_ino; + key.offset = 0; + key.type = BTRFS_INODE_ITEM_KEY; + btrfs_release_path(log, path); + + ret = btrfs_search_slot(trans, log, &key, path, 0, 1); + if (ret == 0) { + struct btrfs_inode_item *item; + u64 i_size; + + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_inode_item); + i_size = btrfs_inode_size(path->nodes[0], item); + if (i_size > bytes_del) + i_size -= bytes_del; + else + i_size = 0; + btrfs_set_inode_size(path->nodes[0], item, i_size); + btrfs_mark_buffer_dirty(path->nodes[0]); + } else + ret = 0; + btrfs_release_path(log, path); + } + + btrfs_free_path(path); + mutex_unlock(&BTRFS_I(dir)->log_mutex); + end_log_trans(root); + + return 0; +} + +/* see comments for btrfs_del_dir_entries_in_log */ +int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + struct inode *inode, u64 dirid) +{ + struct btrfs_root *log; + u64 index; + int ret; + + ret = join_running_log_trans(root); + if (ret) + return 0; + log = root->log_root; + mutex_lock(&BTRFS_I(inode)->log_mutex); + + ret = btrfs_del_inode_ref(trans, log, name, name_len, inode->i_ino, + dirid, &index); + mutex_unlock(&BTRFS_I(inode)->log_mutex); + end_log_trans(root); + + if (ret == 0 || ret == -ENOENT) + return 0; + return ret; +} + +/* + * creates a range item in the log for 'dirid'. first_offset and + * last_offset tell us which parts of the key space the log should + * be considered authoritative for. + */ +static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans, + struct btrfs_root *log, + struct btrfs_path *path, + int key_type, u64 dirid, + u64 first_offset, u64 last_offset) +{ + int ret; + struct btrfs_key key; + struct btrfs_dir_log_item *item; + + key.objectid = dirid; + key.offset = first_offset; + if (key_type == BTRFS_DIR_ITEM_KEY) + key.type = BTRFS_DIR_LOG_ITEM_KEY; + else + key.type = BTRFS_DIR_LOG_INDEX_KEY; + ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item)); + BUG_ON(ret); + + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_dir_log_item); + btrfs_set_dir_log_end(path->nodes[0], item, last_offset); + btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_release_path(log, path); + return 0; +} + +/* + * log all the items included in the current transaction for a given + * directory. This also creates the range items in the log tree required + * to replay anything deleted before the fsync + */ +static noinline int log_dir_items(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + struct btrfs_path *path, + struct btrfs_path *dst_path, int key_type, + u64 min_offset, u64 *last_offset_ret) +{ + struct btrfs_key min_key; + struct btrfs_key max_key; + struct btrfs_root *log = root->log_root; + struct extent_buffer *src; + int ret; + int i; + int nritems; + u64 first_offset = min_offset; + u64 last_offset = (u64)-1; + + log = root->log_root; + max_key.objectid = inode->i_ino; + max_key.offset = (u64)-1; + max_key.type = key_type; + + min_key.objectid = inode->i_ino; + min_key.type = key_type; + min_key.offset = min_offset; + + path->keep_locks = 1; + + ret = btrfs_search_forward(root, &min_key, &max_key, + path, 0, trans->transid); + + /* + * we didn't find anything from this transaction, see if there + * is anything at all + */ + if (ret != 0 || min_key.objectid != inode->i_ino || + min_key.type != key_type) { + min_key.objectid = inode->i_ino; + min_key.type = key_type; + min_key.offset = (u64)-1; + btrfs_release_path(root, path); + ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0); + if (ret < 0) { + btrfs_release_path(root, path); + return ret; + } + ret = btrfs_previous_item(root, path, inode->i_ino, key_type); + + /* if ret == 0 there are items for this type, + * create a range to tell us the last key of this type. + * otherwise, there are no items in this directory after + * *min_offset, and we create a range to indicate that. + */ + if (ret == 0) { + struct btrfs_key tmp; + btrfs_item_key_to_cpu(path->nodes[0], &tmp, + path->slots[0]); + if (key_type == tmp.type) { + first_offset = max(min_offset, tmp.offset) + 1; + } + } + goto done; + } + + /* go backward to find any previous key */ + ret = btrfs_previous_item(root, path, inode->i_ino, key_type); + if (ret == 0) { + struct btrfs_key tmp; + btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]); + if (key_type == tmp.type) { + first_offset = tmp.offset; + ret = overwrite_item(trans, log, dst_path, + path->nodes[0], path->slots[0], + &tmp); + } + } + btrfs_release_path(root, path); + + /* find the first key from this transaction again */ + ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0); + if (ret != 0) { + WARN_ON(1); + goto done; + } + + /* + * we have a block from this transaction, log every item in it + * from our directory + */ + while(1) { + struct btrfs_key tmp; + src = path->nodes[0]; + nritems = btrfs_header_nritems(src); + for (i = path->slots[0]; i < nritems; i++) { + btrfs_item_key_to_cpu(src, &min_key, i); + + if (min_key.objectid != inode->i_ino || + min_key.type != key_type) + goto done; + ret = overwrite_item(trans, log, dst_path, src, i, + &min_key); + BUG_ON(ret); + } + path->slots[0] = nritems; + + /* + * look ahead to the next item and see if it is also + * from this directory and from this transaction + */ + ret = btrfs_next_leaf(root, path); + if (ret == 1) { + last_offset = (u64)-1; + goto done; + } + btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]); + if (tmp.objectid != inode->i_ino || tmp.type != key_type) { + last_offset = (u64)-1; + goto done; + } + if (btrfs_header_generation(path->nodes[0]) != trans->transid) { + ret = overwrite_item(trans, log, dst_path, + path->nodes[0], path->slots[0], + &tmp); + + BUG_ON(ret); + last_offset = tmp.offset; + goto done; + } + } +done: + *last_offset_ret = last_offset; + btrfs_release_path(root, path); + btrfs_release_path(log, dst_path); + + /* insert the log range keys to indicate where the log is valid */ + ret = insert_dir_log_key(trans, log, path, key_type, inode->i_ino, + first_offset, last_offset); + BUG_ON(ret); + return 0; +} + +/* + * logging directories is very similar to logging inodes, We find all the items + * from the current transaction and write them to the log. + * + * The recovery code scans the directory in the subvolume, and if it finds a + * key in the range logged that is not present in the log tree, then it means + * that dir entry was unlinked during the transaction. + * + * In order for that scan to work, we must include one key smaller than + * the smallest logged by this transaction and one key larger than the largest + * key logged by this transaction. + */ +static noinline int log_directory_changes(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + struct btrfs_path *path, + struct btrfs_path *dst_path) +{ + u64 min_key; + u64 max_key; + int ret; + int key_type = BTRFS_DIR_ITEM_KEY; + +again: + min_key = 0; + max_key = 0; + while(1) { + ret = log_dir_items(trans, root, inode, path, + dst_path, key_type, min_key, + &max_key); + BUG_ON(ret); + if (max_key == (u64)-1) + break; + min_key = max_key + 1; + } + + if (key_type == BTRFS_DIR_ITEM_KEY) { + key_type = BTRFS_DIR_INDEX_KEY; + goto again; + } + return 0; +} + +/* + * a helper function to drop items from the log before we relog an + * inode. max_key_type indicates the highest item type to remove. + * This cannot be run for file data extents because it does not + * free the extents they point to. + */ +static int drop_objectid_items(struct btrfs_trans_handle *trans, + struct btrfs_root *log, + struct btrfs_path *path, + u64 objectid, int max_key_type) +{ + int ret; + struct btrfs_key key; + struct btrfs_key found_key; + + key.objectid = objectid; + key.type = max_key_type; + key.offset = (u64)-1; + + while(1) { + ret = btrfs_search_slot(trans, log, &key, path, -1, 1); + + if (ret != 1) + break; + + if (path->slots[0] == 0) + break; + + path->slots[0]--; + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + + if (found_key.objectid != objectid) + break; + + ret = btrfs_del_item(trans, log, path); + BUG_ON(ret); + btrfs_release_path(log, path); + } + btrfs_release_path(log, path); + return 0; +} + +/* log a single inode in the tree log. + * At least one parent directory for this inode must exist in the tree + * or be logged already. + * + * Any items from this inode changed by the current transaction are copied + * to the log tree. An extra reference is taken on any extents in this + * file, allowing us to avoid a whole pile of corner cases around logging + * blocks that have been removed from the tree. + * + * See LOG_INODE_ALL and related defines for a description of what inode_only + * does. + * + * This handles both files and directories. + */ +static int __btrfs_log_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + int inode_only) +{ + struct btrfs_path *path; + struct btrfs_path *dst_path; + struct btrfs_key min_key; + struct btrfs_key max_key; + struct btrfs_root *log = root->log_root; + unsigned long src_offset; + unsigned long dst_offset; + struct extent_buffer *src; + struct btrfs_file_extent_item *extent; + struct btrfs_inode_item *inode_item; + u32 size; + int ret; + + log = root->log_root; + + path = btrfs_alloc_path(); + dst_path = btrfs_alloc_path(); + + min_key.objectid = inode->i_ino; + min_key.type = BTRFS_INODE_ITEM_KEY; + min_key.offset = 0; + + max_key.objectid = inode->i_ino; + if (inode_only == LOG_INODE_EXISTS || S_ISDIR(inode->i_mode)) + max_key.type = BTRFS_XATTR_ITEM_KEY; + else + max_key.type = (u8)-1; + max_key.offset = (u64)-1; + + /* + * if this inode has already been logged and we're in inode_only + * mode, we don't want to delete the things that have already + * been written to the log. + * + * But, if the inode has been through an inode_only log, + * the logged_trans field is not set. This allows us to catch + * any new names for this inode in the backrefs by logging it + * again + */ + if (inode_only == LOG_INODE_EXISTS && + BTRFS_I(inode)->logged_trans == trans->transid) { + btrfs_free_path(path); + btrfs_free_path(dst_path); + goto out; + } + mutex_lock(&BTRFS_I(inode)->log_mutex); + + /* + * a brute force approach to making sure we get the most uptodate + * copies of everything. + */ + if (S_ISDIR(inode->i_mode)) { + int max_key_type = BTRFS_DIR_LOG_INDEX_KEY; + + if (inode_only == LOG_INODE_EXISTS) + max_key_type = BTRFS_XATTR_ITEM_KEY; + ret = drop_objectid_items(trans, log, path, + inode->i_ino, max_key_type); + } else { + ret = btrfs_truncate_inode_items(trans, log, inode, 0, 0); + } + BUG_ON(ret); + path->keep_locks = 1; + + while(1) { + ret = btrfs_search_forward(root, &min_key, &max_key, + path, 0, trans->transid); + if (ret != 0) + break; + + if (min_key.objectid != inode->i_ino) + break; + if (min_key.type > max_key.type) + break; + + src = path->nodes[0]; + size = btrfs_item_size_nr(src, path->slots[0]); + ret = btrfs_insert_empty_item(trans, log, dst_path, &min_key, + size); + if (ret) + BUG(); + + dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0], + dst_path->slots[0]); + + src_offset = btrfs_item_ptr_offset(src, path->slots[0]); + + copy_extent_buffer(dst_path->nodes[0], src, dst_offset, + src_offset, size); + + if (inode_only == LOG_INODE_EXISTS && + min_key.type == BTRFS_INODE_ITEM_KEY) { + inode_item = btrfs_item_ptr(dst_path->nodes[0], + dst_path->slots[0], + struct btrfs_inode_item); + btrfs_set_inode_size(dst_path->nodes[0], inode_item, 0); + + /* set the generation to zero so the recover code + * can tell the difference between an logging + * just to say 'this inode exists' and a logging + * to say 'update this inode with these values' + */ + btrfs_set_inode_generation(dst_path->nodes[0], + inode_item, 0); + } + /* take a reference on file data extents so that truncates + * or deletes of this inode don't have to relog the inode + * again + */ + if (btrfs_key_type(&min_key) == BTRFS_EXTENT_DATA_KEY) { + int found_type; + extent = btrfs_item_ptr(src, path->slots[0], + struct btrfs_file_extent_item); + + found_type = btrfs_file_extent_type(src, extent); + if (found_type == BTRFS_FILE_EXTENT_REG) { + u64 ds = btrfs_file_extent_disk_bytenr(src, + extent); + u64 dl = btrfs_file_extent_disk_num_bytes(src, + extent); + /* ds == 0 is a hole */ + if (ds != 0) { + ret = btrfs_inc_extent_ref(trans, log, + ds, dl, + log->root_key.objectid, + 0, + inode->i_ino, + min_key.offset); + BUG_ON(ret); + } + } + } + + btrfs_mark_buffer_dirty(dst_path->nodes[0]); + btrfs_release_path(root, path); + btrfs_release_path(log, dst_path); + + if (min_key.offset < (u64)-1) + min_key.offset++; + else if (min_key.type < (u8)-1) + min_key.type++; + else if (min_key.objectid < (u64)-1) + min_key.objectid++; + else + break; + } + if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) { + btrfs_release_path(root, path); + btrfs_release_path(log, dst_path); + ret = log_directory_changes(trans, root, inode, path, dst_path); + BUG_ON(ret); + } + mutex_unlock(&BTRFS_I(inode)->log_mutex); + + btrfs_free_path(path); + btrfs_free_path(dst_path); + + mutex_lock(&root->fs_info->tree_log_mutex); + ret = update_log_root(trans, log); + BUG_ON(ret); + mutex_unlock(&root->fs_info->tree_log_mutex); +out: + return 0; +} + +int btrfs_log_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + int inode_only) +{ + int ret; + + start_log_trans(trans, root); + ret = __btrfs_log_inode(trans, root, inode, inode_only); + end_log_trans(root); + return ret; +} + +/* + * helper function around btrfs_log_inode to make sure newly created + * parent directories also end up in the log. A minimal inode and backref + * only logging is done of any parent directories that are older than + * the last committed transaction + */ +int btrfs_log_dentry(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct dentry *dentry) +{ + int inode_only = LOG_INODE_ALL; + struct super_block *sb; + int ret; + + start_log_trans(trans, root); + sb = dentry->d_inode->i_sb; + while(1) { + ret = __btrfs_log_inode(trans, root, dentry->d_inode, + inode_only); + BUG_ON(ret); + inode_only = LOG_INODE_EXISTS; + + dentry = dentry->d_parent; + if (!dentry || !dentry->d_inode || sb != dentry->d_inode->i_sb) + break; + + if (BTRFS_I(dentry->d_inode)->generation <= + root->fs_info->last_trans_committed) + break; + } + end_log_trans(root); + return 0; +} + +/* + * it is not safe to log dentry if the chunk root has added new + * chunks. This returns 0 if the dentry was logged, and 1 otherwise. + * If this returns 1, you must commit the transaction to safely get your + * data on disk. + */ +int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct dentry *dentry) +{ + u64 gen; + gen = root->fs_info->last_trans_new_blockgroup; + if (gen > root->fs_info->last_trans_committed) + return 1; + else + return btrfs_log_dentry(trans, root, dentry); +} + +/* + * should be called during mount to recover any replay any log trees + * from the FS + */ +int btrfs_recover_log_trees(struct btrfs_root *log_root_tree) +{ + int ret; + struct btrfs_path *path; + struct btrfs_trans_handle *trans; + struct btrfs_key key; + struct btrfs_key found_key; + struct btrfs_key tmp_key; + struct btrfs_root *log; + struct btrfs_fs_info *fs_info = log_root_tree->fs_info; + struct walk_control wc = { + .process_func = process_one_buffer, + .stage = 0, + }; + + fs_info->log_root_recovering = 1; + path = btrfs_alloc_path(); + BUG_ON(!path); + + trans = btrfs_start_transaction(fs_info->tree_root, 1); + + wc.trans = trans; + wc.pin = 1; + + walk_log_tree(trans, log_root_tree, &wc); + +again: + key.objectid = BTRFS_TREE_LOG_OBJECTID; + key.offset = (u64)-1; + btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); + + while(1) { + ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0); + if (ret < 0) + break; + if (ret > 0) { + if (path->slots[0] == 0) + break; + path->slots[0]--; + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + btrfs_release_path(log_root_tree, path); + if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID) + break; + + log = btrfs_read_fs_root_no_radix(log_root_tree, + &found_key); + BUG_ON(!log); + + + tmp_key.objectid = found_key.offset; + tmp_key.type = BTRFS_ROOT_ITEM_KEY; + tmp_key.offset = (u64)-1; + + wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key); + + BUG_ON(!wc.replay_dest); + + btrfs_record_root_in_trans(wc.replay_dest); + ret = walk_log_tree(trans, log, &wc); + BUG_ON(ret); + + if (wc.stage == LOG_WALK_REPLAY_ALL) { + ret = fixup_inode_link_counts(trans, wc.replay_dest, + path); + BUG_ON(ret); + } + + key.offset = found_key.offset - 1; + free_extent_buffer(log->node); + kfree(log); + + if (found_key.offset == 0) + break; + } + btrfs_release_path(log_root_tree, path); + + /* step one is to pin it all, step two is to replay just inodes */ + if (wc.pin) { + wc.pin = 0; + wc.process_func = replay_one_buffer; + wc.stage = LOG_WALK_REPLAY_INODES; + goto again; + } + /* step three is to replay everything */ + if (wc.stage < LOG_WALK_REPLAY_ALL) { + wc.stage++; + goto again; + } + + btrfs_free_path(path); + + free_extent_buffer(log_root_tree->node); + log_root_tree->log_root = NULL; + fs_info->log_root_recovering = 0; + + /* step 4: commit the transaction, which also unpins the blocks */ + btrfs_commit_transaction(trans, fs_info->tree_root); + + kfree(log_root_tree); + return 0; +} diff --git a/fs/btrfs/tree-log.h b/fs/btrfs/tree-log.h new file mode 100644 index 00000000000..b9409b32ed0 --- /dev/null +++ b/fs/btrfs/tree-log.h @@ -0,0 +1,41 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __TREE_LOG_ +#define __TREE_LOG_ + +int btrfs_sync_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root); +int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root); +int btrfs_log_dentry(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct dentry *dentry); +int btrfs_recover_log_trees(struct btrfs_root *tree_root); +int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct dentry *dentry); +int btrfs_log_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + int inode_only); +int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + struct inode *dir, u64 index); +int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + const char *name, int name_len, + struct inode *inode, u64 dirid); +#endif -- cgit v1.2.3-70-g09d2 From a237d2a2bd67eb2b57b87c1e1d3d802d078d4ba7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 5 Sep 2008 16:43:29 -0400 Subject: remove unused function btrfs_ilookup btrfs_ilookup is unused, which is good because a normal filesystem should never have to use ilookup anyway. Remove it. Signed-off-by: Christoph Hellwig Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 -- fs/btrfs/inode.c | 13 ------------- 2 files changed, 15 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f720b582fd5..754c738ec2e 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1763,8 +1763,6 @@ void btrfs_destroy_cachep(void); long btrfs_ioctl_trans_end(struct file *file); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); -struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, - u64 root_objectid); struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, struct btrfs_root *root, int *is_new); int btrfs_commit_write(struct file *file, struct page *page, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 65df9d83023..b4dbb5c9764 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1818,19 +1818,6 @@ static int btrfs_find_actor(struct inode *inode, void *opaque) args->root == BTRFS_I(inode)->root); } -struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, - u64 root_objectid) -{ - struct btrfs_iget_args args; - args.ino = objectid; - args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid); - - if (!args.root) - return NULL; - - return ilookup5(s, objectid, btrfs_find_actor, (void *)&args); -} - struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root) { -- cgit v1.2.3-70-g09d2 From 8d5bf1cb35ea29795862ff0ea2f4c4d7e22727f3 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 11 Sep 2008 15:51:21 -0400 Subject: Btrfs: Update the highest objectid in a root after log replay is done Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 3 ++- fs/btrfs/tree-log.c | 7 +++++++ 2 files changed, 9 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b4dbb5c9764..073fdd77c78 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2239,8 +2239,9 @@ static int btrfs_set_inode_index(struct inode *dir, struct inode *inode, if (BTRFS_I(dir)->index_cnt == (u64)-1) { ret = btrfs_set_inode_index_count(dir); - if (ret) + if (ret) { return ret; + } } *index = BTRFS_I(dir)->index_cnt; diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 13d7ee8e0c5..3f4b139b27e 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -1027,6 +1027,7 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans, inode->i_nlink = nlink; btrfs_update_inode(trans, root, inode); } + BTRFS_I(inode)->index_cnt = (u64)-1; return 0; } @@ -2714,6 +2715,7 @@ int btrfs_recover_log_trees(struct btrfs_root *log_root_tree) struct btrfs_key tmp_key; struct btrfs_root *log; struct btrfs_fs_info *fs_info = log_root_tree->fs_info; + u64 highest_inode; struct walk_control wc = { .process_func = process_one_buffer, .stage = 0, @@ -2772,6 +2774,11 @@ again: path); BUG_ON(ret); } + ret = btrfs_find_highest_inode(wc.replay_dest, &highest_inode); + if (ret == 0) { + wc.replay_dest->highest_inode = highest_inode; + wc.replay_dest->last_inode_alloc = highest_inode; + } key.offset = found_key.offset - 1; free_extent_buffer(log->node); -- cgit v1.2.3-70-g09d2 From 98509cfc5a6857bddcfe4b19a9539726655ec9bd Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 11 Sep 2008 15:51:43 -0400 Subject: Btrfs: Fix releasepage to properly keep dirty and writeback pages Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 5 ++++- fs/btrfs/inode.c | 2 ++ 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 42bf9916805..ecb74b72026 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -346,7 +346,7 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); found_start = btrfs_header_bytenr(eb); - if (found_start != start) { + if (0 && found_start != start) { printk("bad tree block start %llu %llu\n", (unsigned long long)found_start, (unsigned long long)eb->start); @@ -592,6 +592,9 @@ static int btree_releasepage(struct page *page, gfp_t gfp_flags) struct extent_map_tree *map; int ret; + if (PageWriteback(page) || PageDirty(page)) + return 0; + tree = &BTRFS_I(page->mapping->host)->io_tree; map = &BTRFS_I(page->mapping->host)->extent_tree; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 073fdd77c78..24b7e97fccb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3060,6 +3060,8 @@ static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags) static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) { + if (PageWriteback(page) || PageDirty(page)) + return 0; return __btrfs_releasepage(page, gfp_flags); } -- cgit v1.2.3-70-g09d2 From 49eb7e46d47ea72a9bd2a5f8cedb04f5159cc277 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 11 Sep 2008 15:53:12 -0400 Subject: Btrfs: Dir fsync optimizations Drop i_mutex during the commit Don't bother doing the fsync at all unless the dir is marked as dirtied and needing fsync in this transaction. For directories, this means that someone has unlinked a file from the dir without fsyncing the file. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 3 +++ fs/btrfs/file.c | 19 ++++++++++++++++++- fs/btrfs/inode.c | 5 ++++- fs/btrfs/tree-log.c | 8 ++++---- 4 files changed, 29 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index fcc8cf27e90..0577fda2168 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -56,6 +56,9 @@ struct btrfs_inode { * transid that last logged this inode */ u64 logged_trans; + + /* trans that last made a change that should be fully fsync'd */ + u64 log_dirty_trans; u64 delalloc_bytes; u64 disk_i_size; u32 flags; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 84ecf3ab851..58b329ddb42 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1061,7 +1061,9 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) } mutex_unlock(&root->fs_info->trans_mutex); + root->fs_info->tree_log_batch++; filemap_fdatawait(inode->i_mapping); + root->fs_info->tree_log_batch++; /* * ok we haven't committed the transaction yet, lets do a commit @@ -1076,14 +1078,29 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) } ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); - if (ret < 0) + if (ret < 0) { goto out; + } + + /* we've logged all the items and now have a consistent + * version of the file in the log. It is possible that + * someone will come in and modify the file, but that's + * fine because the log is consistent on disk, and we + * have references to all of the file's extents + * + * It is possible that someone will come in and log the + * file again, but that will end up using the synchronization + * inside btrfs_sync_log to keep things safe. + */ + mutex_unlock(&file->f_dentry->d_inode->i_mutex); + if (ret > 0) { ret = btrfs_commit_transaction(trans, root); } else { btrfs_sync_log(trans, root); ret = btrfs_end_transaction(trans, root); } + mutex_lock(&file->f_dentry->d_inode->i_mutex); out: return ret > 0 ? EIO : ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 24b7e97fccb..12c1c0530f3 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1187,7 +1187,9 @@ int btrfs_unlink_inode(struct btrfs_trans_handle *trans, ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len, inode, dir->i_ino); - BUG_ON(ret); + BUG_ON(ret != 0 && ret != -ENOENT); + if (ret != -ENOENT) + BTRFS_I(dir)->log_dirty_trans = trans->transid; ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len, dir, index); @@ -1790,6 +1792,7 @@ static noinline void init_btrfs_i(struct inode *inode) bi->disk_i_size = 0; bi->flags = 0; bi->index_cnt = (u64)-1; + bi->log_dirty_trans = 0; extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping, GFP_NOFS); diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 3f4b139b27e..5d49a701bdc 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -1973,10 +1973,10 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans, atomic_set(&log->fs_info->tree_log_commit, 1); while(1) { + batch = log->fs_info->tree_log_batch; mutex_unlock(&log->fs_info->tree_log_mutex); schedule_timeout_uninterruptible(1); mutex_lock(&log->fs_info->tree_log_mutex); - batch = log->fs_info->tree_log_batch; while(atomic_read(&log->fs_info->tree_log_writers)) { DEFINE_WAIT(wait); @@ -2189,8 +2189,6 @@ int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans, mutex_unlock(&BTRFS_I(inode)->log_mutex); end_log_trans(root); - if (ret == 0 || ret == -ENOENT) - return 0; return ret; } @@ -2620,9 +2618,11 @@ static int __btrfs_log_inode(struct btrfs_trans_handle *trans, else break; } - if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) { + if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode) && + BTRFS_I(inode)->log_dirty_trans >= trans->transid) { btrfs_release_path(root, path); btrfs_release_path(log, dst_path); + BTRFS_I(inode)->log_dirty_trans = 0; ret = log_directory_changes(trans, root, inode, path, dst_path); BUG_ON(ret); } -- cgit v1.2.3-70-g09d2 From 0f9dd46cda36b8de3b9f48bc42bd09d20b9c3b52 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 23 Sep 2008 13:14:11 -0400 Subject: Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/ctree.c | 3 +- fs/btrfs/ctree.h | 46 ++- fs/btrfs/disk-io.c | 7 +- fs/btrfs/extent-tree.c | 869 +++++++++++++++++++++++--------------------- fs/btrfs/extent_io.c | 4 + fs/btrfs/free-space-cache.c | 415 +++++++++++++++++++++ fs/btrfs/inode.c | 3 +- fs/btrfs/volumes.c | 11 +- 9 files changed, 925 insertions(+), 435 deletions(-) create mode 100644 fs/btrfs/free-space-cache.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index b7addbfd8c2..eb36ae981bd 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -7,7 +7,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ transaction.o bit-radix.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \ - ref-cache.o export.o tree-log.o acl.o + ref-cache.o export.o tree-log.o acl.o free-space-cache.o else # Normal Makefile diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 18e84472abb..6f467901246 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2725,9 +2725,8 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, total_size = total_data + (nr * sizeof(struct btrfs_item)); ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); - if (ret == 0) { + if (ret == 0) return -EEXIST; - } if (ret < 0) goto out; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index eb65fd80888..730aae3bc18 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -483,7 +483,6 @@ struct btrfs_csum_item { #define BTRFS_BLOCK_GROUP_DUP (1 << 5) #define BTRFS_BLOCK_GROUP_RAID10 (1 << 6) - struct btrfs_block_group_item { __le64 used; __le64 chunk_objectid; @@ -498,17 +497,40 @@ struct btrfs_space_info { int full; int force_alloc; struct list_head list; + + /* for block groups in our same type */ + struct list_head block_groups; + spinlock_t lock; +}; + +struct btrfs_free_space { + struct rb_node bytes_index; + struct rb_node offset_index; + u64 offset; + u64 bytes; }; struct btrfs_block_group_cache { struct btrfs_key key; struct btrfs_block_group_item item; - struct btrfs_space_info *space_info; spinlock_t lock; u64 pinned; u64 flags; int cached; int ro; + int dirty; + + struct btrfs_space_info *space_info; + + /* free space cache stuff */ + struct rb_root free_space_bytes; + struct rb_root free_space_offset; + + /* block group cache stuff */ + struct rb_node cache_node; + + /* for block groups in the same raid type */ + struct list_head list; }; struct btrfs_device; @@ -525,8 +547,10 @@ struct btrfs_fs_info { struct btrfs_root *log_root_tree; struct radix_tree_root fs_roots_radix; - struct extent_io_tree free_space_cache; - struct extent_io_tree block_group_cache; + /* block group cache stuff */ + spinlock_t block_group_cache_lock; + struct rb_root block_group_cache_tree; + struct extent_io_tree pinned_extents; struct extent_io_tree pending_del; struct extent_io_tree extent_ins; @@ -1814,4 +1838,18 @@ int btrfs_sync_fs(struct super_block *sb, int wait); int btrfs_check_acl(struct inode *inode, int mask); int btrfs_init_acl(struct inode *inode, struct inode *dir); int btrfs_acl_chmod(struct inode *inode); + +/* free-space-cache.c */ +int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, + u64 bytenr, u64 size); +int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group, + u64 bytenr, u64 size); +void btrfs_remove_free_space_cache(struct btrfs_block_group_cache + *block_group); +struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache + *block_group, u64 offset, + u64 bytes); +void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group, + u64 bytes); +u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group); #endif diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index f6f7821d43a..535bd0fe1a7 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1410,10 +1410,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; - extent_io_tree_init(&fs_info->free_space_cache, - fs_info->btree_inode->i_mapping, GFP_NOFS); - extent_io_tree_init(&fs_info->block_group_cache, - fs_info->btree_inode->i_mapping, GFP_NOFS); + spin_lock_init(&fs_info->block_group_cache_lock); + fs_info->block_group_cache_tree.rb_node = NULL; + extent_io_tree_init(&fs_info->pinned_extents, fs_info->btree_inode->i_mapping, GFP_NOFS); extent_io_tree_init(&fs_info->pending_del, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 1c10ffc837c..813566acc5d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -29,12 +29,6 @@ #include "locking.h" #include "ref-cache.h" -#define BLOCK_GROUP_DATA EXTENT_WRITEBACK -#define BLOCK_GROUP_METADATA EXTENT_UPTODATE -#define BLOCK_GROUP_SYSTEM EXTENT_NEW - -#define BLOCK_GROUP_DIRTY EXTENT_DIRTY - static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); static int del_pending_extents(struct btrfs_trans_handle *trans, struct @@ -62,6 +56,127 @@ void maybe_unlock_mutex(struct btrfs_root *root) } } +static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) +{ + return (cache->flags & bits) == bits; +} + +/* + * this adds the block group to the fs_info rb tree for the block group + * cache + */ +int btrfs_add_block_group_cache(struct btrfs_fs_info *info, + struct btrfs_block_group_cache *block_group) +{ + struct rb_node **p; + struct rb_node *parent = NULL; + struct btrfs_block_group_cache *cache; + + spin_lock(&info->block_group_cache_lock); + p = &info->block_group_cache_tree.rb_node; + + while (*p) { + parent = *p; + cache = rb_entry(parent, struct btrfs_block_group_cache, + cache_node); + if (block_group->key.objectid < cache->key.objectid) { + p = &(*p)->rb_left; + } else if (block_group->key.objectid > cache->key.objectid) { + p = &(*p)->rb_right; + } else { + spin_unlock(&info->block_group_cache_lock); + return -EEXIST; + } + } + + rb_link_node(&block_group->cache_node, parent, p); + rb_insert_color(&block_group->cache_node, + &info->block_group_cache_tree); + spin_unlock(&info->block_group_cache_lock); + + return 0; +} + +/* + * This will return the block group at or after bytenr if contains is 0, else + * it will return the block group that contains the bytenr + */ +static struct btrfs_block_group_cache * +block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr, + int contains) +{ + struct btrfs_block_group_cache *cache, *ret = NULL; + struct rb_node *n; + u64 end, start; + + spin_lock(&info->block_group_cache_lock); + n = info->block_group_cache_tree.rb_node; + + while (n) { + cache = rb_entry(n, struct btrfs_block_group_cache, + cache_node); + end = cache->key.objectid + cache->key.offset - 1; + start = cache->key.objectid; + + if (bytenr < start) { + if (!contains && (!ret || start < ret->key.objectid)) + ret = cache; + n = n->rb_left; + } else if (bytenr > start) { + if (contains && bytenr <= end) { + ret = cache; + break; + } + n = n->rb_right; + } else { + ret = cache; + break; + } + } + spin_unlock(&info->block_group_cache_lock); + + return ret; +} + +/* + * this is only called by cache_block_group, since we could have freed extents + * we need to check the pinned_extents for any extents that can't be used yet + * since their free space will be released as soon as the transaction commits. + */ +static int add_new_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_fs_info *info, u64 start, u64 end) +{ + u64 extent_start, extent_end, size; + int ret; + + while (start < end) { + ret = find_first_extent_bit(&info->pinned_extents, start, + &extent_start, &extent_end, + EXTENT_DIRTY); + if (ret) + break; + + if (extent_start == start) { + start = extent_end + 1; + } else if (extent_start > start && extent_start < end) { + size = extent_start - start; + ret = btrfs_add_free_space(block_group, start, size); + BUG_ON(ret); + start = extent_end + 1; + } else { + break; + } + } + + if (start < end) { + size = end - start; + ret = btrfs_add_free_space(block_group, start, size); + BUG_ON(ret); + } + + return 0; +} + static int cache_block_group(struct btrfs_root *root, struct btrfs_block_group_cache *block_group) { @@ -69,10 +184,8 @@ static int cache_block_group(struct btrfs_root *root, int ret = 0; struct btrfs_key key; struct extent_buffer *leaf; - struct extent_io_tree *free_space_cache; int slot; u64 last = 0; - u64 hole_size; u64 first_free; int found = 0; @@ -80,7 +193,6 @@ static int cache_block_group(struct btrfs_root *root, return 0; root = root->fs_info->extent_root; - free_space_cache = &root->fs_info->free_space_cache; if (block_group->cached) return 0; @@ -96,7 +208,8 @@ static int cache_block_group(struct btrfs_root *root, * skip the locking here */ path->skip_locking = 1; - first_free = block_group->key.objectid; + first_free = max_t(u64, block_group->key.objectid, + BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); key.objectid = block_group->key.objectid; key.offset = 0; btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); @@ -119,32 +232,28 @@ static int cache_block_group(struct btrfs_root *root, ret = btrfs_next_leaf(root, path); if (ret < 0) goto err; - if (ret == 0) { + if (ret == 0) continue; - } else { + else break; - } } btrfs_item_key_to_cpu(leaf, &key, slot); - if (key.objectid < block_group->key.objectid) { + if (key.objectid < block_group->key.objectid) goto next; - } + if (key.objectid >= block_group->key.objectid + - block_group->key.offset) { + block_group->key.offset) break; - } if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) { if (!found) { last = first_free; found = 1; } - if (key.objectid > last) { - hole_size = key.objectid - last; - set_extent_dirty(free_space_cache, last, - last + hole_size - 1, - GFP_NOFS); - } + + add_new_free_space(block_group, root->fs_info, last, + key.objectid); + last = key.objectid + key.offset; } next: @@ -153,13 +262,11 @@ next: if (!found) last = first_free; - if (block_group->key.objectid + - block_group->key.offset > last) { - hole_size = block_group->key.objectid + - block_group->key.offset - last; - set_extent_dirty(free_space_cache, last, - last + hole_size - 1, GFP_NOFS); - } + + add_new_free_space(block_group, root->fs_info, last, + block_group->key.objectid + + block_group->key.offset); + block_group->cached = 1; ret = 0; err: @@ -167,166 +274,79 @@ err: return ret; } +/* + * return the block group that starts at or after bytenr + */ struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr) { - struct extent_io_tree *block_group_cache; - struct btrfs_block_group_cache *block_group = NULL; - u64 ptr; - u64 start; - u64 end; - int ret; + struct btrfs_block_group_cache *cache; - bytenr = max_t(u64, bytenr, - BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); - block_group_cache = &info->block_group_cache; - ret = find_first_extent_bit(block_group_cache, - bytenr, &start, &end, - BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | - BLOCK_GROUP_SYSTEM); - if (ret) { - return NULL; - } - ret = get_state_private(block_group_cache, start, &ptr); - if (ret) - return NULL; + cache = block_group_cache_tree_search(info, bytenr, 0); - block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; - return block_group; + return cache; } +/* + * return the block group that contains teh given bytenr + */ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, u64 bytenr) { - struct extent_io_tree *block_group_cache; - struct btrfs_block_group_cache *block_group = NULL; - u64 ptr; - u64 start; - u64 end; - int ret; + struct btrfs_block_group_cache *cache; - bytenr = max_t(u64, bytenr, - BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); - block_group_cache = &info->block_group_cache; - ret = find_first_extent_bit(block_group_cache, - bytenr, &start, &end, - BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | - BLOCK_GROUP_SYSTEM); - if (ret) { - return NULL; - } - ret = get_state_private(block_group_cache, start, &ptr); - if (ret) - return NULL; + cache = block_group_cache_tree_search(info, bytenr, 1); - block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; - if (block_group->key.objectid <= bytenr && bytenr < - block_group->key.objectid + block_group->key.offset) - return block_group; - return NULL; + return cache; } -static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) -{ - return (cache->flags & bits) == bits; -} - -static int noinline find_search_start(struct btrfs_root *root, - struct btrfs_block_group_cache **cache_ret, - u64 *start_ret, u64 num, int data) +static int noinline find_free_space(struct btrfs_root *root, + struct btrfs_block_group_cache **cache_ret, + u64 *start_ret, u64 num, int data) { int ret; struct btrfs_block_group_cache *cache = *cache_ret; - struct extent_io_tree *free_space_cache; - struct extent_state *state; + struct btrfs_free_space *info = NULL; u64 last; - u64 start = 0; - u64 cache_miss = 0; u64 total_fs_bytes; u64 search_start = *start_ret; - int wrapped = 0; WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex)); total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); - free_space_cache = &root->fs_info->free_space_cache; if (!cache) goto out; + last = max(search_start, cache->key.objectid); + again: ret = cache_block_group(root, cache); - if (ret) { + if (ret) goto out; - } - last = max(search_start, cache->key.objectid); - if (!block_group_bits(cache, data) || cache->ro) + if (cache->ro || !block_group_bits(cache, data)) goto new_group; - spin_lock_irq(&free_space_cache->lock); - state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY); - while(1) { - if (!state) { - if (!cache_miss) - cache_miss = last; - spin_unlock_irq(&free_space_cache->lock); - goto new_group; - } - - start = max(last, state->start); - last = state->end + 1; - if (last - start < num) { - do { - state = extent_state_next(state); - } while(state && !(state->state & EXTENT_DIRTY)); - continue; - } - spin_unlock_irq(&free_space_cache->lock); - if (cache->ro) { - goto new_group; - } - if (start + num > cache->key.objectid + cache->key.offset) - goto new_group; - if (!block_group_bits(cache, data)) { - printk("block group bits don't match %Lu %d\n", cache->flags, data); - } - *start_ret = start; + info = btrfs_find_free_space(cache, last, num); + if (info) { + *start_ret = info->offset; return 0; } -out: - cache = btrfs_lookup_block_group(root->fs_info, search_start); - if (!cache) { - printk("Unable to find block group for %Lu\n", search_start); - WARN_ON(1); - } - return -ENOSPC; new_group: last = cache->key.objectid + cache->key.offset; -wrapped: + cache = btrfs_lookup_first_block_group(root->fs_info, last); - if (!cache || cache->key.objectid >= total_fs_bytes) { -no_cache: - if (!wrapped) { - wrapped = 1; - last = search_start; - goto wrapped; - } + if (!cache || cache->key.objectid >= total_fs_bytes) goto out; - } - if (cache_miss && !cache->cached) { - cache_block_group(root, cache); - last = cache_miss; - cache = btrfs_lookup_first_block_group(root->fs_info, last); - } - cache_miss = 0; - cache = btrfs_find_block_group(root, cache, last, data, 0); - if (!cache) - goto no_cache; + *cache_ret = cache; goto again; + +out: + return -ENOSPC; } static u64 div_factor(u64 num, int factor) @@ -338,16 +358,19 @@ static u64 div_factor(u64 num, int factor) return num; } -static int block_group_state_bits(u64 flags) +static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, + u64 flags) { - int bits = 0; - if (flags & BTRFS_BLOCK_GROUP_DATA) - bits |= BLOCK_GROUP_DATA; - if (flags & BTRFS_BLOCK_GROUP_METADATA) - bits |= BLOCK_GROUP_METADATA; - if (flags & BTRFS_BLOCK_GROUP_SYSTEM) - bits |= BLOCK_GROUP_SYSTEM; - return bits; + struct list_head *head = &info->space_info; + struct list_head *cur; + struct btrfs_space_info *found; + list_for_each(cur, head) { + found = list_entry(cur, struct btrfs_space_info, list); + if (found->flags == flags) + return found; + } + return NULL; + } static struct btrfs_block_group_cache * @@ -356,28 +379,19 @@ __btrfs_find_block_group(struct btrfs_root *root, u64 search_start, int data, int owner) { struct btrfs_block_group_cache *cache; - struct extent_io_tree *block_group_cache; struct btrfs_block_group_cache *found_group = NULL; struct btrfs_fs_info *info = root->fs_info; + struct btrfs_space_info *sinfo; u64 used; u64 last = 0; - u64 start; - u64 end; u64 free_check; - u64 ptr; - int bit; - int ret; int full_search = 0; int factor = 10; int wrapped = 0; - block_group_cache = &info->block_group_cache; - if (data & BTRFS_BLOCK_GROUP_METADATA) factor = 9; - bit = block_group_state_bits(data); - if (search_start) { struct btrfs_block_group_cache *shint; shint = btrfs_lookup_first_block_group(info, search_start); @@ -408,20 +422,30 @@ __btrfs_find_block_group(struct btrfs_root *root, else last = search_start; } + sinfo = __find_space_info(root->fs_info, data); + if (!sinfo) + goto found; again: while(1) { - ret = find_first_extent_bit(block_group_cache, last, - &start, &end, bit); - if (ret) - break; + struct list_head *l; - ret = get_state_private(block_group_cache, start, &ptr); - if (ret) { - last = end + 1; - continue; + cache = NULL; + + spin_lock(&sinfo->lock); + list_for_each(l, &sinfo->block_groups) { + struct btrfs_block_group_cache *entry; + entry = list_entry(l, struct btrfs_block_group_cache, + list); + if ((entry->key.objectid >= last) && + (!cache || (entry->key.objectid < + cache->key.objectid))) + cache = entry; } + spin_unlock(&sinfo->lock); + + if (!cache) + break; - cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; spin_lock(&cache->lock); last = cache->key.objectid + cache->key.offset; used = btrfs_block_group_used(&cache->item); @@ -462,6 +486,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, ret = __btrfs_find_block_group(root, hint, search_start, data, owner); return ret; } + static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset) { @@ -1175,34 +1200,37 @@ fail: int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - struct extent_io_tree *block_group_cache; - struct btrfs_block_group_cache *cache; - int ret; + struct btrfs_block_group_cache *cache, *entry; + struct rb_node *n; int err = 0; int werr = 0; struct btrfs_path *path; u64 last = 0; - u64 start; - u64 end; - u64 ptr; - block_group_cache = &root->fs_info->block_group_cache; path = btrfs_alloc_path(); if (!path) return -ENOMEM; mutex_lock(&root->fs_info->alloc_mutex); while(1) { - ret = find_first_extent_bit(block_group_cache, last, - &start, &end, BLOCK_GROUP_DIRTY); - if (ret) - break; + cache = NULL; + spin_lock(&root->fs_info->block_group_cache_lock); + for (n = rb_first(&root->fs_info->block_group_cache_tree); + n; n = rb_next(n)) { + entry = rb_entry(n, struct btrfs_block_group_cache, + cache_node); + if (entry->dirty) { + cache = entry; + break; + } + } + spin_unlock(&root->fs_info->block_group_cache_lock); - last = end + 1; - ret = get_state_private(block_group_cache, start, &ptr); - if (ret) + if (!cache) break; - cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; + + last += cache->key.offset; + err = write_one_cache_group(trans, root, path, cache); /* @@ -1214,29 +1242,14 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, werr = err; continue; } - clear_extent_bits(block_group_cache, start, end, - BLOCK_GROUP_DIRTY, GFP_NOFS); + + cache->dirty = 0; } btrfs_free_path(path); mutex_unlock(&root->fs_info->alloc_mutex); return werr; } -static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, - u64 flags) -{ - struct list_head *head = &info->space_info; - struct list_head *cur; - struct btrfs_space_info *found; - list_for_each(cur, head) { - found = list_entry(cur, struct btrfs_space_info, list); - if (found->flags == flags) - return found; - } - return NULL; - -} - static int update_space_info(struct btrfs_fs_info *info, u64 flags, u64 total_bytes, u64 bytes_used, struct btrfs_space_info **space_info) @@ -1256,6 +1269,8 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, return -ENOMEM; list_add(&found->list, &info->space_info); + INIT_LIST_HEAD(&found->block_groups); + spin_lock_init(&found->lock); found->flags = flags; found->total_bytes = total_bytes; found->bytes_used = bytes_used; @@ -1318,7 +1333,7 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, u64 thresh; u64 start; u64 num_bytes; - int ret; + int ret = 0; flags = reduce_alloc_profile(extent_root, flags); @@ -1355,10 +1370,11 @@ printk("space info full %Lu\n", flags); ret = btrfs_make_block_group(trans, extent_root, 0, flags, BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes); BUG_ON(ret); + out_unlock: mutex_unlock(&extent_root->fs_info->chunk_mutex); out: - return 0; + return ret; } static int update_block_group(struct btrfs_trans_handle *trans, @@ -1371,8 +1387,6 @@ static int update_block_group(struct btrfs_trans_handle *trans, u64 total = num_bytes; u64 old_val; u64 byte_in_group; - u64 start; - u64 end; WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex)); while(total) { @@ -1382,12 +1396,9 @@ static int update_block_group(struct btrfs_trans_handle *trans, } byte_in_group = bytenr - cache->key.objectid; WARN_ON(byte_in_group > cache->key.offset); - start = cache->key.objectid; - end = start + cache->key.offset - 1; - set_extent_bits(&info->block_group_cache, start, end, - BLOCK_GROUP_DIRTY, GFP_NOFS); spin_lock(&cache->lock); + cache->dirty = 1; old_val = btrfs_block_group_used(&cache->item); num_bytes = min(total, cache->key.offset - byte_in_group); if (alloc) { @@ -1401,9 +1412,11 @@ static int update_block_group(struct btrfs_trans_handle *trans, btrfs_set_block_group_used(&cache->item, old_val); spin_unlock(&cache->lock); if (mark_free) { - set_extent_dirty(&info->free_space_cache, - bytenr, bytenr + num_bytes - 1, - GFP_NOFS); + int ret; + ret = btrfs_add_free_space(cache, bytenr, + num_bytes); + if (ret) + return -1; } } total -= num_bytes; @@ -1414,16 +1427,13 @@ static int update_block_group(struct btrfs_trans_handle *trans, static u64 first_logical_byte(struct btrfs_root *root, u64 search_start) { - u64 start; - u64 end; - int ret; - ret = find_first_extent_bit(&root->fs_info->block_group_cache, - search_start, &start, &end, - BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | - BLOCK_GROUP_SYSTEM); - if (ret) + struct btrfs_block_group_cache *cache; + + cache = btrfs_lookup_first_block_group(root->fs_info, search_start); + if (!cache) return 0; - return start; + + return cache->key.objectid; } @@ -1501,8 +1511,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, u64 start; u64 end; int ret; - struct extent_io_tree *free_space_cache; - free_space_cache = &root->fs_info->free_space_cache; + struct btrfs_block_group_cache *cache; mutex_lock(&root->fs_info->alloc_mutex); while(1) { @@ -1512,7 +1521,9 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, break; btrfs_update_pinned_extents(root, start, end + 1 - start, 0); clear_extent_dirty(unpin, start, end, GFP_NOFS); - set_extent_dirty(free_space_cache, start, end, GFP_NOFS); + cache = btrfs_lookup_block_group(root->fs_info, start); + if (cache->cached) + btrfs_add_free_space(cache, start, end - start + 1); if (need_resched()) { mutex_unlock(&root->fs_info->alloc_mutex); cond_resched(); @@ -1875,9 +1886,12 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, /* if metadata always pin */ if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) { if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) { + struct btrfs_block_group_cache *cache; + /* btrfs_free_reserved_extent */ - set_extent_dirty(&root->fs_info->free_space_cache, - bytenr, bytenr + num_bytes - 1, GFP_NOFS); + cache = btrfs_lookup_block_group(root->fs_info, bytenr); + BUG_ON(!cache); + btrfs_add_free_space(cache, bytenr, num_bytes); return 0; } pin = 1; @@ -1942,8 +1956,6 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, u64 total_needed = num_bytes; u64 *last_ptr = NULL; struct btrfs_block_group_cache *block_group; - int full_scan = 0; - int wrapped = 0; int chunk_alloc_done = 0; int empty_cluster = 2 * 1024 * 1024; int allowed_chunk_alloc = 0; @@ -1959,9 +1971,9 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, empty_cluster = 256 * 1024; } - if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { + if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) last_ptr = &root->fs_info->last_data_alloc; - } + if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) { last_ptr = &root->fs_info->last_log_alloc; if (!last_ptr == 0 && root->fs_info->last_alloc) { @@ -1972,9 +1984,8 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, if (last_ptr) { if (*last_ptr) hint_byte = *last_ptr; - else { + else empty_size += empty_cluster; - } } search_start = max(search_start, first_logical_byte(root, 0)); @@ -1983,145 +1994,172 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, if (search_end == (u64)-1) search_end = btrfs_super_total_bytes(&info->super_copy); - if (hint_byte) { - block_group = btrfs_lookup_first_block_group(info, hint_byte); - if (!block_group) - hint_byte = search_start; - block_group = btrfs_find_block_group(root, block_group, - hint_byte, data, 1); - if (last_ptr && *last_ptr == 0 && block_group) - hint_byte = block_group->key.objectid; - } else { - block_group = btrfs_find_block_group(root, - trans->block_group, - search_start, data, 1); - } search_start = max(search_start, hint_byte); - total_needed += empty_size; -check_failed: - if (!block_group) { - block_group = btrfs_lookup_first_block_group(info, - search_start); - if (!block_group) - block_group = btrfs_lookup_first_block_group(info, - orig_search_start); - } - if (full_scan && !chunk_alloc_done) { - if (allowed_chunk_alloc) { - do_chunk_alloc(trans, root, - num_bytes + 2 * 1024 * 1024, data, 1); - allowed_chunk_alloc = 0; - } else if (block_group && block_group_bits(block_group, data)) { - block_group->space_info->force_alloc = 1; +new_group: + block_group = btrfs_lookup_block_group(info, search_start); + + /* + * Ok this looks a little tricky, buts its really simple. First if we + * didn't find a block group obviously we want to start over. + * Secondly, if the block group we found does not match the type we + * need, and we have a last_ptr and its not 0, chances are the last + * allocation we made was at the end of the block group, so lets go + * ahead and skip the looking through the rest of the block groups and + * start at the beginning. This helps with metadata allocations, + * since you are likely to have a bunch of data block groups to search + * through first before you realize that you need to start over, so go + * ahead and start over and save the time. + */ + if (!block_group || (!block_group_bits(block_group, data) && + last_ptr && *last_ptr)) { + if (search_start != orig_search_start) { + if (last_ptr && *last_ptr) + *last_ptr = 0; + search_start = orig_search_start; + goto new_group; + } else if (!chunk_alloc_done && allowed_chunk_alloc) { + ret = do_chunk_alloc(trans, root, + num_bytes + 2 * 1024 * 1024, + data, 1); + if (ret < 0) { + struct btrfs_space_info *info; + + info = __find_space_info(root->fs_info, data); + goto error; + } + BUG_ON(ret); + chunk_alloc_done = 1; + search_start = orig_search_start; + goto new_group; + } else { + ret = -ENOSPC; + goto error; } - chunk_alloc_done = 1; - } - ret = find_search_start(root, &block_group, &search_start, - total_needed, data); - if (ret == -ENOSPC && last_ptr && *last_ptr) { - *last_ptr = 0; - block_group = btrfs_lookup_first_block_group(info, - orig_search_start); - search_start = orig_search_start; - ret = find_search_start(root, &block_group, &search_start, - total_needed, data); } - if (ret == -ENOSPC) - goto enospc; - if (ret) - goto error; - if (last_ptr && *last_ptr && search_start != *last_ptr) { - *last_ptr = 0; - if (!empty_size) { - empty_size += empty_cluster; - total_needed += empty_size; + /* + * this is going to seach through all of the existing block groups it + * can find, so if we don't find something we need to see if we can + * allocate what we need. + */ + ret = find_free_space(root, &block_group, &search_start, + total_needed, data); + if (ret == -ENOSPC) { + /* + * instead of allocating, start at the original search start + * and see if there is something to be found, if not then we + * allocate + */ + if (search_start != orig_search_start) { + if (last_ptr && *last_ptr) { + *last_ptr = 0; + total_needed += empty_cluster; + } + search_start = orig_search_start; + goto new_group; } - block_group = btrfs_lookup_first_block_group(info, - orig_search_start); - search_start = orig_search_start; - ret = find_search_start(root, &block_group, - &search_start, total_needed, data); - if (ret == -ENOSPC) - goto enospc; - if (ret) + + /* + * we've already allocated, we're pretty screwed + */ + if (chunk_alloc_done) { goto error; + } else if (!allowed_chunk_alloc && block_group && + block_group_bits(block_group, data)) { + block_group->space_info->force_alloc = 1; + goto error; + } else if (!allowed_chunk_alloc) { + goto error; + } + + ret = do_chunk_alloc(trans, root, num_bytes + 2 * 1024 * 1024, + data, 1); + if (ret < 0) + goto error; + + BUG_ON(ret); + chunk_alloc_done = 1; + if (block_group) + search_start = block_group->key.objectid + + block_group->key.offset; + else + search_start = orig_search_start; + goto new_group; } + if (ret) + goto error; + search_start = stripe_align(root, search_start); ins->objectid = search_start; ins->offset = num_bytes; - if (ins->objectid + num_bytes >= search_end) - goto enospc; + if (ins->objectid + num_bytes >= search_end) { + search_start = orig_search_start; + if (chunk_alloc_done) { + ret = -ENOSPC; + goto error; + } + goto new_group; + } if (ins->objectid + num_bytes > block_group->key.objectid + block_group->key.offset) { + if (search_start == orig_search_start && chunk_alloc_done) { + ret = -ENOSPC; + goto error; + } search_start = block_group->key.objectid + block_group->key.offset; goto new_group; } - if (test_range_bit(&info->extent_ins, ins->objectid, - ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { - search_start = ins->objectid + num_bytes; - goto new_group; - } - - if (test_range_bit(&info->pinned_extents, ins->objectid, - ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { - search_start = ins->objectid + num_bytes; - goto new_group; - } - if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && ins->objectid < exclude_start + exclude_nr)) { search_start = exclude_start + exclude_nr; goto new_group; } - if (!(data & BTRFS_BLOCK_GROUP_DATA)) { - block_group = btrfs_lookup_block_group(info, ins->objectid); - if (block_group) - trans->block_group = block_group; - } + if (!(data & BTRFS_BLOCK_GROUP_DATA)) + trans->block_group = block_group; + ins->offset = num_bytes; if (last_ptr) { *last_ptr = ins->objectid + ins->offset; if (*last_ptr == - btrfs_super_total_bytes(&root->fs_info->super_copy)) { + btrfs_super_total_bytes(&root->fs_info->super_copy)) *last_ptr = 0; - } - } - return 0; - -new_group: - if (search_start + num_bytes >= search_end) { -enospc: - search_start = orig_search_start; - if (full_scan) { - ret = -ENOSPC; - goto error; - } - if (wrapped) { - if (!full_scan) - total_needed -= empty_size; - full_scan = 1; - } else - wrapped = 1; } - block_group = btrfs_lookup_first_block_group(info, search_start); - cond_resched(); - block_group = btrfs_find_block_group(root, block_group, - search_start, data, 0); - goto check_failed; + ret = 0; error: return ret; } +static void dump_space_info(struct btrfs_space_info *info, u64 bytes) +{ + struct btrfs_block_group_cache *cache; + struct list_head *l; + + printk(KERN_INFO "space_info has %Lu free, is %sfull\n", + info->total_bytes - info->bytes_used - info->bytes_pinned, + (info->full) ? "" : "not "); + + spin_lock(&info->lock); + list_for_each(l, &info->block_groups) { + cache = list_entry(l, struct btrfs_block_group_cache, list); + spin_lock(&cache->lock); + printk(KERN_INFO "block group %Lu has %Lu bytes, %Lu used " + "%Lu pinned\n", + cache->key.objectid, cache->key.offset, + btrfs_block_group_used(&cache->item), cache->pinned); + btrfs_dump_free_space(cache, bytes); + spin_unlock(&cache->lock); + } + spin_unlock(&info->lock); +} static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 min_alloc_size, @@ -2133,6 +2171,7 @@ static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans, u64 search_start = 0; u64 alloc_profile; struct btrfs_fs_info *info = root->fs_info; + struct btrfs_block_group_cache *cache; if (data) { alloc_profile = info->avail_data_alloc_bits & @@ -2160,11 +2199,9 @@ again: BTRFS_BLOCK_GROUP_METADATA | (info->metadata_alloc_profile & info->avail_metadata_alloc_bits), 0); - BUG_ON(ret); } ret = do_chunk_alloc(trans, root->fs_info->extent_root, num_bytes + 2 * 1024 * 1024, data, 0); - BUG_ON(ret); } WARN_ON(num_bytes < root->sectorsize); @@ -2175,26 +2212,44 @@ again: if (ret == -ENOSPC && num_bytes > min_alloc_size) { num_bytes = num_bytes >> 1; + num_bytes = num_bytes & ~(root->sectorsize - 1); num_bytes = max(num_bytes, min_alloc_size); do_chunk_alloc(trans, root->fs_info->extent_root, num_bytes, data, 1); goto again; } if (ret) { - printk("allocation failed flags %Lu\n", data); + struct btrfs_space_info *sinfo; + + sinfo = __find_space_info(root->fs_info, data); + printk("allocation failed flags %Lu, wanted %Lu\n", + data, num_bytes); + dump_space_info(sinfo, num_bytes); BUG(); } - clear_extent_dirty(&root->fs_info->free_space_cache, - ins->objectid, ins->objectid + ins->offset - 1, - GFP_NOFS); - return 0; + cache = btrfs_lookup_block_group(root->fs_info, ins->objectid); + if (!cache) { + printk(KERN_ERR "Unable to find block group for %Lu\n", ins->objectid); + return -ENOSPC; + } + + ret = btrfs_remove_free_space(cache, ins->objectid, ins->offset); + + return ret; } int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len) { + struct btrfs_block_group_cache *cache; + maybe_lock_mutex(root); - set_extent_dirty(&root->fs_info->free_space_cache, - start, start + len - 1, GFP_NOFS); + cache = btrfs_lookup_block_group(root->fs_info, start); + if (!cache) { + printk(KERN_ERR "Unable to find block group for %Lu\n", start); + maybe_unlock_mutex(root); + return -ENOSPC; + } + btrfs_add_free_space(cache, start, len); maybe_unlock_mutex(root); return 0; } @@ -2264,8 +2319,8 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, ret = btrfs_insert_empty_items(trans, extent_root, path, keys, sizes, 2); - BUG_ON(ret); + extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_extent_item); btrfs_set_extent_refs(path->nodes[0], extent_item, 1); @@ -2336,9 +2391,9 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid); cache_block_group(root, block_group); - clear_extent_dirty(&root->fs_info->free_space_cache, - ins->objectid, ins->objectid + ins->offset - 1, - GFP_NOFS); + ret = btrfs_remove_free_space(block_group, ins->objectid, ins->offset); + BUG_ON(ret); + ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, ref_generation, owner, owner_offset, ins); @@ -2843,31 +2898,24 @@ out: int btrfs_free_block_groups(struct btrfs_fs_info *info) { - u64 start; - u64 end; - u64 ptr; - int ret; + struct btrfs_block_group_cache *block_group; + struct rb_node *n; mutex_lock(&info->alloc_mutex); - while(1) { - ret = find_first_extent_bit(&info->block_group_cache, 0, - &start, &end, (unsigned int)-1); - if (ret) - break; - ret = get_state_private(&info->block_group_cache, start, &ptr); - if (!ret) - kfree((void *)(unsigned long)ptr); - clear_extent_bits(&info->block_group_cache, start, - end, (unsigned int)-1, GFP_NOFS); - } - while(1) { - ret = find_first_extent_bit(&info->free_space_cache, 0, - &start, &end, EXTENT_DIRTY); - if (ret) - break; - clear_extent_dirty(&info->free_space_cache, start, - end, GFP_NOFS); - } + spin_lock(&info->block_group_cache_lock); + while ((n = rb_last(&info->block_group_cache_tree)) != NULL) { + block_group = rb_entry(n, struct btrfs_block_group_cache, + cache_node); + + btrfs_remove_free_space_cache(block_group); + rb_erase(&block_group->cache_node, + &info->block_group_cache_tree); + spin_lock(&block_group->space_info->lock); + list_del(&block_group->list); + spin_unlock(&block_group->space_info->lock); + kfree(block_group); + } + spin_unlock(&info->block_group_cache_lock); mutex_unlock(&info->alloc_mutex); return 0; } @@ -3386,7 +3434,6 @@ int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) u64 total_found; u64 shrink_last_byte; struct btrfs_block_group_cache *shrink_block_group; - struct btrfs_fs_info *info = root->fs_info; struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; @@ -3542,15 +3589,17 @@ next: goto out; } - clear_extent_bits(&info->block_group_cache, key.objectid, - key.objectid + key.offset - 1, - (unsigned int)-1, GFP_NOFS); - - - clear_extent_bits(&info->free_space_cache, - key.objectid, key.objectid + key.offset - 1, - (unsigned int)-1, GFP_NOFS); + spin_lock(&root->fs_info->block_group_cache_lock); + rb_erase(&shrink_block_group->cache_node, + &root->fs_info->block_group_cache_tree); + spin_unlock(&root->fs_info->block_group_cache_lock); + ret = btrfs_remove_free_space(shrink_block_group, key.objectid, + key.offset); + if (ret) { + btrfs_end_transaction(trans, root); + goto out; + } /* memset(shrink_block_group, 0, sizeof(*shrink_block_group)); kfree(shrink_block_group); @@ -3566,9 +3615,9 @@ next: /* the code to unpin extents might set a few bits in the free * space cache for this range again */ - clear_extent_bits(&info->free_space_cache, - key.objectid, key.objectid + key.offset - 1, - (unsigned int)-1, GFP_NOFS); + /* XXX? */ + ret = btrfs_remove_free_space(shrink_block_group, key.objectid, + key.offset); out: btrfs_free_path(path); mutex_unlock(&root->fs_info->alloc_mutex); @@ -3616,16 +3665,13 @@ int btrfs_read_block_groups(struct btrfs_root *root) { struct btrfs_path *path; int ret; - int bit; struct btrfs_block_group_cache *cache; struct btrfs_fs_info *info = root->fs_info; struct btrfs_space_info *space_info; - struct extent_io_tree *block_group_cache; struct btrfs_key key; struct btrfs_key found_key; struct extent_buffer *leaf; - block_group_cache = &info->block_group_cache; root = info->extent_root; key.objectid = 0; key.offset = 0; @@ -3653,6 +3699,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) } spin_lock_init(&cache->lock); + INIT_LIST_HEAD(&cache->list); read_extent_buffer(leaf, &cache->item, btrfs_item_ptr_offset(leaf, path->slots[0]), sizeof(cache->item)); @@ -3661,31 +3708,19 @@ int btrfs_read_block_groups(struct btrfs_root *root) key.objectid = found_key.objectid + found_key.offset; btrfs_release_path(root, path); cache->flags = btrfs_block_group_flags(&cache->item); - bit = 0; - if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { - bit = BLOCK_GROUP_DATA; - } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { - bit = BLOCK_GROUP_SYSTEM; - } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { - bit = BLOCK_GROUP_METADATA; - } - set_avail_alloc_bits(info, cache->flags); ret = update_space_info(info, cache->flags, found_key.offset, btrfs_block_group_used(&cache->item), &space_info); BUG_ON(ret); cache->space_info = space_info; + spin_lock(&space_info->lock); + list_add(&cache->list, &space_info->block_groups); + spin_unlock(&space_info->lock); + + ret = btrfs_add_block_group_cache(root->fs_info, cache); + BUG_ON(ret); - /* use EXTENT_LOCKED to prevent merging */ - set_extent_bits(block_group_cache, found_key.objectid, - found_key.objectid + found_key.offset - 1, - EXTENT_LOCKED, GFP_NOFS); - set_state_private(block_group_cache, found_key.objectid, - (unsigned long)cache); - set_extent_bits(block_group_cache, found_key.objectid, - found_key.objectid + found_key.offset - 1, - bit | EXTENT_LOCKED, GFP_NOFS); if (key.objectid >= btrfs_super_total_bytes(&info->super_copy)) break; @@ -3703,22 +3738,22 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 size) { int ret; - int bit = 0; struct btrfs_root *extent_root; struct btrfs_block_group_cache *cache; - struct extent_io_tree *block_group_cache; WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex)); extent_root = root->fs_info->extent_root; - block_group_cache = &root->fs_info->block_group_cache; root->fs_info->last_trans_new_blockgroup = trans->transid; cache = kzalloc(sizeof(*cache), GFP_NOFS); - BUG_ON(!cache); + if (!cache) + return -ENOMEM; + cache->key.objectid = chunk_offset; cache->key.offset = size; spin_lock_init(&cache->lock); + INIT_LIST_HEAD(&cache->list); btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); btrfs_set_block_group_used(&cache->item, bytes_used); @@ -3729,16 +3764,12 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, &cache->space_info); BUG_ON(ret); + spin_lock(&cache->space_info->lock); + list_add(&cache->list, &cache->space_info->block_groups); + spin_unlock(&cache->space_info->lock); - bit = block_group_state_bits(type); - set_extent_bits(block_group_cache, chunk_offset, - chunk_offset + size - 1, - EXTENT_LOCKED, GFP_NOFS); - set_state_private(block_group_cache, chunk_offset, - (unsigned long)cache); - set_extent_bits(block_group_cache, chunk_offset, - chunk_offset + size - 1, - bit | EXTENT_LOCKED, GFP_NOFS); + ret = btrfs_add_block_group_cache(root->fs_info, cache); + BUG_ON(ret); ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, sizeof(cache->item)); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 319a0c7a4a5..8624f3e8803 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2634,6 +2634,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, if (eb) { atomic_inc(&eb->refs); spin_unlock(&tree->buffer_lock); + mark_page_accessed(eb->first_page); return eb; } spin_unlock(&tree->buffer_lock); @@ -2713,6 +2714,9 @@ struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, atomic_inc(&eb->refs); spin_unlock(&tree->buffer_lock); + if (eb) + mark_page_accessed(eb->first_page); + return eb; } EXPORT_SYMBOL(find_extent_buffer); diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c new file mode 100644 index 00000000000..01c26e8ae55 --- /dev/null +++ b/fs/btrfs/free-space-cache.c @@ -0,0 +1,415 @@ +/* + * Copyright (C) 2008 Red Hat. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include "ctree.h" + +static int tree_insert_offset(struct rb_root *root, u64 offset, + struct rb_node *node) +{ + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; + struct btrfs_free_space *info; + + while (*p) { + parent = *p; + info = rb_entry(parent, struct btrfs_free_space, offset_index); + + if (offset < info->offset) + p = &(*p)->rb_left; + else if (offset > info->offset) + p = &(*p)->rb_right; + else + return -EEXIST; + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); + + return 0; +} + +static int tree_insert_bytes(struct rb_root *root, u64 bytes, + struct rb_node *node) +{ + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; + struct btrfs_free_space *info; + + while (*p) { + parent = *p; + info = rb_entry(parent, struct btrfs_free_space, bytes_index); + + if (bytes < info->bytes) + p = &(*p)->rb_left; + else + p = &(*p)->rb_right; + } + + rb_link_node(node, parent, p); + rb_insert_color(node, root); + + return 0; +} + +/* + * searches the tree for the given offset. If contains is set we will return + * the free space that contains the given offset. If contains is not set we + * will return the free space that starts at or after the given offset and is + * at least bytes long. + */ +static struct btrfs_free_space *tree_search_offset(struct rb_root *root, + u64 offset, u64 bytes, + int contains) +{ + struct rb_node *n = root->rb_node; + struct btrfs_free_space *entry, *ret = NULL; + + while (n) { + entry = rb_entry(n, struct btrfs_free_space, offset_index); + + if (offset < entry->offset) { + if (!contains && + (!ret || entry->offset < ret->offset) && + (bytes <= entry->bytes)) + ret = entry; + n = n->rb_left; + } else if (offset > entry->offset) { + if (contains && + (entry->offset + entry->bytes - 1) >= offset) { + ret = entry; + break; + } + n = n->rb_right; + } else { + if (bytes > entry->bytes) { + n = n->rb_right; + continue; + } + ret = entry; + break; + } + } + + return ret; +} + +/* + * return a chunk at least bytes size, as close to offset that we can get. + */ +static struct btrfs_free_space *tree_search_bytes(struct rb_root *root, + u64 offset, u64 bytes) +{ + struct rb_node *n = root->rb_node; + struct btrfs_free_space *entry, *ret = NULL; + + while (n) { + entry = rb_entry(n, struct btrfs_free_space, bytes_index); + + if (bytes < entry->bytes) { + /* + * We prefer to get a hole size as close to the size we + * are asking for so we don't take small slivers out of + * huge holes, but we also want to get as close to the + * offset as possible so we don't have a whole lot of + * fragmentation. + */ + if (offset <= entry->offset) { + if (!ret) + ret = entry; + else if (entry->bytes < ret->bytes) + ret = entry; + else if (entry->offset < ret->offset) + ret = entry; + } + n = n->rb_left; + } else if (bytes > entry->bytes) { + n = n->rb_right; + } else { + /* + * Ok we may have multiple chunks of the wanted size, + * so we don't want to take the first one we find, we + * want to take the one closest to our given offset, so + * keep searching just in case theres a better match. + */ + n = n->rb_right; + if (offset > entry->offset) + continue; + else if (!ret || entry->offset < ret->offset) + ret = entry; + } + } + + return ret; +} + +static void unlink_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *info) +{ + rb_erase(&info->offset_index, &block_group->free_space_offset); + rb_erase(&info->bytes_index, &block_group->free_space_bytes); +} + +static int link_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *info) +{ + int ret = 0; + + + ret = tree_insert_offset(&block_group->free_space_offset, info->offset, + &info->offset_index); + if (ret) + return ret; + + ret = tree_insert_bytes(&block_group->free_space_bytes, info->bytes, + &info->bytes_index); + if (ret) + return ret; + + return ret; +} + +int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, + u64 offset, u64 bytes) +{ + struct btrfs_free_space *right_info; + struct btrfs_free_space *left_info; + struct btrfs_free_space *info = NULL; + struct btrfs_free_space *alloc_info; + int ret = 0; + + alloc_info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS); + if (!alloc_info) + return -ENOMEM; + + /* + * first we want to see if there is free space adjacent to the range we + * are adding, if there is remove that struct and add a new one to + * cover the entire range + */ + spin_lock(&block_group->lock); + + right_info = tree_search_offset(&block_group->free_space_offset, + offset+bytes, 0, 1); + left_info = tree_search_offset(&block_group->free_space_offset, + offset-1, 0, 1); + + if (right_info && right_info->offset == offset+bytes) { + unlink_free_space(block_group, right_info); + info = right_info; + info->offset = offset; + info->bytes += bytes; + } else if (right_info && right_info->offset != offset+bytes) { + printk(KERN_ERR "adding space in the middle of an existing " + "free space area. existing: offset=%Lu, bytes=%Lu. " + "new: offset=%Lu, bytes=%Lu\n", right_info->offset, + right_info->bytes, offset, bytes); + BUG(); + } + + if (left_info) { + unlink_free_space(block_group, left_info); + + if (unlikely((left_info->offset + left_info->bytes) != + offset)) { + printk(KERN_ERR "free space to the left of new free " + "space isn't quite right. existing: offset=%Lu," + " bytes=%Lu. new: offset=%Lu, bytes=%Lu\n", + left_info->offset, left_info->bytes, offset, + bytes); + BUG(); + } + + if (info) { + info->offset = left_info->offset; + info->bytes += left_info->bytes; + kfree(left_info); + } else { + info = left_info; + info->bytes += bytes; + } + } + + if (info) { + ret = link_free_space(block_group, info); + if (!ret) + info = NULL; + goto out; + } + + info = alloc_info; + alloc_info = NULL; + info->offset = offset; + info->bytes = bytes; + + ret = link_free_space(block_group, info); + if (ret) + kfree(info); +out: + spin_unlock(&block_group->lock); + if (ret) { + printk(KERN_ERR "btrfs: unable to add free space :%d\n", ret); + if (ret == -EEXIST) + BUG(); + } + + if (alloc_info) + kfree(alloc_info); + + return ret; +} + +int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group, + u64 offset, u64 bytes) +{ + struct btrfs_free_space *info; + int ret = 0; + + spin_lock(&block_group->lock); + info = tree_search_offset(&block_group->free_space_offset, offset, 0, + 1); + + if (info && info->offset == offset) { + if (info->bytes < bytes) { + printk(KERN_ERR "Found free space at %Lu, size %Lu," + "trying to use %Lu\n", + info->offset, info->bytes, bytes); + WARN_ON(1); + ret = -EINVAL; + goto out; + } + + unlink_free_space(block_group, info); + + if (info->bytes == bytes) { + kfree(info); + goto out; + } + + info->offset += bytes; + info->bytes -= bytes; + + ret = link_free_space(block_group, info); + BUG_ON(ret); + } else { + WARN_ON(1); + } +out: + spin_unlock(&block_group->lock); + return ret; +} + +void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group, + u64 bytes) +{ + struct btrfs_free_space *info; + struct rb_node *n; + int count = 0; + + for (n = rb_first(&block_group->free_space_offset); n; n = rb_next(n)) { + info = rb_entry(n, struct btrfs_free_space, offset_index); + if (info->bytes >= bytes) + count++; + //printk(KERN_INFO "offset=%Lu, bytes=%Lu\n", info->offset, + // info->bytes); + } + printk(KERN_INFO "%d blocks of free space at or bigger than bytes is" + "\n", count); +} + +u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group) +{ + struct btrfs_free_space *info; + struct rb_node *n; + u64 ret = 0; + + for (n = rb_first(&block_group->free_space_offset); n; + n = rb_next(n)) { + info = rb_entry(n, struct btrfs_free_space, offset_index); + ret += info->bytes; + } + + return ret; +} + +void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group) +{ + struct btrfs_free_space *info; + struct rb_node *node; + + spin_lock(&block_group->lock); + while ((node = rb_last(&block_group->free_space_bytes)) != NULL) { + info = rb_entry(node, struct btrfs_free_space, bytes_index); + unlink_free_space(block_group, info); + kfree(info); + if (need_resched()) { + spin_unlock(&block_group->lock); + cond_resched(); + spin_lock(&block_group->lock); + } + } + spin_unlock(&block_group->lock); +} + +struct btrfs_free_space *btrfs_find_free_space_offset(struct + btrfs_block_group_cache + *block_group, u64 offset, + u64 bytes) +{ + struct btrfs_free_space *ret; + + spin_lock(&block_group->lock); + ret = tree_search_offset(&block_group->free_space_offset, offset, + bytes, 0); + spin_unlock(&block_group->lock); + + return ret; +} + +struct btrfs_free_space *btrfs_find_free_space_bytes(struct + btrfs_block_group_cache + *block_group, u64 offset, + u64 bytes) +{ + struct btrfs_free_space *ret; + + spin_lock(&block_group->lock); + + ret = tree_search_bytes(&block_group->free_space_bytes, offset, bytes); + spin_unlock(&block_group->lock); + + return ret; +} + +struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache + *block_group, u64 offset, + u64 bytes) +{ + struct btrfs_free_space *ret; + + spin_lock(&block_group->lock); + ret = tree_search_offset(&block_group->free_space_offset, offset, + bytes, 0); + if (!ret) + ret = tree_search_bytes(&block_group->free_space_bytes, + offset, bytes); + + spin_unlock(&block_group->lock); + + return ret; +} diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 12c1c0530f3..65b4f864b0d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -141,7 +141,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) while(num_bytes > 0) { cur_alloc_size = min(num_bytes, root->fs_info->max_extent); ret = btrfs_reserve_extent(trans, root, cur_alloc_size, - root->sectorsize, 0, 0, + root->sectorsize, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { WARN_ON(1); @@ -558,7 +558,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) trans->transid, inode->i_ino, ordered_extent->file_offset, &ins); BUG_ON(ret); - mutex_lock(&BTRFS_I(inode)->extent_mutex); ret = btrfs_drop_extents(trans, root, inode, diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 1546fa6f4f7..b9e5c2d82dd 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -64,8 +64,8 @@ static void lock_chunks(struct btrfs_root *root) static void unlock_chunks(struct btrfs_root *root) { - mutex_unlock(&root->fs_info->alloc_mutex); mutex_unlock(&root->fs_info->chunk_mutex); + mutex_unlock(&root->fs_info->alloc_mutex); } int btrfs_cleanup_fs_uuids(void) @@ -1668,8 +1668,13 @@ again: else min_free = calc_size; - /* we add 1MB because we never use the first 1MB of the device */ - min_free += 1024 * 1024; + /* + * we add 1MB because we never use the first 1MB of the device, unless + * we've looped, then we are likely allocating the maximum amount of + * space left already + */ + if (!looped) + min_free += 1024 * 1024; /* build a private list of devices we will allocate from */ while(index < num_stripes) { -- cgit v1.2.3-70-g09d2 From 31840ae1a6b433ca0e6a8d341756ff478bbf959e Mon Sep 17 00:00:00 2001 From: Zheng Yan Date: Tue, 23 Sep 2008 13:14:14 -0400 Subject: Btrfs: Full back reference support This patch makes the back reference system to explicit record the location of parent node for all types of extents. The location of parent node is placed into the offset field of backref key. Every time a tree block is balanced, the back references for the affected lower level extents are updated. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 228 ++++++---- fs/btrfs/ctree.h | 62 ++- fs/btrfs/disk-io.c | 4 +- fs/btrfs/extent-tree.c | 1143 +++++++++++++++++++++++++++++------------------- fs/btrfs/extent_io.c | 3 +- fs/btrfs/file.c | 121 +++-- fs/btrfs/inode.c | 57 ++- fs/btrfs/ioctl.c | 57 +-- fs/btrfs/print-tree.c | 5 +- fs/btrfs/tree-log.c | 108 ++--- 10 files changed, 1066 insertions(+), 722 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 6f467901246..50aea8cb653 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -125,7 +125,6 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, u32 nritems; int ret = 0; int level; - struct btrfs_key first_key; struct btrfs_root *new_root; new_root = kmalloc(sizeof(*new_root), GFP_NOFS); @@ -141,18 +140,10 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); - if (nritems) { - if (level == 0) - btrfs_item_key_to_cpu(buf, &first_key, 0); - else - btrfs_node_key_to_cpu(buf, &first_key, 0); - } else { - first_key.objectid = 0; - } - cow = btrfs_alloc_free_block(trans, new_root, buf->len, - new_root_objectid, - trans->transid, first_key.objectid, - level, buf->start, 0); + + cow = btrfs_alloc_free_block(trans, new_root, buf->len, 0, + new_root_objectid, trans->transid, + level, buf->start, 0); if (IS_ERR(cow)) { kfree(new_root); return PTR_ERR(cow); @@ -165,7 +156,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN); WARN_ON(btrfs_header_generation(buf) > trans->transid); - ret = btrfs_inc_ref(trans, new_root, buf, 0); + ret = btrfs_inc_ref(trans, new_root, buf, cow, NULL); kfree(new_root); if (ret) @@ -184,39 +175,31 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, u64 search_start, u64 empty_size, u64 prealloc_dest) { - u64 root_gen; + u64 parent_start; struct extent_buffer *cow; u32 nritems; int ret = 0; int different_trans = 0; int level; int unlock_orig = 0; - struct btrfs_key first_key; if (*cow_ret == buf) unlock_orig = 1; WARN_ON(!btrfs_tree_locked(buf)); - if (root->ref_cows) { - root_gen = trans->transid; - } else { - root_gen = 0; - } + if (parent) + parent_start = parent->start; + else + parent_start = 0; + WARN_ON(root->ref_cows && trans->transid != root->fs_info->running_transaction->transid); WARN_ON(root->ref_cows && trans->transid != root->last_trans); level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); - if (nritems) { - if (level == 0) - btrfs_item_key_to_cpu(buf, &first_key, 0); - else - btrfs_node_key_to_cpu(buf, &first_key, 0); - } else { - first_key.objectid = 0; - } + if (prealloc_dest) { struct btrfs_key ins; @@ -224,19 +207,19 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, ins.offset = buf->len; ins.type = BTRFS_EXTENT_ITEM_KEY; - ret = btrfs_alloc_reserved_extent(trans, root, + ret = btrfs_alloc_reserved_extent(trans, root, parent_start, root->root_key.objectid, - root_gen, level, - first_key.objectid, + trans->transid, level, 0, &ins); BUG_ON(ret); cow = btrfs_init_new_buffer(trans, root, prealloc_dest, buf->len); } else { cow = btrfs_alloc_free_block(trans, root, buf->len, + parent_start, root->root_key.objectid, - root_gen, first_key.objectid, - level, search_start, empty_size); + trans->transid, level, + search_start, empty_size); } if (IS_ERR(cow)) return PTR_ERR(cow); @@ -249,17 +232,23 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, WARN_ON(btrfs_header_generation(buf) > trans->transid); if (btrfs_header_generation(buf) != trans->transid) { + u32 nr_extents; different_trans = 1; - ret = btrfs_inc_ref(trans, root, buf, 1); + ret = btrfs_inc_ref(trans, root, buf, cow, &nr_extents); if (ret) return ret; + + ret = btrfs_cache_ref(trans, root, buf, nr_extents); + WARN_ON(ret); } else { + ret = btrfs_update_ref(trans, root, buf, cow, 0, nritems); + if (ret) + return ret; clean_tree_block(trans, root, buf); } if (buf == root->node) { WARN_ON(parent && parent != buf); - root_gen = btrfs_header_generation(buf); spin_lock(&root->node_lock); root->node = cow; @@ -268,13 +257,14 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, if (buf != root->commit_root) { btrfs_free_extent(trans, root, buf->start, - buf->len, root->root_key.objectid, - root_gen, 0, 0, 1); + buf->len, buf->start, + root->root_key.objectid, + btrfs_header_generation(buf), + 0, 0, 1); } free_extent_buffer(buf); add_root_to_dirty_list(root); } else { - root_gen = btrfs_header_generation(parent); btrfs_set_node_blockptr(parent, parent_slot, cow->start); WARN_ON(trans->transid == 0); @@ -283,8 +273,8 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, btrfs_mark_buffer_dirty(parent); WARN_ON(btrfs_header_generation(parent) != trans->transid); btrfs_free_extent(trans, root, buf->start, buf->len, - btrfs_header_owner(parent), root_gen, - 0, 0, 1); + parent_start, btrfs_header_owner(parent), + btrfs_header_generation(parent), 0, 0, 1); } if (unlock_orig) btrfs_tree_unlock(buf); @@ -831,6 +821,12 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, root->node = child; spin_unlock(&root->node_lock); + ret = btrfs_update_extent_ref(trans, root, child->start, + mid->start, child->start, + root->root_key.objectid, + trans->transid, level - 1, 0); + BUG_ON(ret); + add_root_to_dirty_list(root); btrfs_tree_unlock(child); path->locks[level] = 0; @@ -840,7 +836,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, /* once for the path */ free_extent_buffer(mid); ret = btrfs_free_extent(trans, root, mid->start, mid->len, - root->root_key.objectid, + mid->start, root->root_key.objectid, btrfs_header_generation(mid), 0, 0, 1); /* once for the root ptr */ free_extent_buffer(mid); @@ -905,7 +901,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, if (wret) ret = wret; wret = btrfs_free_extent(trans, root, bytenr, - blocksize, + blocksize, parent->start, btrfs_header_owner(parent), generation, 0, 0, 1); if (wret) @@ -954,6 +950,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, if (wret) ret = wret; wret = btrfs_free_extent(trans, root, bytenr, blocksize, + parent->start, btrfs_header_owner(parent), root_gen, 0, 0, 1); if (wret) @@ -1499,6 +1496,41 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, return ret; } +/* + * update item key. + * + * This function isn't completely safe. It's the caller's responsibility + * that the new key won't break the order + */ +int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_path *path, + struct btrfs_key *new_key) +{ + struct btrfs_disk_key disk_key; + struct extent_buffer *eb; + int slot; + + eb = path->nodes[0]; + slot = path->slots[0]; + if (slot > 0) { + btrfs_item_key(eb, &disk_key, slot - 1); + if (comp_keys(&disk_key, new_key) >= 0) + return -1; + } + if (slot < btrfs_header_nritems(eb) - 1) { + btrfs_item_key(eb, &disk_key, slot + 1); + if (comp_keys(&disk_key, new_key) <= 0) + return -1; + } + + btrfs_cpu_key_to_disk(&disk_key, new_key); + btrfs_set_item_key(eb, &disk_key, slot); + btrfs_mark_buffer_dirty(eb); + if (slot == 0) + fixup_low_keys(trans, root, path, &disk_key, 1); + return 0; +} + /* * try to push data from one node into the next node left in the * tree. @@ -1558,6 +1590,10 @@ static int push_node_left(struct btrfs_trans_handle *trans, btrfs_set_header_nritems(dst, dst_nritems + push_items); btrfs_mark_buffer_dirty(src); btrfs_mark_buffer_dirty(dst); + + ret = btrfs_update_ref(trans, root, src, dst, dst_nritems, push_items); + BUG_ON(ret); + return ret; } @@ -1619,6 +1655,10 @@ static int balance_node_right(struct btrfs_trans_handle *trans, btrfs_mark_buffer_dirty(src); btrfs_mark_buffer_dirty(dst); + + ret = btrfs_update_ref(trans, root, src, dst, 0, push_items); + BUG_ON(ret); + return ret; } @@ -1633,30 +1673,24 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { - u64 root_gen; u64 lower_gen; struct extent_buffer *lower; struct extent_buffer *c; struct extent_buffer *old; struct btrfs_disk_key lower_key; + int ret; BUG_ON(path->nodes[level]); BUG_ON(path->nodes[level-1] != root->node); - if (root->ref_cows) - root_gen = trans->transid; - else - root_gen = 0; - lower = path->nodes[level-1]; if (level == 1) btrfs_item_key(lower, &lower_key, 0); else btrfs_node_key(lower, &lower_key, 0); - c = btrfs_alloc_free_block(trans, root, root->nodesize, - root->root_key.objectid, - root_gen, le64_to_cpu(lower_key.objectid), + c = btrfs_alloc_free_block(trans, root, root->nodesize, 0, + root->root_key.objectid, trans->transid, level, root->node->start, 0); if (IS_ERR(c)) return PTR_ERR(c); @@ -1679,7 +1713,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, btrfs_set_node_key(c, &lower_key, 0); btrfs_set_node_blockptr(c, 0, lower->start); lower_gen = btrfs_header_generation(lower); - WARN_ON(lower_gen == 0); + WARN_ON(lower_gen != trans->transid); btrfs_set_node_ptr_generation(c, 0, lower_gen); @@ -1690,6 +1724,12 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, root->node = c; spin_unlock(&root->node_lock); + ret = btrfs_update_extent_ref(trans, root, lower->start, + lower->start, c->start, + root->root_key.objectid, + trans->transid, level - 1, 0); + BUG_ON(ret); + /* the super has an extra ref to root->node */ free_extent_buffer(old); @@ -1698,20 +1738,6 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, path->nodes[level] = c; path->locks[level] = 1; path->slots[level] = 0; - - if (root->ref_cows && lower_gen != trans->transid) { - struct btrfs_path *back_path = btrfs_alloc_path(); - int ret; - mutex_lock(&root->fs_info->alloc_mutex); - ret = btrfs_insert_extent_backref(trans, - root->fs_info->extent_root, - path, lower->start, - root->root_key.objectid, - trans->transid, 0, 0); - BUG_ON(ret); - mutex_unlock(&root->fs_info->alloc_mutex); - btrfs_free_path(back_path); - } return 0; } @@ -1766,7 +1792,6 @@ static noinline int split_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { - u64 root_gen; struct extent_buffer *c; struct extent_buffer *split; struct btrfs_disk_key disk_key; @@ -1793,17 +1818,11 @@ static noinline int split_node(struct btrfs_trans_handle *trans, } c_nritems = btrfs_header_nritems(c); - if (root->ref_cows) - root_gen = trans->transid; - else - root_gen = 0; - btrfs_node_key(c, &disk_key, 0); split = btrfs_alloc_free_block(trans, root, root->nodesize, - root->root_key.objectid, - root_gen, - btrfs_disk_key_objectid(&disk_key), - level, c->start, 0); + path->nodes[level + 1]->start, + root->root_key.objectid, + trans->transid, level, c->start, 0); if (IS_ERR(split)) return PTR_ERR(split); @@ -1840,6 +1859,9 @@ static noinline int split_node(struct btrfs_trans_handle *trans, if (wret) ret = wret; + ret = btrfs_update_ref(trans, root, c, split, 0, c_nritems - mid); + BUG_ON(ret); + if (path->slots[level] >= mid) { path->slots[level] -= mid; btrfs_tree_unlock(c); @@ -1955,10 +1977,23 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root else nr = 1; + if (path->slots[0] >= left_nritems) + push_space += data_size + sizeof(*item); + i = left_nritems - 1; while (i >= nr) { item = btrfs_item_nr(left, i); + if (!empty && push_items > 0) { + if (path->slots[0] > i) + break; + if (path->slots[0] == i) { + int space = btrfs_leaf_free_space(root, left); + if (space + push_space * 2 > free_space) + break; + } + } + if (path->slots[0] == i) push_space += data_size + sizeof(*item); @@ -1973,6 +2008,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root this_item_size = btrfs_item_size(left, item); if (this_item_size + sizeof(*item) + push_space > free_space) break; + push_items++; push_space += this_item_size + sizeof(*item); if (i == 0) @@ -2046,6 +2082,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_mark_buffer_dirty(left); btrfs_mark_buffer_dirty(right); + ret = btrfs_update_ref(trans, root, left, right, 0, push_items); + BUG_ON(ret); + btrfs_item_key(right, &disk_key, 0); btrfs_set_node_key(upper, &disk_key, slot + 1); btrfs_mark_buffer_dirty(upper); @@ -2147,6 +2186,16 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root KM_USER1); } + if (!empty && push_items > 0) { + if (path->slots[0] < i) + break; + if (path->slots[0] == i) { + int space = btrfs_leaf_free_space(root, right); + if (space + push_space * 2 > free_space) + break; + } + } + if (path->slots[0] == i) push_space += data_size + sizeof(*item); @@ -2255,6 +2304,10 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root if (right_nritems) btrfs_mark_buffer_dirty(right); + ret = btrfs_update_ref(trans, root, right, left, + old_left_nritems, push_items); + BUG_ON(ret); + btrfs_item_key(right, &disk_key, 0); wret = fixup_low_keys(trans, root, path, &disk_key, 1); if (wret) @@ -2294,7 +2347,6 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_path *path, int data_size, int extend) { - u64 root_gen; struct extent_buffer *l; u32 nritems; int mid; @@ -2313,11 +2365,6 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans, if (extend) space_needed = data_size; - if (root->ref_cows) - root_gen = trans->transid; - else - root_gen = 0; - /* first try to make some room by pushing left and right */ if (ins_key->type != BTRFS_DIR_ITEM_KEY) { wret = push_leaf_right(trans, root, path, data_size, 0); @@ -2348,13 +2395,10 @@ again: nritems = btrfs_header_nritems(l); mid = (nritems + 1)/ 2; - btrfs_item_key(l, &disk_key, 0); - right = btrfs_alloc_free_block(trans, root, root->leafsize, - root->root_key.objectid, - root_gen, - le64_to_cpu(disk_key.objectid), - 0, l->start, 0); + path->nodes[1]->start, + root->root_key.objectid, + trans->transid, 0, l->start, 0); if (IS_ERR(right)) { BUG_ON(1); return PTR_ERR(right); @@ -2485,6 +2529,9 @@ again: btrfs_mark_buffer_dirty(l); BUG_ON(path->slots[0] != slot); + ret = btrfs_update_ref(trans, root, l, right, 0, nritems); + BUG_ON(ret); + if (mid <= slot) { btrfs_tree_unlock(path->nodes[0]); free_extent_buffer(path->nodes[0]); @@ -2956,6 +3003,7 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, ret = wret; wret = btrfs_free_extent(trans, root, leaf->start, leaf->len, + path->nodes[1]->start, btrfs_header_owner(path->nodes[1]), root_gen, 0, 0, 1); if (wret) @@ -3007,7 +3055,7 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, free_extent_buffer(leaf); wret = btrfs_free_extent(trans, root, bytenr, - blocksize, + blocksize, path->nodes[1]->start, btrfs_header_owner(path->nodes[1]), root_gen, 0, 0, 1); if (wret) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 730aae3bc18..138c157bbc4 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -40,7 +40,7 @@ extern struct kmem_cache *btrfs_bit_radix_cachep; extern struct kmem_cache *btrfs_path_cachep; struct btrfs_ordered_sum; -#define BTRFS_MAGIC "_B8RfS_M" +#define BTRFS_MAGIC "_B9RfS_M" #define BTRFS_ACL_NOT_CACHED ((void *)-1) @@ -81,6 +81,9 @@ struct btrfs_ordered_sum; #define BTRFS_TREE_LOG_OBJECTID -6ULL #define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL +/* dummy objectid represents multiple objectids */ +#define BTRFS_MULTIPLE_OBJECTIDS -255ULL + /* * All files have objectids in this range. */ @@ -369,6 +372,7 @@ struct btrfs_extent_ref { __le64 generation; __le64 objectid; __le64 offset; + __le32 num_refs; } __attribute__ ((__packed__)); /* dev extents record free space on individual devices. The owner @@ -1047,9 +1051,6 @@ btrfs_inode_otime(struct btrfs_inode_item *inode_item) BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64); BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32); -/* struct btrfs_extent_item */ -BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32); - /* struct btrfs_dev_extent */ BTRFS_SETGET_FUNCS(dev_extent_chunk_tree, struct btrfs_dev_extent, chunk_tree, 64); @@ -1070,14 +1071,20 @@ BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64); BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64); BTRFS_SETGET_FUNCS(ref_objectid, struct btrfs_extent_ref, objectid, 64); BTRFS_SETGET_FUNCS(ref_offset, struct btrfs_extent_ref, offset, 64); +BTRFS_SETGET_FUNCS(ref_num_refs, struct btrfs_extent_ref, num_refs, 32); BTRFS_SETGET_STACK_FUNCS(stack_ref_root, struct btrfs_extent_ref, root, 64); BTRFS_SETGET_STACK_FUNCS(stack_ref_generation, struct btrfs_extent_ref, generation, 64); BTRFS_SETGET_STACK_FUNCS(stack_ref_objectid, struct btrfs_extent_ref, objectid, 64); -BTRFS_SETGET_STACK_FUNCS(stack_ref_offset, struct btrfs_extent_ref, offset, 64); +BTRFS_SETGET_STACK_FUNCS(stack_ref_offset, struct btrfs_extent_ref, + offset, 64); +BTRFS_SETGET_STACK_FUNCS(stack_ref_num_refs, struct btrfs_extent_ref, + num_refs, 32); +/* struct btrfs_extent_item */ +BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32); BTRFS_SETGET_STACK_FUNCS(stack_extent_refs, struct btrfs_extent_item, refs, 32); @@ -1474,8 +1481,7 @@ static inline struct dentry *fdentry(struct file *file) { } /* extent-tree.c */ -int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path, - u64 start, u64 len); +int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len); int btrfs_update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin); int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, @@ -1495,10 +1501,9 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, int data, int owner); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u32 blocksize, + u32 blocksize, u64 parent, u64 root_objectid, u64 ref_generation, - u64 first_objectid, int level, u64 hint, u64 empty_size); @@ -1508,23 +1513,24 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size); int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct btrfs_path *path, u64 bytenr, + struct btrfs_path *path, + u64 bytenr, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 num_bytes, u64 min_bytes, + u64 num_bytes, u64 parent, u64 min_bytes, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, u64 data); int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins); int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins); @@ -1535,9 +1541,16 @@ int btrfs_reserve_extent(struct btrfs_trans_handle *trans, u64 search_end, struct btrfs_key *ins, u64 data); int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int cache_ref); -int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, + struct extent_buffer *orig_buf, struct extent_buffer *buf, + u32 *nr_extents); +int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct extent_buffer *buf, u32 nr_extents); +int btrfs_update_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *orig_buf, + struct extent_buffer *buf, int start_slot, int nr); +int btrfs_free_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner_objectid, u64 owner_offset, int pin); int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len); @@ -1545,10 +1558,15 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_io_tree *unpin); int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 bytenr, u64 num_bytes, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset); + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset); +int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 orig_parent, u64 parent, + u64 root_objectid, u64 ref_generation, + u64 owner, u64 owner_offset); int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_free_block_groups(struct btrfs_fs_info *info); @@ -1561,7 +1579,9 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, int btrfs_previous_item(struct btrfs_root *root, struct btrfs_path *path, u64 min_objectid, int type); - +int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_path *path, + struct btrfs_key *new_key); struct extent_buffer *btrfs_root_node(struct btrfs_root *root); struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root); int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 25be96946a2..d35ca6a3f51 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -882,8 +882,8 @@ int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, root->ref_cows = 0; root->node = btrfs_alloc_free_block(trans, root, root->leafsize, - BTRFS_TREE_LOG_OBJECTID, - 0, 0, 0, 0, 0); + 0, BTRFS_TREE_LOG_OBJECTID, + trans->transid, 0, 0, 0); btrfs_set_header_nritems(root->node, 0); btrfs_set_header_level(root->node, 0); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 558fbe40736..5258923d621 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -29,6 +29,21 @@ #include "locking.h" #include "ref-cache.h" +#define PENDING_EXTENT_INSERT 0 +#define PENDING_EXTENT_DELETE 1 +#define PENDING_BACKREF_UPDATE 2 + +struct pending_extent_op { + int type; + u64 bytenr; + u64 num_bytes; + u64 parent; + u64 orig_parent; + u64 generation; + u64 orig_generation; + int level; +}; + static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root); static int del_pending_extents(struct btrfs_trans_handle *trans, struct @@ -487,48 +502,15 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, return ret; } -static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset) -{ - u32 high_crc = ~(u32)0; - u32 low_crc = ~(u32)0; - __le64 lenum; - lenum = cpu_to_le64(root_objectid); - high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum)); - lenum = cpu_to_le64(ref_generation); - low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); - if (owner >= BTRFS_FIRST_FREE_OBJECTID) { - lenum = cpu_to_le64(owner); - low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); - lenum = cpu_to_le64(owner_offset); - low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); - } - return ((u64)high_crc << 32) | (u64)low_crc; -} - -static int match_extent_ref(struct extent_buffer *leaf, - struct btrfs_extent_ref *disk_ref, - struct btrfs_extent_ref *cpu_ref) -{ - int ret; - int len; - - if (cpu_ref->objectid) - len = sizeof(*cpu_ref); - else - len = 2 * sizeof(u64); - ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, - len); - return ret == 0; -} - /* simple helper to search for an existing extent at a given offset */ -int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path, - u64 start, u64 len) +int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len) { int ret; struct btrfs_key key; + struct btrfs_path *path; + path = btrfs_alloc_path(); + BUG_ON(!path); maybe_lock_mutex(root); key.objectid = start; key.offset = len; @@ -536,72 +518,7 @@ int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path, ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path, 0, 0); maybe_unlock_mutex(root); - return ret; -} - -static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, u64 bytenr, - u64 root_objectid, - u64 ref_generation, u64 owner, - u64 owner_offset, int del) -{ - u64 hash; - struct btrfs_key key; - struct btrfs_key found_key; - struct btrfs_extent_ref ref; - struct extent_buffer *leaf; - struct btrfs_extent_ref *disk_ref; - int ret; - int ret2; - - btrfs_set_stack_ref_root(&ref, root_objectid); - btrfs_set_stack_ref_generation(&ref, ref_generation); - btrfs_set_stack_ref_objectid(&ref, owner); - btrfs_set_stack_ref_offset(&ref, owner_offset); - - hash = hash_extent_ref(root_objectid, ref_generation, owner, - owner_offset); - key.offset = hash; - key.objectid = bytenr; - key.type = BTRFS_EXTENT_REF_KEY; - - while (1) { - ret = btrfs_search_slot(trans, root, &key, path, - del ? -1 : 0, del); - if (ret < 0) - goto out; - leaf = path->nodes[0]; - if (ret != 0) { - u32 nritems = btrfs_header_nritems(leaf); - if (path->slots[0] >= nritems) { - ret2 = btrfs_next_leaf(root, path); - if (ret2) - goto out; - leaf = path->nodes[0]; - } - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (found_key.objectid != bytenr || - found_key.type != BTRFS_EXTENT_REF_KEY) - goto out; - key.offset = found_key.offset; - if (del) { - btrfs_release_path(root, path); - continue; - } - } - disk_ref = btrfs_item_ptr(path->nodes[0], - path->slots[0], - struct btrfs_extent_ref); - if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { - ret = 0; - goto out; - } - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - key.offset = found_key.offset + 1; - btrfs_release_path(root, path); - } -out: + btrfs_free_path(path); return ret; } @@ -622,7 +539,7 @@ out: * File extents can be referenced by: * * - multiple snapshots, subvolumes, or different generations in one subvol - * - different files inside a single subvolume (in theory, not implemented yet) + * - different files inside a single subvolume * - different offsets inside a file (bookend extents in file.c) * * The extent ref structure has fields for: @@ -631,119 +548,284 @@ out: * - Generation number of the tree holding the reference * - objectid of the file holding the reference * - offset in the file corresponding to the key holding the reference + * - number of references holding by parent node (alway 1 for tree blocks) + * + * Btree leaf may hold multiple references to a file extent. In most cases, + * these references are from same file and the corresponding offsets inside + * the file are close together. So inode objectid and offset in file are + * just hints, they provide hints about where in the btree the references + * can be found and when we can stop searching. * * When a file extent is allocated the fields are filled in: - * (root_key.objectid, trans->transid, inode objectid, offset in file) + * (root_key.objectid, trans->transid, inode objectid, offset in file, 1) * * When a leaf is cow'd new references are added for every file extent found - * in the leaf. It looks the same as the create case, but trans->transid - * will be different when the block is cow'd. + * in the leaf. It looks similar to the create case, but trans->transid will + * be different when the block is cow'd. * - * (root_key.objectid, trans->transid, inode objectid, offset in file) + * (root_key.objectid, trans->transid, inode objectid, offset in file, + * number of references in the leaf) * - * When a file extent is removed either during snapshot deletion or file - * truncation, the corresponding back reference is found - * by searching for: + * Because inode objectid and offset in file are just hints, they are not + * used when backrefs are deleted. When a file extent is removed either + * during snapshot deletion or file truncation, we find the corresponding + * back back reference and check the following fields. * - * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), - * inode objectid, offset in file) + * (btrfs_header_owner(leaf), btrfs_header_generation(leaf)) * * Btree extents can be referenced by: * * - Different subvolumes * - Different generations of the same subvolume * - * Storing sufficient information for a full reverse mapping of a btree - * block would require storing the lowest key of the block in the backref, - * and it would require updating that lowest key either before write out or - * every time it changed. Instead, the objectid of the lowest key is stored - * along with the level of the tree block. This provides a hint - * about where in the btree the block can be found. Searches through the - * btree only need to look for a pointer to that block, so they stop one - * level higher than the level recorded in the backref. - * - * Some btrees do not do reference counting on their extents. These - * include the extent tree and the tree of tree roots. Backrefs for these - * trees always have a generation of zero. - * * When a tree block is created, back references are inserted: * - * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) + * (root->root_key.objectid, trans->transid, level, 0, 1) * - * When a tree block is cow'd in a reference counted root, - * new back references are added for all the blocks it points to. - * These are of the form (trans->transid will have increased since creation): + * When a tree block is cow'd, new back references are added for all the + * blocks it points to. If the tree block isn't in reference counted root, + * the old back references are removed. These new back references are of + * the form (trans->transid will have increased since creation): * - * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) + * (root->root_key.objectid, trans->transid, level, 0, 1) * - * Because the lowest_key_objectid and the level are just hints - * they are not used when backrefs are deleted. When a backref is deleted: + * When a backref is in deleting, the following fields are checked: * * if backref was for a tree root: - * root_objectid = root->root_key.objectid + * (btrfs_header_owner(itself), btrfs_header_generation(itself)) * else - * root_objectid = btrfs_header_owner(parent) + * (btrfs_header_owner(parent), btrfs_header_generation(parent)) * - * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) + * Back Reference Key composing: * - * Back Reference Key hashing: - * - * Back references have four fields, each 64 bits long. Unfortunately, - * This is hashed into a single 64 bit number and placed into the key offset. - * The key objectid corresponds to the first byte in the extent, and the - * key type is set to BTRFS_EXTENT_REF_KEY + * The key objectid corresponds to the first byte in the extent, the key + * type is set to BTRFS_EXTENT_REF_KEY, and the key offset is the first + * byte of parent extent. If a extent is tree root, the key offset is set + * to the key objectid. */ -int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, u64 bytenr, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset) + +static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr, + u64 parent, u64 ref_root, + u64 ref_generation, int del) { - u64 hash; struct btrfs_key key; - struct btrfs_extent_ref ref; - struct btrfs_extent_ref *disk_ref; + struct btrfs_extent_ref *ref; + struct extent_buffer *leaf; int ret; - btrfs_set_stack_ref_root(&ref, root_objectid); - btrfs_set_stack_ref_generation(&ref, ref_generation); - btrfs_set_stack_ref_objectid(&ref, owner); - btrfs_set_stack_ref_offset(&ref, owner_offset); + key.objectid = bytenr; + key.type = BTRFS_EXTENT_REF_KEY; + key.offset = parent; + + ret = btrfs_search_slot(trans, root, &key, path, del ? -1 : 0, 1); + if (ret < 0) + goto out; + if (ret > 0) { + ret = -ENOENT; + goto out; + } + + leaf = path->nodes[0]; + ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); + if (btrfs_ref_root(leaf, ref) != ref_root || + btrfs_ref_generation(leaf, ref) != ref_generation) { + ret = -EIO; + WARN_ON(1); + goto out; + } + ret = 0; +out: + return ret; +} + +static int noinline insert_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + u64 bytenr, u64 parent, + u64 ref_root, u64 ref_generation, + u64 owner_objectid, u64 owner_offset) +{ + struct btrfs_key key; + struct extent_buffer *leaf; + struct btrfs_extent_ref *ref; + u32 num_refs; + int ret; - hash = hash_extent_ref(root_objectid, ref_generation, owner, - owner_offset); - key.offset = hash; key.objectid = bytenr; key.type = BTRFS_EXTENT_REF_KEY; + key.offset = parent; - ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); - while (ret == -EEXIST) { - disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], - struct btrfs_extent_ref); - if (match_extent_ref(path->nodes[0], disk_ref, &ref)) + ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*ref)); + if (ret == 0) { + leaf = path->nodes[0]; + ref = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_extent_ref); + btrfs_set_ref_root(leaf, ref, ref_root); + btrfs_set_ref_generation(leaf, ref, ref_generation); + btrfs_set_ref_objectid(leaf, ref, owner_objectid); + btrfs_set_ref_offset(leaf, ref, owner_offset); + btrfs_set_ref_num_refs(leaf, ref, 1); + } else if (ret == -EEXIST) { + u64 existing_owner; + BUG_ON(owner_objectid < BTRFS_FIRST_FREE_OBJECTID); + leaf = path->nodes[0]; + ref = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_extent_ref); + if (btrfs_ref_root(leaf, ref) != ref_root || + btrfs_ref_generation(leaf, ref) != ref_generation) { + ret = -EIO; + WARN_ON(1); goto out; - key.offset++; - btrfs_release_path(root, path); - ret = btrfs_insert_empty_item(trans, root, path, &key, - sizeof(ref)); - } - if (ret) + } + + num_refs = btrfs_ref_num_refs(leaf, ref); + BUG_ON(num_refs == 0); + btrfs_set_ref_num_refs(leaf, ref, num_refs + 1); + + existing_owner = btrfs_ref_objectid(leaf, ref); + if (existing_owner == owner_objectid && + btrfs_ref_offset(leaf, ref) > owner_offset) { + btrfs_set_ref_offset(leaf, ref, owner_offset); + } else if (existing_owner != owner_objectid && + existing_owner != BTRFS_MULTIPLE_OBJECTIDS) { + btrfs_set_ref_objectid(leaf, ref, + BTRFS_MULTIPLE_OBJECTIDS); + btrfs_set_ref_offset(leaf, ref, 0); + } + ret = 0; + } else { goto out; - disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], - struct btrfs_extent_ref); - write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, - sizeof(ref)); + } btrfs_mark_buffer_dirty(path->nodes[0]); out: btrfs_release_path(root, path); return ret; } +static int noinline remove_extent_backref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path) +{ + struct extent_buffer *leaf; + struct btrfs_extent_ref *ref; + u32 num_refs; + int ret = 0; + + leaf = path->nodes[0]; + ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); + num_refs = btrfs_ref_num_refs(leaf, ref); + BUG_ON(num_refs == 0); + num_refs -= 1; + if (num_refs == 0) { + ret = btrfs_del_item(trans, root, path); + } else { + btrfs_set_ref_num_refs(leaf, ref, num_refs); + btrfs_mark_buffer_dirty(leaf); + } + btrfs_release_path(root, path); + return ret; +} + +static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 orig_parent, u64 parent, + u64 orig_root, u64 ref_root, + u64 orig_generation, u64 ref_generation, + u64 owner_objectid, u64 owner_offset) +{ + int ret; + struct btrfs_root *extent_root = root->fs_info->extent_root; + struct btrfs_path *path; + + if (root == root->fs_info->extent_root) { + struct pending_extent_op *extent_op; + u64 num_bytes; + + BUG_ON(owner_objectid >= BTRFS_MAX_LEVEL); + num_bytes = btrfs_level_size(root, (int)owner_objectid); + if (test_range_bit(&root->fs_info->extent_ins, bytenr, + bytenr + num_bytes - 1, EXTENT_LOCKED, 0)) { + u64 priv; + ret = get_state_private(&root->fs_info->extent_ins, + bytenr, &priv); + BUG_ON(ret); + extent_op = (struct pending_extent_op *) + (unsigned long)priv; + BUG_ON(extent_op->parent != orig_parent); + BUG_ON(extent_op->generation != orig_generation); + extent_op->parent = parent; + extent_op->generation = ref_generation; + } else { + extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS); + BUG_ON(!extent_op); + + extent_op->type = PENDING_BACKREF_UPDATE; + extent_op->bytenr = bytenr; + extent_op->num_bytes = num_bytes; + extent_op->parent = parent; + extent_op->orig_parent = orig_parent; + extent_op->generation = ref_generation; + extent_op->orig_generation = orig_generation; + extent_op->level = (int)owner_objectid; + + set_extent_bits(&root->fs_info->extent_ins, + bytenr, bytenr + num_bytes - 1, + EXTENT_LOCKED, GFP_NOFS); + set_state_private(&root->fs_info->extent_ins, + bytenr, (unsigned long)extent_op); + } + return 0; + } + + path = btrfs_alloc_path(); + if (!path) + return -ENOMEM; + ret = lookup_extent_backref(trans, extent_root, path, + bytenr, orig_parent, orig_root, + orig_generation, 1); + if (ret) + goto out; + ret = remove_extent_backref(trans, extent_root, path); + if (ret) + goto out; + ret = insert_extent_backref(trans, extent_root, path, bytenr, + parent, ref_root, ref_generation, + owner_objectid, owner_offset); + BUG_ON(ret); + finish_current_insert(trans, extent_root); + del_pending_extents(trans, extent_root); +out: + btrfs_free_path(path); + return ret; +} + +int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 orig_parent, u64 parent, + u64 ref_root, u64 ref_generation, + u64 owner_objectid, u64 owner_offset) +{ + int ret; + if (ref_root == BTRFS_TREE_LOG_OBJECTID && + owner_objectid < BTRFS_FIRST_FREE_OBJECTID) + return 0; + maybe_lock_mutex(root); + ret = __btrfs_update_extent_ref(trans, root, bytenr, orig_parent, + parent, ref_root, ref_root, + ref_generation, ref_generation, + owner_objectid, owner_offset); + maybe_unlock_mutex(root); + return ret; +} + static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 bytenr, u64 num_bytes, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset) + struct btrfs_root *root, u64 bytenr, + u64 orig_parent, u64 parent, + u64 orig_root, u64 ref_root, + u64 orig_generation, u64 ref_generation, + u64 owner_objectid, u64 owner_offset) { struct btrfs_path *path; int ret; @@ -752,24 +834,28 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_extent_item *item; u32 refs; - WARN_ON(num_bytes < root->sectorsize); path = btrfs_alloc_path(); if (!path) return -ENOMEM; path->reada = 1; key.objectid = bytenr; - btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); - key.offset = num_bytes; + key.type = BTRFS_EXTENT_ITEM_KEY; + key.offset = (u64)-1; + ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 0, 1); if (ret < 0) return ret; - if (ret != 0) { - BUG(); - } - BUG_ON(ret != 0); + BUG_ON(ret == 0 || path->slots[0] == 0); + + path->slots[0]--; l = path->nodes[0]; + + btrfs_item_key_to_cpu(l, &key, path->slots[0]); + BUG_ON(key.objectid != bytenr); + BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY); + item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); refs = btrfs_extent_refs(l, item); btrfs_set_extent_refs(l, item, refs + 1); @@ -778,9 +864,10 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, btrfs_release_path(root->fs_info->extent_root, path); path->reada = 1; - ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, - path, bytenr, root_objectid, - ref_generation, owner, owner_offset); + ret = insert_extent_backref(trans, root->fs_info->extent_root, + path, bytenr, parent, + ref_root, ref_generation, + owner_objectid, owner_offset); BUG_ON(ret); finish_current_insert(trans, root->fs_info->extent_root); del_pending_extents(trans, root->fs_info->extent_root); @@ -790,18 +877,20 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, } int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 bytenr, u64 num_bytes, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset) + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, + u64 ref_root, u64 ref_generation, + u64 owner_objectid, u64 owner_offset) { int ret; - - mutex_lock(&root->fs_info->alloc_mutex); - ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, - root_objectid, ref_generation, - owner, owner_offset); - mutex_unlock(&root->fs_info->alloc_mutex); + if (ref_root == BTRFS_TREE_LOG_OBJECTID && + owner_objectid < BTRFS_FIRST_FREE_OBJECTID) + return 0; + maybe_lock_mutex(root); + ret = __btrfs_inc_extent_ref(trans, root, bytenr, 0, parent, + 0, ref_root, 0, ref_generation, + owner_objectid, owner_offset); + maybe_unlock_mutex(root); return ret; } @@ -813,9 +902,9 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans, return 0; } -static int lookup_extent_ref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr, - u64 num_bytes, u32 *refs) +int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 num_bytes, u32 *refs) { struct btrfs_path *path; int ret; @@ -846,7 +935,6 @@ out: return 0; } - static int get_reference_status(struct btrfs_root *root, u64 bytenr, u64 parent_gen, u64 ref_objectid, u64 *min_generation, u32 *ref_count) @@ -863,7 +951,7 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, int ret; key.objectid = bytenr; - key.offset = 0; + key.offset = (u64)-1; key.type = BTRFS_EXTENT_ITEM_KEY; path = btrfs_alloc_path(); @@ -872,7 +960,10 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, if (ret < 0) goto out; BUG_ON(ret == 0); + if (ret < 0 || path->slots[0] == 0) + goto out; + path->slots[0]--; leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); @@ -909,7 +1000,7 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, struct btrfs_extent_ref); ref_generation = btrfs_ref_generation(leaf, ref_item); /* - * For (parent_gen > 0 && parent_gen > ref_gen): + * For (parent_gen > 0 && parent_gen > ref_generation): * * we reach here through the oldest root, therefore * all other reference from same snapshot should have @@ -919,8 +1010,7 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, (parent_gen > 0 && parent_gen > ref_generation) || (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID && ref_objectid != btrfs_ref_objectid(leaf, ref_item))) { - if (ref_count) - *ref_count = 2; + *ref_count = 2; break; } @@ -1020,80 +1110,29 @@ out: return ret; } -int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct extent_buffer *buf, int cache_ref) +int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct extent_buffer *buf, u32 nr_extents) { - u64 bytenr; u32 nritems; struct btrfs_key key; struct btrfs_file_extent_item *fi; int i; int level; - int ret; - int faili; - int nr_file_extents = 0; + int ret = 0; if (!root->ref_cows) return 0; level = btrfs_header_level(buf); nritems = btrfs_header_nritems(buf); - for (i = 0; i < nritems; i++) { - cond_resched(); - if (level == 0) { - u64 disk_bytenr; - btrfs_item_key_to_cpu(buf, &key, i); - if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) - continue; - fi = btrfs_item_ptr(buf, i, - struct btrfs_file_extent_item); - if (btrfs_file_extent_type(buf, fi) == - BTRFS_FILE_EXTENT_INLINE) - continue; - disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); - if (disk_bytenr == 0) - continue; - - if (buf != root->commit_root) - nr_file_extents++; - - mutex_lock(&root->fs_info->alloc_mutex); - ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr, - btrfs_file_extent_disk_num_bytes(buf, fi), - root->root_key.objectid, trans->transid, - key.objectid, key.offset); - mutex_unlock(&root->fs_info->alloc_mutex); - if (ret) { - faili = i; - WARN_ON(1); - goto fail; - } - } else { - bytenr = btrfs_node_blockptr(buf, i); - btrfs_node_key_to_cpu(buf, &key, i); - mutex_lock(&root->fs_info->alloc_mutex); - ret = __btrfs_inc_extent_ref(trans, root, bytenr, - btrfs_level_size(root, level - 1), - root->root_key.objectid, - trans->transid, - level - 1, key.objectid); - mutex_unlock(&root->fs_info->alloc_mutex); - if (ret) { - faili = i; - WARN_ON(1); - goto fail; - } - } - } - /* cache orignal leaf block's references */ - if (level == 0 && cache_ref && buf != root->commit_root) { + if (level == 0) { struct btrfs_leaf_ref *ref; struct btrfs_extent_info *info; - ref = btrfs_alloc_leaf_ref(root, nr_file_extents); + ref = btrfs_alloc_leaf_ref(root, nr_extents); if (!ref) { - WARN_ON(1); + ret = -ENOMEM; goto out; } @@ -1101,10 +1140,10 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, ref->bytenr = buf->start; ref->owner = btrfs_header_owner(buf); ref->generation = btrfs_header_generation(buf); - ref->nritems = nr_file_extents; + ref->nritems = nr_extents; info = ref->extents; - for (i = 0; nr_file_extents > 0 && i < nritems; i++) { + for (i = 0; nr_extents > 0 && i < nritems; i++) { u64 disk_bytenr; btrfs_item_key_to_cpu(buf, &key, i); if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) @@ -1132,13 +1171,52 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, btrfs_free_leaf_ref(root, ref); } out: - return 0; -fail: - WARN_ON(1); -#if 0 - for (i =0; i < faili; i++) { + return ret; +} + +int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, + struct extent_buffer *orig_buf, struct extent_buffer *buf, + u32 *nr_extents) +{ + u64 bytenr; + u64 ref_root; + u64 orig_root; + u64 ref_generation; + u64 orig_generation; + u32 nritems; + u32 nr_file_extents = 0; + struct btrfs_key key; + struct btrfs_file_extent_item *fi; + int i; + int level; + int ret = 0; + int faili = 0; + int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *, + u64, u64, u64, u64, u64, u64, u64, u64, u64); + + ref_root = btrfs_header_owner(buf); + ref_generation = btrfs_header_generation(buf); + orig_root = btrfs_header_owner(orig_buf); + orig_generation = btrfs_header_generation(orig_buf); + + nritems = btrfs_header_nritems(buf); + level = btrfs_header_level(buf); + + if (root->ref_cows) { + process_func = __btrfs_inc_extent_ref; + } else { + if (level == 0 && + root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) + goto out; + if (level != 0 && + root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) + goto out; + process_func = __btrfs_update_extent_ref; + } + + for (i = 0; i < nritems; i++) { + cond_resched(); if (level == 0) { - u64 disk_bytenr; btrfs_item_key_to_cpu(buf, &key, i); if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) continue; @@ -1147,24 +1225,131 @@ fail: if (btrfs_file_extent_type(buf, fi) == BTRFS_FILE_EXTENT_INLINE) continue; - disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); - if (disk_bytenr == 0) + bytenr = btrfs_file_extent_disk_bytenr(buf, fi); + if (bytenr == 0) continue; - err = btrfs_free_extent(trans, root, disk_bytenr, - btrfs_file_extent_disk_num_bytes(buf, - fi), 0); - BUG_ON(err); + + nr_file_extents++; + + maybe_lock_mutex(root); + ret = process_func(trans, root, bytenr, + orig_buf->start, buf->start, + orig_root, ref_root, + orig_generation, ref_generation, + key.objectid, key.offset); + maybe_unlock_mutex(root); + + if (ret) { + faili = i; + WARN_ON(1); + goto fail; + } } else { bytenr = btrfs_node_blockptr(buf, i); - err = btrfs_free_extent(trans, root, bytenr, - btrfs_level_size(root, level - 1), 0); - BUG_ON(err); + maybe_lock_mutex(root); + ret = process_func(trans, root, bytenr, + orig_buf->start, buf->start, + orig_root, ref_root, + orig_generation, ref_generation, + level - 1, 0); + maybe_unlock_mutex(root); + if (ret) { + faili = i; + WARN_ON(1); + goto fail; + } } } -#endif +out: + if (nr_extents) { + if (level == 0) + *nr_extents = nr_file_extents; + else + *nr_extents = nritems; + } + return 0; +fail: + WARN_ON(1); return ret; } +int btrfs_update_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct extent_buffer *orig_buf, + struct extent_buffer *buf, int start_slot, int nr) + +{ + u64 bytenr; + u64 ref_root; + u64 orig_root; + u64 ref_generation; + u64 orig_generation; + struct btrfs_key key; + struct btrfs_file_extent_item *fi; + int i; + int ret; + int slot; + int level; + + BUG_ON(start_slot < 0); + BUG_ON(start_slot + nr > btrfs_header_nritems(buf)); + + ref_root = btrfs_header_owner(buf); + ref_generation = btrfs_header_generation(buf); + orig_root = btrfs_header_owner(orig_buf); + orig_generation = btrfs_header_generation(orig_buf); + level = btrfs_header_level(buf); + + if (!root->ref_cows) { + if (level == 0 && + root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) + return 0; + if (level != 0 && + root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) + return 0; + } + + for (i = 0, slot = start_slot; i < nr; i++, slot++) { + cond_resched(); + if (level == 0) { + btrfs_item_key_to_cpu(buf, &key, slot); + if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) + continue; + fi = btrfs_item_ptr(buf, slot, + struct btrfs_file_extent_item); + if (btrfs_file_extent_type(buf, fi) == + BTRFS_FILE_EXTENT_INLINE) + continue; + bytenr = btrfs_file_extent_disk_bytenr(buf, fi); + if (bytenr == 0) + continue; + maybe_lock_mutex(root); + ret = __btrfs_update_extent_ref(trans, root, bytenr, + orig_buf->start, buf->start, + orig_root, ref_root, + orig_generation, ref_generation, + key.objectid, key.offset); + maybe_unlock_mutex(root); + if (ret) + goto fail; + } else { + bytenr = btrfs_node_blockptr(buf, slot); + maybe_lock_mutex(root); + ret = __btrfs_update_extent_ref(trans, root, bytenr, + orig_buf->start, buf->start, + orig_root, ref_root, + orig_generation, ref_generation, + level - 1, 0); + maybe_unlock_mutex(root); + if (ret) + goto fail; + } + } + return 0; +fail: + WARN_ON(1); + return -1; +} + static int write_one_cache_group(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, @@ -1539,19 +1724,18 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, { u64 start; u64 end; + u64 priv; struct btrfs_fs_info *info = extent_root->fs_info; - struct extent_buffer *eb; struct btrfs_path *path; - struct btrfs_key ins; - struct btrfs_disk_key first; + struct btrfs_extent_ref *ref; + struct pending_extent_op *extent_op; + struct btrfs_key key; struct btrfs_extent_item extent_item; int ret; - int level; int err = 0; WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex)); btrfs_set_stack_extent_refs(&extent_item, 1); - btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); path = btrfs_alloc_path(); while(1) { @@ -1560,37 +1744,54 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, if (ret) break; - ins.objectid = start; - ins.offset = end + 1 - start; - err = btrfs_insert_item(trans, extent_root, &ins, + ret = get_state_private(&info->extent_ins, start, &priv); + BUG_ON(ret); + extent_op = (struct pending_extent_op *)(unsigned long)priv; + + if (extent_op->type == PENDING_EXTENT_INSERT) { + key.objectid = start; + key.offset = end + 1 - start; + key.type = BTRFS_EXTENT_ITEM_KEY; + err = btrfs_insert_item(trans, extent_root, &key, &extent_item, sizeof(extent_item)); - clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, - GFP_NOFS); + BUG_ON(err); - eb = btrfs_find_create_tree_block(extent_root, ins.objectid, - ins.offset); + clear_extent_bits(&info->extent_ins, start, end, + EXTENT_LOCKED, GFP_NOFS); - if (!btrfs_buffer_uptodate(eb, trans->transid)) - btrfs_read_buffer(eb, trans->transid); + err = insert_extent_backref(trans, extent_root, path, + start, extent_op->parent, + extent_root->root_key.objectid, + extent_op->generation, + extent_op->level, 0); + BUG_ON(err); + } else if (extent_op->type == PENDING_BACKREF_UPDATE) { + err = lookup_extent_backref(trans, extent_root, path, + start, extent_op->orig_parent, + extent_root->root_key.objectid, + extent_op->orig_generation, 0); + BUG_ON(err); - btrfs_tree_lock(eb); - level = btrfs_header_level(eb); - if (level == 0) { - btrfs_item_key(eb, &first, 0); + clear_extent_bits(&info->extent_ins, start, end, + EXTENT_LOCKED, GFP_NOFS); + + key.objectid = start; + key.offset = extent_op->parent; + key.type = BTRFS_EXTENT_REF_KEY; + err = btrfs_set_item_key_safe(trans, extent_root, path, + &key); + BUG_ON(err); + ref = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_extent_ref); + btrfs_set_ref_generation(path->nodes[0], ref, + extent_op->generation); + btrfs_mark_buffer_dirty(path->nodes[0]); + btrfs_release_path(extent_root, path); } else { - btrfs_node_key(eb, &first, 0); + BUG_ON(1); } - btrfs_tree_unlock(eb); - free_extent_buffer(eb); - /* - * the first key is just a hint, so the race we've created - * against reading it is fine - */ - err = btrfs_insert_extent_backref(trans, extent_root, path, - start, extent_root->root_key.objectid, - 0, level, - btrfs_disk_key_objectid(&first)); - BUG_ON(err); + kfree(extent_op); + if (need_resched()) { mutex_unlock(&extent_root->fs_info->alloc_mutex); cond_resched(); @@ -1601,52 +1802,44 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, return 0; } -static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, - int is_data, int pending) +static int pin_down_bytes(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, int is_data) { int err = 0; + struct extent_buffer *buf; WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex)); - if (!pending) { - struct extent_buffer *buf; - - if (is_data) - goto pinit; - - buf = btrfs_find_tree_block(root, bytenr, num_bytes); - if (buf) { - /* we can reuse a block if it hasn't been written - * and it is from this transaction. We can't - * reuse anything from the tree log root because - * it has tiny sub-transactions. - */ - if (btrfs_buffer_uptodate(buf, 0) && - btrfs_try_tree_lock(buf)) { - u64 transid = - root->fs_info->running_transaction->transid; - u64 header_transid = - btrfs_header_generation(buf); - if (btrfs_header_owner(buf) != - BTRFS_TREE_LOG_OBJECTID && - header_transid == transid && - !btrfs_header_flag(buf, - BTRFS_HEADER_FLAG_WRITTEN)) { - clean_tree_block(NULL, root, buf); - btrfs_tree_unlock(buf); - free_extent_buffer(buf); - return 1; - } - btrfs_tree_unlock(buf); - } + if (is_data) + goto pinit; + + buf = btrfs_find_tree_block(root, bytenr, num_bytes); + if (!buf) + goto pinit; + + /* we can reuse a block if it hasn't been written + * and it is from this transaction. We can't + * reuse anything from the tree log root because + * it has tiny sub-transactions. + */ + if (btrfs_buffer_uptodate(buf, 0) && + btrfs_try_tree_lock(buf)) { + u64 header_owner = btrfs_header_owner(buf); + u64 header_transid = btrfs_header_generation(buf); + if (header_owner != BTRFS_TREE_LOG_OBJECTID && + header_transid == trans->transid && + !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { + clean_tree_block(NULL, root, buf); + btrfs_tree_unlock(buf); free_extent_buffer(buf); + return 1; } -pinit: - btrfs_update_pinned_extents(root, bytenr, num_bytes, 1); - } else { - set_extent_bits(&root->fs_info->pending_del, - bytenr, bytenr + num_bytes - 1, - EXTENT_LOCKED, GFP_NOFS); + btrfs_tree_unlock(buf); } + free_extent_buffer(buf); +pinit: + btrfs_update_pinned_extents(root, bytenr, num_bytes, 1); + BUG_ON(err < 0); return 0; } @@ -1654,11 +1847,12 @@ pinit: /* * remove an extent from the root, returns 0 on success */ -static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root - *root, u64 bytenr, u64 num_bytes, +static int __free_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, int pin, - int mark_free) + u64 owner_objectid, u64 owner_offset, + int pin, int mark_free) { struct btrfs_path *path; struct btrfs_key key; @@ -1681,10 +1875,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root return -ENOMEM; path->reada = 1; - ret = lookup_extent_backref(trans, extent_root, path, - bytenr, root_objectid, - ref_generation, - owner_objectid, owner_offset, 1); + ret = lookup_extent_backref(trans, extent_root, path, bytenr, parent, + root_objectid, ref_generation, 1); if (ret == 0) { struct btrfs_key found_key; extent_slot = path->slots[0]; @@ -1702,8 +1894,15 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root if (path->slots[0] - extent_slot > 5) break; } - if (!found_extent) - ret = btrfs_del_item(trans, extent_root, path); + if (!found_extent) { + ret = remove_extent_backref(trans, extent_root, path); + BUG_ON(ret); + btrfs_release_path(extent_root, path); + ret = btrfs_search_slot(trans, extent_root, + &key, path, -1, 1); + BUG_ON(ret); + extent_slot = path->slots[0]; + } } else { btrfs_print_leaf(extent_root, path->nodes[0]); WARN_ON(1); @@ -1712,14 +1911,6 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root root_objectid, ref_generation, owner_objectid, owner_offset); } - if (!found_extent) { - btrfs_release_path(extent_root, path); - ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); - if (ret < 0) - return ret; - BUG_ON(ret); - extent_slot = path->slots[0]; - } leaf = path->nodes[0]; ei = btrfs_item_ptr(leaf, extent_slot, @@ -1732,6 +1923,10 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root btrfs_mark_buffer_dirty(leaf); if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) { + struct btrfs_extent_ref *ref; + ref = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_extent_ref); + BUG_ON(btrfs_ref_num_refs(leaf, ref) != 1); /* if the back ref and the extent are next to each other * they get deleted below in one shot */ @@ -1739,15 +1934,13 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root num_to_del = 2; } else if (found_extent) { /* otherwise delete the extent back ref */ - ret = btrfs_del_item(trans, extent_root, path); + ret = remove_extent_backref(trans, extent_root, path); BUG_ON(ret); /* if refs are 0, we need to setup the path for deletion */ if (refs == 0) { btrfs_release_path(extent_root, path); ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); - if (ret < 0) - return ret; BUG_ON(ret); } } @@ -1761,8 +1954,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root #endif if (pin) { - ret = pin_down_bytes(root, bytenr, num_bytes, - owner_objectid >= BTRFS_FIRST_FREE_OBJECTID, 0); + ret = pin_down_bytes(trans, root, bytenr, num_bytes, + owner_objectid >= BTRFS_FIRST_FREE_OBJECTID); if (ret > 0) mark_free = 1; BUG_ON(ret < 0); @@ -1781,9 +1974,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root root_used - num_bytes); ret = btrfs_del_items(trans, extent_root, path, path->slots[0], num_to_del); - if (ret) { - return ret; - } + BUG_ON(ret); ret = update_block_group(trans, root, bytenr, num_bytes, 0, mark_free); BUG_ON(ret); @@ -1822,33 +2013,61 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct { int ret; int err = 0; + int mark_free = 0; u64 start; u64 end; + u64 priv; struct extent_io_tree *pending_del; - struct extent_io_tree *pinned_extents; + struct extent_io_tree *extent_ins; + struct pending_extent_op *extent_op; WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex)); + extent_ins = &extent_root->fs_info->extent_ins; pending_del = &extent_root->fs_info->pending_del; - pinned_extents = &extent_root->fs_info->pinned_extents; while(1) { ret = find_first_extent_bit(pending_del, 0, &start, &end, EXTENT_LOCKED); if (ret) break; + + ret = get_state_private(pending_del, start, &priv); + BUG_ON(ret); + extent_op = (struct pending_extent_op *)(unsigned long)priv; + clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, GFP_NOFS); - if (!test_range_bit(&extent_root->fs_info->extent_ins, - start, end, EXTENT_LOCKED, 0)) { - btrfs_update_pinned_extents(extent_root, start, - end + 1 - start, 1); + + ret = pin_down_bytes(trans, extent_root, start, + end + 1 - start, 0); + mark_free = ret > 0; + if (!test_range_bit(extent_ins, start, end, + EXTENT_LOCKED, 0)) { +free_extent: ret = __free_extent(trans, extent_root, - start, end + 1 - start, - extent_root->root_key.objectid, - 0, 0, 0, 0, 0); + start, end + 1 - start, + extent_op->orig_parent, + extent_root->root_key.objectid, + extent_op->orig_generation, + extent_op->level, 0, 0, mark_free); + kfree(extent_op); } else { - clear_extent_bits(&extent_root->fs_info->extent_ins, - start, end, EXTENT_LOCKED, GFP_NOFS); + kfree(extent_op); + ret = get_state_private(extent_ins, start, &priv); + BUG_ON(ret); + extent_op = (struct pending_extent_op *) + (unsigned long)priv; + + clear_extent_bits(extent_ins, start, end, + EXTENT_LOCKED, GFP_NOFS); + + if (extent_op->type == PENDING_BACKREF_UPDATE) + goto free_extent; + + ret = update_block_group(trans, extent_root, start, + end + 1 - start, 0, mark_free); + BUG_ON(ret); + kfree(extent_op); } if (ret) err = ret; @@ -1866,21 +2085,36 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct * remove an extent from the root, returns 0 on success */ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr, - u64 num_bytes, u64 root_objectid, - u64 ref_generation, u64 owner_objectid, - u64 owner_offset, int pin) + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, + u64 root_objectid, u64 ref_generation, + u64 owner_objectid, u64 owner_offset, int pin) { struct btrfs_root *extent_root = root->fs_info->extent_root; int pending_ret; int ret; WARN_ON(num_bytes < root->sectorsize); - if (!root->ref_cows) - ref_generation = 0; - if (root == extent_root) { - pin_down_bytes(root, bytenr, num_bytes, 0, 1); + struct pending_extent_op *extent_op; + + extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS); + BUG_ON(!extent_op); + + extent_op->type = PENDING_EXTENT_DELETE; + extent_op->bytenr = bytenr; + extent_op->num_bytes = num_bytes; + extent_op->parent = parent; + extent_op->orig_parent = parent; + extent_op->generation = ref_generation; + extent_op->orig_generation = ref_generation; + extent_op->level = (int)owner_objectid; + + set_extent_bits(&root->fs_info->pending_del, + bytenr, bytenr + num_bytes - 1, + EXTENT_LOCKED, GFP_NOFS); + set_state_private(&root->fs_info->pending_del, + bytenr, (unsigned long)extent_op); return 0; } /* if metadata always pin */ @@ -1901,9 +2135,9 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, if (ref_generation != trans->transid) pin = 1; - ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, - ref_generation, owner_objectid, owner_offset, - pin, pin == 0); + ret = __free_extent(trans, root, bytenr, num_bytes, parent, + root_objectid, ref_generation, owner_objectid, + owner_offset, pin, pin == 0); finish_current_insert(trans, root->fs_info->extent_root); pending_ret = del_pending_extents(trans, root->fs_info->extent_root); @@ -1911,15 +2145,15 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, } int btrfs_free_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr, - u64 num_bytes, u64 root_objectid, - u64 ref_generation, u64 owner_objectid, - u64 owner_offset, int pin) + struct btrfs_root *root, + u64 bytenr, u64 num_bytes, u64 parent, + u64 root_objectid, u64 ref_generation, + u64 owner_objectid, u64 owner_offset, int pin) { int ret; maybe_lock_mutex(root); - ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, + ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, parent, root_objectid, ref_generation, owner_objectid, owner_offset, pin); maybe_unlock_mutex(root); @@ -2271,7 +2505,7 @@ int btrfs_reserve_extent(struct btrfs_trans_handle *trans, } static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins) @@ -2289,6 +2523,9 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path; struct btrfs_key keys[2]; + if (parent == 0) + parent = ins->objectid; + /* block accounting for super block */ spin_lock_irq(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); @@ -2300,17 +2537,32 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, btrfs_set_root_used(&root->root_item, root_used + num_bytes); if (root == extent_root) { + struct pending_extent_op *extent_op; + + extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS); + BUG_ON(!extent_op); + + extent_op->type = PENDING_EXTENT_INSERT; + extent_op->bytenr = ins->objectid; + extent_op->num_bytes = ins->offset; + extent_op->parent = parent; + extent_op->orig_parent = 0; + extent_op->generation = ref_generation; + extent_op->orig_generation = 0; + extent_op->level = (int)owner; + set_extent_bits(&root->fs_info->extent_ins, ins->objectid, ins->objectid + ins->offset - 1, EXTENT_LOCKED, GFP_NOFS); + set_state_private(&root->fs_info->extent_ins, + ins->objectid, (unsigned long)extent_op); goto update_block; } memcpy(&keys[0], ins, sizeof(*ins)); - keys[1].offset = hash_extent_ref(root_objectid, ref_generation, - owner, owner_offset); keys[1].objectid = ins->objectid; keys[1].type = BTRFS_EXTENT_REF_KEY; + keys[1].offset = parent; sizes[0] = sizeof(*extent_item); sizes[1] = sizeof(*ref); @@ -2331,6 +2583,7 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); btrfs_set_ref_objectid(path->nodes[0], ref, owner); btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); + btrfs_set_ref_num_refs(path->nodes[0], ref, 1); btrfs_mark_buffer_dirty(path->nodes[0]); @@ -2359,7 +2612,7 @@ out: } int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins) @@ -2369,9 +2622,9 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, if (root_objectid == BTRFS_TREE_LOG_OBJECTID) return 0; maybe_lock_mutex(root); - ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, - ref_generation, owner, - owner_offset, ins); + ret = __btrfs_alloc_reserved_extent(trans, root, parent, + root_objectid, ref_generation, + owner, owner_offset, ins); maybe_unlock_mutex(root); return ret; } @@ -2382,7 +2635,7 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, * space cache bits as well */ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, + struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, u64 owner, u64 owner_offset, struct btrfs_key *ins) @@ -2396,10 +2649,9 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, ret = btrfs_remove_free_space(block_group, ins->objectid, ins->offset); BUG_ON(ret); - - ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, - ref_generation, owner, - owner_offset, ins); + ret = __btrfs_alloc_reserved_extent(trans, root, parent, + root_objectid, ref_generation, + owner, owner_offset, ins); maybe_unlock_mutex(root); return ret; } @@ -2413,9 +2665,9 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, */ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 num_bytes, u64 min_alloc_size, + u64 num_bytes, u64 parent, u64 min_alloc_size, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, + u64 owner_objectid, u64 owner_offset, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, u64 data) { @@ -2428,9 +2680,9 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, search_end, ins, data); BUG_ON(ret); if (root_objectid != BTRFS_TREE_LOG_OBJECTID) { - ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid, - ref_generation, owner, - owner_offset, ins); + ret = __btrfs_alloc_reserved_extent(trans, root, parent, + root_objectid, ref_generation, + owner_objectid, owner_offset, ins); BUG_ON(ret); } @@ -2468,10 +2720,9 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, */ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u32 blocksize, + u32 blocksize, u64 parent, u64 root_objectid, u64 ref_generation, - u64 first_objectid, int level, u64 hint, u64 empty_size) @@ -2480,10 +2731,9 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, int ret; struct extent_buffer *buf; - ret = btrfs_alloc_extent(trans, root, blocksize, blocksize, - root_objectid, ref_generation, - level, first_objectid, empty_size, hint, - (u64)-1, &ins, 0); + ret = btrfs_alloc_extent(trans, root, blocksize, parent, blocksize, + root_objectid, ref_generation, level, 0, + empty_size, hint, (u64)-1, &ins, 0); if (ret) { BUG_ON(ret > 0); return ERR_PTR(ret); @@ -2531,15 +2781,14 @@ int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(leaf, fi), - leaf_owner, leaf_generation, + leaf->start, leaf_owner, leaf_generation, key.objectid, key.offset, 0); mutex_unlock(&root->fs_info->alloc_mutex); + BUG_ON(ret); atomic_inc(&root->fs_info->throttle_gen); wake_up(&root->fs_info->transaction_throttle); cond_resched(); - - BUG_ON(ret); } return 0; } @@ -2554,10 +2803,10 @@ static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, for (i = 0; i < ref->nritems; i++) { mutex_lock(&root->fs_info->alloc_mutex); - ret = __btrfs_free_extent(trans, root, - info->bytenr, info->num_bytes, - ref->owner, ref->generation, - info->objectid, info->offset, 0); + ret = __btrfs_free_extent(trans, root, info->bytenr, + info->num_bytes, ref->bytenr, + ref->owner, ref->generation, + info->objectid, info->offset, 0); mutex_unlock(&root->fs_info->alloc_mutex); atomic_inc(&root->fs_info->throttle_gen); @@ -2576,7 +2825,7 @@ int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len, { int ret; - ret = lookup_extent_ref(NULL, root, start, len, refs); + ret = btrfs_lookup_extent_ref(NULL, root, start, len, refs); BUG_ON(ret); #if 0 // some debugging code in case we see problems here @@ -2672,8 +2921,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, bytenr, - blocksize, root_owner, - root_gen, 0, 0, 1); + blocksize, parent->start, + root_owner, root_gen, 0, 0, 1); BUG_ON(ret); mutex_unlock(&root->fs_info->alloc_mutex); @@ -2690,8 +2939,6 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, * So, we don't need to check it again */ if (*level == 1) { - struct btrfs_key key; - btrfs_node_key_to_cpu(cur, &key, path->slots[*level]); ref = btrfs_lookup_leaf_ref(root, bytenr); if (ref) { ret = cache_drop_leaf_ref(trans, root, ref); @@ -2750,12 +2997,13 @@ out: mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, bytenr, blocksize, - root_owner, root_gen, 0, 0, 1); + parent->start, root_owner, root_gen, + 0, 0, 1); + mutex_unlock(&root->fs_info->alloc_mutex); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; *level += 1; BUG_ON(ret); - mutex_unlock(&root->fs_info->alloc_mutex); cond_resched(); return 0; @@ -2792,19 +3040,18 @@ static int noinline walk_up_tree(struct btrfs_trans_handle *trans, root_item->drop_level = i; return 0; } else { - if (path->nodes[*level] == root->node) { - root_owner = root->root_key.objectid; - root_gen = - btrfs_header_generation(path->nodes[*level]); - } else { - struct extent_buffer *node; - node = path->nodes[*level + 1]; - root_owner = btrfs_header_owner(node); - root_gen = btrfs_header_generation(node); - } + struct extent_buffer *parent; + if (path->nodes[*level] == root->node) + parent = path->nodes[*level]; + else + parent = path->nodes[*level + 1]; + + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, path->nodes[*level]->len, + parent->start, root_owner, root_gen, 0, 0, 1); BUG_ON(ret); free_extent_buffer(path->nodes[*level]); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8624f3e8803..58ad25838a4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2201,9 +2201,10 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; - +#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; +#endif return ret; } EXPORT_SYMBOL(extent_write_cache_pages); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 58b329ddb42..48a702d41c8 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -524,6 +524,9 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, { u64 extent_end = 0; u64 search_start = start; + u64 leaf_start; + u64 root_gen; + u64 root_owner; struct extent_buffer *leaf; struct btrfs_file_extent_item *extent; struct btrfs_path *path; @@ -562,6 +565,9 @@ next_slot: bookend = 0; found_extent = 0; found_inline = 0; + leaf_start = 0; + root_gen = 0; + root_owner = 0; extent = NULL; leaf = path->nodes[0]; slot = path->slots[0]; @@ -628,27 +634,18 @@ next_slot: search_start = extent_end; if (end <= extent_end && start >= key.offset && found_inline) { *hint_byte = EXTENT_MAP_INLINE; - continue; + goto out; + } + + if (found_extent) { + read_extent_buffer(leaf, &old, (unsigned long)extent, + sizeof(old)); + root_gen = btrfs_header_generation(leaf); + root_owner = btrfs_header_owner(leaf); + leaf_start = leaf->start; } + if (end < extent_end && end >= key.offset) { - if (found_extent) { - u64 disk_bytenr = - btrfs_file_extent_disk_bytenr(leaf, extent); - u64 disk_num_bytes = - btrfs_file_extent_disk_num_bytes(leaf, - extent); - read_extent_buffer(leaf, &old, - (unsigned long)extent, - sizeof(old)); - if (disk_bytenr != 0) { - ret = btrfs_inc_extent_ref(trans, root, - disk_bytenr, disk_num_bytes, - root->root_key.objectid, - trans->transid, - key.objectid, end); - BUG_ON(ret); - } - } bookend = 1; if (found_inline && start <= key.offset) keep = 1; @@ -687,49 +684,12 @@ next_slot: } /* delete the entire extent */ if (!keep) { - u64 disk_bytenr = 0; - u64 disk_num_bytes = 0; - u64 extent_num_bytes = 0; - u64 root_gen; - u64 root_owner; - - root_gen = btrfs_header_generation(leaf); - root_owner = btrfs_header_owner(leaf); - if (found_extent) { - disk_bytenr = - btrfs_file_extent_disk_bytenr(leaf, - extent); - disk_num_bytes = - btrfs_file_extent_disk_num_bytes(leaf, - extent); - extent_num_bytes = - btrfs_file_extent_num_bytes(leaf, extent); - *hint_byte = - btrfs_file_extent_disk_bytenr(leaf, - extent); - } ret = btrfs_del_item(trans, root, path); /* TODO update progress marker and return */ BUG_ON(ret); - btrfs_release_path(root, path); extent = NULL; - if (found_extent && disk_bytenr != 0) { - dec_i_blocks(inode, extent_num_bytes); - ret = btrfs_free_extent(trans, root, - disk_bytenr, - disk_num_bytes, - root_owner, - root_gen, inode->i_ino, - key.offset, 0); - } - - BUG_ON(ret); - if (!bookend && search_start >= end) { - ret = 0; - goto out; - } - if (!bookend) - continue; + btrfs_release_path(root, path); + /* the extent will be freed later */ } if (bookend && found_inline && start <= key.offset) { u32 new_size; @@ -737,10 +697,13 @@ next_slot: extent_end - end); dec_i_blocks(inode, (extent_end - key.offset) - (extent_end - end)); - btrfs_truncate_item(trans, root, path, new_size, 0); + ret = btrfs_truncate_item(trans, root, path, + new_size, 0); + BUG_ON(ret); } /* create bookend, splitting the extent in two */ if (bookend && found_extent) { + u64 disk_bytenr; struct btrfs_key ins; ins.objectid = inode->i_ino; ins.offset = end; @@ -748,13 +711,9 @@ next_slot: btrfs_release_path(root, path); ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*extent)); + BUG_ON(ret); leaf = path->nodes[0]; - if (ret) { - btrfs_print_leaf(root, leaf); - printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.type, ins.offset, start, end, key.offset, extent_end, keep); - } - BUG_ON(ret); extent = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); write_extent_buffer(leaf, &old, @@ -770,11 +729,43 @@ next_slot: BTRFS_FILE_EXTENT_REG); btrfs_mark_buffer_dirty(path->nodes[0]); - if (le64_to_cpu(old.disk_bytenr) != 0) { + + disk_bytenr = le64_to_cpu(old.disk_bytenr); + if (disk_bytenr != 0) { + ret = btrfs_inc_extent_ref(trans, root, + disk_bytenr, + le64_to_cpu(old.disk_num_bytes), + leaf->start, + root->root_key.objectid, + trans->transid, + ins.objectid, ins.offset); + BUG_ON(ret); + } + btrfs_release_path(root, path); + if (disk_bytenr != 0) { inode->i_blocks += btrfs_file_extent_num_bytes(leaf, extent) >> 9; } + } + + if (found_extent && !keep) { + u64 disk_bytenr = le64_to_cpu(old.disk_bytenr); + + if (disk_bytenr != 0) { + dec_i_blocks(inode, le64_to_cpu(old.num_bytes)); + ret = btrfs_free_extent(trans, root, + disk_bytenr, + le64_to_cpu(old.disk_num_bytes), + leaf_start, root_owner, + root_gen, key.objectid, + key.offset, 0); + BUG_ON(ret); + *hint_byte = disk_bytenr; + } + } + + if (search_start >= end) { ret = 0; goto out; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 65b4f864b0d..2e7d82ec5d1 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -528,6 +528,9 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; struct btrfs_ordered_extent *ordered_extent; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct btrfs_file_extent_item *extent_item; + struct btrfs_path *path = NULL; + struct extent_buffer *leaf; u64 alloc_hint = 0; struct list_head list; struct btrfs_key ins; @@ -544,20 +547,15 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) goto nocow; + path = btrfs_alloc_path(); + BUG_ON(!path); + lock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, GFP_NOFS); INIT_LIST_HEAD(&list); - ins.objectid = ordered_extent->start; - ins.offset = ordered_extent->len; - ins.type = BTRFS_EXTENT_ITEM_KEY; - - ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid, - trans->transid, inode->i_ino, - ordered_extent->file_offset, &ins); - BUG_ON(ret); mutex_lock(&BTRFS_I(inode)->extent_mutex); ret = btrfs_drop_extents(trans, root, inode, @@ -566,18 +564,42 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ordered_extent->len, ordered_extent->file_offset, &alloc_hint); BUG_ON(ret); - ret = btrfs_insert_file_extent(trans, root, inode->i_ino, - ordered_extent->file_offset, - ordered_extent->start, - ordered_extent->len, - ordered_extent->len, 0); + + ins.objectid = inode->i_ino; + ins.offset = ordered_extent->file_offset; + ins.type = BTRFS_EXTENT_DATA_KEY; + ret = btrfs_insert_empty_item(trans, root, path, &ins, + sizeof(*extent_item)); BUG_ON(ret); + leaf = path->nodes[0]; + extent_item = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, extent_item, trans->transid); + btrfs_set_file_extent_type(leaf, extent_item, BTRFS_FILE_EXTENT_REG); + btrfs_set_file_extent_disk_bytenr(leaf, extent_item, + ordered_extent->start); + btrfs_set_file_extent_disk_num_bytes(leaf, extent_item, + ordered_extent->len); + btrfs_set_file_extent_offset(leaf, extent_item, 0); + btrfs_set_file_extent_num_bytes(leaf, extent_item, + ordered_extent->len); + btrfs_mark_buffer_dirty(leaf); btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1); mutex_unlock(&BTRFS_I(inode)->extent_mutex); + ins.objectid = ordered_extent->start; + ins.offset = ordered_extent->len; + ins.type = BTRFS_EXTENT_ITEM_KEY; + ret = btrfs_alloc_reserved_extent(trans, root, leaf->start, + root->root_key.objectid, + trans->transid, inode->i_ino, + ordered_extent->file_offset, &ins); + BUG_ON(ret); + btrfs_release_path(root, path); + inode->i_blocks += ordered_extent->len >> 9; unlock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, @@ -596,6 +618,8 @@ nocow: btrfs_put_ordered_extent(ordered_extent); btrfs_end_transaction(trans, root); + if (path) + btrfs_free_path(path); return 0; } @@ -1433,10 +1457,7 @@ search_again: if (root->ref_cows) dec_i_blocks(inode, num_dec); } - if (root->ref_cows) { - root_gen = - btrfs_header_generation(leaf); - } + root_gen = btrfs_header_generation(leaf); root_owner = btrfs_header_owner(leaf); } } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { @@ -1477,7 +1498,7 @@ delete: if (found_extent) { ret = btrfs_free_extent(trans, root, extent_start, extent_num_bytes, - root_owner, + leaf->start, root_owner, root_gen, inode->i_ino, found_key.offset, 0); BUG_ON(ret); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index f84b5f6991c..4c6e0c15754 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -76,9 +76,8 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, if (ret) goto fail; - leaf = btrfs_alloc_free_block(trans, root, root->leafsize, - objectid, trans->transid, 0, 0, - 0, 0); + leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, + objectid, trans->transid, 0, 0, 0); if (IS_ERR(leaf)) { ret = PTR_ERR(leaf); goto fail; @@ -525,13 +524,10 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) struct file *src_file; struct inode *src; struct btrfs_trans_handle *trans; - struct btrfs_ordered_extent *ordered; struct btrfs_path *path; struct extent_buffer *leaf; char *buf; struct btrfs_key key; - struct btrfs_key new_key; - u32 size; u32 nritems; int slot; int ret; @@ -576,6 +572,7 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) /* do any pending delalloc/csum calc on src, one way or another, and lock file content */ while (1) { + struct btrfs_ordered_extent *ordered; lock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS); ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered) @@ -619,6 +616,32 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) key.objectid != src->i_ino) break; + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY || + btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { + u32 size; + struct btrfs_key new_key; + + size = btrfs_item_size_nr(leaf, slot); + read_extent_buffer(leaf, buf, + btrfs_item_ptr_offset(leaf, slot), + size); + btrfs_release_path(root, path); + + memcpy(&new_key, &key, sizeof(new_key)); + new_key.objectid = inode->i_ino; + ret = btrfs_insert_empty_item(trans, root, path, + &new_key, size); + if (ret) + goto out; + + leaf = path->nodes[0]; + slot = path->slots[0]; + write_extent_buffer(leaf, buf, + btrfs_item_ptr_offset(leaf, slot), + size); + btrfs_mark_buffer_dirty(leaf); + } + if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) { struct btrfs_file_extent_item *extent; int found_type; @@ -634,31 +657,15 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) /* ds == 0 means there's a hole */ if (ds != 0) { ret = btrfs_inc_extent_ref(trans, root, - ds, dl, + ds, dl, leaf->start, root->root_key.objectid, trans->transid, inode->i_ino, key.offset); - if (ret) - goto out; + BUG_ON(ret); } } } - - if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY || - btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { - size = btrfs_item_size_nr(leaf, slot); - read_extent_buffer(leaf, buf, - btrfs_item_ptr_offset(leaf, slot), - size); - btrfs_release_path(root, path); - memcpy(&new_key, &key, sizeof(new_key)); - new_key.objectid = inode->i_ino; - ret = btrfs_insert_item(trans, root, &new_key, - buf, size); - BUG_ON(ret); - } else { - btrfs_release_path(root, path); - } + btrfs_release_path(root, path); key.offset++; } ret = 0; diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index f1374d597a1..3577badfa5b 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -102,11 +102,12 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) case BTRFS_EXTENT_REF_KEY: ref = btrfs_item_ptr(l, i, struct btrfs_extent_ref); printk("\t\textent back ref root %llu gen %llu " - "owner %llu offset %llu\n", + "owner %llu offset %llu num_refs %lu\n", (unsigned long long)btrfs_ref_root(l, ref), (unsigned long long)btrfs_ref_generation(l, ref), (unsigned long long)btrfs_ref_objectid(l, ref), - (unsigned long long)btrfs_ref_offset(l, ref)); + (unsigned long long)btrfs_ref_offset(l, ref), + (unsigned long)btrfs_ref_num_refs(l, ref)); break; case BTRFS_EXTENT_DATA_KEY: diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 56de3fb2d8d..88bbfd959f1 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -89,9 +89,9 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans, int ret; u64 objectid = root->root_key.objectid; - leaf = btrfs_alloc_free_block(trans, root, root->leafsize, + leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, BTRFS_TREE_LOG_OBJECTID, - 0, 0, 0, 0, 0); + trans->transid, 0, 0, 0); if (IS_ERR(leaf)) { ret = PTR_ERR(leaf); return ret; @@ -433,6 +433,49 @@ insert: trans->transid); } } + + if (overwrite_root && + key->type == BTRFS_EXTENT_DATA_KEY) { + int extent_type; + struct btrfs_file_extent_item *fi; + + fi = (struct btrfs_file_extent_item *)dst_ptr; + extent_type = btrfs_file_extent_type(path->nodes[0], fi); + if (extent_type == BTRFS_FILE_EXTENT_REG) { + struct btrfs_key ins; + ins.objectid = btrfs_file_extent_disk_bytenr( + path->nodes[0], fi); + ins.offset = btrfs_file_extent_disk_num_bytes( + path->nodes[0], fi); + ins.type = BTRFS_EXTENT_ITEM_KEY; + + /* + * is this extent already allocated in the extent + * allocation tree? If so, just add a reference + */ + ret = btrfs_lookup_extent(root, ins.objectid, + ins.offset); + if (ret == 0) { + ret = btrfs_inc_extent_ref(trans, root, + ins.objectid, ins.offset, + path->nodes[0]->start, + root->root_key.objectid, + trans->transid, + key->objectid, key->offset); + } else { + /* + * insert the extent pointer in the extent + * allocation tree + */ + ret = btrfs_alloc_logged_extent(trans, root, + path->nodes[0]->start, + root->root_key.objectid, + trans->transid, key->objectid, + key->offset, &ins); + BUG_ON(ret); + } + } + } no_copy: btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_release_path(root, path); @@ -551,45 +594,10 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, start, extent_end, start, &alloc_hint); BUG_ON(ret); + /* insert the extent */ + ret = overwrite_item(trans, root, path, eb, slot, key); BUG_ON(ret); - if (found_type == BTRFS_FILE_EXTENT_REG) { - struct btrfs_key ins; - - ins.objectid = btrfs_file_extent_disk_bytenr(eb, item); - ins.offset = btrfs_file_extent_disk_num_bytes(eb, item); - ins.type = BTRFS_EXTENT_ITEM_KEY; - - /* insert the extent pointer in the file */ - ret = overwrite_item(trans, root, path, eb, slot, key); - BUG_ON(ret); - /* - * is this extent already allocated in the extent - * allocation tree? If so, just add a reference - */ - ret = btrfs_lookup_extent(root, path, ins.objectid, ins.offset); - btrfs_release_path(root, path); - if (ret == 0) { - ret = btrfs_inc_extent_ref(trans, root, - ins.objectid, ins.offset, - root->root_key.objectid, - trans->transid, key->objectid, start); - } else { - /* - * insert the extent pointer in the extent - * allocation tree - */ - ret = btrfs_alloc_logged_extent(trans, root, - root->root_key.objectid, - trans->transid, key->objectid, - start, &ins); - BUG_ON(ret); - } - } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - /* inline extents are easy, we just overwrite them */ - ret = overwrite_item(trans, root, path, eb, slot, key); - BUG_ON(ret); - } /* btrfs_drop_extents changes i_blocks, update it here */ inode->i_blocks += (extent_end - start) >> 9; btrfs_update_inode(trans, root, inode); @@ -1806,16 +1814,14 @@ static int noinline walk_up_log_tree(struct btrfs_trans_handle *trans, WARN_ON(*level == 0); return 0; } else { - if (path->nodes[*level] == root->node) { - root_owner = root->root_key.objectid; - root_gen = - btrfs_header_generation(path->nodes[*level]); - } else { - struct extent_buffer *node; - node = path->nodes[*level + 1]; - root_owner = btrfs_header_owner(node); - root_gen = btrfs_header_generation(node); - } + struct extent_buffer *parent; + if (path->nodes[*level] == root->node) + parent = path->nodes[*level]; + else + parent = path->nodes[*level + 1]; + + root_owner = btrfs_header_owner(parent); + root_gen = btrfs_header_generation(parent); wc->process_func(root, path->nodes[*level], wc, btrfs_header_generation(path->nodes[*level])); if (wc->free) { @@ -2525,8 +2531,10 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, if (ds != 0) { ret = btrfs_inc_extent_ref(trans, log, ds, dl, + dst_path->nodes[0]->start, BTRFS_TREE_LOG_OBJECTID, - 0, ins_keys[i].objectid, + trans->transid, + ins_keys[i].objectid, ins_keys[i].offset); BUG_ON(ret); } -- cgit v1.2.3-70-g09d2 From 34353029534a08e41cfb8be647d734b9ce9ebff8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 23 Sep 2008 20:19:49 -0400 Subject: Btrfs: Fix race against disk_i_size updates The code to update the on disk i_size happens before the ordered_extent record is removed. So, it is possible for multiple ordered_extent completion routines to run at the same time, and to find each other in the ordered tree. The end result is they both decide not to update disk_i_size, leaving it too small. This temporary fix just puts the updates inside the extent_mutex. A real solution would be stronger ordering of disk_i_size updates against removing the ordered extent from the tree. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2e7d82ec5d1..adb169d739c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -608,9 +608,11 @@ nocow: add_pending_csums(trans, inode, ordered_extent->file_offset, &ordered_extent->list); + mutex_lock(&BTRFS_I(inode)->extent_mutex); btrfs_ordered_update_i_size(inode, ordered_extent); btrfs_update_inode(trans, root, inode); btrfs_remove_ordered_extent(inode, ordered_extent); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); /* once for us */ btrfs_put_ordered_extent(ordered_extent); -- cgit v1.2.3-70-g09d2 From 2b1f55b0f0d0d1a66470ef4ea2696cd5dd741a12 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 24 Sep 2008 11:48:04 -0400 Subject: Remove Btrfs compat code for older kernels Btrfs had compatibility code for kernels back to 2.6.18. These have been removed, and will be maintained in a separate backport git tree from now on. Signed-off-by: Chris Mason --- fs/btrfs/async-thread.c | 6 --- fs/btrfs/compat.h | 40 +------------- fs/btrfs/crc32c.h | 6 --- fs/btrfs/ctree.h | 7 +-- fs/btrfs/disk-io.c | 28 ---------- fs/btrfs/export.c | 6 --- fs/btrfs/extent_io.c | 66 ++--------------------- fs/btrfs/file.c | 24 +-------- fs/btrfs/inode.c | 135 +----------------------------------------------- fs/btrfs/super.c | 7 --- fs/btrfs/sysfs.c | 33 ------------ fs/btrfs/volumes.c | 20 ------- 12 files changed, 10 insertions(+), 368 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index 2ee30174019..4e780b279de 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -20,13 +20,7 @@ #include #include #include - -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) # include -#else -# include -#endif - #include "async-thread.h" /* diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h index b0ed1887d9b..cd6598b169d 100644 --- a/fs/btrfs/compat.h +++ b/fs/btrfs/compat.h @@ -1,9 +1,8 @@ #ifndef _COMPAT_H_ #define _COMPAT_H_ -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,26) -#define trylock_page(page) (!TestSetPageLocked(page)) -#endif +#define btrfs_drop_nlink(inode) drop_nlink(inode) +#define btrfs_inc_nlink(inode) inc_nlink(inode) #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,27) static inline struct dentry *d_obtain_alias(struct inode *inode) @@ -22,39 +21,4 @@ static inline struct dentry *d_obtain_alias(struct inode *inode) } #endif -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) -static inline void btrfs_drop_nlink(struct inode *inode) -{ - inode->i_nlink--; -} - -static inline void btrfs_inc_nlink(struct inode *inode) -{ - inode->i_nlink++; -} -#else -# define btrfs_drop_nlink(inode) drop_nlink(inode) -# define btrfs_inc_nlink(inode) inc_nlink(inode) -#endif - -/* - * Even if AppArmor isn't enabled, it still has different prototypes. - * Add more distro/version pairs here to declare which has AppArmor applied. - */ -#if defined(CONFIG_SUSE_KERNEL) -# if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) -# define REMOVE_SUID_PATH 1 -# endif -#endif - -/* - * catch any other distros that have patched in apparmor. This isn't - * 100% reliable because it won't catch people that hand compile their - * own distro kernels without apparmor compiled in. But, it is better - * than nothing. - */ -#ifdef CONFIG_SECURITY_APPARMOR -# define REMOVE_SUID_PATH 1 -#endif - #endif /* _COMPAT_H_ */ diff --git a/fs/btrfs/crc32c.h b/fs/btrfs/crc32c.h index bf6c12e8573..4f0fefed132 100644 --- a/fs/btrfs/crc32c.h +++ b/fs/btrfs/crc32c.h @@ -96,13 +96,7 @@ static inline u32 __btrfs_crc32c(u32 crc, unsigned char const *address, * We must workaround older implementations of crc32c_le() * found on older kernel versions. */ -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) -#define btrfs_crc32c(seed, data, length) \ - __cpu_to_le32( __btrfs_crc32c( __le32_to_cpu(seed), \ - (unsigned char const *)data, length) ) -#else #define btrfs_crc32c(seed, data, length) \ __btrfs_crc32c(seed, (unsigned char const *)data, length) #endif -#endif diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 138c157bbc4..3b3c1ca50c5 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1472,12 +1472,9 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { ((unsigned long)(btrfs_leaf_data(leaf) + \ btrfs_item_offset_nr(leaf, slot))) -static inline struct dentry *fdentry(struct file *file) { -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - return file->f_dentry; -#else +static inline struct dentry *fdentry(struct file *file) +{ return file->f_path.dentry; -#endif } /* extent-tree.c */ diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index d35ca6a3f51..dffb8dabd53 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -26,11 +26,7 @@ #include // for block_sync_page #include #include -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) # include -#else -# include -#endif #include "crc32c.h" #include "ctree.h" #include "disk-io.h" @@ -373,21 +369,11 @@ out: return ret; } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_workqueue_bio(struct bio *bio, int err) -#else -static int end_workqueue_bio(struct bio *bio, - unsigned int bytes_done, int err) -#endif { struct end_io_wq *end_io_wq = bio->bi_private; struct btrfs_fs_info *fs_info; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - fs_info = end_io_wq->info; end_io_wq->error = err; end_io_wq->work.func = end_workqueue_fn; @@ -397,10 +383,6 @@ static int end_workqueue_bio(struct bio *bio, &end_io_wq->work); else btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); - -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, @@ -1161,9 +1143,7 @@ void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) { -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) bdi_init(bdi); -#endif bdi->ra_pages = default_backing_dev_info.ra_pages; bdi->state = 0; bdi->capabilities = default_backing_dev_info.capabilities; @@ -1242,11 +1222,7 @@ static void end_workqueue_fn(struct btrfs_work *work) bio->bi_private = end_io_wq->private; bio->bi_end_io = end_io_wq->end_io; kfree(end_io_wq); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - bio_endio(bio, bio->bi_size, error); -#else bio_endio(bio, error); -#endif } static int cleaner_kthread(void *arg) @@ -1673,9 +1649,7 @@ fail: kfree(extent_root); kfree(tree_root); -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) bdi_destroy(&fs_info->bdi); -#endif kfree(fs_info); return ERR_PTR(err); } @@ -1936,9 +1910,7 @@ int close_ctree(struct btrfs_root *root) btrfs_close_devices(fs_info->fs_devices); btrfs_mapping_tree_free(&fs_info->mapping_tree); -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) bdi_destroy(&fs_info->bdi); -#endif kfree(fs_info->extent_root); kfree(fs_info->tree_root); diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c index 2b357a6d240..48b82cd7583 100644 --- a/fs/btrfs/export.c +++ b/fs/btrfs/export.c @@ -7,12 +7,6 @@ #include "export.h" #include "compat.h" -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) -#define FILEID_BTRFS_WITHOUT_PARENT 0x4d -#define FILEID_BTRFS_WITH_PARENT 0x4e -#define FILEID_BTRFS_WITH_PARENT_ROOT 0x4f -#endif - #define BTRFS_FID_SIZE_NON_CONNECTABLE (offsetof(struct btrfs_fid, parent_objectid)/4) #define BTRFS_FID_SIZE_CONNECTABLE (offsetof(struct btrfs_fid, parent_root_objectid)/4) #define BTRFS_FID_SIZE_CONNECTABLE_ROOT (sizeof(struct btrfs_fid)/4) diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 58ad25838a4..e3a25be5c66 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1397,12 +1397,7 @@ static int check_page_writeback(struct extent_io_tree *tree, * Scheduling is not allowed, so the extent state tree is expected * to have one and only one object corresponding to this IO. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_writepage(struct bio *bio, int err) -#else -static int end_bio_extent_writepage(struct bio *bio, - unsigned int bytes_done, int err) -#endif { int uptodate = err == 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; @@ -1412,10 +1407,6 @@ static int end_bio_extent_writepage(struct bio *bio, int whole_page; int ret; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif do { struct page *page = bvec->bv_page; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1461,10 +1452,8 @@ static int end_bio_extent_writepage(struct bio *bio, else check_page_writeback(tree, page); } while (bvec >= bio->bi_io_vec); + bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } /* @@ -1478,12 +1467,7 @@ static int end_bio_extent_writepage(struct bio *bio, * Scheduling is not allowed, so the extent state tree is expected * to have one and only one object corresponding to this IO. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_readpage(struct bio *bio, int err) -#else -static int end_bio_extent_readpage(struct bio *bio, - unsigned int bytes_done, int err) -#endif { int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; @@ -1493,11 +1477,6 @@ static int end_bio_extent_readpage(struct bio *bio, int whole_page; int ret; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - do { struct page *page = bvec->bv_page; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1556,9 +1535,6 @@ static int end_bio_extent_readpage(struct bio *bio, } while (bvec >= bio->bi_io_vec); bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } /* @@ -1566,12 +1542,7 @@ static int end_bio_extent_readpage(struct bio *bio, * the structs in the extent tree when done, and set the uptodate bits * as appropriate. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_extent_preparewrite(struct bio *bio, int err) -#else -static int end_bio_extent_preparewrite(struct bio *bio, - unsigned int bytes_done, int err) -#endif { const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; @@ -1579,11 +1550,6 @@ static int end_bio_extent_preparewrite(struct bio *bio, u64 start; u64 end; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif - do { struct page *page = bvec->bv_page; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1607,9 +1573,6 @@ static int end_bio_extent_preparewrite(struct bio *bio, } while (bvec >= bio->bi_io_vec); bio_put(bio); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } static struct bio * @@ -2079,12 +2042,6 @@ done: return 0; } -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) -/* Taken directly from 2.6.23 with a mod for a lockpage hook */ -typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc, - void *data); -#endif - /** * write_cache_pages - walk the list of dirty pages of the given address space and write all of them. * @mapping: address space structure to write @@ -2201,10 +2158,9 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; -#endif return ret; } EXPORT_SYMBOL(extent_write_cache_pages); @@ -2560,18 +2516,10 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, * by increasing the reference count. So we know the page must * be in the radix tree. */ -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) rcu_read_lock(); -#else - read_lock_irq(&mapping->tree_lock); -#endif p = radix_tree_lookup(&mapping->page_tree, i); - -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) rcu_read_unlock(); -#else - read_unlock_irq(&mapping->tree_lock); -#endif + return p; } @@ -2773,21 +2721,13 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, } } clear_page_dirty_for_io(page); -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) spin_lock_irq(&page->mapping->tree_lock); -#else - read_lock_irq(&page->mapping->tree_lock); -#endif if (!PageDirty(page)) { radix_tree_tag_clear(&page->mapping->page_tree, page_index(page), PAGECACHE_TAG_DIRTY); } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) spin_unlock_irq(&page->mapping->tree_lock); -#else - read_unlock_irq(&page->mapping->tree_lock); -#endif unlock_page(page); } return 0; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 48a702d41c8..8856570a0eb 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -871,15 +871,8 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, goto out_nolock; if (count == 0) goto out_nolock; -#ifdef REMOVE_SUID_PATH - err = remove_suid(&file->f_path); -#else -# if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) + err = file_remove_suid(file); -# else - err = remove_suid(fdentry(file)); -# endif -#endif if (err) goto out_nolock; file_update_time(file); @@ -1003,17 +996,10 @@ out_nolock: btrfs_commit_transaction(trans, root); } } else if (num_written > 0 && (file->f_flags & O_DIRECT)) { -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22) - do_sync_file_range(file, start_pos, - start_pos + num_written - 1, - SYNC_FILE_RANGE_WRITE | - SYNC_FILE_RANGE_WAIT_AFTER); -#else do_sync_mapping_range(inode->i_mapping, start_pos, start_pos + num_written - 1, SYNC_FILE_RANGE_WRITE | SYNC_FILE_RANGE_WAIT_AFTER); -#endif invalidate_mapping_pages(inode->i_mapping, start_pos >> PAGE_CACHE_SHIFT, (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); @@ -1097,12 +1083,7 @@ out: } static struct vm_operations_struct btrfs_file_vm_ops = { -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) - .nopage = filemap_nopage, - .populate = filemap_populate, -#else .fault = filemap_fault, -#endif .page_mkwrite = btrfs_page_mkwrite, }; @@ -1118,9 +1099,6 @@ struct file_operations btrfs_file_operations = { .read = do_sync_read, .aio_read = generic_file_aio_read, .splice_read = generic_file_splice_read, -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) - .sendfile = generic_file_sendfile, -#endif .write = btrfs_file_write, .mmap = btrfs_file_mmap, .open = generic_file_open, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index adb169d739c..48a3dc03080 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2073,104 +2073,6 @@ err: return ret; } -/* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd - will call the file system's ->lookup() method from within its - filldir callback, which in turn was called from the file system's - ->readdir() method. And will deadlock for many file systems. */ -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) - -struct nfshack_dirent { - u64 ino; - loff_t offset; - int namlen; - unsigned int d_type; - char name[]; -}; - -struct nfshack_readdir { - char *dirent; - size_t used; - int full; -}; - - - -static int btrfs_nfshack_filldir(void *__buf, const char *name, int namlen, - loff_t offset, u64 ino, unsigned int d_type) -{ - struct nfshack_readdir *buf = __buf; - struct nfshack_dirent *de = (void *)(buf->dirent + buf->used); - unsigned int reclen; - - reclen = ALIGN(sizeof(struct nfshack_dirent) + namlen, sizeof(u64)); - if (buf->used + reclen > PAGE_SIZE) { - buf->full = 1; - return -EINVAL; - } - - de->namlen = namlen; - de->offset = offset; - de->ino = ino; - de->d_type = d_type; - memcpy(de->name, name, namlen); - buf->used += reclen; - - return 0; -} - -static int btrfs_nfshack_readdir(struct file *file, void *dirent, - filldir_t filldir) -{ - struct nfshack_readdir buf; - struct nfshack_dirent *de; - int err; - int size; - loff_t offset; - - buf.dirent = (void *)__get_free_page(GFP_KERNEL); - if (!buf.dirent) - return -ENOMEM; - - offset = file->f_pos; - - do { - unsigned int reclen; - - buf.used = 0; - buf.full = 0; - err = btrfs_real_readdir(file, &buf, btrfs_nfshack_filldir); - if (err) - break; - - size = buf.used; - - if (!size) - break; - - de = (struct nfshack_dirent *)buf.dirent; - while (size > 0) { - offset = de->offset; - - if (filldir(dirent, de->name, de->namlen, de->offset, - de->ino, de->d_type)) - goto done; - offset = file->f_pos; - - reclen = ALIGN(sizeof(*de) + de->namlen, - sizeof(u64)); - size -= reclen; - de = (struct nfshack_dirent *)((char *)de + reclen); - } - } while (buf.full); - - done: - free_page((unsigned long)buf.dirent); - file->f_pos = offset; - - return err; -} -#endif - int btrfs_write_inode(struct inode *inode, int wait) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -3311,13 +3213,8 @@ unsigned long btrfs_force_ra(struct address_space *mapping, { pgoff_t req_size = last_index - offset + 1; -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) - offset = page_cache_readahead(mapping, ra, file, offset, req_size); - return offset; -#else page_cache_sync_readahead(mapping, ra, file, offset, req_size); return offset + req_size; -#endif } struct inode *btrfs_alloc_inode(struct super_block *sb) @@ -3373,14 +3270,7 @@ void btrfs_destroy_inode(struct inode *inode) kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) static void init_once(void *foo) -#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) -static void init_once(struct kmem_cache * cachep, void *foo) -#else -static void init_once(void * foo, struct kmem_cache * cachep, - unsigned long flags) -#endif { struct btrfs_inode *ei = (struct btrfs_inode *) foo; @@ -3403,22 +3293,10 @@ void btrfs_destroy_cachep(void) struct kmem_cache *btrfs_cache_create(const char *name, size_t size, unsigned long extra_flags, -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) - void (*ctor)(void *) -#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) - void (*ctor)(struct kmem_cache *, void *) -#else - void (*ctor)(void *, struct kmem_cache *, - unsigned long) -#endif - ) + void (*ctor)(void *)) { return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT | - SLAB_MEM_SPREAD | extra_flags), ctor -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) - ,NULL -#endif - ); + SLAB_MEM_SPREAD | extra_flags), ctor); } int btrfs_init_cachep(void) @@ -3666,12 +3544,7 @@ static int btrfs_set_page_dirty(struct page *page) return __set_page_dirty_nobuffers(page); } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) static int btrfs_permission(struct inode *inode, int mask) -#else -static int btrfs_permission(struct inode *inode, int mask, - struct nameidata *nd) -#endif { if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE)) return -EACCES; @@ -3702,11 +3575,7 @@ static struct inode_operations btrfs_dir_ro_inode_operations = { static struct file_operations btrfs_dir_file_operations = { .llseek = generic_file_llseek, .read = generic_read_dir, -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) - .readdir = btrfs_nfshack_readdir, -#else /* NFSd readdir/lookup deadlock is fixed */ .readdir = btrfs_real_readdir, -#endif .unlocked_ioctl = btrfs_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = btrfs_ioctl, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index f7b3eac7ac6..8399d6d05d6 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -349,10 +349,7 @@ static int btrfs_fill_super(struct super_block * sb, sb->s_root = root_dentry; -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25) save_mount_options(sb, data); -#endif - return 0; fail_close: @@ -566,11 +563,7 @@ static struct super_operations btrfs_super_ops = { .put_super = btrfs_put_super, .write_super = btrfs_write_super, .sync_fs = btrfs_sync_fs, -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25) - .read_inode = btrfs_read_locked_inode, -#else .show_options = generic_show_options, -#endif .write_inode = btrfs_write_inode, .dirty_inode = btrfs_dirty_inode, .alloc_inode = btrfs_alloc_inode, diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index 61af5d8446e..300076e6676 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -28,7 +28,6 @@ #include "disk-io.h" #include "transaction.h" -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25) static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf) { return snprintf(buf, PAGE_SIZE, "%llu\n", @@ -267,35 +266,3 @@ void btrfs_exit_sysfs(void) kset_unregister(btrfs_kset); } -#else - -int btrfs_sysfs_add_super(struct btrfs_fs_info *fs) -{ - return 0; -} - -int btrfs_sysfs_add_root(struct btrfs_root *root) -{ - return 0; -} - -void btrfs_sysfs_del_root(struct btrfs_root *root) -{ - return; -} - -void btrfs_sysfs_del_super(struct btrfs_fs_info *fs) -{ - return; -} - -int btrfs_init_sysfs(void) -{ - return 0; -} - -void btrfs_exit_sysfs(void) -{ - return; -} -#endif diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index b9e5c2d82dd..ddf89626498 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2080,20 +2080,11 @@ int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, } -#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) static void end_bio_multi_stripe(struct bio *bio, int err) -#else -static int end_bio_multi_stripe(struct bio *bio, - unsigned int bytes_done, int err) -#endif { struct btrfs_multi_bio *multi = bio->bi_private; int is_orig_bio = 0; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - if (bio->bi_size) - return 1; -#endif if (err) atomic_inc(&multi->error); @@ -2122,17 +2113,10 @@ static int end_bio_multi_stripe(struct bio *bio, } kfree(multi); -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - bio_endio(bio, bio->bi_size, err); -#else bio_endio(bio, err); -#endif } else if (!is_orig_bio) { bio_put(bio); } -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - return 0; -#endif } struct async_sched { @@ -2248,11 +2232,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, } else { bio->bi_bdev = root->fs_info->fs_devices->latest_bdev; bio->bi_sector = logical >> 9; -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) - bio_endio(bio, bio->bi_size, -EIO); -#else bio_endio(bio, -EIO); -#endif } dev_nr++; } -- cgit v1.2.3-70-g09d2 From 5b21f2ed3f2947b5195b65c9fdbdd9e52904cc03 Mon Sep 17 00:00:00 2001 From: Zheng Yan Date: Fri, 26 Sep 2008 10:05:38 -0400 Subject: Btrfs: extent_map and data=ordered fixes for space balancing * Add an EXTENT_BOUNDARY state bit to keep the writepage code from merging data extents that are in the process of being relocated. This allows us to do accounting for them properly. * The balancing code relocates data extents indepdent of the underlying inode. The extent_map code was modified to properly account for things moving around (invalidating extent_map caches in the inode). * Don't take the drop_mutex in the create_subvol ioctl. It isn't required. * Fix walking of the ordered extent list to avoid races with sys_unlink * Change the lock ordering rules. Transaction start goes outside the drop_mutex. This allows btrfs_commit_transaction to directly drop the relocation trees. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 9 +++++---- fs/btrfs/ctree.h | 11 ++++++++++- fs/btrfs/extent_io.c | 13 +++++++++---- fs/btrfs/extent_io.h | 1 + fs/btrfs/file.c | 31 +++++++++++++++++++++++------ fs/btrfs/inode-map.c | 4 ++++ fs/btrfs/inode.c | 52 +++++++++++++++++++++++++++++++++++-------------- fs/btrfs/ioctl.c | 2 -- fs/btrfs/ordered-data.c | 26 ++++++++++++------------- fs/btrfs/transaction.c | 8 ++++---- 10 files changed, 108 insertions(+), 49 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 50aea8cb653..f9cd40967d0 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -290,7 +290,6 @@ int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, struct extent_buffer **cow_ret, u64 prealloc_dest) { u64 search_start; - u64 header_trans; int ret; if (trans->transaction != root->fs_info->running_transaction) { @@ -304,9 +303,9 @@ int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, WARN_ON(1); } - header_trans = btrfs_header_generation(buf); spin_lock(&root->fs_info->hash_lock); - if (header_trans == trans->transid && + if (btrfs_header_generation(buf) == trans->transid && + btrfs_header_owner(buf) == root->root_key.objectid && !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { *cow_ret = buf; spin_unlock(&root->fs_info->hash_lock); @@ -1300,6 +1299,7 @@ again: /* is a cow on this block not required */ spin_lock(&root->fs_info->hash_lock); if (btrfs_header_generation(b) == trans->transid && + btrfs_header_owner(b) == root->root_key.objectid && !btrfs_header_flag(b, BTRFS_HEADER_FLAG_WRITTEN)) { spin_unlock(&root->fs_info->hash_lock); goto cow_done; @@ -1396,7 +1396,8 @@ cow_done: /* this is only true while dropping a snapshot */ if (level == lowest_level) { - break; + ret = 0; + goto done; } blocknr = btrfs_node_blockptr(b, slot); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b9f9f815ed0..3e62a1b0a1f 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1486,6 +1486,9 @@ static inline struct dentry *fdentry(struct file *file) /* extent-tree.c */ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len); +int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr, + u64 num_bytes, u32 *refs); int btrfs_update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin); int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, @@ -1812,6 +1815,8 @@ void btrfs_destroy_inode(struct inode *inode); int btrfs_init_cachep(void); void btrfs_destroy_cachep(void); long btrfs_ioctl_trans_end(struct file *file); +struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, + struct btrfs_root *root, int wait); struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root); struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, @@ -1824,13 +1829,17 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, int btrfs_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode); +int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode); +int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode); +void btrfs_orphan_cleanup(struct btrfs_root *root); /* ioctl.c */ long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); /* file.c */ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync); -int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end); +int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, + int skip_pinned); int btrfs_check_file(struct btrfs_root *root, struct inode *inode); extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index e3a25be5c66..8bd1b402f3f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -292,7 +292,7 @@ static int merge_state(struct extent_io_tree *tree, struct extent_state *other; struct rb_node *other_node; - if (state->state & EXTENT_IOBITS) + if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) return 0; other_node = rb_prev(&state->rb_node); @@ -1070,7 +1070,8 @@ search_again: while(1) { state = rb_entry(node, struct extent_state, rb_node); - if (found && state->start != cur_start) { + if (found && (state->start != cur_start || + (state->state & EXTENT_BOUNDARY))) { goto out; } if (!(state->state & EXTENT_DELALLOC)) { @@ -1078,7 +1079,7 @@ search_again: *end = state->end; goto out; } - if (!found) { + if (!found && !(state->state & EXTENT_BOUNDARY)) { struct extent_state *prev_state; struct rb_node *prev_node = node; while(1) { @@ -1088,7 +1089,11 @@ search_again: prev_state = rb_entry(prev_node, struct extent_state, rb_node); - if (!(prev_state->state & EXTENT_DELALLOC)) + if ((prev_state->end + 1 != state->start) || + !(prev_state->state & EXTENT_DELALLOC)) + break; + if ((cur_start - prev_state->start) * 2 > + max_bytes) break; state = prev_state; node = prev_node; diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 3cb411a5f4d..c9d1908a1ae 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -15,6 +15,7 @@ #define EXTENT_BUFFER_FILLED (1 << 8) #define EXTENT_ORDERED (1 << 9) #define EXTENT_ORDERED_METADATA (1 << 10) +#define EXTENT_BOUNDARY (1 << 11) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) /* diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 8856570a0eb..1b7e51a9db0 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -294,7 +294,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, last_pos_in_file, 0, 0, hole_size, 0); btrfs_drop_extent_cache(inode, last_pos_in_file, - last_pos_in_file + hole_size -1); + last_pos_in_file + hole_size - 1, 0); mutex_unlock(&BTRFS_I(inode)->extent_mutex); btrfs_check_file(root, inode); } @@ -337,7 +337,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, inline_size -= start_pos; err = insert_inline_extent(trans, root, inode, start_pos, inline_size, pages, 0, num_pages); - btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1); + btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1, 0); BUG_ON(err); mutex_unlock(&BTRFS_I(inode)->extent_mutex); @@ -362,7 +362,8 @@ out_unlock: return err; } -int noinline btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) +int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, + int skip_pinned) { struct extent_map *em; struct extent_map *split = NULL; @@ -371,6 +372,7 @@ int noinline btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) u64 len = end - start + 1; int ret; int testend = 1; + unsigned long flags; WARN_ON(end < start); if (end == (u64)-1) { @@ -389,6 +391,23 @@ int noinline btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) spin_unlock(&em_tree->lock); break; } + flags = em->flags; + if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) { + spin_unlock(&em_tree->lock); + if (em->start <= start && + (!testend || em->start + em->len >= start + len)) { + free_extent_map(em); + break; + } + if (start < em->start) { + len = em->start - start; + } else { + len = start + len - (em->start + em->len); + start = em->start + em->len; + } + free_extent_map(em); + continue; + } clear_bit(EXTENT_FLAG_PINNED, &em->flags); remove_extent_mapping(em_tree, em); @@ -398,7 +417,7 @@ int noinline btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) split->len = start - em->start; split->block_start = em->block_start; split->bdev = em->bdev; - split->flags = em->flags; + split->flags = flags; ret = add_extent_mapping(em_tree, split); BUG_ON(ret); free_extent_map(split); @@ -412,7 +431,7 @@ int noinline btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end) split->start = start + len; split->len = em->start + em->len - (start + len); split->bdev = em->bdev; - split->flags = em->flags; + split->flags = flags; split->block_start = em->block_start + diff; @@ -541,7 +560,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, int recow; int ret; - btrfs_drop_extent_cache(inode, start, end - 1); + btrfs_drop_extent_cache(inode, start, end - 1, 0); path = btrfs_alloc_path(); if (!path) diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index cd6171c2da4..80038c5ef7c 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -117,10 +117,14 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, *objectid = last_ino; goto found; } + } else if (key.objectid > search_start) { + *objectid = search_start; + goto found; } } if (key.objectid >= BTRFS_LAST_FREE_OBJECTID) break; + start_found = 1; last_ino = key.objectid + 1; path->slots[0]++; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 48a3dc03080..4516fbf0167 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -135,7 +135,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); mutex_lock(&BTRFS_I(inode)->extent_mutex); - btrfs_drop_extent_cache(inode, start, start + num_bytes - 1); + btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); mutex_unlock(&BTRFS_I(inode)->extent_mutex); while(num_bytes > 0) { @@ -163,7 +163,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) break; } btrfs_drop_extent_cache(inode, start, - start + ins.offset - 1); + start + ins.offset - 1, 0); } mutex_unlock(&BTRFS_I(inode)->extent_mutex); @@ -587,7 +587,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + - ordered_extent->len - 1); + ordered_extent->len - 1, 0); mutex_unlock(&BTRFS_I(inode)->extent_mutex); ins.objectid = ordered_extent->start; @@ -880,7 +880,7 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) int ret = 0, nr_unlink = 0, nr_truncate = 0; /* don't do orphan cleanup if the fs is readonly. */ - if (root->inode->i_sb->s_flags & MS_RDONLY) + if (root->fs_info->sb->s_flags & MS_RDONLY) return; path = btrfs_alloc_path(); @@ -892,8 +892,6 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); key.offset = (u64)-1; - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, root->inode); while (1) { ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); @@ -933,7 +931,7 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) * crossing root thing. we store the inode number in the * offset of the orphan item. */ - inode = btrfs_iget_locked(root->inode->i_sb, + inode = btrfs_iget_locked(root->fs_info->sb, found_key.offset, root); if (!inode) break; @@ -965,7 +963,9 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) * do a destroy_inode */ if (is_bad_inode(inode)) { + trans = btrfs_start_transaction(root, 1); btrfs_orphan_del(trans, inode); + btrfs_end_transaction(trans, root); iput(inode); continue; } @@ -988,7 +988,6 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate); btrfs_free_path(path); - btrfs_end_transaction(trans, root); } void btrfs_read_locked_inode(struct inode *inode) @@ -1343,8 +1342,7 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, u64 mask = root->sectorsize - 1; if (root->ref_cows) - btrfs_drop_extent_cache(inode, - new_size & (~mask), (u64)-1); + btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0); path = btrfs_alloc_path(); path->reada = -1; BUG_ON(!path); @@ -1677,7 +1675,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) hole_start, 0, 0, hole_size, 0); btrfs_drop_extent_cache(inode, hole_start, - (u64)-1); + (u64)-1, 0); btrfs_check_file(root, inode); } mutex_unlock(&BTRFS_I(inode)->extent_mutex); @@ -1843,6 +1841,24 @@ static int btrfs_find_actor(struct inode *inode, void *opaque) args->root == BTRFS_I(inode)->root); } +struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, + struct btrfs_root *root, int wait) +{ + struct inode *inode; + struct btrfs_iget_args args; + args.ino = objectid; + args.root = root; + + if (wait) { + inode = ilookup5(s, objectid, btrfs_find_actor, + (void *)&args); + } else { + inode = ilookup5_nowait(s, objectid, btrfs_find_actor, + (void *)&args); + } + return inode; +} + struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, struct btrfs_root *root) { @@ -3266,7 +3282,7 @@ void btrfs_destroy_inode(struct inode *inode) btrfs_put_ordered_extent(ordered); } } - btrfs_drop_extent_cache(inode, 0, (u64)-1); + btrfs_drop_extent_cache(inode, 0, (u64)-1, 0); kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); } @@ -3412,16 +3428,22 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) { struct list_head *head = &root->fs_info->delalloc_inodes; struct btrfs_inode *binode; + struct inode *inode; unsigned long flags; spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); while(!list_empty(head)) { binode = list_entry(head->next, struct btrfs_inode, delalloc_inodes); - atomic_inc(&binode->vfs_inode.i_count); + inode = igrab(&binode->vfs_inode); + if (!inode) + list_del_init(&binode->delalloc_inodes); spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); - filemap_write_and_wait(binode->vfs_inode.i_mapping); - iput(&binode->vfs_inode); + if (inode) { + filemap_write_and_wait(inode->i_mapping); + iput(inode); + } + cond_resched(); spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); } spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 4c6e0c15754..04de767a8db 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -444,12 +444,10 @@ static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, goto out; } - mutex_lock(&root->fs_info->drop_mutex); if (root == root->fs_info->tree_root) ret = create_subvol(root, vol_args->name, namelen); else ret = create_snapshot(root, vol_args->name, namelen); - mutex_unlock(&root->fs_info->drop_mutex); out: kfree(vol_args); return ret; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index da6d43eb41d..951eacff242 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -309,7 +309,6 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only) { struct list_head splice; struct list_head *cur; - struct list_head *tmp; struct btrfs_ordered_extent *ordered; struct inode *inode; @@ -317,37 +316,38 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only) spin_lock(&root->fs_info->ordered_extent_lock); list_splice_init(&root->fs_info->ordered_extents, &splice); - list_for_each_safe(cur, tmp, &splice) { + while (!list_empty(&splice)) { cur = splice.next; ordered = list_entry(cur, struct btrfs_ordered_extent, root_extent_list); if (nocow_only && !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) { + list_move(&ordered->root_extent_list, + &root->fs_info->ordered_extents); cond_resched_lock(&root->fs_info->ordered_extent_lock); continue; } list_del_init(&ordered->root_extent_list); atomic_inc(&ordered->refs); - inode = ordered->inode; /* - * the inode can't go away until all the pages are gone - * and the pages won't go away while there is still - * an ordered extent and the ordered extent won't go - * away until it is off this list. So, we can safely - * increment i_count here and call iput later + * the inode may be getting freed (in sys_unlink path). */ - atomic_inc(&inode->i_count); + inode = igrab(ordered->inode); + spin_unlock(&root->fs_info->ordered_extent_lock); - btrfs_start_ordered_extent(inode, ordered, 1); - btrfs_put_ordered_extent(ordered); - iput(inode); + if (inode) { + btrfs_start_ordered_extent(inode, ordered, 1); + btrfs_put_ordered_extent(ordered); + iput(inode); + } else { + btrfs_put_ordered_extent(ordered); + } spin_lock(&root->fs_info->ordered_extent_lock); } - list_splice_init(&splice, &root->fs_info->ordered_extents); spin_unlock(&root->fs_info->ordered_extent_lock); return 0; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 656baefa525..8c83cf464c8 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -109,6 +109,7 @@ noinline int btrfs_record_root_in_trans(struct btrfs_root *root) spin_lock_init(&dirty->root->node_lock); spin_lock_init(&dirty->root->list_lock); mutex_init(&dirty->root->objectid_mutex); + mutex_init(&dirty->root->log_mutex); INIT_LIST_HEAD(&dirty->root->dead_list); dirty->root->node = root->commit_root; dirty->root->commit_root = NULL; @@ -590,13 +591,14 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, root = dirty->latest_root; atomic_inc(&root->fs_info->throttles); - mutex_lock(&root->fs_info->drop_mutex); while(1) { trans = btrfs_start_transaction(tree_root, 1); + mutex_lock(&root->fs_info->drop_mutex); ret = btrfs_drop_snapshot(trans, dirty->root); if (ret != -EAGAIN) { break; } + mutex_unlock(&root->fs_info->drop_mutex); err = btrfs_update_root(trans, tree_root, @@ -608,10 +610,8 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, ret = btrfs_end_transaction(trans, tree_root); BUG_ON(ret); - mutex_unlock(&root->fs_info->drop_mutex); btrfs_btree_balance_dirty(tree_root, nr); cond_resched(); - mutex_lock(&root->fs_info->drop_mutex); } BUG_ON(ret); atomic_dec(&root->fs_info->throttles); @@ -689,7 +689,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, memcpy(new_root_item, &root->root_item, sizeof(*new_root_item)); key.objectid = objectid; - key.offset = 1; + key.offset = trans->transid; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); old = btrfs_lock_root_node(root); -- cgit v1.2.3-70-g09d2 From 8c8bee1d7ca47fc75b6bd24a8085c525a2394c02 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 29 Sep 2008 11:19:10 -0400 Subject: Btrfs: Wait for IO on the block device inodes of newly added devices btrfs-vol -a /dev/xxx will zero the first and last two MB of the device. The kernel code needs to wait for this IO to finish before it adds the device. btrfs metadata IO does not happen through the block device inode. A separate address space is used, allowing the zero filled buffer heads in the block device inode to be written to disk after FS metadata starts going down to the disk via the btrfs metadata inode. The end result is zero filled metadata blocks after adding new devices into the filesystem. The fix is a simple filemap_write_and_wait on the block device inode before actually inserting it into the pool of available devices. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 11 ++++++++--- fs/btrfs/inode.c | 13 ++++++++++++- fs/btrfs/volumes.c | 1 + 4 files changed, 22 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 2775e270881..0079b60b18f 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -610,6 +610,7 @@ struct btrfs_fs_info { struct list_head dead_roots; atomic_t nr_async_submits; + atomic_t async_submit_draining; atomic_t nr_async_bios; atomic_t tree_log_writers; atomic_t tree_log_commit; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 45bc3132b05..45b4f728527 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -460,6 +460,13 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, async->submit_bio_hook = submit_bio_hook; async->work.func = run_one_async_submit; async->work.flags = 0; + + while(atomic_read(&fs_info->async_submit_draining) && + atomic_read(&fs_info->nr_async_submits)) { + wait_event(fs_info->async_submit_wait, + (atomic_read(&fs_info->nr_async_submits) == 0)); + } + atomic_inc(&fs_info->nr_async_submits); btrfs_queue_worker(&fs_info->workers, &async->work); @@ -495,11 +502,8 @@ static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { struct btrfs_root *root = BTRFS_I(inode)->root; - u64 offset; int ret; - offset = bio->bi_sector << 9; - /* * when we're called for a write, we're already in the async * submission context. Just jump into btrfs_map_bio @@ -1360,6 +1364,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->space_info); btrfs_mapping_init(&fs_info->mapping_tree); atomic_set(&fs_info->nr_async_submits, 0); + atomic_set(&fs_info->async_submit_draining, 0); atomic_set(&fs_info->nr_async_bios, 0); atomic_set(&fs_info->throttles, 0); atomic_set(&fs_info->throttle_gen, 0); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4516fbf0167..404704d2682 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3440,13 +3440,24 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) list_del_init(&binode->delalloc_inodes); spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); if (inode) { - filemap_write_and_wait(inode->i_mapping); + filemap_flush(inode->i_mapping); iput(inode); } cond_resched(); spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); } spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + + /* the filemap_flush will queue IO into the worker threads, but + * we have to make sure the IO is actually started and that + * ordered extents get created before we return + */ + atomic_inc(&root->fs_info->async_submit_draining); + while(atomic_read(&root->fs_info->nr_async_submits)) { + wait_event(root->fs_info->async_submit_wait, + (atomic_read(&root->fs_info->nr_async_submits) == 0)); + } + atomic_dec(&root->fs_info->async_submit_draining); return 0; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 51f113119b2..f63cf7621a0 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1038,6 +1038,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) return -EIO; } + filemap_write_and_wait(bdev->bd_inode->i_mapping); mutex_lock(&root->fs_info->volume_mutex); trans = btrfs_start_transaction(root, 1); -- cgit v1.2.3-70-g09d2 From d352ac68148b69937d39ca5d48bcc4478e118dbf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 29 Sep 2008 15:18:18 -0400 Subject: Btrfs: add and improve comments This improves the comments at the top of many functions. It didn't dive into the guts of functions because I was trying to avoid merging problems with the new allocator and back reference work. extent-tree.c and volumes.c were both skipped, and there is definitely more work todo in cleaning and commenting the code. Signed-off-by: Chris Mason --- fs/btrfs/Makefile | 2 +- fs/btrfs/TODO | 20 ----- fs/btrfs/async-thread.c | 10 ++- fs/btrfs/async-thread.h | 7 +- fs/btrfs/bit-radix.c | 130 --------------------------------- fs/btrfs/bit-radix.h | 33 --------- fs/btrfs/btrfs_inode.h | 54 +++++++++++++- fs/btrfs/crc32c.h | 18 +++++ fs/btrfs/ctree.c | 127 +++++++++++++++++++++++++++++--- fs/btrfs/ctree.h | 1 - fs/btrfs/dir-item.c | 41 +++++++++++ fs/btrfs/disk-io.c | 33 ++++++++- fs/btrfs/extent_io.c | 34 ++++++++- fs/btrfs/extent_map.c | 10 +++ fs/btrfs/file.c | 44 ++++++++++- fs/btrfs/inode.c | 189 +++++++++++++++++++++++++++++++----------------- fs/btrfs/locking.c | 13 ++++ fs/btrfs/ordered-data.c | 19 ++++- fs/btrfs/ref-cache.c | 26 +++++++ fs/btrfs/ref-cache.h | 3 + fs/btrfs/root-tree.c | 21 +++++- fs/btrfs/struct-funcs.c | 21 ++++++ fs/btrfs/super.c | 3 + fs/btrfs/transaction.c | 67 ++++++++++++++++- fs/btrfs/tree-defrag.c | 4 + 25 files changed, 653 insertions(+), 277 deletions(-) delete mode 100644 fs/btrfs/TODO delete mode 100644 fs/btrfs/bit-radix.c delete mode 100644 fs/btrfs/bit-radix.h (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index d5c28557fba..48b7909ca8d 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -4,7 +4,7 @@ ifneq ($(KERNELRELEASE),) obj-m := btrfs.o btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ file-item.o inode-item.o inode-map.o disk-io.o \ - transaction.o bit-radix.o inode.o file.o tree-defrag.o \ + transaction.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \ ref-cache.o export.o tree-log.o acl.o free-space-cache.o diff --git a/fs/btrfs/TODO b/fs/btrfs/TODO deleted file mode 100644 index d9b6d38c603..00000000000 --- a/fs/btrfs/TODO +++ /dev/null @@ -1,20 +0,0 @@ -* cleanup, add more error checking, get rid of BUG_ONs -* Fix ENOSPC handling -* Make allocator smarter -* add a block group to struct inode -* Do actual block accounting -* Check compat and incompat flags on the inode -* Get rid of struct ctree_path, limiting tree levels held at one time -* Add generation number to key pointer in nodes -* Add generation number to inode -* forbid cross subvolume renames and hardlinks -* Release -* Do real tree locking -* Add extent mirroring (backup copies of blocks) -* Add fancy interface to get access to incremental backups -* Add fancy striped extents to make big reads faster -* Use relocation to try and fix write errors -* Make allocator much smarter -* xattrs (directory streams for regular files) -* Scrub & defrag - diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index 4e780b279de..04fb9702d14 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -231,17 +231,25 @@ static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers) /* * if we pick a busy task, move the task to the end of the list. - * hopefully this will keep things somewhat evenly balanced + * hopefully this will keep things somewhat evenly balanced. + * Do the move in batches based on the sequence number. This groups + * requests submitted at roughly the same time onto the same worker. */ next = workers->worker_list.next; worker = list_entry(next, struct btrfs_worker_thread, worker_list); atomic_inc(&worker->num_pending); worker->sequence++; + if (worker->sequence % workers->idle_thresh == 0) list_move_tail(next, &workers->worker_list); return worker; } +/* + * selects a worker thread to take the next job. This will either find + * an idle worker, start a new worker up to the max count, or just return + * one of the existing busy workers. + */ static struct btrfs_worker_thread *find_worker(struct btrfs_workers *workers) { struct btrfs_worker_thread *worker; diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h index 43e44d115dd..4ec9a2ee0f9 100644 --- a/fs/btrfs/async-thread.h +++ b/fs/btrfs/async-thread.h @@ -63,14 +63,17 @@ struct btrfs_workers { /* once a worker has this many requests or fewer, it is idle */ int idle_thresh; - /* list with all the work threads */ + /* list with all the work threads. The workers on the idle thread + * may be actively servicing jobs, but they haven't yet hit the + * idle thresh limit above. + */ struct list_head worker_list; struct list_head idle_list; /* lock for finding the next worker thread to queue on */ spinlock_t lock; - /* extra name for this worker */ + /* extra name for this worker, used for current->name */ char *name; }; diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c deleted file mode 100644 index e8bf876db39..00000000000 --- a/fs/btrfs/bit-radix.c +++ /dev/null @@ -1,130 +0,0 @@ -/* - * Copyright (C) 2007 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License v2 as published by the Free Software Foundation. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#include "bit-radix.h" - -#define BIT_ARRAY_BYTES 256 -#define BIT_RADIX_BITS_PER_ARRAY ((BIT_ARRAY_BYTES - sizeof(unsigned long)) * 8) - -extern struct kmem_cache *btrfs_bit_radix_cachep; -int set_radix_bit(struct radix_tree_root *radix, unsigned long bit) -{ - unsigned long *bits; - unsigned long slot; - int bit_slot; - int ret; - - slot = bit / BIT_RADIX_BITS_PER_ARRAY; - bit_slot = bit % BIT_RADIX_BITS_PER_ARRAY; - - bits = radix_tree_lookup(radix, slot); - if (!bits) { - bits = kmem_cache_alloc(btrfs_bit_radix_cachep, GFP_NOFS); - if (!bits) - return -ENOMEM; - memset(bits + 1, 0, BIT_ARRAY_BYTES - sizeof(unsigned long)); - bits[0] = slot; - ret = radix_tree_insert(radix, slot, bits); - if (ret) - return ret; - } - ret = test_and_set_bit(bit_slot, bits + 1); - if (ret < 0) - ret = 1; - return ret; -} - -int test_radix_bit(struct radix_tree_root *radix, unsigned long bit) -{ - unsigned long *bits; - unsigned long slot; - int bit_slot; - - slot = bit / BIT_RADIX_BITS_PER_ARRAY; - bit_slot = bit % BIT_RADIX_BITS_PER_ARRAY; - - bits = radix_tree_lookup(radix, slot); - if (!bits) - return 0; - return test_bit(bit_slot, bits + 1); -} - -int clear_radix_bit(struct radix_tree_root *radix, unsigned long bit) -{ - unsigned long *bits; - unsigned long slot; - int bit_slot; - int i; - int empty = 1; - - slot = bit / BIT_RADIX_BITS_PER_ARRAY; - bit_slot = bit % BIT_RADIX_BITS_PER_ARRAY; - - bits = radix_tree_lookup(radix, slot); - if (!bits) - return 0; - clear_bit(bit_slot, bits + 1); - for (i = 1; i < BIT_ARRAY_BYTES / sizeof(unsigned long); i++) { - if (bits[i]) { - empty = 0; - break; - } - } - if (empty) { - bits = radix_tree_delete(radix, slot); - BUG_ON(!bits); - kmem_cache_free(btrfs_bit_radix_cachep, bits); - } - return 0; -} - -int find_first_radix_bit(struct radix_tree_root *radix, unsigned long *retbits, - unsigned long start, int nr) -{ - unsigned long *bits; - unsigned long *gang[4]; - int found; - int ret; - int i; - int total_found = 0; - unsigned long slot; - - slot = start / BIT_RADIX_BITS_PER_ARRAY; - ret = radix_tree_gang_lookup(radix, (void **)gang, slot, - ARRAY_SIZE(gang)); - found = start % BIT_RADIX_BITS_PER_ARRAY; - for (i = 0; i < ret && nr > 0; i++) { - bits = gang[i]; - while(nr > 0) { - found = find_next_bit(bits + 1, - BIT_RADIX_BITS_PER_ARRAY, - found); - if (found < BIT_RADIX_BITS_PER_ARRAY) { - *retbits = bits[0] * - BIT_RADIX_BITS_PER_ARRAY + found; - retbits++; - nr--; - total_found++; - found++; - } else - break; - } - found = 0; - } - return total_found; -} diff --git a/fs/btrfs/bit-radix.h b/fs/btrfs/bit-radix.h deleted file mode 100644 index c100f54d5c3..00000000000 --- a/fs/btrfs/bit-radix.h +++ /dev/null @@ -1,33 +0,0 @@ -/* - * Copyright (C) 2007 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License v2 as published by the Free Software Foundation. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. - */ - -#ifndef __BIT_RADIX__ -#define __BIT_RADIX__ -#include - -int set_radix_bit(struct radix_tree_root *radix, unsigned long bit); -int test_radix_bit(struct radix_tree_root *radix, unsigned long bit); -int clear_radix_bit(struct radix_tree_root *radix, unsigned long bit); -int find_first_radix_bit(struct radix_tree_root *radix, unsigned long *retbits, - unsigned long start, int nr); - -static inline void init_bit_radix(struct radix_tree_root *radix) -{ - INIT_RADIX_TREE(radix, GFP_NOFS); -} -#endif diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 0577fda2168..0b2e623cf42 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -25,27 +25,58 @@ /* in memory btrfs inode */ struct btrfs_inode { + /* which subvolume this inode belongs to */ struct btrfs_root *root; + + /* the block group preferred for allocations. This pointer is buggy + * and needs to be replaced with a bytenr instead + */ struct btrfs_block_group_cache *block_group; + + /* key used to find this inode on disk. This is used by the code + * to read in roots of subvolumes + */ struct btrfs_key location; + + /* the extent_tree has caches of all the extent mappings to disk */ struct extent_map_tree extent_tree; + + /* the io_tree does range state (DIRTY, LOCKED etc) */ struct extent_io_tree io_tree; + + /* special utility tree used to record which mirrors have already been + * tried when checksums fail for a given block + */ struct extent_io_tree io_failure_tree; + + /* held while inserting checksums to avoid races */ struct mutex csum_mutex; + + /* held while inesrting or deleting extents from files */ struct mutex extent_mutex; + + /* held while logging the inode in tree-log.c */ struct mutex log_mutex; - struct inode vfs_inode; + + /* used to order data wrt metadata */ struct btrfs_ordered_inode_tree ordered_tree; + /* standard acl pointers */ struct posix_acl *i_acl; struct posix_acl *i_default_acl; /* for keeping track of orphaned inodes */ struct list_head i_orphan; + /* list of all the delalloc inodes in the FS. There are times we need + * to write all the delalloc pages to disk, and this list is used + * to walk them all. + */ struct list_head delalloc_inodes; - /* full 64 bit generation number */ + /* full 64 bit generation number, struct vfs_inode doesn't have a big + * enough field for this. + */ u64 generation; /* @@ -57,10 +88,25 @@ struct btrfs_inode { */ u64 logged_trans; - /* trans that last made a change that should be fully fsync'd */ + /* + * trans that last made a change that should be fully fsync'd. This + * gets reset to zero each time the inode is logged + */ u64 log_dirty_trans; + + /* total number of bytes pending delalloc, used by stat to calc the + * real block usage of the file + */ u64 delalloc_bytes; + + /* + * the size of the file stored in the metadata on disk. data=ordered + * means the in-memory i_size might be larger than the size on disk + * because not all the blocks are written yet. + */ u64 disk_i_size; + + /* flags field from the on disk inode */ u32 flags; /* @@ -68,6 +114,8 @@ struct btrfs_inode { * number for new files that are created */ u64 index_cnt; + + struct inode vfs_inode; }; static inline struct btrfs_inode *BTRFS_I(struct inode *inode) diff --git a/fs/btrfs/crc32c.h b/fs/btrfs/crc32c.h index 4f0fefed132..1eaf11d334f 100644 --- a/fs/btrfs/crc32c.h +++ b/fs/btrfs/crc32c.h @@ -1,3 +1,21 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + #ifndef __BTRFS_CRC32C__ #define __BTRFS_CRC32C__ #include diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 50e81f43e6d..ff3261ff2e1 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2007 Oracle. All rights reserved. + * Copyright (C) 2007,2008 Oracle. All rights reserved. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public @@ -54,12 +54,19 @@ struct btrfs_path *btrfs_alloc_path(void) return path; } +/* this also releases the path */ void btrfs_free_path(struct btrfs_path *p) { btrfs_release_path(NULL, p); kmem_cache_free(btrfs_path_cachep, p); } +/* + * path release drops references on the extent buffers in the path + * and it drops any locks held by this path + * + * It is safe to call this on paths that no locks or extent buffers held. + */ void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) { int i; @@ -77,6 +84,16 @@ void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) } } +/* + * safely gets a reference on the root node of a tree. A lock + * is not taken, so a concurrent writer may put a different node + * at the root of the tree. See btrfs_lock_root_node for the + * looping required. + * + * The extent buffer returned by this has a reference taken, so + * it won't disappear. It may stop being the root of the tree + * at any time because there are no locks held. + */ struct extent_buffer *btrfs_root_node(struct btrfs_root *root) { struct extent_buffer *eb; @@ -87,6 +104,10 @@ struct extent_buffer *btrfs_root_node(struct btrfs_root *root) return eb; } +/* loop around taking references on and locking the root node of the + * tree until you end up with a lock on the root. A locked buffer + * is returned, with a reference held. + */ struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root) { struct extent_buffer *eb; @@ -108,6 +129,10 @@ struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root) return eb; } +/* cowonly root (everything not a reference counted cow subvolume), just get + * put onto a simple dirty list. transaction.c walks this to make sure they + * get properly updated on disk. + */ static void add_root_to_dirty_list(struct btrfs_root *root) { if (root->track_dirty && list_empty(&root->dirty_list)) { @@ -116,6 +141,11 @@ static void add_root_to_dirty_list(struct btrfs_root *root) } } +/* + * used by snapshot creation to make a copy of a root for a tree with + * a given objectid. The buffer with the new root node is returned in + * cow_ret, and this func returns zero on success or a negative error code. + */ int btrfs_copy_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, @@ -167,6 +197,22 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, return 0; } +/* + * does the dirty work in cow of a single block. The parent block + * (if supplied) is updated to point to the new cow copy. The new + * buffer is marked dirty and returned locked. If you modify the block + * it needs to be marked dirty again. + * + * search_start -- an allocation hint for the new block + * + * empty_size -- a hint that you plan on doing more cow. This is the size in bytes + * the allocator should try to find free next to the block it returns. This is + * just a hint and may be ignored by the allocator. + * + * prealloc_dest -- if you have already reserved a destination for the cow, + * this uses that block instead of allocating a new one. btrfs_alloc_reserved_extent + * is used to finish the allocation. + */ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, @@ -311,6 +357,11 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, return 0; } +/* + * cows a single block, see __btrfs_cow_block for the real work. + * This version of it has extra checks so that a block isn't cow'd more than + * once per transaction, as long as it hasn't been written yet + */ int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, @@ -347,6 +398,10 @@ int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, return ret; } +/* + * helper function for defrag to decide if two blocks pointed to by a + * node are actually close by + */ static int close_blocks(u64 blocknr, u64 other, u32 blocksize) { if (blocknr < other && other - (blocknr + blocksize) < 32768) @@ -381,6 +436,11 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2) } +/* + * this is used by the defrag code to go through all the + * leaves pointed to by a node and reallocate them so that + * disk order is close to key order + */ int btrfs_realloc_node(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *parent, int start_slot, int cache_only, u64 *last_ret, @@ -521,6 +581,10 @@ static inline unsigned int leaf_data_end(struct btrfs_root *root, return btrfs_item_offset_nr(leaf, nr - 1); } +/* + * extra debugging checks to make sure all the items in a key are + * well formed and in the proper order + */ static int check_node(struct btrfs_root *root, struct btrfs_path *path, int level) { @@ -561,6 +625,10 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path, return 0; } +/* + * extra checking to make sure all the items in a leaf are + * well formed and in the proper order + */ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, int level) { @@ -782,6 +850,10 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key, return -1; } +/* given a node and slot number, this reads the blocks it points to. The + * extent buffer is returned with a reference taken (but unlocked). + * NULL is returned on error. + */ static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root, struct extent_buffer *parent, int slot) { @@ -798,6 +870,11 @@ static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root, btrfs_node_ptr_generation(parent, slot)); } +/* + * node level balancing, used to make sure nodes are in proper order for + * item deletion. We balance from the top down, so we have to make sure + * that a deletion won't leave an node completely empty later on. + */ static noinline int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) @@ -1024,7 +1101,10 @@ enospc: return ret; } -/* returns zero if the push worked, non-zero otherwise */ +/* Node balancing for insertion. Here we only split or push nodes around + * when they are completely full. This is also done top down, so we + * have to be pessimistic. + */ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) @@ -1150,7 +1230,8 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, } /* - * readahead one full node of leaves + * readahead one full node of leaves, finding things that are close + * to the block in 'slot', and triggering ra on them. */ static noinline void reada_for_search(struct btrfs_root *root, struct btrfs_path *path, @@ -1226,6 +1307,19 @@ static noinline void reada_for_search(struct btrfs_root *root, } } +/* + * when we walk down the tree, it is usually safe to unlock the higher layers in + * the tree. The exceptions are when our path goes through slot 0, because operations + * on the tree might require changing key pointers higher up in the tree. + * + * callers might also have set path->keep_locks, which tells this code to + * keep the lock if the path points to the last slot in the block. This is + * part of walking through the tree, and selecting the next slot in the higher + * block. + * + * lowest_unlock sets the lowest level in the tree we're allowed to unlock. + * so if lowest_unlock is 1, level 0 won't be unlocked + */ static noinline void unlock_up(struct btrfs_path *path, int level, int lowest_unlock) { @@ -2705,6 +2799,12 @@ again: return ret; } +/* + * make the item pointed to by the path smaller. new_size indicates + * how small to make it, and from_end tells us if we just chop bytes + * off the end of the item or if we shift the item to chop bytes off + * the front. + */ int btrfs_truncate_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, @@ -2818,6 +2918,9 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, return ret; } +/* + * make the item pointed to by the path bigger, data_size is the new size. + */ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u32 data_size) @@ -2897,7 +3000,7 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, } /* - * Given a key and some data, insert an item into the tree. + * Given a key and some data, insert items into the tree. * This does all the path init required, making room in the tree if needed. */ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, @@ -3046,9 +3149,8 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root /* * delete the pointer from a given node. * - * If the delete empties a node, the node is removed from the tree, - * continuing all the way the root if required. The root is converted into - * a leaf if all the nodes are emptied. + * the tree should have been previously balanced so the deletion does not + * empty a node. */ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level, int slot) @@ -3233,6 +3335,9 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, * search the tree again to find a leaf with lesser keys * returns 0 if it found something or 1 if there are no lesser leaves. * returns < 0 on io errors. + * + * This may release the path, and so you may lose any locks held at the + * time you call it. */ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) { @@ -3265,9 +3370,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) /* * A helper function to walk down the tree starting at min_key, and looking * for nodes or leaves that are either in cache or have a minimum - * transaction id. This is used by the btree defrag code, but could - * also be used to search for blocks that have changed since a given - * transaction id. + * transaction id. This is used by the btree defrag code, and tree logging * * This does not cow, but it does stuff the starting key it finds back * into min_key, so you can call btrfs_search_slot with cow=1 on the @@ -3279,6 +3382,10 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) * This honors path->lowest_level to prevent descent past a given level * of the tree. * + * min_trans indicates the oldest transaction that you are interested + * in walking through. Any nodes or leaves older than min_trans are + * skipped over (without reading them). + * * returns zero if something useful was found, < 0 on error and 1 if there * was nothing in the tree that matched the search criteria. */ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 0079b60b18f..ded1643c027 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -27,7 +27,6 @@ #include #include #include -#include "bit-radix.h" #include "extent_io.h" #include "extent_map.h" #include "async-thread.h" diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index e4f30090d64..5040b71f190 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -21,6 +21,14 @@ #include "hash.h" #include "transaction.h" +/* + * insert a name into a directory, doing overflow properly if there is a hash + * collision. data_size indicates how big the item inserted should be. On + * success a struct btrfs_dir_item pointer is returned, otherwise it is + * an ERR_PTR. + * + * The name is not copied into the dir item, you have to do that yourself. + */ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -55,6 +63,10 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle return (struct btrfs_dir_item *)ptr; } +/* + * xattrs work a lot like directories, this inserts an xattr item + * into the tree + */ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, u16 name_len, const void *data, u16 data_len, @@ -109,6 +121,13 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans, return ret; } +/* + * insert a directory item in the tree, doing all the magic for + * both indexes. 'dir' indicates which objectid to insert it into, + * 'location' is the key to stuff into the directory item, 'type' is the + * type of the inode we're pointing to, and 'index' is the sequence number + * to use for the second index (if one is created). + */ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, const char *name, int name_len, u64 dir, struct btrfs_key *location, u8 type, u64 index) @@ -184,6 +203,11 @@ out: return 0; } +/* + * lookup a directory item based on name. 'dir' is the objectid + * we're searching in, and 'mod' tells us if you plan on deleting the + * item (use mod < 0) or changing the options (use mod > 0) + */ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 dir, @@ -222,6 +246,14 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, return btrfs_match_dir_item_name(root, path, name, name_len); } +/* + * lookup a directory item based on index. 'dir' is the objectid + * we're searching in, and 'mod' tells us if you plan on deleting the + * item (use mod < 0) or changing the options (use mod > 0) + * + * The name is used to make sure the index really points to the name you were + * looking for. + */ struct btrfs_dir_item * btrfs_lookup_dir_index_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -282,6 +314,11 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans, return btrfs_match_dir_item_name(root, path, name, name_len); } +/* + * helper function to look at the directory item pointed to by 'path' + * this walks through all the entries in a dir item and finds one + * for a specific name. + */ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path, const char *name, int name_len) @@ -313,6 +350,10 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, return NULL; } +/* + * given a pointer into a directory item, delete it. This + * handles items that have more than one entry in them. + */ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 45b4f728527..5ee10d3136f 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -55,6 +55,11 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) static struct extent_io_ops btree_extent_io_ops; static void end_workqueue_fn(struct btrfs_work *work); +/* + * end_io_wq structs are used to do processing in task context when an IO is + * complete. This is used during reads to verify checksums, and it is used + * by writes to insert metadata for new file extents after IO is complete. + */ struct end_io_wq { struct bio *bio; bio_end_io_t *end_io; @@ -66,6 +71,11 @@ struct end_io_wq { struct btrfs_work work; }; +/* + * async submit bios are used to offload expensive checksumming + * onto the worker threads. They checksum file and metadata bios + * just before they are sent down the IO stack. + */ struct async_submit_bio { struct inode *inode; struct bio *bio; @@ -76,6 +86,10 @@ struct async_submit_bio { struct btrfs_work work; }; +/* + * extents on the btree inode are pretty simple, there's one extent + * that covers the entire device + */ struct extent_map *btree_get_extent(struct inode *inode, struct page *page, size_t page_offset, u64 start, u64 len, int create) @@ -151,6 +165,10 @@ void btrfs_csum_final(u32 crc, char *result) *(__le32 *)result = ~cpu_to_le32(crc); } +/* + * compute the csum for a btree block, and either verify it or write it + * into the csum field of the block. + */ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, int verify) { @@ -204,6 +222,12 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, return 0; } +/* + * we can't consider a given block up to date unless the transid of the + * block matches the transid in the parent node's pointer. This is how we + * detect blocks that either didn't get written at all or got written + * in the wrong place. + */ static int verify_parent_transid(struct extent_io_tree *io_tree, struct extent_buffer *eb, u64 parent_transid) { @@ -228,9 +252,12 @@ out: unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); return ret; - } +/* + * helper to read a given tree block, doing retries as required when + * the checksums don't match and we have alternate mirrors to try. + */ static int btree_read_extent_buffer_pages(struct btrfs_root *root, struct extent_buffer *eb, u64 start, u64 parent_transid) @@ -260,6 +287,10 @@ printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror return -EIO; } +/* + * checksum a dirty tree block before IO. This has extra checks to make + * sure we only fill in the checksum field in the first page of a multi-page block + */ int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { struct extent_io_tree *tree; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 8bd1b402f3f..563b2d12f4f 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -914,6 +914,10 @@ int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(wait_on_extent_writeback); +/* + * either insert or lock state struct between start and end use mask to tell + * us if waiting is desired. + */ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { int err; @@ -982,6 +986,13 @@ int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(set_range_writeback); +/* + * find the first offset in the io tree with 'bits' set. zero is + * returned if we find something, and *start_ret and *end_ret are + * set to reflect the state struct that was found. + * + * If nothing was found, 1 is returned, < 0 on error + */ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, u64 *start_ret, u64 *end_ret, int bits) { @@ -1017,6 +1028,10 @@ out: } EXPORT_SYMBOL(find_first_extent_bit); +/* find the first state struct with 'bits' set after 'start', and + * return it. tree->lock must be held. NULL will returned if + * nothing was found after 'start' + */ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, u64 start, int bits) { @@ -1046,8 +1061,14 @@ out: } EXPORT_SYMBOL(find_first_extent_bit_state); -u64 find_lock_delalloc_range(struct extent_io_tree *tree, - u64 *start, u64 *end, u64 max_bytes) +/* + * find a contiguous range of bytes in the file marked as delalloc, not + * more than 'max_bytes'. start and end are used to return the range, + * + * 1 is returned if we find something, 0 if nothing was in the tree + */ +static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) { struct rb_node *node; struct extent_state *state; @@ -1130,6 +1151,11 @@ out: return found; } +/* + * count the number of bytes in the tree that have a given bit(s) + * set. This can be fairly slow, except for EXTENT_DIRTY which is + * cached. The total number found is returned. + */ u64 count_range_bits(struct extent_io_tree *tree, u64 *start, u64 search_end, u64 max_bytes, unsigned long bits) @@ -1245,6 +1271,10 @@ int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) } EXPORT_SYMBOL(unlock_range); +/* + * set the private field for a given byte offset in the tree. If there isn't + * an extent_state there already, this does nothing. + */ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) { struct rb_node *node; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 78ced11d18c..74b2a29880d 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -114,6 +114,10 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset, return NULL; } +/* + * search through the tree for an extent_map with a given offset. If + * it can't be found, try to find some neighboring extents + */ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, struct rb_node **prev_ret, struct rb_node **next_ret) @@ -160,6 +164,10 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, return NULL; } +/* + * look for an offset in the tree, and if it can't be found, return + * the first offset we can find smaller than 'offset'. + */ static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) { struct rb_node *prev; @@ -170,6 +178,7 @@ static inline struct rb_node *tree_search(struct rb_root *root, u64 offset) return ret; } +/* check to see if two extent_map structs are adjacent and safe to merge */ static int mergable_maps(struct extent_map *prev, struct extent_map *next) { if (test_bit(EXTENT_FLAG_PINNED, &prev->flags)) @@ -250,6 +259,7 @@ out: } EXPORT_SYMBOL(add_extent_mapping); +/* simple helper to do math around the end of an extent, handling wrap */ static u64 range_end(u64 start, u64 len) { if (start + len < start) diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1b7e51a9db0..3088a118448 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -41,6 +41,9 @@ #include "compat.h" +/* simple helper to fault in pages and copy. This should go away + * and be replaced with calls into generic code. + */ static int noinline btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes, struct page **prepared_pages, @@ -72,12 +75,19 @@ static int noinline btrfs_copy_from_user(loff_t pos, int num_pages, return page_fault ? -EFAULT : 0; } +/* + * unlocks pages after btrfs_file_write is done with them + */ static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) { size_t i; for (i = 0; i < num_pages; i++) { if (!pages[i]) break; + /* page checked is some magic around finding pages that + * have been modified without going through btrfs_set_page_dirty + * clear it here + */ ClearPageChecked(pages[i]); unlock_page(pages[i]); mark_page_accessed(pages[i]); @@ -85,6 +95,10 @@ static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) } } +/* this does all the hard work for inserting an inline extent into + * the btree. Any existing inline extent is extended as required to make room, + * otherwise things are inserted as required into the btree + */ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 offset, size_t size, @@ -228,6 +242,14 @@ fail: return err; } +/* + * after copy_from_user, pages need to be dirtied and we need to make + * sure holes are created between the current EOF and the start of + * any next extents (if required). + * + * this also makes the decision about creating an inline extent vs + * doing real data extents, marking pages dirty and delalloc as required. + */ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct file *file, @@ -362,6 +384,10 @@ out_unlock: return err; } +/* + * this drops all the extents in the cache that intersect the range + * [start, end]. Existing extents are split as required. + */ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, int skip_pinned) { @@ -536,6 +562,9 @@ out: * If an extent intersects the range but is not entirely inside the range * it is either truncated or split. Anything entirely inside the range * is deleted from the tree. + * + * inline_limit is used to tell this code which offsets in the file to keep + * if they contain inline extents. */ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, @@ -796,7 +825,9 @@ out: } /* - * this gets pages into the page cache and locks them down + * this gets pages into the page cache and locks them down, it also properly + * waits for data=ordered extents to finish before allowing the pages to be + * modified. */ static int noinline prepare_pages(struct btrfs_root *root, struct file *file, struct page **pages, size_t num_pages, @@ -1034,6 +1065,17 @@ int btrfs_release_file(struct inode * inode, struct file * filp) return 0; } +/* + * fsync call for both files and directories. This logs the inode into + * the tree log instead of forcing full commits whenever possible. + * + * It needs to call filemap_fdatawait so that all ordered extent updates are + * in the metadata btree are up to date for copying to the log. + * + * It drops the inode mutex before doing the tree log commit. This is an + * important optimization for directories because holding the mutex prevents + * new operations on the dir while we write to disk. + */ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) { struct inode *inode = dentry->d_inode; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 404704d2682..f3abecc2d14 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -83,6 +83,10 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { static void btrfs_truncate(struct inode *inode); +/* + * a very lame attempt at stopping writes when the FS is 85% full. There + * are countless ways this is incorrect, but it is better than nothing. + */ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, int for_del) { @@ -108,6 +112,12 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, return ret; } +/* + * when extent_io.c finds a delayed allocation range in the file, + * the call backs end up in this code. The basic idea is to + * allocate extents on disk for the range, and create ordered data structs + * in ram to track those extents. + */ static int cow_file_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -185,6 +195,13 @@ out: return ret; } +/* + * when nowcow writeback call back. This checks for snapshots or COW copies + * of the extents that exist in the file, and COWs the file as required. + * + * If no cow copies or snapshots exist, we write directly to the existing + * blocks on disk + */ static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) { u64 extent_start; @@ -291,6 +308,9 @@ out: return err; } +/* + * extent_io.c call back to do delayed allocation processing + */ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -305,6 +325,11 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end) return ret; } +/* + * extent_io.c set_bit_hook, used to track delayed allocation + * bytes in this file, and to maintain the list of inodes that + * have pending delalloc work to be done. + */ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { @@ -323,6 +348,9 @@ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, return 0; } +/* + * extent_io.c clear_bit_hook, see set_bit_hook for why + */ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { @@ -349,6 +377,10 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, return 0; } +/* + * extent_io.c merge_bio_hook, this must check the chunk tree to make sure + * we don't create bios that span stripes or chunks + */ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio) { @@ -371,6 +403,14 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, return 0; } +/* + * in order to insert checksums into the metadata in large chunks, + * we wait until bio submission time. All the pages in the bio are + * checksummed and sums are attached onto the ordered extent record. + * + * At IO completion time the cums attached on the ordered extent record + * are inserted into the btree + */ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { @@ -383,6 +423,10 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, return btrfs_map_bio(root, rw, bio, mirror_num, 1); } +/* + * extent_io.c submission hook. This does the right thing for csum calculation on write, + * or reading the csums from the tree before a read + */ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num) { @@ -408,6 +452,10 @@ mapit: return btrfs_map_bio(root, rw, bio, mirror_num, 0); } +/* + * given a list of ordered sums record them in the inode. This happens + * at IO completion time based on sums calculated at bio submission time. + */ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, struct inode *inode, u64 file_offset, struct list_head *list) @@ -430,12 +478,12 @@ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end) GFP_NOFS); } +/* see btrfs_writepage_start_hook for details on why this is required */ struct btrfs_writepage_fixup { struct page *page; struct btrfs_work work; }; -/* see btrfs_writepage_start_hook for details on why this is required */ void btrfs_writepage_fixup_worker(struct btrfs_work *work) { struct btrfs_writepage_fixup *fixup; @@ -522,6 +570,10 @@ int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) return -EAGAIN; } +/* as ordered data IO finishes, this gets called so we can finish + * an ordered extent if the range of bytes in the file it covers are + * fully written. + */ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -631,6 +683,14 @@ int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, return btrfs_finish_ordered_io(page->mapping->host, start, end); } +/* + * When IO fails, either with EIO or csum verification fails, we + * try other mirrors that might have a good copy of the data. This + * io_failure_record is used to record state as we go through all the + * mirrors. If another mirror has good data, the page is set up to date + * and things continue. If a good mirror can't be found, the original + * bio end_io callback is called to indicate things have failed. + */ struct io_failure_record { struct page *page; u64 start; @@ -725,6 +785,10 @@ int btrfs_io_failed_hook(struct bio *failed_bio, return 0; } +/* + * each time an IO finishes, we do a fast check in the IO failure tree + * to see if we need to process or clean up an io_failure_record + */ int btrfs_clean_io_failures(struct inode *inode, u64 start) { u64 private; @@ -753,6 +817,11 @@ int btrfs_clean_io_failures(struct inode *inode, u64 start) return 0; } +/* + * when reads are done, we need to check csums to verify the data is correct + * if there's a match, we allow the bio to finish. If not, we go through + * the io_failure_record routines to find good copies + */ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state) { @@ -990,6 +1059,9 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) btrfs_free_path(path); } +/* + * read an inode from the btree into the in-memory inode + */ void btrfs_read_locked_inode(struct inode *inode) { struct btrfs_path *path; @@ -1083,6 +1155,9 @@ make_bad: make_bad_inode(inode); } +/* + * given a leaf and an inode, copy the inode fields into the leaf + */ static void fill_inode_item(struct btrfs_trans_handle *trans, struct extent_buffer *leaf, struct btrfs_inode_item *item, @@ -1118,6 +1193,9 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, BTRFS_I(inode)->block_group->key.objectid); } +/* + * copy everything in the in-memory inode into the btree. + */ int noinline btrfs_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) @@ -1151,6 +1229,11 @@ failed: } +/* + * unlink helper that gets used here in inode.c and in the tree logging + * recovery code. It remove a link in a directory with a given name, and + * also drops the back refs in the inode to the directory + */ int btrfs_unlink_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *dir, struct inode *inode, @@ -1309,7 +1392,7 @@ fail: /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find - * any higher than i_size. + * any higher than new_size * * csum items that cross the new i_size are truncated to the new size * as well. @@ -2123,6 +2206,11 @@ void btrfs_dirty_inode(struct inode *inode) btrfs_end_transaction(trans, root); } +/* + * find the highest existing sequence number in a directory + * and then set the in-memory index_cnt variable to reflect + * free sequence numbers + */ static int btrfs_set_inode_index_count(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -2175,6 +2263,10 @@ out: return ret; } +/* + * helper to find a free sequence number in a given directory. This current + * code is very simple, later versions will do smarter things in the btree + */ static int btrfs_set_inode_index(struct inode *dir, struct inode *inode, u64 *index) { @@ -2305,6 +2397,12 @@ static inline u8 btrfs_inode_type(struct inode *inode) return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT]; } +/* + * utility function to add 'inode' into 'parent_inode' with + * a give name and a given sequence number. + * if 'add_backref' is true, also insert a backref from the + * inode to the parent directory. + */ int btrfs_add_link(struct btrfs_trans_handle *trans, struct inode *parent_inode, struct inode *inode, const char *name, int name_len, int add_backref, u64 index) @@ -2611,6 +2709,10 @@ out_unlock: return err; } +/* helper for btfs_get_extent. Given an existing extent in the tree, + * and an extent that you want to insert, deal with overlap and insert + * the new extent into the tree. + */ static int merge_extent_mapping(struct extent_map_tree *em_tree, struct extent_map *existing, struct extent_map *em, @@ -2627,6 +2729,14 @@ static int merge_extent_mapping(struct extent_map_tree *em_tree, return add_extent_mapping(em_tree, em); } +/* + * a bit scary, this does extent mapping from logical file offset to the disk. + * the ugly parts come from merging extents from the disk with the + * in-ram representation. This gets more complex because of the data=ordered code, + * where the in-ram extents might be locked pending data=ordered completion. + * + * This also copies inline extents directly into the page. + */ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, size_t pg_offset, u64 start, u64 len, int create) @@ -2869,76 +2979,11 @@ out: return em; } -#if 0 /* waiting for O_DIRECT reads */ -static int btrfs_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *bh_result, int create) -{ - struct extent_map *em; - u64 start = (u64)iblock << inode->i_blkbits; - struct btrfs_multi_bio *multi = NULL; - struct btrfs_root *root = BTRFS_I(inode)->root; - u64 len; - u64 logical; - u64 map_length; - int ret = 0; - - em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0); - - if (!em || IS_ERR(em)) - goto out; - - if (em->start > start || em->start + em->len <= start) { - goto out; - } - - if (em->block_start == EXTENT_MAP_INLINE) { - ret = -EINVAL; - goto out; - } - - len = em->start + em->len - start; - len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size))); - - if (em->block_start == EXTENT_MAP_HOLE || - em->block_start == EXTENT_MAP_DELALLOC) { - bh_result->b_size = len; - goto out; - } - - logical = start - em->start; - logical = em->block_start + logical; - - map_length = len; - ret = btrfs_map_block(&root->fs_info->mapping_tree, READ, - logical, &map_length, &multi, 0); - BUG_ON(ret); - bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits; - bh_result->b_size = min(map_length, len); - - bh_result->b_bdev = multi->stripes[0].dev->bdev; - set_buffer_mapped(bh_result); - kfree(multi); -out: - free_extent_map(em); - return ret; -} -#endif - static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t offset, unsigned long nr_segs) { return -EINVAL; -#if 0 - struct file *file = iocb->ki_filp; - struct inode *inode = file->f_mapping->host; - - if (rw == WRITE) - return -EINVAL; - - return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, - offset, nr_segs, btrfs_get_block, NULL); -#endif } static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) @@ -3202,6 +3247,9 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, } } +/* + * create a new subvolume directory/inode (helper for the ioctl). + */ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct btrfs_trans_handle *trans, u64 new_dirid, struct btrfs_block_group_cache *block_group) @@ -3223,6 +3271,9 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, return btrfs_update_inode(trans, new_root, inode); } +/* helper function for file defrag and space balancing. This + * forces readahead on a given range of bytes in an inode + */ unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, pgoff_t offset, pgoff_t last_index) @@ -3424,6 +3475,10 @@ out_unlock: return ret; } +/* + * some fairly slow code that needs optimization. This walks the list + * of all the inodes with pending delalloc and forces them to disk. + */ int btrfs_start_delalloc_inodes(struct btrfs_root *root) { struct list_head *head = &root->fs_info->delalloc_inodes; diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c index 0cc314c10d6..e30aa6e2958 100644 --- a/fs/btrfs/locking.c +++ b/fs/btrfs/locking.c @@ -25,6 +25,15 @@ #include "extent_io.h" #include "locking.h" +/* + * locks the per buffer mutex in an extent buffer. This uses adaptive locks + * and the spin is not tuned very extensively. The spinning does make a big + * difference in almost every workload, but spinning for the right amount of + * time needs some help. + * + * In general, we want to spin as long as the lock holder is doing btree searches, + * and we should give up if they are in more expensive code. + */ int btrfs_tree_lock(struct extent_buffer *eb) { int i; @@ -57,6 +66,10 @@ int btrfs_tree_locked(struct extent_buffer *eb) return mutex_is_locked(&eb->mutex); } +/* + * btrfs_search_slot uses this to decide if it should drop its locks + * before doing something expensive like allocating free blocks for cow. + */ int btrfs_path_lock_waiting(struct btrfs_path *path, int level) { int i; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 951eacff242..dcc1730dd83 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -26,7 +26,6 @@ #include "btrfs_inode.h" #include "extent_io.h" - static u64 entry_end(struct btrfs_ordered_extent *entry) { if (entry->file_offset + entry->len < entry->file_offset) @@ -34,6 +33,9 @@ static u64 entry_end(struct btrfs_ordered_extent *entry) return entry->file_offset + entry->len; } +/* returns NULL if the insertion worked, or it returns the node it did find + * in the tree + */ static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, struct rb_node *node) { @@ -58,6 +60,10 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, return NULL; } +/* + * look for a given offset in the tree, and if it can't be found return the + * first lesser offset + */ static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, struct rb_node **prev_ret) { @@ -108,6 +114,9 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, return NULL; } +/* + * helper to check if a given offset is inside a given entry + */ static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset) { if (file_offset < entry->file_offset || @@ -116,6 +125,10 @@ static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset) return 1; } +/* + * look find the first ordered struct that has this offset, otherwise + * the first one less than this offset + */ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, u64 file_offset) { @@ -305,6 +318,10 @@ int btrfs_remove_ordered_extent(struct inode *inode, return 0; } +/* + * wait for all the ordered extents in a root. This is done when balancing + * space between drives. + */ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only) { struct list_head splice; diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c index 30fcb7aea5b..a50ebb67055 100644 --- a/fs/btrfs/ref-cache.c +++ b/fs/btrfs/ref-cache.c @@ -21,6 +21,16 @@ #include "ref-cache.h" #include "transaction.h" +/* + * leaf refs are used to cache the information about which extents + * a given leaf has references on. This allows us to process that leaf + * in btrfs_drop_snapshot without needing to read it back from disk. + */ + +/* + * kmalloc a leaf reference struct and update the counters for the + * total ref cache size + */ struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root, int nr_extents) { @@ -40,6 +50,10 @@ struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root, return ref; } +/* + * free a leaf reference struct and update the counters for the + * total ref cache size + */ void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { if (!ref) @@ -135,6 +149,10 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen, return 0; } +/* + * find the leaf ref for a given extent. This returns the ref struct with + * a usage reference incremented + */ struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root, u64 bytenr) { @@ -160,6 +178,10 @@ again: return NULL; } +/* + * add a fully filled in leaf ref struct + * remove all the refs older than a given root generation + */ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref, int shared) { @@ -184,6 +206,10 @@ int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref, return ret; } +/* + * remove a single leaf ref from the tree. This drops the ref held by the tree + * only + */ int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) { struct btrfs_leaf_ref_tree *tree; diff --git a/fs/btrfs/ref-cache.h b/fs/btrfs/ref-cache.h index 617564787f5..16f3183d7c5 100644 --- a/fs/btrfs/ref-cache.h +++ b/fs/btrfs/ref-cache.h @@ -19,8 +19,11 @@ #define __REFCACHE__ struct btrfs_extent_info { + /* bytenr and num_bytes find the extent in the extent allocation tree */ u64 bytenr; u64 num_bytes; + + /* objectid and offset find the back reference for the file */ u64 objectid; u64 offset; }; diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index 0091c01abb0..eb7f7655e9d 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -22,8 +22,10 @@ #include "print-tree.h" /* - * returns 0 on finding something, 1 if no more roots are there - * and < 0 on error + * search forward for a root, starting with objectid 'search_start' + * if a root key is found, the objectid we find is filled into 'found_objectid' + * and 0 is returned. < 0 is returned on error, 1 if there is nothing + * left in the tree. */ int btrfs_search_root(struct btrfs_root *root, u64 search_start, u64 *found_objectid) @@ -66,6 +68,11 @@ out: return ret; } +/* + * lookup the root with the highest offset for a given objectid. The key we do + * find is copied into 'key'. If we find something return 0, otherwise 1, < 0 + * on error. + */ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct btrfs_root_item *item, struct btrfs_key *key) { @@ -104,6 +111,9 @@ out: return ret; } +/* + * copy the data in 'item' into the btree + */ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_root_item *item) @@ -147,6 +157,12 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root return ret; } +/* + * at mount time we want to find all the old transaction snapshots that were in + * the process of being deleted if we crashed. This is any root item with an offset + * lower than the latest root. They need to be queued for deletion to finish + * what was happening when we crashed. + */ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, struct btrfs_root *latest) { @@ -227,6 +243,7 @@ err: return ret; } +/* drop the root item for 'key' from 'root' */ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key) { diff --git a/fs/btrfs/struct-funcs.c b/fs/btrfs/struct-funcs.c index ad03a32d111..cdedbe144d4 100644 --- a/fs/btrfs/struct-funcs.c +++ b/fs/btrfs/struct-funcs.c @@ -17,6 +17,27 @@ */ #include + +/* this is some deeply nasty code. ctree.h has a different + * definition for this BTRFS_SETGET_FUNCS macro, behind a #ifndef + * + * The end result is that anyone who #includes ctree.h gets a + * declaration for the btrfs_set_foo functions and btrfs_foo functions + * + * This file declares the macros and then #includes ctree.h, which results + * in cpp creating the function here based on the template below. + * + * These setget functions do all the extent_buffer related mapping + * required to efficiently read and write specific fields in the extent + * buffers. Every pointer to metadata items in btrfs is really just + * an unsigned long offset into the extent buffer which has been + * cast to a specific type. This gives us all the gcc type checking. + * + * The extent buffer api is used to do all the kmapping and page + * spanning work required to get extent buffers in highmem and have + * a metadata blocksize different from the page size. + */ + #define BTRFS_SETGET_FUNCS(name, type, member, bits) \ u##bits btrfs_##name(struct extent_buffer *eb, \ type *s) \ diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 8399d6d05d6..2e6039825b7 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -519,6 +519,9 @@ static struct file_system_type btrfs_fs_type = { .fs_flags = FS_REQUIRES_DEV, }; +/* + * used by btrfsctl to scan devices when no FS is mounted + */ static long btrfs_control_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 444abe0796a..11266d68a6c 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -46,6 +46,9 @@ static noinline void put_transaction(struct btrfs_transaction *transaction) } } +/* + * either allocate a new transaction or hop into the existing one + */ static noinline int join_transaction(struct btrfs_root *root) { struct btrfs_transaction *cur_trans; @@ -85,6 +88,12 @@ static noinline int join_transaction(struct btrfs_root *root) return 0; } +/* + * this does all the record keeping required to make sure that a + * reference counted root is properly recorded in a given transaction. + * This is required to make sure the old root from before we joined the transaction + * is deleted when the transaction commits + */ noinline int btrfs_record_root_in_trans(struct btrfs_root *root) { struct btrfs_dirty_root *dirty; @@ -127,6 +136,10 @@ noinline int btrfs_record_root_in_trans(struct btrfs_root *root) return 0; } +/* wait for commit against the current transaction to become unblocked + * when this is done, it is safe to start a new transaction, but the current + * transaction might not be fully on disk. + */ static void wait_current_trans(struct btrfs_root *root) { struct btrfs_transaction *cur_trans; @@ -198,7 +211,7 @@ struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r, return start_transaction(r, num_blocks, 2); } - +/* wait for a transaction commit to be fully complete */ static noinline int wait_for_commit(struct btrfs_root *root, struct btrfs_transaction *commit) { @@ -218,6 +231,10 @@ static noinline int wait_for_commit(struct btrfs_root *root, return 0; } +/* + * rate limit against the drop_snapshot code. This helps to slow down new operations + * if the drop_snapshot code isn't able to keep up. + */ static void throttle_on_drops(struct btrfs_root *root) { struct btrfs_fs_info *info = root->fs_info; @@ -302,7 +319,11 @@ int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans, return __btrfs_end_transaction(trans, root, 1); } - +/* + * when btree blocks are allocated, they have some corresponding bits set for + * them in one of two extent_io trees. This is used to make sure all of + * those extents are on disk for transaction or log commit + */ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, struct extent_io_tree *dirty_pages) { @@ -393,6 +414,16 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, &trans->transaction->dirty_pages); } +/* + * this is used to update the root pointer in the tree of tree roots. + * + * But, in the case of the extent allocation tree, updating the root + * pointer may allocate blocks which may change the root of the extent + * allocation tree. + * + * So, this loops and repeats and makes sure the cowonly root didn't + * change while the root pointer was being updated in the metadata. + */ static int update_cowonly_root(struct btrfs_trans_handle *trans, struct btrfs_root *root) { @@ -418,6 +449,9 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans, return 0; } +/* + * update all the cowonly tree roots on disk + */ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, struct btrfs_root *root) { @@ -433,6 +467,11 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, return 0; } +/* + * dead roots are old snapshots that need to be deleted. This allocates + * a dirty root struct and adds it into the list of dead roots that need to + * be deleted + */ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest) { struct btrfs_dirty_root *dirty; @@ -449,6 +488,12 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest) return 0; } +/* + * at transaction commit time we need to schedule the old roots for + * deletion via btrfs_drop_snapshot. This runs through all the + * reference counted roots that were modified in the current + * transaction and puts them into the drop list + */ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, struct radix_tree_root *radix, struct list_head *list) @@ -541,6 +586,10 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, return err; } +/* + * defrag a given btree. If cacheonly == 1, this won't read from the disk, + * otherwise every leaf in the btree is read and defragged. + */ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) { struct btrfs_fs_info *info = root->fs_info; @@ -570,6 +619,10 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) return 0; } +/* + * Given a list of roots that need to be deleted, call btrfs_drop_snapshot on + * all of them + */ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, struct list_head *list) { @@ -664,6 +717,10 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, return ret; } +/* + * new snapshots need to be created at a very specific time in the + * transaction commit. This does the actual creation + */ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, struct btrfs_pending_snapshot *pending) @@ -734,6 +791,9 @@ fail: return ret; } +/* + * create all the snapshots we've scheduled for creation + */ static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info) { @@ -944,6 +1004,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, return ret; } +/* + * interface function to delete all the snapshots we have scheduled for deletion + */ int btrfs_clean_old_snapshots(struct btrfs_root *root) { struct list_head dirty_roots; diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index b3bb5bbad76..6f57d0889b1 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -23,6 +23,10 @@ #include "transaction.h" #include "locking.h" +/* defrag all the leaves in a given btree. If cache_only == 1, don't read things + * from disk, otherwise read all the leaves and try to get key order to + * better reflect disk order + */ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only) { -- cgit v1.2.3-70-g09d2 From 323ac95bce442bbde514e3ce57e840402f80d909 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 1 Oct 2008 19:05:46 -0400 Subject: Btrfs: don't read leaf blocks containing only checksums during truncate Checksum items take up a significant portion of the metadata for large files. It is possible to avoid reading them during truncates by checking the keys in the higher level nodes. If a given leaf is followed by another leaf where the lowest key is a checksum item from the same file, we know we can safely delete the leaf without reading it. For a 32GB file on a 6 drive raid0 array, Btrfs needs 8s to delete the file with a cold cache. It is read bound during the run. With this change, Btrfs is able to delete the file in 0.5s Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 69 +++++++++++++------------ fs/btrfs/ctree.h | 4 +- fs/btrfs/inode.c | 152 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 193 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index ff3261ff2e1..2eab4643dcb 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1388,7 +1388,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root struct btrfs_key prealloc_block; lowest_level = p->lowest_level; - WARN_ON(lowest_level && ins_len); + WARN_ON(lowest_level && ins_len > 0); WARN_ON(p->nodes[0] != NULL); WARN_ON(cow && root == root->fs_info->extent_root && !mutex_is_locked(&root->fs_info->alloc_mutex)); @@ -3186,6 +3186,36 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, return ret; } +/* + * a helper function to delete the leaf pointed to by path->slots[1] and + * path->nodes[1]. bytenr is the node block pointer, but since the callers + * already know it, it is faster to have them pass it down than to + * read it out of the node again. + * + * This deletes the pointer in path->nodes[1] and frees the leaf + * block extent. zero is returned if it all worked out, < 0 otherwise. + * + * The path must have already been setup for deleting the leaf, including + * all the proper balancing. path->nodes[1] must be locked. + */ +noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr) +{ + int ret; + u64 root_gen = btrfs_header_generation(path->nodes[1]); + + ret = del_ptr(trans, root, path, 1, path->slots[1]); + if (ret) + return ret; + + ret = btrfs_free_extent(trans, root, bytenr, + btrfs_level_size(root, 0), + path->nodes[1]->start, + btrfs_header_owner(path->nodes[1]), + root_gen, 0, 0, 1); + return ret; +} /* * delete the item at the leaf level in path. If that empties * the leaf, remove it from the tree @@ -3251,17 +3281,8 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (leaf == root->node) { btrfs_set_header_level(leaf, 0); } else { - u64 root_gen = btrfs_header_generation(path->nodes[1]); - wret = del_ptr(trans, root, path, 1, path->slots[1]); - if (wret) - ret = wret; - wret = btrfs_free_extent(trans, root, - leaf->start, leaf->len, - path->nodes[1]->start, - btrfs_header_owner(path->nodes[1]), - root_gen, 0, 0, 1); - if (wret) - ret = wret; + ret = btrfs_del_leaf(trans, root, path, leaf->start); + BUG_ON(ret); } } else { int used = leaf_space_used(leaf, 0, nritems); @@ -3296,24 +3317,10 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, } if (btrfs_header_nritems(leaf) == 0) { - u64 root_gen; - u64 bytenr = leaf->start; - u32 blocksize = leaf->len; - - root_gen = btrfs_header_generation( - path->nodes[1]); - - wret = del_ptr(trans, root, path, 1, slot); - if (wret) - ret = wret; - + path->slots[1] = slot; + ret = btrfs_del_leaf(trans, root, path, leaf->start); + BUG_ON(ret); free_extent_buffer(leaf); - wret = btrfs_free_extent(trans, root, bytenr, - blocksize, path->nodes[1]->start, - btrfs_header_owner(path->nodes[1]), - root_gen, 0, 0, 1); - if (wret) - ret = wret; } else { /* if we're still in the path, make sure * we're dirty. Otherwise, one of the @@ -3418,8 +3425,8 @@ again: level = btrfs_header_level(cur); sret = bin_search(cur, min_key, level, &slot); - /* at level = 0, we're done, setup the path and exit */ - if (level == 0) { + /* at the lowest level, we're done, setup the path and exit */ + if (level == path->lowest_level) { if (slot >= nritems) goto find_next_key; ret = 0; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ded1643c027..94e0cdfddc0 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1649,7 +1649,9 @@ void btrfs_free_path(struct btrfs_path *p); void btrfs_init_path(struct btrfs_path *p); int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int slot, int nr); - +int btrfs_del_leaf(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, u64 bytenr); static inline int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f3abecc2d14..e5c9261dcba 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1389,6 +1389,154 @@ fail: return err; } +/* + * when truncating bytes in a file, it is possible to avoid reading + * the leaves that contain only checksum items. This can be the + * majority of the IO required to delete a large file, but it must + * be done carefully. + * + * The keys in the level just above the leaves are checked to make sure + * the lowest key in a given leaf is a csum key, and starts at an offset + * after the new size. + * + * Then the key for the next leaf is checked to make sure it also has + * a checksum item for the same file. If it does, we know our target leaf + * contains only checksum items, and it can be safely freed without reading + * it. + * + * This is just an optimization targeted at large files. It may do + * nothing. It will return 0 unless things went badly. + */ +static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_path *path, + struct inode *inode, u64 new_size) +{ + struct btrfs_key key; + int ret; + int nritems; + struct btrfs_key found_key; + struct btrfs_key other_key; + + path->lowest_level = 1; + key.objectid = inode->i_ino; + key.type = BTRFS_CSUM_ITEM_KEY; + key.offset = new_size; +again: + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret < 0) + goto out; + + if (path->nodes[1] == NULL) { + ret = 0; + goto out; + } + ret = 0; + btrfs_node_key_to_cpu(path->nodes[1], &found_key, path->slots[1]); + nritems = btrfs_header_nritems(path->nodes[1]); + + if (!nritems) + goto out; + + if (path->slots[1] >= nritems) + goto next_node; + + /* did we find a key greater than anything we want to delete? */ + if (found_key.objectid > inode->i_ino || + (found_key.objectid == inode->i_ino && found_key.type > key.type)) + goto out; + + /* we check the next key in the node to make sure the leave contains + * only checksum items. This comparison doesn't work if our + * leaf is the last one in the node + */ + if (path->slots[1] + 1 >= nritems) { +next_node: + /* search forward from the last key in the node, this + * will bring us into the next node in the tree + */ + btrfs_node_key_to_cpu(path->nodes[1], &found_key, nritems - 1); + + /* unlikely, but we inc below, so check to be safe */ + if (found_key.offset == (u64)-1) + goto out; + + /* search_forward needs a path with locks held, do the + * search again for the original key. It is possible + * this will race with a balance and return a path that + * we could modify, but this drop is just an optimization + * and is allowed to miss some leaves. + */ + btrfs_release_path(root, path); + found_key.offset++; + + /* setup a max key for search_forward */ + other_key.offset = (u64)-1; + other_key.type = key.type; + other_key.objectid = key.objectid; + + path->keep_locks = 1; + ret = btrfs_search_forward(root, &found_key, &other_key, + path, 0, 0); + path->keep_locks = 0; + if (ret || found_key.objectid != key.objectid || + found_key.type != key.type) { + ret = 0; + goto out; + } + + key.offset = found_key.offset; + btrfs_release_path(root, path); + cond_resched(); + goto again; + } + + /* we know there's one more slot after us in the tree, + * read that key so we can verify it is also a checksum item + */ + btrfs_node_key_to_cpu(path->nodes[1], &other_key, path->slots[1] + 1); + + if (found_key.objectid < inode->i_ino) + goto next_key; + + if (found_key.type != key.type || found_key.offset < new_size) + goto next_key; + + /* + * if the key for the next leaf isn't a csum key from this objectid, + * we can't be sure there aren't good items inside this leaf. + * Bail out + */ + if (other_key.objectid != inode->i_ino || other_key.type != key.type) + goto out; + + /* + * it is safe to delete this leaf, it contains only + * csum items from this inode at an offset >= new_size + */ + ret = btrfs_del_leaf(trans, root, path, + btrfs_node_blockptr(path->nodes[1], + path->slots[1])); + BUG_ON(ret); + +next_key: + btrfs_release_path(root, path); + + if (other_key.objectid == inode->i_ino && + other_key.type == key.type && other_key.offset > key.offset) { + key.offset = other_key.offset; + cond_resched(); + goto again; + } + ret = 0; +out: + /* fixup any changes we've made to the path */ + path->lowest_level = 0; + path->keep_locks = 0; + btrfs_release_path(root, path); + return ret; +} + /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find @@ -1436,6 +1584,10 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, key.type = (u8)-1; btrfs_init_path(path); + + ret = drop_csum_leaves(trans, root, path, inode, new_size); + BUG_ON(ret); + search_again: ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) { -- cgit v1.2.3-70-g09d2 From a62b940160d8125016e85046e68ae621c99e751f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Oct 2008 16:31:08 -0400 Subject: Btrfs: cast bio->bi_sector to a u64 before shifting On 32 bit machines without CONFIG_LBD, the bi_sector field is only 32 bits. Btrfs needs to cast it before shifting up, or we end up doing IO into the wrong place. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- fs/btrfs/volumes.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e5c9261dcba..ff0c3597665 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -386,7 +386,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, { struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; struct btrfs_mapping_tree *map_tree; - u64 logical = bio->bi_sector << 9; + u64 logical = (u64)bio->bi_sector << 9; u64 length = 0; u64 map_length; int ret; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index f63cf7621a0..2eed7f91f51 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2187,7 +2187,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, struct btrfs_mapping_tree *map_tree; struct btrfs_device *dev; struct bio *first_bio = bio; - u64 logical = bio->bi_sector << 9; + u64 logical = (u64)bio->bi_sector << 9; u64 length = 0; u64 map_length; struct btrfs_multi_bio *multi = NULL; -- cgit v1.2.3-70-g09d2 From a76a3cd40c1127ca199d4f7f37bf0d541bf44eb2 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 9 Oct 2008 11:46:29 -0400 Subject: Btrfs: Count space allocated to file in bytes This patch makes btrfs count space allocated to file in bytes instead of 512 byte sectors. Everything else in btrfs uses a byte count instead of sector sizes or blocks sizes, so this fits better. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.h | 13 ++----------- fs/btrfs/extent-tree.c | 2 +- fs/btrfs/file.c | 24 +++++++++++++----------- fs/btrfs/inode.c | 23 ++++++++++++----------- fs/btrfs/ioctl.c | 4 ++-- fs/btrfs/tree-log.c | 6 +++--- 6 files changed, 33 insertions(+), 39 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8566eb30f56..50fbcc9ec45 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -411,7 +411,7 @@ struct btrfs_inode_item { /* transid that last touched this inode */ __le64 transid; __le64 size; - __le64 nblocks; + __le64 nbytes; __le64 block_group; __le32 nlink; __le32 uid; @@ -1017,7 +1017,7 @@ BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64); BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); BTRFS_SETGET_FUNCS(inode_transid, struct btrfs_inode_item, transid, 64); BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64); -BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64); +BTRFS_SETGET_FUNCS(inode_nbytes, struct btrfs_inode_item, nbytes, 64); BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64); BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32); BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32); @@ -1814,15 +1814,6 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio); -static inline void dec_i_blocks(struct inode *inode, u64 dec) -{ - dec = dec >> 9; - if (dec <= inode->i_blocks) - inode->i_blocks -= dec; - else - inode->i_blocks = 0; -} - unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, pgoff_t offset, pgoff_t last_index); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 29380467b67..69db54e09fb 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3930,7 +3930,7 @@ next: BUG_ON(ret); btrfs_release_path(root, path); - inode->i_blocks += extent_len >> 9; + inode_add_bytes(inode, extent_len); ext_offset = 0; num_bytes -= extent_len; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index a03d1bbb19a..18dfdf5f91d 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -193,7 +193,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, leaf = path->nodes[0]; ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - inode->i_blocks += (offset + size - found_end) >> 9; + inode_add_bytes(inode, offset + size - found_end); } if (found_end < offset) { ptr = btrfs_file_extent_inline_start(ei) + found_size; @@ -203,7 +203,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, insert: btrfs_release_path(root, path); datasize = offset + size - key.offset; - inode->i_blocks += datasize >> 9; + inode_add_bytes(inode, datasize); datasize = btrfs_file_extent_calc_inline_size(datasize); ret = btrfs_insert_empty_item(trans, root, path, &key, datasize); @@ -713,7 +713,8 @@ next_slot: extent); if (btrfs_file_extent_disk_bytenr(leaf, extent)) { - dec_i_blocks(inode, old_num - new_num); + inode_sub_bytes(inode, old_num - + new_num); } btrfs_set_file_extent_num_bytes(leaf, extent, new_num); @@ -724,14 +725,17 @@ next_slot: u32 new_size; new_size = btrfs_file_extent_calc_inline_size( inline_limit - key.offset); - dec_i_blocks(inode, (extent_end - key.offset) - - (inline_limit - key.offset)); + inode_sub_bytes(inode, extent_end - + inline_limit); btrfs_truncate_item(trans, root, path, new_size, 1); } } /* delete the entire extent */ if (!keep) { + if (found_inline) + inode_sub_bytes(inode, extent_end - + key.offset); ret = btrfs_del_item(trans, root, path); /* TODO update progress marker and return */ BUG_ON(ret); @@ -743,8 +747,7 @@ next_slot: u32 new_size; new_size = btrfs_file_extent_calc_inline_size( extent_end - end); - dec_i_blocks(inode, (extent_end - key.offset) - - (extent_end - end)); + inode_sub_bytes(inode, end - key.offset); ret = btrfs_truncate_item(trans, root, path, new_size, 0); BUG_ON(ret); @@ -791,9 +794,7 @@ next_slot: } btrfs_release_path(root, path); if (disk_bytenr != 0) { - inode->i_blocks += - btrfs_file_extent_num_bytes(leaf, - extent) >> 9; + inode_add_bytes(inode, extent_end - end); } } @@ -801,7 +802,8 @@ next_slot: u64 disk_bytenr = le64_to_cpu(old.disk_bytenr); if (disk_bytenr != 0) { - dec_i_blocks(inode, le64_to_cpu(old.num_bytes)); + inode_sub_bytes(inode, + le64_to_cpu(old.num_bytes)); ret = btrfs_free_extent(trans, root, disk_bytenr, le64_to_cpu(old.disk_num_bytes), diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index ff0c3597665..f9df89c5fdf 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -652,7 +652,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) BUG_ON(ret); btrfs_release_path(root, path); - inode->i_blocks += ordered_extent->len >> 9; + inode_add_bytes(inode, ordered_extent->len); unlock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, GFP_NOFS); @@ -1104,7 +1104,7 @@ void btrfs_read_locked_inode(struct inode *inode) inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec); inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); - inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item); + inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item)); BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item); inode->i_generation = BTRFS_I(inode)->generation; inode->i_rdev = 0; @@ -1184,7 +1184,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item), inode->i_ctime.tv_nsec); - btrfs_set_inode_nblocks(leaf, item, inode->i_blocks); + btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode)); btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation); btrfs_set_inode_transid(leaf, item, trans->transid); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); @@ -1679,7 +1679,7 @@ search_again: num_dec = (orig_num_bytes - extent_num_bytes); if (root->ref_cows && extent_start != 0) - dec_i_blocks(inode, num_dec); + inode_sub_bytes(inode, num_dec); btrfs_mark_buffer_dirty(leaf); } else { extent_num_bytes = @@ -1690,7 +1690,7 @@ search_again: if (extent_start != 0) { found_extent = 1; if (root->ref_cows) - dec_i_blocks(inode, num_dec); + inode_sub_bytes(inode, num_dec); } root_gen = btrfs_header_generation(leaf); root_owner = btrfs_header_owner(leaf); @@ -1700,8 +1700,8 @@ search_again: u32 size = new_size - found_key.offset; if (root->ref_cows) { - dec_i_blocks(inode, item_end + 1 - - found_key.offset - size); + inode_sub_bytes(inode, item_end + 1 - + new_size); } size = btrfs_file_extent_calc_inline_size(size); @@ -1709,8 +1709,8 @@ search_again: size, 1); BUG_ON(ret); } else if (root->ref_cows) { - dec_i_blocks(inode, item_end + 1 - - found_key.offset); + inode_sub_bytes(inode, item_end + 1 - + found_key.offset); } } delete: @@ -2514,7 +2514,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, inode->i_gid = current->fsgid; inode->i_mode = mode; inode->i_ino = objectid; - inode->i_blocks = 0; + inode_set_bytes(inode, 0); inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_inode_item); @@ -3557,7 +3557,8 @@ static int btrfs_getattr(struct vfsmount *mnt, struct inode *inode = dentry->d_inode; generic_fillattr(inode, stat); stat->blksize = PAGE_CACHE_SIZE; - stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9); + stat->blocks = (inode_get_bytes(inode) + + BTRFS_I(inode)->delalloc_bytes) >> 9; return 0; } diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 04de767a8db..ab7a0f61ded 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -99,7 +99,7 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, inode_item->generation = cpu_to_le64(1); inode_item->size = cpu_to_le64(3); inode_item->nlink = cpu_to_le32(1); - inode_item->nblocks = cpu_to_le64(1); + inode_item->nbytes = cpu_to_le64(root->leafsize); inode_item->mode = cpu_to_le32(S_IFDIR | 0755); btrfs_set_root_bytenr(&root_item, leaf->start); @@ -671,7 +671,7 @@ out: btrfs_release_path(root, path); if (ret == 0) { inode->i_mtime = inode->i_ctime = CURRENT_TIME; - inode->i_blocks = src->i_blocks; + inode_set_bytes(inode, inode_get_bytes(src)); btrfs_i_size_write(inode, src->i_size); BTRFS_I(inode)->flags = BTRFS_I(src)->flags; ret = btrfs_update_inode(trans, root, inode); diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 88bbfd959f1..8df719a73d8 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -113,7 +113,7 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans, inode_item->generation = cpu_to_le64(1); inode_item->size = cpu_to_le64(3); inode_item->nlink = cpu_to_le32(1); - inode_item->nblocks = cpu_to_le64(1); + inode_item->nbytes = cpu_to_le64(root->leafsize); inode_item->mode = cpu_to_le32(S_IFDIR | 0755); btrfs_set_root_bytenr(&root_item, leaf->start); @@ -598,8 +598,8 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, ret = overwrite_item(trans, root, path, eb, slot, key); BUG_ON(ret); - /* btrfs_drop_extents changes i_blocks, update it here */ - inode->i_blocks += (extent_end - start) >> 9; + /* btrfs_drop_extents changes i_bytes & i_blocks, update it here */ + inode_add_bytes(inode, extent_end - start); btrfs_update_inode(trans, root, inode); out: if (inode) -- cgit v1.2.3-70-g09d2 From 3bb1a1bc42f2ae9582c28adf620484efcd4da38d Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 9 Oct 2008 11:46:24 -0400 Subject: Btrfs: Remove offset field from struct btrfs_extent_ref The offset field in struct btrfs_extent_ref records the position inside file that file extent is referenced by. In the new back reference system, tree leaves holding references to file extent are recorded explicitly. We can scan these tree leaves very quickly, so the offset field is not required. This patch also makes the back reference system check the objectid when extents are in deleting. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.c | 24 +++---- fs/btrfs/ctree.h | 25 ++----- fs/btrfs/extent-tree.c | 185 +++++++++++++++++++++++-------------------------- fs/btrfs/file.c | 6 +- fs/btrfs/inode.c | 6 +- fs/btrfs/ioctl.c | 2 +- fs/btrfs/print-tree.c | 3 +- fs/btrfs/tree-log.c | 8 +-- 8 files changed, 115 insertions(+), 144 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 2eab4643dcb..9caeb377de6 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -254,8 +254,7 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, ret = btrfs_alloc_reserved_extent(trans, root, parent_start, root->root_key.objectid, - trans->transid, level, 0, - &ins); + trans->transid, level, &ins); BUG_ON(ret); cow = btrfs_init_new_buffer(trans, root, prealloc_dest, buf->len); @@ -333,7 +332,7 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, buf->len, buf->start, root->root_key.objectid, btrfs_header_generation(buf), - 0, 0, 1); + level, 1); } free_extent_buffer(buf); add_root_to_dirty_list(root); @@ -347,7 +346,7 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, WARN_ON(btrfs_header_generation(parent) != trans->transid); btrfs_free_extent(trans, root, buf->start, buf->len, parent_start, btrfs_header_owner(parent), - btrfs_header_generation(parent), 0, 0, 1); + btrfs_header_generation(parent), level, 1); } if (unlock_orig) btrfs_tree_unlock(buf); @@ -927,7 +926,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, ret = btrfs_update_extent_ref(trans, root, child->start, mid->start, child->start, root->root_key.objectid, - trans->transid, level - 1, 0); + trans->transid, level - 1); BUG_ON(ret); add_root_to_dirty_list(root); @@ -940,7 +939,8 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, free_extent_buffer(mid); ret = btrfs_free_extent(trans, root, mid->start, mid->len, mid->start, root->root_key.objectid, - btrfs_header_generation(mid), 0, 0, 1); + btrfs_header_generation(mid), + level, 1); /* once for the root ptr */ free_extent_buffer(mid); return ret; @@ -1006,7 +1006,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, wret = btrfs_free_extent(trans, root, bytenr, blocksize, parent->start, btrfs_header_owner(parent), - generation, 0, 0, 1); + generation, level, 1); if (wret) ret = wret; } else { @@ -1055,7 +1055,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, wret = btrfs_free_extent(trans, root, bytenr, blocksize, parent->start, btrfs_header_owner(parent), - root_gen, 0, 0, 1); + root_gen, level, 1); if (wret) ret = wret; } else { @@ -1691,13 +1691,13 @@ next_level: blocksize, parent->start, btrfs_header_owner(parent), btrfs_header_generation(parent), - level - 1, 0); + level - 1); BUG_ON(ret); ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent->start, btrfs_header_owner(parent), btrfs_header_generation(parent), - level - 1, 0, 1); + level - 1, 1); BUG_ON(ret); if (generation == trans->transid) { @@ -1973,7 +1973,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans, ret = btrfs_update_extent_ref(trans, root, lower->start, lower->start, c->start, root->root_key.objectid, - trans->transid, level - 1, 0); + trans->transid, level - 1); BUG_ON(ret); /* the super has an extra ref to root->node */ @@ -3213,7 +3213,7 @@ noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans, btrfs_level_size(root, 0), path->nodes[1]->start, btrfs_header_owner(path->nodes[1]), - root_gen, 0, 0, 1); + root_gen, 0, 1); return ret; } /* diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 50fbcc9ec45..a37fd783407 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -374,7 +374,6 @@ struct btrfs_extent_ref { __le64 root; __le64 generation; __le64 objectid; - __le64 offset; __le32 num_refs; } __attribute__ ((__packed__)); @@ -1082,7 +1081,6 @@ static inline u8 *btrfs_dev_extent_chunk_tree_uuid(struct btrfs_dev_extent *dev) BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64); BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64); BTRFS_SETGET_FUNCS(ref_objectid, struct btrfs_extent_ref, objectid, 64); -BTRFS_SETGET_FUNCS(ref_offset, struct btrfs_extent_ref, offset, 64); BTRFS_SETGET_FUNCS(ref_num_refs, struct btrfs_extent_ref, num_refs, 32); BTRFS_SETGET_STACK_FUNCS(stack_ref_root, struct btrfs_extent_ref, root, 64); @@ -1090,8 +1088,6 @@ BTRFS_SETGET_STACK_FUNCS(stack_ref_generation, struct btrfs_extent_ref, generation, 64); BTRFS_SETGET_STACK_FUNCS(stack_ref_objectid, struct btrfs_extent_ref, objectid, 64); -BTRFS_SETGET_STACK_FUNCS(stack_ref_offset, struct btrfs_extent_ref, - offset, 64); BTRFS_SETGET_STACK_FUNCS(stack_ref_num_refs, struct btrfs_extent_ref, num_refs, 32); @@ -1522,29 +1518,20 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u32 blocksize); -int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, - u64 bytenr, u64 parent, - u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 parent, u64 min_bytes, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - u64 empty_size, u64 hint_byte, + u64 owner, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, u64 data); int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - struct btrfs_key *ins); + u64 owner, struct btrfs_key *ins); int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - struct btrfs_key *ins); + u64 owner, struct btrfs_key *ins); int btrfs_reserve_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 min_alloc_size, @@ -1563,7 +1550,7 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, int pin); + u64 owner_objectid, int pin); int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len); int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct btrfs_root *root, @@ -1572,12 +1559,12 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset); + u64 owner_objectid); int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 orig_parent, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset); + u64 owner_objectid); int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_free_block_groups(struct btrfs_fs_info *info); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 69db54e09fb..ab36769c356 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -525,31 +525,28 @@ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len) * - Objectid of the subvolume root * - Generation number of the tree holding the reference * - objectid of the file holding the reference - * - offset in the file corresponding to the key holding the reference * - number of references holding by parent node (alway 1 for tree blocks) * * Btree leaf may hold multiple references to a file extent. In most cases, * these references are from same file and the corresponding offsets inside - * the file are close together. So inode objectid and offset in file are - * just hints, they provide hints about where in the btree the references - * can be found and when we can stop searching. + * the file are close together. * * When a file extent is allocated the fields are filled in: - * (root_key.objectid, trans->transid, inode objectid, offset in file, 1) + * (root_key.objectid, trans->transid, inode objectid, 1) * * When a leaf is cow'd new references are added for every file extent found * in the leaf. It looks similar to the create case, but trans->transid will * be different when the block is cow'd. * - * (root_key.objectid, trans->transid, inode objectid, offset in file, + * (root_key.objectid, trans->transid, inode objectid, * number of references in the leaf) * - * Because inode objectid and offset in file are just hints, they are not - * used when backrefs are deleted. When a file extent is removed either - * during snapshot deletion or file truncation, we find the corresponding - * back back reference and check the following fields. + * When a file extent is removed either during snapshot deletion or + * file truncation, we find the corresponding back reference and check + * the following fields: * - * (btrfs_header_owner(leaf), btrfs_header_generation(leaf)) + * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), + * inode objectid) * * Btree extents can be referenced by: * @@ -558,21 +555,21 @@ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len) * * When a tree block is created, back references are inserted: * - * (root->root_key.objectid, trans->transid, level, 0, 1) + * (root->root_key.objectid, trans->transid, level, 1) * * When a tree block is cow'd, new back references are added for all the * blocks it points to. If the tree block isn't in reference counted root, * the old back references are removed. These new back references are of * the form (trans->transid will have increased since creation): * - * (root->root_key.objectid, trans->transid, level, 0, 1) + * (root->root_key.objectid, trans->transid, level, 1) * * When a backref is in deleting, the following fields are checked: * * if backref was for a tree root: - * (btrfs_header_owner(itself), btrfs_header_generation(itself)) + * (btrfs_header_owner(itself), btrfs_header_generation(itself), level) * else - * (btrfs_header_owner(parent), btrfs_header_generation(parent)) + * (btrfs_header_owner(parent), btrfs_header_generation(parent), level) * * Back Reference Key composing: * @@ -584,13 +581,15 @@ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len) static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, - struct btrfs_path *path, u64 bytenr, - u64 parent, u64 ref_root, - u64 ref_generation, int del) + struct btrfs_path *path, + u64 bytenr, u64 parent, + u64 ref_root, u64 ref_generation, + u64 owner_objectid, int del) { struct btrfs_key key; struct btrfs_extent_ref *ref; struct extent_buffer *leaf; + u64 ref_objectid; int ret; key.objectid = bytenr; @@ -607,8 +606,11 @@ static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, leaf = path->nodes[0]; ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); + ref_objectid = btrfs_ref_objectid(leaf, ref); if (btrfs_ref_root(leaf, ref) != ref_root || - btrfs_ref_generation(leaf, ref) != ref_generation) { + btrfs_ref_generation(leaf, ref) != ref_generation || + (ref_objectid != owner_objectid && + ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) { ret = -EIO; WARN_ON(1); goto out; @@ -623,7 +625,7 @@ static int noinline insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_path *path, u64 bytenr, u64 parent, u64 ref_root, u64 ref_generation, - u64 owner_objectid, u64 owner_offset) + u64 owner_objectid) { struct btrfs_key key; struct extent_buffer *leaf; @@ -643,7 +645,6 @@ static int noinline insert_extent_backref(struct btrfs_trans_handle *trans, btrfs_set_ref_root(leaf, ref, ref_root); btrfs_set_ref_generation(leaf, ref, ref_generation); btrfs_set_ref_objectid(leaf, ref, owner_objectid); - btrfs_set_ref_offset(leaf, ref, owner_offset); btrfs_set_ref_num_refs(leaf, ref, 1); } else if (ret == -EEXIST) { u64 existing_owner; @@ -663,14 +664,10 @@ static int noinline insert_extent_backref(struct btrfs_trans_handle *trans, btrfs_set_ref_num_refs(leaf, ref, num_refs + 1); existing_owner = btrfs_ref_objectid(leaf, ref); - if (existing_owner == owner_objectid && - btrfs_ref_offset(leaf, ref) > owner_offset) { - btrfs_set_ref_offset(leaf, ref, owner_offset); - } else if (existing_owner != owner_objectid && - existing_owner != BTRFS_MULTIPLE_OBJECTIDS) { + if (existing_owner != owner_objectid && + existing_owner != BTRFS_MULTIPLE_OBJECTIDS) { btrfs_set_ref_objectid(leaf, ref, BTRFS_MULTIPLE_OBJECTIDS); - btrfs_set_ref_offset(leaf, ref, 0); } ret = 0; } else { @@ -711,7 +708,7 @@ static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans, u64 orig_parent, u64 parent, u64 orig_root, u64 ref_root, u64 orig_generation, u64 ref_generation, - u64 owner_objectid, u64 owner_offset) + u64 owner_objectid) { int ret; struct btrfs_root *extent_root = root->fs_info->extent_root; @@ -762,7 +759,7 @@ static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans, return -ENOMEM; ret = lookup_extent_backref(trans, extent_root, path, bytenr, orig_parent, orig_root, - orig_generation, 1); + orig_generation, owner_objectid, 1); if (ret) goto out; ret = remove_extent_backref(trans, extent_root, path); @@ -770,7 +767,7 @@ static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans, goto out; ret = insert_extent_backref(trans, extent_root, path, bytenr, parent, ref_root, ref_generation, - owner_objectid, owner_offset); + owner_objectid); BUG_ON(ret); finish_current_insert(trans, extent_root); del_pending_extents(trans, extent_root); @@ -783,7 +780,7 @@ int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 orig_parent, u64 parent, u64 ref_root, u64 ref_generation, - u64 owner_objectid, u64 owner_offset) + u64 owner_objectid) { int ret; if (ref_root == BTRFS_TREE_LOG_OBJECTID && @@ -793,7 +790,7 @@ int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, ret = __btrfs_update_extent_ref(trans, root, bytenr, orig_parent, parent, ref_root, ref_root, ref_generation, ref_generation, - owner_objectid, owner_offset); + owner_objectid); maybe_unlock_mutex(root); return ret; } @@ -803,7 +800,7 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, u64 orig_parent, u64 parent, u64 orig_root, u64 ref_root, u64 orig_generation, u64 ref_generation, - u64 owner_objectid, u64 owner_offset) + u64 owner_objectid) { struct btrfs_path *path; int ret; @@ -845,7 +842,7 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, ret = insert_extent_backref(trans, root->fs_info->extent_root, path, bytenr, parent, ref_root, ref_generation, - owner_objectid, owner_offset); + owner_objectid); BUG_ON(ret); finish_current_insert(trans, root->fs_info->extent_root); del_pending_extents(trans, root->fs_info->extent_root); @@ -858,7 +855,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root, u64 ref_generation, - u64 owner_objectid, u64 owner_offset) + u64 owner_objectid) { int ret; if (ref_root == BTRFS_TREE_LOG_OBJECTID && @@ -867,7 +864,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, maybe_lock_mutex(root); ret = __btrfs_inc_extent_ref(trans, root, bytenr, 0, parent, 0, ref_root, 0, ref_generation, - owner_objectid, owner_offset); + owner_objectid); maybe_unlock_mutex(root); return ret; } @@ -1179,7 +1176,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, int ret = 0; int faili = 0; int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *, - u64, u64, u64, u64, u64, u64, u64, u64, u64); + u64, u64, u64, u64, u64, u64, u64, u64); ref_root = btrfs_header_owner(buf); ref_generation = btrfs_header_generation(buf); @@ -1223,7 +1220,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, orig_buf->start, buf->start, orig_root, ref_root, orig_generation, ref_generation, - key.objectid, key.offset); + key.objectid); maybe_unlock_mutex(root); if (ret) { @@ -1238,7 +1235,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, orig_buf->start, buf->start, orig_root, ref_root, orig_generation, ref_generation, - level - 1, 0); + level - 1); maybe_unlock_mutex(root); if (ret) { faili = i; @@ -1314,7 +1311,7 @@ int btrfs_update_ref(struct btrfs_trans_handle *trans, orig_buf->start, buf->start, orig_root, ref_root, orig_generation, ref_generation, - key.objectid, key.offset); + key.objectid); maybe_unlock_mutex(root); if (ret) goto fail; @@ -1325,7 +1322,7 @@ int btrfs_update_ref(struct btrfs_trans_handle *trans, orig_buf->start, buf->start, orig_root, ref_root, orig_generation, ref_generation, - level - 1, 0); + level - 1); maybe_unlock_mutex(root); if (ret) goto fail; @@ -1781,13 +1778,14 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, start, extent_op->parent, extent_root->root_key.objectid, extent_op->generation, - extent_op->level, 0); + extent_op->level); BUG_ON(err); } else if (extent_op->type == PENDING_BACKREF_UPDATE) { err = lookup_extent_backref(trans, extent_root, path, start, extent_op->orig_parent, extent_root->root_key.objectid, - extent_op->orig_generation, 0); + extent_op->orig_generation, + extent_op->level, 0); BUG_ON(err); clear_extent_bits(&info->extent_ins, start, end, @@ -1870,8 +1868,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, - int pin, int mark_free) + u64 owner_objectid, int pin, int mark_free) { struct btrfs_path *path; struct btrfs_key key; @@ -1894,8 +1891,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, return -ENOMEM; path->reada = 1; - ret = lookup_extent_backref(trans, extent_root, path, bytenr, parent, - root_objectid, ref_generation, 1); + ret = lookup_extent_backref(trans, extent_root, path, + bytenr, parent, root_objectid, + ref_generation, owner_objectid, 1); if (ret == 0) { struct btrfs_key found_key; extent_slot = path->slots[0]; @@ -1926,9 +1924,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, btrfs_print_leaf(extent_root, path->nodes[0]); WARN_ON(1); printk("Unable to find ref byte nr %Lu root %Lu " - " gen %Lu owner %Lu offset %Lu\n", bytenr, - root_objectid, ref_generation, owner_objectid, - owner_offset); + "gen %Lu owner %Lu\n", bytenr, + root_objectid, ref_generation, owner_objectid); } leaf = path->nodes[0]; @@ -2068,7 +2065,7 @@ free_extent: extent_op->orig_parent, extent_root->root_key.objectid, extent_op->orig_generation, - extent_op->level, 0, 0, mark_free); + extent_op->level, 0, mark_free); kfree(extent_op); } else { kfree(extent_op); @@ -2107,7 +2104,7 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, int pin) + u64 owner_objectid, int pin) { struct btrfs_root *extent_root = root->fs_info->extent_root; int pending_ret; @@ -2156,8 +2153,8 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, pin = 1; ret = __free_extent(trans, root, bytenr, num_bytes, parent, - root_objectid, ref_generation, owner_objectid, - owner_offset, pin, pin == 0); + root_objectid, ref_generation, + owner_objectid, pin, pin == 0); finish_current_insert(trans, root->fs_info->extent_root); pending_ret = del_pending_extents(trans, root->fs_info->extent_root); @@ -2168,14 +2165,14 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, int pin) + u64 owner_objectid, int pin) { int ret; maybe_lock_mutex(root); ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, parent, root_objectid, ref_generation, - owner_objectid, owner_offset, pin); + owner_objectid, pin); maybe_unlock_mutex(root); return ret; } @@ -2522,8 +2519,7 @@ int btrfs_reserve_extent(struct btrfs_trans_handle *trans, static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - struct btrfs_key *ins) + u64 owner, struct btrfs_key *ins) { int ret; int pending_ret; @@ -2597,7 +2593,6 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, btrfs_set_ref_root(path->nodes[0], ref, root_objectid); btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); btrfs_set_ref_objectid(path->nodes[0], ref, owner); - btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); btrfs_set_ref_num_refs(path->nodes[0], ref, 1); btrfs_mark_buffer_dirty(path->nodes[0]); @@ -2629,17 +2624,15 @@ out: int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - struct btrfs_key *ins) + u64 owner, struct btrfs_key *ins) { int ret; if (root_objectid == BTRFS_TREE_LOG_OBJECTID) return 0; maybe_lock_mutex(root); - ret = __btrfs_alloc_reserved_extent(trans, root, parent, - root_objectid, ref_generation, - owner, owner_offset, ins); + ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid, + ref_generation, owner, ins); update_reserved_extents(root, ins->objectid, ins->offset, 0); maybe_unlock_mutex(root); return ret; @@ -2653,8 +2646,7 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 parent, u64 root_objectid, u64 ref_generation, - u64 owner, u64 owner_offset, - struct btrfs_key *ins) + u64 owner, struct btrfs_key *ins) { int ret; struct btrfs_block_group_cache *block_group; @@ -2665,9 +2657,8 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, ret = btrfs_remove_free_space(block_group, ins->objectid, ins->offset); BUG_ON(ret); - ret = __btrfs_alloc_reserved_extent(trans, root, parent, - root_objectid, ref_generation, - owner, owner_offset, ins); + ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid, + ref_generation, owner, ins); maybe_unlock_mutex(root); return ret; } @@ -2683,8 +2674,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 parent, u64 min_alloc_size, u64 root_objectid, u64 ref_generation, - u64 owner_objectid, u64 owner_offset, - u64 empty_size, u64 hint_byte, + u64 owner_objectid, u64 empty_size, u64 hint_byte, u64 search_end, struct btrfs_key *ins, u64 data) { int ret; @@ -2698,7 +2688,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, if (root_objectid != BTRFS_TREE_LOG_OBJECTID) { ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid, ref_generation, - owner_objectid, owner_offset, ins); + owner_objectid, ins); BUG_ON(ret); } else { @@ -2750,7 +2740,7 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct extent_buffer *buf; ret = btrfs_alloc_extent(trans, root, blocksize, parent, blocksize, - root_objectid, ref_generation, level, 0, + root_objectid, ref_generation, level, empty_size, hint, (u64)-1, &ins, 0); if (ret) { BUG_ON(ret > 0); @@ -2800,7 +2790,7 @@ int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, ret = __btrfs_free_extent(trans, root, disk_bytenr, btrfs_file_extent_disk_num_bytes(leaf, fi), leaf->start, leaf_owner, leaf_generation, - key.objectid, key.offset, 0); + key.objectid, 0); mutex_unlock(&root->fs_info->alloc_mutex); BUG_ON(ret); @@ -2824,7 +2814,7 @@ static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, ret = __btrfs_free_extent(trans, root, info->bytenr, info->num_bytes, ref->bytenr, ref->owner, ref->generation, - info->objectid, info->offset, 0); + info->objectid, 0); mutex_unlock(&root->fs_info->alloc_mutex); atomic_inc(&root->fs_info->throttle_gen); @@ -2940,7 +2930,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, bytenr, blocksize, parent->start, - root_owner, root_gen, 0, 0, 1); + root_owner, root_gen, + *level - 1, 1); BUG_ON(ret); mutex_unlock(&root->fs_info->alloc_mutex); @@ -2970,9 +2961,10 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, *level = 0; break; } - if (printk_ratelimit()) + if (printk_ratelimit()) { printk("leaf ref miss for bytenr %llu\n", (unsigned long long)bytenr); + } } next = btrfs_find_tree_block(root, bytenr, blocksize); if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) { @@ -3020,7 +3012,7 @@ out: mutex_lock(&root->fs_info->alloc_mutex); ret = __btrfs_free_extent(trans, root, bytenr, blocksize, parent->start, root_owner, root_gen, - 0, 0, 1); + *level, 1); mutex_unlock(&root->fs_info->alloc_mutex); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; @@ -3073,8 +3065,8 @@ static int noinline walk_up_tree(struct btrfs_trans_handle *trans, ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, path->nodes[*level]->len, - parent->start, - root_owner, root_gen, 0, 0, 1); + parent->start, root_owner, + root_gen, *level, 1); BUG_ON(ret); free_extent_buffer(path->nodes[*level]); path->nodes[*level] = NULL; @@ -3308,7 +3300,6 @@ struct btrfs_ref_path { u64 root_objectid; u64 root_generation; u64 owner_objectid; - u64 owner_offset; u32 num_refs; int lowest_level; int current_level; @@ -3480,7 +3471,6 @@ found: if (ref_path->lowest_level == level) { ref_path->owner_objectid = ref_objectid; - ref_path->owner_offset = btrfs_ref_offset(leaf, ref); ref_path->num_refs = btrfs_ref_num_refs(leaf, ref); } @@ -3686,16 +3676,20 @@ static int noinline replace_one_extent(struct btrfs_trans_handle *trans, u64 ext_offset; u64 first_pos; u32 nritems; + int nr_scaned = 0; int extent_locked = 0; int ret; - first_pos = ref_path->owner_offset; + memcpy(&key, leaf_key, sizeof(key)); + first_pos = INT_LIMIT(loff_t) - extent_key->offset; if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) { - key.objectid = ref_path->owner_objectid; - key.offset = ref_path->owner_offset; - key.type = BTRFS_EXTENT_DATA_KEY; - } else { - memcpy(&key, leaf_key, sizeof(key)); + if (key.objectid < ref_path->owner_objectid || + (key.objectid == ref_path->owner_objectid && + key.type < BTRFS_EXTENT_DATA_KEY)) { + key.objectid = ref_path->owner_objectid; + key.type = BTRFS_EXTENT_DATA_KEY; + key.offset = 0; + } } while (1) { @@ -3718,8 +3712,7 @@ next: } if (path->slots[0] >= nritems) { - if (ref_path->owner_objectid == - BTRFS_MULTIPLE_OBJECTIDS) + if (++nr_scaned > 2) break; BUG_ON(extent_locked); @@ -3858,7 +3851,7 @@ next: leaf->start, root->root_key.objectid, trans->transid, - key.objectid, key.offset); + key.objectid); BUG_ON(ret); ret = btrfs_free_extent(trans, root, @@ -3867,7 +3860,7 @@ next: leaf->start, btrfs_header_owner(leaf), btrfs_header_generation(leaf), - key.objectid, key.offset, 0); + key.objectid, 0); BUG_ON(ret); btrfs_release_path(root, path); @@ -3925,8 +3918,7 @@ next: new_extents[i].disk_num_bytes, leaf->start, root->root_key.objectid, - trans->transid, - key.objectid, key.offset); + trans->transid, key.objectid); BUG_ON(ret); btrfs_release_path(root, path); @@ -4182,14 +4174,13 @@ static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans, new_extent->disk_num_bytes, leaf->start, root->root_key.objectid, - trans->transid, - key.objectid, key.offset); + trans->transid, key.objectid); BUG_ON(ret); ret = btrfs_free_extent(trans, root, bytenr, num_bytes, leaf->start, btrfs_header_owner(leaf), btrfs_header_generation(leaf), - key.objectid, key.offset, 0); + key.objectid, 0); BUG_ON(ret); cond_resched(); } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 18dfdf5f91d..69abbe19add 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -788,8 +788,7 @@ next_slot: le64_to_cpu(old.disk_num_bytes), leaf->start, root->root_key.objectid, - trans->transid, - ins.objectid, ins.offset); + trans->transid, ins.objectid); BUG_ON(ret); } btrfs_release_path(root, path); @@ -808,8 +807,7 @@ next_slot: disk_bytenr, le64_to_cpu(old.disk_num_bytes), leaf_start, root_owner, - root_gen, key.objectid, - key.offset, 0); + root_gen, key.objectid, 0); BUG_ON(ret); *hint_byte = disk_bytenr; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f9df89c5fdf..3ab147dc3c0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -647,8 +647,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ins.type = BTRFS_EXTENT_ITEM_KEY; ret = btrfs_alloc_reserved_extent(trans, root, leaf->start, root->root_key.objectid, - trans->transid, inode->i_ino, - ordered_extent->file_offset, &ins); + trans->transid, inode->i_ino, &ins); BUG_ON(ret); btrfs_release_path(root, path); @@ -1734,8 +1733,7 @@ delete: ret = btrfs_free_extent(trans, root, extent_start, extent_num_bytes, leaf->start, root_owner, - root_gen, inode->i_ino, - found_key.offset, 0); + root_gen, inode->i_ino, 0); BUG_ON(ret); } next: diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index ab7a0f61ded..50c8a066d1f 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -658,7 +658,7 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) ds, dl, leaf->start, root->root_key.objectid, trans->transid, - inode->i_ino, key.offset); + inode->i_ino); BUG_ON(ret); } } diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 3577badfa5b..bd9ab3e9a7f 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -102,11 +102,10 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) case BTRFS_EXTENT_REF_KEY: ref = btrfs_item_ptr(l, i, struct btrfs_extent_ref); printk("\t\textent back ref root %llu gen %llu " - "owner %llu offset %llu num_refs %lu\n", + "owner %llu num_refs %lu\n", (unsigned long long)btrfs_ref_root(l, ref), (unsigned long long)btrfs_ref_generation(l, ref), (unsigned long long)btrfs_ref_objectid(l, ref), - (unsigned long long)btrfs_ref_offset(l, ref), (unsigned long)btrfs_ref_num_refs(l, ref)); break; diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 8df719a73d8..cf618cc8b34 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -460,8 +460,7 @@ insert: ins.objectid, ins.offset, path->nodes[0]->start, root->root_key.objectid, - trans->transid, - key->objectid, key->offset); + trans->transid, key->objectid); } else { /* * insert the extent pointer in the extent @@ -471,7 +470,7 @@ insert: path->nodes[0]->start, root->root_key.objectid, trans->transid, key->objectid, - key->offset, &ins); + &ins); BUG_ON(ret); } } @@ -2534,8 +2533,7 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, dst_path->nodes[0]->start, BTRFS_TREE_LOG_OBJECTID, trans->transid, - ins_keys[i].objectid, - ins_keys[i].offset); + ins_keys[i].objectid); BUG_ON(ret); } } -- cgit v1.2.3-70-g09d2 From 5b84e8d6eef82fcf7b1b16d92e29375b85b6549a Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 9 Oct 2008 11:46:19 -0400 Subject: Btrfs: Fix leaf reference cache miss Due to the optimization for truncate, tree leaves only containing checksum items can be deleted without being COW'ed first. This causes reference cache misses. The way to fix the miss is create cache entries for tree leaves only contain checksum. This patch also fixes a -EEXIST issue in shared reference cache. Signed-off-by: Yan Zheng --- fs/btrfs/extent-tree.c | 8 ++++++++ fs/btrfs/inode.c | 26 +++++++++++++++++++++++--- 2 files changed, 31 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index ab36769c356..280ac1aa9b6 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1151,6 +1151,14 @@ int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, } ret = btrfs_add_leaf_ref(root, ref, shared); + if (ret == -EEXIST && shared) { + struct btrfs_leaf_ref *old; + old = btrfs_lookup_leaf_ref(root, ref->bytenr); + BUG_ON(!old); + btrfs_remove_leaf_ref(root, old); + btrfs_free_leaf_ref(root, old); + ret = btrfs_add_leaf_ref(root, ref, shared); + } WARN_ON(ret); btrfs_free_leaf_ref(root, ref); } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3ab147dc3c0..11bfe131fde 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -48,6 +48,7 @@ #include "xattr.h" #include "compat.h" #include "tree-log.h" +#include "ref-cache.h" struct btrfs_iget_args { u64 ino; @@ -1416,6 +1417,9 @@ static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans, int nritems; struct btrfs_key found_key; struct btrfs_key other_key; + struct btrfs_leaf_ref *ref; + u64 leaf_gen; + u64 leaf_start; path->lowest_level = 1; key.objectid = inode->i_ino; @@ -1509,15 +1513,31 @@ next_node: if (other_key.objectid != inode->i_ino || other_key.type != key.type) goto out; + leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]); + leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]); /* * it is safe to delete this leaf, it contains only * csum items from this inode at an offset >= new_size */ - ret = btrfs_del_leaf(trans, root, path, - btrfs_node_blockptr(path->nodes[1], - path->slots[1])); + ret = btrfs_del_leaf(trans, root, path, leaf_start); BUG_ON(ret); + if (root->ref_cows && leaf_gen < trans->transid) { + ref = btrfs_alloc_leaf_ref(root, 0); + if (ref) { + ref->root_gen = root->root_key.offset; + ref->bytenr = leaf_start; + ref->owner = 0; + ref->generation = leaf_gen; + ref->nritems = 0; + + ret = btrfs_add_leaf_ref(root, ref, 0); + WARN_ON(ret); + btrfs_free_leaf_ref(root, ref); + } else { + WARN_ON(1); + } + } next_key: btrfs_release_path(root, path); -- cgit v1.2.3-70-g09d2 From cb8e70901d36f32017614f16d2cf7cc089544574 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 9 Oct 2008 13:39:39 -0400 Subject: Btrfs: Fix subvolume creation locking rules Creating a subvolume is in many ways like a normal VFS ->mkdir, and we really need to play with the VFS topology locking rules. So instead of just creating the snapshot on disk and then later getting rid of confliting aliases do it correctly from the start. This will become especially important once we allow for subvolumes anywhere in the tree, and not just below a hidden root. Note that snapshots will need the same treatment, but do to the delay in creating them we can't do it currently. Chris promised to fix that issue, so I'll wait on that. Signed-off-by: Christoph Hellwig --- fs/btrfs/ctree.h | 2 +- fs/btrfs/inode.c | 10 ++++-- fs/btrfs/ioctl.c | 102 +++++++++++++++++++++++++++++++++++++++++++++++++------ 3 files changed, 100 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 3fa9b8d6751..8559f39fd47 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1791,7 +1791,7 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root); int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end); int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc); -int btrfs_create_subvol_root(struct btrfs_root *new_root, +int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, struct btrfs_trans_handle *trans, u64 new_dirid, struct btrfs_block_group_cache *block_group); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 11bfe131fde..bf4bed6ca4d 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3420,11 +3420,12 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, /* * create a new subvolume directory/inode (helper for the ioctl). */ -int btrfs_create_subvol_root(struct btrfs_root *new_root, +int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, struct btrfs_trans_handle *trans, u64 new_dirid, struct btrfs_block_group_cache *block_group) { struct inode *inode; + int error; u64 index = 0; inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, @@ -3438,7 +3439,12 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, inode->i_nlink = 1; btrfs_i_size_write(inode, 0); - return btrfs_update_inode(trans, new_root, inode); + error = btrfs_update_inode(trans, new_root, inode); + if (error) + return error; + + d_instantiate(dentry, inode); + return 0; } /* helper function for file defrag and space balancing. This diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 50c8a066d1f..3d85f18bbba 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include @@ -28,12 +29,15 @@ #include #include #include +#include #include +#include #include #include #include #include #include +#include #include #include #include @@ -48,8 +52,9 @@ -static noinline int create_subvol(struct btrfs_root *root, char *name, - int namelen) +static noinline int create_subvol(struct btrfs_root *root, + struct dentry *dentry, + char *name, int namelen) { struct btrfs_trans_handle *trans; struct btrfs_key key; @@ -151,14 +156,11 @@ static noinline int create_subvol(struct btrfs_root *root, char *name, trans = btrfs_start_transaction(new_root, 1); BUG_ON(!trans); - ret = btrfs_create_subvol_root(new_root, trans, new_dirid, + ret = btrfs_create_subvol_root(new_root, dentry, trans, new_dirid, BTRFS_I(dir)->block_group); if (ret) goto fail; - /* Invalidate existing dcache entry for new subvolume. */ - btrfs_invalidate_dcache_root(root, name, namelen); - fail: nr = trans->blocks_used; err = btrfs_commit_transaction(trans, new_root); @@ -210,6 +212,79 @@ fail_unlock: return ret; } +/* copy of may_create in fs/namei.c() */ +static inline int btrfs_may_create(struct inode *dir, struct dentry *child) +{ + if (child->d_inode) + return -EEXIST; + if (IS_DEADDIR(dir)) + return -ENOENT; + return inode_permission(dir, MAY_WRITE | MAY_EXEC); +} + +/* + * Create a new subvolume below @parent. This is largely modeled after + * sys_mkdirat and vfs_mkdir, but we only do a single component lookup + * inside this filesystem so it's quite a bit simpler. + */ +static noinline int btrfs_mksubvol(struct path *parent, char *name, + int mode, int namelen) +{ + struct dentry *dentry; + int error; + + mutex_lock_nested(&parent->dentry->d_inode->i_mutex, I_MUTEX_PARENT); + + dentry = lookup_one_len(name, parent->dentry, namelen); + error = PTR_ERR(dentry); + if (IS_ERR(dentry)) + goto out_unlock; + + error = -EEXIST; + if (dentry->d_inode) + goto out_dput; + + if (!IS_POSIXACL(parent->dentry->d_inode)) + mode &= ~current->fs->umask; + error = mnt_want_write(parent->mnt); + if (error) + goto out_dput; + + error = btrfs_may_create(parent->dentry->d_inode, dentry); + if (error) + goto out_drop_write; + + mode &= (S_IRWXUGO|S_ISVTX); + error = security_inode_mkdir(parent->dentry->d_inode, dentry, mode); + if (error) + goto out_drop_write; + + /* + * Actually perform the low-level subvolume creation after all + * this VFS fuzz. + * + * Eventually we want to pass in an inode under which we create this + * subvolume, but for now all are under the filesystem root. + * + * Also we should pass on the mode eventually to allow creating new + * subvolume with specific mode bits. + */ + error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, dentry, + name, namelen); + if (error) + goto out_drop_write; + + fsnotify_mkdir(parent->dentry->d_inode, dentry); +out_drop_write: + mnt_drop_write(parent->mnt); +out_dput: + dput(dentry); +out_unlock: + mutex_unlock(&parent->dentry->d_inode->i_mutex); + return error; +} + + int btrfs_defrag_file(struct file *file) { struct inode *inode = fdentry(file)->d_inode; @@ -395,9 +470,10 @@ out: return ret; } -static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, +static noinline int btrfs_ioctl_snap_create(struct file *file, void __user *arg) { + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; struct btrfs_ioctl_vol_args *vol_args; struct btrfs_dir_item *di; struct btrfs_path *path; @@ -444,10 +520,14 @@ static noinline int btrfs_ioctl_snap_create(struct btrfs_root *root, goto out; } - if (root == root->fs_info->tree_root) - ret = create_subvol(root, vol_args->name, namelen); - else + if (root == root->fs_info->tree_root) { + ret = btrfs_mksubvol(&file->f_path, vol_args->name, + file->f_path.dentry->d_inode->i_mode, + namelen); + } else { ret = create_snapshot(root, vol_args->name, namelen); + } + out: kfree(vol_args); return ret; @@ -761,7 +841,7 @@ long btrfs_ioctl(struct file *file, unsigned int switch (cmd) { case BTRFS_IOC_SNAP_CREATE: - return btrfs_ioctl_snap_create(root, (void __user *)arg); + return btrfs_ioctl_snap_create(file, (void __user *)arg); case BTRFS_IOC_DEFRAG: return btrfs_ioctl_defrag(file); case BTRFS_IOC_RESIZE: -- cgit v1.2.3-70-g09d2 From c8b978188c9a0fd3d535c13debd19d522b726f1f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 29 Oct 2008 14:49:59 -0400 Subject: Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason --- fs/Kconfig | 2 + fs/btrfs/Makefile | 3 +- fs/btrfs/compression.c | 454 ++++++++++++++++++++++++++++++++++ fs/btrfs/compression.h | 47 ++++ fs/btrfs/ctree.h | 99 ++++++-- fs/btrfs/disk-io.c | 18 +- fs/btrfs/disk-io.h | 1 + fs/btrfs/extent-tree.c | 27 +- fs/btrfs/extent_io.c | 411 ++++++++++++++++++++++++------- fs/btrfs/extent_io.h | 17 +- fs/btrfs/extent_map.c | 9 + fs/btrfs/extent_map.h | 6 +- fs/btrfs/file-item.c | 75 +++++- fs/btrfs/file.c | 263 +++++--------------- fs/btrfs/inode.c | 584 ++++++++++++++++++++++++++++++++++++++++---- fs/btrfs/ordered-data.c | 9 +- fs/btrfs/ordered-data.h | 10 +- fs/btrfs/print-tree.c | 7 +- fs/btrfs/super.c | 10 +- fs/btrfs/tree-log.c | 3 +- fs/btrfs/volumes.c | 2 + fs/btrfs/zlib.c | 637 ++++++++++++++++++++++++++++++++++++++++++++++++ 22 files changed, 2315 insertions(+), 379 deletions(-) create mode 100644 fs/btrfs/compression.c create mode 100644 fs/btrfs/compression.h create mode 100644 fs/btrfs/zlib.c (limited to 'fs/btrfs/inode.c') diff --git a/fs/Kconfig b/fs/Kconfig index 18f5a85b47c..31cce5d88b1 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -501,6 +501,8 @@ config BTRFS_FS tristate "Btrfs filesystem (EXPERIMENTAL) Unstable disk format" depends on EXPERIMENTAL select LIBCRC32C + select ZLIB_INFLATE + select ZLIB_DEFLATE help Btrfs is a new filesystem with extents, writable snapshotting, support for multiple devices and many more features. diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile index 7125716e142..d2cf5a54a4b 100644 --- a/fs/btrfs/Makefile +++ b/fs/btrfs/Makefile @@ -7,7 +7,8 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \ transaction.o inode.o file.o tree-defrag.o \ extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \ extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \ - ref-cache.o export.o tree-log.o acl.o free-space-cache.o + ref-cache.o export.o tree-log.o acl.o free-space-cache.o zlib.o \ + compression.o else # Normal Makefile diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c new file mode 100644 index 00000000000..c5470367ca5 --- /dev/null +++ b/fs/btrfs/compression.c @@ -0,0 +1,454 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "ctree.h" +#include "disk-io.h" +#include "transaction.h" +#include "btrfs_inode.h" +#include "volumes.h" +#include "ordered-data.h" +#include "compat.h" +#include "compression.h" +#include "extent_io.h" +#include "extent_map.h" + +struct compressed_bio { + /* number of bios pending for this compressed extent */ + atomic_t pending_bios; + + /* the pages with the compressed data on them */ + struct page **compressed_pages; + + /* inode that owns this data */ + struct inode *inode; + + /* starting offset in the inode for our pages */ + u64 start; + + /* number of bytes in the inode we're working on */ + unsigned long len; + + /* number of bytes on disk */ + unsigned long compressed_len; + + /* number of compressed pages in the array */ + unsigned long nr_pages; + + /* IO errors */ + int errors; + + /* for reads, this is the bio we are copying the data into */ + struct bio *orig_bio; +}; + +static struct bio *compressed_bio_alloc(struct block_device *bdev, + u64 first_byte, gfp_t gfp_flags) +{ + struct bio *bio; + int nr_vecs; + + nr_vecs = bio_get_nr_vecs(bdev); + bio = bio_alloc(gfp_flags, nr_vecs); + + if (bio == NULL && (current->flags & PF_MEMALLOC)) { + while (!bio && (nr_vecs /= 2)) + bio = bio_alloc(gfp_flags, nr_vecs); + } + + if (bio) { + bio->bi_size = 0; + bio->bi_bdev = bdev; + bio->bi_sector = first_byte >> 9; + } + return bio; +} + +/* when we finish reading compressed pages from the disk, we + * decompress them and then run the bio end_io routines on the + * decompressed pages (in the inode address space). + * + * This allows the checksumming and other IO error handling routines + * to work normally + * + * The compressed pages are freed here, and it must be run + * in process context + */ +static void end_compressed_bio_read(struct bio *bio, int err) +{ + struct extent_io_tree *tree; + struct compressed_bio *cb = bio->bi_private; + struct inode *inode; + struct page *page; + unsigned long index; + int ret; + + if (err) + cb->errors = 1; + + /* if there are more bios still pending for this compressed + * extent, just exit + */ + if (!atomic_dec_and_test(&cb->pending_bios)) + goto out; + + /* ok, we're the last bio for this extent, lets start + * the decompression. + */ + inode = cb->inode; + tree = &BTRFS_I(inode)->io_tree; + ret = btrfs_zlib_decompress_biovec(cb->compressed_pages, + cb->start, + cb->orig_bio->bi_io_vec, + cb->orig_bio->bi_vcnt, + cb->compressed_len); + if (ret) + cb->errors = 1; + + /* release the compressed pages */ + index = 0; + for (index = 0; index < cb->nr_pages; index++) { + page = cb->compressed_pages[index]; + page->mapping = NULL; + page_cache_release(page); + } + + /* do io completion on the original bio */ + if (cb->errors) + bio_io_error(cb->orig_bio); + else + bio_endio(cb->orig_bio, 0); + + /* finally free the cb struct */ + kfree(cb->compressed_pages); + kfree(cb); +out: + bio_put(bio); +} + +/* + * Clear the writeback bits on all of the file + * pages for a compressed write + */ +static noinline int end_compressed_writeback(struct inode *inode, u64 start, + unsigned long ram_size) +{ + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = (start + ram_size - 1) >> PAGE_CACHE_SHIFT; + struct page *pages[16]; + unsigned long nr_pages = end_index - index + 1; + int i; + int ret; + + while(nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min(nr_pages, ARRAY_SIZE(pages)), pages); + if (ret == 0) { + nr_pages -= 1; + index += 1; + continue; + } + for (i = 0; i < ret; i++) { + end_page_writeback(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + } + /* the inode may be gone now */ + return 0; +} + +/* + * do the cleanup once all the compressed pages hit the disk. + * This will clear writeback on the file pages and free the compressed + * pages. + * + * This also calls the writeback end hooks for the file pages so that + * metadata and checksums can be updated in the file. + */ +static void end_compressed_bio_write(struct bio *bio, int err) +{ + struct extent_io_tree *tree; + struct compressed_bio *cb = bio->bi_private; + struct inode *inode; + struct page *page; + unsigned long index; + + if (err) + cb->errors = 1; + + /* if there are more bios still pending for this compressed + * extent, just exit + */ + if (!atomic_dec_and_test(&cb->pending_bios)) + goto out; + + /* ok, we're the last bio for this extent, step one is to + * call back into the FS and do all the end_io operations + */ + inode = cb->inode; + tree = &BTRFS_I(inode)->io_tree; + tree->ops->writepage_end_io_hook(cb->compressed_pages[0], + cb->start, + cb->start + cb->len - 1, + NULL, 1); + + end_compressed_writeback(inode, cb->start, cb->len); + /* note, our inode could be gone now */ + + /* + * release the compressed pages, these came from alloc_page and + * are not attached to the inode at all + */ + index = 0; + for (index = 0; index < cb->nr_pages; index++) { + page = cb->compressed_pages[index]; + page->mapping = NULL; + page_cache_release(page); + } + + /* finally free the cb struct */ + kfree(cb->compressed_pages); + kfree(cb); +out: + bio_put(bio); +} + +/* + * worker function to build and submit bios for previously compressed pages. + * The corresponding pages in the inode should be marked for writeback + * and the compressed pages should have a reference on them for dropping + * when the IO is complete. + * + * This also checksums the file bytes and gets things ready for + * the end io hooks. + */ +int btrfs_submit_compressed_write(struct inode *inode, u64 start, + unsigned long len, u64 disk_start, + unsigned long compressed_len, + struct page **compressed_pages, + unsigned long nr_pages) +{ + struct bio *bio = NULL; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct compressed_bio *cb; + unsigned long bytes_left; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + int page_index = 0; + struct page *page; + u64 first_byte = disk_start; + struct block_device *bdev; + int ret; + + WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1)); + cb = kmalloc(sizeof(*cb), GFP_NOFS); + atomic_set(&cb->pending_bios, 0); + cb->errors = 0; + cb->inode = inode; + cb->start = start; + cb->len = len; + cb->compressed_pages = compressed_pages; + cb->compressed_len = compressed_len; + cb->orig_bio = NULL; + cb->nr_pages = nr_pages; + + bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; + + ret = btrfs_csum_file_bytes(root, inode, start, len); + BUG_ON(ret); + + bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS); + bio->bi_private = cb; + bio->bi_end_io = end_compressed_bio_write; + atomic_inc(&cb->pending_bios); + + /* create and submit bios for the compressed pages */ + bytes_left = compressed_len; + while(bytes_left > 0) { + page = compressed_pages[page_index]; + page->mapping = inode->i_mapping; + if (bio->bi_size) + ret = io_tree->ops->merge_bio_hook(page, 0, + PAGE_CACHE_SIZE, + bio, 0); + else + ret = 0; + + if (ret || bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < + PAGE_CACHE_SIZE) { + bio_get(bio); + + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); + BUG_ON(ret); + + ret = btrfs_map_bio(root, WRITE, bio, 0, 1); + BUG_ON(ret); + + bio_put(bio); + + bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS); + atomic_inc(&cb->pending_bios); + bio->bi_private = cb; + bio->bi_end_io = end_compressed_bio_write; + bio_add_page(bio, page, PAGE_CACHE_SIZE, 0); + } + page_index++; + bytes_left -= PAGE_CACHE_SIZE; + first_byte += PAGE_CACHE_SIZE; + } + bio_get(bio); + + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); + BUG_ON(ret); + + ret = btrfs_map_bio(root, WRITE, bio, 0, 1); + BUG_ON(ret); + + bio_put(bio); + return 0; +} + +/* + * for a compressed read, the bio we get passed has all the inode pages + * in it. We don't actually do IO on those pages but allocate new ones + * to hold the compressed pages on disk. + * + * bio->bi_sector points to the compressed extent on disk + * bio->bi_io_vec points to all of the inode pages + * bio->bi_vcnt is a count of pages + * + * After the compressed pages are read, we copy the bytes into the + * bio we were passed and then call the bio end_io calls + */ +int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, + int mirror_num, unsigned long bio_flags) +{ + struct extent_io_tree *tree; + struct extent_map_tree *em_tree; + struct compressed_bio *cb; + struct btrfs_root *root = BTRFS_I(inode)->root; + unsigned long uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE; + unsigned long compressed_len; + unsigned long nr_pages; + unsigned long page_index; + struct page *page; + struct block_device *bdev; + struct bio *comp_bio; + u64 cur_disk_byte = (u64)bio->bi_sector << 9; + struct extent_map *em; + int ret; + + tree = &BTRFS_I(inode)->io_tree; + em_tree = &BTRFS_I(inode)->extent_tree; + + /* we need the actual starting offset of this extent in the file */ + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, + page_offset(bio->bi_io_vec->bv_page), + PAGE_CACHE_SIZE); + spin_unlock(&em_tree->lock); + + cb = kmalloc(sizeof(*cb), GFP_NOFS); + atomic_set(&cb->pending_bios, 0); + cb->errors = 0; + cb->inode = inode; + + cb->start = em->start; + compressed_len = em->block_len; + free_extent_map(em); + + cb->len = uncompressed_len; + cb->compressed_len = compressed_len; + cb->orig_bio = bio; + + nr_pages = (compressed_len + PAGE_CACHE_SIZE - 1) / + PAGE_CACHE_SIZE; + cb->compressed_pages = kmalloc(sizeof(struct page *) * nr_pages, + GFP_NOFS); + bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; + + for (page_index = 0; page_index < nr_pages; page_index++) { + cb->compressed_pages[page_index] = alloc_page(GFP_NOFS | + __GFP_HIGHMEM); + } + cb->nr_pages = nr_pages; + + comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, GFP_NOFS); + comp_bio->bi_private = cb; + comp_bio->bi_end_io = end_compressed_bio_read; + atomic_inc(&cb->pending_bios); + + for (page_index = 0; page_index < nr_pages; page_index++) { + page = cb->compressed_pages[page_index]; + page->mapping = inode->i_mapping; + if (comp_bio->bi_size) + ret = tree->ops->merge_bio_hook(page, 0, + PAGE_CACHE_SIZE, + comp_bio, 0); + else + ret = 0; + + if (ret || bio_add_page(comp_bio, page, PAGE_CACHE_SIZE, 0) < + PAGE_CACHE_SIZE) { + bio_get(comp_bio); + + ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); + BUG_ON(ret); + + ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); + BUG_ON(ret); + + bio_put(comp_bio); + + comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, + GFP_NOFS); + atomic_inc(&cb->pending_bios); + bio->bi_private = cb; + bio->bi_end_io = end_compressed_bio_write; + bio_add_page(bio, page, PAGE_CACHE_SIZE, 0); + } + cur_disk_byte += PAGE_CACHE_SIZE; + } + bio_get(comp_bio); + + ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); + BUG_ON(ret); + + ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); + BUG_ON(ret); + + bio_put(comp_bio); + return 0; +} diff --git a/fs/btrfs/compression.h b/fs/btrfs/compression.h new file mode 100644 index 00000000000..421f5b4aa71 --- /dev/null +++ b/fs/btrfs/compression.h @@ -0,0 +1,47 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + */ + +#ifndef __BTRFS_COMPRESSION_ +#define __BTRFS_COMPRESSION_ + +int btrfs_zlib_decompress(unsigned char *data_in, + struct page *dest_page, + unsigned long start_byte, + size_t srclen, size_t destlen); +int btrfs_zlib_compress_pages(struct address_space *mapping, + u64 start, unsigned long len, + struct page **pages, + unsigned long nr_dest_pages, + unsigned long *out_pages, + unsigned long *total_in, + unsigned long *total_out, + unsigned long max_out); +int btrfs_zlib_decompress_biovec(struct page **pages_in, + u64 disk_start, + struct bio_vec *bvec, + int vcnt, + size_t srclen); +void btrfs_zlib_exit(void); +int btrfs_submit_compressed_write(struct inode *inode, u64 start, + unsigned long len, u64 disk_start, + unsigned long compressed_len, + struct page **compressed_pages, + unsigned long nr_pages); +int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, + int mirror_num, unsigned long bio_flags); +#endif diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8559f39fd47..793d8fdda24 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -400,10 +400,18 @@ struct btrfs_timespec { __le32 nsec; } __attribute__ ((__packed__)); -/* - * there is no padding here on purpose. If you want to extent the inode, - * make a new item type - */ +typedef enum { + BTRFS_COMPRESS_NONE = 0, + BTRFS_COMPRESS_ZLIB = 1, + BTRFS_COMPRESS_LAST = 2, +} btrfs_compression_type; + +/* we don't understand any encryption methods right now */ +typedef enum { + BTRFS_ENCRYPTION_NONE = 0, + BTRFS_ENCRYPTION_LAST = 1, +} btrfs_encryption_type; + struct btrfs_inode_item { /* nfs style generation number */ __le64 generation; @@ -419,6 +427,7 @@ struct btrfs_inode_item { __le64 rdev; __le16 flags; __le16 compat_flags; + struct btrfs_timespec atime; struct btrfs_timespec ctime; struct btrfs_timespec mtime; @@ -454,8 +463,33 @@ struct btrfs_root_item { #define BTRFS_FILE_EXTENT_INLINE 1 struct btrfs_file_extent_item { + /* + * transaction id that created this extent + */ __le64 generation; + /* + * max number of bytes to hold this extent in ram + * when we split a compressed extent we can't know how big + * each of the resulting pieces will be. So, this is + * an upper limit on the size of the extent in ram instead of + * an exact limit. + */ + __le64 ram_bytes; + + /* + * 32 bits for the various ways we might encode the data, + * including compression and encryption. If any of these + * are set to something a given disk format doesn't understand + * it is treated like an incompat flag for reading and writing, + * but not for stat. + */ + u8 compression; + u8 encryption; + __le16 other_encoding; /* spare for later use */ + + /* are we inline data or a real extent? */ u8 type; + /* * disk space consumed by the extent, checksum blocks are included * in these numbers @@ -471,9 +505,11 @@ struct btrfs_file_extent_item { */ __le64 offset; /* - * the logical number of file blocks (no csums included) + * the logical number of file blocks (no csums included). This + * always reflects the size uncompressed and without encoding. */ __le64 num_bytes; + } __attribute__ ((__packed__)); struct btrfs_csum_item { @@ -814,6 +850,7 @@ struct btrfs_root { #define BTRFS_MOUNT_NOBARRIER (1 << 2) #define BTRFS_MOUNT_SSD (1 << 3) #define BTRFS_MOUNT_DEGRADED (1 << 4) +#define BTRFS_MOUNT_COMPRESS (1 << 5) #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) @@ -825,6 +862,7 @@ struct btrfs_root { #define BTRFS_INODE_NODATASUM (1 << 0) #define BTRFS_INODE_NODATACOW (1 << 1) #define BTRFS_INODE_READONLY (1 << 2) +#define BTRFS_INODE_NOCOMPRESS (1 << 3) #define btrfs_clear_flag(inode, flag) (BTRFS_I(inode)->flags &= \ ~BTRFS_INODE_##flag) #define btrfs_set_flag(inode, flag) (BTRFS_I(inode)->flags |= \ @@ -1424,14 +1462,6 @@ static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize) return offsetof(struct btrfs_file_extent_item, disk_bytenr) + datasize; } -static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb, - struct btrfs_item *e) -{ - unsigned long offset; - offset = offsetof(struct btrfs_file_extent_item, disk_bytenr); - return btrfs_item_size(eb, e) - offset; -} - BTRFS_SETGET_FUNCS(file_extent_disk_bytenr, struct btrfs_file_extent_item, disk_bytenr, 64); BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item, @@ -1442,6 +1472,36 @@ BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item, offset, 64); BTRFS_SETGET_FUNCS(file_extent_num_bytes, struct btrfs_file_extent_item, num_bytes, 64); +BTRFS_SETGET_FUNCS(file_extent_ram_bytes, struct btrfs_file_extent_item, + ram_bytes, 64); +BTRFS_SETGET_FUNCS(file_extent_compression, struct btrfs_file_extent_item, + compression, 8); +BTRFS_SETGET_FUNCS(file_extent_encryption, struct btrfs_file_extent_item, + encryption, 8); +BTRFS_SETGET_FUNCS(file_extent_other_encoding, struct btrfs_file_extent_item, + other_encoding, 16); + +/* this returns the number of file bytes represented by the inline item. + * If an item is compressed, this is the uncompressed size + */ +static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb, + struct btrfs_file_extent_item *e) +{ + return btrfs_file_extent_ram_bytes(eb, e); +} + +/* + * this returns the number of bytes used by the item on disk, minus the + * size of any extent headers. If a file is compressed on disk, this is + * the compressed size + */ +static inline u32 btrfs_file_extent_inline_item_len(struct extent_buffer *eb, + struct btrfs_item *e) +{ + unsigned long offset; + offset = offsetof(struct btrfs_file_extent_item, disk_bytenr); + return btrfs_item_size(eb, e) - offset; +} static inline struct btrfs_root *btrfs_sb(struct super_block *sb) { @@ -1745,10 +1805,11 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct bio *bio); int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - u64 objectid, u64 pos, u64 disk_offset, - u64 disk_num_bytes, - u64 num_bytes, u64 offset); + struct btrfs_root *root, + u64 objectid, u64 pos, + u64 disk_offset, u64 disk_num_bytes, + u64 num_bytes, u64 offset, u64 ram_bytes, + u8 compression, u8 encryption, u16 other_encoding); int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 objectid, @@ -1758,6 +1819,8 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, struct btrfs_ordered_sum *sums); int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio); +int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode, + u64 start, unsigned long len); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, @@ -1799,7 +1862,7 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, int namelen); int btrfs_merge_bio_hook(struct page *page, unsigned long offset, - size_t size, struct bio *bio); + size_t size, struct bio *bio, unsigned long bio_flags); unsigned long btrfs_force_ra(struct address_space *mapping, struct file_ra_state *ra, struct file *file, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 0be044bb619..dc95f636a11 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -83,6 +83,7 @@ struct async_submit_bio { extent_submit_bio_hook_t *submit_bio_hook; int rw; int mirror_num; + unsigned long bio_flags; struct btrfs_work work; }; @@ -115,6 +116,7 @@ struct extent_map *btree_get_extent(struct inode *inode, struct page *page, } em->start = 0; em->len = (u64)-1; + em->block_len = (u64)-1; em->block_start = 0; em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; @@ -469,12 +471,13 @@ static void run_one_async_submit(struct btrfs_work *work) wake_up(&fs_info->async_submit_wait); async->submit_bio_hook(async->inode, async->rw, async->bio, - async->mirror_num); + async->mirror_num, async->bio_flags); kfree(async); } int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, int rw, struct bio *bio, int mirror_num, + unsigned long bio_flags, extent_submit_bio_hook_t *submit_bio_hook) { struct async_submit_bio *async; @@ -491,6 +494,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, async->submit_bio_hook = submit_bio_hook; async->work.func = run_one_async_submit; async->work.flags = 0; + async->bio_flags = bio_flags; while(atomic_read(&fs_info->async_submit_draining) && atomic_read(&fs_info->nr_async_submits)) { @@ -530,7 +534,7 @@ static int btree_csum_one_bio(struct bio *bio) } static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, - int mirror_num) + int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret; @@ -556,17 +560,17 @@ static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, } static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, - int mirror_num) + int mirror_num, unsigned long bio_flags) { /* * kthread helpers are used to submit writes so that checksumming * can happen in parallel across all CPUs */ if (!(rw & (1 << BIO_RW))) { - return __btree_submit_bio_hook(inode, rw, bio, mirror_num); + return __btree_submit_bio_hook(inode, rw, bio, mirror_num, 0); } return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, - inode, rw, bio, mirror_num, + inode, rw, bio, mirror_num, 0, __btree_submit_bio_hook); } @@ -1407,6 +1411,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->btree_inode = new_inode(sb); fs_info->btree_inode->i_ino = 1; fs_info->btree_inode->i_nlink = 1; + fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); INIT_LIST_HEAD(&fs_info->ordered_extents); @@ -1508,6 +1513,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, */ btrfs_init_workers(&fs_info->workers, "worker", fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->submit_workers, "submit", min_t(u64, fs_devices->num_devices, fs_info->thread_pool_size)); @@ -1559,6 +1565,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, } fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); + fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, + 4 * 1024 * 1024 / PAGE_CACHE_SIZE); nodesize = btrfs_super_nodesize(disk_super); leafsize = btrfs_super_leafsize(disk_super); diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index f84f5058dbb..4eb1f1408d2 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -71,6 +71,7 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, int metadata); int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, int rw, struct bio *bio, int mirror_num, + unsigned long bio_flags, extent_submit_bio_hook_t *submit_bio_hook); int btrfs_congested_async(struct btrfs_fs_info *info, int iodone); unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 280ac1aa9b6..bbf04e80a1a 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3278,6 +3278,7 @@ static int noinline relocate_data_extent(struct inode *reloc_inode, em->start = extent_key->objectid - offset; em->len = extent_key->offset; + em->block_len = extent_key->offset; em->block_start = extent_key->objectid; em->bdev = root->fs_info->fs_devices->latest_bdev; set_bit(EXTENT_FLAG_PINNED, &em->flags); @@ -3314,10 +3315,14 @@ struct btrfs_ref_path { }; struct disk_extent { + u64 ram_bytes; u64 disk_bytenr; u64 disk_num_bytes; u64 offset; u64 num_bytes; + u8 compression; + u8 encryption; + u16 other_encoding; }; static int is_cowonly_root(u64 root_objectid) @@ -3631,6 +3636,11 @@ static int noinline get_new_locations(struct inode *reloc_inode, btrfs_file_extent_disk_num_bytes(leaf, fi); exts[nr].offset = btrfs_file_extent_offset(leaf, fi); exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi); + exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi); + exts[nr].compression = btrfs_file_extent_compression(leaf, fi); + exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi); + exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf, + fi); WARN_ON(exts[nr].offset > 0); WARN_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes); @@ -3846,6 +3856,8 @@ next: new_extents[0].disk_bytenr); btrfs_set_file_extent_disk_num_bytes(leaf, fi, new_extents[0].disk_num_bytes); + btrfs_set_file_extent_ram_bytes(leaf, fi, + new_extents[0].ram_bytes); ext_offset += new_extents[0].offset; btrfs_set_file_extent_offset(leaf, fi, ext_offset); btrfs_mark_buffer_dirty(leaf); @@ -3911,6 +3923,16 @@ next: new_extents[i].disk_bytenr); btrfs_set_file_extent_disk_num_bytes(leaf, fi, new_extents[i].disk_num_bytes); + btrfs_set_file_extent_ram_bytes(leaf, fi, + new_extents[i].ram_bytes); + + btrfs_set_file_extent_compression(leaf, fi, + new_extents[i].compression); + btrfs_set_file_extent_encryption(leaf, fi, + new_extents[i].encryption); + btrfs_set_file_extent_other_encoding(leaf, fi, + new_extents[i].other_encoding); + btrfs_set_file_extent_num_bytes(leaf, fi, extent_len); ext_offset += new_extents[i].offset; @@ -4169,6 +4191,8 @@ static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans, ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes; btrfs_set_file_extent_generation(leaf, fi, trans->transid); + btrfs_set_file_extent_ram_bytes(leaf, fi, + new_extent->ram_bytes); btrfs_set_file_extent_disk_bytenr(leaf, fi, new_extent->disk_bytenr); btrfs_set_file_extent_disk_num_bytes(leaf, fi, @@ -4847,7 +4871,8 @@ static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info, BUG_ON(err); err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0, - group->key.offset, 0); + group->key.offset, 0, group->key.offset, + 0, 0, 0); BUG_ON(err); inode = btrfs_iget_locked(root->fs_info->sb, objectid, root); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 563b2d12f4f..314041fdfa4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -30,6 +30,7 @@ static struct kmem_cache *extent_buffer_cache; static LIST_HEAD(buffers); static LIST_HEAD(states); +#define LEAK_DEBUG 1 #ifdef LEAK_DEBUG static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; #endif @@ -1067,8 +1068,8 @@ EXPORT_SYMBOL(find_first_extent_bit_state); * * 1 is returned if we find something, 0 if nothing was in the tree */ -static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree, - u64 *start, u64 *end, u64 max_bytes) +static noinline u64 find_delalloc_range(struct extent_io_tree *tree, + u64 *start, u64 *end, u64 max_bytes) { struct rb_node *node; struct extent_state *state; @@ -1077,11 +1078,11 @@ static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree, u64 total_bytes = 0; spin_lock_irq(&tree->lock); + /* * this search will find all the extents that end after * our range starts. */ -search_again: node = tree_search(tree, cur_start); if (!node) { if (!found) @@ -1100,40 +1101,6 @@ search_again: *end = state->end; goto out; } - if (!found && !(state->state & EXTENT_BOUNDARY)) { - struct extent_state *prev_state; - struct rb_node *prev_node = node; - while(1) { - prev_node = rb_prev(prev_node); - if (!prev_node) - break; - prev_state = rb_entry(prev_node, - struct extent_state, - rb_node); - if ((prev_state->end + 1 != state->start) || - !(prev_state->state & EXTENT_DELALLOC)) - break; - if ((cur_start - prev_state->start) * 2 > - max_bytes) - break; - state = prev_state; - node = prev_node; - } - } - if (state->state & EXTENT_LOCKED) { - DEFINE_WAIT(wait); - atomic_inc(&state->refs); - prepare_to_wait(&state->wq, &wait, - TASK_UNINTERRUPTIBLE); - spin_unlock_irq(&tree->lock); - schedule(); - spin_lock_irq(&tree->lock); - finish_wait(&state->wq, &wait); - free_extent_state(state); - goto search_again; - } - set_state_cb(tree, state, EXTENT_LOCKED); - state->state |= EXTENT_LOCKED; if (!found) *start = state->start; found++; @@ -1151,6 +1118,208 @@ out: return found; } +static noinline int __unlock_for_delalloc(struct inode *inode, + struct page *locked_page, + u64 start, u64 end) +{ + int ret; + struct page *pages[16]; + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + unsigned long nr_pages = end_index - index + 1; + int i; + + if (index == locked_page->index && end_index == index) + return 0; + + while(nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min(nr_pages, ARRAY_SIZE(pages)), pages); + for (i = 0; i < ret; i++) { + if (pages[i] != locked_page) + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + cond_resched(); + } + return 0; +} + +static noinline int lock_delalloc_pages(struct inode *inode, + struct page *locked_page, + u64 delalloc_start, + u64 delalloc_end) +{ + unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT; + unsigned long start_index = index; + unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT; + unsigned long pages_locked = 0; + struct page *pages[16]; + unsigned long nrpages; + int ret; + int i; + + /* the caller is responsible for locking the start index */ + if (index == locked_page->index && index == end_index) + return 0; + + /* skip the page at the start index */ + nrpages = end_index - index + 1; + while(nrpages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min(nrpages, ARRAY_SIZE(pages)), pages); + if (ret == 0) { + ret = -EAGAIN; + goto done; + } + /* now we have an array of pages, lock them all */ + for (i = 0; i < ret; i++) { + /* + * the caller is taking responsibility for + * locked_page + */ + if (pages[i] != locked_page) + lock_page(pages[i]); + page_cache_release(pages[i]); + } + pages_locked += ret; + nrpages -= ret; + index += ret; + cond_resched(); + } + ret = 0; +done: + if (ret && pages_locked) { + __unlock_for_delalloc(inode, locked_page, + delalloc_start, + ((u64)(start_index + pages_locked - 1)) << + PAGE_CACHE_SHIFT); + } + return ret; +} + +/* + * find a contiguous range of bytes in the file marked as delalloc, not + * more than 'max_bytes'. start and end are used to return the range, + * + * 1 is returned if we find something, 0 if nothing was in the tree + */ +static noinline u64 find_lock_delalloc_range(struct inode *inode, + struct extent_io_tree *tree, + struct page *locked_page, + u64 *start, u64 *end, + u64 max_bytes) +{ + u64 delalloc_start; + u64 delalloc_end; + u64 found; + int ret; + int loops = 0; + +again: + /* step one, find a bunch of delalloc bytes starting at start */ + delalloc_start = *start; + delalloc_end = 0; + found = find_delalloc_range(tree, &delalloc_start, &delalloc_end, + max_bytes); + if (!found) { + *start = delalloc_start; + *end = delalloc_end; + return found; + } + + /* + * make sure to limit the number of pages we try to lock down + * if we're looping. + */ + if (delalloc_end + 1 - delalloc_start > max_bytes && loops) { + delalloc_end = (delalloc_start + PAGE_CACHE_SIZE - 1) & + ~((u64)PAGE_CACHE_SIZE - 1); + } + /* step two, lock all the pages after the page that has start */ + ret = lock_delalloc_pages(inode, locked_page, + delalloc_start, delalloc_end); + if (ret == -EAGAIN) { + /* some of the pages are gone, lets avoid looping by + * shortening the size of the delalloc range we're searching + */ + if (!loops) { + unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1); + max_bytes = PAGE_CACHE_SIZE - offset; + loops = 1; + goto again; + } else { + found = 0; + goto out_failed; + } + } + BUG_ON(ret); + + /* step three, lock the state bits for the whole range */ + lock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS); + + /* then test to make sure it is all still delalloc */ + ret = test_range_bit(tree, delalloc_start, delalloc_end, + EXTENT_DELALLOC, 1); + if (!ret) { + unlock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS); + __unlock_for_delalloc(inode, locked_page, + delalloc_start, delalloc_end); + cond_resched(); + goto again; + } + *start = delalloc_start; + *end = delalloc_end; +out_failed: + return found; +} + +int extent_clear_unlock_delalloc(struct inode *inode, + struct extent_io_tree *tree, + u64 start, u64 end, struct page *locked_page, + int clear_dirty, int set_writeback, + int end_writeback) +{ + int ret; + struct page *pages[16]; + unsigned long index = start >> PAGE_CACHE_SHIFT; + unsigned long end_index = end >> PAGE_CACHE_SHIFT; + unsigned long nr_pages = end_index - index + 1; + int i; + int clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC; + + if (clear_dirty) + clear_bits |= EXTENT_DIRTY; + + clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS); + + while(nr_pages > 0) { + ret = find_get_pages_contig(inode->i_mapping, index, + min(nr_pages, ARRAY_SIZE(pages)), pages); + for (i = 0; i < ret; i++) { + if (pages[i] == locked_page) { + page_cache_release(pages[i]); + continue; + } + if (clear_dirty) + clear_page_dirty_for_io(pages[i]); + if (set_writeback) + set_page_writeback(pages[i]); + if (end_writeback) + end_page_writeback(pages[i]); + unlock_page(pages[i]); + page_cache_release(pages[i]); + } + nr_pages -= ret; + index += ret; + cond_resched(); + } + return 0; +} +EXPORT_SYMBOL(extent_clear_unlock_delalloc); + /* * count the number of bytes in the tree that have a given bit(s) * set. This can be fairly slow, except for EXTENT_DIRTY which is @@ -1631,38 +1800,26 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, return bio; } -static int submit_one_bio(int rw, struct bio *bio, int mirror_num) +static int submit_one_bio(int rw, struct bio *bio, int mirror_num, + unsigned long bio_flags) { int ret = 0; struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; struct page *page = bvec->bv_page; struct extent_io_tree *tree = bio->bi_private; - struct rb_node *node; - struct extent_state *state; u64 start; u64 end; start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; end = start + bvec->bv_len - 1; - spin_lock_irq(&tree->lock); - node = __etree_search(tree, start, NULL, NULL); - BUG_ON(!node); - state = rb_entry(node, struct extent_state, rb_node); - while(state->end < end) { - node = rb_next(node); - state = rb_entry(node, struct extent_state, rb_node); - } - BUG_ON(state->end != end); - spin_unlock_irq(&tree->lock); - bio->bi_private = NULL; bio_get(bio); if (tree->ops && tree->ops->submit_bio_hook) tree->ops->submit_bio_hook(page->mapping->host, rw, bio, - mirror_num); + mirror_num, bio_flags); else submit_bio(rw, bio); if (bio_flagged(bio, BIO_EOPNOTSUPP)) @@ -1678,39 +1835,56 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, struct bio **bio_ret, unsigned long max_pages, bio_end_io_t end_io_func, - int mirror_num) + int mirror_num, + unsigned long prev_bio_flags, + unsigned long bio_flags) { int ret = 0; struct bio *bio; int nr; + int contig = 0; + int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED; + int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED; + size_t page_size = min(size, PAGE_CACHE_SIZE); if (bio_ret && *bio_ret) { bio = *bio_ret; - if (bio->bi_sector + (bio->bi_size >> 9) != sector || + if (old_compressed) + contig = bio->bi_sector == sector; + else + contig = bio->bi_sector + (bio->bi_size >> 9) == + sector; + + if (prev_bio_flags != bio_flags || !contig || (tree->ops && tree->ops->merge_bio_hook && - tree->ops->merge_bio_hook(page, offset, size, bio)) || - bio_add_page(bio, page, size, offset) < size) { - ret = submit_one_bio(rw, bio, mirror_num); + tree->ops->merge_bio_hook(page, offset, page_size, bio, + bio_flags)) || + bio_add_page(bio, page, page_size, offset) < page_size) { + ret = submit_one_bio(rw, bio, mirror_num, + prev_bio_flags); bio = NULL; } else { return 0; } } - nr = bio_get_nr_vecs(bdev); + if (this_compressed) + nr = BIO_MAX_PAGES; + else + nr = bio_get_nr_vecs(bdev); + bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); if (!bio) { printk("failed to allocate bio nr %d\n", nr); } - - bio_add_page(bio, page, size, offset); + bio_add_page(bio, page, page_size, offset); bio->bi_end_io = end_io_func; bio->bi_private = tree; if (bio_ret) { *bio_ret = bio; } else { - ret = submit_one_bio(rw, bio, mirror_num); + ret = submit_one_bio(rw, bio, mirror_num, bio_flags); } return ret; @@ -1738,7 +1912,8 @@ void set_page_extent_head(struct page *page, unsigned long len) static int __extent_read_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, - struct bio **bio, int mirror_num) + struct bio **bio, int mirror_num, + unsigned long *bio_flags) { struct inode *inode = page->mapping->host; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -1756,13 +1931,27 @@ static int __extent_read_full_page(struct extent_io_tree *tree, int nr = 0; size_t page_offset = 0; size_t iosize; + size_t disk_io_size; size_t blocksize = inode->i_sb->s_blocksize; + unsigned long this_bio_flag = 0; set_page_extent_mapped(page); end = page_end; lock_extent(tree, start, end, GFP_NOFS); + if (page->index == last_byte >> PAGE_CACHE_SHIFT) { + char *userpage; + size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1); + + if (zero_offset) { + iosize = PAGE_CACHE_SIZE - zero_offset; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + zero_offset, 0, iosize); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + } + } while (cur <= end) { if (cur >= last_byte) { char *userpage; @@ -1793,10 +1982,19 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); } BUG_ON(end < cur); + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) + this_bio_flag = EXTENT_BIO_COMPRESSED; + iosize = min(extent_map_end(em) - cur, end - cur + 1); cur_end = min(extent_map_end(em) - 1, end); iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); - sector = (em->block_start + extent_offset) >> 9; + if (this_bio_flag & EXTENT_BIO_COMPRESSED) { + disk_io_size = em->block_len; + sector = em->block_start >> 9; + } else { + sector = (em->block_start + extent_offset) >> 9; + disk_io_size = iosize; + } bdev = em->bdev; block_start = em->block_start; free_extent_map(em); @@ -1845,10 +2043,13 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; pnr -= page->index; ret = submit_extent_page(READ, tree, page, - sector, iosize, page_offset, + sector, disk_io_size, page_offset, bdev, bio, pnr, - end_bio_extent_readpage, mirror_num); + end_bio_extent_readpage, mirror_num, + *bio_flags, + this_bio_flag); nr++; + *bio_flags = this_bio_flag; } if (ret) SetPageError(page); @@ -1867,11 +2068,13 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent) { struct bio *bio = NULL; + unsigned long bio_flags = 0; int ret; - ret = __extent_read_full_page(tree, page, get_extent, &bio, 0); + ret = __extent_read_full_page(tree, page, get_extent, &bio, 0, + &bio_flags); if (bio) - submit_one_bio(READ, bio, 0); + submit_one_bio(READ, bio, 0, bio_flags); return ret; } EXPORT_SYMBOL(extent_read_full_page); @@ -1909,6 +2112,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; u64 nr_delalloc; u64 delalloc_end; + int page_started; + int compressed; WARN_ON(!PageLocked(page)); pg_offset = i_size & (PAGE_CACHE_SIZE - 1); @@ -1934,27 +2139,33 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = start; delalloc_end = 0; + page_started = 0; while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start, + nr_delalloc = find_lock_delalloc_range(inode, tree, + page, + &delalloc_start, &delalloc_end, 128 * 1024 * 1024); if (nr_delalloc == 0) { delalloc_start = delalloc_end + 1; continue; } - tree->ops->fill_delalloc(inode, delalloc_start, - delalloc_end); - clear_extent_bit(tree, delalloc_start, - delalloc_end, - EXTENT_LOCKED | EXTENT_DELALLOC, - 1, 0, GFP_NOFS); + tree->ops->fill_delalloc(inode, page, delalloc_start, + delalloc_end, &page_started); delalloc_start = delalloc_end + 1; } + + /* did the fill delalloc function already unlock and start the IO? */ + if (page_started) { + return 0; + } + lock_extent(tree, start, page_end, GFP_NOFS); unlock_start = start; if (tree->ops && tree->ops->writepage_start_hook) { - ret = tree->ops->writepage_start_hook(page, start, page_end); + ret = tree->ops->writepage_start_hook(page, start, + page_end); if (ret == -EAGAIN) { unlock_extent(tree, start, page_end, GFP_NOFS); redirty_page_for_writepage(wbc, page); @@ -2006,10 +2217,15 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, sector = (em->block_start + extent_offset) >> 9; bdev = em->bdev; block_start = em->block_start; + compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags); free_extent_map(em); em = NULL; - if (block_start == EXTENT_MAP_HOLE || + /* + * compressed and inline extents are written through other + * paths in the FS + */ + if (compressed || block_start == EXTENT_MAP_HOLE || block_start == EXTENT_MAP_INLINE) { clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); @@ -2017,16 +2233,28 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_extent(tree, unlock_start, cur + iosize -1, GFP_NOFS); - if (tree->ops && tree->ops->writepage_end_io_hook) + /* + * end_io notification does not happen here for + * compressed extents + */ + if (!compressed && tree->ops && + tree->ops->writepage_end_io_hook) tree->ops->writepage_end_io_hook(page, cur, cur + iosize - 1, NULL, 1); - cur = cur + iosize; + else if (compressed) { + /* we don't want to end_page_writeback on + * a compressed extent. this happens + * elsewhere + */ + nr++; + } + + cur += iosize; pg_offset += iosize; unlock_start = cur; continue; } - /* leave this out until we have a page_mkwrite call */ if (0 && !test_range_bit(tree, cur, cur + iosize - 1, EXTENT_DIRTY, 0)) { @@ -2034,6 +2262,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, pg_offset += iosize; continue; } + clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); if (tree->ops && tree->ops->writepage_io_hook) { ret = tree->ops->writepage_io_hook(page, cur, @@ -2057,7 +2286,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, ret = submit_extent_page(WRITE, tree, page, sector, iosize, pg_offset, bdev, &epd->bio, max_nr, - end_bio_extent_writepage, 0); + end_bio_extent_writepage, + 0, 0, 0); if (ret) SetPageError(page); } @@ -2226,7 +2456,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, extent_write_cache_pages(tree, mapping, &wbc_writepages, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); + submit_one_bio(WRITE, epd.bio, 0, 0); } return ret; } @@ -2248,7 +2478,7 @@ int extent_writepages(struct extent_io_tree *tree, ret = extent_write_cache_pages(tree, mapping, wbc, __extent_writepage, &epd); if (epd.bio) { - submit_one_bio(WRITE, epd.bio, 0); + submit_one_bio(WRITE, epd.bio, 0, 0); } return ret; } @@ -2262,6 +2492,7 @@ int extent_readpages(struct extent_io_tree *tree, struct bio *bio = NULL; unsigned page_idx; struct pagevec pvec; + unsigned long bio_flags = 0; pagevec_init(&pvec, 0); for (page_idx = 0; page_idx < nr_pages; page_idx++) { @@ -2281,7 +2512,7 @@ int extent_readpages(struct extent_io_tree *tree, if (!pagevec_add(&pvec, page)) __pagevec_lru_add(&pvec); __extent_read_full_page(tree, page, get_extent, - &bio, 0); + &bio, 0, &bio_flags); } page_cache_release(page); } @@ -2289,7 +2520,7 @@ int extent_readpages(struct extent_io_tree *tree, __pagevec_lru_add(&pvec); BUG_ON(!list_empty(pages)); if (bio) - submit_one_bio(READ, bio, 0); + submit_one_bio(READ, bio, 0, bio_flags); return 0; } EXPORT_SYMBOL(extent_readpages); @@ -2414,7 +2645,8 @@ int extent_prepare_write(struct extent_io_tree *tree, ret = submit_extent_page(READ, tree, page, sector, iosize, page_offset, em->bdev, NULL, 1, - end_bio_extent_preparewrite, 0); + end_bio_extent_preparewrite, 0, + 0, 0); iocount++; block_start = block_start + iosize; } else { @@ -2495,7 +2727,9 @@ int try_release_extent_mapping(struct extent_map_tree *map, } if (!test_range_bit(tree, em->start, extent_map_end(em) - 1, - EXTENT_LOCKED, 0)) { + EXTENT_LOCKED | EXTENT_WRITEBACK | + EXTENT_ORDERED, + 0)) { remove_extent_mapping(map, em); /* once for the rb tree */ free_extent_map(em); @@ -2923,6 +3157,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, int inc_all_pages = 0; unsigned long num_pages; struct bio *bio = NULL; + unsigned long bio_flags = 0; if (eb->flags & EXTENT_UPTODATE) return 0; @@ -2973,7 +3208,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, ClearPageError(page); err = __extent_read_full_page(tree, page, get_extent, &bio, - mirror_num); + mirror_num, &bio_flags); if (err) { ret = err; printk("err %d from __extent_read_full_page\n", ret); @@ -2984,7 +3219,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, } if (bio) - submit_one_bio(READ, bio, mirror_num); + submit_one_bio(READ, bio, mirror_num, bio_flags); if (ret || !wait) { if (ret) diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index c9d1908a1ae..86f859b87a6 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -18,6 +18,9 @@ #define EXTENT_BOUNDARY (1 << 11) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) +/* flags for bio submission */ +#define EXTENT_BIO_COMPRESSED 1 + /* * page->private values. Every page that is controlled by the extent * map has page->private set to one. @@ -28,14 +31,17 @@ struct extent_state; typedef int (extent_submit_bio_hook_t)(struct inode *inode, int rw, - struct bio *bio, int mirror_num); + struct bio *bio, int mirror_num, + unsigned long bio_flags); struct extent_io_ops { - int (*fill_delalloc)(struct inode *inode, u64 start, u64 end); + int (*fill_delalloc)(struct inode *inode, struct page *locked_page, + u64 start, u64 end, int *page_started); int (*writepage_start_hook)(struct page *page, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); extent_submit_bio_hook_t *submit_bio_hook; int (*merge_bio_hook)(struct page *page, unsigned long offset, - size_t size, struct bio *bio); + size_t size, struct bio *bio, + unsigned long bio_flags); int (*readpage_io_hook)(struct page *page, u64 start, u64 end); int (*readpage_io_failed_hook)(struct bio *bio, struct page *page, u64 start, u64 end, @@ -245,4 +251,9 @@ void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km); int release_extent_buffer_tail_pages(struct extent_buffer *eb); int extent_range_uptodate(struct extent_io_tree *tree, u64 start, u64 end); +int extent_clear_unlock_delalloc(struct inode *inode, + struct extent_io_tree *tree, + u64 start, u64 end, struct page *locked_page, + int clear_dirty, int set_writeback, + int clear_writeback); #endif diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index 74b2a29880d..fd3ebfb8c3c 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -184,6 +184,13 @@ static int mergable_maps(struct extent_map *prev, struct extent_map *next) if (test_bit(EXTENT_FLAG_PINNED, &prev->flags)) return 0; + /* + * don't merge compressed extents, we need to know their + * actual size + */ + if (test_bit(EXTENT_FLAG_COMPRESSED, &prev->flags)) + return 0; + if (extent_map_end(prev) == next->start && prev->flags == next->flags && prev->bdev == next->bdev && @@ -239,6 +246,7 @@ int add_extent_mapping(struct extent_map_tree *tree, if (rb && mergable_maps(merge, em)) { em->start = merge->start; em->len += merge->len; + em->block_len += merge->block_len; em->block_start = merge->block_start; merge->in_tree = 0; rb_erase(&merge->rb_node, &tree->map); @@ -250,6 +258,7 @@ int add_extent_mapping(struct extent_map_tree *tree, merge = rb_entry(rb, struct extent_map, rb_node); if (rb && mergable_maps(em, merge)) { em->len += merge->len; + em->block_len += merge->len; rb_erase(&merge->rb_node, &tree->map); merge->in_tree = 0; free_extent_map(merge); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index 26ac6fe0b26..abbcbeb28c7 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -10,6 +10,7 @@ /* bits for the flags field */ #define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */ +#define EXTENT_FLAG_COMPRESSED 1 struct extent_map { struct rb_node rb_node; @@ -18,6 +19,7 @@ struct extent_map { u64 start; u64 len; u64 block_start; + u64 block_len; unsigned long flags; struct block_device *bdev; atomic_t refs; @@ -38,9 +40,9 @@ static inline u64 extent_map_end(struct extent_map *em) static inline u64 extent_map_block_end(struct extent_map *em) { - if (em->block_start + em->len < em->block_start) + if (em->block_start + em->block_len < em->block_start) return (u64)-1; - return em->block_start + em->len; + return em->block_start + em->block_len; } void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask); diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 6dbe88b9d7d..f4d3fa71bc4 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -31,7 +31,8 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, u64 disk_offset, u64 disk_num_bytes, - u64 num_bytes, u64 offset) + u64 num_bytes, u64 offset, u64 ram_bytes, + u8 compression, u8 encryption, u16 other_encoding) { int ret = 0; struct btrfs_file_extent_item *item; @@ -57,8 +58,13 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes); btrfs_set_file_extent_offset(leaf, item, offset); btrfs_set_file_extent_num_bytes(leaf, item, num_bytes); + btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes); btrfs_set_file_extent_generation(leaf, item, trans->transid); btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG); + btrfs_set_file_extent_compression(leaf, item, compression); + btrfs_set_file_extent_encryption(leaf, item, encryption); + btrfs_set_file_extent_other_encoding(leaf, item, other_encoding); + btrfs_mark_buffer_dirty(leaf); out: btrfs_free_path(path); @@ -213,6 +219,73 @@ found: return 0; } +int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode, + u64 start, unsigned long len) +{ + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; + struct btrfs_ordered_extent *ordered; + char *data; + struct page *page; + unsigned long total_bytes = 0; + unsigned long this_sum_bytes = 0; + + sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS); + if (!sums) + return -ENOMEM; + + sector_sum = sums->sums; + sums->file_offset = start; + sums->len = len; + INIT_LIST_HEAD(&sums->list); + ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset); + BUG_ON(!ordered); + + while(len > 0) { + if (start >= ordered->file_offset + ordered->len || + start < ordered->file_offset) { + sums->len = this_sum_bytes; + this_sum_bytes = 0; + btrfs_add_ordered_sum(inode, ordered, sums); + btrfs_put_ordered_extent(ordered); + + sums = kzalloc(btrfs_ordered_sum_size(root, len), + GFP_NOFS); + BUG_ON(!sums); + sector_sum = sums->sums; + sums->len = len; + sums->file_offset = start; + ordered = btrfs_lookup_ordered_extent(inode, + sums->file_offset); + BUG_ON(!ordered); + } + + page = find_get_page(inode->i_mapping, + start >> PAGE_CACHE_SHIFT); + + data = kmap_atomic(page, KM_USER0); + sector_sum->sum = ~(u32)0; + sector_sum->sum = btrfs_csum_data(root, data, sector_sum->sum, + PAGE_CACHE_SIZE); + kunmap_atomic(data, KM_USER0); + btrfs_csum_final(sector_sum->sum, + (char *)§or_sum->sum); + sector_sum->offset = page_offset(page); + page_cache_release(page); + + sector_sum++; + total_bytes += PAGE_CACHE_SIZE; + this_sum_bytes += PAGE_CACHE_SIZE; + start += PAGE_CACHE_SIZE; + + WARN_ON(len < PAGE_CACHE_SIZE); + len -= PAGE_CACHE_SIZE; + } + btrfs_add_ordered_sum(inode, ordered, sums); + btrfs_put_ordered_extent(ordered); + return 0; +} + int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 69abbe19add..0aa15436590 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -95,153 +95,6 @@ static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) } } -/* this does all the hard work for inserting an inline extent into - * the btree. Any existing inline extent is extended as required to make room, - * otherwise things are inserted as required into the btree - */ -static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct inode *inode, - u64 offset, size_t size, - struct page **pages, size_t page_offset, - int num_pages) -{ - struct btrfs_key key; - struct btrfs_path *path; - struct extent_buffer *leaf; - char *kaddr; - unsigned long ptr; - struct btrfs_file_extent_item *ei; - struct page *page; - u32 datasize; - int err = 0; - int ret; - int i; - ssize_t cur_size; - - path = btrfs_alloc_path(); - if (!path) - return -ENOMEM; - - btrfs_set_trans_block_group(trans, inode); - - key.objectid = inode->i_ino; - key.offset = offset; - btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); - - ret = btrfs_search_slot(trans, root, &key, path, 0, 1); - if (ret < 0) { - err = ret; - goto fail; - } - if (ret == 1) { - struct btrfs_key found_key; - - if (path->slots[0] == 0) - goto insert; - - path->slots[0]--; - leaf = path->nodes[0]; - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - - if (found_key.objectid != inode->i_ino) - goto insert; - - if (found_key.type != BTRFS_EXTENT_DATA_KEY) - goto insert; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - - if (btrfs_file_extent_type(leaf, ei) != - BTRFS_FILE_EXTENT_INLINE) { - goto insert; - } - btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); - ret = 0; - } - if (ret == 0) { - u32 found_size; - u64 found_end; - - leaf = path->nodes[0]; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - - if (btrfs_file_extent_type(leaf, ei) != - BTRFS_FILE_EXTENT_INLINE) { - err = ret; - btrfs_print_leaf(root, leaf); - printk("found wasn't inline offset %Lu inode %lu\n", - offset, inode->i_ino); - goto fail; - } - found_size = btrfs_file_extent_inline_len(leaf, - btrfs_item_nr(leaf, path->slots[0])); - found_end = key.offset + found_size; - - if (found_end < offset + size) { - btrfs_release_path(root, path); - ret = btrfs_search_slot(trans, root, &key, path, - offset + size - found_end, 1); - BUG_ON(ret != 0); - - ret = btrfs_extend_item(trans, root, path, - offset + size - found_end); - if (ret) { - err = ret; - goto fail; - } - leaf = path->nodes[0]; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - inode_add_bytes(inode, offset + size - found_end); - } - if (found_end < offset) { - ptr = btrfs_file_extent_inline_start(ei) + found_size; - memset_extent_buffer(leaf, 0, ptr, offset - found_end); - } - } else { -insert: - btrfs_release_path(root, path); - datasize = offset + size - key.offset; - inode_add_bytes(inode, datasize); - datasize = btrfs_file_extent_calc_inline_size(datasize); - ret = btrfs_insert_empty_item(trans, root, path, &key, - datasize); - if (ret) { - err = ret; - printk("got bad ret %d\n", ret); - goto fail; - } - leaf = path->nodes[0]; - ei = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(leaf, ei, trans->transid); - btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); - } - ptr = btrfs_file_extent_inline_start(ei) + offset - key.offset; - - cur_size = size; - i = 0; - while (size > 0) { - page = pages[i]; - kaddr = kmap_atomic(page, KM_USER0); - cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size); - write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size); - kunmap_atomic(kaddr, KM_USER0); - page_offset = 0; - ptr += cur_size; - size -= cur_size; - if (i >= num_pages) { - printk("i %d num_pages %d\n", i, num_pages); - } - i++; - } - btrfs_mark_buffer_dirty(leaf); -fail: - btrfs_free_path(path); - return err; -} - /* * after copy_from_user, pages need to be dirtied and we need to make * sure holes are created between the current EOF and the start of @@ -267,8 +120,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, u64 start_pos; u64 end_of_last_block; u64 end_pos = pos + write_bytes; - u64 inline_size; - int did_inline = 0; loff_t isize = i_size_read(inode); start_pos = pos & ~((u64)root->sectorsize - 1); @@ -314,7 +165,8 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, err = btrfs_insert_file_extent(trans, root, inode->i_ino, last_pos_in_file, - 0, 0, hole_size, 0); + 0, 0, hole_size, 0, + hole_size, 0, 0, 0); btrfs_drop_extent_cache(inode, last_pos_in_file, last_pos_in_file + hole_size - 1, 0); mutex_unlock(&BTRFS_I(inode)->extent_mutex); @@ -324,57 +176,19 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, goto failed; } - /* - * either allocate an extent for the new bytes or setup the key - * to show we are doing inline data in the extent + /* check for reserved extents on each page, we don't want + * to reset the delalloc bit on things that already have + * extents reserved. */ - inline_size = end_pos; - if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) || - inline_size > root->fs_info->max_inline || - (inline_size & (root->sectorsize -1)) == 0 || - inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) { - /* check for reserved extents on each page, we don't want - * to reset the delalloc bit on things that already have - * extents reserved. - */ - btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block); - for (i = 0; i < num_pages; i++) { - struct page *p = pages[i]; - SetPageUptodate(p); - ClearPageChecked(p); - set_page_dirty(p); - } - } else { - u64 aligned_end; - /* step one, delete the existing extents in this range */ - aligned_end = (pos + write_bytes + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); - mutex_lock(&BTRFS_I(inode)->extent_mutex); - err = btrfs_drop_extents(trans, root, inode, start_pos, - aligned_end, aligned_end, &hint_byte); - if (err) - goto failed; - if (isize > inline_size) - inline_size = min_t(u64, isize, aligned_end); - inline_size -= start_pos; - err = insert_inline_extent(trans, root, inode, start_pos, - inline_size, pages, 0, num_pages); - btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1, 0); - BUG_ON(err); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); - - /* - * an ugly way to do all the prop accounting around - * the page bits and mapping tags - */ - set_page_writeback(pages[0]); - end_page_writeback(pages[0]); - did_inline = 1; + btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block); + for (i = 0; i < num_pages; i++) { + struct page *p = pages[i]; + SetPageUptodate(p); + ClearPageChecked(p); + set_page_dirty(p); } if (end_pos > isize) { i_size_write(inode, end_pos); - if (did_inline) - BTRFS_I(inode)->disk_i_size = end_pos; btrfs_update_inode(trans, root, inode); } failed: @@ -399,6 +213,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, int ret; int testend = 1; unsigned long flags; + int compressed = 0; WARN_ON(end < start); if (end == (u64)-1) { @@ -434,6 +249,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, free_extent_map(em); continue; } + compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags); clear_bit(EXTENT_FLAG_PINNED, &em->flags); remove_extent_mapping(em_tree, em); @@ -442,6 +258,12 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, split->start = em->start; split->len = start - em->start; split->block_start = em->block_start; + + if (compressed) + split->block_len = em->block_len; + else + split->block_len = split->len; + split->bdev = em->bdev; split->flags = flags; ret = add_extent_mapping(em_tree, split); @@ -459,7 +281,13 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, split->bdev = em->bdev; split->flags = flags; - split->block_start = em->block_start + diff; + if (compressed) { + split->block_len = em->block_len; + split->block_start = em->block_start; + } else { + split->block_len = split->len; + split->block_start = em->block_start + diff; + } ret = add_extent_mapping(em_tree, split); BUG_ON(ret); @@ -533,7 +361,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) struct btrfs_item *item; item = btrfs_item_nr(leaf, slot); extent_end = found_key.offset + - btrfs_file_extent_inline_len(leaf, item); + btrfs_file_extent_inline_len(leaf, extent); extent_end = (extent_end + root->sectorsize - 1) & ~((u64)root->sectorsize -1 ); } @@ -573,6 +401,10 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, u64 extent_end = 0; u64 search_start = start; u64 leaf_start; + u64 ram_bytes = 0; + u8 compression = 0; + u8 encryption = 0; + u16 other_encoding = 0; u64 root_gen; u64 root_owner; struct extent_buffer *leaf; @@ -589,6 +421,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, int recow; int ret; + inline_limit = 0; btrfs_drop_extent_cache(inode, start, end - 1, 0); path = btrfs_alloc_path(); @@ -637,6 +470,12 @@ next_slot: extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); found_type = btrfs_file_extent_type(leaf, extent); + compression = btrfs_file_extent_compression(leaf, + extent); + encryption = btrfs_file_extent_encryption(leaf, + extent); + other_encoding = btrfs_file_extent_other_encoding(leaf, + extent); if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = btrfs_file_extent_disk_bytenr(leaf, @@ -646,13 +485,13 @@ next_slot: extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, extent); + ram_bytes = btrfs_file_extent_ram_bytes(leaf, + extent); found_extent = 1; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - struct btrfs_item *item; - item = btrfs_item_nr(leaf, slot); found_inline = 1; extent_end = key.offset + - btrfs_file_extent_inline_len(leaf, item); + btrfs_file_extent_inline_len(leaf, extent); } } else { extent_end = search_start; @@ -680,10 +519,9 @@ next_slot: search_start = (extent_end + mask) & ~mask; } else search_start = extent_end; - if (end <= extent_end && start >= key.offset && found_inline) { + + if (end <= extent_end && start >= key.offset && found_inline) *hint_byte = EXTENT_MAP_INLINE; - goto out; - } if (found_extent) { read_extent_buffer(leaf, &old, (unsigned long)extent, @@ -770,12 +608,27 @@ next_slot: write_extent_buffer(leaf, &old, (unsigned long)extent, sizeof(old)); + btrfs_set_file_extent_compression(leaf, extent, + compression); + btrfs_set_file_extent_encryption(leaf, extent, + encryption); + btrfs_set_file_extent_other_encoding(leaf, extent, + other_encoding); btrfs_set_file_extent_offset(leaf, extent, le64_to_cpu(old.offset) + end - key.offset); WARN_ON(le64_to_cpu(old.num_bytes) < (extent_end - end)); btrfs_set_file_extent_num_bytes(leaf, extent, extent_end - end); + + /* + * set the ram bytes to the size of the full extent + * before splitting. This is a worst case flag, + * but its the best we can do because we don't know + * how splitting affects compression + */ + btrfs_set_file_extent_ram_bytes(leaf, extent, + ram_bytes); btrfs_set_file_extent_type(leaf, extent, BTRFS_FILE_EXTENT_REG); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bf4bed6ca4d..9797592dc86 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -49,6 +49,7 @@ #include "compat.h" #include "tree-log.h" #include "ref-cache.h" +#include "compression.h" struct btrfs_iget_args { u64 ino; @@ -83,6 +84,7 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { }; static void btrfs_truncate(struct inode *inode); +static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end); /* * a very lame attempt at stopping writes when the FS is 85% full. There @@ -113,58 +115,375 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, return ret; } +/* + * this does all the hard work for inserting an inline extent into + * the btree. The caller should have done a btrfs_drop_extents so that + * no overlapping inline items exist in the btree + */ +static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode, + u64 start, size_t size, size_t compressed_size, + struct page **compressed_pages) +{ + struct btrfs_key key; + struct btrfs_path *path; + struct extent_buffer *leaf; + struct page *page = NULL; + char *kaddr; + unsigned long ptr; + struct btrfs_file_extent_item *ei; + int err = 0; + int ret; + size_t cur_size = size; + size_t datasize; + unsigned long offset; + int use_compress = 0; + + if (compressed_size && compressed_pages) { + use_compress = 1; + cur_size = compressed_size; + } + + path = btrfs_alloc_path(); if (!path) + return -ENOMEM; + + btrfs_set_trans_block_group(trans, inode); + + key.objectid = inode->i_ino; + key.offset = start; + btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); + inode_add_bytes(inode, size); + datasize = btrfs_file_extent_calc_inline_size(cur_size); + + inode_add_bytes(inode, size); + ret = btrfs_insert_empty_item(trans, root, path, &key, + datasize); + BUG_ON(ret); + if (ret) { + err = ret; + printk("got bad ret %d\n", ret); + goto fail; + } + leaf = path->nodes[0]; + ei = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, ei, trans->transid); + btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); + btrfs_set_file_extent_encryption(leaf, ei, 0); + btrfs_set_file_extent_other_encoding(leaf, ei, 0); + btrfs_set_file_extent_ram_bytes(leaf, ei, size); + ptr = btrfs_file_extent_inline_start(ei); + + if (use_compress) { + struct page *cpage; + int i = 0; + while(compressed_size > 0) { + cpage = compressed_pages[i]; + cur_size = min(compressed_size, + PAGE_CACHE_SIZE); + + kaddr = kmap(cpage); + write_extent_buffer(leaf, kaddr, ptr, cur_size); + kunmap(cpage); + + i++; + ptr += cur_size; + compressed_size -= cur_size; + } + btrfs_set_file_extent_compression(leaf, ei, + BTRFS_COMPRESS_ZLIB); + } else { + page = find_get_page(inode->i_mapping, + start >> PAGE_CACHE_SHIFT); + btrfs_set_file_extent_compression(leaf, ei, 0); + kaddr = kmap_atomic(page, KM_USER0); + offset = start & (PAGE_CACHE_SIZE - 1); + write_extent_buffer(leaf, kaddr + offset, ptr, size); + kunmap_atomic(kaddr, KM_USER0); + page_cache_release(page); + } + btrfs_mark_buffer_dirty(leaf); + btrfs_free_path(path); + + BTRFS_I(inode)->disk_i_size = inode->i_size; + btrfs_update_inode(trans, root, inode); + return 0; +fail: + btrfs_free_path(path); + return err; +} + + +/* + * conditionally insert an inline extent into the file. This + * does the checks required to make sure the data is small enough + * to fit as an inline extent. + */ +static int cow_file_range_inline(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode, u64 start, u64 end, + size_t compressed_size, + struct page **compressed_pages) +{ + u64 isize = i_size_read(inode); + u64 actual_end = min(end + 1, isize); + u64 inline_len = actual_end - start; + u64 aligned_end = (end + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); + u64 hint_byte; + u64 data_len = inline_len; + int ret; + + if (compressed_size) + data_len = compressed_size; + + if (start > 0 || + data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) || + (!compressed_size && + (actual_end & (root->sectorsize - 1)) == 0) || + end + 1 < isize || + data_len > root->fs_info->max_inline) { + return 1; + } + + mutex_lock(&BTRFS_I(inode)->extent_mutex); + ret = btrfs_drop_extents(trans, root, inode, start, + aligned_end, aligned_end, &hint_byte); + BUG_ON(ret); + + if (isize > actual_end) + inline_len = min_t(u64, isize, actual_end); + ret = insert_inline_extent(trans, root, inode, start, + inline_len, compressed_size, + compressed_pages); + BUG_ON(ret); + btrfs_drop_extent_cache(inode, start, aligned_end, 0); + mutex_unlock(&BTRFS_I(inode)->extent_mutex); + return 0; +} + /* * when extent_io.c finds a delayed allocation range in the file, * the call backs end up in this code. The basic idea is to * allocate extents on disk for the range, and create ordered data structs * in ram to track those extents. + * + * locked_page is the page that writepage had locked already. We use + * it to make sure we don't do extra locks or unlocks. + * + * *page_started is set to one if we unlock locked_page and do everything + * required to start IO on it. It may be clean and already done with + * IO when we return. */ -static int cow_file_range(struct inode *inode, u64 start, u64 end) +static int cow_file_range(struct inode *inode, struct page *locked_page, + u64 start, u64 end, int *page_started) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; u64 alloc_hint = 0; u64 num_bytes; + unsigned long ram_size; + u64 orig_start; + u64 disk_num_bytes; u64 cur_alloc_size; u64 blocksize = root->sectorsize; - u64 orig_num_bytes; + u64 actual_end; struct btrfs_key ins; struct extent_map *em; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; int ret = 0; + struct page **pages = NULL; + unsigned long nr_pages; + unsigned long nr_pages_ret = 0; + unsigned long total_compressed = 0; + unsigned long total_in = 0; + unsigned long max_compressed = 128 * 1024; + unsigned long max_uncompressed = 256 * 1024; + int i; + int will_compress; trans = btrfs_join_transaction(root, 1); BUG_ON(!trans); btrfs_set_trans_block_group(trans, inode); + orig_start = start; + + /* + * compression made this loop a bit ugly, but the basic idea is to + * compress some pages but keep the total size of the compressed + * extent relatively small. If compression is off, this goto target + * is never used. + */ +again: + will_compress = 0; + nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1; + nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE); + actual_end = min_t(u64, i_size_read(inode), end + 1); + total_compressed = actual_end - start; + + /* we want to make sure that amount of ram required to uncompress + * an extent is reasonable, so we limit the total size in ram + * of a compressed extent to 256k + */ + total_compressed = min(total_compressed, max_uncompressed); num_bytes = (end - start + blocksize) & ~(blocksize - 1); num_bytes = max(blocksize, num_bytes); - orig_num_bytes = num_bytes; + disk_num_bytes = num_bytes; + total_in = 0; + ret = 0; - if (alloc_hint == EXTENT_MAP_INLINE) - goto out; + /* we do compression for mount -o compress and when the + * inode has not been flagged as nocompress + */ + if (!btrfs_test_flag(inode, NOCOMPRESS) && + btrfs_test_opt(root, COMPRESS)) { + WARN_ON(pages); + pages = kmalloc(sizeof(struct page *) * nr_pages, GFP_NOFS); + + /* we want to make sure the amount of IO required to satisfy + * a random read is reasonably small, so we limit the size + * of a compressed extent to 128k + */ + ret = btrfs_zlib_compress_pages(inode->i_mapping, start, + total_compressed, pages, + nr_pages, &nr_pages_ret, + &total_in, + &total_compressed, + max_compressed); + + if (!ret) { + unsigned long offset = total_compressed & + (PAGE_CACHE_SIZE - 1); + struct page *page = pages[nr_pages_ret - 1]; + char *kaddr; + + /* zero the tail end of the last page, we might be + * sending it down to disk + */ + if (offset) { + kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr + offset, 0, + PAGE_CACHE_SIZE - offset); + kunmap_atomic(kaddr, KM_USER0); + } + will_compress = 1; + } + } + if (start == 0) { + /* lets try to make an inline extent */ + if (ret || total_in < (end - start + 1)) { + /* we didn't compress the entire range, try + * to make an uncompressed inline extent. This + * is almost sure to fail, but maybe inline sizes + * will get bigger later + */ + ret = cow_file_range_inline(trans, root, inode, + start, end, 0, NULL); + } else { + ret = cow_file_range_inline(trans, root, inode, + start, end, + total_compressed, pages); + } + if (ret == 0) { + extent_clear_unlock_delalloc(inode, + &BTRFS_I(inode)->io_tree, + start, end, NULL, + 1, 1, 1); + *page_started = 1; + ret = 0; + goto free_pages_out; + } + } + + if (will_compress) { + /* + * we aren't doing an inline extent round the compressed size + * up to a block size boundary so the allocator does sane + * things + */ + total_compressed = (total_compressed + blocksize - 1) & + ~(blocksize - 1); + + /* + * one last check to make sure the compression is really a + * win, compare the page count read with the blocks on disk + */ + total_in = (total_in + PAGE_CACHE_SIZE - 1) & + ~(PAGE_CACHE_SIZE - 1); + if (total_compressed >= total_in) { + will_compress = 0; + } else { + disk_num_bytes = total_compressed; + num_bytes = total_in; + } + } + if (!will_compress && pages) { + /* + * the compression code ran but failed to make things smaller, + * free any pages it allocated and our page pointer array + */ + for (i = 0; i < nr_pages_ret; i++) { + page_cache_release(pages[i]); + } + kfree(pages); + pages = NULL; + total_compressed = 0; + nr_pages_ret = 0; + + /* flag the file so we don't compress in the future */ + btrfs_set_flag(inode, NOCOMPRESS); + } + + BUG_ON(disk_num_bytes > + btrfs_super_total_bytes(&root->fs_info->super_copy)); - BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); mutex_lock(&BTRFS_I(inode)->extent_mutex); btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); mutex_unlock(&BTRFS_I(inode)->extent_mutex); - while(num_bytes > 0) { - cur_alloc_size = min(num_bytes, root->fs_info->max_extent); + while(disk_num_bytes > 0) { + unsigned long min_bytes; + + /* + * the max size of a compressed extent is pretty small, + * make the code a little less complex by forcing + * the allocator to find a whole compressed extent at once + */ + if (will_compress) + min_bytes = disk_num_bytes; + else + min_bytes = root->sectorsize; + + cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent); ret = btrfs_reserve_extent(trans, root, cur_alloc_size, - root->sectorsize, 0, alloc_hint, + min_bytes, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { WARN_ON(1); - goto out; + goto free_pages_out_fail; } em = alloc_extent_map(GFP_NOFS); em->start = start; - em->len = ins.offset; + + if (will_compress) { + ram_size = num_bytes; + em->len = num_bytes; + } else { + /* ramsize == disk size */ + ram_size = ins.offset; + em->len = ins.offset; + } + em->block_start = ins.objectid; + em->block_len = ins.offset; em->bdev = root->fs_info->fs_devices->latest_bdev; + mutex_lock(&BTRFS_I(inode)->extent_mutex); set_bit(EXTENT_FLAG_PINNED, &em->flags); + + if (will_compress) + set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); + while(1) { spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); @@ -174,26 +493,95 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end) break; } btrfs_drop_extent_cache(inode, start, - start + ins.offset - 1, 0); + start + ram_size - 1, 0); } mutex_unlock(&BTRFS_I(inode)->extent_mutex); cur_alloc_size = ins.offset; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, - ins.offset, 0); + ram_size, cur_alloc_size, 0, + will_compress); BUG_ON(ret); - if (num_bytes < cur_alloc_size) { - printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes, + + if (disk_num_bytes < cur_alloc_size) { + printk("num_bytes %Lu cur_alloc %Lu\n", disk_num_bytes, cur_alloc_size); break; } + + if (will_compress) { + /* + * we're doing compression, we and we need to + * submit the compressed extents down to the device. + * + * We lock down all the file pages, clearing their + * dirty bits and setting them writeback. Everyone + * that wants to modify the page will wait on the + * ordered extent above. + * + * The writeback bits on the file pages are + * cleared when the compressed pages are on disk + */ + btrfs_end_transaction(trans, root); + + if (start <= page_offset(locked_page) && + page_offset(locked_page) < start + ram_size) { + *page_started = 1; + } + + extent_clear_unlock_delalloc(inode, + &BTRFS_I(inode)->io_tree, + start, + start + ram_size - 1, + NULL, 1, 1, 0); + + ret = btrfs_submit_compressed_write(inode, start, + ram_size, ins.objectid, + cur_alloc_size, pages, + nr_pages_ret); + + BUG_ON(ret); + trans = btrfs_join_transaction(root, 1); + if (start + ram_size < end) { + start += ram_size; + alloc_hint = ins.objectid + ins.offset; + /* pages will be freed at end_bio time */ + pages = NULL; + goto again; + } else { + /* we've written everything, time to go */ + break; + } + } + /* we're not doing compressed IO, don't unlock the first + * page (which the caller expects to stay locked), don't + * clear any dirty bits and don't set any writeback bits + */ + extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, + start, start + ram_size - 1, + locked_page, 0, 0, 0); + disk_num_bytes -= cur_alloc_size; num_bytes -= cur_alloc_size; alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; } + + ret = 0; out: btrfs_end_transaction(trans, root); + return ret; + +free_pages_out_fail: + extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, + start, end, locked_page, 0, 0, 0); +free_pages_out: + for (i = 0; i < nr_pages_ret; i++) + page_cache_release(pages[i]); + if (pages) + kfree(pages); + + goto out; } /* @@ -203,7 +591,8 @@ out: * If no cow copies or snapshots exist, we write directly to the existing * blocks on disk */ -static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) +static int run_delalloc_nocow(struct inode *inode, struct page *locked_page, + u64 start, u64 end, int *page_started) { u64 extent_start; u64 extent_end; @@ -260,6 +649,11 @@ again: extent_end = extent_start + extent_num_bytes; err = 0; + if (btrfs_file_extent_compression(leaf, item) || + btrfs_file_extent_encryption(leaf,item) || + btrfs_file_extent_other_encoding(leaf, item)) + goto not_found; + if (loops && start != extent_start) goto not_found; @@ -284,7 +678,8 @@ again: bytenr += btrfs_file_extent_offset(leaf, item); extent_num_bytes = min(end + 1, extent_end) - start; ret = btrfs_add_ordered_extent(inode, start, bytenr, - extent_num_bytes, 1); + extent_num_bytes, + extent_num_bytes, 1, 0); if (ret) { err = ret; goto out; @@ -300,7 +695,8 @@ again: not_found: btrfs_end_transaction(trans, root); btrfs_free_path(path); - return cow_file_range(inode, start, end); + return cow_file_range(inode, locked_page, start, end, + page_started); } out: WARN_ON(err); @@ -312,16 +708,19 @@ out: /* * extent_io.c call back to do delayed allocation processing */ -static int run_delalloc_range(struct inode *inode, u64 start, u64 end) +static int run_delalloc_range(struct inode *inode, struct page *locked_page, + u64 start, u64 end, int *page_started) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret; if (btrfs_test_opt(root, NODATACOW) || btrfs_test_flag(inode, NODATACOW)) - ret = run_delalloc_nocow(inode, start, end); + ret = run_delalloc_nocow(inode, locked_page, start, end, + page_started); else - ret = cow_file_range(inode, start, end); + ret = cow_file_range(inode, locked_page, start, end, + page_started); return ret; } @@ -383,7 +782,8 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, * we don't create bios that span stripes or chunks */ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, - size_t size, struct bio *bio) + size_t size, struct bio *bio, + unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; struct btrfs_mapping_tree *map_tree; @@ -413,7 +813,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, * are inserted into the btree */ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, - int mirror_num) + int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; @@ -429,7 +829,7 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, * or reading the csums from the tree before a read */ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, - int mirror_num) + int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; @@ -444,11 +844,17 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, if (!(rw & (1 << BIO_RW))) { btrfs_lookup_bio_sums(root, inode, bio); + + if (bio_flags & EXTENT_BIO_COMPRESSED) { + return btrfs_submit_compressed_read(inode, bio, + mirror_num, bio_flags); + } + goto mapit; } return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, - __btrfs_submit_bio_hook); + bio_flags, __btrfs_submit_bio_hook); mapit: return btrfs_map_bio(root, rw, bio, mirror_num, 0); } @@ -539,7 +945,7 @@ out_page: * good idea. This causes problems because we want to make sure COW * properly happens and the data=ordered rules are followed. * - * In our case any range that doesn't have the EXTENT_ORDERED bit set + * In our case any range that doesn't have the ORDERED bit set * hasn't been properly setup for IO. We kick off an async process * to fix it up. The async helper will wait for ordered extents, set * the delalloc bit and make it safe to write the page. @@ -632,10 +1038,21 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) btrfs_set_file_extent_disk_bytenr(leaf, extent_item, ordered_extent->start); btrfs_set_file_extent_disk_num_bytes(leaf, extent_item, - ordered_extent->len); + ordered_extent->disk_len); btrfs_set_file_extent_offset(leaf, extent_item, 0); + + if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) + btrfs_set_file_extent_compression(leaf, extent_item, 1); + else + btrfs_set_file_extent_compression(leaf, extent_item, 0); + btrfs_set_file_extent_encryption(leaf, extent_item, 0); + btrfs_set_file_extent_other_encoding(leaf, extent_item, 0); + + /* ram bytes = extent_num_bytes for now */ btrfs_set_file_extent_num_bytes(leaf, extent_item, ordered_extent->len); + btrfs_set_file_extent_ram_bytes(leaf, extent_item, + ordered_extent->len); btrfs_mark_buffer_dirty(leaf); btrfs_drop_extent_cache(inode, ordered_extent->file_offset, @@ -644,7 +1061,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) mutex_unlock(&BTRFS_I(inode)->extent_mutex); ins.objectid = ordered_extent->start; - ins.offset = ordered_extent->len; + ins.offset = ordered_extent->disk_len; ins.type = BTRFS_EXTENT_ITEM_KEY; ret = btrfs_alloc_reserved_extent(trans, root, leaf->start, root->root_key.objectid, @@ -714,6 +1131,7 @@ int btrfs_io_failed_hook(struct bio *failed_bio, int ret; int rw; u64 logical; + unsigned long bio_flags = 0; ret = get_state_private(failure_tree, start, &private); if (ret) { @@ -738,6 +1156,8 @@ int btrfs_io_failed_hook(struct bio *failed_bio, } logical = start - em->start; logical = em->block_start + logical; + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) + bio_flags = EXTENT_BIO_COMPRESSED; failrec->logical = logical; free_extent_map(em); set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | @@ -781,7 +1201,8 @@ int btrfs_io_failed_hook(struct bio *failed_bio, rw = READ; BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio, - failrec->last_mirror); + failrec->last_mirror, + bio_flags); return 0; } @@ -1644,10 +2065,8 @@ search_again: item_end += btrfs_file_extent_num_bytes(leaf, fi); } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { - struct btrfs_item *item = btrfs_item_nr(leaf, - path->slots[0]); item_end += btrfs_file_extent_inline_len(leaf, - item); + fi); } item_end--; } @@ -1715,7 +2134,14 @@ search_again: root_owner = btrfs_header_owner(leaf); } } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { - if (!del_item) { + /* + * we can't truncate inline items that have had + * special encodings + */ + if (!del_item && + btrfs_file_extent_compression(leaf, fi) == 0 && + btrfs_file_extent_encryption(leaf, fi) == 0 && + btrfs_file_extent_other_encoding(leaf, fi) == 0) { u32 size = new_size - found_key.offset; if (root->ref_cows) { @@ -1926,7 +2352,8 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) err = btrfs_insert_file_extent(trans, root, inode->i_ino, hole_start, 0, 0, - hole_size, 0); + hole_size, 0, hole_size, + 0, 0, 0); btrfs_drop_extent_cache(inode, hole_start, (u64)-1, 0); btrfs_check_file(root, inode); @@ -2894,11 +3321,50 @@ static int merge_extent_mapping(struct extent_map_tree *em_tree, start_diff = map_start - em->start; em->start = map_start; em->len = map_len; - if (em->block_start < EXTENT_MAP_LAST_BYTE) + if (em->block_start < EXTENT_MAP_LAST_BYTE && + !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) { em->block_start += start_diff; + em->block_len -= start_diff; + } return add_extent_mapping(em_tree, em); } +static noinline int uncompress_inline(struct btrfs_path *path, + struct inode *inode, struct page *page, + size_t pg_offset, u64 extent_offset, + struct btrfs_file_extent_item *item) +{ + int ret; + struct extent_buffer *leaf = path->nodes[0]; + char *tmp; + size_t max_size; + unsigned long inline_size; + unsigned long ptr; + + WARN_ON(pg_offset != 0); + max_size = btrfs_file_extent_ram_bytes(leaf, item); + inline_size = btrfs_file_extent_inline_item_len(leaf, + btrfs_item_nr(leaf, path->slots[0])); + tmp = kmalloc(inline_size, GFP_NOFS); + ptr = btrfs_file_extent_inline_start(item); + + read_extent_buffer(leaf, tmp, ptr, inline_size); + + max_size = min(PAGE_CACHE_SIZE, max_size); + ret = btrfs_zlib_decompress(tmp, page, extent_offset, + inline_size, max_size); + if (ret) { + char *kaddr = kmap_atomic(page, KM_USER0); + unsigned long copy_size = min_t(u64, + PAGE_CACHE_SIZE - pg_offset, + max_size - extent_offset); + memset(kaddr + pg_offset, 0, copy_size); + kunmap_atomic(kaddr, KM_USER0); + } + kfree(tmp); + return 0; +} + /* * a bit scary, this does extent mapping from logical file offset to the disk. * the ugly parts come from merging extents from the disk with the @@ -2927,6 +3393,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; struct btrfs_trans_handle *trans = NULL; + int compressed; again: spin_lock(&em_tree->lock); @@ -2951,6 +3418,7 @@ again: em->bdev = root->fs_info->fs_devices->latest_bdev; em->start = EXTENT_MAP_HOLE; em->len = (u64)-1; + em->block_len = (u64)-1; if (!path) { path = btrfs_alloc_path(); @@ -2983,6 +3451,7 @@ again: found_type = btrfs_file_extent_type(leaf, item); extent_start = found_key.offset; + compressed = btrfs_file_extent_compression(leaf, item); if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = extent_start + btrfs_file_extent_num_bytes(leaf, item); @@ -3005,10 +3474,18 @@ again: em->block_start = EXTENT_MAP_HOLE; goto insert; } - bytenr += btrfs_file_extent_offset(leaf, item); - em->block_start = bytenr; em->start = extent_start; em->len = extent_end - extent_start; + if (compressed) { + set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); + em->block_start = bytenr; + em->block_len = btrfs_file_extent_disk_num_bytes(leaf, + item); + } else { + bytenr += btrfs_file_extent_offset(leaf, item); + em->block_start = bytenr; + em->block_len = em->len; + } goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { u64 page_start; @@ -3018,8 +3495,7 @@ again: size_t extent_offset; size_t copy_size; - size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, - path->slots[0])); + size = btrfs_file_extent_inline_len(leaf, item); extent_end = (extent_start + size + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); if (start < extent_start || start >= extent_end) { @@ -3035,9 +3511,10 @@ again: } em->block_start = EXTENT_MAP_INLINE; - if (!page) { + if (!page || create) { em->start = extent_start; - em->len = size; + em->len = (size + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); goto out; } @@ -3048,11 +3525,22 @@ again: em->start = extent_start + extent_offset; em->len = (copy_size + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); - map = kmap(page); + if (compressed) + set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); ptr = btrfs_file_extent_inline_start(item) + extent_offset; if (create == 0 && !PageUptodate(page)) { - read_extent_buffer(leaf, map + pg_offset, ptr, - copy_size); + if (btrfs_file_extent_compression(leaf, item) == + BTRFS_COMPRESS_ZLIB) { + ret = uncompress_inline(path, inode, page, + pg_offset, + extent_offset, item); + BUG_ON(ret); + } else { + map = kmap(page); + read_extent_buffer(leaf, map + pg_offset, ptr, + copy_size); + kunmap(page); + } flush_dcache_page(page); } else if (create && PageUptodate(page)) { if (!trans) { @@ -3063,11 +3551,12 @@ again: trans = btrfs_join_transaction(root, 1); goto again; } + map = kmap(page); write_extent_buffer(leaf, map + pg_offset, ptr, copy_size); + kunmap(page); btrfs_mark_buffer_dirty(leaf); } - kunmap(page); set_extent_uptodate(io_tree, em->start, extent_map_end(em) - 1, GFP_NOFS); goto insert; @@ -3779,6 +4268,11 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, btrfs_set_file_extent_generation(leaf, ei, trans->transid); btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE); + btrfs_set_file_extent_encryption(leaf, ei, 0); + btrfs_set_file_extent_compression(leaf, ei, 0); + btrfs_set_file_extent_other_encoding(leaf, ei, 0); + btrfs_set_file_extent_ram_bytes(leaf, ei, name_len); + ptr = btrfs_file_extent_inline_start(ei); write_extent_buffer(leaf, symname, ptr, name_len); btrfs_mark_buffer_dirty(leaf); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 2eb6caba57c..b5745bb96d4 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -165,7 +165,8 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, * inserted. */ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, int nocow) + u64 start, u64 len, u64 disk_len, int nocow, + int compressed) { struct btrfs_ordered_inode_tree *tree; struct rb_node *node; @@ -180,9 +181,12 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, entry->file_offset = file_offset; entry->start = start; entry->len = len; + entry->disk_len = disk_len; entry->inode = inode; if (nocow) set_bit(BTRFS_ORDERED_NOCOW, &entry->flags); + if (compressed) + set_bit(BTRFS_ORDERED_COMPRESSED, &entry->flags); /* one ref for the tree */ atomic_set(&entry->refs, 1); @@ -389,9 +393,10 @@ void btrfs_start_ordered_extent(struct inode *inode, * for pdflush to find them */ btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_NONE); - if (wait) + if (wait) { wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); + } } /* diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index f50f8870a14..1ef464145d2 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -66,6 +66,8 @@ struct btrfs_ordered_sum { #define BTRFS_ORDERED_NOCOW 2 /* set when we want to write in place */ +#define BTRFS_ORDERED_COMPRESSED 3 /* writing a compressed extent */ + struct btrfs_ordered_extent { /* logical offset in the file */ u64 file_offset; @@ -73,9 +75,12 @@ struct btrfs_ordered_extent { /* disk byte number */ u64 start; - /* length of the extent in bytes */ + /* ram length of the extent in bytes */ u64 len; + /* extent length on disk */ + u64 disk_len; + /* flags (described above) */ unsigned long flags; @@ -127,7 +132,8 @@ int btrfs_remove_ordered_extent(struct inode *inode, int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size); int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, int nocow); + u64 start, u64 len, u64 disk_len, int nocow, + int compressed); int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_extent *entry, struct btrfs_ordered_sum *sum); diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index bd9ab3e9a7f..64725c13aa1 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -115,15 +115,16 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) if (btrfs_file_extent_type(l, fi) == BTRFS_FILE_EXTENT_INLINE) { printk("\t\tinline extent data size %u\n", - btrfs_file_extent_inline_len(l, item)); + btrfs_file_extent_inline_len(l, fi)); break; } printk("\t\textent data disk bytenr %llu nr %llu\n", (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi), (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi)); - printk("\t\textent data offset %llu nr %llu\n", + printk("\t\textent data offset %llu nr %llu ram %llu\n", (unsigned long long)btrfs_file_extent_offset(l, fi), - (unsigned long long)btrfs_file_extent_num_bytes(l, fi)); + (unsigned long long)btrfs_file_extent_num_bytes(l, fi), + (unsigned long long)btrfs_file_extent_ram_bytes(l, fi)); break; case BTRFS_BLOCK_GROUP_ITEM_KEY: bi = btrfs_item_ptr(l, i, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 2e6039825b7..431fdf144b5 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -47,6 +47,7 @@ #include "volumes.h" #include "version.h" #include "export.h" +#include "compression.h" #define BTRFS_SUPER_MAGIC 0x9123683E @@ -69,7 +70,7 @@ static void btrfs_put_super (struct super_block * sb) enum { Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, - Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_err, + Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_compress, Opt_err, }; static match_table_t tokens = { @@ -83,6 +84,7 @@ static match_table_t tokens = { {Opt_max_inline, "max_inline=%s"}, {Opt_alloc_start, "alloc_start=%s"}, {Opt_thread_pool, "thread_pool=%d"}, + {Opt_compress, "compress"}, {Opt_ssd, "ssd"}, {Opt_noacl, "noacl"}, {Opt_err, NULL}, @@ -163,6 +165,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) btrfs_set_opt(info->mount_opt, NODATACOW); btrfs_set_opt(info->mount_opt, NODATASUM); break; + case Opt_compress: + printk(KERN_INFO "btrfs: use compression\n"); + btrfs_set_opt(info->mount_opt, COMPRESS); + break; case Opt_ssd: printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); btrfs_set_opt(info->mount_opt, SSD); @@ -622,6 +628,7 @@ static int __init init_btrfs_fs(void) err = btrfs_interface_init(); if (err) goto free_extent_map; + err = register_filesystem(&btrfs_fs_type); if (err) goto unregister_ioctl; @@ -651,6 +658,7 @@ static void __exit exit_btrfs_fs(void) unregister_filesystem(&btrfs_fs_type); btrfs_exit_sysfs(); btrfs_cleanup_fs_uuids(); + btrfs_zlib_exit(); } module_init(init_btrfs_fs) diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index cf618cc8b34..e6d579053a4 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -540,8 +540,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, if (found_type == BTRFS_FILE_EXTENT_REG) extent_end = start + btrfs_file_extent_num_bytes(eb, item); else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - size = btrfs_file_extent_inline_len(eb, - btrfs_item_nr(eb, slot)); + size = btrfs_file_extent_inline_len(eb, item); extent_end = (start + size + mask) & ~mask; } else { ret = 0; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 2eed7f91f51..7db4cfd03a9 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -1816,6 +1816,7 @@ again: em->start = key.offset; em->len = *num_bytes; em->block_start = 0; + em->block_len = em->len; if (type & BTRFS_BLOCK_GROUP_SYSTEM) { ret = btrfs_add_system_chunk(trans, chunk_root, &key, @@ -2323,6 +2324,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, em->start = logical; em->len = length; em->block_start = 0; + em->block_len = em->len; map->num_stripes = num_stripes; map->io_width = btrfs_chunk_io_width(leaf, chunk); diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c new file mode 100644 index 00000000000..e99309180a1 --- /dev/null +++ b/fs/btrfs/zlib.c @@ -0,0 +1,637 @@ +/* + * Copyright (C) 2008 Oracle. All rights reserved. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public + * License v2 as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public + * License along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA. + * + * Based on jffs2 zlib code: + * Copyright © 2001-2007 Red Hat, Inc. + * Created by David Woodhouse + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Plan: call deflate() with avail_in == *sourcelen, + avail_out = *dstlen - 12 and flush == Z_FINISH. + If it doesn't manage to finish, call it again with + avail_in == 0 and avail_out set to the remaining 12 + bytes for it to clean up. + Q: Is 12 bytes sufficient? +*/ +#define STREAM_END_SPACE 12 + +struct workspace { + z_stream inf_strm; + z_stream def_strm; + char *buf; + struct list_head list; +}; + +static LIST_HEAD(idle_workspace); +static DEFINE_SPINLOCK(workspace_lock); +static unsigned long num_workspace; +static atomic_t alloc_workspace = ATOMIC_INIT(0); +static DECLARE_WAIT_QUEUE_HEAD(workspace_wait); + +/* + * this finds an available zlib workspace or allocates a new one + * NULL or an ERR_PTR is returned if things go bad. + */ +static struct workspace *find_zlib_workspace(void) +{ + struct workspace *workspace; + int ret; + int cpus = num_online_cpus(); + +again: + spin_lock(&workspace_lock); + if (!list_empty(&idle_workspace)) { + workspace = list_entry(idle_workspace.next, struct workspace, + list); + list_del(&workspace->list); + num_workspace--; + spin_unlock(&workspace_lock); + return workspace; + + } + spin_unlock(&workspace_lock); + if (atomic_read(&alloc_workspace) > cpus) { + DEFINE_WAIT(wait); + prepare_to_wait(&workspace_wait, &wait, TASK_UNINTERRUPTIBLE); + if (atomic_read(&alloc_workspace) > cpus) + schedule(); + finish_wait(&workspace_wait, &wait); + goto again; + } + atomic_inc(&alloc_workspace); + workspace = kzalloc(sizeof(*workspace), GFP_NOFS); + if (!workspace) { + ret = -ENOMEM; + goto fail; + } + + workspace->def_strm.workspace = vmalloc(zlib_deflate_workspacesize()); + if (!workspace->def_strm.workspace) { + ret = -ENOMEM; + goto fail; + } + workspace->inf_strm.workspace = vmalloc(zlib_inflate_workspacesize()); + if (!workspace->inf_strm.workspace) { + ret = -ENOMEM; + goto fail_inflate; + } + workspace->buf = kmalloc(PAGE_CACHE_SIZE, GFP_NOFS); + if (!workspace->buf) { + ret = -ENOMEM; + goto fail_kmalloc; + } + return workspace; + +fail_kmalloc: + vfree(workspace->inf_strm.workspace); +fail_inflate: + vfree(workspace->def_strm.workspace); +fail: + kfree(workspace); + atomic_dec(&alloc_workspace); + wake_up(&workspace_wait); + return ERR_PTR(ret); +} + +/* + * put a workspace struct back on the list or free it if we have enough + * idle ones sitting around + */ +static int free_workspace(struct workspace *workspace) +{ + spin_lock(&workspace_lock); + if (num_workspace < num_online_cpus()) { + list_add_tail(&workspace->list, &idle_workspace); + num_workspace++; + spin_unlock(&workspace_lock); + if (waitqueue_active(&workspace_wait)) + wake_up(&workspace_wait); + return 0; + } + spin_unlock(&workspace_lock); + vfree(workspace->def_strm.workspace); + vfree(workspace->inf_strm.workspace); + kfree(workspace->buf); + kfree(workspace); + + atomic_dec(&alloc_workspace); + if (waitqueue_active(&workspace_wait)) + wake_up(&workspace_wait); + return 0; +} + +/* + * cleanup function for module exit + */ +static void free_workspaces(void) +{ + struct workspace *workspace; + while(!list_empty(&idle_workspace)) { + workspace = list_entry(idle_workspace.next, struct workspace, + list); + list_del(&workspace->list); + vfree(workspace->def_strm.workspace); + vfree(workspace->inf_strm.workspace); + kfree(workspace->buf); + kfree(workspace); + atomic_dec(&alloc_workspace); + } +} + +/* + * given an address space and start/len, compress the bytes. + * + * pages are allocated to hold the compressed result and stored + * in 'pages' + * + * out_pages is used to return the number of pages allocated. There + * may be pages allocated even if we return an error + * + * total_in is used to return the number of bytes actually read. It + * may be smaller then len if we had to exit early because we + * ran out of room in the pages array or because we cross the + * max_out threshold. + * + * total_out is used to return the total number of compressed bytes + * + * max_out tells us the max number of bytes that we're allowed to + * stuff into pages + */ +int btrfs_zlib_compress_pages(struct address_space *mapping, + u64 start, unsigned long len, + struct page **pages, + unsigned long nr_dest_pages, + unsigned long *out_pages, + unsigned long *total_in, + unsigned long *total_out, + unsigned long max_out) +{ + int ret; + struct workspace *workspace; + char *data_in; + char *cpage_out; + int nr_pages = 0; + struct page *in_page = NULL; + struct page *out_page = NULL; + int out_written = 0; + int in_read = 0; + unsigned long bytes_left; + + *out_pages = 0; + *total_out = 0; + *total_in = 0; + + workspace = find_zlib_workspace(); + if (!workspace) + return -1; + + if (Z_OK != zlib_deflateInit(&workspace->def_strm, 3)) { + printk(KERN_WARNING "deflateInit failed\n"); + ret = -1; + goto out; + } + + workspace->def_strm.total_in = 0; + workspace->def_strm.total_out = 0; + + in_page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT); + data_in = kmap(in_page); + + out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM); + cpage_out = kmap(out_page); + pages[0] = out_page; + nr_pages = 1; + + workspace->def_strm.next_in = data_in; + workspace->def_strm.next_out = cpage_out; + workspace->def_strm.avail_out = PAGE_CACHE_SIZE; + workspace->def_strm.avail_in = min(len, PAGE_CACHE_SIZE); + + out_written = 0; + in_read = 0; + + while (workspace->def_strm.total_in < len) { + ret = zlib_deflate(&workspace->def_strm, Z_SYNC_FLUSH); + if (ret != Z_OK) { + printk(KERN_DEBUG "btrfs deflate in loop returned %d\n", + ret); + zlib_deflateEnd(&workspace->def_strm); + ret = -1; + goto out; + } + + /* we're making it bigger, give up */ + if (workspace->def_strm.total_in > 8192 && + workspace->def_strm.total_in < + workspace->def_strm.total_out) { + ret = -1; + goto out; + } + /* we need another page for writing out. Test this + * before the total_in so we will pull in a new page for + * the stream end if required + */ + if (workspace->def_strm.avail_out == 0) { + kunmap(out_page); + if (nr_pages == nr_dest_pages) { + out_page = NULL; + ret = -1; + goto out; + } + out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM); + cpage_out = kmap(out_page); + pages[nr_pages] = out_page; + nr_pages++; + workspace->def_strm.avail_out = PAGE_CACHE_SIZE; + workspace->def_strm.next_out = cpage_out; + } + /* we're all done */ + if (workspace->def_strm.total_in >= len) + break; + + /* we've read in a full page, get a new one */ + if (workspace->def_strm.avail_in == 0) { + if (workspace->def_strm.total_out > max_out) + break; + + bytes_left = len - workspace->def_strm.total_in; + kunmap(in_page); + page_cache_release(in_page); + + start += PAGE_CACHE_SIZE; + in_page = find_get_page(mapping, + start >> PAGE_CACHE_SHIFT); + data_in = kmap(in_page); + workspace->def_strm.avail_in = min(bytes_left, + PAGE_CACHE_SIZE); + workspace->def_strm.next_in = data_in; + } + } + workspace->def_strm.avail_in = 0; + ret = zlib_deflate(&workspace->def_strm, Z_FINISH); + zlib_deflateEnd(&workspace->def_strm); + + if (ret != Z_STREAM_END) { + ret = -1; + goto out; + } + + if (workspace->def_strm.total_out >= workspace->def_strm.total_in) { + ret = -1; + goto out; + } + + ret = 0; + *total_out = workspace->def_strm.total_out; + *total_in = workspace->def_strm.total_in; +out: + *out_pages = nr_pages; + if (out_page) + kunmap(out_page); + + if (in_page) { + kunmap(in_page); + page_cache_release(in_page); + } + free_workspace(workspace); + return ret; +} + +/* + * pages_in is an array of pages with compressed data. + * + * disk_start is the starting logical offset of this array in the file + * + * bvec is a bio_vec of pages from the file that we want to decompress into + * + * vcnt is the count of pages in the biovec + * + * srclen is the number of bytes in pages_in + * + * The basic idea is that we have a bio that was created by readpages. + * The pages in the bio are for the uncompressed data, and they may not + * be contiguous. They all correspond to the range of bytes covered by + * the compressed extent. + */ +int btrfs_zlib_decompress_biovec(struct page **pages_in, + u64 disk_start, + struct bio_vec *bvec, + int vcnt, + size_t srclen) +{ + int ret = 0; + int wbits = MAX_WBITS; + struct workspace *workspace; + char *data_in; + size_t total_out = 0; + unsigned long page_bytes_left; + unsigned long page_in_index = 0; + unsigned long page_out_index = 0; + struct page *page_out; + unsigned long total_pages_in = (srclen + PAGE_CACHE_SIZE - 1) / + PAGE_CACHE_SIZE; + unsigned long buf_start; + unsigned long buf_offset; + unsigned long bytes; + unsigned long working_bytes; + unsigned long pg_offset; + unsigned long start_byte; + unsigned long current_buf_start; + char *kaddr; + + workspace = find_zlib_workspace(); + if (!workspace) + return -ENOMEM; + + data_in = kmap(pages_in[page_in_index]); + workspace->inf_strm.next_in = data_in; + workspace->inf_strm.avail_in = min(srclen, PAGE_CACHE_SIZE); + workspace->inf_strm.total_in = 0; + + workspace->inf_strm.total_out = 0; + workspace->inf_strm.next_out = workspace->buf; + workspace->inf_strm.avail_out = PAGE_CACHE_SIZE; + page_out = bvec[page_out_index].bv_page; + page_bytes_left = PAGE_CACHE_SIZE; + pg_offset = 0; + + /* If it's deflate, and it's got no preset dictionary, then + we can tell zlib to skip the adler32 check. */ + if (srclen > 2 && !(data_in[1] & PRESET_DICT) && + ((data_in[0] & 0x0f) == Z_DEFLATED) && + !(((data_in[0]<<8) + data_in[1]) % 31)) { + + wbits = -((data_in[0] >> 4) + 8); + workspace->inf_strm.next_in += 2; + workspace->inf_strm.avail_in -= 2; + } + + if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) { + printk(KERN_WARNING "inflateInit failed\n"); + ret = -1; + goto out; + } + while(workspace->inf_strm.total_in < srclen) { + ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH); + if (ret != Z_OK && ret != Z_STREAM_END) { + break; + } + + /* + * buf start is the byte offset we're of the start of + * our workspace buffer + */ + buf_start = total_out; + + /* total_out is the last byte of the workspace buffer */ + total_out = workspace->inf_strm.total_out; + + working_bytes = total_out - buf_start; + + /* + * start byte is the first byte of the page we're currently + * copying into relative to the start of the compressed data. + */ + start_byte = page_offset(page_out) - disk_start; + + if (working_bytes == 0) { + /* we didn't make progress in this inflate + * call, we're done + */ + if (ret != Z_STREAM_END) + ret = -1; + break; + } + + /* we haven't yet hit data corresponding to this page */ + if (total_out <= start_byte) { + goto next; + } + + /* + * the start of the data we care about is offset into + * the middle of our working buffer + */ + if (total_out > start_byte && buf_start < start_byte) { + buf_offset = start_byte - buf_start; + working_bytes -= buf_offset; + } else { + buf_offset = 0; + } + current_buf_start = buf_start; + + /* copy bytes from the working buffer into the pages */ + while(working_bytes > 0) { + bytes = min(PAGE_CACHE_SIZE - pg_offset, + PAGE_CACHE_SIZE - buf_offset); + bytes = min(bytes, working_bytes); + kaddr = kmap_atomic(page_out, KM_USER0); + memcpy(kaddr + pg_offset, workspace->buf + buf_offset, + bytes); + kunmap_atomic(kaddr, KM_USER0); + flush_dcache_page(page_out); + + pg_offset += bytes; + page_bytes_left -= bytes; + buf_offset += bytes; + working_bytes -= bytes; + current_buf_start += bytes; + + /* check if we need to pick another page */ + if (page_bytes_left == 0) { + page_out_index++; + if (page_out_index >= vcnt) { + ret = 0; + goto done; + } + page_out = bvec[page_out_index].bv_page; + pg_offset = 0; + page_bytes_left = PAGE_CACHE_SIZE; + start_byte = page_offset(page_out) - disk_start; + + /* + * make sure our new page is covered by this + * working buffer + */ + if (total_out <= start_byte) { + goto next; + } + + /* the next page in the biovec might not + * be adjacent to the last page, but it + * might still be found inside this working + * buffer. bump our offset pointer + */ + if (total_out > start_byte && + current_buf_start < start_byte) { + buf_offset = start_byte - buf_start; + working_bytes = total_out - start_byte; + current_buf_start = buf_start + + buf_offset; + } + } + } +next: + workspace->inf_strm.next_out = workspace->buf; + workspace->inf_strm.avail_out = PAGE_CACHE_SIZE; + + if (workspace->inf_strm.avail_in == 0) { + unsigned long tmp; + kunmap(pages_in[page_in_index]); + page_in_index++; + if (page_in_index >= total_pages_in) { + data_in = NULL; + break; + } + data_in = kmap(pages_in[page_in_index]); + workspace->inf_strm.next_in = data_in; + tmp = srclen - workspace->inf_strm.total_in; + workspace->inf_strm.avail_in = min(tmp, + PAGE_CACHE_SIZE); + } + } + if (ret != Z_STREAM_END) { + ret = -1; + } else { + ret = 0; + } +done: + zlib_inflateEnd(&workspace->inf_strm); + if (data_in) + kunmap(pages_in[page_in_index]); +out: + free_workspace(workspace); + return ret; +} + +/* + * a less complex decompression routine. Our compressed data fits in a + * single page, and we want to read a single page out of it. + * start_byte tells us the offset into the compressed data we're interested in + */ +int btrfs_zlib_decompress(unsigned char *data_in, + struct page *dest_page, + unsigned long start_byte, + size_t srclen, size_t destlen) +{ + int ret = 0; + int wbits = MAX_WBITS; + struct workspace *workspace; + unsigned long bytes_left = destlen; + unsigned long total_out = 0; + char *kaddr; + + if (destlen > PAGE_CACHE_SIZE) + return -ENOMEM; + + workspace = find_zlib_workspace(); + if (!workspace) + return -ENOMEM; + + workspace->inf_strm.next_in = data_in; + workspace->inf_strm.avail_in = srclen; + workspace->inf_strm.total_in = 0; + + workspace->inf_strm.next_out = workspace->buf; + workspace->inf_strm.avail_out = PAGE_CACHE_SIZE; + workspace->inf_strm.total_out = 0; + /* If it's deflate, and it's got no preset dictionary, then + we can tell zlib to skip the adler32 check. */ + if (srclen > 2 && !(data_in[1] & PRESET_DICT) && + ((data_in[0] & 0x0f) == Z_DEFLATED) && + !(((data_in[0]<<8) + data_in[1]) % 31)) { + + wbits = -((data_in[0] >> 4) + 8); + workspace->inf_strm.next_in += 2; + workspace->inf_strm.avail_in -= 2; + } + + if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) { + printk(KERN_WARNING "inflateInit failed\n"); + ret = -1; + goto out; + } + + while(bytes_left > 0) { + unsigned long buf_start; + unsigned long buf_offset; + unsigned long bytes; + unsigned long pg_offset = 0; + + ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH); + if (ret != Z_OK && ret != Z_STREAM_END) { + break; + } + + buf_start = total_out; + total_out = workspace->inf_strm.total_out; + + if (total_out == buf_start) { + ret = -1; + break; + } + + if (total_out <= start_byte) { + goto next; + } + + if (total_out > start_byte && buf_start < start_byte) { + buf_offset = start_byte - buf_start; + } else { + buf_offset = 0; + } + + bytes = min(PAGE_CACHE_SIZE - pg_offset, + PAGE_CACHE_SIZE - buf_offset); + bytes = min(bytes, bytes_left); + + kaddr = kmap_atomic(dest_page, KM_USER0); + memcpy(kaddr + pg_offset, workspace->buf + buf_offset, bytes); + kunmap_atomic(kaddr, KM_USER0); + + pg_offset += bytes; + bytes_left -= bytes; +next: + workspace->inf_strm.next_out = workspace->buf; + workspace->inf_strm.avail_out = PAGE_CACHE_SIZE; + } + if (ret != Z_STREAM_END && bytes_left != 0) { + ret = -1; + } else { + ret = 0; + } + zlib_inflateEnd(&workspace->inf_strm); +out: + free_workspace(workspace); + return ret; +} + +void btrfs_zlib_exit(void) +{ + free_workspaces(); +} -- cgit v1.2.3-70-g09d2 From cfbc246eaae2a1089911016094b74b3055e8a906 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 30 Oct 2008 13:22:14 -0400 Subject: Btrfs: walk compressed pages based on the nr_pages count instead of bytes The byte walk counting was awkward and error prone. This uses the number of pages sent the higher layer to build bios. Signed-off-by: Chris Mason --- fs/btrfs/compression.c | 7 +++++-- fs/btrfs/inode.c | 2 +- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index c5470367ca5..9adaa79adad 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -296,7 +296,7 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, /* create and submit bios for the compressed pages */ bytes_left = compressed_len; - while(bytes_left > 0) { + for (page_index = 0; page_index < cb->nr_pages; page_index++) { page = compressed_pages[page_index]; page->mapping = inode->i_mapping; if (bio->bi_size) @@ -324,7 +324,10 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, bio->bi_end_io = end_compressed_bio_write; bio_add_page(bio, page, PAGE_CACHE_SIZE, 0); } - page_index++; + if (bytes_left < PAGE_CACHE_SIZE) { + printk("bytes left %lu compress len %lu nr %lu\n", + bytes_left, cb->compressed_len, cb->nr_pages); + } bytes_left -= PAGE_CACHE_SIZE; first_byte += PAGE_CACHE_SIZE; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 9797592dc86..6739424c0fe 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -338,7 +338,7 @@ again: if (!btrfs_test_flag(inode, NOCOMPRESS) && btrfs_test_opt(root, COMPRESS)) { WARN_ON(pages); - pages = kmalloc(sizeof(struct page *) * nr_pages, GFP_NOFS); + pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS); /* we want to make sure the amount of IO required to satisfy * a random read is reasonably small, so we limit the size -- cgit v1.2.3-70-g09d2 From 19b9bdb054895ba07086f0264641c9f80e0eb2c4 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 30 Oct 2008 14:23:13 -0400 Subject: Btrfs: Fix logic to avoid reading checksums for -o nodatasum,compress When compression was on, we were improperly ignoring -o nodatasum. This reworks the logic a bit to properly honor all the flags. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6739424c0fe..dd9cd01042b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -833,28 +833,29 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, { struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; + int skip_sum; ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) { - goto mapit; - } + skip_sum = btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM); if (!(rw & (1 << BIO_RW))) { - btrfs_lookup_bio_sums(root, inode, bio); + if (!skip_sum) + btrfs_lookup_bio_sums(root, inode, bio); - if (bio_flags & EXTENT_BIO_COMPRESSED) { + if (bio_flags & EXTENT_BIO_COMPRESSED) return btrfs_submit_compressed_read(inode, bio, mirror_num, bio_flags); - } - goto mapit; - } - return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, + } else if (!skip_sum) { + /* we're doing a write, do the async checksumming */ + return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, bio_flags, __btrfs_submit_bio_hook); + } + mapit: return btrfs_map_bio(root, rw, bio, mirror_num, 0); } -- cgit v1.2.3-70-g09d2 From 9036c10208e1fc496cef7692ba66a78699b360dc Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:19:41 -0400 Subject: Btrfs: update hole handling v2 This patch splits the hole insertion code out of btrfs_setattr into btrfs_cont_expand and updates btrfs_get_extent to properly handle the case that file extent items are not continuous. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.h | 1 + fs/btrfs/extent_map.h | 1 + fs/btrfs/file.c | 41 ++--------- fs/btrfs/inode.c | 189 ++++++++++++++++++++++++++------------------------ 4 files changed, 108 insertions(+), 124 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index abb27332c91..ca5547af609 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1908,6 +1908,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans, int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode); int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode); void btrfs_orphan_cleanup(struct btrfs_root *root); +int btrfs_cont_expand(struct inode *inode, loff_t size); /* ioctl.c */ long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index abbcbeb28c7..e693e1b4ac4 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -11,6 +11,7 @@ /* bits for the flags field */ #define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */ #define EXTENT_FLAG_COMPRESSED 1 +#define EXTENT_FLAG_VACANCY 2 /* no file extent item found */ struct extent_map { struct rb_node rb_node; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 0aa15436590..b8a7637e14a 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -142,40 +142,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, } set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); - /* FIXME...EIEIO, ENOSPC and more */ - /* insert any holes we need to create */ - if (isize < start_pos) { - u64 last_pos_in_file; - u64 hole_size; - u64 mask = root->sectorsize - 1; - last_pos_in_file = (isize + mask) & ~mask; - hole_size = (start_pos - last_pos_in_file + mask) & ~mask; - if (hole_size > 0) { - btrfs_wait_ordered_range(inode, last_pos_in_file, - last_pos_in_file + hole_size); - mutex_lock(&BTRFS_I(inode)->extent_mutex); - err = btrfs_drop_extents(trans, root, inode, - last_pos_in_file, - last_pos_in_file + hole_size, - last_pos_in_file, - &hint_byte); - if (err) - goto failed; - - err = btrfs_insert_file_extent(trans, root, - inode->i_ino, - last_pos_in_file, - 0, 0, hole_size, 0, - hole_size, 0, 0, 0); - btrfs_drop_extent_cache(inode, last_pos_in_file, - last_pos_in_file + hole_size - 1, 0); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); - btrfs_check_file(root, inode); - } - if (err) - goto failed; - } - /* check for reserved extents on each page, we don't want * to reset the delalloc bit on things that already have * extents reserved. @@ -191,7 +157,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, i_size_write(inode, end_pos); btrfs_update_inode(trans, root, inode); } -failed: err = btrfs_end_transaction(trans, root); out_unlock: unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS); @@ -697,6 +662,12 @@ static int noinline prepare_pages(struct btrfs_root *root, struct file *file, start_pos = pos & ~((u64)root->sectorsize - 1); last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT; + if (start_pos > inode->i_size) { + err = btrfs_cont_expand(inode, start_pos); + if (err) + return err; + } + memset(pages, 0, num_pages * sizeof(struct page *)); again: for (i = 0; i < num_pages; i++) { diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index dd9cd01042b..8254d6fa691 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2296,81 +2296,91 @@ out: return ret; } -static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) +int btrfs_cont_expand(struct inode *inode, loff_t size) { - struct inode *inode = dentry->d_inode; + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + struct extent_map *em; + u64 mask = root->sectorsize - 1; + u64 hole_start = (inode->i_size + mask) & ~mask; + u64 block_end = (size + mask) & ~mask; + u64 last_byte; + u64 cur_offset; + u64 hole_size; int err; - err = inode_change_ok(inode, attr); + if (size <= hole_start) + return 0; + + err = btrfs_check_free_space(root, 1, 0); if (err) return err; - if (S_ISREG(inode->i_mode) && - attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { - struct btrfs_trans_handle *trans; - struct btrfs_root *root = BTRFS_I(inode)->root; - struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; + btrfs_truncate_page(inode->i_mapping, inode->i_size); - u64 mask = root->sectorsize - 1; - u64 hole_start = (inode->i_size + mask) & ~mask; - u64 block_end = (attr->ia_size + mask) & ~mask; - u64 hole_size; - u64 alloc_hint = 0; + while (1) { + struct btrfs_ordered_extent *ordered; + btrfs_wait_ordered_range(inode, hole_start, + block_end - hole_start); + lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + ordered = btrfs_lookup_ordered_extent(inode, hole_start); + if (!ordered) + break; + unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + btrfs_put_ordered_extent(ordered); + } - if (attr->ia_size <= hole_start) - goto out; + trans = btrfs_start_transaction(root, 1); + btrfs_set_trans_block_group(trans, inode); - err = btrfs_check_free_space(root, 1, 0); - if (err) - goto fail; + cur_offset = hole_start; + while (1) { + em = btrfs_get_extent(inode, NULL, 0, cur_offset, + block_end - cur_offset, 0); + BUG_ON(IS_ERR(em) || !em); + last_byte = min(extent_map_end(em), block_end); + last_byte = (last_byte + mask) & ~mask; + if (test_bit(EXTENT_FLAG_VACANCY, &em->flags)) { + hole_size = last_byte - cur_offset; + err = btrfs_insert_file_extent(trans, root, + inode->i_ino, cur_offset, 0, + 0, hole_size, 0, hole_size, + 0, 0, 0); + btrfs_drop_extent_cache(inode, hole_start, + last_byte - 1, 0); + } + free_extent_map(em); + cur_offset = last_byte; + if (err || cur_offset >= block_end) + break; + } - btrfs_truncate_page(inode->i_mapping, inode->i_size); + btrfs_end_transaction(trans, root); + unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + return err; +} - hole_size = block_end - hole_start; - while(1) { - struct btrfs_ordered_extent *ordered; - btrfs_wait_ordered_range(inode, hole_start, hole_size); - - lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); - ordered = btrfs_lookup_ordered_extent(inode, hole_start); - if (ordered) { - unlock_extent(io_tree, hole_start, - block_end - 1, GFP_NOFS); - btrfs_put_ordered_extent(ordered); - } else { - break; - } - } +static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) +{ + struct inode *inode = dentry->d_inode; + int err; - trans = btrfs_start_transaction(root, 1); - btrfs_set_trans_block_group(trans, inode); - mutex_lock(&BTRFS_I(inode)->extent_mutex); - err = btrfs_drop_extents(trans, root, inode, - hole_start, block_end, hole_start, - &alloc_hint); + err = inode_change_ok(inode, attr); + if (err) + return err; - if (alloc_hint != EXTENT_MAP_INLINE) { - err = btrfs_insert_file_extent(trans, root, - inode->i_ino, - hole_start, 0, 0, - hole_size, 0, hole_size, - 0, 0, 0); - btrfs_drop_extent_cache(inode, hole_start, - (u64)-1, 0); - btrfs_check_file(root, inode); - } - mutex_unlock(&BTRFS_I(inode)->extent_mutex); - btrfs_end_transaction(trans, root); - unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); + if (S_ISREG(inode->i_mode) && + attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { + err = btrfs_cont_expand(inode, attr->ia_size); if (err) return err; } -out: + err = inode_setattr(inode, attr); if (!err && ((attr->ia_valid & ATTR_MODE))) err = btrfs_acl_chmod(inode); -fail: return err; } @@ -3456,27 +3466,44 @@ again: if (found_type == BTRFS_FILE_EXTENT_REG) { extent_end = extent_start + btrfs_file_extent_num_bytes(leaf, item); - err = 0; - if (start < extent_start || start >= extent_end) { - em->start = start; - if (start < extent_start) { - if (start + len <= extent_start) - goto not_found; - em->len = extent_end - extent_start; - } else { - em->len = len; + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + size_t size; + size = btrfs_file_extent_inline_len(leaf, item); + extent_end = (extent_start + size + root->sectorsize - 1) & + ~((u64)root->sectorsize - 1); + } + + if (start >= extent_end) { + path->slots[0]++; + if (path->slots[0] >= btrfs_header_nritems(leaf)) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) { + err = ret; + goto out; } - goto not_found_em; + if (ret > 0) + goto not_found; + leaf = path->nodes[0]; } + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); + if (found_key.objectid != objectid || + found_key.type != BTRFS_EXTENT_DATA_KEY) + goto not_found; + if (start + len <= found_key.offset) + goto not_found; + em->start = start; + em->len = found_key.offset - start; + goto not_found_em; + } + + if (found_type == BTRFS_FILE_EXTENT_REG) { + em->start = extent_start; + em->len = extent_end - extent_start; bytenr = btrfs_file_extent_disk_bytenr(leaf, item); if (bytenr == 0) { - em->start = extent_start; - em->len = extent_end - extent_start; em->block_start = EXTENT_MAP_HOLE; goto insert; } - em->start = extent_start; - em->len = extent_end - extent_start; if (compressed) { set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); em->block_start = bytenr; @@ -3489,38 +3516,21 @@ again: } goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { - u64 page_start; unsigned long ptr; char *map; size_t size; size_t extent_offset; size_t copy_size; - size = btrfs_file_extent_inline_len(leaf, item); - extent_end = (extent_start + size + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); - if (start < extent_start || start >= extent_end) { - em->start = start; - if (start < extent_start) { - if (start + len <= extent_start) - goto not_found; - em->len = extent_end - extent_start; - } else { - em->len = len; - } - goto not_found_em; - } em->block_start = EXTENT_MAP_INLINE; - if (!page || create) { em->start = extent_start; - em->len = (size + root->sectorsize - 1) & - ~((u64)root->sectorsize - 1); + em->len = extent_end - extent_start; goto out; } - page_start = page_offset(page) + pg_offset; - extent_offset = page_start - extent_start; + size = btrfs_file_extent_inline_len(leaf, item); + extent_offset = page_offset(page) + pg_offset - extent_start; copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset, size - extent_offset); em->start = extent_start + extent_offset; @@ -3570,6 +3580,7 @@ not_found: em->len = len; not_found_em: em->block_start = EXTENT_MAP_HOLE; + set_bit(EXTENT_FLAG_VACANCY, &em->flags); insert: btrfs_release_path(root, path); if (em->start > start || extent_map_end(em) <= start) { -- cgit v1.2.3-70-g09d2 From 6643558db29006825dbb10012b3f8890aca4bcd5 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:19:50 -0400 Subject: Btrfs: Fix bookend extent race v2 When dropping middle part of an extent, btrfs_drop_extents truncates the extent at first, then inserts a bookend extent. Since truncation and insertion can't be done atomically, there is a small period that the bookend extent isn't in the tree. This causes problem for functions that search the tree for file extent item. The way to fix this is lock the range of the bookend extent before truncation. Signed-off-by: Yan Zheng --- fs/btrfs/extent-tree.c | 28 +++++++++++++--------------- fs/btrfs/extent_io.c | 6 +++++- fs/btrfs/file.c | 31 +++++++++++++++++++++++++------ fs/btrfs/inode.c | 10 ---------- 4 files changed, 43 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index fada9c22a02..535cee47fcf 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3379,11 +3379,13 @@ static int noinline relocate_data_extent(struct inode *reloc_inode, struct btrfs_root *root = BTRFS_I(reloc_inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree; struct extent_map *em; + u64 start = extent_key->objectid - offset; + u64 end = start + extent_key->offset - 1; em = alloc_extent_map(GFP_NOFS); BUG_ON(!em || IS_ERR(em)); - em->start = extent_key->objectid - offset; + em->start = start; em->len = extent_key->offset; em->block_len = extent_key->offset; em->block_start = extent_key->objectid; @@ -3391,7 +3393,7 @@ static int noinline relocate_data_extent(struct inode *reloc_inode, set_bit(EXTENT_FLAG_PINNED, &em->flags); /* setup extent map to cheat btrfs_readpage */ - mutex_lock(&BTRFS_I(reloc_inode)->extent_mutex); + lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS); while (1) { int ret; spin_lock(&em_tree->lock); @@ -3401,13 +3403,11 @@ static int noinline relocate_data_extent(struct inode *reloc_inode, free_extent_map(em); break; } - btrfs_drop_extent_cache(reloc_inode, em->start, - em->start + em->len - 1, 0); + btrfs_drop_extent_cache(reloc_inode, start, end, 0); } - mutex_unlock(&BTRFS_I(reloc_inode)->extent_mutex); + unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS); - return relocate_inode_pages(reloc_inode, extent_key->objectid - offset, - extent_key->offset); + return relocate_inode_pages(reloc_inode, start, extent_key->offset); } struct btrfs_ref_path { @@ -3831,7 +3831,6 @@ next: * the file extent item was modified by someone * before the extent got locked. */ - mutex_unlock(&BTRFS_I(inode)->extent_mutex); unlock_extent(&BTRFS_I(inode)->io_tree, lock_start, lock_end, GFP_NOFS); extent_locked = 0; @@ -3896,8 +3895,12 @@ next: lock_start = key.offset; lock_end = lock_start + num_bytes - 1; } else { - BUG_ON(lock_start != key.offset); - BUG_ON(lock_end - lock_start + 1 < num_bytes); + if (lock_start > key.offset || + lock_end + 1 < key.offset + num_bytes) { + unlock_extent(&BTRFS_I(inode)->io_tree, + lock_start, lock_end, GFP_NOFS); + extent_locked = 0; + } } if (!inode) { @@ -3951,7 +3954,6 @@ next: if (ordered) btrfs_put_ordered_extent(ordered); - mutex_lock(&BTRFS_I(inode)->extent_mutex); extent_locked = 1; continue; } @@ -4073,7 +4075,6 @@ next: } if (extent_locked) { - mutex_unlock(&BTRFS_I(inode)->extent_mutex); unlock_extent(&BTRFS_I(inode)->io_tree, lock_start, lock_end, GFP_NOFS); extent_locked = 0; @@ -4091,7 +4092,6 @@ out: if (inode) { mutex_unlock(&inode->i_mutex); if (extent_locked) { - mutex_unlock(&BTRFS_I(inode)->extent_mutex); unlock_extent(&BTRFS_I(inode)->io_tree, lock_start, lock_end, GFP_NOFS); } @@ -4180,10 +4180,8 @@ static int noinline invalidate_extent_cache(struct btrfs_root *root, lock_extent(&BTRFS_I(inode)->io_tree, key.offset, key.offset + num_bytes - 1, GFP_NOFS); - mutex_lock(&BTRFS_I(inode)->extent_mutex); btrfs_drop_extent_cache(inode, key.offset, key.offset + num_bytes - 1, 1); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); unlock_extent(&BTRFS_I(inode)->io_tree, key.offset, key.offset + num_bytes - 1, GFP_NOFS); cond_resched(); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 7503bd46819..65a0583027e 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -946,8 +946,12 @@ int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1, &failed_start, mask); - if (err == -EEXIST) + if (err == -EEXIST) { + if (failed_start > start) + clear_extent_bit(tree, start, failed_start - 1, + EXTENT_LOCKED, 1, 0, mask); return 0; + } return 1; } EXPORT_SYMBOL(try_lock_extent); diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index b8a7637e14a..1a0510ad030 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -364,6 +364,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, u64 start, u64 end, u64 inline_limit, u64 *hint_byte) { u64 extent_end = 0; + u64 locked_end = end; u64 search_start = start; u64 leaf_start; u64 ram_bytes = 0; @@ -479,12 +480,6 @@ next_slot: goto next_slot; } - if (found_inline) { - u64 mask = root->sectorsize - 1; - search_start = (extent_end + mask) & ~mask; - } else - search_start = extent_end; - if (end <= extent_end && start >= key.offset && found_inline) *hint_byte = EXTENT_MAP_INLINE; @@ -501,6 +496,26 @@ next_slot: if (found_inline && start <= key.offset) keep = 1; } + + if (bookend && found_extent && locked_end < extent_end) { + ret = try_lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, GFP_NOFS); + if (!ret) { + btrfs_release_path(root, path); + lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, GFP_NOFS); + locked_end = extent_end; + continue; + } + locked_end = extent_end; + } + + if (found_inline) { + u64 mask = root->sectorsize - 1; + search_start = (extent_end + mask) & ~mask; + } else + search_start = extent_end; + /* truncate existing extent */ if (start > key.offset) { u64 new_num; @@ -638,6 +653,10 @@ next_slot: } out: btrfs_free_path(path); + if (locked_end > end) { + unlock_extent(&BTRFS_I(inode)->io_tree, end, locked_end - 1, + GFP_NOFS); + } btrfs_check_file(root, inode); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8254d6fa691..e8511d14b11 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -246,7 +246,6 @@ static int cow_file_range_inline(struct btrfs_trans_handle *trans, return 1; } - mutex_lock(&BTRFS_I(inode)->extent_mutex); ret = btrfs_drop_extents(trans, root, inode, start, aligned_end, aligned_end, &hint_byte); BUG_ON(ret); @@ -258,7 +257,6 @@ static int cow_file_range_inline(struct btrfs_trans_handle *trans, compressed_pages); BUG_ON(ret); btrfs_drop_extent_cache(inode, start, aligned_end, 0); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); return 0; } @@ -437,9 +435,7 @@ again: BUG_ON(disk_num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); - mutex_lock(&BTRFS_I(inode)->extent_mutex); btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); while(disk_num_bytes > 0) { unsigned long min_bytes; @@ -477,8 +473,6 @@ again: em->block_start = ins.objectid; em->block_len = ins.offset; em->bdev = root->fs_info->fs_devices->latest_bdev; - - mutex_lock(&BTRFS_I(inode)->extent_mutex); set_bit(EXTENT_FLAG_PINNED, &em->flags); if (will_compress) @@ -495,7 +489,6 @@ again: btrfs_drop_extent_cache(inode, start, start + ram_size - 1, 0); } - mutex_unlock(&BTRFS_I(inode)->extent_mutex); cur_alloc_size = ins.offset; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, @@ -1016,8 +1009,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) INIT_LIST_HEAD(&list); - mutex_lock(&BTRFS_I(inode)->extent_mutex); - ret = btrfs_drop_extents(trans, root, inode, ordered_extent->file_offset, ordered_extent->file_offset + @@ -1059,7 +1050,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) btrfs_drop_extent_cache(inode, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, 0); - mutex_unlock(&BTRFS_I(inode)->extent_mutex); ins.objectid = ordered_extent->start; ins.offset = ordered_extent->disk_len; -- cgit v1.2.3-70-g09d2 From 80ff385665b7fca29fefe358a60ab0d09f9b8e87 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:20:02 -0400 Subject: Btrfs: update nodatacow code v2 This patch simplifies the nodatacow checker. If all references were created after the latest snapshot, then we can avoid COW safely. This patch also updates run_delalloc_nocow to do more fine-grained checking. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.h | 8 +- fs/btrfs/extent-tree.c | 131 ++++------------------------- fs/btrfs/inode.c | 213 ++++++++++++++++++++++++++++-------------------- fs/btrfs/ioctl.c | 1 + fs/btrfs/ordered-data.c | 9 +- fs/btrfs/ordered-data.h | 3 +- fs/btrfs/transaction.c | 2 + 7 files changed, 154 insertions(+), 213 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ca5547af609..8bf6a085a73 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -454,6 +454,7 @@ struct btrfs_root_item { __le64 bytenr; __le64 byte_limit; __le64 bytes_used; + __le64 last_snapshot; __le32 flags; __le32 refs; struct btrfs_disk_key drop_progress; @@ -1413,6 +1414,8 @@ BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32); BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32); BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, bytes_used, 64); BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64); +BTRFS_SETGET_STACK_FUNCS(root_last_snapshot, struct btrfs_root_item, + last_snapshot, 64); /* struct btrfs_super_block */ BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64); @@ -1564,9 +1567,8 @@ int btrfs_update_pinned_extents(struct btrfs_root *root, u64 bytenr, u64 num, int pin); int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *leaf); -int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_key *key, u64 bytenr); +int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 535cee47fcf..1eb69a91b72 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -848,9 +848,8 @@ out: return 0; } -static int get_reference_status(struct btrfs_root *root, u64 bytenr, - u64 parent_gen, u64 ref_objectid, - u64 *min_generation, u32 *ref_count) +int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, + struct btrfs_root *root, u64 bytenr) { struct btrfs_root *extent_root = root->fs_info->extent_root; struct btrfs_path *path; @@ -858,8 +857,8 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, struct btrfs_extent_ref *ref_item; struct btrfs_key key; struct btrfs_key found_key; - u64 root_objectid = root->root_key.objectid; - u64 ref_generation; + u64 ref_root; + u64 last_snapshot; u32 nritems; int ret; @@ -872,7 +871,9 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, if (ret < 0) goto out; BUG_ON(ret == 0); - if (ret < 0 || path->slots[0] == 0) + + ret = -ENOENT; + if (path->slots[0] == 0) goto out; path->slots[0]--; @@ -880,14 +881,10 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (found_key.objectid != bytenr || - found_key.type != BTRFS_EXTENT_ITEM_KEY) { - ret = 1; + found_key.type != BTRFS_EXTENT_ITEM_KEY) goto out; - } - - *ref_count = 0; - *min_generation = (u64)-1; + last_snapshot = btrfs_root_last_snapshot(&root->root_item); while (1) { leaf = path->nodes[0]; nritems = btrfs_header_nritems(leaf); @@ -910,114 +907,22 @@ static int get_reference_status(struct btrfs_root *root, u64 bytenr, ref_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); - ref_generation = btrfs_ref_generation(leaf, ref_item); - /* - * For (parent_gen > 0 && parent_gen > ref_generation): - * - * we reach here through the oldest root, therefore - * all other reference from same snapshot should have - * a larger generation. - */ - if ((root_objectid != btrfs_ref_root(leaf, ref_item)) || - (parent_gen > 0 && parent_gen > ref_generation) || - (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID && - ref_objectid != btrfs_ref_objectid(leaf, ref_item))) { - *ref_count = 2; - break; - } - - *ref_count = 1; - if (*min_generation > ref_generation) - *min_generation = ref_generation; - - path->slots[0]++; - } - ret = 0; -out: - btrfs_free_path(path); - return ret; -} - -int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_key *key, u64 bytenr) -{ - struct btrfs_root *old_root; - struct btrfs_path *path = NULL; - struct extent_buffer *eb; - struct btrfs_file_extent_item *item; - u64 ref_generation; - u64 min_generation; - u64 extent_start; - u32 ref_count; - int level; - int ret; - - BUG_ON(trans == NULL); - BUG_ON(key->type != BTRFS_EXTENT_DATA_KEY); - ret = get_reference_status(root, bytenr, 0, key->objectid, - &min_generation, &ref_count); - if (ret) - return ret; - - if (ref_count != 1) - return 1; - - old_root = root->dirty_root->root; - ref_generation = old_root->root_key.offset; - - /* all references are created in running transaction */ - if (min_generation > ref_generation) { - ret = 0; - goto out; - } - - path = btrfs_alloc_path(); - if (!path) { - ret = -ENOMEM; - goto out; - } - - path->skip_locking = 1; - /* if no item found, the extent is referenced by other snapshot */ - ret = btrfs_search_slot(NULL, old_root, key, path, 0, 0); - if (ret) - goto out; - - eb = path->nodes[0]; - item = btrfs_item_ptr(eb, path->slots[0], - struct btrfs_file_extent_item); - if (btrfs_file_extent_type(eb, item) != BTRFS_FILE_EXTENT_REG || - btrfs_file_extent_disk_bytenr(eb, item) != bytenr) { - ret = 1; - goto out; - } - - for (level = BTRFS_MAX_LEVEL - 1; level >= -1; level--) { - if (level >= 0) { - eb = path->nodes[level]; - if (!eb) - continue; - extent_start = eb->start; - } else - extent_start = bytenr; - - ret = get_reference_status(root, extent_start, ref_generation, - 0, &min_generation, &ref_count); - if (ret) + ref_root = btrfs_ref_root(leaf, ref_item); + if (ref_root != root->root_key.objectid && + ref_root != BTRFS_TREE_LOG_OBJECTID) { + ret = 1; goto out; - - if (ref_count != 1) { + } + if (btrfs_ref_generation(leaf, ref_item) <= last_snapshot) { ret = 1; goto out; } - if (level >= 0) - ref_generation = btrfs_header_generation(eb); + + path->slots[0]++; } ret = 0; out: - if (path) - btrfs_free_path(path); + btrfs_free_path(path); return ret; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e8511d14b11..3e6f0568fdb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -298,6 +298,7 @@ static int cow_file_range(struct inode *inode, struct page *locked_page, unsigned long max_compressed = 128 * 1024; unsigned long max_uncompressed = 256 * 1024; int i; + int ordered_type; int will_compress; trans = btrfs_join_transaction(root, 1); @@ -491,9 +492,10 @@ again: } cur_alloc_size = ins.offset; + ordered_type = will_compress ? BTRFS_ORDERED_COMPRESSED : 0; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, - ram_size, cur_alloc_size, 0, - will_compress); + ram_size, cur_alloc_size, + ordered_type); BUG_ON(ret); if (disk_num_bytes < cur_alloc_size) { @@ -587,115 +589,148 @@ free_pages_out: static int run_delalloc_nocow(struct inode *inode, struct page *locked_page, u64 start, u64 end, int *page_started) { - u64 extent_start; - u64 extent_end; - u64 bytenr; - u64 loops = 0; - u64 total_fs_bytes; struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_block_group_cache *block_group; struct btrfs_trans_handle *trans; struct extent_buffer *leaf; - int found_type; struct btrfs_path *path; - struct btrfs_file_extent_item *item; - int ret; - int err = 0; + struct btrfs_file_extent_item *fi; struct btrfs_key found_key; + u64 cow_start; + u64 cur_offset; + u64 extent_end; + u64 disk_bytenr; + u64 num_bytes; + int extent_type; + int ret; + int nocow; + int check_prev = 1; - total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); path = btrfs_alloc_path(); BUG_ON(!path); trans = btrfs_join_transaction(root, 1); BUG_ON(!trans); -again: - ret = btrfs_lookup_file_extent(NULL, root, path, - inode->i_ino, start, 0); - if (ret < 0) { - err = ret; - goto out; - } - - if (ret != 0) { - if (path->slots[0] == 0) - goto not_found; - path->slots[0]--; - } - - leaf = path->nodes[0]; - item = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - - /* are we inside the extent that was found? */ - btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - found_type = btrfs_key_type(&found_key); - if (found_key.objectid != inode->i_ino || - found_type != BTRFS_EXTENT_DATA_KEY) - goto not_found; - - found_type = btrfs_file_extent_type(leaf, item); - extent_start = found_key.offset; - if (found_type == BTRFS_FILE_EXTENT_REG) { - u64 extent_num_bytes; - - extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item); - extent_end = extent_start + extent_num_bytes; - err = 0; - if (btrfs_file_extent_compression(leaf, item) || - btrfs_file_extent_encryption(leaf,item) || - btrfs_file_extent_other_encoding(leaf, item)) - goto not_found; + cow_start = (u64)-1; + cur_offset = start; + while (1) { + ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, + cur_offset, 0); + BUG_ON(ret < 0); + if (ret > 0 && path->slots[0] > 0 && check_prev) { + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &found_key, + path->slots[0] - 1); + if (found_key.objectid == inode->i_ino && + found_key.type == BTRFS_EXTENT_DATA_KEY) + path->slots[0]--; + } + check_prev = 0; +next_slot: + leaf = path->nodes[0]; + if (path->slots[0] >= btrfs_header_nritems(leaf)) { + ret = btrfs_next_leaf(root, path); + if (ret < 0) + BUG_ON(1); + if (ret > 0) + break; + leaf = path->nodes[0]; + } - if (loops && start != extent_start) - goto not_found; + nocow = 0; + disk_bytenr = 0; + btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (start < extent_start || start >= extent_end) - goto not_found; + if (found_key.objectid > inode->i_ino || + found_key.type > BTRFS_EXTENT_DATA_KEY || + found_key.offset > end) + break; - bytenr = btrfs_file_extent_disk_bytenr(leaf, item); - if (bytenr == 0) - goto not_found; + if (found_key.offset > cur_offset) { + extent_end = found_key.offset; + goto out_check; + } - if (btrfs_cross_ref_exists(trans, root, &found_key, bytenr)) - goto not_found; - /* - * we may be called by the resizer, make sure we're inside - * the limits of the FS - */ - block_group = btrfs_lookup_block_group(root->fs_info, - bytenr); - if (!block_group || block_group->ro) - goto not_found; + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + extent_type = btrfs_file_extent_type(leaf, fi); - bytenr += btrfs_file_extent_offset(leaf, item); - extent_num_bytes = min(end + 1, extent_end) - start; - ret = btrfs_add_ordered_extent(inode, start, bytenr, - extent_num_bytes, - extent_num_bytes, 1, 0); - if (ret) { - err = ret; - goto out; + if (extent_type == BTRFS_FILE_EXTENT_REG) { + struct btrfs_block_group_cache *block_group; + disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); + extent_end = found_key.offset + + btrfs_file_extent_num_bytes(leaf, fi); + if (extent_end <= start) { + path->slots[0]++; + goto next_slot; + } + if (btrfs_file_extent_compression(leaf, fi) || + btrfs_file_extent_encryption(leaf, fi) || + btrfs_file_extent_other_encoding(leaf, fi)) + goto out_check; + if (disk_bytenr == 0) + goto out_check; + if (btrfs_cross_ref_exist(trans, root, disk_bytenr)) + goto out_check; + block_group = btrfs_lookup_block_group(root->fs_info, + disk_bytenr); + if (!block_group || block_group->ro) + goto out_check; + disk_bytenr += btrfs_file_extent_offset(leaf, fi); + nocow = 1; + } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { + extent_end = found_key.offset + + btrfs_file_extent_inline_len(leaf, fi); + extent_end = ALIGN(extent_end, root->sectorsize); + } else { + BUG_ON(1); + } +out_check: + if (extent_end <= start) { + path->slots[0]++; + goto next_slot; + } + if (!nocow) { + if (cow_start == (u64)-1) + cow_start = cur_offset; + cur_offset = extent_end; + if (cur_offset > end) + break; + path->slots[0]++; + goto next_slot; } btrfs_release_path(root, path); - start = extent_end; - if (start <= end) { - loops++; - goto again; + if (cow_start != (u64)-1) { + ret = cow_file_range(inode, locked_page, cow_start, + found_key.offset - 1, page_started); + BUG_ON(ret); + cow_start = (u64)-1; } - } else { -not_found: - btrfs_end_transaction(trans, root); - btrfs_free_path(path); - return cow_file_range(inode, locked_page, start, end, - page_started); + + disk_bytenr += cur_offset - found_key.offset; + num_bytes = min(end + 1, extent_end) - cur_offset; + + ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr, + num_bytes, num_bytes, + BTRFS_ORDERED_NOCOW); + cur_offset = extent_end; + if (cur_offset > end) + break; } -out: - WARN_ON(err); - btrfs_end_transaction(trans, root); + btrfs_release_path(root, path); + + if (cur_offset <= end && cow_start == (u64)-1) + cow_start = cur_offset; + if (cow_start != (u64)-1) { + ret = cow_file_range(inode, locked_page, cow_start, end, + page_started); + BUG_ON(ret); + } + + ret = btrfs_end_transaction(trans, root); + BUG_ON(ret); btrfs_free_path(path); - return err; + return 0; } /* diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index fd3c8b5676c..7f915d47839 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -112,6 +112,7 @@ static noinline int create_subvol(struct btrfs_root *root, btrfs_set_root_level(&root_item, 0); btrfs_set_root_refs(&root_item, 1); btrfs_set_root_used(&root_item, 0); + btrfs_set_root_last_snapshot(&root_item, 0); memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress)); root_item.drop_level = 0; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index b5745bb96d4..e7317c8fda2 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -165,8 +165,7 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, * inserted. */ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, u64 disk_len, int nocow, - int compressed) + u64 start, u64 len, u64 disk_len, int type) { struct btrfs_ordered_inode_tree *tree; struct rb_node *node; @@ -183,10 +182,8 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, entry->len = len; entry->disk_len = disk_len; entry->inode = inode; - if (nocow) - set_bit(BTRFS_ORDERED_NOCOW, &entry->flags); - if (compressed) - set_bit(BTRFS_ORDERED_COMPRESSED, &entry->flags); + if (type == BTRFS_ORDERED_NOCOW || type == BTRFS_ORDERED_COMPRESSED) + set_bit(type, &entry->flags); /* one ref for the tree */ atomic_set(&entry->refs, 1); diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 1ef464145d2..e6d9bc54c2b 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -132,8 +132,7 @@ int btrfs_remove_ordered_extent(struct inode *inode, int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size); int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, u64 disk_len, int nocow, - int compressed); + u64 start, u64 len, u64 disk_len, int type); int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_extent *entry, struct btrfs_ordered_sum *sum); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 968b84f17a1..e72a013d24b 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -763,6 +763,8 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, if (ret) goto fail; + btrfs_record_root_in_trans(root); + btrfs_set_root_last_snapshot(&root->root_item, trans->transid); memcpy(new_root_item, &root->root_item, sizeof(*new_root_item)); key.objectid = objectid; -- cgit v1.2.3-70-g09d2 From d899e05215178fed903ad0e7fc1cb4d8e0cc0a88 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 30 Oct 2008 14:25:28 -0400 Subject: Btrfs: Add fallocate support v2 This patch updates btrfs-progs for fallocate support. fallocate is a little different in Btrfs because we need to tell the COW system that a given preallocated extent doesn't need to be cow'd as long as there are no snapshots of it. This leverages the -o nodatacow checks. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.h | 9 +- fs/btrfs/extent-tree.c | 65 +++++----- fs/btrfs/extent_io.c | 12 +- fs/btrfs/extent_map.h | 1 + fs/btrfs/file.c | 245 +++++++++++++++++++++++++++++++++++- fs/btrfs/inode.c | 323 +++++++++++++++++++++++++++++++++++++----------- fs/btrfs/ioctl.c | 3 +- fs/btrfs/ordered-data.c | 5 +- fs/btrfs/ordered-data.h | 4 +- fs/btrfs/tree-log.c | 13 +- 10 files changed, 563 insertions(+), 117 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8bf6a085a73..d5ba3d1aaf9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -462,8 +462,9 @@ struct btrfs_root_item { u8 level; } __attribute__ ((__packed__)); -#define BTRFS_FILE_EXTENT_REG 0 -#define BTRFS_FILE_EXTENT_INLINE 1 +#define BTRFS_FILE_EXTENT_INLINE 0 +#define BTRFS_FILE_EXTENT_REG 1 +#define BTRFS_FILE_EXTENT_PREALLOC 2 struct btrfs_file_extent_item { /* @@ -868,6 +869,7 @@ struct btrfs_root { #define BTRFS_INODE_NODATACOW (1 << 1) #define BTRFS_INODE_READONLY (1 << 2) #define BTRFS_INODE_NOCOMPRESS (1 << 3) +#define BTRFS_INODE_PREALLOC (1 << 4) #define btrfs_clear_flag(inode, flag) (BTRFS_I(inode)->flags &= \ ~BTRFS_INODE_##flag) #define btrfs_set_flag(inode, flag) (BTRFS_I(inode)->flags |= \ @@ -1924,6 +1926,9 @@ extern struct file_operations btrfs_file_operations; int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 start, u64 end, u64 inline_limit, u64 *hint_block); +int btrfs_mark_extent_written(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode, u64 start, u64 end); int btrfs_release_file(struct inode *inode, struct file *file); /* tree-defrag.c */ diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 1eb69a91b72..8af39521eb7 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2147,6 +2147,9 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, total_needed += empty_size; block_group = btrfs_lookup_block_group(root->fs_info, search_start); + if (!block_group) + block_group = btrfs_lookup_first_block_group(root->fs_info, + search_start); space_info = __find_space_info(root->fs_info, data); down_read(&space_info->groups_sem); @@ -3426,9 +3429,7 @@ walk_down: next: level--; btrfs_release_path(extent_root, path); - if (need_resched()) { - cond_resched(); - } + cond_resched(); } /* reached lowest level */ ret = 1; @@ -3539,9 +3540,7 @@ found: } btrfs_release_path(extent_root, path); - if (need_resched()) { - cond_resched(); - } + cond_resched(); } /* reached max tree level, but no tree root found. */ BUG(); @@ -3654,8 +3653,9 @@ static int noinline get_new_locations(struct inode *reloc_inode, exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi); exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf, fi); - WARN_ON(exts[nr].offset > 0); - WARN_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes); + BUG_ON(exts[nr].offset > 0); + BUG_ON(exts[nr].compression || exts[nr].encryption); + BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes); cur_pos += exts[nr].num_bytes; nr++; @@ -3709,6 +3709,7 @@ static int noinline replace_one_extent(struct btrfs_trans_handle *trans, u32 nritems; int nr_scaned = 0; int extent_locked = 0; + int extent_type; int ret; memcpy(&key, leaf_key, sizeof(key)); @@ -3781,8 +3782,9 @@ next: } fi = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); - if ((btrfs_file_extent_type(leaf, fi) != - BTRFS_FILE_EXTENT_REG) || + extent_type = btrfs_file_extent_type(leaf, fi); + if ((extent_type != BTRFS_FILE_EXTENT_REG && + extent_type != BTRFS_FILE_EXTENT_PREALLOC) || (btrfs_file_extent_disk_bytenr(leaf, fi) != extent_key->objectid)) { path->slots[0]++; @@ -3865,16 +3867,10 @@ next: if (nr_extents == 1) { /* update extent pointer in place */ - btrfs_set_file_extent_generation(leaf, fi, - trans->transid); btrfs_set_file_extent_disk_bytenr(leaf, fi, new_extents[0].disk_bytenr); btrfs_set_file_extent_disk_num_bytes(leaf, fi, new_extents[0].disk_num_bytes); - btrfs_set_file_extent_ram_bytes(leaf, fi, - new_extents[0].ram_bytes); - ext_offset += new_extents[0].offset; - btrfs_set_file_extent_offset(leaf, fi, ext_offset); btrfs_mark_buffer_dirty(leaf); btrfs_drop_extent_cache(inode, key.offset, @@ -3901,6 +3897,8 @@ next: btrfs_release_path(root, path); key.offset += num_bytes; } else { + BUG_ON(1); +#if 0 u64 alloc_hint; u64 extent_len; int i; @@ -3977,6 +3975,7 @@ next: break; } BUG_ON(i >= nr_extents); +#endif } if (extent_locked) { @@ -4156,15 +4155,10 @@ static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans, ref->extents[ext_index].bytenr = new_extent->disk_bytenr; ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes; - btrfs_set_file_extent_generation(leaf, fi, trans->transid); - btrfs_set_file_extent_ram_bytes(leaf, fi, - new_extent->ram_bytes); btrfs_set_file_extent_disk_bytenr(leaf, fi, new_extent->disk_bytenr); btrfs_set_file_extent_disk_num_bytes(leaf, fi, new_extent->disk_num_bytes); - new_extent->offset += btrfs_file_extent_offset(leaf, fi); - btrfs_set_file_extent_offset(leaf, fi, new_extent->offset); btrfs_mark_buffer_dirty(leaf); ret = btrfs_inc_extent_ref(trans, root, @@ -4625,12 +4619,15 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, */ if (!new_extents) { u64 group_start = group->key.objectid; + new_extents = kmalloc(sizeof(*new_extents), + GFP_NOFS); + nr_extents = 1; ret = get_new_locations(reloc_inode, extent_key, - group_start, 0, + group_start, 1, &new_extents, &nr_extents); - if (ret < 0) + if (ret) goto out; } btrfs_record_root_in_trans(found_root); @@ -4762,7 +4759,8 @@ static int __insert_orphan_inode(struct btrfs_trans_handle *trans, btrfs_set_inode_generation(leaf, item, 1); btrfs_set_inode_size(leaf, item, size); btrfs_set_inode_mode(leaf, item, S_IFREG | 0600); - btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM); + btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM | + BTRFS_INODE_NOCOMPRESS); btrfs_mark_buffer_dirty(leaf); btrfs_release_path(root, path); out: @@ -4835,6 +4833,7 @@ int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start) struct inode *reloc_inode; struct btrfs_block_group_cache *block_group; struct btrfs_key key; + u64 skipped; u64 cur_byte; u64 total_found; u32 nritems; @@ -4864,6 +4863,7 @@ int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start) btrfs_start_delalloc_inodes(info->tree_root); btrfs_wait_ordered_extents(info->tree_root, 0); again: + skipped = 0; total_found = 0; progress = 0; key.objectid = block_group->key.objectid; @@ -4926,6 +4926,8 @@ next: ret = relocate_one_extent(root, path, &key, block_group, reloc_inode, pass); BUG_ON(ret < 0); + if (ret > 0) + skipped++; key.objectid = cur_byte; key.type = 0; @@ -4944,6 +4946,11 @@ next: printk("btrfs found %llu extents in pass %d\n", (unsigned long long)total_found, pass); pass++; + if (total_found == skipped && pass > 2) { + iput(reloc_inode); + reloc_inode = create_reloc_inode(info, block_group); + pass = 0; + } goto again; } @@ -5011,17 +5018,17 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) while ((n = rb_last(&info->block_group_cache_tree)) != NULL) { block_group = rb_entry(n, struct btrfs_block_group_cache, cache_node); - - spin_unlock(&info->block_group_cache_lock); - btrfs_remove_free_space_cache(block_group); - spin_lock(&info->block_group_cache_lock); - rb_erase(&block_group->cache_node, &info->block_group_cache_tree); + spin_unlock(&info->block_group_cache_lock); + + btrfs_remove_free_space_cache(block_group); down_write(&block_group->space_info->groups_sem); list_del(&block_group->list); up_write(&block_group->space_info->groups_sem); kfree(block_group); + + spin_lock(&info->block_group_cache_lock); } spin_unlock(&info->block_group_cache_lock); return 0; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 65a0583027e..eb3c12e7bea 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2015,6 +2015,8 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur); } bdev = em->bdev; block_start = em->block_start; + if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) + block_start = EXTENT_MAP_HOLE; free_extent_map(em); em = NULL; @@ -2769,14 +2771,18 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock, struct inode *inode = mapping->host; u64 start = iblock << inode->i_blkbits; sector_t sector = 0; + size_t blksize = (1 << inode->i_blkbits); struct extent_map *em; - em = get_extent(inode, NULL, 0, start, (1 << inode->i_blkbits), 0); + lock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1, + GFP_NOFS); + em = get_extent(inode, NULL, 0, start, blksize, 0); + unlock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1, + GFP_NOFS); if (!em || IS_ERR(em)) return 0; - if (em->block_start == EXTENT_MAP_INLINE || - em->block_start == EXTENT_MAP_HOLE) + if (em->block_start > EXTENT_MAP_LAST_BYTE) goto out; sector = (em->block_start + start - em->start) >> inode->i_blkbits; diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index e693e1b4ac4..accfedaeb51 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -12,6 +12,7 @@ #define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */ #define EXTENT_FLAG_COMPRESSED 1 #define EXTENT_FLAG_VACANCY 2 /* no file extent item found */ +#define EXTENT_FLAG_PREALLOC 3 /* pre-allocated extent */ struct extent_map { struct rb_node rb_node; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1a0510ad030..238a8e215eb 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -381,7 +381,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, int keep; int slot; int bookend; - int found_type; + int found_type = 0; int found_extent; int found_inline; int recow; @@ -442,7 +442,8 @@ next_slot: extent); other_encoding = btrfs_file_extent_other_encoding(leaf, extent); - if (found_type == BTRFS_FILE_EXTENT_REG) { + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { extent_end = btrfs_file_extent_disk_bytenr(leaf, extent); @@ -609,8 +610,7 @@ next_slot: */ btrfs_set_file_extent_ram_bytes(leaf, extent, ram_bytes); - btrfs_set_file_extent_type(leaf, extent, - BTRFS_FILE_EXTENT_REG); + btrfs_set_file_extent_type(leaf, extent, found_type); btrfs_mark_buffer_dirty(path->nodes[0]); @@ -661,6 +661,243 @@ out: return ret; } +static int extent_mergeable(struct extent_buffer *leaf, int slot, + u64 objectid, u64 bytenr, u64 *start, u64 *end) +{ + struct btrfs_file_extent_item *fi; + struct btrfs_key key; + u64 extent_end; + + if (slot < 0 || slot >= btrfs_header_nritems(leaf)) + return 0; + + btrfs_item_key_to_cpu(leaf, &key, slot); + if (key.objectid != objectid || key.type != BTRFS_EXTENT_DATA_KEY) + return 0; + + fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); + if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG || + btrfs_file_extent_disk_bytenr(leaf, fi) != bytenr || + btrfs_file_extent_compression(leaf, fi) || + btrfs_file_extent_encryption(leaf, fi) || + btrfs_file_extent_other_encoding(leaf, fi)) + return 0; + + extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi); + if ((*start && *start != key.offset) || (*end && *end != extent_end)) + return 0; + + *start = key.offset; + *end = extent_end; + return 1; +} + +/* + * Mark extent in the range start - end as written. + * + * This changes extent type from 'pre-allocated' to 'regular'. If only + * part of extent is marked as written, the extent will be split into + * two or three. + */ +int btrfs_mark_extent_written(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct inode *inode, u64 start, u64 end) +{ + struct extent_buffer *leaf; + struct btrfs_path *path; + struct btrfs_file_extent_item *fi; + struct btrfs_key key; + u64 bytenr; + u64 num_bytes; + u64 extent_end; + u64 extent_offset; + u64 other_start; + u64 other_end; + u64 split = start; + u64 locked_end = end; + int extent_type; + int split_end = 1; + int ret; + + btrfs_drop_extent_cache(inode, start, end - 1, 0); + + path = btrfs_alloc_path(); + BUG_ON(!path); +again: + key.objectid = inode->i_ino; + key.type = BTRFS_EXTENT_DATA_KEY; + if (split == start) + key.offset = split; + else + key.offset = split - 1; + + ret = btrfs_search_slot(trans, root, &key, path, -1, 1); + if (ret > 0 && path->slots[0] > 0) + path->slots[0]--; + + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); + BUG_ON(key.objectid != inode->i_ino || + key.type != BTRFS_EXTENT_DATA_KEY); + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + extent_type = btrfs_file_extent_type(leaf, fi); + BUG_ON(extent_type != BTRFS_FILE_EXTENT_PREALLOC); + extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi); + BUG_ON(key.offset > start || extent_end < end); + + bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); + num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi); + extent_offset = btrfs_file_extent_offset(leaf, fi); + + if (key.offset == start) + split = end; + + if (key.offset == start && extent_end == end) { + int del_nr = 0; + int del_slot = 0; + u64 leaf_owner = btrfs_header_owner(leaf); + u64 leaf_gen = btrfs_header_generation(leaf); + other_start = end; + other_end = 0; + if (extent_mergeable(leaf, path->slots[0] + 1, inode->i_ino, + bytenr, &other_start, &other_end)) { + extent_end = other_end; + del_slot = path->slots[0] + 1; + del_nr++; + ret = btrfs_free_extent(trans, root, bytenr, num_bytes, + leaf->start, leaf_owner, + leaf_gen, inode->i_ino, 0); + BUG_ON(ret); + } + other_start = 0; + other_end = start; + if (extent_mergeable(leaf, path->slots[0] - 1, inode->i_ino, + bytenr, &other_start, &other_end)) { + key.offset = other_start; + del_slot = path->slots[0]; + del_nr++; + ret = btrfs_free_extent(trans, root, bytenr, num_bytes, + leaf->start, leaf_owner, + leaf_gen, inode->i_ino, 0); + BUG_ON(ret); + } + split_end = 0; + if (del_nr == 0) { + btrfs_set_file_extent_type(leaf, fi, + BTRFS_FILE_EXTENT_REG); + goto done; + } + + fi = btrfs_item_ptr(leaf, del_slot - 1, + struct btrfs_file_extent_item); + btrfs_set_file_extent_type(leaf, fi, BTRFS_FILE_EXTENT_REG); + btrfs_set_file_extent_num_bytes(leaf, fi, + extent_end - key.offset); + btrfs_mark_buffer_dirty(leaf); + + ret = btrfs_del_items(trans, root, path, del_slot, del_nr); + BUG_ON(ret); + goto done; + } else if (split == start) { + if (locked_end < extent_end) { + ret = try_lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, GFP_NOFS); + if (!ret) { + btrfs_release_path(root, path); + lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, GFP_NOFS); + locked_end = extent_end; + goto again; + } + locked_end = extent_end; + } + btrfs_set_file_extent_num_bytes(leaf, fi, split - key.offset); + extent_offset += split - key.offset; + } else { + BUG_ON(key.offset != start); + btrfs_set_file_extent_offset(leaf, fi, extent_offset + + split - key.offset); + btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - split); + key.offset = split; + btrfs_set_item_key_safe(trans, root, path, &key); + extent_end = split; + } + + if (extent_end == end) { + split_end = 0; + extent_type = BTRFS_FILE_EXTENT_REG; + } + if (extent_end == end && split == start) { + other_start = end; + other_end = 0; + if (extent_mergeable(leaf, path->slots[0] + 1, inode->i_ino, + bytenr, &other_start, &other_end)) { + path->slots[0]++; + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + key.offset = split; + btrfs_set_item_key_safe(trans, root, path, &key); + btrfs_set_file_extent_offset(leaf, fi, extent_offset); + btrfs_set_file_extent_num_bytes(leaf, fi, + other_end - split); + goto done; + } + } + if (extent_end == end && split == end) { + other_start = 0; + other_end = start; + if (extent_mergeable(leaf, path->slots[0] - 1 , inode->i_ino, + bytenr, &other_start, &other_end)) { + path->slots[0]--; + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - + other_start); + goto done; + } + } + + btrfs_mark_buffer_dirty(leaf); + btrfs_release_path(root, path); + + key.offset = start; + ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*fi)); + BUG_ON(ret); + + leaf = path->nodes[0]; + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, fi, trans->transid); + btrfs_set_file_extent_type(leaf, fi, extent_type); + btrfs_set_file_extent_disk_bytenr(leaf, fi, bytenr); + btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes); + btrfs_set_file_extent_offset(leaf, fi, extent_offset); + btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - key.offset); + btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes); + btrfs_set_file_extent_compression(leaf, fi, 0); + btrfs_set_file_extent_encryption(leaf, fi, 0); + btrfs_set_file_extent_other_encoding(leaf, fi, 0); + + ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, + leaf->start, root->root_key.objectid, + trans->transid, inode->i_ino); + BUG_ON(ret); +done: + btrfs_mark_buffer_dirty(leaf); + btrfs_release_path(root, path); + if (split_end && split == start) { + split = end; + goto again; + } + if (locked_end > end) { + unlock_extent(&BTRFS_I(inode)->io_tree, end, locked_end - 1, + GFP_NOFS); + } + btrfs_free_path(path); + return 0; +} + /* * this gets pages into the page cache and locks them down, it also properly * waits for data=ordered extents to finish before allowing the pages to be diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3e6f0568fdb..789c376157f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -37,6 +37,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -587,7 +588,7 @@ free_pages_out: * blocks on disk */ static int run_delalloc_nocow(struct inode *inode, struct page *locked_page, - u64 start, u64 end, int *page_started) + u64 start, u64 end, int *page_started, int force) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; @@ -602,6 +603,7 @@ static int run_delalloc_nocow(struct inode *inode, struct page *locked_page, u64 num_bytes; int extent_type; int ret; + int type; int nocow; int check_prev = 1; @@ -654,7 +656,8 @@ next_slot: struct btrfs_file_extent_item); extent_type = btrfs_file_extent_type(leaf, fi); - if (extent_type == BTRFS_FILE_EXTENT_REG) { + if (extent_type == BTRFS_FILE_EXTENT_REG || + extent_type == BTRFS_FILE_EXTENT_PREALLOC) { struct btrfs_block_group_cache *block_group; disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); extent_end = found_key.offset + @@ -669,6 +672,8 @@ next_slot: goto out_check; if (disk_bytenr == 0) goto out_check; + if (extent_type == BTRFS_FILE_EXTENT_REG && !force) + goto out_check; if (btrfs_cross_ref_exist(trans, root, disk_bytenr)) goto out_check; block_group = btrfs_lookup_block_group(root->fs_info, @@ -709,10 +714,39 @@ out_check: disk_bytenr += cur_offset - found_key.offset; num_bytes = min(end + 1, extent_end) - cur_offset; + if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) { + struct extent_map *em; + struct extent_map_tree *em_tree; + em_tree = &BTRFS_I(inode)->extent_tree; + em = alloc_extent_map(GFP_NOFS); + em->start = cur_offset; + em->len = num_bytes; + em->block_len = num_bytes; + em->block_start = disk_bytenr; + em->bdev = root->fs_info->fs_devices->latest_bdev; + set_bit(EXTENT_FLAG_PINNED, &em->flags); + while (1) { + spin_lock(&em_tree->lock); + ret = add_extent_mapping(em_tree, em); + spin_unlock(&em_tree->lock); + if (ret != -EEXIST) { + free_extent_map(em); + break; + } + btrfs_drop_extent_cache(inode, em->start, + em->start + em->len - 1, 0); + } + type = BTRFS_ORDERED_PREALLOC; + } else { + type = BTRFS_ORDERED_NOCOW; + } ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr, - num_bytes, num_bytes, - BTRFS_ORDERED_NOCOW); + num_bytes, num_bytes, type); + BUG_ON(ret); + extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, + cur_offset, cur_offset + num_bytes - 1, + locked_page, 0, 0, 0); cur_offset = extent_end; if (cur_offset > end) break; @@ -745,7 +779,10 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, if (btrfs_test_opt(root, NODATACOW) || btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started); + page_started, 0); + else if (btrfs_test_flag(inode, PREALLOC)) + ret = run_delalloc_nocow(inode, locked_page, start, end, + page_started, 1); else ret = cow_file_range(inode, locked_page, start, end, page_started); @@ -1006,6 +1043,63 @@ int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) return -EAGAIN; } +static int insert_reserved_file_extent(struct btrfs_trans_handle *trans, + struct inode *inode, u64 file_pos, + u64 disk_bytenr, u64 disk_num_bytes, + u64 num_bytes, u64 ram_bytes, + u8 compression, u8 encryption, + u16 other_encoding, int extent_type) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_file_extent_item *fi; + struct btrfs_path *path; + struct extent_buffer *leaf; + struct btrfs_key ins; + u64 hint; + int ret; + + path = btrfs_alloc_path(); + BUG_ON(!path); + + ret = btrfs_drop_extents(trans, root, inode, file_pos, + file_pos + num_bytes, file_pos, &hint); + BUG_ON(ret); + + ins.objectid = inode->i_ino; + ins.offset = file_pos; + ins.type = BTRFS_EXTENT_DATA_KEY; + ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi)); + BUG_ON(ret); + leaf = path->nodes[0]; + fi = btrfs_item_ptr(leaf, path->slots[0], + struct btrfs_file_extent_item); + btrfs_set_file_extent_generation(leaf, fi, trans->transid); + btrfs_set_file_extent_type(leaf, fi, extent_type); + btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr); + btrfs_set_file_extent_disk_num_bytes(leaf, fi, disk_num_bytes); + btrfs_set_file_extent_offset(leaf, fi, 0); + btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes); + btrfs_set_file_extent_ram_bytes(leaf, fi, ram_bytes); + btrfs_set_file_extent_compression(leaf, fi, compression); + btrfs_set_file_extent_encryption(leaf, fi, encryption); + btrfs_set_file_extent_other_encoding(leaf, fi, other_encoding); + btrfs_mark_buffer_dirty(leaf); + + inode_add_bytes(inode, num_bytes); + btrfs_drop_extent_cache(inode, file_pos, file_pos + num_bytes - 1, 0); + + ins.objectid = disk_bytenr; + ins.offset = disk_num_bytes; + ins.type = BTRFS_EXTENT_ITEM_KEY; + ret = btrfs_alloc_reserved_extent(trans, root, leaf->start, + root->root_key.objectid, + trans->transid, inode->i_ino, &ins); + BUG_ON(ret); + + btrfs_free_path(path); + return 0; +} + /* as ordered data IO finishes, this gets called so we can finish * an ordered extent if the range of bytes in the file it covers are * fully written. @@ -1016,12 +1110,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) struct btrfs_trans_handle *trans; struct btrfs_ordered_extent *ordered_extent; struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; - struct btrfs_file_extent_item *extent_item; - struct btrfs_path *path = NULL; - struct extent_buffer *leaf; - u64 alloc_hint = 0; - struct list_head list; - struct btrfs_key ins; + int compressed = 0; int ret; ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); @@ -1035,67 +1124,30 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) goto nocow; - path = btrfs_alloc_path(); - BUG_ON(!path); - lock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, GFP_NOFS); - INIT_LIST_HEAD(&list); - - ret = btrfs_drop_extents(trans, root, inode, - ordered_extent->file_offset, - ordered_extent->file_offset + - ordered_extent->len, - ordered_extent->file_offset, &alloc_hint); - BUG_ON(ret); - - ins.objectid = inode->i_ino; - ins.offset = ordered_extent->file_offset; - ins.type = BTRFS_EXTENT_DATA_KEY; - ret = btrfs_insert_empty_item(trans, root, path, &ins, - sizeof(*extent_item)); - BUG_ON(ret); - leaf = path->nodes[0]; - extent_item = btrfs_item_ptr(leaf, path->slots[0], - struct btrfs_file_extent_item); - btrfs_set_file_extent_generation(leaf, extent_item, trans->transid); - btrfs_set_file_extent_type(leaf, extent_item, BTRFS_FILE_EXTENT_REG); - btrfs_set_file_extent_disk_bytenr(leaf, extent_item, - ordered_extent->start); - btrfs_set_file_extent_disk_num_bytes(leaf, extent_item, - ordered_extent->disk_len); - btrfs_set_file_extent_offset(leaf, extent_item, 0); - if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) - btrfs_set_file_extent_compression(leaf, extent_item, 1); - else - btrfs_set_file_extent_compression(leaf, extent_item, 0); - btrfs_set_file_extent_encryption(leaf, extent_item, 0); - btrfs_set_file_extent_other_encoding(leaf, extent_item, 0); - - /* ram bytes = extent_num_bytes for now */ - btrfs_set_file_extent_num_bytes(leaf, extent_item, - ordered_extent->len); - btrfs_set_file_extent_ram_bytes(leaf, extent_item, - ordered_extent->len); - btrfs_mark_buffer_dirty(leaf); - - btrfs_drop_extent_cache(inode, ordered_extent->file_offset, - ordered_extent->file_offset + - ordered_extent->len - 1, 0); - - ins.objectid = ordered_extent->start; - ins.offset = ordered_extent->disk_len; - ins.type = BTRFS_EXTENT_ITEM_KEY; - ret = btrfs_alloc_reserved_extent(trans, root, leaf->start, - root->root_key.objectid, - trans->transid, inode->i_ino, &ins); - BUG_ON(ret); - btrfs_release_path(root, path); - - inode_add_bytes(inode, ordered_extent->len); + compressed = 1; + if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) { + BUG_ON(compressed); + ret = btrfs_mark_extent_written(trans, root, inode, + ordered_extent->file_offset, + ordered_extent->file_offset + + ordered_extent->len); + BUG_ON(ret); + } else { + ret = insert_reserved_file_extent(trans, inode, + ordered_extent->file_offset, + ordered_extent->start, + ordered_extent->disk_len, + ordered_extent->len, + ordered_extent->len, + compressed, 0, 0, + BTRFS_FILE_EXTENT_REG); + BUG_ON(ret); + } unlock_extent(io_tree, ordered_extent->file_offset, ordered_extent->file_offset + ordered_extent->len - 1, GFP_NOFS); @@ -1115,8 +1167,6 @@ nocow: btrfs_put_ordered_extent(ordered_extent); btrfs_end_transaction(trans, root); - if (path) - btrfs_free_path(path); return 0; } @@ -3488,7 +3538,8 @@ again: found_type = btrfs_file_extent_type(leaf, item); extent_start = found_key.offset; compressed = btrfs_file_extent_compression(leaf, item); - if (found_type == BTRFS_FILE_EXTENT_REG) { + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { extent_end = extent_start + btrfs_file_extent_num_bytes(leaf, item); } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { @@ -3521,7 +3572,8 @@ again: goto not_found_em; } - if (found_type == BTRFS_FILE_EXTENT_REG) { + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { em->start = extent_start; em->len = extent_end - extent_start; bytenr = btrfs_file_extent_disk_bytenr(leaf, item); @@ -3538,6 +3590,8 @@ again: bytenr += btrfs_file_extent_offset(leaf, item); em->block_start = bytenr; em->block_len = em->len; + if (found_type == BTRFS_FILE_EXTENT_PREALLOC) + set_bit(EXTENT_FLAG_PREALLOC, &em->flags); } goto insert; } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { @@ -3969,6 +4023,7 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, if (error) return error; + atomic_inc(&inode->i_count); d_instantiate(dentry, inode); return 0; } @@ -4318,6 +4373,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_symlink_inode_operations; inode->i_mapping->a_ops = &btrfs_symlink_aops; inode->i_mapping->backing_dev_info = &root->fs_info->bdi; + inode_set_bytes(inode, name_len); btrfs_i_size_write(inode, name_len - 1); err = btrfs_update_inode(trans, root, inode); if (err) @@ -4335,6 +4391,130 @@ out_fail: return err; } +static int prealloc_file_range(struct inode *inode, u64 start, u64 end, + u64 alloc_hint, int mode) +{ + struct btrfs_trans_handle *trans; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_key ins; + u64 alloc_size; + u64 cur_offset = start; + u64 num_bytes = end - start; + int ret = 0; + + trans = btrfs_join_transaction(root, 1); + BUG_ON(!trans); + btrfs_set_trans_block_group(trans, inode); + + while (num_bytes > 0) { + alloc_size = min(num_bytes, root->fs_info->max_extent); + ret = btrfs_reserve_extent(trans, root, alloc_size, + root->sectorsize, 0, alloc_hint, + (u64)-1, &ins, 1); + if (ret) { + WARN_ON(1); + goto out; + } + ret = insert_reserved_file_extent(trans, inode, + cur_offset, ins.objectid, + ins.offset, ins.offset, + ins.offset, 0, 0, 0, + BTRFS_FILE_EXTENT_PREALLOC); + BUG_ON(ret); + num_bytes -= ins.offset; + cur_offset += ins.offset; + alloc_hint = ins.objectid + ins.offset; + } +out: + if (cur_offset > start) { + inode->i_ctime = CURRENT_TIME; + btrfs_set_flag(inode, PREALLOC); + if (!(mode & FALLOC_FL_KEEP_SIZE) && + cur_offset > i_size_read(inode)) + btrfs_i_size_write(inode, cur_offset); + ret = btrfs_update_inode(trans, root, inode); + BUG_ON(ret); + } + + btrfs_end_transaction(trans, root); + return ret; +} + +static long btrfs_fallocate(struct inode *inode, int mode, + loff_t offset, loff_t len) +{ + u64 cur_offset; + u64 last_byte; + u64 alloc_start; + u64 alloc_end; + u64 alloc_hint = 0; + u64 mask = BTRFS_I(inode)->root->sectorsize - 1; + struct extent_map *em; + int ret; + + alloc_start = offset & ~mask; + alloc_end = (offset + len + mask) & ~mask; + + mutex_lock(&inode->i_mutex); + if (alloc_start > inode->i_size) { + ret = btrfs_cont_expand(inode, alloc_start); + if (ret) + goto out; + } + + while (1) { + struct btrfs_ordered_extent *ordered; + lock_extent(&BTRFS_I(inode)->io_tree, alloc_start, + alloc_end - 1, GFP_NOFS); + ordered = btrfs_lookup_first_ordered_extent(inode, + alloc_end - 1); + if (ordered && + ordered->file_offset + ordered->len > alloc_start && + ordered->file_offset < alloc_end) { + btrfs_put_ordered_extent(ordered); + unlock_extent(&BTRFS_I(inode)->io_tree, + alloc_start, alloc_end - 1, GFP_NOFS); + btrfs_wait_ordered_range(inode, alloc_start, + alloc_end - alloc_start); + } else { + if (ordered) + btrfs_put_ordered_extent(ordered); + break; + } + } + + cur_offset = alloc_start; + while (1) { + em = btrfs_get_extent(inode, NULL, 0, cur_offset, + alloc_end - cur_offset, 0); + BUG_ON(IS_ERR(em) || !em); + last_byte = min(extent_map_end(em), alloc_end); + last_byte = (last_byte + mask) & ~mask; + if (em->block_start == EXTENT_MAP_HOLE) { + ret = prealloc_file_range(inode, cur_offset, + last_byte, alloc_hint, mode); + if (ret < 0) { + free_extent_map(em); + break; + } + } + if (em->block_start <= EXTENT_MAP_LAST_BYTE) + alloc_hint = em->block_start; + free_extent_map(em); + + cur_offset = last_byte; + if (cur_offset >= alloc_end) { + ret = 0; + break; + } + } + unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, alloc_end - 1, + GFP_NOFS); +out: + mutex_unlock(&inode->i_mutex); + return ret; +} + static int btrfs_set_page_dirty(struct page *page) { return __set_page_dirty_nobuffers(page); @@ -4421,6 +4601,7 @@ static struct inode_operations btrfs_file_inode_operations = { .listxattr = btrfs_listxattr, .removexattr = btrfs_removexattr, .permission = btrfs_permission, + .fallocate = btrfs_fallocate, }; static struct inode_operations btrfs_special_inode_operations = { .getattr = btrfs_getattr, diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 7f915d47839..9ff2b4e0e92 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -724,7 +724,8 @@ long btrfs_ioctl_clone(struct file *file, unsigned long src_fd) extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); found_type = btrfs_file_extent_type(leaf, extent); - if (found_type == BTRFS_FILE_EXTENT_REG) { + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { u64 ds = btrfs_file_extent_disk_bytenr(leaf, extent); u64 dl = btrfs_file_extent_disk_num_bytes(leaf, diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index e7317c8fda2..370bb428559 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -182,7 +182,7 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, entry->len = len; entry->disk_len = disk_len; entry->inode = inode; - if (type == BTRFS_ORDERED_NOCOW || type == BTRFS_ORDERED_COMPRESSED) + if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE) set_bit(type, &entry->flags); /* one ref for the tree */ @@ -339,7 +339,8 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nocow_only) ordered = list_entry(cur, struct btrfs_ordered_extent, root_extent_list); if (nocow_only && - !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) { + !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags) && + !test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags)) { list_move(&ordered->root_extent_list, &root->fs_info->ordered_extents); cond_resched_lock(&root->fs_info->ordered_extent_lock); diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index e6d9bc54c2b..260bf95dfe0 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -68,6 +68,8 @@ struct btrfs_ordered_sum { #define BTRFS_ORDERED_COMPRESSED 3 /* writing a compressed extent */ +#define BTRFS_ORDERED_PREALLOC 4 /* set when writing to prealloced extent */ + struct btrfs_ordered_extent { /* logical offset in the file */ u64 file_offset; @@ -132,7 +134,7 @@ int btrfs_remove_ordered_extent(struct inode *inode, int btrfs_dec_test_ordered_pending(struct inode *inode, u64 file_offset, u64 io_size); int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, u64 disk_len, int type); + u64 start, u64 len, u64 disk_len, int tyep); int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_extent *entry, struct btrfs_ordered_sum *sum); diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index e0201c3a7dc..be4fc30a30e 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -442,7 +442,8 @@ insert: fi = (struct btrfs_file_extent_item *)dst_ptr; extent_type = btrfs_file_extent_type(path->nodes[0], fi); - if (extent_type == BTRFS_FILE_EXTENT_REG) { + if (extent_type == BTRFS_FILE_EXTENT_REG || + extent_type == BTRFS_FILE_EXTENT_PREALLOC) { struct btrfs_key ins; ins.objectid = btrfs_file_extent_disk_bytenr( path->nodes[0], fi); @@ -538,7 +539,8 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); found_type = btrfs_file_extent_type(eb, item); - if (found_type == BTRFS_FILE_EXTENT_REG) + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) extent_end = start + btrfs_file_extent_num_bytes(eb, item); else if (found_type == BTRFS_FILE_EXTENT_INLINE) { size = btrfs_file_extent_inline_len(eb, item); @@ -562,7 +564,9 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, start, 0); - if (ret == 0 && found_type == BTRFS_FILE_EXTENT_REG) { + if (ret == 0 && + (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC)) { struct btrfs_file_extent_item cmp1; struct btrfs_file_extent_item cmp2; struct btrfs_file_extent_item *existing; @@ -2522,7 +2526,8 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, struct btrfs_file_extent_item); found_type = btrfs_file_extent_type(src, extent); - if (found_type == BTRFS_FILE_EXTENT_REG) { + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { u64 ds = btrfs_file_extent_disk_bytenr(src, extent); u64 dl = btrfs_file_extent_disk_num_bytes(src, -- cgit v1.2.3-70-g09d2 From 70b99e6959a4c28ae1b314985eca731f3db72f1d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 31 Oct 2008 12:46:39 -0400 Subject: Btrfs: Compression corner fixes Make sure we keep page->mapping NULL on the pages we're getting via alloc_page. It gets set so a few of the callbacks can do the right thing, but in general these pages don't have a mapping. Don't try to truncate compressed inline items in btrfs_drop_extents. The whole compressed item must be preserved. Don't try to create multipage inline compressed items. When we try to overwrite just the first page of the file, we would have to read in and recow all the pages after it in the same compressed inline items. For now, only create single page inline items. Make sure we lock pages in the correct order during delalloc. The search into the state tree for delalloc bytes can return bytes before the page we already have locked. Signed-off-by: Chris Mason --- fs/btrfs/compression.c | 4 ++++ fs/btrfs/extent_io.c | 11 ++++++++++- fs/btrfs/file.c | 21 +++++++++++++++------ fs/btrfs/inode.c | 8 ++++++-- 4 files changed, 35 insertions(+), 9 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 9adaa79adad..354913177ba 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -220,10 +220,12 @@ static void end_compressed_bio_write(struct bio *bio, int err) */ inode = cb->inode; tree = &BTRFS_I(inode)->io_tree; + cb->compressed_pages[0]->mapping = cb->inode->i_mapping; tree->ops->writepage_end_io_hook(cb->compressed_pages[0], cb->start, cb->start + cb->len - 1, NULL, 1); + cb->compressed_pages[0]->mapping = NULL; end_compressed_writeback(inode, cb->start, cb->len); /* note, our inode could be gone now */ @@ -306,6 +308,7 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, else ret = 0; + page->mapping = NULL; if (ret || bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) { bio_get(bio); @@ -423,6 +426,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, else ret = 0; + page->mapping = NULL; if (ret || bio_add_page(comp_bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) { bio_get(comp_bio); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index eb3c12e7bea..9b37ce6e516 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1242,12 +1242,21 @@ again: delalloc_end = 0; found = find_delalloc_range(tree, &delalloc_start, &delalloc_end, max_bytes); - if (!found) { + if (!found || delalloc_end <= *start) { *start = delalloc_start; *end = delalloc_end; return found; } + /* + * start comes from the offset of locked_page. We have to lock + * pages in order, so we can't process delalloc bytes before + * locked_page + */ + if (delalloc_start < *start) { + delalloc_start = *start; + } + /* * make sure to limit the number of pages we try to lock down * if we're looping. diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 238a8e215eb..0c8cc35a8b9 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -368,8 +368,8 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, u64 search_start = start; u64 leaf_start; u64 ram_bytes = 0; - u8 compression = 0; - u8 encryption = 0; + u8 compression; + u8 encryption; u16 other_encoding = 0; u64 root_gen; u64 root_owner; @@ -415,6 +415,8 @@ next_slot: leaf_start = 0; root_gen = 0; root_owner = 0; + compression = 0; + encryption = 0; extent = NULL; leaf = path->nodes[0]; slot = path->slots[0]; @@ -546,8 +548,12 @@ next_slot: inline_limit - key.offset); inode_sub_bytes(inode, extent_end - inline_limit); - btrfs_truncate_item(trans, root, path, - new_size, 1); + btrfs_set_file_extent_ram_bytes(leaf, extent, + new_size); + if (!compression && !encryption) { + btrfs_truncate_item(trans, root, path, + new_size, 1); + } } } /* delete the entire extent */ @@ -567,8 +573,11 @@ next_slot: new_size = btrfs_file_extent_calc_inline_size( extent_end - end); inode_sub_bytes(inode, end - key.offset); - ret = btrfs_truncate_item(trans, root, path, - new_size, 0); + btrfs_set_file_extent_ram_bytes(leaf, extent, + new_size); + if (!compression && !encryption) + ret = btrfs_truncate_item(trans, root, path, + new_size, 0); BUG_ON(ret); } /* create bookend, splitting the extent in two */ diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 789c376157f..806708dd7e3 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -239,6 +239,7 @@ static int cow_file_range_inline(struct btrfs_trans_handle *trans, data_len = compressed_size; if (start > 0 || + actual_end >= PAGE_CACHE_SIZE || data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) || (!compressed_size && (actual_end & (root->sectorsize - 1)) == 0) || @@ -248,7 +249,7 @@ static int cow_file_range_inline(struct btrfs_trans_handle *trans, } ret = btrfs_drop_extents(trans, root, inode, start, - aligned_end, aligned_end, &hint_byte); + aligned_end, start, &hint_byte); BUG_ON(ret); if (isize > actual_end) @@ -423,6 +424,7 @@ again: * free any pages it allocated and our page pointer array */ for (i = 0; i < nr_pages_ret; i++) { + WARN_ON(pages[i]->mapping); page_cache_release(pages[i]); } kfree(pages); @@ -572,8 +574,10 @@ free_pages_out_fail: extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, start, end, locked_page, 0, 0, 0); free_pages_out: - for (i = 0; i < nr_pages_ret; i++) + for (i = 0; i < nr_pages_ret; i++) { + WARN_ON(pages[i]->mapping); page_cache_release(pages[i]); + } if (pages) kfree(pages); -- cgit v1.2.3-70-g09d2 From 4a69a41009c4ac691f7d9c289f5f37fabeddce46 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 6 Nov 2008 22:03:00 -0500 Subject: Btrfs: Add ordered async work queues Btrfs uses kernel threads to create async work queues for cpu intensive operations such as checksumming and decompression. These work well, but they make it difficult to keep IO order intact. A single writepages call from pdflush or fsync will turn into a number of bios, and each bio is checksummed in parallel. Once the checksum is computed, the bio is sent down to the disk, and since we don't control the order in which the parallel operations happen, they might go down to the disk in almost any order. The code deals with this somewhat by having deep work queues for a single kernel thread, making it very likely that a single thread will process all the bios for a single inode. This patch introduces an explicitly ordered work queue. As work structs are placed into the queue they are put onto the tail of a list. They have three callbacks: ->func (cpu intensive processing here) ->ordered_func (order sensitive processing here) ->ordered_free (free the work struct, all processing is done) The work struct has three callbacks. The func callback does the cpu intensive work, and when it completes the work struct is marked as done. Every time a work struct completes, the list is checked to see if the head is marked as done. If so the ordered_func callback is used to do the order sensitive processing and the ordered_free callback is used to do any cleanup. Then we loop back and check the head of the list again. This patch also changes the checksumming code to use the ordered workqueues. One a 4 drive array, it increases streaming writes from 280MB/s to 350MB/s. Signed-off-by: Chris Mason --- fs/btrfs/async-thread.c | 67 ++++++++++++++++++++++++++++++++++++++-- fs/btrfs/async-thread.h | 18 ++++++++++- fs/btrfs/disk-io.c | 81 ++++++++++++++++++++++++++++++++++--------------- fs/btrfs/disk-io.h | 4 ++- fs/btrfs/inode.c | 19 ++++++++++-- 5 files changed, 158 insertions(+), 31 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index d82efd722a4..e1e49715459 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -23,6 +23,10 @@ # include #include "async-thread.h" +#define WORK_QUEUED_BIT 0 +#define WORK_DONE_BIT 1 +#define WORK_ORDER_DONE_BIT 2 + /* * container for the kthread task pointer and the list of pending work * One of these is allocated per thread. @@ -88,6 +92,47 @@ static void check_busy_worker(struct btrfs_worker_thread *worker) } } +static noinline int run_ordered_completions(struct btrfs_workers *workers, + struct btrfs_work *work) +{ + unsigned long flags; + + if (!workers->ordered) + return 0; + + set_bit(WORK_DONE_BIT, &work->flags); + + spin_lock_irqsave(&workers->lock, flags); + + while(!list_empty(&workers->order_list)) { + work = list_entry(workers->order_list.next, + struct btrfs_work, order_list); + + if (!test_bit(WORK_DONE_BIT, &work->flags)) + break; + + /* we are going to call the ordered done function, but + * we leave the work item on the list as a barrier so + * that later work items that are done don't have their + * functions called before this one returns + */ + if (test_and_set_bit(WORK_ORDER_DONE_BIT, &work->flags)) + break; + + spin_unlock_irqrestore(&workers->lock, flags); + + work->ordered_func(work); + + /* now take the lock again and call the freeing code */ + spin_lock_irqsave(&workers->lock, flags); + list_del(&work->order_list); + work->ordered_free(work); + } + + spin_unlock_irqrestore(&workers->lock, flags); + return 0; +} + /* * main loop for servicing work items */ @@ -102,7 +147,7 @@ static int worker_loop(void *arg) cur = worker->pending.next; work = list_entry(cur, struct btrfs_work, list); list_del(&work->list); - clear_bit(0, &work->flags); + clear_bit(WORK_QUEUED_BIT, &work->flags); work->worker = worker; spin_unlock_irq(&worker->lock); @@ -110,8 +155,15 @@ static int worker_loop(void *arg) work->func(work); atomic_dec(&worker->num_pending); + /* + * unless this is an ordered work queue, + * 'work' was probably freed by func above. + */ + run_ordered_completions(worker->workers, work); + spin_lock_irq(&worker->lock); check_idle_worker(worker); + } worker->working = 0; if (freezing(current)) { @@ -154,10 +206,12 @@ void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max) workers->num_workers = 0; INIT_LIST_HEAD(&workers->worker_list); INIT_LIST_HEAD(&workers->idle_list); + INIT_LIST_HEAD(&workers->order_list); spin_lock_init(&workers->lock); workers->max_workers = max; workers->idle_thresh = 32; workers->name = name; + workers->ordered = 0; } /* @@ -296,7 +350,7 @@ int btrfs_requeue_work(struct btrfs_work *work) struct btrfs_worker_thread *worker = work->worker; unsigned long flags; - if (test_and_set_bit(0, &work->flags)) + if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags)) goto out; spin_lock_irqsave(&worker->lock, flags); @@ -330,10 +384,17 @@ int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work) int wake = 0; /* don't requeue something already on a list */ - if (test_and_set_bit(0, &work->flags)) + if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags)) goto out; worker = find_worker(workers); + if (workers->ordered) { + spin_lock_irqsave(&workers->lock, flags); + list_add_tail(&work->order_list, &workers->order_list); + spin_unlock_irqrestore(&workers->lock, flags); + } else { + INIT_LIST_HEAD(&work->order_list); + } spin_lock_irqsave(&worker->lock, flags); atomic_inc(&worker->num_pending); diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h index 4ec9a2ee0f9..31be4ed8b63 100644 --- a/fs/btrfs/async-thread.h +++ b/fs/btrfs/async-thread.h @@ -37,10 +37,16 @@ struct btrfs_worker_thread; */ struct btrfs_work { /* - * only func should be set to the function you want called + * func should be set to the function you want called * your work struct is passed as the only arg + * + * ordered_func must be set for work sent to an ordered work queue, + * and it is called to complete a given work item in the same + * order they were sent to the queue. */ void (*func)(struct btrfs_work *work); + void (*ordered_func)(struct btrfs_work *work); + void (*ordered_free)(struct btrfs_work *work); /* * flags should be set to zero. It is used to make sure the @@ -51,6 +57,7 @@ struct btrfs_work { /* don't touch these */ struct btrfs_worker_thread *worker; struct list_head list; + struct list_head order_list; }; struct btrfs_workers { @@ -63,6 +70,9 @@ struct btrfs_workers { /* once a worker has this many requests or fewer, it is idle */ int idle_thresh; + /* force completions in the order they were queued */ + int ordered; + /* list with all the work threads. The workers on the idle thread * may be actively servicing jobs, but they haven't yet hit the * idle thresh limit above. @@ -70,6 +80,12 @@ struct btrfs_workers { struct list_head worker_list; struct list_head idle_list; + /* + * when operating in ordered mode, this maintains the list + * of work items waiting for completion + */ + struct list_head order_list; + /* lock for finding the next worker thread to queue on */ spinlock_t lock; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 94b4e50f6b2..e0a28f705a6 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -80,7 +80,8 @@ struct async_submit_bio { struct inode *inode; struct bio *bio; struct list_head list; - extent_submit_bio_hook_t *submit_bio_hook; + extent_submit_bio_hook_t *submit_bio_start; + extent_submit_bio_hook_t *submit_bio_done; int rw; int mirror_num; unsigned long bio_flags; @@ -452,7 +453,18 @@ int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) btrfs_async_submit_limit(info); } -static void run_one_async_submit(struct btrfs_work *work) +static void run_one_async_start(struct btrfs_work *work) +{ + struct btrfs_fs_info *fs_info; + struct async_submit_bio *async; + + async = container_of(work, struct async_submit_bio, work); + fs_info = BTRFS_I(async->inode)->root->fs_info; + async->submit_bio_start(async->inode, async->rw, async->bio, + async->mirror_num, async->bio_flags); +} + +static void run_one_async_done(struct btrfs_work *work) { struct btrfs_fs_info *fs_info; struct async_submit_bio *async; @@ -470,15 +482,23 @@ static void run_one_async_submit(struct btrfs_work *work) waitqueue_active(&fs_info->async_submit_wait)) wake_up(&fs_info->async_submit_wait); - async->submit_bio_hook(async->inode, async->rw, async->bio, + async->submit_bio_done(async->inode, async->rw, async->bio, async->mirror_num, async->bio_flags); +} + +static void run_one_async_free(struct btrfs_work *work) +{ + struct async_submit_bio *async; + + async = container_of(work, struct async_submit_bio, work); kfree(async); } int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags, - extent_submit_bio_hook_t *submit_bio_hook) + extent_submit_bio_hook_t *submit_bio_start, + extent_submit_bio_hook_t *submit_bio_done) { struct async_submit_bio *async; int limit = btrfs_async_submit_limit(fs_info); @@ -491,8 +511,13 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, async->rw = rw; async->bio = bio; async->mirror_num = mirror_num; - async->submit_bio_hook = submit_bio_hook; - async->work.func = run_one_async_submit; + async->submit_bio_start = submit_bio_start; + async->submit_bio_done = submit_bio_done; + + async->work.func = run_one_async_start; + async->work.ordered_func = run_one_async_done; + async->work.ordered_free = run_one_async_free; + async->work.flags = 0; async->bio_flags = bio_flags; @@ -533,29 +558,25 @@ static int btree_csum_one_bio(struct bio *bio) return 0; } -static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, - int mirror_num, unsigned long bio_flags) +static int __btree_submit_bio_start(struct inode *inode, int rw, + struct bio *bio, int mirror_num, + unsigned long bio_flags) { - struct btrfs_root *root = BTRFS_I(inode)->root; - int ret; - /* * when we're called for a write, we're already in the async * submission context. Just jump into btrfs_map_bio */ - if (rw & (1 << BIO_RW)) { - btree_csum_one_bio(bio); - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, - mirror_num, 1); - } + btree_csum_one_bio(bio); + return 0; +} +static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, + int mirror_num, unsigned long bio_flags) +{ /* - * called for a read, do the setup so that checksum validation - * can happen in the async kernel threads + * when we're called for a write, we're already in the async + * submission context. Just jump into btrfs_map_bio */ - ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); - BUG_ON(ret); - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); } @@ -567,11 +588,22 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, * can happen in parallel across all CPUs */ if (!(rw & (1 << BIO_RW))) { - return __btree_submit_bio_hook(inode, rw, bio, mirror_num, 0); + int ret; + /* + * called for a read, do the setup so that checksum validation + * can happen in the async kernel threads + */ + ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, + bio, 1); + BUG_ON(ret); + + return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, + mirror_num, 1); } return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, 0, - __btree_submit_bio_hook); + __btree_submit_bio_start, + __btree_submit_bio_done); } static int btree_writepage(struct page *page, struct writeback_control *wbc) @@ -1534,7 +1566,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, * were sent by the writeback daemons, improving overall locality * of the IO going down the pipe. */ - fs_info->workers.idle_thresh = 128; + fs_info->workers.idle_thresh = 8; + fs_info->workers.ordered = 1; btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); btrfs_init_workers(&fs_info->endio_workers, "endio", diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 4eb1f1408d2..b8d5948fa27 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -72,7 +72,9 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags, - extent_submit_bio_hook_t *submit_bio_hook); + extent_submit_bio_hook_t *submit_bio_start, + extent_submit_bio_hook_t *submit_bio_done); + int btrfs_congested_async(struct btrfs_fs_info *info, int iodone); unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info); int btrfs_write_tree_block(struct extent_buffer *buf); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 806708dd7e3..3df0ffad976 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -881,7 +881,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, * At IO completion time the cums attached on the ordered extent record * are inserted into the btree */ -int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, +int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -889,7 +889,21 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, ret = btrfs_csum_one_bio(root, inode, bio); BUG_ON(ret); + return 0; +} +/* + * in order to insert checksums into the metadata in large chunks, + * we wait until bio submission time. All the pages in the bio are + * checksummed and sums are attached onto the ordered extent record. + * + * At IO completion time the cums attached on the ordered extent record + * are inserted into the btree + */ +int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, + int mirror_num, unsigned long bio_flags) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; return btrfs_map_bio(root, rw, bio, mirror_num, 1); } @@ -922,7 +936,8 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, /* we're doing a write, do the async checksumming */ return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, - bio_flags, __btrfs_submit_bio_hook); + bio_flags, __btrfs_submit_bio_start, + __btrfs_submit_bio_done); } mapit: -- cgit v1.2.3-70-g09d2 From 771ed689d2cd53439e28e095bc38fbe40a71429e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 6 Nov 2008 22:02:51 -0500 Subject: Btrfs: Optimize compressed writeback and reads When reading compressed extents, try to put pages into the page cache for any pages covered by the compressed extent that readpages didn't already preload. Add an async work queue to handle transformations at delayed allocation processing time. Right now this is just compression. The workflow is: 1) Find offsets in the file marked for delayed allocation 2) Lock the pages 3) Lock the state bits 4) Call the async delalloc code The async delalloc code clears the state lock bits and delalloc bits. It is important this happens before the range goes into the work queue because otherwise it might deadlock with other work queue items that try to lock those extent bits. The file pages are compressed, and if the compression doesn't work the pages are written back directly. An ordered work queue is used to make sure the inodes are written in the same order that pdflush or writepages sent them down. This changes extent_write_cache_pages to let the writepage function update the wbc nr_written count. Signed-off-by: Chris Mason --- fs/btrfs/compression.c | 150 ++++++++++- fs/btrfs/ctree.h | 4 +- fs/btrfs/disk-io.c | 27 +- fs/btrfs/extent-tree.c | 6 +- fs/btrfs/extent_io.c | 140 +++++++++-- fs/btrfs/extent_io.h | 13 +- fs/btrfs/file.c | 53 ++-- fs/btrfs/inode.c | 643 +++++++++++++++++++++++++++++++++++++----------- fs/btrfs/ordered-data.c | 13 +- fs/btrfs/super.c | 4 + fs/btrfs/zlib.c | 3 +- 11 files changed, 849 insertions(+), 207 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 354913177ba..284f21025bc 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -33,6 +33,7 @@ #include #include #include +#include #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -145,9 +146,9 @@ static void end_compressed_bio_read(struct bio *bio, int err) } /* do io completion on the original bio */ - if (cb->errors) + if (cb->errors) { bio_io_error(cb->orig_bio); - else + } else bio_endio(cb->orig_bio, 0); /* finally free the cb struct */ @@ -333,6 +334,7 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, } bytes_left -= PAGE_CACHE_SIZE; first_byte += PAGE_CACHE_SIZE; + cond_resched(); } bio_get(bio); @@ -346,6 +348,130 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, return 0; } +static noinline int add_ra_bio_pages(struct inode *inode, + u64 compressed_end, + struct compressed_bio *cb) +{ + unsigned long end_index; + unsigned long page_index; + u64 last_offset; + u64 isize = i_size_read(inode); + int ret; + struct page *page; + unsigned long nr_pages = 0; + struct extent_map *em; + struct address_space *mapping = inode->i_mapping; + struct pagevec pvec; + struct extent_map_tree *em_tree; + struct extent_io_tree *tree; + u64 end; + int misses = 0; + + page = cb->orig_bio->bi_io_vec[cb->orig_bio->bi_vcnt - 1].bv_page; + last_offset = (page_offset(page) + PAGE_CACHE_SIZE); + em_tree = &BTRFS_I(inode)->extent_tree; + tree = &BTRFS_I(inode)->io_tree; + + if (isize == 0) + return 0; + + end_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT; + + pagevec_init(&pvec, 0); + while(last_offset < compressed_end) { + page_index = last_offset >> PAGE_CACHE_SHIFT; + + if (page_index > end_index) + break; + + rcu_read_lock(); + page = radix_tree_lookup(&mapping->page_tree, page_index); + rcu_read_unlock(); + if (page) { + misses++; + if (misses > 4) + break; + goto next; + } + + page = alloc_page(mapping_gfp_mask(mapping) | GFP_NOFS); + if (!page) + break; + + page->index = page_index; + /* + * what we want to do here is call add_to_page_cache_lru, + * but that isn't exported, so we reproduce it here + */ + if (add_to_page_cache(page, mapping, + page->index, GFP_NOFS)) { + page_cache_release(page); + goto next; + } + + /* open coding of lru_cache_add, also not exported */ + page_cache_get(page); + if (!pagevec_add(&pvec, page)) + __pagevec_lru_add(&pvec); + + end = last_offset + PAGE_CACHE_SIZE - 1; + /* + * at this point, we have a locked page in the page cache + * for these bytes in the file. But, we have to make + * sure they map to this compressed extent on disk. + */ + set_page_extent_mapped(page); + lock_extent(tree, last_offset, end, GFP_NOFS); + spin_lock(&em_tree->lock); + em = lookup_extent_mapping(em_tree, last_offset, + PAGE_CACHE_SIZE); + spin_unlock(&em_tree->lock); + + if (!em || last_offset < em->start || + (last_offset + PAGE_CACHE_SIZE > extent_map_end(em)) || + (em->block_start >> 9) != cb->orig_bio->bi_sector) { + free_extent_map(em); + unlock_extent(tree, last_offset, end, GFP_NOFS); + unlock_page(page); + page_cache_release(page); + break; + } + free_extent_map(em); + + if (page->index == end_index) { + char *userpage; + size_t zero_offset = isize & (PAGE_CACHE_SIZE - 1); + + if (zero_offset) { + int zeros; + zeros = PAGE_CACHE_SIZE - zero_offset; + userpage = kmap_atomic(page, KM_USER0); + memset(userpage + zero_offset, 0, zeros); + flush_dcache_page(page); + kunmap_atomic(userpage, KM_USER0); + } + } + + ret = bio_add_page(cb->orig_bio, page, + PAGE_CACHE_SIZE, 0); + + if (ret == PAGE_CACHE_SIZE) { + nr_pages++; + page_cache_release(page); + } else { + unlock_extent(tree, last_offset, end, GFP_NOFS); + unlock_page(page); + page_cache_release(page); + break; + } +next: + last_offset += PAGE_CACHE_SIZE; + } + if (pagevec_count(&pvec)) + __pagevec_lru_add(&pvec); + return 0; +} + /* * for a compressed read, the bio we get passed has all the inode pages * in it. We don't actually do IO on those pages but allocate new ones @@ -373,6 +499,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, struct block_device *bdev; struct bio *comp_bio; u64 cur_disk_byte = (u64)bio->bi_sector << 9; + u64 em_len; struct extent_map *em; int ret; @@ -393,6 +520,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, cb->start = em->start; compressed_len = em->block_len; + em_len = em->len; free_extent_map(em); cb->len = uncompressed_len; @@ -411,6 +539,17 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, } cb->nr_pages = nr_pages; + add_ra_bio_pages(inode, cb->start + em_len, cb); + + if (!btrfs_test_opt(root, NODATASUM) && + !btrfs_test_flag(inode, NODATASUM)) { + btrfs_lookup_bio_sums(root, inode, cb->orig_bio); + } + + /* include any pages we added in add_ra-bio_pages */ + uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE; + cb->len = uncompressed_len; + comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, GFP_NOFS); comp_bio->bi_private = cb; comp_bio->bi_end_io = end_compressed_bio_read; @@ -442,9 +581,10 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, GFP_NOFS); atomic_inc(&cb->pending_bios); - bio->bi_private = cb; - bio->bi_end_io = end_compressed_bio_write; - bio_add_page(bio, page, PAGE_CACHE_SIZE, 0); + comp_bio->bi_private = cb; + comp_bio->bi_end_io = end_compressed_bio_read; + + bio_add_page(comp_bio, page, PAGE_CACHE_SIZE, 0); } cur_disk_byte += PAGE_CACHE_SIZE; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 689df070c8e..c83cc5b2ded 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -625,8 +625,8 @@ struct btrfs_fs_info { struct btrfs_transaction *running_transaction; wait_queue_head_t transaction_throttle; wait_queue_head_t transaction_wait; - wait_queue_head_t async_submit_wait; + wait_queue_head_t async_submit_wait; wait_queue_head_t tree_log_wait; struct btrfs_super_block super_copy; @@ -653,6 +653,7 @@ struct btrfs_fs_info { atomic_t nr_async_submits; atomic_t async_submit_draining; atomic_t nr_async_bios; + atomic_t async_delalloc_pages; atomic_t tree_log_writers; atomic_t tree_log_commit; unsigned long tree_log_batch; @@ -677,6 +678,7 @@ struct btrfs_fs_info { * two */ struct btrfs_workers workers; + struct btrfs_workers delalloc_workers; struct btrfs_workers endio_workers; struct btrfs_workers endio_write_workers; struct btrfs_workers submit_workers; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e0a28f705a6..8efc123d222 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -539,6 +539,13 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, (atomic_read(&fs_info->nr_async_bios) < limit), HZ/10); } + + while(atomic_read(&fs_info->async_submit_draining) && + atomic_read(&fs_info->nr_async_submits)) { + wait_event(fs_info->async_submit_wait, + (atomic_read(&fs_info->nr_async_submits) == 0)); + } + return 0; } @@ -1437,6 +1444,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->space_info); btrfs_mapping_init(&fs_info->mapping_tree); atomic_set(&fs_info->nr_async_submits, 0); + atomic_set(&fs_info->async_delalloc_pages, 0); atomic_set(&fs_info->async_submit_draining, 0); atomic_set(&fs_info->nr_async_bios, 0); atomic_set(&fs_info->throttles, 0); @@ -1550,6 +1558,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_init_workers(&fs_info->workers, "worker", fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", + fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->submit_workers, "submit", min_t(u64, fs_devices->num_devices, fs_info->thread_pool_size)); @@ -1560,15 +1571,12 @@ struct btrfs_root *open_ctree(struct super_block *sb, */ fs_info->submit_workers.idle_thresh = 64; - /* fs_info->workers is responsible for checksumming file data - * blocks and metadata. Using a larger idle thresh allows each - * worker thread to operate on things in roughly the order they - * were sent by the writeback daemons, improving overall locality - * of the IO going down the pipe. - */ - fs_info->workers.idle_thresh = 8; + fs_info->workers.idle_thresh = 16; fs_info->workers.ordered = 1; + fs_info->delalloc_workers.idle_thresh = 2; + fs_info->delalloc_workers.ordered = 1; + btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); btrfs_init_workers(&fs_info->endio_workers, "endio", fs_info->thread_pool_size); @@ -1584,6 +1592,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_start_workers(&fs_info->workers, 1); btrfs_start_workers(&fs_info->submit_workers, 1); + btrfs_start_workers(&fs_info->delalloc_workers, 1); btrfs_start_workers(&fs_info->fixup_workers, 1); btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); btrfs_start_workers(&fs_info->endio_write_workers, @@ -1732,6 +1741,7 @@ fail_tree_root: fail_sys_array: fail_sb_buffer: btrfs_stop_workers(&fs_info->fixup_workers); + btrfs_stop_workers(&fs_info->delalloc_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_write_workers); @@ -1988,6 +1998,7 @@ int close_ctree(struct btrfs_root *root) truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); btrfs_stop_workers(&fs_info->fixup_workers); + btrfs_stop_workers(&fs_info->delalloc_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_write_workers); @@ -2062,7 +2073,7 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) struct extent_io_tree *tree; u64 num_dirty; u64 start = 0; - unsigned long thresh = 96 * 1024 * 1024; + unsigned long thresh = 32 * 1024 * 1024; tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; if (current_is_pdflush() || current->flags & PF_MEMALLOC) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 8af39521eb7..ebd8275a193 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -768,7 +768,11 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, l = path->nodes[0]; btrfs_item_key_to_cpu(l, &key, path->slots[0]); - BUG_ON(key.objectid != bytenr); + if (key.objectid != bytenr) { + btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]); + printk("wanted %Lu found %Lu\n", bytenr, key.objectid); + BUG(); + } BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY); item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 9b37ce6e516..bbe3bcfcf4a 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -47,6 +47,11 @@ struct extent_page_data { struct bio *bio; struct extent_io_tree *tree; get_extent_t *get_extent; + + /* tells writepage not to lock the state bits for this range + * it still does the unlocking + */ + int extent_locked; }; int __init extent_io_init(void) @@ -1198,11 +1203,18 @@ static noinline int lock_delalloc_pages(struct inode *inode, * the caller is taking responsibility for * locked_page */ - if (pages[i] != locked_page) + if (pages[i] != locked_page) { lock_page(pages[i]); + if (pages[i]->mapping != inode->i_mapping) { + ret = -EAGAIN; + unlock_page(pages[i]); + page_cache_release(pages[i]); + goto done; + } + } page_cache_release(pages[i]); + pages_locked++; } - pages_locked += ret; nrpages -= ret; index += ret; cond_resched(); @@ -1262,8 +1274,7 @@ again: * if we're looping. */ if (delalloc_end + 1 - delalloc_start > max_bytes && loops) { - delalloc_end = (delalloc_start + PAGE_CACHE_SIZE - 1) & - ~((u64)PAGE_CACHE_SIZE - 1); + delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1; } /* step two, lock all the pages after the page that has start */ ret = lock_delalloc_pages(inode, locked_page, @@ -1306,7 +1317,10 @@ out_failed: int extent_clear_unlock_delalloc(struct inode *inode, struct extent_io_tree *tree, u64 start, u64 end, struct page *locked_page, - int clear_dirty, int set_writeback, + int unlock_pages, + int clear_unlock, + int clear_delalloc, int clear_dirty, + int set_writeback, int end_writeback) { int ret; @@ -1315,12 +1329,19 @@ int extent_clear_unlock_delalloc(struct inode *inode, unsigned long end_index = end >> PAGE_CACHE_SHIFT; unsigned long nr_pages = end_index - index + 1; int i; - int clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC; + int clear_bits = 0; + if (clear_unlock) + clear_bits |= EXTENT_LOCKED; if (clear_dirty) clear_bits |= EXTENT_DIRTY; + if (clear_delalloc) + clear_bits |= EXTENT_DELALLOC; + clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS); + if (!(unlock_pages || clear_dirty || set_writeback || end_writeback)) + return 0; while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, @@ -1336,7 +1357,8 @@ int extent_clear_unlock_delalloc(struct inode *inode, set_page_writeback(pages[i]); if (end_writeback) end_page_writeback(pages[i]); - unlock_page(pages[i]); + if (unlock_pages) + unlock_page(pages[i]); page_cache_release(pages[i]); } nr_pages -= ret; @@ -1741,9 +1763,10 @@ static void end_bio_extent_readpage(struct bio *bio, int err) } } - if (uptodate) + if (uptodate) { set_extent_uptodate(tree, start, end, GFP_ATOMIC); + } unlock_extent(tree, start, end, GFP_ATOMIC); if (whole_page) { @@ -1925,6 +1948,7 @@ void set_page_extent_mapped(struct page *page) set_page_private(page, EXTENT_PAGE_PRIVATE); } } +EXPORT_SYMBOL(set_page_extent_mapped); void set_page_extent_head(struct page *page, unsigned long len) { @@ -2143,12 +2167,17 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, u64 delalloc_end; int page_started; int compressed; + unsigned long nr_written = 0; WARN_ON(!PageLocked(page)); pg_offset = i_size & (PAGE_CACHE_SIZE - 1); if (page->index > end_index || (page->index == end_index && !pg_offset)) { - page->mapping->a_ops->invalidatepage(page, 0); + if (epd->extent_locked) { + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, + page_end, NULL, 1); + } unlock_page(page); return 0; } @@ -2169,27 +2198,33 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_start = start; delalloc_end = 0; page_started = 0; - while(delalloc_end < page_end) { - nr_delalloc = find_lock_delalloc_range(inode, tree, + if (!epd->extent_locked) { + while(delalloc_end < page_end) { + nr_delalloc = find_lock_delalloc_range(inode, tree, page, &delalloc_start, &delalloc_end, 128 * 1024 * 1024); - if (nr_delalloc == 0) { + if (nr_delalloc == 0) { + delalloc_start = delalloc_end + 1; + continue; + } + tree->ops->fill_delalloc(inode, page, delalloc_start, + delalloc_end, &page_started, + &nr_written); delalloc_start = delalloc_end + 1; - continue; } - tree->ops->fill_delalloc(inode, page, delalloc_start, - delalloc_end, &page_started); - delalloc_start = delalloc_end + 1; - } - /* did the fill delalloc function already unlock and start the IO? */ - if (page_started) { - return 0; + /* did the fill delalloc function already unlock and start + * the IO? + */ + if (page_started) { + ret = 0; + goto update_nr_written; + } } - lock_extent(tree, start, page_end, GFP_NOFS); + unlock_start = start; if (tree->ops && tree->ops->writepage_start_hook) { @@ -2199,10 +2234,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, unlock_extent(tree, start, page_end, GFP_NOFS); redirty_page_for_writepage(wbc, page); unlock_page(page); - return 0; + ret = 0; + goto update_nr_written; } } + nr_written++; + end = page_end; if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { printk("found delalloc bits after lock_extent\n"); @@ -2333,6 +2371,12 @@ done: if (unlock_start <= page_end) unlock_extent(tree, unlock_start, page_end, GFP_NOFS); unlock_page(page); + +update_nr_written: + wbc->nr_to_write -= nr_written; + if (wbc->range_cyclic || (wbc->nr_to_write > 0 && + wbc->range_start == 0 && wbc->range_end == LLONG_MAX)) + page->mapping->writeback_index = page->index + nr_written; return 0; } @@ -2431,7 +2475,7 @@ retry: unlock_page(page); ret = 0; } - if (ret || (--(wbc->nr_to_write) <= 0)) + if (ret || wbc->nr_to_write <= 0) done = 1; if (wbc->nonblocking && bdi_write_congested(bdi)) { wbc->encountered_congestion = 1; @@ -2452,6 +2496,8 @@ retry: } if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) mapping->writeback_index = index; + if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) + range_whole = 1; if (wbc->range_cont) wbc->range_start = index << PAGE_CACHE_SHIFT; @@ -2469,6 +2515,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, }; struct writeback_control wbc_writepages = { .bdi = wbc->bdi, @@ -2491,6 +2538,52 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, } EXPORT_SYMBOL(extent_write_full_page); +int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, + u64 start, u64 end, get_extent_t *get_extent, + int mode) +{ + int ret = 0; + struct address_space *mapping = inode->i_mapping; + struct page *page; + unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >> + PAGE_CACHE_SHIFT; + + struct extent_page_data epd = { + .bio = NULL, + .tree = tree, + .get_extent = get_extent, + .extent_locked = 1, + }; + struct writeback_control wbc_writepages = { + .bdi = inode->i_mapping->backing_dev_info, + .sync_mode = mode, + .older_than_this = NULL, + .nr_to_write = nr_pages * 2, + .range_start = start, + .range_end = end + 1, + }; + + while(start <= end) { + page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT); + if (clear_page_dirty_for_io(page)) + ret = __extent_writepage(page, &wbc_writepages, &epd); + else { + if (tree->ops && tree->ops->writepage_end_io_hook) + tree->ops->writepage_end_io_hook(page, start, + start + PAGE_CACHE_SIZE - 1, + NULL, 1); + unlock_page(page); + } + page_cache_release(page); + start += PAGE_CACHE_SIZE; + } + + if (epd.bio) + submit_one_bio(WRITE, epd.bio, 0, 0); + return ret; +} +EXPORT_SYMBOL(extent_write_locked_range); + int extent_writepages(struct extent_io_tree *tree, struct address_space *mapping, @@ -2502,6 +2595,7 @@ int extent_writepages(struct extent_io_tree *tree, .bio = NULL, .tree = tree, .get_extent = get_extent, + .extent_locked = 0, }; ret = extent_write_cache_pages(tree, mapping, wbc, diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 283110ec4ee..2d5f67065b6 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -35,7 +35,8 @@ typedef int (extent_submit_bio_hook_t)(struct inode *inode, int rw, unsigned long bio_flags); struct extent_io_ops { int (*fill_delalloc)(struct inode *inode, struct page *locked_page, - u64 start, u64 end, int *page_started); + u64 start, u64 end, int *page_started, + unsigned long *nr_written); int (*writepage_start_hook)(struct page *page, u64 start, u64 end); int (*writepage_io_hook)(struct page *page, u64 start, u64 end); extent_submit_bio_hook_t *submit_bio_hook; @@ -172,6 +173,9 @@ int extent_invalidatepage(struct extent_io_tree *tree, int extent_write_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, struct writeback_control *wbc); +int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, + u64 start, u64 end, get_extent_t *get_extent, + int mode); int extent_writepages(struct extent_io_tree *tree, struct address_space *mapping, get_extent_t *get_extent, @@ -256,6 +260,9 @@ int extent_range_uptodate(struct extent_io_tree *tree, int extent_clear_unlock_delalloc(struct inode *inode, struct extent_io_tree *tree, u64 start, u64 end, struct page *locked_page, - int clear_dirty, int set_writeback, - int clear_writeback); + int unlock_page, + int clear_unlock, + int clear_delalloc, int clear_dirty, + int set_writeback, + int end_writeback); #endif diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 0c8cc35a8b9..337221ecca2 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -368,6 +368,8 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, u64 search_start = start; u64 leaf_start; u64 ram_bytes = 0; + u64 orig_parent = 0; + u64 disk_bytenr = 0; u8 compression; u8 encryption; u16 other_encoding = 0; @@ -500,17 +502,31 @@ next_slot: keep = 1; } - if (bookend && found_extent && locked_end < extent_end) { - ret = try_lock_extent(&BTRFS_I(inode)->io_tree, - locked_end, extent_end - 1, GFP_NOFS); - if (!ret) { - btrfs_release_path(root, path); - lock_extent(&BTRFS_I(inode)->io_tree, - locked_end, extent_end - 1, GFP_NOFS); + if (bookend && found_extent) { + if (locked_end < extent_end) { + ret = try_lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, + GFP_NOFS); + if (!ret) { + btrfs_release_path(root, path); + lock_extent(&BTRFS_I(inode)->io_tree, + locked_end, extent_end - 1, + GFP_NOFS); + locked_end = extent_end; + continue; + } locked_end = extent_end; - continue; } - locked_end = extent_end; + orig_parent = path->nodes[0]->start; + disk_bytenr = le64_to_cpu(old.disk_bytenr); + if (disk_bytenr != 0) { + ret = btrfs_inc_extent_ref(trans, root, + disk_bytenr, + le64_to_cpu(old.disk_num_bytes), + orig_parent, root->root_key.objectid, + trans->transid, inode->i_ino); + BUG_ON(ret); + } } if (found_inline) { @@ -537,8 +553,12 @@ next_slot: inode_sub_bytes(inode, old_num - new_num); } - btrfs_set_file_extent_num_bytes(leaf, extent, - new_num); + if (!compression && !encryption) { + btrfs_set_file_extent_ram_bytes(leaf, + extent, new_num); + } + btrfs_set_file_extent_num_bytes(leaf, + extent, new_num); btrfs_mark_buffer_dirty(leaf); } else if (key.offset < inline_limit && (end > extent_end) && @@ -582,11 +602,11 @@ next_slot: } /* create bookend, splitting the extent in two */ if (bookend && found_extent) { - u64 disk_bytenr; struct btrfs_key ins; ins.objectid = inode->i_ino; ins.offset = end; btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY); + btrfs_release_path(root, path); ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*extent)); @@ -623,14 +643,13 @@ next_slot: btrfs_mark_buffer_dirty(path->nodes[0]); - disk_bytenr = le64_to_cpu(old.disk_bytenr); if (disk_bytenr != 0) { - ret = btrfs_inc_extent_ref(trans, root, - disk_bytenr, - le64_to_cpu(old.disk_num_bytes), - leaf->start, + ret = btrfs_update_extent_ref(trans, root, + disk_bytenr, orig_parent, + leaf->start, root->root_key.objectid, trans->transid, ins.objectid); + BUG_ON(ret); } btrfs_release_path(root, path); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3df0ffad976..e01c0d0310a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -86,6 +86,10 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { static void btrfs_truncate(struct inode *inode); static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end); +static noinline int cow_file_range(struct inode *inode, + struct page *locked_page, + u64 start, u64 end, int *page_started, + unsigned long *nr_written, int unlock); /* * a very lame attempt at stopping writes when the FS is 85% full. There @@ -262,35 +266,72 @@ static int cow_file_range_inline(struct btrfs_trans_handle *trans, return 0; } +struct async_extent { + u64 start; + u64 ram_size; + u64 compressed_size; + struct page **pages; + unsigned long nr_pages; + struct list_head list; +}; + +struct async_cow { + struct inode *inode; + struct btrfs_root *root; + struct page *locked_page; + u64 start; + u64 end; + struct list_head extents; + struct btrfs_work work; +}; + +static noinline int add_async_extent(struct async_cow *cow, + u64 start, u64 ram_size, + u64 compressed_size, + struct page **pages, + unsigned long nr_pages) +{ + struct async_extent *async_extent; + + async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS); + async_extent->start = start; + async_extent->ram_size = ram_size; + async_extent->compressed_size = compressed_size; + async_extent->pages = pages; + async_extent->nr_pages = nr_pages; + list_add_tail(&async_extent->list, &cow->extents); + return 0; +} + /* - * when extent_io.c finds a delayed allocation range in the file, - * the call backs end up in this code. The basic idea is to - * allocate extents on disk for the range, and create ordered data structs - * in ram to track those extents. + * we create compressed extents in two phases. The first + * phase compresses a range of pages that have already been + * locked (both pages and state bits are locked). * - * locked_page is the page that writepage had locked already. We use - * it to make sure we don't do extra locks or unlocks. + * This is done inside an ordered work queue, and the compression + * is spread across many cpus. The actual IO submission is step + * two, and the ordered work queue takes care of making sure that + * happens in the same order things were put onto the queue by + * writepages and friends. * - * *page_started is set to one if we unlock locked_page and do everything - * required to start IO on it. It may be clean and already done with - * IO when we return. + * If this code finds it can't get good compression, it puts an + * entry onto the work queue to write the uncompressed bytes. This + * makes sure that both compressed inodes and uncompressed inodes + * are written in the same order that pdflush sent them down. */ -static int cow_file_range(struct inode *inode, struct page *locked_page, - u64 start, u64 end, int *page_started) +static noinline int compress_file_range(struct inode *inode, + struct page *locked_page, + u64 start, u64 end, + struct async_cow *async_cow, + int *num_added) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; - u64 alloc_hint = 0; u64 num_bytes; - unsigned long ram_size; u64 orig_start; u64 disk_num_bytes; - u64 cur_alloc_size; u64 blocksize = root->sectorsize; u64 actual_end; - struct btrfs_key ins; - struct extent_map *em; - struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; int ret = 0; struct page **pages = NULL; unsigned long nr_pages; @@ -298,22 +339,12 @@ static int cow_file_range(struct inode *inode, struct page *locked_page, unsigned long total_compressed = 0; unsigned long total_in = 0; unsigned long max_compressed = 128 * 1024; - unsigned long max_uncompressed = 256 * 1024; + unsigned long max_uncompressed = 128 * 1024; int i; - int ordered_type; int will_compress; - trans = btrfs_join_transaction(root, 1); - BUG_ON(!trans); - btrfs_set_trans_block_group(trans, inode); orig_start = start; - /* - * compression made this loop a bit ugly, but the basic idea is to - * compress some pages but keep the total size of the compressed - * extent relatively small. If compression is off, this goto target - * is never used. - */ again: will_compress = 0; nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1; @@ -324,7 +355,13 @@ again: /* we want to make sure that amount of ram required to uncompress * an extent is reasonable, so we limit the total size in ram - * of a compressed extent to 256k + * of a compressed extent to 128k. This is a crucial number + * because it also controls how easily we can spread reads across + * cpus for decompression. + * + * We also want to make sure the amount of IO required to do + * a random read is reasonably small, so we limit the size of + * a compressed extent to 128k. */ total_compressed = min(total_compressed, max_uncompressed); num_bytes = (end - start + blocksize) & ~(blocksize - 1); @@ -333,18 +370,16 @@ again: total_in = 0; ret = 0; - /* we do compression for mount -o compress and when the - * inode has not been flagged as nocompress + /* + * we do compression for mount -o compress and when the + * inode has not been flagged as nocompress. This flag can + * change at any time if we discover bad compression ratios. */ if (!btrfs_test_flag(inode, NOCOMPRESS) && btrfs_test_opt(root, COMPRESS)) { WARN_ON(pages); pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS); - /* we want to make sure the amount of IO required to satisfy - * a random read is reasonably small, so we limit the size - * of a compressed extent to 128k - */ ret = btrfs_zlib_compress_pages(inode->i_mapping, start, total_compressed, pages, nr_pages, &nr_pages_ret, @@ -371,26 +406,34 @@ again: } } if (start == 0) { + trans = btrfs_join_transaction(root, 1); + BUG_ON(!trans); + btrfs_set_trans_block_group(trans, inode); + /* lets try to make an inline extent */ - if (ret || total_in < (end - start + 1)) { + if (ret || total_in < (actual_end - start)) { /* we didn't compress the entire range, try - * to make an uncompressed inline extent. This - * is almost sure to fail, but maybe inline sizes - * will get bigger later + * to make an uncompressed inline extent. */ ret = cow_file_range_inline(trans, root, inode, start, end, 0, NULL); } else { + /* try making a compressed inline extent */ ret = cow_file_range_inline(trans, root, inode, start, end, total_compressed, pages); } + btrfs_end_transaction(trans, root); if (ret == 0) { + /* + * inline extent creation worked, we don't need + * to create any more async work items. Unlock + * and free up our temp pages. + */ extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, - start, end, NULL, - 1, 1, 1); - *page_started = 1; + start, end, NULL, 1, 0, + 0, 1, 1, 1); ret = 0; goto free_pages_out; } @@ -435,53 +478,280 @@ again: /* flag the file so we don't compress in the future */ btrfs_set_flag(inode, NOCOMPRESS); } + if (will_compress) { + *num_added += 1; - BUG_ON(disk_num_bytes > - btrfs_super_total_bytes(&root->fs_info->super_copy)); + /* the async work queues will take care of doing actual + * allocation on disk for these compressed pages, + * and will submit them to the elevator. + */ + add_async_extent(async_cow, start, num_bytes, + total_compressed, pages, nr_pages_ret); - btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); + if (start + num_bytes < end) { + start += num_bytes; + pages = NULL; + cond_resched(); + goto again; + } + } else { + /* + * No compression, but we still need to write the pages in + * the file we've been given so far. redirty the locked + * page if it corresponds to our extent and set things up + * for the async work queue to run cow_file_range to do + * the normal delalloc dance + */ + if (page_offset(locked_page) >= start && + page_offset(locked_page) <= end) { + __set_page_dirty_nobuffers(locked_page); + /* unlocked later on in the async handlers */ + } + add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0); + *num_added += 1; + } - while(disk_num_bytes > 0) { - unsigned long min_bytes; +out: + return 0; + +free_pages_out: + for (i = 0; i < nr_pages_ret; i++) { + WARN_ON(pages[i]->mapping); + page_cache_release(pages[i]); + } + if (pages) + kfree(pages); + + goto out; +} + +/* + * phase two of compressed writeback. This is the ordered portion + * of the code, which only gets called in the order the work was + * queued. We walk all the async extents created by compress_file_range + * and send them down to the disk. + */ +static noinline int submit_compressed_extents(struct inode *inode, + struct async_cow *async_cow) +{ + struct async_extent *async_extent; + u64 alloc_hint = 0; + struct btrfs_trans_handle *trans; + struct btrfs_key ins; + struct extent_map *em; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + struct extent_io_tree *io_tree; + int ret; + + if (list_empty(&async_cow->extents)) + return 0; + + trans = btrfs_join_transaction(root, 1); + + while(!list_empty(&async_cow->extents)) { + async_extent = list_entry(async_cow->extents.next, + struct async_extent, list); + list_del(&async_extent->list); + io_tree = &BTRFS_I(inode)->io_tree; + + /* did the compression code fall back to uncompressed IO? */ + if (!async_extent->pages) { + int page_started = 0; + unsigned long nr_written = 0; + + lock_extent(io_tree, async_extent->start, + async_extent->start + async_extent->ram_size - 1, + GFP_NOFS); + + /* allocate blocks */ + cow_file_range(inode, async_cow->locked_page, + async_extent->start, + async_extent->start + + async_extent->ram_size - 1, + &page_started, &nr_written, 0); + + /* + * if page_started, cow_file_range inserted an + * inline extent and took care of all the unlocking + * and IO for us. Otherwise, we need to submit + * all those pages down to the drive. + */ + if (!page_started) + extent_write_locked_range(io_tree, + inode, async_extent->start, + async_extent->start + + async_extent->ram_size - 1, + btrfs_get_extent, + WB_SYNC_ALL); + kfree(async_extent); + cond_resched(); + continue; + } + + lock_extent(io_tree, async_extent->start, + async_extent->start + async_extent->ram_size - 1, + GFP_NOFS); /* - * the max size of a compressed extent is pretty small, - * make the code a little less complex by forcing - * the allocator to find a whole compressed extent at once + * here we're doing allocation and writeback of the + * compressed pages */ - if (will_compress) - min_bytes = disk_num_bytes; - else - min_bytes = root->sectorsize; + btrfs_drop_extent_cache(inode, async_extent->start, + async_extent->start + + async_extent->ram_size - 1, 0); + + ret = btrfs_reserve_extent(trans, root, + async_extent->compressed_size, + async_extent->compressed_size, + 0, alloc_hint, + (u64)-1, &ins, 1); + BUG_ON(ret); + em = alloc_extent_map(GFP_NOFS); + em->start = async_extent->start; + em->len = async_extent->ram_size; + + em->block_start = ins.objectid; + em->block_len = ins.offset; + em->bdev = root->fs_info->fs_devices->latest_bdev; + set_bit(EXTENT_FLAG_PINNED, &em->flags); + set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); + + while(1) { + spin_lock(&em_tree->lock); + ret = add_extent_mapping(em_tree, em); + spin_unlock(&em_tree->lock); + if (ret != -EEXIST) { + free_extent_map(em); + break; + } + btrfs_drop_extent_cache(inode, async_extent->start, + async_extent->start + + async_extent->ram_size - 1, 0); + } + + ret = btrfs_add_ordered_extent(inode, async_extent->start, + ins.objectid, + async_extent->ram_size, + ins.offset, + BTRFS_ORDERED_COMPRESSED); + BUG_ON(ret); + + btrfs_end_transaction(trans, root); + + /* + * clear dirty, set writeback and unlock the pages. + */ + extent_clear_unlock_delalloc(inode, + &BTRFS_I(inode)->io_tree, + async_extent->start, + async_extent->start + + async_extent->ram_size - 1, + NULL, 1, 1, 0, 1, 1, 0); + + ret = btrfs_submit_compressed_write(inode, + async_extent->start, + async_extent->ram_size, + ins.objectid, + ins.offset, async_extent->pages, + async_extent->nr_pages); + + BUG_ON(ret); + trans = btrfs_join_transaction(root, 1); + alloc_hint = ins.objectid + ins.offset; + kfree(async_extent); + cond_resched(); + } + + btrfs_end_transaction(trans, root); + return 0; +} + +/* + * when extent_io.c finds a delayed allocation range in the file, + * the call backs end up in this code. The basic idea is to + * allocate extents on disk for the range, and create ordered data structs + * in ram to track those extents. + * + * locked_page is the page that writepage had locked already. We use + * it to make sure we don't do extra locks or unlocks. + * + * *page_started is set to one if we unlock locked_page and do everything + * required to start IO on it. It may be clean and already done with + * IO when we return. + */ +static noinline int cow_file_range(struct inode *inode, + struct page *locked_page, + u64 start, u64 end, int *page_started, + unsigned long *nr_written, + int unlock) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_trans_handle *trans; + u64 alloc_hint = 0; + u64 num_bytes; + unsigned long ram_size; + u64 disk_num_bytes; + u64 cur_alloc_size; + u64 blocksize = root->sectorsize; + u64 actual_end; + struct btrfs_key ins; + struct extent_map *em; + struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; + int ret = 0; + + trans = btrfs_join_transaction(root, 1); + BUG_ON(!trans); + btrfs_set_trans_block_group(trans, inode); + actual_end = min_t(u64, i_size_read(inode), end + 1); + + num_bytes = (end - start + blocksize) & ~(blocksize - 1); + num_bytes = max(blocksize, num_bytes); + disk_num_bytes = num_bytes; + ret = 0; + + if (start == 0) { + /* lets try to make an inline extent */ + ret = cow_file_range_inline(trans, root, inode, + start, end, 0, NULL); + if (ret == 0) { + extent_clear_unlock_delalloc(inode, + &BTRFS_I(inode)->io_tree, + start, end, NULL, 1, 1, + 1, 1, 1, 1); + *nr_written = *nr_written + + (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE; + *page_started = 1; + ret = 0; + goto out; + } + } + + BUG_ON(disk_num_bytes > + btrfs_super_total_bytes(&root->fs_info->super_copy)); + + btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); + + while(disk_num_bytes > 0) { cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent); ret = btrfs_reserve_extent(trans, root, cur_alloc_size, - min_bytes, 0, alloc_hint, + root->sectorsize, 0, alloc_hint, (u64)-1, &ins, 1); if (ret) { - WARN_ON(1); - goto free_pages_out_fail; + BUG(); } em = alloc_extent_map(GFP_NOFS); em->start = start; - if (will_compress) { - ram_size = num_bytes; - em->len = num_bytes; - } else { - /* ramsize == disk size */ - ram_size = ins.offset; - em->len = ins.offset; - } + ram_size = ins.offset; + em->len = ins.offset; em->block_start = ins.objectid; em->block_len = ins.offset; em->bdev = root->fs_info->fs_devices->latest_bdev; set_bit(EXTENT_FLAG_PINNED, &em->flags); - if (will_compress) - set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); - while(1) { spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); @@ -495,10 +765,8 @@ again: } cur_alloc_size = ins.offset; - ordered_type = will_compress ? BTRFS_ORDERED_COMPRESSED : 0; ret = btrfs_add_ordered_extent(inode, start, ins.objectid, - ram_size, cur_alloc_size, - ordered_type); + ram_size, cur_alloc_size, 0); BUG_ON(ret); if (disk_num_bytes < cur_alloc_size) { @@ -506,82 +774,145 @@ again: cur_alloc_size); break; } - - if (will_compress) { - /* - * we're doing compression, we and we need to - * submit the compressed extents down to the device. - * - * We lock down all the file pages, clearing their - * dirty bits and setting them writeback. Everyone - * that wants to modify the page will wait on the - * ordered extent above. - * - * The writeback bits on the file pages are - * cleared when the compressed pages are on disk - */ - btrfs_end_transaction(trans, root); - - if (start <= page_offset(locked_page) && - page_offset(locked_page) < start + ram_size) { - *page_started = 1; - } - - extent_clear_unlock_delalloc(inode, - &BTRFS_I(inode)->io_tree, - start, - start + ram_size - 1, - NULL, 1, 1, 0); - - ret = btrfs_submit_compressed_write(inode, start, - ram_size, ins.objectid, - cur_alloc_size, pages, - nr_pages_ret); - - BUG_ON(ret); - trans = btrfs_join_transaction(root, 1); - if (start + ram_size < end) { - start += ram_size; - alloc_hint = ins.objectid + ins.offset; - /* pages will be freed at end_bio time */ - pages = NULL; - goto again; - } else { - /* we've written everything, time to go */ - break; - } - } /* we're not doing compressed IO, don't unlock the first * page (which the caller expects to stay locked), don't * clear any dirty bits and don't set any writeback bits */ extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, start, start + ram_size - 1, - locked_page, 0, 0, 0); + locked_page, unlock, 1, + 1, 0, 0, 0); disk_num_bytes -= cur_alloc_size; num_bytes -= cur_alloc_size; alloc_hint = ins.objectid + ins.offset; start += cur_alloc_size; } - - ret = 0; out: + ret = 0; btrfs_end_transaction(trans, root); return ret; +} -free_pages_out_fail: - extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, - start, end, locked_page, 0, 0, 0); -free_pages_out: - for (i = 0; i < nr_pages_ret; i++) { - WARN_ON(pages[i]->mapping); - page_cache_release(pages[i]); +/* + * work queue call back to started compression on a file and pages + */ +static noinline void async_cow_start(struct btrfs_work *work) +{ + struct async_cow *async_cow; + int num_added = 0; + async_cow = container_of(work, struct async_cow, work); + + compress_file_range(async_cow->inode, async_cow->locked_page, + async_cow->start, async_cow->end, async_cow, + &num_added); + if (num_added == 0) + async_cow->inode = NULL; +} + +/* + * work queue call back to submit previously compressed pages + */ +static noinline void async_cow_submit(struct btrfs_work *work) +{ + struct async_cow *async_cow; + struct btrfs_root *root; + unsigned long nr_pages; + + async_cow = container_of(work, struct async_cow, work); + + root = async_cow->root; + nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >> + PAGE_CACHE_SHIFT; + + atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages); + + if (atomic_read(&root->fs_info->async_delalloc_pages) < + 5 * 1042 * 1024 && + waitqueue_active(&root->fs_info->async_submit_wait)) + wake_up(&root->fs_info->async_submit_wait); + + if (async_cow->inode) { + submit_compressed_extents(async_cow->inode, async_cow); } - if (pages) - kfree(pages); +} - goto out; +static noinline void async_cow_free(struct btrfs_work *work) +{ + struct async_cow *async_cow; + async_cow = container_of(work, struct async_cow, work); + kfree(async_cow); +} + +static int cow_file_range_async(struct inode *inode, struct page *locked_page, + u64 start, u64 end, int *page_started, + unsigned long *nr_written) +{ + struct async_cow *async_cow; + struct btrfs_root *root = BTRFS_I(inode)->root; + unsigned long nr_pages; + u64 cur_end; + int limit = 10 * 1024 * 1042; + + if (!btrfs_test_opt(root, COMPRESS)) { + return cow_file_range(inode, locked_page, start, end, + page_started, nr_written, 1); + } + + clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED | + EXTENT_DELALLOC, 1, 0, GFP_NOFS); + while(start < end) { + async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS); + async_cow->inode = inode; + async_cow->root = root; + async_cow->locked_page = locked_page; + async_cow->start = start; + + if (btrfs_test_flag(inode, NOCOMPRESS)) + cur_end = end; + else + cur_end = min(end, start + 512 * 1024 - 1); + + async_cow->end = cur_end; + INIT_LIST_HEAD(&async_cow->extents); + + async_cow->work.func = async_cow_start; + async_cow->work.ordered_func = async_cow_submit; + async_cow->work.ordered_free = async_cow_free; + async_cow->work.flags = 0; + + while(atomic_read(&root->fs_info->async_submit_draining) && + atomic_read(&root->fs_info->async_delalloc_pages)) { + wait_event(root->fs_info->async_submit_wait, + (atomic_read(&root->fs_info->async_delalloc_pages) + == 0)); + } + + nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >> + PAGE_CACHE_SHIFT; + atomic_add(nr_pages, &root->fs_info->async_delalloc_pages); + + btrfs_queue_worker(&root->fs_info->delalloc_workers, + &async_cow->work); + + if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) { + wait_event(root->fs_info->async_submit_wait, + (atomic_read(&root->fs_info->async_delalloc_pages) < + limit)); + } + + while(atomic_read(&root->fs_info->async_submit_draining) && + atomic_read(&root->fs_info->async_delalloc_pages)) { + wait_event(root->fs_info->async_submit_wait, + (atomic_read(&root->fs_info->async_delalloc_pages) == + 0)); + } + + *nr_written += nr_pages; + start = cur_end + 1; + } + *page_started = 1; + return 0; } /* @@ -592,7 +923,8 @@ free_pages_out: * blocks on disk */ static int run_delalloc_nocow(struct inode *inode, struct page *locked_page, - u64 start, u64 end, int *page_started, int force) + u64 start, u64 end, int *page_started, int force, + unsigned long *nr_written) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; @@ -711,7 +1043,8 @@ out_check: btrfs_release_path(root, path); if (cow_start != (u64)-1) { ret = cow_file_range(inode, locked_page, cow_start, - found_key.offset - 1, page_started); + found_key.offset - 1, page_started, + nr_written, 1); BUG_ON(ret); cow_start = (u64)-1; } @@ -748,9 +1081,10 @@ out_check: ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr, num_bytes, num_bytes, type); BUG_ON(ret); + extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, cur_offset, cur_offset + num_bytes - 1, - locked_page, 0, 0, 0); + locked_page, 1, 1, 1, 0, 0, 0); cur_offset = extent_end; if (cur_offset > end) break; @@ -761,7 +1095,7 @@ out_check: cow_start = cur_offset; if (cow_start != (u64)-1) { ret = cow_file_range(inode, locked_page, cow_start, end, - page_started); + page_started, nr_written, 1); BUG_ON(ret); } @@ -775,7 +1109,8 @@ out_check: * extent_io.c call back to do delayed allocation processing */ static int run_delalloc_range(struct inode *inode, struct page *locked_page, - u64 start, u64 end, int *page_started) + u64 start, u64 end, int *page_started, + unsigned long *nr_written) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret; @@ -783,13 +1118,13 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, if (btrfs_test_opt(root, NODATACOW) || btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 0); + page_started, 0, nr_written); else if (btrfs_test_flag(inode, PREALLOC)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 1); + page_started, 1, nr_written); else - ret = cow_file_range(inode, locked_page, start, end, - page_started); + ret = cow_file_range_async(inode, locked_page, start, end, + page_started, nr_written); return ret; } @@ -861,6 +1196,9 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, u64 map_length; int ret; + if (bio_flags & EXTENT_BIO_COMPRESSED) + return 0; + length = bio->bi_size; map_tree = &root->fs_info->mapping_tree; map_length = length; @@ -925,12 +1263,12 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, btrfs_test_flag(inode, NODATASUM); if (!(rw & (1 << BIO_RW))) { - if (!skip_sum) - btrfs_lookup_bio_sums(root, inode, bio); if (bio_flags & EXTENT_BIO_COMPRESSED) return btrfs_submit_compressed_read(inode, bio, mirror_num, bio_flags); + else if (!skip_sum) + btrfs_lookup_bio_sums(root, inode, bio); goto mapit; } else if (!skip_sum) { /* we're doing a write, do the async checksumming */ @@ -966,6 +1304,9 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end) { + if ((end & (PAGE_CACHE_SIZE - 1)) == 0) { + WARN_ON(1); + } return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS); } @@ -2105,6 +2446,7 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, int pending_del_nr = 0; int pending_del_slot = 0; int extent_type = -1; + int encoding; u64 mask = root->sectorsize - 1; if (root->ref_cows) @@ -2144,6 +2486,7 @@ search_again: leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); found_type = btrfs_key_type(&found_key); + encoding = 0; if (found_key.objectid != inode->i_ino) break; @@ -2156,6 +2499,10 @@ search_again: fi = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_file_extent_item); extent_type = btrfs_file_extent_type(leaf, fi); + encoding = btrfs_file_extent_compression(leaf, fi); + encoding |= btrfs_file_extent_encryption(leaf, fi); + encoding |= btrfs_file_extent_other_encoding(leaf, fi); + if (extent_type != BTRFS_FILE_EXTENT_INLINE) { item_end += btrfs_file_extent_num_bytes(leaf, fi); @@ -2200,7 +2547,7 @@ search_again: if (extent_type != BTRFS_FILE_EXTENT_INLINE) { u64 num_dec; extent_start = btrfs_file_extent_disk_bytenr(leaf, fi); - if (!del_item) { + if (!del_item && !encoding) { u64 orig_num_bytes = btrfs_file_extent_num_bytes(leaf, fi); extent_num_bytes = new_size - @@ -2436,7 +2783,14 @@ int btrfs_cont_expand(struct inode *inode, loff_t size) last_byte = min(extent_map_end(em), block_end); last_byte = (last_byte + mask) & ~mask; if (test_bit(EXTENT_FLAG_VACANCY, &em->flags)) { + u64 hint_byte = 0; hole_size = last_byte - cur_offset; + err = btrfs_drop_extents(trans, root, inode, + cur_offset, + cur_offset + hole_size, + cur_offset, &hint_byte); + if (err) + break; err = btrfs_insert_file_extent(trans, root, inode->i_ino, cur_offset, 0, 0, hole_size, 0, hole_size, @@ -3785,6 +4139,7 @@ int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc) { struct extent_io_tree *tree; + tree = &BTRFS_I(mapping->host)->io_tree; return extent_writepages(tree, mapping, btrfs_get_extent, wbc); } @@ -4285,9 +4640,11 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) * ordered extents get created before we return */ atomic_inc(&root->fs_info->async_submit_draining); - while(atomic_read(&root->fs_info->nr_async_submits)) { + while(atomic_read(&root->fs_info->nr_async_submits) || + atomic_read(&root->fs_info->async_delalloc_pages)) { wait_event(root->fs_info->async_submit_wait, - (atomic_read(&root->fs_info->nr_async_submits) == 0)); + (atomic_read(&root->fs_info->nr_async_submits) == 0 && + atomic_read(&root->fs_info->async_delalloc_pages) == 0)); } atomic_dec(&root->fs_info->async_submit_draining); return 0; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 370bb428559..027ad6b3839 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -390,7 +390,7 @@ void btrfs_start_ordered_extent(struct inode *inode, * start IO on any dirty ones so the wait doesn't stall waiting * for pdflush to find them */ - btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_NONE); + btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_ALL); if (wait) { wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); @@ -421,6 +421,12 @@ again: */ btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_NONE); + /* The compression code will leave pages locked but return from + * writepage without setting the page writeback. Starting again + * with WB_SYNC_ALL will end up waiting for the IO to actually start. + */ + btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_ALL); + btrfs_wait_on_page_writeback_range(inode->i_mapping, start >> PAGE_CACHE_SHIFT, orig_end >> PAGE_CACHE_SHIFT); @@ -448,10 +454,7 @@ again: } if (test_range_bit(&BTRFS_I(inode)->io_tree, start, orig_end, EXTENT_ORDERED | EXTENT_DELALLOC, 0)) { - printk("inode %lu still ordered or delalloc after wait " - "%llu %llu\n", inode->i_ino, - (unsigned long long)start, - (unsigned long long)orig_end); + schedule_timeout(1); goto again; } return 0; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 431fdf144b5..ab9d5e89ed1 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -375,6 +375,10 @@ int btrfs_sync_fs(struct super_block *sb, int wait) filemap_flush(root->fs_info->btree_inode->i_mapping); return 0; } + + btrfs_start_delalloc_inodes(root); + btrfs_wait_ordered_extents(root, 0); + btrfs_clean_old_snapshots(root); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c index e99309180a1..ba2527d0873 100644 --- a/fs/btrfs/zlib.c +++ b/fs/btrfs/zlib.c @@ -423,8 +423,9 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, /* we didn't make progress in this inflate * call, we're done */ - if (ret != Z_STREAM_END) + if (ret != Z_STREAM_END) { ret = -1; + } break; } -- cgit v1.2.3-70-g09d2 From ff5b7ee33d82414bf4baf299c21fb703bcc89629 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Mon, 10 Nov 2008 07:34:43 -0500 Subject: Btrfs: Fix csum error for compressed data The decompress code doesn't take the logical offset in extent pointer into account. If the logical offset isn't zero, data will be decompressed into wrong pages. The solution used here is to record the starting offset of the extent in the file separately from the logical start of the extent_map struct. This allows us to avoid problems inserting overlapping extents. Signed-off-by: Yan Zheng --- fs/btrfs/compression.c | 6 ++---- fs/btrfs/extent_map.h | 1 + fs/btrfs/file.c | 2 ++ fs/btrfs/inode.c | 3 +++ 4 files changed, 8 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 7397c622fb6..8e7a78acf81 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -505,7 +505,6 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, struct block_device *bdev; struct bio *comp_bio; u64 cur_disk_byte = (u64)bio->bi_sector << 9; - u64 em_len; struct extent_map *em; int ret; @@ -524,9 +523,8 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, cb->errors = 0; cb->inode = inode; - cb->start = em->start; + cb->start = em->orig_start; compressed_len = em->block_len; - em_len = em->len; free_extent_map(em); cb->len = uncompressed_len; @@ -545,7 +543,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, } cb->nr_pages = nr_pages; - add_ra_bio_pages(inode, cb->start + em_len, cb); + add_ra_bio_pages(inode, em->start + em->len, cb); if (!btrfs_test_opt(root, NODATASUM) && !btrfs_test_flag(inode, NODATASUM)) { diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h index accfedaeb51..fb6eeef06bb 100644 --- a/fs/btrfs/extent_map.h +++ b/fs/btrfs/extent_map.h @@ -20,6 +20,7 @@ struct extent_map { /* all of these are in bytes */ u64 start; u64 len; + u64 orig_start; u64 block_start; u64 block_len; unsigned long flags; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 337221ecca2..85841c53880 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -222,6 +222,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, em->start < start) { split->start = em->start; split->len = start - em->start; + split->orig_start = em->orig_start; split->block_start = em->block_start; if (compressed) @@ -243,6 +244,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, split->start = start + len; split->len = em->start + em->len - (start + len); + split->orig_start = em->orig_start; split->bdev = em->bdev; split->flags = flags; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e01c0d0310a..59660293d29 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3949,6 +3949,8 @@ again: found_type == BTRFS_FILE_EXTENT_PREALLOC) { em->start = extent_start; em->len = extent_end - extent_start; + em->orig_start = extent_start - + btrfs_file_extent_offset(leaf, item); bytenr = btrfs_file_extent_disk_bytenr(leaf, item); if (bytenr == 0) { em->block_start = EXTENT_MAP_HOLE; @@ -3988,6 +3990,7 @@ again: em->start = extent_start + extent_offset; em->len = (copy_size + root->sectorsize - 1) & ~((u64)root->sectorsize - 1); + em->orig_start = EXTENT_MAP_INLINE; if (compressed) set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); ptr = btrfs_file_extent_inline_start(item) + extent_offset; -- cgit v1.2.3-70-g09d2 From 445a69449994a37615cd47e47bcab2e42a070adf Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 10 Nov 2008 11:53:33 -0500 Subject: Btrfs: Fix usage of struct extent_map->orig_start This makes sure the orig_start field in struct extent_map gets set everywhere the extent_map structs are created or modified. Signed-off-by: Chris Mason --- fs/btrfs/file.c | 3 ++- fs/btrfs/inode.c | 4 ++++ 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 85841c53880..4119f9a9532 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -244,16 +244,17 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, split->start = start + len; split->len = em->start + em->len - (start + len); - split->orig_start = em->orig_start; split->bdev = em->bdev; split->flags = flags; if (compressed) { split->block_len = em->block_len; split->block_start = em->block_start; + split->orig_start = em->orig_start; } else { split->block_len = split->len; split->block_start = em->block_start + diff; + split->orig_start = split->start; } ret = add_extent_mapping(em_tree, split); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 59660293d29..5ca9c067237 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -610,6 +610,7 @@ static noinline int submit_compressed_extents(struct inode *inode, em = alloc_extent_map(GFP_NOFS); em->start = async_extent->start; em->len = async_extent->ram_size; + em->orig_start = em->start; em->block_start = ins.objectid; em->block_len = ins.offset; @@ -743,6 +744,7 @@ static noinline int cow_file_range(struct inode *inode, } em = alloc_extent_map(GFP_NOFS); em->start = start; + em->orig_start = em->start; ram_size = ins.offset; em->len = ins.offset; @@ -1057,6 +1059,7 @@ out_check: em_tree = &BTRFS_I(inode)->extent_tree; em = alloc_extent_map(GFP_NOFS); em->start = cur_offset; + em->orig_start = em->start; em->len = num_bytes; em->block_len = num_bytes; em->block_start = disk_bytenr; @@ -3876,6 +3879,7 @@ again: } em->bdev = root->fs_info->fs_devices->latest_bdev; em->start = EXTENT_MAP_HOLE; + em->orig_start = EXTENT_MAP_HOLE; em->len = (u64)-1; em->block_len = (u64)-1; -- cgit v1.2.3-70-g09d2 From 5b050f04c8ce911c5b6831305a24d70eab95e732 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 11 Nov 2008 09:34:41 -0500 Subject: Btrfs: Fix compile warnings on 32 bit machines Simple casting here and there to fix things up. Signed-off-by: Chris Mason --- fs/btrfs/compression.c | 3 ++- fs/btrfs/extent_io.c | 11 +++++++---- fs/btrfs/inode.c | 4 ++-- fs/btrfs/zlib.c | 2 +- 4 files changed, 12 insertions(+), 8 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index b582c6fd80f..bfd1512cce0 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -174,7 +174,8 @@ static noinline int end_compressed_writeback(struct inode *inode, u64 start, while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nr_pages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, + nr_pages, ARRAY_SIZE(pages)), pages); if (ret == 0) { nr_pages -= 1; index += 1; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 5cc0082379c..54d013c3bb8 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1157,7 +1157,8 @@ static noinline int __unlock_for_delalloc(struct inode *inode, while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nr_pages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, nr_pages, + ARRAY_SIZE(pages)), pages); for (i = 0; i < ret; i++) { if (pages[i] != locked_page) unlock_page(pages[i]); @@ -1192,7 +1193,8 @@ static noinline int lock_delalloc_pages(struct inode *inode, nrpages = end_index - index + 1; while(nrpages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nrpages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, + nrpages, ARRAY_SIZE(pages)), pages); if (ret == 0) { ret = -EAGAIN; goto done; @@ -1346,7 +1348,8 @@ int extent_clear_unlock_delalloc(struct inode *inode, while(nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, - min(nr_pages, ARRAY_SIZE(pages)), pages); + min_t(unsigned long, + nr_pages, ARRAY_SIZE(pages)), pages); for (i = 0; i < ret; i++) { if (pages[i] == locked_page) { page_cache_release(pages[i]); @@ -1896,7 +1899,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, int contig = 0; int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED; int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED; - size_t page_size = min(size, PAGE_CACHE_SIZE); + size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE); if (bio_ret && *bio_ret) { bio = *bio_ret; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5ca9c067237..2ed2deacde9 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -184,7 +184,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, int i = 0; while(compressed_size > 0) { cpage = compressed_pages[i]; - cur_size = min(compressed_size, + cur_size = min_t(unsigned long, compressed_size, PAGE_CACHE_SIZE); kaddr = kmap(cpage); @@ -3812,7 +3812,7 @@ static noinline int uncompress_inline(struct btrfs_path *path, read_extent_buffer(leaf, tmp, ptr, inline_size); - max_size = min(PAGE_CACHE_SIZE, max_size); + max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size); ret = btrfs_zlib_decompress(tmp, page, extent_offset, inline_size, max_size); if (ret) { diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c index ba2527d0873..5b9f7002513 100644 --- a/fs/btrfs/zlib.c +++ b/fs/btrfs/zlib.c @@ -370,7 +370,7 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, data_in = kmap(pages_in[page_in_index]); workspace->inf_strm.next_in = data_in; - workspace->inf_strm.avail_in = min(srclen, PAGE_CACHE_SIZE); + workspace->inf_strm.avail_in = min_t(size_t, srclen, PAGE_CACHE_SIZE); workspace->inf_strm.total_in = 0; workspace->inf_strm.total_out = 0; -- cgit v1.2.3-70-g09d2 From c146afad2c7fea6a366d4945c1bab9b03880f526 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Wed, 12 Nov 2008 14:34:12 -0500 Subject: Btrfs: mount ro and remount support This patch adds mount ro and remount support. The main changes in patch are: adding btrfs_remount and related helper function; splitting the transaction related code out of close_ctree into btrfs_commit_super; updating allocator to properly handle read only block group. Signed-off-by: Yan Zheng --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 83 ++++++++++++++++++++++++++++++++++++++------------ fs/btrfs/disk-io.h | 2 ++ fs/btrfs/extent-tree.c | 61 +++++++++++++++++++++---------------- fs/btrfs/inode.c | 15 +++------ fs/btrfs/ioctl.c | 22 +++++++++++++ fs/btrfs/super.c | 37 ++++++++++++++++++++-- 7 files changed, 163 insertions(+), 58 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f575939e025..c4c6c127323 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -541,6 +541,7 @@ struct btrfs_space_info { u64 bytes_used; u64 bytes_pinned; u64 bytes_reserved; + u64 bytes_readonly; int full; int force_alloc; struct list_head list; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3b0e974a9e9..c599f0ee997 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1075,10 +1075,12 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, kfree(root); return ERR_PTR(ret); } - ret = btrfs_find_dead_roots(fs_info->tree_root, - root->root_key.objectid, root); - BUG_ON(ret); - + if (!(fs_info->sb->s_flags & MS_RDONLY)) { + ret = btrfs_find_dead_roots(fs_info->tree_root, + root->root_key.objectid, root); + BUG_ON(ret); + btrfs_orphan_cleanup(root); + } return root; } @@ -1700,7 +1702,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_read_block_groups(extent_root); - fs_info->generation = btrfs_super_generation(disk_super) + 1; + fs_info->generation = generation + 1; + fs_info->last_trans_committed = generation; fs_info->data_alloc_profile = (u64)-1; fs_info->metadata_alloc_profile = (u64)-1; fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; @@ -1715,6 +1718,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, if (!fs_info->transaction_kthread) goto fail_cleaner; + if (sb->s_flags & MS_RDONLY) + return tree_root; + if (btrfs_super_log_root(disk_super) != 0) { u32 blocksize; u64 bytenr = btrfs_super_log_root(disk_super); @@ -1735,7 +1741,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, ret = btrfs_recover_log_trees(log_tree_root); BUG_ON(ret); } - fs_info->last_trans_committed = btrfs_super_generation(disk_super); ret = btrfs_cleanup_reloc_trees(tree_root); BUG_ON(ret); @@ -1955,28 +1960,69 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info) return 0; } -int close_ctree(struct btrfs_root *root) +int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) { + u64 root_objectid = 0; + struct btrfs_root *gang[8]; + int i; int ret; - struct btrfs_trans_handle *trans; - struct btrfs_fs_info *fs_info = root->fs_info; - fs_info->closing = 1; - smp_mb(); + while (1) { + ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, + (void **)gang, root_objectid, + ARRAY_SIZE(gang)); + if (!ret) + break; + for (i = 0; i < ret; i++) { + root_objectid = gang[i]->root_key.objectid; + ret = btrfs_find_dead_roots(fs_info->tree_root, + root_objectid, gang[i]); + BUG_ON(ret); + btrfs_orphan_cleanup(gang[i]); + } + root_objectid++; + } + return 0; +} - kthread_stop(root->fs_info->transaction_kthread); - kthread_stop(root->fs_info->cleaner_kthread); +int btrfs_commit_super(struct btrfs_root *root) +{ + struct btrfs_trans_handle *trans; + int ret; + mutex_lock(&root->fs_info->cleaner_mutex); btrfs_clean_old_snapshots(root); + mutex_unlock(&root->fs_info->cleaner_mutex); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); - /* run commit again to drop the original snapshot */ + BUG_ON(ret); + /* run commit again to drop the original snapshot */ trans = btrfs_start_transaction(root, 1); btrfs_commit_transaction(trans, root); ret = btrfs_write_and_wait_transaction(NULL, root); BUG_ON(ret); - write_ctree_super(NULL, root); + ret = write_ctree_super(NULL, root); + return ret; +} + +int close_ctree(struct btrfs_root *root) +{ + struct btrfs_fs_info *fs_info = root->fs_info; + int ret; + + fs_info->closing = 1; + smp_mb(); + + kthread_stop(root->fs_info->transaction_kthread); + kthread_stop(root->fs_info->cleaner_kthread); + + if (!(fs_info->sb->s_flags & MS_RDONLY)) { + ret = btrfs_commit_super(root); + if (ret) { + printk("btrfs: commit super returns %d\n", ret); + } + } if (fs_info->delalloc_bytes) { printk("btrfs: at unmount delalloc count %Lu\n", @@ -2000,12 +2046,10 @@ int close_ctree(struct btrfs_root *root) free_extent_buffer(root->fs_info->dev_root->node); btrfs_free_block_groups(root->fs_info); - fs_info->closing = 2; - del_fs_roots(fs_info); - filemap_write_and_wait(fs_info->btree_inode->i_mapping); + del_fs_roots(fs_info); - truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); + iput(fs_info->btree_inode); btrfs_stop_workers(&fs_info->fixup_workers); btrfs_stop_workers(&fs_info->delalloc_workers); @@ -2014,7 +2058,6 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); - iput(fs_info->btree_inode); #if 0 while(!list_empty(&fs_info->hashers)) { struct btrfs_hasher *hasher; diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index b8d5948fa27..717e94811e4 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -38,6 +38,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, int close_ctree(struct btrfs_root *root); int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root); +int btrfs_commit_super(struct btrfs_root *root); struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize); struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, @@ -49,6 +50,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, struct btrfs_key *location); struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, struct btrfs_key *location); +int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info); int btrfs_insert_dev_radix(struct btrfs_root *root, struct block_device *bdev, u64 device_id, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index e785f0a0632..af2de30dbea 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1794,7 +1794,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, *space_info = found; return 0; } - found = kmalloc(sizeof(*found), GFP_NOFS); + found = kzalloc(sizeof(*found), GFP_NOFS); if (!found) return -ENOMEM; @@ -1807,6 +1807,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags, found->bytes_used = bytes_used; found->bytes_pinned = 0; found->bytes_reserved = 0; + found->bytes_readonly = 0; found->full = 0; found->force_alloc = 0; *space_info = found; @@ -1829,6 +1830,19 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) } } +static void set_block_group_readonly(struct btrfs_block_group_cache *cache) +{ + spin_lock(&cache->space_info->lock); + spin_lock(&cache->lock); + if (!cache->ro) { + cache->space_info->bytes_readonly += cache->key.offset - + btrfs_block_group_used(&cache->item); + cache->ro = 1; + } + spin_unlock(&cache->lock); + spin_unlock(&cache->space_info->lock); +} + static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags) { u64 num_devices = root->fs_info->fs_devices->num_devices; @@ -1865,7 +1879,9 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, u64 thresh; u64 start; u64 num_bytes; - int ret = 0, waited = 0; + int ret = 0; + + mutex_lock(&extent_root->fs_info->chunk_mutex); flags = reduce_alloc_profile(extent_root, flags); @@ -1887,46 +1903,28 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, goto out; } - thresh = div_factor(space_info->total_bytes, 6); + thresh = space_info->total_bytes - space_info->bytes_readonly; + thresh = div_factor(thresh, 6); if (!force && (space_info->bytes_used + space_info->bytes_pinned + space_info->bytes_reserved + alloc_bytes) < thresh) { spin_unlock(&space_info->lock); goto out; } - spin_unlock(&space_info->lock); - ret = mutex_trylock(&extent_root->fs_info->chunk_mutex); - if (!ret && !force) { - goto out; - } else if (!ret) { - mutex_lock(&extent_root->fs_info->chunk_mutex); - waited = 1; - } - - if (waited) { - spin_lock(&space_info->lock); - if (space_info->full) { - spin_unlock(&space_info->lock); - goto out_unlock; - } - spin_unlock(&space_info->lock); - } - ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); if (ret) { printk("space info full %Lu\n", flags); space_info->full = 1; - goto out_unlock; + goto out; } ret = btrfs_make_block_group(trans, extent_root, 0, flags, BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes); BUG_ON(ret); -out_unlock: - mutex_unlock(&extent_root->fs_info->chunk_mutex); out: + mutex_unlock(&extent_root->fs_info->chunk_mutex); return ret; } @@ -1956,12 +1954,18 @@ static int update_block_group(struct btrfs_trans_handle *trans, if (alloc) { old_val += num_bytes; cache->space_info->bytes_used += num_bytes; + if (cache->ro) { + cache->space_info->bytes_readonly -= num_bytes; + WARN_ON(1); + } btrfs_set_block_group_used(&cache->item, old_val); spin_unlock(&cache->lock); spin_unlock(&cache->space_info->lock); } else { old_val -= num_bytes; cache->space_info->bytes_used -= num_bytes; + if (cache->ro) + cache->space_info->bytes_readonly += num_bytes; btrfs_set_block_group_used(&cache->item, old_val); spin_unlock(&cache->lock); spin_unlock(&cache->space_info->lock); @@ -5560,8 +5564,7 @@ int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start) BUG_ON(IS_ERR(reloc_inode)); __alloc_chunk_for_shrink(root, block_group, 1); - block_group->ro = 1; - block_group->space_info->total_bytes -= block_group->key.offset; + set_block_group_readonly(block_group); btrfs_start_delalloc_inodes(info->tree_root); btrfs_wait_ordered_extents(info->tree_root, 0); @@ -5868,6 +5871,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, block_group = btrfs_lookup_block_group(root->fs_info, group_start); BUG_ON(!block_group); + BUG_ON(!block_group->ro); memcpy(&key, &block_group->key, sizeof(key)); @@ -5881,6 +5885,11 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, list_del(&block_group->list); up_write(&block_group->space_info->groups_sem); + spin_lock(&block_group->space_info->lock); + block_group->space_info->total_bytes -= block_group->key.offset; + block_group->space_info->bytes_readonly -= block_group->key.offset; + spin_unlock(&block_group->space_info->lock); + /* memset(shrink_block_group, 0, sizeof(*shrink_block_group)); kfree(shrink_block_group); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2ed2deacde9..3e3620e69bb 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1808,10 +1808,6 @@ void btrfs_orphan_cleanup(struct btrfs_root *root) struct inode *inode; int ret = 0, nr_unlink = 0, nr_truncate = 0; - /* don't do orphan cleanup if the fs is readonly. */ - if (root->fs_info->sb->s_flags & MS_RDONLY) - return; - path = btrfs_alloc_path(); if (!path) return; @@ -3050,7 +3046,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, struct btrfs_root *root = bi->root; struct btrfs_root *sub_root = root; struct btrfs_key location; - int ret, new, do_orphan = 0; + int ret, new; if (dentry->d_name.len > BTRFS_NAME_LEN) return ERR_PTR(-ENAMETOOLONG); @@ -3076,13 +3072,9 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, if (new && root != sub_root) { igrab(inode); sub_root->inode = inode; - do_orphan = 1; } } - if (unlikely(do_orphan)) - btrfs_orphan_cleanup(sub_root); - return d_splice_alias(inode, dentry); } @@ -3237,7 +3229,7 @@ int btrfs_write_inode(struct inode *inode, int wait) struct btrfs_trans_handle *trans; int ret = 0; - if (root->fs_info->closing > 1) + if (root->fs_info->btree_inode == inode) return 0; if (wait) { @@ -4625,6 +4617,9 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) struct inode *inode; unsigned long flags; + if (root->fs_info->sb->s_flags & MS_RDONLY) + return -EROFS; + spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); while(!list_empty(head)) { binode = list_entry(head->next, struct btrfs_inode, diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 4d7cc7c504d..52863cebd59 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -378,6 +378,9 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg) int namelen; int mod = 0; + if (root->fs_info->sb->s_flags & MS_RDONLY) + return -EROFS; + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); if (!vol_args) @@ -478,6 +481,9 @@ static noinline int btrfs_ioctl_snap_create(struct file *file, int namelen; int ret; + if (root->fs_info->sb->s_flags & MS_RDONLY) + return -EROFS; + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); if (!vol_args) @@ -534,6 +540,11 @@ static int btrfs_ioctl_defrag(struct file *file) { struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; + int ret; + + ret = mnt_want_write(file->f_path.mnt); + if (ret) + return ret; switch (inode->i_mode & S_IFMT) { case S_IFDIR: @@ -575,6 +586,9 @@ long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) struct btrfs_ioctl_vol_args *vol_args; int ret; + if (root->fs_info->sb->s_flags & MS_RDONLY) + return -EROFS; + vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS); if (!vol_args) @@ -621,6 +635,10 @@ long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, u64 off, * they don't overlap)? */ + ret = mnt_want_write(file->f_path.mnt); + if (ret) + return ret; + src_file = fget(srcfd); if (!src_file) return -EBADF; @@ -958,6 +976,10 @@ long btrfs_ioctl_trans_start(struct file *file) goto out; } + ret = mnt_want_write(file->f_path.mnt); + if (ret) + goto out; + mutex_lock(&root->fs_info->trans_mutex); root->fs_info->open_ioctl_trans++; mutex_unlock(&root->fs_info->trans_mutex); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index ab9d5e89ed1..04a3bf81650 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -370,6 +370,9 @@ int btrfs_sync_fs(struct super_block *sb, int wait) int ret; root = btrfs_sb(sb); + if (sb->s_flags & MS_RDONLY) + return 0; + sb->s_dirt = 0; if (!wait) { filemap_flush(root->fs_info->btree_inode->i_mapping); @@ -438,7 +441,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags, up_write(&s->s_umount); deactivate_super(s); error = -EBUSY; - goto error_bdev; + goto error_close_devices; } } else { @@ -487,7 +490,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags, error_s: error = PTR_ERR(s); -error_bdev: +error_close_devices: btrfs_close_devices(fs_devices); error_free_subvol_name: kfree(subvol_name); @@ -495,6 +498,35 @@ error: return error; } +static int btrfs_remount(struct super_block *sb, int *flags, char *data) +{ + struct btrfs_root *root = btrfs_sb(sb); + int ret; + + if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) + return 0; + + if (*flags & MS_RDONLY) { + sb->s_flags |= MS_RDONLY; + + ret = btrfs_commit_super(root); + WARN_ON(ret); + } else { + if (btrfs_super_log_root(&root->fs_info->super_copy) != 0) + return -EINVAL; + + ret = btrfs_cleanup_reloc_trees(root); + WARN_ON(ret); + + ret = btrfs_cleanup_fs_roots(root->fs_info); + WARN_ON(ret); + + sb->s_flags &= ~MS_RDONLY; + } + + return 0; +} + static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) { struct btrfs_root *root = btrfs_sb(dentry->d_sb); @@ -582,6 +614,7 @@ static struct super_operations btrfs_super_ops = { .alloc_inode = btrfs_alloc_inode, .destroy_inode = btrfs_destroy_inode, .statfs = btrfs_statfs, + .remount_fs = btrfs_remount, .write_super_lockfs = btrfs_write_super_lockfs, .unlockfs = btrfs_unlockfs, }; -- cgit v1.2.3-70-g09d2 From 3de4586c5278a28107030c336956381f69ff7a9d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Nov 2008 21:02:50 -0500 Subject: Btrfs: Allow subvolumes and snapshots anywhere in the directory tree Before, all snapshots and subvolumes lived in a single flat directory. This was awkward and confusing because the single flat directory was only writable with the ioctls. This commit changes the ioctls to create subvols and snapshots at any point in the directory tree. This requires making separate ioctls for snapshot and subvol creation instead of a combining them into one. The subvol ioctl does: btrfsctl -S subvol_name parent_dir After the ioctl is done subvol_name lives inside parent_dir. The snapshot ioctl does: btrfsctl -s path_for_snapshot root_to_snapshot path_for_snapshot can be an absolute or relative path. btrfsctl breaks it up into directory and basename components. root_to_snapshot can be any file or directory in the FS. The snapshot is taken of the entire root where that file lives. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 7 +++-- fs/btrfs/disk-io.c | 14 +++++++--- fs/btrfs/inode.c | 50 ++++++++++++++++++++++------------- fs/btrfs/ioctl.c | 71 ++++++++++++++++++++++++++++++++++++-------------- fs/btrfs/ioctl.h | 7 +++-- fs/btrfs/super.c | 10 +++---- fs/btrfs/transaction.c | 66 +++++++++++++++++++++++++++++++++++++--------- fs/btrfs/transaction.h | 2 ++ 8 files changed, 162 insertions(+), 65 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5ff74282a62..5611f8e035a 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -606,6 +606,7 @@ struct btrfs_fs_info { struct btrfs_root *tree_root; struct btrfs_root *chunk_root; struct btrfs_root *dev_root; + struct btrfs_root *fs_root; /* the log root tree is a directory of all the other log roots */ struct btrfs_root *log_root_tree; @@ -758,7 +759,6 @@ struct btrfs_root { struct btrfs_root_item root_item; struct btrfs_key root_key; struct btrfs_fs_info *fs_info; - struct inode *inode; struct extent_io_tree dirty_log_pages; struct kobject root_kobj; @@ -1876,6 +1876,8 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans, #define PageChecked PageFsMisc #endif +struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry); +int btrfs_set_inode_index(struct inode *dir, u64 *index); int btrfs_unlink_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *dir, struct inode *inode, @@ -1896,9 +1898,6 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, struct btrfs_trans_handle *trans, u64 new_dirid, struct btrfs_block_group_cache *block_group); -void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, - int namelen); - int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio, unsigned long bio_flags); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 82833e5d84b..0a5350573f6 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -838,7 +838,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, u64 objectid) { root->node = NULL; - root->inode = NULL; root->commit_root = NULL; root->ref_tree = NULL; root->sectorsize = sectorsize; @@ -1430,6 +1429,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, u32 blocksize; u32 stripesize; u64 generation; + struct btrfs_key location; struct buffer_head *bh; struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); @@ -1729,7 +1729,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, goto fail_cleaner; if (sb->s_flags & MS_RDONLY) - return tree_root; + goto read_fs_root; if (btrfs_super_log_root(disk_super) != 0) { u32 blocksize; @@ -1755,6 +1755,14 @@ struct btrfs_root *open_ctree(struct super_block *sb, ret = btrfs_cleanup_reloc_trees(tree_root); BUG_ON(ret); + location.objectid = BTRFS_FS_TREE_OBJECTID; + location.type = BTRFS_ROOT_ITEM_KEY; + location.offset = (u64)-1; + +read_fs_root: + fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); + if (!fs_info->fs_root) + goto fail_cleaner; return tree_root; fail_cleaner: @@ -1944,8 +1952,6 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) (unsigned long)root->root_key.objectid); if (root->in_sysfs) btrfs_sysfs_del_root(root); - if (root->inode) - iput(root->inode); if (root->node) free_extent_buffer(root->node); if (root->commit_root) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3e3620e69bb..e163b1b7470 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3038,8 +3038,7 @@ struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, return inode; } -static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, - struct nameidata *nd) +struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry) { struct inode * inode; struct btrfs_inode *bi = BTRFS_I(dir); @@ -3067,13 +3066,21 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, inode = btrfs_iget(dir->i_sb, &location, sub_root, &new); if (IS_ERR(inode)) return ERR_CAST(inode); - - /* the inode and parent dir are two different roots */ - if (new && root != sub_root) { - igrab(inode); - sub_root->inode = inode; - } } + return inode; +} + +static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, + struct nameidata *nd) +{ + struct inode *inode; + + if (dentry->d_name.len > BTRFS_NAME_LEN) + return ERR_PTR(-ENAMETOOLONG); + + inode = btrfs_lookup_dentry(dir, dentry); + if (IS_ERR(inode)) + return ERR_CAST(inode); return d_splice_alias(inode, dentry); } @@ -3129,7 +3136,6 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, return 0; filp->f_pos = 2; } - path = btrfs_alloc_path(); path->reada = 2; @@ -3159,6 +3165,7 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, path->slots[0]++; } } + advance = 1; item = btrfs_item_nr(leaf, slot); btrfs_item_key_to_cpu(leaf, &found_key, slot); @@ -3194,16 +3201,25 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)]; btrfs_dir_item_key_to_cpu(leaf, di, &location); + + /* is this a reference to our own snapshot? If so + * skip it + */ + if (location.type == BTRFS_ROOT_ITEM_KEY && + location.objectid == root->root_key.objectid) { + over = 0; + goto skip; + } over = filldir(dirent, name_ptr, name_len, found_key.offset, location.objectid, d_type); +skip: if (name_ptr != tmp_name) kfree(name_ptr); if (over) goto nopos; - di_len = btrfs_dir_name_len(leaf, di) + btrfs_dir_data_len(leaf, di) + sizeof(*di); di_cur += di_len; @@ -3318,8 +3334,7 @@ out: * helper to find a free sequence number in a given directory. This current * code is very simple, later versions will do smarter things in the btree */ -static int btrfs_set_inode_index(struct inode *dir, struct inode *inode, - u64 *index) +int btrfs_set_inode_index(struct inode *dir, u64 *index) { int ret = 0; @@ -3365,7 +3380,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, return ERR_PTR(-ENOMEM); if (dir) { - ret = btrfs_set_inode_index(dir, inode, index); + ret = btrfs_set_inode_index(dir, index); if (ret) return ERR_PTR(ret); } @@ -3651,7 +3666,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, err = btrfs_check_free_space(root, 1, 0); if (err) goto fail; - err = btrfs_set_inode_index(dir, inode, &index); + err = btrfs_set_inode_index(dir, &index); if (err) goto fail; @@ -4349,13 +4364,13 @@ out: * Invalidate a single dcache entry at the root of the filesystem. * Needed after creation of snapshot or subvolume. */ -void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, +void btrfs_invalidate_dcache_root(struct inode *dir, char *name, int namelen) { struct dentry *alias, *entry; struct qstr qstr; - alias = d_find_alias(root->fs_info->sb->s_root->d_inode); + alias = d_find_alias(dir); if (alias) { qstr.name = name; qstr.len = namelen; @@ -4387,7 +4402,6 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, return PTR_ERR(inode); inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; - new_root->inode = inode; inode->i_nlink = 1; btrfs_i_size_write(inode, 0); @@ -4590,7 +4604,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, } } - ret = btrfs_set_inode_index(new_dir, old_inode, &index); + ret = btrfs_set_inode_index(new_dir, &index); if (ret) goto out_fail; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index f43df72b0e1..ec45b308613 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -67,6 +67,7 @@ static noinline int create_subvol(struct btrfs_root *root, int err; u64 objectid; u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID; + u64 index = 0; unsigned long nr = 1; ret = btrfs_check_free_space(root, 1, 0); @@ -126,6 +127,7 @@ static noinline int create_subvol(struct btrfs_root *root, key.objectid = objectid; key.offset = 1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); +printk("inserting root objectid %Lu\n", objectid); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &root_item); if (ret) @@ -135,24 +137,27 @@ static noinline int create_subvol(struct btrfs_root *root, * insert the directory item */ key.offset = (u64)-1; - dir = root->fs_info->sb->s_root->d_inode; - ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, + dir = dentry->d_parent->d_inode; + ret = btrfs_set_inode_index(dir, &index); + BUG_ON(ret); + + ret = btrfs_insert_dir_item(trans, root, name, namelen, dir->i_ino, &key, - BTRFS_FT_DIR, 0); + BTRFS_FT_DIR, index); if (ret) goto fail; - +#if 0 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, name, namelen, objectid, root->fs_info->sb->s_root->d_inode->i_ino, 0); if (ret) goto fail; - +#endif ret = btrfs_commit_transaction(trans, root); if (ret) goto fail_commit; - new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen); + new_root = btrfs_read_fs_root_no_name(root->fs_info, &key); BUG_ON(!new_root); trans = btrfs_start_transaction(new_root, 1); @@ -170,14 +175,16 @@ fail: ret = err; fail_commit: btrfs_btree_balance_dirty(root, nr); +printk("all done ret %d\n", ret); return ret; } -static int create_snapshot(struct btrfs_root *root, char *name, int namelen) +static int create_snapshot(struct btrfs_root *root, struct dentry *dentry, + char *name, int namelen) { struct btrfs_pending_snapshot *pending_snapshot; struct btrfs_trans_handle *trans; - int ret; + int ret = 0; int err; unsigned long nr = 0; @@ -188,7 +195,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) if (ret) goto fail_unlock; - pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS); + pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS); if (!pending_snapshot) { ret = -ENOMEM; goto fail_unlock; @@ -201,12 +208,12 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen) } memcpy(pending_snapshot->name, name, namelen); pending_snapshot->name[namelen] = '\0'; + pending_snapshot->dentry = dentry; trans = btrfs_start_transaction(root, 1); BUG_ON(!trans); pending_snapshot->root = root; list_add(&pending_snapshot->list, &trans->transaction->pending_snapshots); - ret = btrfs_update_inode(trans, root, root->inode); err = btrfs_commit_transaction(trans, root); fail_unlock: @@ -230,7 +237,8 @@ static inline int btrfs_may_create(struct inode *dir, struct dentry *child) * inside this filesystem so it's quite a bit simpler. */ static noinline int btrfs_mksubvol(struct path *parent, char *name, - int mode, int namelen) + int mode, int namelen, + struct btrfs_root *snap_src) { struct dentry *dentry; int error; @@ -248,6 +256,7 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name, if (!IS_POSIXACL(parent->dentry->d_inode)) mode &= ~current->fs->umask; + error = mnt_want_write(parent->mnt); if (error) goto out_dput; @@ -266,8 +275,12 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name, * Also we should pass on the mode eventually to allow creating new * subvolume with specific mode bits. */ - error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, dentry, - name, namelen); + if (snap_src) { + error = create_snapshot(snap_src, dentry, name, namelen); + } else { + error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root, + dentry, name, namelen); + } if (error) goto out_drop_write; @@ -471,15 +484,16 @@ out: } static noinline int btrfs_ioctl_snap_create(struct file *file, - void __user *arg) + void __user *arg, int subvol) { struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; struct btrfs_ioctl_vol_args *vol_args; struct btrfs_dir_item *di; struct btrfs_path *path; + struct file *src_file; u64 root_dirid; int namelen; - int ret; + int ret = 0; if (root->fs_info->sb->s_flags & MS_RDONLY) return -EROFS; @@ -523,12 +537,29 @@ static noinline int btrfs_ioctl_snap_create(struct file *file, goto out; } - if (root == root->fs_info->tree_root) { + if (subvol) { ret = btrfs_mksubvol(&file->f_path, vol_args->name, file->f_path.dentry->d_inode->i_mode, - namelen); + namelen, NULL); } else { - ret = create_snapshot(root, vol_args->name, namelen); + struct inode *src_inode; + src_file = fget(vol_args->fd); + if (!src_file) { + ret = -EINVAL; + goto out; + } + + src_inode = src_file->f_path.dentry->d_inode; + if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { + printk("btrfs: Snapshot src from another FS\n"); + ret = -EINVAL; + fput(src_file); + goto out; + } + ret = btrfs_mksubvol(&file->f_path, vol_args->name, + file->f_path.dentry->d_inode->i_mode, + namelen, BTRFS_I(src_inode)->root); + fput(src_file); } out: @@ -1030,7 +1061,9 @@ long btrfs_ioctl(struct file *file, unsigned int switch (cmd) { case BTRFS_IOC_SNAP_CREATE: - return btrfs_ioctl_snap_create(file, (void __user *)arg); + return btrfs_ioctl_snap_create(file, (void __user *)arg, 0); + case BTRFS_IOC_SUBVOL_CREATE: + return btrfs_ioctl_snap_create(file, (void __user *)arg, 1); case BTRFS_IOC_DEFRAG: return btrfs_ioctl_defrag(file); case BTRFS_IOC_RESIZE: diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h index 989ba8a0121..78049ea208d 100644 --- a/fs/btrfs/ioctl.h +++ b/fs/btrfs/ioctl.h @@ -22,9 +22,10 @@ #define BTRFS_IOCTL_MAGIC 0x94 #define BTRFS_VOL_NAME_MAX 255 -#define BTRFS_PATH_NAME_MAX 4095 +#define BTRFS_PATH_NAME_MAX 3072 struct btrfs_ioctl_vol_args { + __s64 fd; char name[BTRFS_PATH_NAME_MAX + 1]; }; @@ -51,7 +52,6 @@ struct btrfs_ioctl_vol_args { struct btrfs_ioctl_vol_args) #define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \ struct btrfs_ioctl_vol_args) - struct btrfs_ioctl_clone_range_args { __s64 src_fd; __u64 src_offset, src_length; @@ -61,4 +61,7 @@ struct btrfs_ioctl_clone_range_args { #define BTRFS_IOC_CLONE_RANGE _IOW(BTRFS_IOCTL_MAGIC, 13, \ struct btrfs_ioctl_clone_range_args) +#define BTRFS_IOC_SUBVOL_CREATE _IOW(BTRFS_IOCTL_MAGIC, 14, \ + struct btrfs_ioctl_vol_args) + #endif diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 92393cc60d0..77c5eff3e20 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -285,11 +285,11 @@ static int btrfs_parse_early_options(const char *options, int flags, out: /* * If no subvolume name is specified we use the default one. Allocate - * a copy of the string "default" here so that code later in the + * a copy of the string "." here so that code later in the * mount path doesn't care if it's the default volume or another one. */ if (!*subvol_name) { - *subvol_name = kstrdup("default", GFP_KERNEL); + *subvol_name = kstrdup(".", GFP_KERNEL); if (!*subvol_name) return -ENOMEM; } @@ -323,12 +323,12 @@ static int btrfs_fill_super(struct super_block * sb, } sb->s_fs_info = tree_root; disk_super = &tree_root->fs_info->super_copy; - inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super), - tree_root); + inode = btrfs_iget_locked(sb, BTRFS_FIRST_FREE_OBJECTID, + tree_root->fs_info->fs_root); bi = BTRFS_I(inode); bi->location.objectid = inode->i_ino; bi->location.offset = 0; - bi->root = tree_root; + bi->root = tree_root->fs_info->fs_root; btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 202c1b6df4a..eec8b246503 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -779,7 +779,6 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, struct extent_buffer *tmp; struct extent_buffer *old; int ret; - int namelen; u64 objectid; new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS); @@ -816,28 +815,48 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, if (ret) goto fail; + key.offset = (u64)-1; + memcpy(&pending->root_key, &key, sizeof(key)); +fail: + kfree(new_root_item); + return ret; +} + +static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info, + struct btrfs_pending_snapshot *pending) +{ + int ret; + int namelen; + u64 index = 0; + struct btrfs_trans_handle *trans; + struct inode *parent_inode; + struct inode *inode; + + trans = btrfs_start_transaction(fs_info->fs_root, 1); + /* * insert the directory item */ - key.offset = (u64)-1; namelen = strlen(pending->name); - ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, - pending->name, namelen, - root->fs_info->sb->s_root->d_inode->i_ino, - &key, BTRFS_FT_DIR, 0); + parent_inode = pending->dentry->d_parent->d_inode; + ret = btrfs_set_inode_index(parent_inode, &index); + ret = btrfs_insert_dir_item(trans, + BTRFS_I(parent_inode)->root, + pending->name, namelen, + parent_inode->i_ino, + &pending->root_key, BTRFS_FT_DIR, index); if (ret) goto fail; - +#if 0 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, pending->name, strlen(pending->name), objectid, root->fs_info->sb->s_root->d_inode->i_ino, 0); - - /* Invalidate existing dcache entry for new snapshot. */ - btrfs_invalidate_dcache_root(root, pending->name, namelen); - +#endif + inode = btrfs_lookup_dentry(parent_inode, pending->dentry); + d_instantiate(pending->dentry, inode); fail: - kfree(new_root_item); + btrfs_end_transaction(trans, fs_info->fs_root); return ret; } @@ -846,6 +865,22 @@ fail: */ static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info) +{ + struct btrfs_pending_snapshot *pending; + struct list_head *head = &trans->transaction->pending_snapshots; + struct list_head *cur; + int ret; + + list_for_each(cur, head) { + pending = list_entry(cur, struct btrfs_pending_snapshot, list); + ret = create_pending_snapshot(trans, fs_info, pending); + BUG_ON(ret); + } + return 0; +} + +static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans, + struct btrfs_fs_info *fs_info) { struct btrfs_pending_snapshot *pending; struct list_head *head = &trans->transaction->pending_snapshots; @@ -854,7 +889,7 @@ static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans, while(!list_empty(head)) { pending = list_entry(head->next, struct btrfs_pending_snapshot, list); - ret = create_pending_snapshot(trans, fs_info, pending); + ret = finish_pending_snapshot(fs_info, pending); BUG_ON(ret); list_del(&pending->list); kfree(pending->name); @@ -1033,11 +1068,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_drop_dead_reloc_roots(root); mutex_unlock(&root->fs_info->tree_reloc_mutex); + /* do the directory inserts of any pending snapshot creations */ + finish_pending_snapshots(trans, root->fs_info); + mutex_lock(&root->fs_info->trans_mutex); cur_trans->commit_done = 1; root->fs_info->last_trans_committed = cur_trans->transid; wake_up(&cur_trans->commit_wait); + put_transaction(cur_trans); put_transaction(cur_trans); @@ -1046,6 +1085,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); mutex_unlock(&root->fs_info->trans_mutex); + kmem_cache_free(btrfs_trans_handle_cachep, trans); if (root->fs_info->closing) { diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index eef2cb7d7e7..202c8be6c05 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -47,8 +47,10 @@ struct btrfs_trans_handle { }; struct btrfs_pending_snapshot { + struct dentry *dentry; struct btrfs_root *root; char *name; + struct btrfs_key root_key; struct list_head list; }; -- cgit v1.2.3-70-g09d2 From 3394e1607eaf870ebba37d303fbd590a4c569908 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Nov 2008 20:42:26 -0500 Subject: Btrfs: Give each subvol and snapshot their own anonymous devid Each subvolume has its own private inode number space, and so we need to fill in different device numbers for each subvolume to avoid confusing applications. This commit puts a struct super_block into struct btrfs_root so it can call set_anon_super() and get a different device number generated for each root. btrfs_rename is changed to prevent renames across subvols. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 6 ++++++ fs/btrfs/disk-io.c | 13 +++++++++++++ fs/btrfs/inode.c | 15 +++++++++++++-- fs/btrfs/ioctl.c | 2 -- fs/btrfs/transaction.c | 4 ++-- 5 files changed, 34 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5611f8e035a..b3bc65b08c6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -799,6 +799,12 @@ struct btrfs_root { spinlock_t list_lock; struct list_head dead_list; struct list_head orphan_list; + + /* + * right now this just gets used so that a root has its own devid + * for stat. It may be used for more later + */ + struct super_block anon_super; }; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 0a5350573f6..8d7866b733d 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -877,6 +877,12 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, root->defrag_running = 0; root->defrag_level = 0; root->root_key.objectid = objectid; + root->anon_super.s_root = NULL; + root->anon_super.s_dev = 0; + INIT_LIST_HEAD(&root->anon_super.s_list); + INIT_LIST_HEAD(&root->anon_super.s_instances); + init_rwsem(&root->anon_super.s_umount); + return 0; } @@ -1083,6 +1089,9 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); if (IS_ERR(root)) return root; + + set_anon_super(&root->anon_super, NULL); + ret = radix_tree_insert(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, root); @@ -1950,6 +1959,10 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) { radix_tree_delete(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid); + if (root->anon_super.s_dev) { + down_write(&root->anon_super.s_umount); + kill_anon_super(&root->anon_super); + } if (root->in_sysfs) btrfs_sysfs_del_root(root); if (root->node) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e163b1b7470..7ef79ce86e2 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2212,7 +2212,12 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) struct btrfs_trans_handle *trans; unsigned long nr = 0; - if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) { + /* + * the FIRST_FREE_OBJECTID check makes sure we don't try to rmdir + * the root of a subvolume or snapshot + */ + if (inode->i_size > BTRFS_EMPTY_DIR_SIZE || + inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) { return -ENOTEMPTY; } @@ -4410,7 +4415,6 @@ int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, if (error) return error; - atomic_inc(&inode->i_count); d_instantiate(dentry, inode); return 0; } @@ -4548,6 +4552,7 @@ static int btrfs_getattr(struct vfsmount *mnt, { struct inode *inode = dentry->d_inode; generic_fillattr(inode, stat); + stat->dev = BTRFS_I(inode)->root->anon_super.s_dev; stat->blksize = PAGE_CACHE_SIZE; stat->blocks = (inode_get_bytes(inode) + BTRFS_I(inode)->delalloc_bytes) >> 9; @@ -4565,6 +4570,11 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, u64 index = 0; int ret; + /* we're not allowed to rename between subvolumes */ + if (BTRFS_I(old_inode)->root->root_key.objectid != + BTRFS_I(new_dir)->root->root_key.objectid) + return -EXDEV; + if (S_ISDIR(old_inode->i_mode) && new_inode && new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) { return -ENOTEMPTY; @@ -4920,6 +4930,7 @@ static int btrfs_permission(struct inode *inode, int mask) } static struct inode_operations btrfs_dir_inode_operations = { + .getattr = btrfs_getattr, .lookup = btrfs_lookup, .create = btrfs_create, .unlink = btrfs_unlink, diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index ec45b308613..773db07b5f7 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -127,7 +127,6 @@ static noinline int create_subvol(struct btrfs_root *root, key.objectid = objectid; key.offset = 1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); -printk("inserting root objectid %Lu\n", objectid); ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, &root_item); if (ret) @@ -175,7 +174,6 @@ fail: ret = err; fail_commit: btrfs_btree_balance_dirty(root, nr); -printk("all done ret %d\n", ret); return ret; } diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index eec8b246503..93f23a456a3 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -832,13 +832,13 @@ static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info, struct inode *parent_inode; struct inode *inode; - trans = btrfs_start_transaction(fs_info->fs_root, 1); + parent_inode = pending->dentry->d_parent->d_inode; + trans = btrfs_start_transaction(BTRFS_I(parent_inode)->root, 1); /* * insert the directory item */ namelen = strlen(pending->name); - parent_inode = pending->dentry->d_parent->d_inode; ret = btrfs_set_inode_index(parent_inode, &index); ret = btrfs_insert_dir_item(trans, BTRFS_I(parent_inode)->root, -- cgit v1.2.3-70-g09d2 From 0660b5af3f7ac0fac69de975914e1f4a3a586fb3 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 17 Nov 2008 20:37:39 -0500 Subject: Btrfs: Add backrefs and forward refs for subvols and snapshots Subvols and snapshots can now be referenced from any point in the directory tree. We need to maintain back refs for them so we can find lost subvols. Forward refs are added so that we know all of the subvols and snapshots referenced anywhere in the directory tree of a single subvol. This can be used to do recursive snapshotting (but they aren't yet) and it is also used to detect and prevent directory loops when creating new snapshots. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 67 +++++++++++++++++++++++++++++++++----------- fs/btrfs/inode.c | 6 ++++ fs/btrfs/ioctl.c | 24 +++++++++++----- fs/btrfs/root-tree.c | 76 +++++++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/transaction.c | 31 ++++++++++++++------ 5 files changed, 172 insertions(+), 32 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b3bc65b08c6..ad2cbe63503 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -469,6 +469,15 @@ struct btrfs_root_item { u8 level; } __attribute__ ((__packed__)); +/* + * this is used for both forward and backward root refs + */ +struct btrfs_root_ref { + __le64 dirid; + __le64 sequence; + __le16 name_len; +} __attribute__ ((__packed__)); + #define BTRFS_FILE_EXTENT_INLINE 0 #define BTRFS_FILE_EXTENT_REG 1 #define BTRFS_FILE_EXTENT_PREALLOC 2 @@ -814,27 +823,27 @@ struct btrfs_root { * the FS */ #define BTRFS_INODE_ITEM_KEY 1 -#define BTRFS_INODE_REF_KEY 2 -#define BTRFS_XATTR_ITEM_KEY 8 -#define BTRFS_ORPHAN_ITEM_KEY 9 +#define BTRFS_INODE_REF_KEY 12 +#define BTRFS_XATTR_ITEM_KEY 24 +#define BTRFS_ORPHAN_ITEM_KEY 48 /* reserve 2-15 close to the inode for later flexibility */ /* * dir items are the name -> inode pointers in a directory. There is one * for every name in a directory. */ -#define BTRFS_DIR_LOG_ITEM_KEY 14 -#define BTRFS_DIR_LOG_INDEX_KEY 15 -#define BTRFS_DIR_ITEM_KEY 16 -#define BTRFS_DIR_INDEX_KEY 17 +#define BTRFS_DIR_LOG_ITEM_KEY 60 +#define BTRFS_DIR_LOG_INDEX_KEY 72 +#define BTRFS_DIR_ITEM_KEY 84 +#define BTRFS_DIR_INDEX_KEY 96 /* * extent data is for file data */ -#define BTRFS_EXTENT_DATA_KEY 18 +#define BTRFS_EXTENT_DATA_KEY 108 /* * csum items have the checksums for data in the extents */ -#define BTRFS_CSUM_ITEM_KEY 19 +#define BTRFS_CSUM_ITEM_KEY 120 /* reserve 21-31 for other file/dir stuff */ @@ -843,23 +852,37 @@ struct btrfs_root { * root items point to tree roots. There are typically in the root * tree used by the super block to find all the other trees */ -#define BTRFS_ROOT_ITEM_KEY 32 +#define BTRFS_ROOT_ITEM_KEY 132 + +/* + * root backrefs tie subvols and snapshots to the directory entries that + * reference them + */ +#define BTRFS_ROOT_BACKREF_KEY 144 + +/* + * root refs make a fast index for listing all of the snapshots and + * subvolumes referenced by a given root. They point directly to the + * directory item in the root that references the subvol + */ +#define BTRFS_ROOT_REF_KEY 156 + /* * extent items are in the extent map tree. These record which blocks * are used, and how many references there are to each block */ -#define BTRFS_EXTENT_ITEM_KEY 33 -#define BTRFS_EXTENT_REF_KEY 34 +#define BTRFS_EXTENT_ITEM_KEY 168 +#define BTRFS_EXTENT_REF_KEY 180 /* * block groups give us hints into the extent allocation trees. Which * blocks are free etc etc */ -#define BTRFS_BLOCK_GROUP_ITEM_KEY 50 +#define BTRFS_BLOCK_GROUP_ITEM_KEY 192 -#define BTRFS_DEV_EXTENT_KEY 75 -#define BTRFS_DEV_ITEM_KEY 76 -#define BTRFS_CHUNK_ITEM_KEY 77 +#define BTRFS_DEV_EXTENT_KEY 204 +#define BTRFS_DEV_ITEM_KEY 216 +#define BTRFS_CHUNK_ITEM_KEY 228 /* * string items are for debugging. They just store a short string of @@ -1274,6 +1297,13 @@ static inline void btrfs_set_item_key(struct extent_buffer *eb, BTRFS_SETGET_FUNCS(dir_log_end, struct btrfs_dir_log_item, end, 64); +/* + * struct btrfs_root_ref + */ +BTRFS_SETGET_FUNCS(root_ref_dirid, struct btrfs_root_ref, dirid, 64); +BTRFS_SETGET_FUNCS(root_ref_sequence, struct btrfs_root_ref, sequence, 64); +BTRFS_SETGET_FUNCS(root_ref_name_len, struct btrfs_root_ref, name_len, 16); + /* struct btrfs_dir_item */ BTRFS_SETGET_FUNCS(dir_data_len, struct btrfs_dir_item, data_len, 16); BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8); @@ -1771,6 +1801,11 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans, struct extent_buffer *node, struct extent_buffer *parent); /* root-item.c */ +int btrfs_add_root_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *tree_root, + u64 root_id, u8 type, u64 ref_id, + u64 dirid, u64 sequence, + const char *name, int name_len); int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key); int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7ef79ce86e2..6854bf41856 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4580,6 +4580,12 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, return -ENOTEMPTY; } + /* to rename a snapshot or subvolume, we need to juggle the + * backrefs. This isn't coded yet + */ + if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) + return -EXDEV; + ret = btrfs_check_free_space(root, 1, 0); if (ret) goto out_unlock; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 773db07b5f7..536ae883780 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -145,13 +145,23 @@ static noinline int create_subvol(struct btrfs_root *root, BTRFS_FT_DIR, index); if (ret) goto fail; -#if 0 - ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, - name, namelen, objectid, - root->fs_info->sb->s_root->d_inode->i_ino, 0); - if (ret) - goto fail; -#endif + + /* add the backref first */ + ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, + objectid, BTRFS_ROOT_BACKREF_KEY, + root->root_key.objectid, + dir->i_ino, index, name, namelen); + + BUG_ON(ret); + + /* now add the forward ref */ + ret = btrfs_add_root_ref(trans, root->fs_info->tree_root, + root->root_key.objectid, BTRFS_ROOT_REF_KEY, + objectid, + dir->i_ino, index, name, namelen); + + BUG_ON(ret); + ret = btrfs_commit_transaction(trans, root); if (ret) goto fail_commit; diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index eb7f7655e9d..e9be3abcb36 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -84,7 +84,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, int slot; search_key.objectid = objectid; - search_key.type = (u8)-1; + search_key.type = BTRFS_ROOT_ITEM_KEY; search_key.offset = (u64)-1; path = btrfs_alloc_path(); @@ -275,3 +275,77 @@ out: btrfs_free_path(path); return ret; } + +int btrfs_del_root_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *tree_root, + u64 root_id, u8 type, u64 ref_id) +{ + struct btrfs_key key; + int ret; + struct btrfs_path *path; + + path = btrfs_alloc_path(); + + key.objectid = root_id; + key.type = type; + key.offset = ref_id; + + ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1); + BUG_ON(ret); + + ret = btrfs_del_item(trans, tree_root, path); + BUG_ON(ret); + + btrfs_free_path(path); + return ret; +} + +/* + * add a btrfs_root_ref item. type is either BTRFS_ROOT_REF_KEY + * or BTRFS_ROOT_BACKREF_KEY. + * + * The dirid, sequence, name and name_len refer to the directory entry + * that is referencing the root. + * + * For a forward ref, the root_id is the id of the tree referencing + * the root and ref_id is the id of the subvol or snapshot. + * + * For a back ref the root_id is the id of the subvol or snapshot and + * ref_id is the id of the tree referencing it. + */ +int btrfs_add_root_ref(struct btrfs_trans_handle *trans, + struct btrfs_root *tree_root, + u64 root_id, u8 type, u64 ref_id, + u64 dirid, u64 sequence, + const char *name, int name_len) +{ + struct btrfs_key key; + int ret; + struct btrfs_path *path; + struct btrfs_root_ref *ref; + struct extent_buffer *leaf; + unsigned long ptr; + + + path = btrfs_alloc_path(); + + key.objectid = root_id; + key.type = type; + key.offset = ref_id; + + ret = btrfs_insert_empty_item(trans, tree_root, path, &key, + sizeof(*ref) + name_len); + BUG_ON(ret); + + leaf = path->nodes[0]; + ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref); + btrfs_set_root_ref_dirid(leaf, ref, dirid); + btrfs_set_root_ref_sequence(leaf, ref, sequence); + btrfs_set_root_ref_name_len(leaf, ref, name_len); + ptr = (unsigned long)(ref + 1); + write_extent_buffer(leaf, name, ptr, name_len); + btrfs_mark_buffer_dirty(leaf); + + btrfs_free_path(path); + return ret; +} diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 93f23a456a3..e9c8ebeedd7 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -831,28 +831,43 @@ static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info, struct btrfs_trans_handle *trans; struct inode *parent_inode; struct inode *inode; + struct btrfs_root *parent_root; parent_inode = pending->dentry->d_parent->d_inode; - trans = btrfs_start_transaction(BTRFS_I(parent_inode)->root, 1); + parent_root = BTRFS_I(parent_inode)->root; + trans = btrfs_start_transaction(parent_root, 1); /* * insert the directory item */ namelen = strlen(pending->name); ret = btrfs_set_inode_index(parent_inode, &index); - ret = btrfs_insert_dir_item(trans, - BTRFS_I(parent_inode)->root, + ret = btrfs_insert_dir_item(trans, parent_root, pending->name, namelen, parent_inode->i_ino, &pending->root_key, BTRFS_FT_DIR, index); if (ret) goto fail; -#if 0 - ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, - pending->name, strlen(pending->name), objectid, - root->fs_info->sb->s_root->d_inode->i_ino, 0); -#endif + + /* add the backref first */ + ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root, + pending->root_key.objectid, + BTRFS_ROOT_BACKREF_KEY, + parent_root->root_key.objectid, + parent_inode->i_ino, index, pending->name, + namelen); + + BUG_ON(ret); + + /* now add the forward ref */ + ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root, + parent_root->root_key.objectid, + BTRFS_ROOT_REF_KEY, + pending->root_key.objectid, + parent_inode->i_ino, index, pending->name, + namelen); + inode = btrfs_lookup_dentry(parent_inode, pending->dentry); d_instantiate(pending->dentry, inode); fail: -- cgit v1.2.3-70-g09d2 From d2c3f4f695edac4d75c1b3eb01a1d16072de63bb Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 19 Nov 2008 12:44:22 -0500 Subject: Btrfs: Avoid writeback stalls While building large bios in writepages, btrfs may end up waiting for other page writeback to finish if WB_SYNC_ALL is used. While it is waiting, the bio it is building has a number of pages with the writeback bit set and they aren't getting to the disk any time soon. This lowers the latencies of writeback in general by sending down the bio being built before waiting for other pages. The bio submission code tries to limit the total number of async bios in flight by waiting when we're over a certain number of async bios. But, the waits are happening while writepages is building bios, and this can easily lead to stalls and other problems for people calling wait_on_page_writeback. The current fix is to let the congestion tests take care of waiting. sync() and others make sure to drain the current async requests to make sure that everything that was pending when the sync was started really get to disk. The code would drain pending requests both before and after submitting a new request. But, if one of the requests is waiting for page writeback to finish, the draining waits might block that page writeback. This changes the draining code to only wait after submitting the bio being processed. Signed-off-by: Chris Mason --- fs/btrfs/disk-io.c | 12 +++--------- fs/btrfs/extent_io.c | 24 ++++++++++++++++++++---- fs/btrfs/inode.c | 7 ------- 3 files changed, 23 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 0e8d31274c9..8d03e4a3c4e 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -538,15 +538,9 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, async->work.flags = 0; async->bio_flags = bio_flags; - while(atomic_read(&fs_info->async_submit_draining) && - atomic_read(&fs_info->nr_async_submits)) { - wait_event(fs_info->async_submit_wait, - (atomic_read(&fs_info->nr_async_submits) == 0)); - } - atomic_inc(&fs_info->nr_async_submits); btrfs_queue_worker(&fs_info->workers, &async->work); - +#if 0 if (atomic_read(&fs_info->nr_async_submits) > limit) { wait_event_timeout(fs_info->async_submit_wait, (atomic_read(&fs_info->nr_async_submits) < limit), @@ -556,7 +550,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, (atomic_read(&fs_info->nr_async_bios) < limit), HZ/10); } - +#endif while(atomic_read(&fs_info->async_submit_draining) && atomic_read(&fs_info->nr_async_submits)) { wait_event(fs_info->async_submit_wait, @@ -1765,11 +1759,11 @@ struct btrfs_root *open_ctree(struct super_block *sb, ret = btrfs_cleanup_reloc_trees(tree_root); BUG_ON(ret); +read_fs_root: location.objectid = BTRFS_FS_TREE_OBJECTID; location.type = BTRFS_ROOT_ITEM_KEY; location.offset = (u64)-1; -read_fs_root: fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); if (!fs_info->fs_root) goto fail_cleaner; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 54d013c3bb8..a0f3804efe4 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2398,7 +2398,8 @@ update_nr_written: int extent_write_cache_pages(struct extent_io_tree *tree, struct address_space *mapping, struct writeback_control *wbc, - writepage_t writepage, void *data) + writepage_t writepage, void *data, + void (*flush_fn)(void *)) { struct backing_dev_info *bdi = mapping->backing_dev_info; int ret = 0; @@ -2460,8 +2461,10 @@ retry: continue; } - if (wbc->sync_mode != WB_SYNC_NONE) + if (wbc->sync_mode != WB_SYNC_NONE) { + flush_fn(data); wait_on_page_writeback(page); + } if (PageWriteback(page) || !clear_page_dirty_for_io(page)) { @@ -2498,6 +2501,15 @@ retry: } EXPORT_SYMBOL(extent_write_cache_pages); +static noinline void flush_write_bio(void *data) +{ + struct extent_page_data *epd = data; + if (epd->bio) { + submit_one_bio(WRITE, epd->bio, 0, 0); + epd->bio = NULL; + } +} + int extent_write_full_page(struct extent_io_tree *tree, struct page *page, get_extent_t *get_extent, struct writeback_control *wbc) @@ -2523,7 +2535,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, ret = __extent_writepage(page, wbc, &epd); extent_write_cache_pages(tree, mapping, &wbc_writepages, - __extent_writepage, &epd); + __extent_writepage, &epd, flush_write_bio); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0, 0); } @@ -2592,7 +2604,8 @@ int extent_writepages(struct extent_io_tree *tree, }; ret = extent_write_cache_pages(tree, mapping, wbc, - __extent_writepage, &epd); + __extent_writepage, &epd, + flush_write_bio); if (epd.bio) { submit_one_bio(WRITE, epd.bio, 0, 0); } @@ -3087,6 +3100,9 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree, for (i = 0; i < num_pages; i++) { page = extent_buffer_page(eb, i); + if (!set && !PageDirty(page)) + continue; + lock_page(page); if (i == 0) set_page_extent_head(page, eb->len); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6854bf41856..806caacff86 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -883,13 +883,6 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, async_cow->work.ordered_free = async_cow_free; async_cow->work.flags = 0; - while(atomic_read(&root->fs_info->async_submit_draining) && - atomic_read(&root->fs_info->async_delalloc_pages)) { - wait_event(root->fs_info->async_submit_wait, - (atomic_read(&root->fs_info->async_delalloc_pages) - == 0)); - } - nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >> PAGE_CACHE_SHIFT; atomic_add(nr_pages, &root->fs_info->async_delalloc_pages); -- cgit v1.2.3-70-g09d2 From 79683f2d685cfb6ef9c97c5194e3ce3319e80cac Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 19 Nov 2008 22:00:53 -0500 Subject: Btrfs: Use current_fsuid/gid This fixes compile problems with linux-next Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 806caacff86..2c77e0957f7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3422,8 +3422,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, if (objectid > root->highest_inode) root->highest_inode = objectid; - inode->i_uid = current->fsuid; - inode->i_gid = current->fsgid; + inode->i_uid = current_fsuid(); + inode->i_gid = current_fsgid(); inode->i_mode = mode; inode->i_ino = objectid; inode_set_bytes(inode, 0); -- cgit v1.2.3-70-g09d2 From 4b4e25f2a6ddb070bab7f7dd2bd2926fb8db9e04 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 20 Nov 2008 10:22:27 -0500 Subject: Btrfs: compat code fixes The btrfs git kernel trees is used to build a standalone tree for compiling against older kernels. This commit makes the standalone tree work with 2.6.27 Signed-off-by: Chris Mason --- fs/btrfs/compat.h | 1 + fs/btrfs/compression.c | 2 +- fs/btrfs/disk-io.c | 3 ++- fs/btrfs/extent-tree.c | 5 ++++- fs/btrfs/inode.c | 2 +- fs/btrfs/ioctl.c | 1 + fs/btrfs/super.c | 2 ++ fs/btrfs/volumes.c | 2 ++ 8 files changed, 14 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h index dd1defdbfa3..75e4426d6fb 100644 --- a/fs/btrfs/compat.h +++ b/fs/btrfs/compat.h @@ -25,6 +25,7 @@ static inline struct dentry *d_obtain_alias(struct inode *inode) # define __pagevec_lru_add_file __pagevec_lru_add # define open_bdev_exclusive open_bdev_excl # define close_bdev_exclusive(bdev, mode) close_bdev_excl(bdev) +typedef unsigned __bitwise__ fmode_t; #endif diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index df05f513e1e..4febe2eb0b8 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -34,13 +34,13 @@ #include #include #include +#include "compat.h" #include "ctree.h" #include "disk-io.h" #include "transaction.h" #include "btrfs_inode.h" #include "volumes.h" #include "ordered-data.h" -#include "compat.h" #include "compression.h" #include "extent_io.h" #include "extent_map.h" diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index c8dcb47b6d7..981652233f7 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -26,7 +26,8 @@ #include // for block_sync_page #include #include -# include +#include +#include "compat.h" #include "crc32c.h" #include "ctree.h" #include "disk-io.h" diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 62d49705d14..b33e0bfb99e 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -19,6 +19,8 @@ #include #include #include +#include +#include "compat.h" #include "hash.h" #include "crc32c.h" #include "ctree.h" @@ -900,6 +902,7 @@ static int noinline remove_extent_backref(struct btrfs_trans_handle *trans, return ret; } +#ifdef BIO_RW_DISCARD static void btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len) { @@ -909,7 +912,7 @@ static void btrfs_issue_discard(struct block_device *bdev, blkdev_issue_discard(bdev, start >> 9, len >> 9); #endif } - +#endif static int noinline free_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 2c77e0957f7..7a8ad4292f7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -38,6 +38,7 @@ #include #include #include +#include "compat.h" #include "ctree.h" #include "disk-io.h" #include "transaction.h" @@ -47,7 +48,6 @@ #include "volumes.h" #include "ordered-data.h" #include "xattr.h" -#include "compat.h" #include "tree-log.h" #include "ref-cache.h" #include "compression.h" diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 8828109fa58..f3d68457e66 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -41,6 +41,7 @@ #include #include #include +#include "compat.h" #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 77c5eff3e20..1975ea273dc 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -37,6 +37,8 @@ #include #include #include +#include +#include "compat.h" #include "ctree.h" #include "disk-io.h" #include "transaction.h" diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c3ee63f92a5..724ead54529 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -20,7 +20,9 @@ #include #include #include +#include #include +#include "compat.h" #include "ctree.h" #include "extent_map.h" #include "disk-io.h" -- cgit v1.2.3-70-g09d2 From ce397c0616de9bfb678569f1dfa655838a711594 Mon Sep 17 00:00:00 2001 From: Liu Hui Date: Mon, 1 Dec 2008 20:31:40 -0500 Subject: Btrfs: Fix cow semantic in run_delalloc_nocow() The file preallocation code reversed the logic to force nodatacow. This fixes it. --- fs/btrfs/inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7a8ad4292f7..b3d4078b69a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1114,10 +1114,10 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, if (btrfs_test_opt(root, NODATACOW) || btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 0, nr_written); + page_started, 1, nr_written); else if (btrfs_test_flag(inode, PREALLOC)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 1, nr_written); + page_started, 0, nr_written); else ret = cow_file_range_async(inode, locked_page, start, end, page_started, nr_written); -- cgit v1.2.3-70-g09d2 From b2950863c61bc24cf0f63bc05947d9d50663c4c0 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Dec 2008 09:54:17 -0500 Subject: Btrfs: make things static and include the right headers Shut up various sparse warnings about symbols that should be either static or have their declarations in scope. Signed-off-by: Christoph Hellwig --- fs/btrfs/ctree.c | 2 +- fs/btrfs/disk-io.c | 16 ++++++++-------- fs/btrfs/extent-tree.c | 12 ++++++------ fs/btrfs/extent_io.c | 35 +++++++++++++++-------------------- fs/btrfs/free-space-cache.c | 6 ++++-- fs/btrfs/inode-item.c | 2 +- fs/btrfs/inode.c | 26 +++++++++++++------------- fs/btrfs/ioctl.c | 14 +++++++------- fs/btrfs/root-tree.c | 2 ++ fs/btrfs/super.c | 2 +- fs/btrfs/tree-log.c | 5 +++-- fs/btrfs/volumes.c | 12 ++++++------ fs/btrfs/zlib.c | 1 + 13 files changed, 68 insertions(+), 67 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 71ef0a2e2da..a83cbdf1d8c 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -217,7 +217,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, * this uses that block instead of allocating a new one. btrfs_alloc_reserved_extent * is used to finish the allocation. */ -int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, +static int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index fb04665e500..8a2bcc7024f 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -93,9 +93,9 @@ struct async_submit_bio { * extents on the btree inode are pretty simple, there's one extent * that covers the entire device */ -struct extent_map *btree_get_extent(struct inode *inode, struct page *page, - size_t page_offset, u64 start, u64 len, - int create) +static struct extent_map *btree_get_extent(struct inode *inode, + struct page *page, size_t page_offset, u64 start, u64 len, + int create) { struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct extent_map *em; @@ -295,7 +295,7 @@ printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror * checksum a dirty tree block before IO. This has extra checks to make * sure we only fill in the checksum field in the first page of a multi-page block */ -int csum_dirty_buffer(struct btrfs_root *root, struct page *page) +static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { struct extent_io_tree *tree; u64 start = (u64)page->index << PAGE_CACHE_SHIFT; @@ -365,7 +365,7 @@ static int check_tree_block_fsid(struct btrfs_root *root, return ret; } -int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, +static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state) { struct extent_io_tree *tree; @@ -660,7 +660,7 @@ static int btree_writepages(struct address_space *mapping, return extent_writepages(tree, mapping, btree_get_extent, wbc); } -int btree_readpage(struct file *file, struct page *page) +static int btree_readpage(struct file *file, struct page *page) { struct extent_io_tree *tree; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1200,7 +1200,7 @@ static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) } } -void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) +static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) { struct inode *inode; struct extent_map_tree *em_tree; @@ -1842,7 +1842,7 @@ static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) put_bh(bh); } -int write_all_supers(struct btrfs_root *root) +static int write_all_supers(struct btrfs_root *root) { struct list_head *cur; struct list_head *head = &root->fs_info->fs_devices->devices; diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a970472eab1..d1563852938 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -74,7 +74,7 @@ static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) * this adds the block group to the fs_info rb tree for the block group * cache */ -int btrfs_add_block_group_cache(struct btrfs_fs_info *info, +static int btrfs_add_block_group_cache(struct btrfs_fs_info *info, struct btrfs_block_group_cache *block_group) { struct rb_node **p; @@ -289,7 +289,7 @@ err: /* * return the block group that starts at or after bytenr */ -struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct +static struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr) { @@ -3445,7 +3445,7 @@ static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, return 0; } -int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len, +static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len, u32 *refs) { int ret; @@ -5434,7 +5434,7 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) return flags; } -int __alloc_chunk_for_shrink(struct btrfs_root *root, +static int __alloc_chunk_for_shrink(struct btrfs_root *root, struct btrfs_block_group_cache *shrink_block_group, int force) { @@ -5703,8 +5703,8 @@ out: return ret; } -int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, - struct btrfs_key *key) +static int find_first_block_group(struct btrfs_root *root, + struct btrfs_path *path, struct btrfs_key *key) { int ret = 0; struct btrfs_key found_key; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index d79ccdbfdd9..c3dfe2a0ec8 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -112,7 +112,7 @@ void extent_io_tree_init(struct extent_io_tree *tree, } EXPORT_SYMBOL(extent_io_tree_init); -struct extent_state *alloc_extent_state(gfp_t mask) +static struct extent_state *alloc_extent_state(gfp_t mask) { struct extent_state *state; #ifdef LEAK_DEBUG @@ -136,7 +136,7 @@ struct extent_state *alloc_extent_state(gfp_t mask) } EXPORT_SYMBOL(alloc_extent_state); -void free_extent_state(struct extent_state *state) +static void free_extent_state(struct extent_state *state) { if (!state) return; @@ -662,7 +662,7 @@ static void set_state_bits(struct extent_io_tree *tree, * [start, end] is inclusive * This takes the tree lock. */ -int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, +static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, int exclusive, u64 *failed_start, gfp_t mask) { struct extent_state *state; @@ -879,12 +879,11 @@ int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_new); -int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, +static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask); } -EXPORT_SYMBOL(clear_extent_new); int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) @@ -894,27 +893,24 @@ int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_uptodate); -int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, +static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); } -EXPORT_SYMBOL(clear_extent_uptodate); -int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, +static int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return set_extent_bit(tree, start, end, EXTENT_WRITEBACK, 0, NULL, mask); } -EXPORT_SYMBOL(set_extent_writeback); -int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, +static int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); } -EXPORT_SYMBOL(clear_extent_writeback); int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) { @@ -994,7 +990,7 @@ EXPORT_SYMBOL(set_range_dirty); /* * helper function to set both pages and extents in the tree writeback */ -int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) +static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) { unsigned long index = start >> PAGE_CACHE_SHIFT; unsigned long end_index = end >> PAGE_CACHE_SHIFT; @@ -1010,7 +1006,6 @@ int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) set_extent_writeback(tree, start, end, GFP_NOFS); return 0; } -EXPORT_SYMBOL(set_range_writeback); /* * find the first offset in the io tree with 'bits' set. zero is @@ -1432,11 +1427,13 @@ out: spin_unlock_irq(&tree->lock); return total_bytes; } + +#if 0 /* * helper function to lock both pages and extents in the tree. * pages must be locked first. */ -int lock_range(struct extent_io_tree *tree, u64 start, u64 end) +static int lock_range(struct extent_io_tree *tree, u64 start, u64 end) { unsigned long index = start >> PAGE_CACHE_SHIFT; unsigned long end_index = end >> PAGE_CACHE_SHIFT; @@ -1473,12 +1470,11 @@ failed: } return err; } -EXPORT_SYMBOL(lock_range); /* * helper function to unlock both pages and extents in the tree. */ -int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) +static int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) { unsigned long index = start >> PAGE_CACHE_SHIFT; unsigned long end_index = end >> PAGE_CACHE_SHIFT; @@ -1493,7 +1489,7 @@ int unlock_range(struct extent_io_tree *tree, u64 start, u64 end) unlock_extent(tree, start, end, GFP_NOFS); return 0; } -EXPORT_SYMBOL(unlock_range); +#endif /* * set the private field for a given byte offset in the tree. If there isn't @@ -1956,7 +1952,7 @@ void set_page_extent_mapped(struct page *page) } EXPORT_SYMBOL(set_page_extent_mapped); -void set_page_extent_head(struct page *page, unsigned long len) +static void set_page_extent_head(struct page *page, unsigned long len) { set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); } @@ -2397,7 +2393,7 @@ update_nr_written: * WB_SYNC_ALL then we were called for data integrity and we must wait for * existing IO to complete. */ -int extent_write_cache_pages(struct extent_io_tree *tree, +static int extent_write_cache_pages(struct extent_io_tree *tree, struct address_space *mapping, struct writeback_control *wbc, writepage_t writepage, void *data, @@ -2502,7 +2498,6 @@ retry: } return ret; } -EXPORT_SYMBOL(extent_write_cache_pages); static noinline void flush_write_bio(void *data) { diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index f4926c0f3c8..09462adfbe3 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -443,7 +443,8 @@ void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group) mutex_unlock(&block_group->alloc_mutex); } -struct btrfs_free_space *btrfs_find_free_space_offset(struct +#if 0 +static struct btrfs_free_space *btrfs_find_free_space_offset(struct btrfs_block_group_cache *block_group, u64 offset, u64 bytes) @@ -458,7 +459,7 @@ struct btrfs_free_space *btrfs_find_free_space_offset(struct return ret; } -struct btrfs_free_space *btrfs_find_free_space_bytes(struct +static struct btrfs_free_space *btrfs_find_free_space_bytes(struct btrfs_block_group_cache *block_group, u64 offset, u64 bytes) @@ -472,6 +473,7 @@ struct btrfs_free_space *btrfs_find_free_space_bytes(struct return ret; } +#endif struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache *block_group, u64 offset, diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c index d93451c66ba..3d46fa1f29a 100644 --- a/fs/btrfs/inode-item.c +++ b/fs/btrfs/inode-item.c @@ -20,7 +20,7 @@ #include "disk-io.h" #include "transaction.h" -int find_name_in_backref(struct btrfs_path *path, const char * name, +static int find_name_in_backref(struct btrfs_path *path, const char *name, int name_len, struct btrfs_inode_ref **ref_ret) { struct extent_buffer *leaf; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index b3d4078b69a..bd58ba655a4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1130,7 +1130,7 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, * bytes in this file, and to maintain the list of inodes that * have pending delalloc work to be done. */ -int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, +static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { unsigned long flags; @@ -1151,7 +1151,7 @@ int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, /* * extent_io.c clear_bit_hook, see set_bit_hook for why */ -int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, +static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { @@ -1215,7 +1215,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, * At IO completion time the cums attached on the ordered extent record * are inserted into the btree */ -int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio, +static int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1234,7 +1234,7 @@ int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio, * At IO completion time the cums attached on the ordered extent record * are inserted into the btree */ -int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, +static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1245,7 +1245,7 @@ int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, * extent_io.c submission hook. This does the right thing for csum calculation on write, * or reading the csums from the tree before a read */ -int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, +static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1313,7 +1313,7 @@ struct btrfs_writepage_fixup { struct btrfs_work work; }; -void btrfs_writepage_fixup_worker(struct btrfs_work *work) +static void btrfs_writepage_fixup_worker(struct btrfs_work *work) { struct btrfs_writepage_fixup *fixup; struct btrfs_ordered_extent *ordered; @@ -1372,7 +1372,7 @@ out_page: * to fix it up. The async helper will wait for ordered extents, set * the delalloc bit and make it safe to write the page. */ -int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) +static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) { struct inode *inode = page->mapping->host; struct btrfs_writepage_fixup *fixup; @@ -1526,7 +1526,7 @@ nocow: return 0; } -int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, +static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state, int uptodate) { return btrfs_finish_ordered_io(page->mapping->host, start, end); @@ -1548,7 +1548,7 @@ struct io_failure_record { int last_mirror; }; -int btrfs_io_failed_hook(struct bio *failed_bio, +static int btrfs_io_failed_hook(struct bio *failed_bio, struct page *page, u64 start, u64 end, struct extent_state *state) { @@ -1642,7 +1642,7 @@ int btrfs_io_failed_hook(struct bio *failed_bio, * each time an IO finishes, we do a fast check in the IO failure tree * to see if we need to process or clean up an io_failure_record */ -int btrfs_clean_io_failures(struct inode *inode, u64 start) +static int btrfs_clean_io_failures(struct inode *inode, u64 start) { u64 private; u64 private_failure; @@ -1675,7 +1675,7 @@ int btrfs_clean_io_failures(struct inode *inode, u64 start) * if there's a match, we allow the bio to finish. If not, we go through * the io_failure_record routines to find good copies */ -int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, +static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state) { size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT); @@ -4362,8 +4362,8 @@ out: * Invalidate a single dcache entry at the root of the filesystem. * Needed after creation of snapshot or subvolume. */ -void btrfs_invalidate_dcache_root(struct inode *dir, char *name, - int namelen) +static void btrfs_invalidate_dcache_root(struct inode *dir, + char *name, int namelen) { struct dentry *alias, *entry; struct qstr qstr; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 35f650e183e..cc7c5161e26 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -354,7 +354,7 @@ out_unlock: } -int btrfs_defrag_file(struct file *file) +static int btrfs_defrag_file(struct file *file) { struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -649,7 +649,7 @@ static int btrfs_ioctl_defrag(struct file *file) return 0; } -long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) +static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg) { struct btrfs_ioctl_vol_args *vol_args; int ret; @@ -671,7 +671,7 @@ out: return ret; } -long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) +static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg) { struct btrfs_ioctl_vol_args *vol_args; int ret; @@ -696,8 +696,8 @@ out: return ret; } -long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, u64 off, - u64 olen, u64 destoff) +static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, + u64 off, u64 olen, u64 destoff) { struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; @@ -1035,7 +1035,7 @@ out_fput: return ret; } -long btrfs_ioctl_clone_range(struct file *file, unsigned long argptr) +static long btrfs_ioctl_clone_range(struct file *file, unsigned long argptr) { struct btrfs_ioctl_clone_range_args args; @@ -1051,7 +1051,7 @@ long btrfs_ioctl_clone_range(struct file *file, unsigned long argptr) * basically own the machine, and have a very in depth understanding * of all the possible deadlocks and enospc problems. */ -long btrfs_ioctl_trans_start(struct file *file) +static long btrfs_ioctl_trans_start(struct file *file) { struct inode *inode = fdentry(file)->d_inode; struct btrfs_root *root = BTRFS_I(inode)->root; diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index dbe20d4c6ea..f99335a999d 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -276,6 +276,7 @@ out: return ret; } +#if 0 /* this will get used when snapshot deletion is implemented */ int btrfs_del_root_ref(struct btrfs_trans_handle *trans, struct btrfs_root *tree_root, u64 root_id, u8 type, u64 ref_id) @@ -299,6 +300,7 @@ int btrfs_del_root_ref(struct btrfs_trans_handle *trans, btrfs_free_path(path); return ret; } +#endif int btrfs_find_root_ref(struct btrfs_root *tree_root, struct btrfs_path *path, diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 1975ea273dc..93a21c77064 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -647,7 +647,7 @@ static int btrfs_interface_init(void) return misc_register(&btrfs_misc); } -void btrfs_interface_exit(void) +static void btrfs_interface_exit(void) { if (misc_deregister(&btrfs_misc) < 0) printk("misc_deregister failed for control device"); diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index be4fc30a30e..4fcfc8b1189 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -23,6 +23,7 @@ #include "locking.h" #include "print-tree.h" #include "compat.h" +#include "tree-log.h" /* magic values for the inode_only field in btrfs_log_inode: * @@ -78,7 +79,7 @@ static int __btrfs_log_inode(struct btrfs_trans_handle *trans, * tree of log tree roots. This must be called with a tree log transaction * running (see start_log_trans). */ -int btrfs_add_log_tree(struct btrfs_trans_handle *trans, +static int btrfs_add_log_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root) { struct btrfs_key key; @@ -1934,7 +1935,7 @@ static int walk_log_tree(struct btrfs_trans_handle *trans, return ret; } -int wait_log_commit(struct btrfs_root *log) +static int wait_log_commit(struct btrfs_root *log) { DEFINE_WAIT(wait); u64 transid = log->fs_info->tree_log_transid; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 724ead54529..769f2c5d9e9 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -238,7 +238,7 @@ done: return 0; } -void pending_bios_fn(struct btrfs_work *work) +static void pending_bios_fn(struct btrfs_work *work) { struct btrfs_device *device; @@ -686,7 +686,7 @@ error: return ret; } -int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, +static int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, struct btrfs_device *device, u64 start) { @@ -1393,7 +1393,7 @@ error: goto out; } -int noinline btrfs_update_device(struct btrfs_trans_handle *trans, +static int noinline btrfs_update_device(struct btrfs_trans_handle *trans, struct btrfs_device *device) { int ret; @@ -1497,7 +1497,7 @@ static int btrfs_free_chunk(struct btrfs_trans_handle *trans, return 0; } -int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 +static int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 chunk_offset) { struct btrfs_super_block *super_copy = &root->fs_info->super_copy; @@ -1543,7 +1543,7 @@ int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 return ret; } -int btrfs_relocate_chunk(struct btrfs_root *root, +static int btrfs_relocate_chunk(struct btrfs_root *root, u64 chunk_tree, u64 chunk_objectid, u64 chunk_offset) { @@ -1884,7 +1884,7 @@ done: return ret; } -int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, +static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_key *key, struct btrfs_chunk *chunk, int item_size) diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c index 5b9f7002513..c4617cde6c7 100644 --- a/fs/btrfs/zlib.c +++ b/fs/btrfs/zlib.c @@ -30,6 +30,7 @@ #include #include #include +#include "compression.h" /* Plan: call deflate() with avail_in == *sourcelen, avail_out = *dstlen - 12 and flush == Z_FINISH. -- cgit v1.2.3-70-g09d2 From 4022abf4498186de2e893ce614333216f2add461 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 2 Dec 2008 09:57:03 -0500 Subject: Btrfs: delete unused function: btrfs_invalidate_dcache_root Snapshot and subvolume creation no longer need this helper. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 25 ------------------------- 1 file changed, 25 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bd58ba655a4..09efc9473a3 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4358,31 +4358,6 @@ out: btrfs_btree_balance_dirty(root, nr); } -/* - * Invalidate a single dcache entry at the root of the filesystem. - * Needed after creation of snapshot or subvolume. - */ -static void btrfs_invalidate_dcache_root(struct inode *dir, - char *name, int namelen) -{ - struct dentry *alias, *entry; - struct qstr qstr; - - alias = d_find_alias(dir); - if (alias) { - qstr.name = name; - qstr.len = namelen; - /* change me if btrfs ever gets a d_hash operation */ - qstr.hash = full_name_hash(qstr.name, qstr.len); - entry = d_lookup(alias, &qstr); - dput(alias); - if (entry) { - d_invalidate(entry); - dput(entry); - } - } -} - /* * create a new subvolume directory/inode (helper for the ioctl). */ -- cgit v1.2.3-70-g09d2 From d20f7043fa65659136c1a7c3c456eeeb5c6f431f Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 8 Dec 2008 16:58:54 -0500 Subject: Btrfs: move data checksumming into a dedicated tree Btrfs stores checksums for each data block. Until now, they have been stored in the subvolume trees, indexed by the inode that is referencing the data block. This means that when we read the inode, we've probably read in at least some checksums as well. But, this has a few problems: * The checksums are indexed by logical offset in the file. When compression is on, this means we have to do the expensive checksumming on the uncompressed data. It would be faster if we could checksum the compressed data instead. * If we implement encryption, we'll be checksumming the plain text and storing that on disk. This is significantly less secure. * For either compression or encryption, we have to get the plain text back before we can verify the checksum as correct. This makes the raid layer balancing and extent moving much more expensive. * It makes the front end caching code more complex, as we have touch the subvolume and inodes as we cache extents. * There is potentitally one copy of the checksum in each subvolume referencing an extent. The solution used here is to store the extent checksums in a dedicated tree. This allows us to index the checksums by phyiscal extent start and length. It means: * The checksum is against the data stored on disk, after any compression or encryption is done. * The checksum is stored in a central location, and can be verified without following back references, or reading inodes. This makes compression significantly faster by reducing the amount of data that needs to be checksummed. It will also allow much faster raid management code in general. The checksums are indexed by a key with a fixed objectid (a magic value in ctree.h) and offset set to the starting byte of the extent. This allows us to copy the checksum items into the fsync log tree directly (or any other tree), without having to invent a second format for them. Signed-off-by: Chris Mason --- fs/btrfs/compression.c | 124 ++++++++++++++++++++++++++++---- fs/btrfs/ctree.h | 30 +++++--- fs/btrfs/disk-io.c | 45 ++++++++++-- fs/btrfs/extent_io.c | 5 ++ fs/btrfs/file-item.c | 185 ++++++++++++++++++------------------------------ fs/btrfs/inode.c | 45 ++++++------ fs/btrfs/ioctl.c | 55 +------------- fs/btrfs/ordered-data.c | 7 +- fs/btrfs/ordered-data.h | 10 +-- fs/btrfs/tree-log.c | 121 +++++++++++++++++++++++++++---- fs/btrfs/volumes.c | 1 + 11 files changed, 387 insertions(+), 241 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 4febe2eb0b8..ad727413730 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -69,11 +69,27 @@ struct compressed_bio { /* IO errors */ int errors; + int mirror_num; /* for reads, this is the bio we are copying the data into */ struct bio *orig_bio; + + /* + * the start of a variable length array of checksums only + * used by reads + */ + u32 sums; }; +static inline int compressed_bio_size(struct btrfs_root *root, + unsigned long disk_size) +{ + u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy); + return sizeof(struct compressed_bio) + + ((disk_size + root->sectorsize - 1) / root->sectorsize) * + csum_size; +} + static struct bio *compressed_bio_alloc(struct block_device *bdev, u64 first_byte, gfp_t gfp_flags) { @@ -96,6 +112,47 @@ static struct bio *compressed_bio_alloc(struct block_device *bdev, return bio; } +static int check_compressed_csum(struct inode *inode, + struct compressed_bio *cb, + u64 disk_start) +{ + int ret; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct page *page; + unsigned long i; + char *kaddr; + u32 csum; + u32 *cb_sum = &cb->sums; + + if (btrfs_test_opt(root, NODATASUM) || + btrfs_test_flag(inode, NODATASUM)) + return 0; + + for (i = 0; i < cb->nr_pages; i++) { + page = cb->compressed_pages[i]; + csum = ~(u32)0; + + kaddr = kmap_atomic(page, KM_USER0); + csum = btrfs_csum_data(root, kaddr, csum, PAGE_CACHE_SIZE); + btrfs_csum_final(csum, (char *)&csum); + kunmap_atomic(kaddr, KM_USER0); + + if (csum != *cb_sum) { + printk("btrfs csum failed ino %lu extent %llu csum %u " + "wanted %u mirror %d\n", inode->i_ino, + (unsigned long long)disk_start, + csum, *cb_sum, cb->mirror_num); + ret = -EIO; + goto fail; + } + cb_sum++; + + } + ret = 0; +fail: + return ret; +} + /* when we finish reading compressed pages from the disk, we * decompress them and then run the bio end_io routines on the * decompressed pages (in the inode address space). @@ -124,16 +181,21 @@ static void end_compressed_bio_read(struct bio *bio, int err) if (!atomic_dec_and_test(&cb->pending_bios)) goto out; + inode = cb->inode; + ret = check_compressed_csum(inode, cb, (u64)bio->bi_sector << 9); + if (ret) + goto csum_failed; + /* ok, we're the last bio for this extent, lets start * the decompression. */ - inode = cb->inode; tree = &BTRFS_I(inode)->io_tree; ret = btrfs_zlib_decompress_biovec(cb->compressed_pages, cb->start, cb->orig_bio->bi_io_vec, cb->orig_bio->bi_vcnt, cb->compressed_len); +csum_failed: if (ret) cb->errors = 1; @@ -148,8 +210,21 @@ static void end_compressed_bio_read(struct bio *bio, int err) /* do io completion on the original bio */ if (cb->errors) { bio_io_error(cb->orig_bio); - } else + } else { + int bio_index = 0; + struct bio_vec *bvec = cb->orig_bio->bi_io_vec; + + /* + * we have verified the checksum already, set page + * checked so the end_io handlers know about it + */ + while(bio_index < cb->orig_bio->bi_vcnt) { + SetPageChecked(bvec->bv_page); + bvec++; + bio_index++; + } bio_endio(cb->orig_bio, 0); + } /* finally free the cb struct */ kfree(cb->compressed_pages); @@ -277,12 +352,13 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, int ret; WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1)); - cb = kmalloc(sizeof(*cb), GFP_NOFS); + cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS); atomic_set(&cb->pending_bios, 0); cb->errors = 0; cb->inode = inode; cb->start = start; cb->len = len; + cb->mirror_num = 0; cb->compressed_pages = compressed_pages; cb->compressed_len = compressed_len; cb->orig_bio = NULL; @@ -290,9 +366,6 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; - ret = btrfs_csum_file_bytes(root, inode, start, len); - BUG_ON(ret); - bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS); bio->bi_private = cb; bio->bi_end_io = end_compressed_bio_write; @@ -325,6 +398,9 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); + ret = btrfs_csum_one_bio(root, inode, bio, start, 1); + BUG_ON(ret); + ret = btrfs_map_bio(root, WRITE, bio, 0, 1); BUG_ON(ret); @@ -348,6 +424,9 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start, ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); + ret = btrfs_csum_one_bio(root, inode, bio, start, 1); + BUG_ON(ret); + ret = btrfs_map_bio(root, WRITE, bio, 0, 1); BUG_ON(ret); @@ -510,6 +589,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, u64 em_start; struct extent_map *em; int ret; + u32 *sums; tree = &BTRFS_I(inode)->io_tree; em_tree = &BTRFS_I(inode)->extent_tree; @@ -521,15 +601,18 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, PAGE_CACHE_SIZE); spin_unlock(&em_tree->lock); - cb = kmalloc(sizeof(*cb), GFP_NOFS); + compressed_len = em->block_len; + cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS); atomic_set(&cb->pending_bios, 0); cb->errors = 0; cb->inode = inode; + cb->mirror_num = mirror_num; + sums = &cb->sums; cb->start = em->orig_start; - compressed_len = em->block_len; em_len = em->len; em_start = em->start; + free_extent_map(em); em = NULL; @@ -551,11 +634,6 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, add_ra_bio_pages(inode, em_start + em_len, cb); - if (!btrfs_test_opt(root, NODATASUM) && - !btrfs_test_flag(inode, NODATASUM)) { - btrfs_lookup_bio_sums(root, inode, cb->orig_bio); - } - /* include any pages we added in add_ra-bio_pages */ uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE; cb->len = uncompressed_len; @@ -568,6 +646,8 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, for (page_index = 0; page_index < nr_pages; page_index++) { page = cb->compressed_pages[page_index]; page->mapping = inode->i_mapping; + page->index = em_start >> PAGE_CACHE_SHIFT; + if (comp_bio->bi_size) ret = tree->ops->merge_bio_hook(page, 0, PAGE_CACHE_SIZE, @@ -591,7 +671,16 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, */ atomic_inc(&cb->pending_bios); - ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); + if (!btrfs_test_opt(root, NODATASUM) && + !btrfs_test_flag(inode, NODATASUM)) { + btrfs_lookup_bio_sums(root, inode, comp_bio, + sums); + } + sums += (comp_bio->bi_size + root->sectorsize - 1) / + root->sectorsize; + + ret = btrfs_map_bio(root, READ, comp_bio, + mirror_num, 0); BUG_ON(ret); bio_put(comp_bio); @@ -610,7 +699,12 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); BUG_ON(ret); - ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); + if (!btrfs_test_opt(root, NODATASUM) && + !btrfs_test_flag(inode, NODATASUM)) { + btrfs_lookup_bio_sums(root, inode, comp_bio, sums); + } + + ret = btrfs_map_bio(root, READ, comp_bio, mirror_num, 0); BUG_ON(ret); bio_put(comp_bio); diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 96f2ec7ad5b..242b961ae6d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -73,6 +73,9 @@ struct btrfs_ordered_sum; /* directory objectid inside the root tree */ #define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL +/* holds checksums of all the data extents */ +#define BTRFS_CSUM_TREE_OBJECTID 7ULL + /* orhpan objectid for tracking unlinked/truncated files */ #define BTRFS_ORPHAN_OBJECTID -5ULL @@ -84,6 +87,13 @@ struct btrfs_ordered_sum; #define BTRFS_TREE_RELOC_OBJECTID -8ULL #define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL +/* + * extent checksums all have this objectid + * this allows them to share the logging tree + * for fsyncs + */ +#define BTRFS_EXTENT_CSUM_OBJECTID -10ULL + /* dummy objectid represents multiple objectids */ #define BTRFS_MULTIPLE_OBJECTIDS -255ULL @@ -634,6 +644,7 @@ struct btrfs_fs_info { struct btrfs_root *chunk_root; struct btrfs_root *dev_root; struct btrfs_root *fs_root; + struct btrfs_root *csum_root; /* the log root tree is a directory of all the other log roots */ struct btrfs_root *log_root_tree; @@ -716,6 +727,7 @@ struct btrfs_fs_info { struct btrfs_workers workers; struct btrfs_workers delalloc_workers; struct btrfs_workers endio_workers; + struct btrfs_workers endio_meta_workers; struct btrfs_workers endio_write_workers; struct btrfs_workers submit_workers; /* @@ -858,13 +870,12 @@ struct btrfs_root { * extent data is for file data */ #define BTRFS_EXTENT_DATA_KEY 108 + /* - * csum items have the checksums for data in the extents + * extent csums are stored in a separate tree and hold csums for + * an entire extent on disk. */ -#define BTRFS_CSUM_ITEM_KEY 120 - - -/* reserve 21-31 for other file/dir stuff */ +#define BTRFS_EXTENT_CSUM_KEY 128 /* * root items point to tree roots. There are typically in the root @@ -1917,7 +1928,7 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root /* file-item.c */ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, - struct bio *bio); + struct bio *bio, u32 *dst); int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, @@ -1929,17 +1940,16 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, struct btrfs_path *path, u64 objectid, u64 bytenr, int mod); int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct inode *inode, + struct btrfs_root *root, struct btrfs_ordered_sum *sums); int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, - struct bio *bio); + struct bio *bio, u64 file_start, int contig); int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode, u64 start, unsigned long len); struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, - u64 objectid, u64 offset, - int cow); + u64 bytenr, int cow); int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3eb7c2576fe..61dc3b2c834 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -445,11 +445,18 @@ static void end_workqueue_bio(struct bio *bio, int err) end_io_wq->error = err; end_io_wq->work.func = end_workqueue_fn; end_io_wq->work.flags = 0; - if (bio->bi_rw & (1 << BIO_RW)) + + if (bio->bi_rw & (1 << BIO_RW)) { btrfs_queue_worker(&fs_info->endio_write_workers, &end_io_wq->work); - else - btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); + } else { + if (end_io_wq->metadata) + btrfs_queue_worker(&fs_info->endio_meta_workers, + &end_io_wq->work); + else + btrfs_queue_worker(&fs_info->endio_workers, + &end_io_wq->work); + } } int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, @@ -1208,6 +1215,9 @@ static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) info = (struct btrfs_fs_info *)bdi->unplug_io_data; list_for_each(cur, &info->fs_devices->devices) { device = list_entry(cur, struct btrfs_device, dev_list); + if (!device->bdev) + continue; + bdi = blk_get_backing_dev_info(device->bdev); if (bdi->unplug_io_fn) { bdi->unplug_io_fn(bdi, page); @@ -1344,7 +1354,7 @@ static void end_workqueue_fn(struct btrfs_work *work) * blocksize <= pagesize, it is basically a noop */ if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { - btrfs_queue_worker(&fs_info->endio_workers, + btrfs_queue_worker(&fs_info->endio_meta_workers, &end_io_wq->work); return; } @@ -1454,6 +1464,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, struct buffer_head *bh; struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); + struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), + GFP_NOFS); struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), @@ -1470,7 +1482,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, struct btrfs_super_block *disk_super; if (!extent_root || !tree_root || !fs_info || - !chunk_root || !dev_root) { + !chunk_root || !dev_root || !csum_root) { err = -ENOMEM; goto fail; } @@ -1487,6 +1499,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, init_completion(&fs_info->kobj_unregister); fs_info->tree_root = tree_root; fs_info->extent_root = extent_root; + fs_info->csum_root = csum_root; fs_info->chunk_root = chunk_root; fs_info->dev_root = dev_root; fs_info->fs_devices = fs_devices; @@ -1652,6 +1665,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); btrfs_init_workers(&fs_info->endio_workers, "endio", fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", + fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", fs_info->thread_pool_size); @@ -1667,6 +1682,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_start_workers(&fs_info->delalloc_workers, 1); btrfs_start_workers(&fs_info->fixup_workers, 1); btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); + btrfs_start_workers(&fs_info->endio_meta_workers, + fs_info->thread_pool_size); btrfs_start_workers(&fs_info->endio_write_workers, fs_info->thread_pool_size); @@ -1751,6 +1768,13 @@ struct btrfs_root *open_ctree(struct super_block *sb, if (ret) goto fail_extent_root; + ret = find_and_setup_root(tree_root, fs_info, + BTRFS_CSUM_TREE_OBJECTID, csum_root); + if (ret) + goto fail_extent_root; + + csum_root->track_dirty = 1; + btrfs_read_block_groups(extent_root); fs_info->generation = generation + 1; @@ -1761,7 +1785,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, "btrfs-cleaner"); if (!fs_info->cleaner_kthread) - goto fail_extent_root; + goto fail_csum_root; fs_info->transaction_kthread = kthread_run(transaction_kthread, tree_root, @@ -1825,6 +1849,8 @@ fail_cleaner: filemap_write_and_wait(fs_info->btree_inode->i_mapping); invalidate_inode_pages2(fs_info->btree_inode->i_mapping); +fail_csum_root: + free_extent_buffer(csum_root->node); fail_extent_root: free_extent_buffer(extent_root->node); fail_tree_root: @@ -1838,6 +1864,7 @@ fail_sb_buffer: btrfs_stop_workers(&fs_info->delalloc_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); + btrfs_stop_workers(&fs_info->endio_meta_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); fail_iput: @@ -1853,6 +1880,7 @@ fail: kfree(fs_info); kfree(chunk_root); kfree(dev_root); + kfree(csum_root); return ERR_PTR(err); } @@ -2131,6 +2159,9 @@ int close_ctree(struct btrfs_root *root) if (root->fs_info->dev_root->node); free_extent_buffer(root->fs_info->dev_root->node); + if (root->fs_info->csum_root->node); + free_extent_buffer(root->fs_info->csum_root->node); + btrfs_free_block_groups(root->fs_info); del_fs_roots(fs_info); @@ -2141,6 +2172,7 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->delalloc_workers); btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); + btrfs_stop_workers(&fs_info->endio_meta_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); @@ -2163,6 +2195,7 @@ int close_ctree(struct btrfs_root *root) kfree(fs_info->tree_root); kfree(fs_info->chunk_root); kfree(fs_info->dev_root); + kfree(fs_info->csum_root); return 0; } diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index c3dfe2a0ec8..7449ecf32c5 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1732,6 +1732,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err) int whole_page; int ret; + if (err) + uptodate = 0; + do { struct page *page = bvec->bv_page; tree = &BTRFS_I(page->mapping->host)->io_tree; @@ -1761,6 +1764,8 @@ static void end_bio_extent_readpage(struct bio *bio, int err) if (ret == 0) { uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); + if (err) + uptodate = 0; continue; } } diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 234ed441736..a3ad2ce0011 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -74,8 +74,7 @@ out: struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, - u64 objectid, u64 offset, - int cow) + u64 bytenr, int cow) { int ret; struct btrfs_key file_key; @@ -87,9 +86,9 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, btrfs_super_csum_size(&root->fs_info->super_copy); int csums_in_item; - file_key.objectid = objectid; - file_key.offset = offset; - btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); + file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID; + file_key.offset = bytenr; + btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY); ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow); if (ret < 0) goto fail; @@ -100,11 +99,10 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, goto fail; path->slots[0]--; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || - found_key.objectid != objectid) { + if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY) goto fail; - } - csum_offset = (offset - found_key.offset) >> + + csum_offset = (bytenr - found_key.offset) >> root->fs_info->sb->s_blocksize_bits; csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]); csums_in_item /= csum_size; @@ -143,7 +141,7 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, } int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, - struct bio *bio) + struct bio *bio, u32 *dst) { u32 sum; struct bio_vec *bvec = bio->bi_io_vec; @@ -151,6 +149,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, u64 offset; u64 item_start_offset = 0; u64 item_last_offset = 0; + u64 disk_bytenr; u32 diff; u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy); @@ -165,21 +164,22 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, WARN_ON(bio->bi_vcnt <= 0); + disk_bytenr = (u64)bio->bi_sector << 9; while(bio_index < bio->bi_vcnt) { offset = page_offset(bvec->bv_page) + bvec->bv_offset; - ret = btrfs_find_ordered_sum(inode, offset, &sum); + ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum); if (ret == 0) goto found; - if (!item || offset < item_start_offset || - offset >= item_last_offset) { + if (!item || disk_bytenr < item_start_offset || + disk_bytenr >= item_last_offset) { struct btrfs_key found_key; u32 item_size; if (item) btrfs_release_path(root, path); - item = btrfs_lookup_csum(NULL, root, path, - inode->i_ino, offset, 0); + item = btrfs_lookup_csum(NULL, root->fs_info->csum_root, + path, disk_bytenr, 0); if (IS_ERR(item)) { ret = PTR_ERR(item); if (ret == -ENOENT || ret == -EFBIG) @@ -208,7 +208,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, * this byte range must be able to fit inside * a single leaf so it will also fit inside a u32 */ - diff = offset - item_start_offset; + diff = disk_bytenr - item_start_offset; diff = diff / root->sectorsize; diff = diff * csum_size; @@ -216,7 +216,11 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, ((unsigned long)item) + diff, csum_size); found: - set_state_private(io_tree, offset, sum); + if (dst) + *dst++ = sum; + else + set_state_private(io_tree, offset, sum); + disk_bytenr += bvec->bv_len; bio_index++; bvec++; } @@ -224,75 +228,8 @@ found: return 0; } -int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode, - u64 start, unsigned long len) -{ - struct btrfs_ordered_sum *sums; - struct btrfs_sector_sum *sector_sum; - struct btrfs_ordered_extent *ordered; - char *data; - struct page *page; - unsigned long total_bytes = 0; - unsigned long this_sum_bytes = 0; - - sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS); - if (!sums) - return -ENOMEM; - - sector_sum = sums->sums; - sums->file_offset = start; - sums->len = len; - INIT_LIST_HEAD(&sums->list); - ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset); - BUG_ON(!ordered); - - while(len > 0) { - if (start >= ordered->file_offset + ordered->len || - start < ordered->file_offset) { - sums->len = this_sum_bytes; - this_sum_bytes = 0; - btrfs_add_ordered_sum(inode, ordered, sums); - btrfs_put_ordered_extent(ordered); - - sums = kzalloc(btrfs_ordered_sum_size(root, len), - GFP_NOFS); - BUG_ON(!sums); - sector_sum = sums->sums; - sums->len = len; - sums->file_offset = start; - ordered = btrfs_lookup_ordered_extent(inode, - sums->file_offset); - BUG_ON(!ordered); - } - - page = find_get_page(inode->i_mapping, - start >> PAGE_CACHE_SHIFT); - - data = kmap_atomic(page, KM_USER0); - sector_sum->sum = ~(u32)0; - sector_sum->sum = btrfs_csum_data(root, data, sector_sum->sum, - PAGE_CACHE_SIZE); - kunmap_atomic(data, KM_USER0); - btrfs_csum_final(sector_sum->sum, - (char *)§or_sum->sum); - sector_sum->offset = page_offset(page); - page_cache_release(page); - - sector_sum++; - total_bytes += PAGE_CACHE_SIZE; - this_sum_bytes += PAGE_CACHE_SIZE; - start += PAGE_CACHE_SIZE; - - WARN_ON(len < PAGE_CACHE_SIZE); - len -= PAGE_CACHE_SIZE; - } - btrfs_add_ordered_sum(inode, ordered, sums); - btrfs_put_ordered_extent(ordered); - return 0; -} - int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, - struct bio *bio) + struct bio *bio, u64 file_start, int contig) { struct btrfs_ordered_sum *sums; struct btrfs_sector_sum *sector_sum; @@ -303,6 +240,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, unsigned long total_bytes = 0; unsigned long this_sum_bytes = 0; u64 offset; + u64 disk_bytenr; WARN_ON(bio->bi_vcnt <= 0); sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS); @@ -310,16 +248,25 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, return -ENOMEM; sector_sum = sums->sums; - sums->file_offset = page_offset(bvec->bv_page) + bvec->bv_offset; + disk_bytenr = (u64)bio->bi_sector << 9; sums->len = bio->bi_size; INIT_LIST_HEAD(&sums->list); - ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset); + + if (contig) + offset = file_start; + else + offset = page_offset(bvec->bv_page) + bvec->bv_offset; + + ordered = btrfs_lookup_ordered_extent(inode, offset); BUG_ON(!ordered); + sums->bytenr = ordered->start; while(bio_index < bio->bi_vcnt) { - offset = page_offset(bvec->bv_page) + bvec->bv_offset; - if (offset >= ordered->file_offset + ordered->len || - offset < ordered->file_offset) { + if (!contig) + offset = page_offset(bvec->bv_page) + bvec->bv_offset; + + if (!contig && (offset >= ordered->file_offset + ordered->len || + offset < ordered->file_offset)) { unsigned long bytes_left; sums->len = this_sum_bytes; this_sum_bytes = 0; @@ -333,10 +280,9 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, BUG_ON(!sums); sector_sum = sums->sums; sums->len = bytes_left; - sums->file_offset = offset; - ordered = btrfs_lookup_ordered_extent(inode, - sums->file_offset); + ordered = btrfs_lookup_ordered_extent(inode, offset); BUG_ON(!ordered); + sums->bytenr = ordered->start; } data = kmap_atomic(bvec->bv_page, KM_USER0); @@ -348,13 +294,14 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, kunmap_atomic(data, KM_USER0); btrfs_csum_final(sector_sum->sum, (char *)§or_sum->sum); - sector_sum->offset = page_offset(bvec->bv_page) + - bvec->bv_offset; + sector_sum->bytenr = disk_bytenr; sector_sum++; bio_index++; total_bytes += bvec->bv_len; this_sum_bytes += bvec->bv_len; + disk_bytenr += bvec->bv_len; + offset += bvec->bv_len; bvec++; } this_sum_bytes = 0; @@ -364,11 +311,10 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, } int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct inode *inode, + struct btrfs_root *root, struct btrfs_ordered_sum *sums) { - u64 objectid = inode->i_ino; - u64 offset; + u64 bytenr; int ret; struct btrfs_key file_key; struct btrfs_key found_key; @@ -396,13 +342,12 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, again: next_offset = (u64)-1; found_next = 0; - offset = sector_sum->offset; - file_key.objectid = objectid; - file_key.offset = offset; - btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); + file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID; + file_key.offset = sector_sum->bytenr; + bytenr = sector_sum->bytenr; + btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY); - mutex_lock(&BTRFS_I(inode)->csum_mutex); - item = btrfs_lookup_csum(trans, root, path, objectid, offset, 1); + item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1); if (!IS_ERR(item)) { leaf = path->nodes[0]; ret = 0; @@ -432,8 +377,8 @@ again: slot = 0; } btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot); - if (found_key.objectid != objectid || - found_key.type != BTRFS_CSUM_ITEM_KEY) { + if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID || + found_key.type != BTRFS_EXTENT_CSUM_KEY) { found_next = 1; goto insert; } @@ -460,10 +405,10 @@ again: path->slots[0]--; leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - csum_offset = (offset - found_key.offset) >> + csum_offset = (bytenr - found_key.offset) >> root->fs_info->sb->s_blocksize_bits; - if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || - found_key.objectid != objectid || + if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY || + found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID || csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) { goto insert; } @@ -482,8 +427,18 @@ insert: btrfs_release_path(root, path); csum_offset = 0; if (found_next) { - u64 tmp = min((u64)i_size_read(inode), next_offset); - tmp -= offset & ~((u64)root->sectorsize -1); + u64 tmp = total_bytes + root->sectorsize; + u64 next_sector = sector_sum->bytenr; + struct btrfs_sector_sum *next = sector_sum + 1; + + while(tmp < sums->len) { + if (next_sector + root->sectorsize != next->bytenr) + break; + tmp += root->sectorsize; + next_sector = next->bytenr; + next++; + } + tmp = min(tmp, next_offset - file_key.offset); tmp >>= root->fs_info->sb->s_blocksize_bits; tmp = max((u64)1, tmp); tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size)); @@ -510,7 +465,6 @@ found: item_end = (struct btrfs_csum_item *)((unsigned char *)item_end + btrfs_item_size_nr(leaf, path->slots[0])); eb_token = NULL; - mutex_unlock(&BTRFS_I(inode)->csum_mutex); cond_resched(); next_sector: @@ -541,9 +495,9 @@ next_sector: if (total_bytes < sums->len) { item = (struct btrfs_csum_item *)((char *)item + csum_size); - if (item < item_end && offset + PAGE_CACHE_SIZE == - sector_sum->offset) { - offset = sector_sum->offset; + if (item < item_end && bytenr + PAGE_CACHE_SIZE == + sector_sum->bytenr) { + bytenr = sector_sum->bytenr; goto next_sector; } } @@ -562,7 +516,6 @@ out: return ret; fail_unlock: - mutex_unlock(&BTRFS_I(inode)->csum_mutex); goto out; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 09efc9473a3..c03d847b8c4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1221,7 +1221,7 @@ static int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; - ret = btrfs_csum_one_bio(root, inode, bio); + ret = btrfs_csum_one_bio(root, inode, bio, 0, 0); BUG_ON(ret); return 0; } @@ -1259,12 +1259,11 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, btrfs_test_flag(inode, NODATASUM); if (!(rw & (1 << BIO_RW))) { - - if (bio_flags & EXTENT_BIO_COMPRESSED) + if (bio_flags & EXTENT_BIO_COMPRESSED) { return btrfs_submit_compressed_read(inode, bio, mirror_num, bio_flags); - else if (!skip_sum) - btrfs_lookup_bio_sums(root, inode, bio); + } else if (!skip_sum) + btrfs_lookup_bio_sums(root, inode, bio, NULL); goto mapit; } else if (!skip_sum) { /* we're doing a write, do the async checksumming */ @@ -1292,8 +1291,8 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, btrfs_set_trans_block_group(trans, inode); list_for_each(cur, list) { sum = list_entry(cur, struct btrfs_ordered_sum, list); - btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, - inode, sum); + btrfs_csum_file_blocks(trans, + BTRFS_I(inode)->root->fs_info->csum_root, sum); } return 0; } @@ -1545,6 +1544,7 @@ struct io_failure_record { u64 start; u64 len; u64 logical; + unsigned long bio_flags; int last_mirror; }; @@ -1563,7 +1563,6 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, int ret; int rw; u64 logical; - unsigned long bio_flags = 0; ret = get_state_private(failure_tree, start, &private); if (ret) { @@ -1573,6 +1572,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, failrec->start = start; failrec->len = end - start + 1; failrec->last_mirror = 0; + failrec->bio_flags = 0; spin_lock(&em_tree->lock); em = lookup_extent_mapping(em_tree, start, failrec->len); @@ -1588,8 +1588,10 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, } logical = start - em->start; logical = em->block_start + logical; - if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) - bio_flags = EXTENT_BIO_COMPRESSED; + if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) { + logical = em->block_start; + failrec->bio_flags = EXTENT_BIO_COMPRESSED; + } failrec->logical = logical; free_extent_map(em); set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | @@ -1626,6 +1628,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, bio->bi_sector = failrec->logical >> 9; bio->bi_bdev = failed_bio->bi_bdev; bio->bi_size = 0; + bio_add_page(bio, page, failrec->len, start - page_offset(page)); if (failed_bio->bi_rw & (1 << BIO_RW)) rw = WRITE; @@ -1634,7 +1637,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio, failrec->last_mirror, - bio_flags); + failrec->bio_flags); return 0; } @@ -1688,9 +1691,14 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, u32 csum = ~(u32)0; unsigned long flags; + if (PageChecked(page)) { + ClearPageChecked(page); + goto good; + } if (btrfs_test_opt(root, NODATASUM) || btrfs_test_flag(inode, NODATASUM)) return 0; + if (state && state->start == start) { private = state->private; ret = 0; @@ -1709,7 +1717,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, } kunmap_atomic(kaddr, KM_IRQ0); local_irq_restore(flags); - +good: /* if the io failure tree for this inode is non-empty, * check to see if we've recovered from a failed IO */ @@ -2243,6 +2251,7 @@ fail: return err; } +#if 0 /* * when truncating bytes in a file, it is possible to avoid reading * the leaves that contain only checksum items. This can be the @@ -2410,6 +2419,8 @@ out: return ret; } +#endif + /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find @@ -2459,9 +2470,6 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, btrfs_init_path(path); - ret = drop_csum_leaves(trans, root, path, inode, new_size); - BUG_ON(ret); - search_again: ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) { @@ -2509,16 +2517,11 @@ search_again: } item_end--; } - if (found_type == BTRFS_CSUM_ITEM_KEY) { - ret = btrfs_csum_truncate(trans, root, path, - new_size); - BUG_ON(ret); - } if (item_end < new_size) { if (found_type == BTRFS_DIR_ITEM_KEY) { found_type = BTRFS_INODE_ITEM_KEY; } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { - found_type = BTRFS_CSUM_ITEM_KEY; + found_type = BTRFS_EXTENT_DATA_KEY; } else if (found_type == BTRFS_EXTENT_DATA_KEY) { found_type = BTRFS_XATTR_ITEM_KEY; } else if (found_type == BTRFS_XATTR_ITEM_KEY) { diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index b4da53d55c8..6228b69c2b9 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -714,8 +714,7 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, u64 len = olen; u64 bs = root->fs_info->sb->s_blocksize; u64 hint_byte; - u16 csum_size = - btrfs_super_csum_size(&root->fs_info->super_copy); + /* * TODO: * - split compressed inline extents. annoying: we need to @@ -833,7 +832,7 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, slot = path->slots[0]; btrfs_item_key_to_cpu(leaf, &key, slot); - if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || + if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY || key.objectid != src->i_ino) break; @@ -958,56 +957,6 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, btrfs_mark_buffer_dirty(leaf); } - if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) { - u32 size; - struct btrfs_key new_key; - u64 coverslen; - int coff, clen; - - size = btrfs_item_size_nr(leaf, slot); - coverslen = (size / csum_size) << - root->fs_info->sb->s_blocksize_bits; - printk("csums for %llu~%llu\n", - key.offset, coverslen); - if (key.offset + coverslen < off || - key.offset >= off+len) - goto next; - - read_extent_buffer(leaf, buf, - btrfs_item_ptr_offset(leaf, slot), - size); - btrfs_release_path(root, path); - - coff = 0; - if (off > key.offset) - coff = ((off - key.offset) >> - root->fs_info->sb->s_blocksize_bits) * - csum_size; - clen = size - coff; - if (key.offset + coverslen > off+len) - clen -= ((key.offset+coverslen-off-len) >> - root->fs_info->sb->s_blocksize_bits) * - csum_size; - printk(" will dup %d~%d of %d\n", - coff, clen, size); - - memcpy(&new_key, &key, sizeof(new_key)); - new_key.objectid = inode->i_ino; - new_key.offset = key.offset + destoff - off; - - ret = btrfs_insert_empty_item(trans, root, path, - &new_key, clen); - if (ret) - goto out; - - leaf = path->nodes[0]; - slot = path->slots[0]; - write_extent_buffer(leaf, buf + coff, - btrfs_item_ptr_offset(leaf, slot), - clen); - btrfs_mark_buffer_dirty(leaf); - } - next: btrfs_release_path(root, path); key.offset++; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 027ad6b3839..d9e232227da 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -610,7 +610,8 @@ out: * try to find a checksum. This is used because we allow pages to * be reclaimed before their checksum is actually put into the btree */ -int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) +int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, + u32 *sum) { struct btrfs_ordered_sum *ordered_sum; struct btrfs_sector_sum *sector_sums; @@ -629,11 +630,11 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) mutex_lock(&tree->mutex); list_for_each_prev(cur, &ordered->list) { ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); - if (offset >= ordered_sum->file_offset) { + if (disk_bytenr >= ordered_sum->bytenr) { num_sectors = ordered_sum->len / sectorsize; sector_sums = ordered_sum->sums; for (i = 0; i < num_sectors; i++) { - if (sector_sums[i].offset == offset) { + if (sector_sums[i].bytenr == disk_bytenr) { *sum = sector_sums[i].sum; ret = 0; goto out; diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 260bf95dfe0..ab66d5e8d6d 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -33,15 +33,17 @@ struct btrfs_ordered_inode_tree { * the ordered extent are on disk */ struct btrfs_sector_sum { - u64 offset; + /* bytenr on disk */ + u64 bytenr; u32 sum; }; struct btrfs_ordered_sum { - u64 file_offset; + /* bytenr is the start of this extent on disk */ + u64 bytenr; + /* * this is the length in bytes covered by the sums array below. - * But, the sums array may not be contiguous in the file. */ unsigned long len; struct list_head list; @@ -147,7 +149,7 @@ struct btrfs_ordered_extent * btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); int btrfs_ordered_update_i_size(struct inode *inode, struct btrfs_ordered_extent *ordered); -int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum); +int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, u32 *sum); int btrfs_wait_on_page_writeback_range(struct address_space *mapping, pgoff_t start, pgoff_t end); int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index c766649ad45..08469ec0585 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -934,24 +934,17 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans, unsigned long file_bytes; struct btrfs_ordered_sum *sums; struct btrfs_sector_sum *sector_sum; - struct inode *inode; unsigned long ptr; file_bytes = (item_size / csum_size) * root->sectorsize; - inode = read_one_inode(root, key->objectid); - if (!inode) { - return -EIO; - } - sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS); if (!sums) { - iput(inode); return -ENOMEM; } INIT_LIST_HEAD(&sums->list); sums->len = file_bytes; - sums->file_offset = key->offset; + sums->bytenr = key->offset; /* * copy all the sums into the ordered sum struct @@ -960,7 +953,7 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans, cur_offset = key->offset; ptr = btrfs_item_ptr_offset(eb, slot); while(item_size > 0) { - sector_sum->offset = cur_offset; + sector_sum->bytenr = cur_offset; read_extent_buffer(eb, §or_sum->sum, ptr, csum_size); sector_sum++; item_size -= csum_size; @@ -969,11 +962,9 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans, } /* let btrfs_csum_file_blocks add them into the file */ - ret = btrfs_csum_file_blocks(trans, root, inode, sums); + ret = btrfs_csum_file_blocks(trans, root->fs_info->csum_root, sums); BUG_ON(ret); kfree(sums); - iput(inode); - return 0; } /* @@ -1670,7 +1661,7 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb, ret = replay_one_extent(wc->trans, root, path, eb, i, &key); BUG_ON(ret); - } else if (key.type == BTRFS_CSUM_ITEM_KEY) { + } else if (key.type == BTRFS_EXTENT_CSUM_KEY) { ret = replay_one_csum(wc->trans, root, path, eb, i, &key); BUG_ON(ret); @@ -2466,6 +2457,85 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans, return 0; } +static noinline int copy_extent_csums(struct btrfs_trans_handle *trans, + struct list_head *list, + struct btrfs_root *root, + u64 disk_bytenr, u64 len) +{ + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; + int ret; + struct btrfs_path *path; + struct btrfs_csum_item *item = NULL; + u64 end = disk_bytenr + len; + u64 item_start_offset = 0; + u64 item_last_offset = 0; + u32 diff; + u32 sum; + u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy); + + sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS); + + sector_sum = sums->sums; + sums->bytenr = disk_bytenr; + sums->len = len; + list_add_tail(&sums->list, list); + + path = btrfs_alloc_path(); + while(disk_bytenr < end) { + if (!item || disk_bytenr < item_start_offset || + disk_bytenr >= item_last_offset) { + struct btrfs_key found_key; + u32 item_size; + + if (item) + btrfs_release_path(root, path); + item = btrfs_lookup_csum(NULL, root, path, + disk_bytenr, 0); + if (IS_ERR(item)) { + ret = PTR_ERR(item); + if (ret == -ENOENT || ret == -EFBIG) + ret = 0; + sum = 0; + printk("log no csum found for byte %llu\n", + (unsigned long long)disk_bytenr); + item = NULL; + btrfs_release_path(root, path); + goto found; + } + btrfs_item_key_to_cpu(path->nodes[0], &found_key, + path->slots[0]); + + item_start_offset = found_key.offset; + item_size = btrfs_item_size_nr(path->nodes[0], + path->slots[0]); + item_last_offset = item_start_offset + + (item_size / csum_size) * + root->sectorsize; + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_csum_item); + } + /* + * this byte range must be able to fit inside + * a single leaf so it will also fit inside a u32 + */ + diff = disk_bytenr - item_start_offset; + diff = diff / root->sectorsize; + diff = diff * csum_size; + + read_extent_buffer(path->nodes[0], &sum, + ((unsigned long)item) + diff, + csum_size); +found: + sector_sum->bytenr = disk_bytenr; + sector_sum->sum = sum; + disk_bytenr += root->sectorsize; + sector_sum++; + } + btrfs_free_path(path); + return 0; +} + static noinline int copy_items(struct btrfs_trans_handle *trans, struct btrfs_root *log, struct btrfs_path *dst_path, @@ -2481,6 +2551,9 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, u32 *ins_sizes; char *ins_data; int i; + struct list_head ordered_sums; + + INIT_LIST_HEAD(&ordered_sums); ins_data = kmalloc(nr * sizeof(struct btrfs_key) + nr * sizeof(u32), GFP_NOFS); @@ -2535,6 +2608,9 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, extent); u64 dl = btrfs_file_extent_disk_num_bytes(src, extent); + u64 cs = btrfs_file_extent_offset(src, extent); + u64 cl = btrfs_file_extent_num_bytes(src, + extent);; /* ds == 0 is a hole */ if (ds != 0) { ret = btrfs_inc_extent_ref(trans, log, @@ -2544,6 +2620,11 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, trans->transid, ins_keys[i].objectid); BUG_ON(ret); + ret = copy_extent_csums(trans, + &ordered_sums, + log->fs_info->csum_root, + ds + cs, cl); + BUG_ON(ret); } } } @@ -2553,6 +2634,20 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, btrfs_mark_buffer_dirty(dst_path->nodes[0]); btrfs_release_path(log, dst_path); kfree(ins_data); + + /* + * we have to do this after the loop above to avoid changing the + * log tree while trying to change the log tree. + */ + while(!list_empty(&ordered_sums)) { + struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next, + struct btrfs_ordered_sum, + list); + ret = btrfs_csum_file_blocks(trans, log, sums); + BUG_ON(ret); + list_del(&sums->list); + kfree(sums); + } return 0; } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 6c523b3360f..2049d179ccd 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2771,6 +2771,7 @@ static struct btrfs_device *add_missing_dev(struct btrfs_root *root, device->work.func = pending_bios_fn; fs_devices->num_devices++; spin_lock_init(&device->io_lock); + INIT_LIST_HEAD(&device->dev_alloc_list); memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE); return device; } -- cgit v1.2.3-70-g09d2 From c3027eb5523d6983f12628f3fe13d8a7576db701 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 8 Dec 2008 16:40:21 -0500 Subject: Btrfs: Add inode sequence number for NFS and reserved space in a few structs This adds a sequence number to the btrfs inode that is increased on every update. NFS will be able to use that to detect when an inode has changed, without relying on inaccurate time fields. While we're here, this also: Puts reserved space into the super block and inode Adds a log root transid to the super so we can pick the newest super based on the fsync log as well as the main transaction ID. For now the log root transid is always zero, but that'll get fixed. Adds a starting offset to the dev_item. This will let us do better alignment calculations if we know the start of a partition on the disk. Signed-off-by: Chris Mason --- fs/btrfs/btrfs_inode.h | 6 +++--- fs/btrfs/ctree.h | 26 ++++++++++++++++++++++++++ fs/btrfs/file.c | 1 + fs/btrfs/inode.c | 4 +++- fs/btrfs/volumes.c | 1 + 5 files changed, 34 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 0b2e623cf42..1b9ec1ab1f6 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -49,9 +49,6 @@ struct btrfs_inode { */ struct extent_io_tree io_failure_tree; - /* held while inserting checksums to avoid races */ - struct mutex csum_mutex; - /* held while inesrting or deleting extents from files */ struct mutex extent_mutex; @@ -79,6 +76,9 @@ struct btrfs_inode { */ u64 generation; + /* sequence number for NFS changes */ + u64 sequence; + /* * transid of the trans_handle that last modified this inode */ diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 242b961ae6d..f72b4381934 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -196,6 +196,12 @@ struct btrfs_dev_item { /* expected generation for this device */ __le64 generation; + /* + * starting byte of this partition on the device, + * to allowr for stripe alignment in the future + */ + __le64 start_offset; + /* grouping information for allocation decisions */ __le32 dev_group; @@ -311,6 +317,9 @@ struct btrfs_super_block { __le64 root; __le64 chunk_root; __le64 log_root; + + /* this will help find the new super based on the log root */ + __le64 log_root_transid; __le64 total_bytes; __le64 bytes_used; __le64 root_dir_objectid; @@ -329,7 +338,11 @@ struct btrfs_super_block { u8 chunk_root_level; u8 log_root_level; struct btrfs_dev_item dev_item; + char label[BTRFS_LABEL_SIZE]; + + /* future expansion */ + __le64 reserved[32]; u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE]; } __attribute__ ((__packed__)); @@ -463,6 +476,14 @@ struct btrfs_inode_item { __le64 rdev; __le64 flags; + /* modification sequence number for NFS */ + __le64 sequence; + + /* + * a little future expansion, for more than this we can + * just grow the inode item and version it + */ + __le64 reserved[4]; struct btrfs_timespec atime; struct btrfs_timespec ctime; struct btrfs_timespec mtime; @@ -1001,6 +1022,8 @@ BTRFS_SETGET_FUNCS(device_total_bytes, struct btrfs_dev_item, total_bytes, 64); BTRFS_SETGET_FUNCS(device_bytes_used, struct btrfs_dev_item, bytes_used, 64); BTRFS_SETGET_FUNCS(device_io_align, struct btrfs_dev_item, io_align, 32); BTRFS_SETGET_FUNCS(device_io_width, struct btrfs_dev_item, io_width, 32); +BTRFS_SETGET_FUNCS(device_start_offset, struct btrfs_dev_item, + start_offset, 64); BTRFS_SETGET_FUNCS(device_sector_size, struct btrfs_dev_item, sector_size, 32); BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64); BTRFS_SETGET_FUNCS(device_group, struct btrfs_dev_item, dev_group, 32); @@ -1135,6 +1158,7 @@ BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64); /* struct btrfs_inode_item */ BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64); +BTRFS_SETGET_FUNCS(inode_sequence, struct btrfs_inode_item, sequence, 64); BTRFS_SETGET_FUNCS(inode_transid, struct btrfs_inode_item, transid, 64); BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64); BTRFS_SETGET_FUNCS(inode_nbytes, struct btrfs_inode_item, nbytes, 64); @@ -1519,6 +1543,8 @@ BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block, chunk_root_level, 8); BTRFS_SETGET_STACK_FUNCS(super_log_root, struct btrfs_super_block, log_root, 64); +BTRFS_SETGET_STACK_FUNCS(super_log_root_transid, struct btrfs_super_block, + log_root_transid, 64); BTRFS_SETGET_STACK_FUNCS(super_log_root_level, struct btrfs_super_block, log_root_level, 8); BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block, diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 1c9243560ea..b5a6a2b6f66 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1055,6 +1055,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL); mutex_lock(&inode->i_mutex); + BTRFS_I(inode)->sequence++; first_index = pos >> PAGE_CACHE_SHIFT; last_index = (pos + count) >> PAGE_CACHE_SHIFT; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c03d847b8c4..932d8c0b2c0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1963,6 +1963,7 @@ void btrfs_read_locked_inode(struct inode *inode) inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item)); BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item); + BTRFS_I(inode)->sequence = btrfs_inode_sequence(leaf, inode_item); inode->i_generation = BTRFS_I(inode)->generation; inode->i_rdev = 0; rdev = btrfs_inode_rdev(leaf, inode_item); @@ -2043,6 +2044,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode)); btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation); + btrfs_set_inode_sequence(leaf, item, BTRFS_I(inode)->sequence); btrfs_set_inode_transid(leaf, item, trans->transid); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); @@ -2945,6 +2947,7 @@ static noinline void init_btrfs_i(struct inode *inode) bi->i_default_acl = NULL; bi->generation = 0; + bi->sequence = 0; bi->last_trans = 0; bi->logged_trans = 0; bi->delalloc_bytes = 0; @@ -2959,7 +2962,6 @@ static noinline void init_btrfs_i(struct inode *inode) inode->i_mapping, GFP_NOFS); INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); - mutex_init(&BTRFS_I(inode)->csum_mutex); mutex_init(&BTRFS_I(inode)->extent_mutex); mutex_init(&BTRFS_I(inode)->log_mutex); } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 825364fae69..4d210a731d4 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -890,6 +890,7 @@ int btrfs_add_device(struct btrfs_trans_handle *trans, btrfs_set_device_group(leaf, dev_item, 0); btrfs_set_device_seek_speed(leaf, dev_item, 0); btrfs_set_device_bandwidth(leaf, dev_item, 0); + btrfs_set_device_start_offset(leaf, dev_item, 0); ptr = (unsigned long)btrfs_device_uuid(dev_item); write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); -- cgit v1.2.3-70-g09d2 From d2fb3437e4d8d12c73c587615ad187d5288547ec Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 11 Dec 2008 16:30:39 -0500 Subject: Btrfs: fix leaking block group on balance The block group structs are referenced in many different places, and it's not safe to free while balancing. So, those block group structs were simply leaked instead. This patch replaces the block group pointer in the inode with the starting byte offset of the block group and adds reference counting to the block group struct. Signed-off-by: Yan Zheng --- fs/btrfs/btrfs_inode.h | 8 ++- fs/btrfs/ctree.h | 17 ++++--- fs/btrfs/extent-tree.c | 132 ++++++++++++++++++++++--------------------------- fs/btrfs/inode.c | 43 +++++----------- fs/btrfs/ioctl.c | 2 +- fs/btrfs/transaction.c | 2 +- fs/btrfs/transaction.h | 2 +- 7 files changed, 88 insertions(+), 118 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 1b9ec1ab1f6..a8c9693b75a 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -28,11 +28,6 @@ struct btrfs_inode { /* which subvolume this inode belongs to */ struct btrfs_root *root; - /* the block group preferred for allocations. This pointer is buggy - * and needs to be replaced with a bytenr instead - */ - struct btrfs_block_group_cache *block_group; - /* key used to find this inode on disk. This is used by the code * to read in roots of subvolumes */ @@ -115,6 +110,9 @@ struct btrfs_inode { */ u64 index_cnt; + /* the start of block group preferred for allocations. */ + u64 block_group; + struct inode vfs_inode; }; diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 5b0c79d22c0..8733081d97a 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -653,6 +653,9 @@ struct btrfs_block_group_cache { /* for block groups in the same raid type */ struct list_head list; + + /* usage count */ + atomic_t count; }; struct btrfs_leaf_ref_tree { @@ -1706,10 +1709,8 @@ int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); struct btrfs_block_group_cache *btrfs_lookup_block_group(struct btrfs_fs_info *info, u64 bytenr); -struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, - struct btrfs_block_group_cache - *hint, u64 search_start, - int data, int owner); +u64 btrfs_find_block_group(struct btrfs_root *root, + u64 search_start, u64 search_hint, int owner); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, u32 blocksize, u64 parent, @@ -1770,6 +1771,7 @@ int btrfs_update_extent_ref(struct btrfs_trans_handle *trans, u64 owner_objectid); int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, struct btrfs_root *root); +int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr); int btrfs_free_block_groups(struct btrfs_fs_info *info); int btrfs_read_block_groups(struct btrfs_root *root); int btrfs_make_block_group(struct btrfs_trans_handle *trans, @@ -2019,10 +2021,9 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root); int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end); int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc); -int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, - struct btrfs_trans_handle *trans, u64 new_dirid, - struct btrfs_block_group_cache *block_group); - +int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, + struct btrfs_root *new_root, struct dentry *dentry, + u64 new_dirid, u64 alloc_hint); int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio, unsigned long bio_flags); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 673ff59c288..1cc89246ee2 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -53,10 +53,6 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, int all); static int del_pending_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, int all); -static struct btrfs_block_group_cache * -__btrfs_find_block_group(struct btrfs_root *root, - struct btrfs_block_group_cache *hint, - u64 search_start, int data, int owner); static int pin_down_bytes(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, int is_data); @@ -142,6 +138,8 @@ block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr, break; } } + if (ret) + atomic_inc(&ret->count); spin_unlock(&info->block_group_cache_lock); return ret; @@ -318,6 +316,12 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct return cache; } +static inline void put_block_group(struct btrfs_block_group_cache *cache) +{ + if (atomic_dec_and_test(&cache->count)) + kfree(cache); +} + static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, u64 flags) { @@ -341,54 +345,16 @@ static u64 div_factor(u64 num, int factor) return num; } -static struct btrfs_block_group_cache * -__btrfs_find_block_group(struct btrfs_root *root, - struct btrfs_block_group_cache *hint, - u64 search_start, int data, int owner) +u64 btrfs_find_block_group(struct btrfs_root *root, + u64 search_start, u64 search_hint, int owner) { struct btrfs_block_group_cache *cache; - struct btrfs_block_group_cache *found_group = NULL; - struct btrfs_fs_info *info = root->fs_info; u64 used; - u64 last = 0; - u64 free_check; + u64 last = max(search_hint, search_start); + u64 group_start = 0; int full_search = 0; - int factor = 10; + int factor = 9; int wrapped = 0; - - if (data & BTRFS_BLOCK_GROUP_METADATA) - factor = 9; - - if (search_start) { - struct btrfs_block_group_cache *shint; - shint = btrfs_lookup_first_block_group(info, search_start); - if (shint && block_group_bits(shint, data)) { - spin_lock(&shint->lock); - used = btrfs_block_group_used(&shint->item); - if (used + shint->pinned + shint->reserved < - div_factor(shint->key.offset, factor)) { - spin_unlock(&shint->lock); - return shint; - } - spin_unlock(&shint->lock); - } - } - if (hint && block_group_bits(hint, data)) { - spin_lock(&hint->lock); - used = btrfs_block_group_used(&hint->item); - if (used + hint->pinned + hint->reserved < - div_factor(hint->key.offset, factor)) { - spin_unlock(&hint->lock); - return hint; - } - spin_unlock(&hint->lock); - last = hint->key.objectid + hint->key.offset; - } else { - if (hint) - last = max(hint->key.objectid, search_start); - else - last = search_start; - } again: while (1) { cache = btrfs_lookup_first_block_group(root->fs_info, last); @@ -399,16 +365,18 @@ again: last = cache->key.objectid + cache->key.offset; used = btrfs_block_group_used(&cache->item); - if (block_group_bits(cache, data)) { - free_check = div_factor(cache->key.offset, factor); + if ((full_search || !cache->ro) && + block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) { if (used + cache->pinned + cache->reserved < - free_check) { - found_group = cache; + div_factor(cache->key.offset, factor)) { + group_start = cache->key.objectid; spin_unlock(&cache->lock); + put_block_group(cache); goto found; } } spin_unlock(&cache->lock); + put_block_group(cache); cond_resched(); } if (!wrapped) { @@ -423,18 +391,7 @@ again: goto again; } found: - return found_group; -} - -struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, - struct btrfs_block_group_cache - *hint, u64 search_start, - int data, int owner) -{ - - struct btrfs_block_group_cache *ret; - ret = __btrfs_find_block_group(root, hint, search_start, data, owner); - return ret; + return group_start; } /* simple helper to search for an existing extent at a given offset */ @@ -1809,6 +1766,19 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, return werr; } +int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr) +{ + struct btrfs_block_group_cache *block_group; + int readonly = 0; + + block_group = btrfs_lookup_block_group(root->fs_info, bytenr); + if (!block_group || block_group->ro) + readonly = 1; + if (block_group) + put_block_group(block_group); + return readonly; +} + static int update_space_info(struct btrfs_fs_info *info, u64 flags, u64 total_bytes, u64 bytes_used, struct btrfs_space_info **space_info) @@ -1995,10 +1965,10 @@ static int update_block_group(struct btrfs_trans_handle *trans, int ret; ret = btrfs_add_free_space(cache, bytenr, num_bytes); - if (ret) - return -1; + WARN_ON(ret); } } + put_block_group(cache); total -= num_bytes; bytenr += num_bytes; } @@ -2008,12 +1978,16 @@ static int update_block_group(struct btrfs_trans_handle *trans, static u64 first_logical_byte(struct btrfs_root *root, u64 search_start) { struct btrfs_block_group_cache *cache; + u64 bytenr; cache = btrfs_lookup_first_block_group(root->fs_info, search_start); if (!cache) return 0; - return cache->key.objectid; + bytenr = cache->key.objectid; + put_block_group(cache); + + return bytenr; } int btrfs_update_pinned_extents(struct btrfs_root *root, @@ -2055,6 +2029,7 @@ int btrfs_update_pinned_extents(struct btrfs_root *root, if (cache->cached) btrfs_add_free_space(cache, bytenr, len); } + put_block_group(cache); bytenr += len; num -= len; } @@ -2085,6 +2060,7 @@ static int update_reserved_extents(struct btrfs_root *root, } spin_unlock(&cache->lock); spin_unlock(&cache->space_info->lock); + put_block_group(cache); bytenr += len; num -= len; } @@ -2724,6 +2700,7 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, cache = btrfs_lookup_block_group(root->fs_info, bytenr); BUG_ON(!cache); btrfs_add_free_space(cache, bytenr, num_bytes); + put_block_group(cache); update_reserved_extents(root, bytenr, num_bytes, 0); return 0; } @@ -2928,6 +2905,8 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, } new_group: mutex_unlock(&block_group->alloc_mutex); + put_block_group(block_group); + block_group = NULL; new_group_no_lock: /* don't try to compare new allocations against the * last allocation any more @@ -2997,6 +2976,8 @@ loop_check: block_group = list_entry(cur, struct btrfs_block_group_cache, list); + atomic_inc(&block_group->count); + search_start = block_group->key.objectid; cur = cur->next; } @@ -3004,7 +2985,7 @@ loop_check: /* we found what we needed */ if (ins->objectid) { if (!(data & BTRFS_BLOCK_GROUP_DATA)) - trans->block_group = block_group; + trans->block_group = block_group->key.objectid; if (last_ptr) *last_ptr = ins->objectid + ins->offset; @@ -3015,6 +2996,8 @@ loop_check: loop, allowed_chunk_alloc); ret = -ENOSPC; } + if (block_group) + put_block_group(block_group); up_read(&space_info->groups_sem); return ret; @@ -3124,6 +3107,7 @@ int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len) return -ENOSPC; } btrfs_add_free_space(cache, start, len); + put_block_group(cache); update_reserved_extents(root, start, len, 0); return 0; } @@ -3288,6 +3272,7 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans, ret = btrfs_remove_free_space(block_group, ins->objectid, ins->offset); BUG_ON(ret); + put_block_group(block_group); ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid, ref_generation, owner, ins); return ret; @@ -5703,6 +5688,7 @@ next: WARN_ON(block_group->reserved > 0); WARN_ON(btrfs_block_group_used(&block_group->item) > 0); spin_unlock(&block_group->lock); + put_block_group(block_group); ret = 0; out: btrfs_free_path(path); @@ -5763,6 +5749,8 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info) down_write(&block_group->space_info->groups_sem); list_del(&block_group->list); up_write(&block_group->space_info->groups_sem); + + WARN_ON(atomic_read(&block_group->count) != 1); kfree(block_group); spin_lock(&info->block_group_cache_lock); @@ -5807,6 +5795,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) break; } + atomic_set(&cache->count, 1); spin_lock_init(&cache->lock); mutex_init(&cache->alloc_mutex); mutex_init(&cache->cache_mutex); @@ -5861,11 +5850,12 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans, cache->key.objectid = chunk_offset; cache->key.offset = size; + cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY; + atomic_set(&cache->count, 1); spin_lock_init(&cache->lock); mutex_init(&cache->alloc_mutex); mutex_init(&cache->cache_mutex); INIT_LIST_HEAD(&cache->list); - btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); btrfs_set_block_group_used(&cache->item, bytes_used); btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); @@ -5926,10 +5916,8 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans, spin_unlock(&block_group->space_info->lock); block_group->space_info->full = 0; - /* - memset(shrink_block_group, 0, sizeof(*shrink_block_group)); - kfree(shrink_block_group); - */ + put_block_group(block_group); + put_block_group(block_group); ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret > 0) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 932d8c0b2c0..0a28b770631 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -989,7 +989,6 @@ next_slot: if (extent_type == BTRFS_FILE_EXTENT_REG || extent_type == BTRFS_FILE_EXTENT_PREALLOC) { - struct btrfs_block_group_cache *block_group; disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); extent_end = found_key.offset + btrfs_file_extent_num_bytes(leaf, fi); @@ -1007,9 +1006,7 @@ next_slot: goto out_check; if (btrfs_cross_ref_exist(trans, root, disk_bytenr)) goto out_check; - block_group = btrfs_lookup_block_group(root->fs_info, - disk_bytenr); - if (!block_group || block_group->ro) + if (btrfs_extent_readonly(root, disk_bytenr)) goto out_check; disk_bytenr += btrfs_file_extent_offset(leaf, fi); nocow = 1; @@ -1969,16 +1966,11 @@ void btrfs_read_locked_inode(struct inode *inode) rdev = btrfs_inode_rdev(leaf, inode_item); BTRFS_I(inode)->index_cnt = (u64)-1; + BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); alloc_group_block = btrfs_inode_block_group(leaf, inode_item); - BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, - alloc_group_block); - BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); - if (!BTRFS_I(inode)->block_group) { - BTRFS_I(inode)->block_group = btrfs_find_block_group(root, - NULL, 0, - BTRFS_BLOCK_GROUP_METADATA, 0); - } + BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0, + alloc_group_block, 0); btrfs_free_path(path); inode_item = NULL; @@ -2048,8 +2040,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, btrfs_set_inode_transid(leaf, item, trans->transid); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); - btrfs_set_inode_block_group(leaf, item, - BTRFS_I(inode)->block_group->key.objectid); + btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group); } /* @@ -3358,14 +3349,11 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *dir, const char *name, int name_len, - u64 ref_objectid, - u64 objectid, - struct btrfs_block_group_cache *group, - int mode, u64 *index) + u64 ref_objectid, u64 objectid, + u64 alloc_hint, int mode, u64 *index) { struct inode *inode; struct btrfs_inode_item *inode_item; - struct btrfs_block_group_cache *new_inode_group; struct btrfs_key *location; struct btrfs_path *path; struct btrfs_inode_ref *ref; @@ -3401,13 +3389,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 0; else owner = 1; - new_inode_group = btrfs_find_block_group(root, group, 0, - BTRFS_BLOCK_GROUP_METADATA, owner); - if (!new_inode_group) { - printk("find_block group failed\n"); - new_inode_group = group; - } - BTRFS_I(inode)->block_group = new_inode_group; + BTRFS_I(inode)->block_group = + btrfs_find_block_group(root, 0, alloc_hint, owner); key[0].objectid = objectid; btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); @@ -4366,16 +4349,16 @@ out: /* * create a new subvolume directory/inode (helper for the ioctl). */ -int btrfs_create_subvol_root(struct btrfs_root *new_root, struct dentry *dentry, - struct btrfs_trans_handle *trans, u64 new_dirid, - struct btrfs_block_group_cache *block_group) +int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, + struct btrfs_root *new_root, struct dentry *dentry, + u64 new_dirid, u64 alloc_hint) { struct inode *inode; int error; u64 index = 0; inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, - new_dirid, block_group, S_IFDIR | 0700, &index); + new_dirid, alloc_hint, S_IFDIR | 0700, &index); if (IS_ERR(inode)) return PTR_ERR(inode); inode->i_op = &btrfs_dir_inode_operations; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 69c4a07f586..5d67858ce99 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -173,7 +173,7 @@ static noinline int create_subvol(struct btrfs_root *root, trans = btrfs_start_transaction(new_root, 1); BUG_ON(!trans); - ret = btrfs_create_subvol_root(new_root, dentry, trans, new_dirid, + ret = btrfs_create_subvol_root(trans, new_root, dentry, new_dirid, BTRFS_I(dir)->block_group); if (ret) goto fail; diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 47cd5fcad2c..4604178a43a 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -182,7 +182,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, h->transaction = root->fs_info->running_transaction; h->blocks_reserved = num_blocks; h->blocks_used = 0; - h->block_group = NULL; + h->block_group = 0; h->alloc_exclude_nr = 0; h->alloc_exclude_start = 0; root->fs_info->running_transaction->use_count++; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 202c8be6c05..ffe7f639732 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -41,7 +41,7 @@ struct btrfs_trans_handle { unsigned long blocks_reserved; unsigned long blocks_used; struct btrfs_transaction *transaction; - struct btrfs_block_group_cache *block_group; + u64 block_group; u64 alloc_exclude_start; u64 alloc_exclude_nr; }; -- cgit v1.2.3-70-g09d2 From 17d217fe970d34720f4f1633dca73a6aa2f3d9d1 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Fri, 12 Dec 2008 10:03:38 -0500 Subject: Btrfs: fix nodatasum handling in balancing code Checksums on data can be disabled by mount option, so it's possible some data extents don't have checksums or have invalid checksums. This causes trouble for data relocation. This patch contains following things to make data relocation work. 1) make nodatasum/nodatacow mount option only affects new files. Checksums and COW on data are only controlled by the inode flags. 2) check the existence of checksum in the nodatacow checker. If checksums exist, force COW the data extent. This ensure that checksum for a given block is either valid or does not exist. 3) update data relocation code to properly handle the case of checksum missing. Signed-off-by: Yan Zheng --- fs/btrfs/compression.c | 9 ++-- fs/btrfs/ctree.h | 5 ++- fs/btrfs/extent-tree.c | 50 ++++++++++++++++++++-- fs/btrfs/extent_io.h | 1 + fs/btrfs/file-item.c | 114 +++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/file.c | 8 ---- fs/btrfs/inode.c | 74 ++++++++++++++++++++++++++------ 7 files changed, 226 insertions(+), 35 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index ad727413730..2436163d543 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -124,8 +124,7 @@ static int check_compressed_csum(struct inode *inode, u32 csum; u32 *cb_sum = &cb->sums; - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) + if (btrfs_test_flag(inode, NODATASUM)) return 0; for (i = 0; i < cb->nr_pages; i++) { @@ -671,8 +670,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, */ atomic_inc(&cb->pending_bios); - if (!btrfs_test_opt(root, NODATASUM) && - !btrfs_test_flag(inode, NODATASUM)) { + if (!btrfs_test_flag(inode, NODATASUM)) { btrfs_lookup_bio_sums(root, inode, comp_bio, sums); } @@ -699,8 +697,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); BUG_ON(ret); - if (!btrfs_test_opt(root, NODATASUM) && - !btrfs_test_flag(inode, NODATASUM)) { + if (!btrfs_test_flag(inode, NODATASUM)) { btrfs_lookup_bio_sums(root, inode, comp_bio, sums); } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8733081d97a..b89999de456 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1702,7 +1702,7 @@ int btrfs_update_pinned_extents(struct btrfs_root *root, int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *leaf); int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr); + struct btrfs_root *root, u64 objectid, u64 bytenr); int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); @@ -1789,6 +1789,7 @@ int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans, struct extent_buffer *buf, u64 orig_start); int btrfs_add_dead_reloc_root(struct btrfs_root *root); int btrfs_cleanup_reloc_trees(struct btrfs_root *root); +int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len); u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags); /* ctree.c */ int btrfs_previous_item(struct btrfs_root *root, @@ -1994,6 +1995,8 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, int btrfs_csum_truncate(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 isize); +int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, + u64 end, struct list_head *list); /* inode.c */ /* RHEL and EL kernels have a patch that renames PG_checked to FsMisc */ diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 171057a3267..8004695d24d 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1359,7 +1359,7 @@ out: } int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, - struct btrfs_root *root, u64 bytenr) + struct btrfs_root *root, u64 objectid, u64 bytenr) { struct btrfs_root *extent_root = root->fs_info->extent_root; struct btrfs_path *path; @@ -1418,8 +1418,9 @@ int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, ref_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref); ref_root = btrfs_ref_root(leaf, ref_item); - if (ref_root != root->root_key.objectid && - ref_root != BTRFS_TREE_LOG_OBJECTID) { + if ((ref_root != root->root_key.objectid && + ref_root != BTRFS_TREE_LOG_OBJECTID) || + objectid != btrfs_ref_objectid(leaf, ref_item)) { ret = 1; goto out; } @@ -5367,7 +5368,6 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, if (ret) goto out; } - btrfs_record_root_in_trans(found_root); ret = replace_one_extent(trans, found_root, path, extent_key, &first_key, ref_path, @@ -5534,6 +5534,7 @@ static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info, } else { BUG_ON(1); } + BTRFS_I(inode)->index_cnt = group->key.objectid; err = btrfs_orphan_add(trans, inode); out: @@ -5546,6 +5547,47 @@ out: return inode; } +int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len) +{ + + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; + struct btrfs_ordered_extent *ordered; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct list_head list; + size_t offset; + int ret; + u64 disk_bytenr; + + INIT_LIST_HEAD(&list); + + ordered = btrfs_lookup_ordered_extent(inode, file_pos); + BUG_ON(ordered->file_offset != file_pos || ordered->len != len); + + disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt; + ret = btrfs_lookup_csums_range(root, disk_bytenr, + disk_bytenr + len - 1, &list); + + while (!list_empty(&list)) { + sums = list_entry(list.next, struct btrfs_ordered_sum, list); + list_del_init(&sums->list); + + sector_sum = sums->sums; + sums->bytenr = ordered->start; + + offset = 0; + while (offset < sums->len) { + sector_sum->bytenr += ordered->start - disk_bytenr; + sector_sum++; + offset += root->sectorsize; + } + + btrfs_add_ordered_sum(inode, ordered, sums); + } + btrfs_put_ordered_extent(ordered); + return 0; +} + int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start) { struct btrfs_trans_handle *trans; diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h index 2d5f67065b6..c5b483a7913 100644 --- a/fs/btrfs/extent_io.h +++ b/fs/btrfs/extent_io.h @@ -16,6 +16,7 @@ #define EXTENT_ORDERED (1 << 9) #define EXTENT_ORDERED_METADATA (1 << 10) #define EXTENT_BOUNDARY (1 << 11) +#define EXTENT_NODATASUM (1 << 12) #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) /* flags for bio submission */ diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index 3ebef871ee6..df0447632db 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -140,6 +140,7 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans, return ret; } + int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, struct bio *bio, u32 *dst) { @@ -185,9 +186,16 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, if (ret == -ENOENT || ret == -EFBIG) ret = 0; sum = 0; - printk("no csum found for inode %lu start " - "%llu\n", inode->i_ino, - (unsigned long long)offset); + if (BTRFS_I(inode)->root->root_key.objectid == + BTRFS_DATA_RELOC_TREE_OBJECTID) { + set_extent_bits(io_tree, offset, + offset + bvec->bv_len - 1, + EXTENT_NODATASUM, GFP_NOFS); + } else { + printk("no csum found for inode %lu " + "start %llu\n", inode->i_ino, + (unsigned long long)offset); + } item = NULL; btrfs_release_path(root, path); goto found; @@ -228,6 +236,106 @@ found: return 0; } +int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end, + struct list_head *list) +{ + struct btrfs_key key; + struct btrfs_path *path; + struct extent_buffer *leaf; + struct btrfs_ordered_sum *sums; + struct btrfs_sector_sum *sector_sum; + struct btrfs_csum_item *item; + unsigned long offset; + int ret; + size_t size; + u64 csum_end; + u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy); + + path = btrfs_alloc_path(); + BUG_ON(!path); + + key.objectid = BTRFS_EXTENT_CSUM_OBJECTID; + key.offset = start; + key.type = BTRFS_EXTENT_CSUM_KEY; + + ret = btrfs_search_slot(NULL, root->fs_info->csum_root, + &key, path, 0, 0); + if (ret < 0) + goto fail; + if (ret > 0 && path->slots[0] > 0) { + leaf = path->nodes[0]; + btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1); + if (key.objectid == BTRFS_EXTENT_CSUM_OBJECTID && + key.type == BTRFS_EXTENT_CSUM_KEY) { + offset = (start - key.offset) >> + root->fs_info->sb->s_blocksize_bits; + if (offset * csum_size < + btrfs_item_size_nr(leaf, path->slots[0] - 1)) + path->slots[0]--; + } + } + + while (start <= end) { + leaf = path->nodes[0]; + if (path->slots[0] >= btrfs_header_nritems(leaf)) { + ret = btrfs_next_leaf(root->fs_info->csum_root, path); + if (ret < 0) + goto fail; + if (ret > 0) + break; + leaf = path->nodes[0]; + } + + btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); + if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID || + key.type != BTRFS_EXTENT_CSUM_KEY) + break; + + btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); + if (key.offset > end) + break; + + if (key.offset > start) + start = key.offset; + + size = btrfs_item_size_nr(leaf, path->slots[0]); + csum_end = key.offset + (size / csum_size) * root->sectorsize; + + size = min(csum_end, end + 1) - start; + sums = kzalloc(btrfs_ordered_sum_size(root, size), GFP_NOFS); + BUG_ON(!sums); + + sector_sum = sums->sums; + sums->bytenr = start; + sums->len = size; + + offset = (start - key.offset) >> + root->fs_info->sb->s_blocksize_bits; + offset *= csum_size; + + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_csum_item); + while (size > 0) { + read_extent_buffer(path->nodes[0], §or_sum->sum, + ((unsigned long)item) + offset, + csum_size); + sector_sum->bytenr = start; + + size -= root->sectorsize; + start += root->sectorsize; + offset += csum_size; + sector_sum++; + } + list_add_tail(&sums->list, list); + + path->slots[0]++; + } + ret = 0; +fail: + btrfs_free_path(path); + return ret; +} + int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, struct bio *bio, u64 file_start, int contig) { diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 71bfe3a6a44..507081059d9 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1059,14 +1059,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, first_index = pos >> PAGE_CACHE_SHIFT; last_index = (pos + count) >> PAGE_CACHE_SHIFT; - /* - * if this is a nodatasum mount, force summing off for the inode - * all the time. That way a later mount with summing on won't - * get confused - */ - if (btrfs_test_opt(root, NODATASUM)) - btrfs_set_flag(inode, NODATASUM); - /* * there are lots of better ways to do this, but this code * makes sure the first and last page in the file range are diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0a28b770631..e64a4fe19a6 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -771,6 +771,13 @@ static noinline int cow_file_range(struct inode *inode, ram_size, cur_alloc_size, 0); BUG_ON(ret); + if (root->root_key.objectid == + BTRFS_DATA_RELOC_TREE_OBJECTID) { + ret = btrfs_reloc_clone_csums(inode, start, + cur_alloc_size); + BUG_ON(ret); + } + if (disk_num_bytes < cur_alloc_size) { printk("num_bytes %Lu cur_alloc %Lu\n", disk_num_bytes, cur_alloc_size); @@ -910,6 +917,26 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, return 0; } +static int noinline csum_exist_in_range(struct btrfs_root *root, + u64 bytenr, u64 num_bytes) +{ + int ret; + struct btrfs_ordered_sum *sums; + LIST_HEAD(list); + + ret = btrfs_lookup_csums_range(root, bytenr, bytenr + num_bytes - 1, + &list); + if (ret == 0 && list_empty(&list)) + return 0; + + while (!list_empty(&list)) { + sums = list_entry(list.next, struct btrfs_ordered_sum, list); + list_del(&sums->list); + kfree(sums); + } + return 1; +} + /* * when nowcow writeback call back. This checks for snapshots or COW copies * of the extents that exist in the file, and COWs the file as required. @@ -971,6 +998,7 @@ next_slot: nocow = 0; disk_bytenr = 0; + num_bytes = 0; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); if (found_key.objectid > inode->i_ino || @@ -996,19 +1024,29 @@ next_slot: path->slots[0]++; goto next_slot; } + if (disk_bytenr == 0) + goto out_check; if (btrfs_file_extent_compression(leaf, fi) || btrfs_file_extent_encryption(leaf, fi) || btrfs_file_extent_other_encoding(leaf, fi)) goto out_check; - if (disk_bytenr == 0) - goto out_check; if (extent_type == BTRFS_FILE_EXTENT_REG && !force) goto out_check; - if (btrfs_cross_ref_exist(trans, root, disk_bytenr)) - goto out_check; if (btrfs_extent_readonly(root, disk_bytenr)) goto out_check; + if (btrfs_cross_ref_exist(trans, root, inode->i_ino, + disk_bytenr)) + goto out_check; disk_bytenr += btrfs_file_extent_offset(leaf, fi); + disk_bytenr += cur_offset - found_key.offset; + num_bytes = min(end + 1, extent_end) - cur_offset; + /* + * force cow if csum exists in the range. + * this ensure that csum for a given extent are + * either valid or do not exist. + */ + if (csum_exist_in_range(root, disk_bytenr, num_bytes)) + goto out_check; nocow = 1; } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { extent_end = found_key.offset + @@ -1041,8 +1079,6 @@ out_check: cow_start = (u64)-1; } - disk_bytenr += cur_offset - found_key.offset; - num_bytes = min(end + 1, extent_end) - cur_offset; if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) { struct extent_map *em; struct extent_map_tree *em_tree; @@ -1105,11 +1141,9 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, u64 start, u64 end, int *page_started, unsigned long *nr_written) { - struct btrfs_root *root = BTRFS_I(inode)->root; int ret; - if (btrfs_test_opt(root, NODATACOW) || - btrfs_test_flag(inode, NODATACOW)) + if (btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, locked_page, start, end, page_started, 1, nr_written); else if (btrfs_test_flag(inode, PREALLOC)) @@ -1252,8 +1286,7 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); - skip_sum = btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM); + skip_sum = btrfs_test_flag(inode, NODATASUM); if (!(rw & (1 << BIO_RW))) { if (bio_flags & EXTENT_BIO_COMPRESSED) { @@ -1263,6 +1296,9 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, btrfs_lookup_bio_sums(root, inode, bio, NULL); goto mapit; } else if (!skip_sum) { + /* csum items have already been cloned */ + if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID) + goto mapit; /* we're doing a write, do the async checksumming */ return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, @@ -1692,9 +1728,15 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, ClearPageChecked(page); goto good; } - if (btrfs_test_opt(root, NODATASUM) || - btrfs_test_flag(inode, NODATASUM)) + if (btrfs_test_flag(inode, NODATASUM)) + return 0; + + if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID && + test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1)) { + clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM, + GFP_NOFS); return 0; + } if (state && state->start == start) { private = state->private; @@ -3391,6 +3433,12 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 1; BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0, alloc_hint, owner); + if ((mode & S_IFREG)) { + if (btrfs_test_opt(root, NODATASUM)) + btrfs_set_flag(inode, NODATASUM); + if (btrfs_test_opt(root, NODATACOW)) + btrfs_set_flag(inode, NODATACOW); + } key[0].objectid = objectid; btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); -- cgit v1.2.3-70-g09d2 From 42dc7babdcc99feadb04d461592ce5898a362550 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Dec 2008 11:44:56 -0500 Subject: Btrfs: Fix compressed writes on truncated pages The compression code was using isize to limit the amount of data it sent through zlib. But, it wasn't properly limiting the looping to just the pages inside i_size. The end result was trying to compress too many pages, including those that had not been setup and properly locked down. This made the compression code oops while trying find_get_page on a page that didn't exist. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 2 +- fs/btrfs/inode.c | 8 +++++--- 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index c0c95cccbb5..f6f7a6af035 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -2887,8 +2887,8 @@ int btrfs_split_item(struct btrfs_trans_handle *trans, path->keep_locks = 0; BUG_ON(ret); - BUG_ON(btrfs_leaf_free_space(root, leaf) < sizeof(struct btrfs_item)); leaf = path->nodes[0]; + BUG_ON(btrfs_leaf_free_space(root, leaf) < sizeof(struct btrfs_item)); split: item = btrfs_item_nr(leaf, path->slots[0]); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e64a4fe19a6..5313a13a998 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -332,6 +332,7 @@ static noinline int compress_file_range(struct inode *inode, u64 disk_num_bytes; u64 blocksize = root->sectorsize; u64 actual_end; + u64 isize = i_size_read(inode); int ret = 0; struct page **pages = NULL; unsigned long nr_pages; @@ -345,12 +346,12 @@ static noinline int compress_file_range(struct inode *inode, orig_start = start; + actual_end = min_t(u64, isize, end + 1); again: will_compress = 0; nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1; nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE); - actual_end = min_t(u64, i_size_read(inode), end + 1); total_compressed = actual_end - start; /* we want to make sure that amount of ram required to uncompress @@ -488,7 +489,7 @@ again: add_async_extent(async_cow, start, num_bytes, total_compressed, pages, nr_pages_ret); - if (start + num_bytes < end) { + if (start + num_bytes < end && start + num_bytes < actual_end) { start += num_bytes; pages = NULL; cond_resched(); @@ -696,6 +697,7 @@ static noinline int cow_file_range(struct inode *inode, u64 cur_alloc_size; u64 blocksize = root->sectorsize; u64 actual_end; + u64 isize = i_size_read(inode); struct btrfs_key ins; struct extent_map *em; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; @@ -705,7 +707,7 @@ static noinline int cow_file_range(struct inode *inode, BUG_ON(!trans); btrfs_set_trans_block_group(trans, inode); - actual_end = min_t(u64, i_size_read(inode), end + 1); + actual_end = min_t(u64, isize, end + 1); num_bytes = (end - start + blocksize) & ~(blocksize - 1); num_bytes = max(blocksize, num_bytes); -- cgit v1.2.3-70-g09d2 From 75eff68ea6959efd999b125fc3dcf73931a3b30b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 15 Dec 2008 15:54:40 -0500 Subject: Btrfs: Don't use spin*lock_irq for the delalloc lock The delalloc lock doesn't need to have irqs disabled, nobody that changes the number of delalloc bytes in the FS is running with irqs off. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 12 ++++++------ fs/btrfs/inode.c | 34 ++++++++++++++++++++-------------- 2 files changed, 26 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 8004695d24d..fbd6a8f28b5 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1047,11 +1047,11 @@ search: end = pos; /* update the free space counters */ - spin_lock_irq(&info->delalloc_lock); + spin_lock(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); btrfs_set_super_bytes_used(&info->super_copy, super_used - bytes_freed); - spin_unlock_irq(&info->delalloc_lock); + spin_unlock(&info->delalloc_lock); root_used = btrfs_root_used(&extent_root->root_item); btrfs_set_root_used(&extent_root->root_item, @@ -2463,11 +2463,11 @@ static int __free_extent(struct btrfs_trans_handle *trans, BUG_ON(ret < 0); } /* block accounting for super block */ - spin_lock_irq(&info->delalloc_lock); + spin_lock(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); btrfs_set_super_bytes_used(&info->super_copy, super_used - num_bytes); - spin_unlock_irq(&info->delalloc_lock); + spin_unlock(&info->delalloc_lock); /* block accounting for root item */ root_used = btrfs_root_used(&root->root_item); @@ -3151,10 +3151,10 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, parent = ins->objectid; /* block accounting for super block */ - spin_lock_irq(&info->delalloc_lock); + spin_lock(&info->delalloc_lock); super_used = btrfs_super_bytes_used(&info->super_copy); btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); - spin_unlock_irq(&info->delalloc_lock); + spin_unlock(&info->delalloc_lock); /* block accounting for root item */ root_used = btrfs_root_used(&root->root_item); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5313a13a998..0577e77e661 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -101,10 +101,9 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, u64 total; u64 used; u64 thresh; - unsigned long flags; int ret = 0; - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + spin_lock(&root->fs_info->delalloc_lock); total = btrfs_super_total_bytes(&root->fs_info->super_copy); used = btrfs_super_bytes_used(&root->fs_info->super_copy); if (for_del) @@ -116,7 +115,7 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, if (used + root->fs_info->delalloc_bytes + num_required > thresh) ret = -ENOSPC; - spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + spin_unlock(&root->fs_info->delalloc_lock); return ret; } @@ -1166,17 +1165,21 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { - unsigned long flags; + /* + * set_bit and clear bit hooks normally require _irqsave/restore + * but in this case, we are only testeing for the DELALLOC + * bit, which is only set or cleared with irqs on + */ if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + spin_lock(&root->fs_info->delalloc_lock); BTRFS_I(inode)->delalloc_bytes += end - start + 1; root->fs_info->delalloc_bytes += end - start + 1; if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) { list_add_tail(&BTRFS_I(inode)->delalloc_inodes, &root->fs_info->delalloc_inodes); } - spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + spin_unlock(&root->fs_info->delalloc_lock); } return 0; } @@ -1187,11 +1190,15 @@ static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, unsigned long old, unsigned long bits) { + /* + * set_bit and clear bit hooks normally require _irqsave/restore + * but in this case, we are only testeing for the DELALLOC + * bit, which is only set or cleared with irqs on + */ if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { struct btrfs_root *root = BTRFS_I(inode)->root; - unsigned long flags; - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + spin_lock(&root->fs_info->delalloc_lock); if (end - start + 1 > root->fs_info->delalloc_bytes) { printk("warning: delalloc account %Lu %Lu\n", end - start + 1, root->fs_info->delalloc_bytes); @@ -1205,7 +1212,7 @@ static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, !list_empty(&BTRFS_I(inode)->delalloc_inodes)) { list_del_init(&BTRFS_I(inode)->delalloc_inodes); } - spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + spin_unlock(&root->fs_info->delalloc_lock); } return 0; } @@ -4651,27 +4658,26 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) struct list_head *head = &root->fs_info->delalloc_inodes; struct btrfs_inode *binode; struct inode *inode; - unsigned long flags; if (root->fs_info->sb->s_flags & MS_RDONLY) return -EROFS; - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + spin_lock(&root->fs_info->delalloc_lock); while(!list_empty(head)) { binode = list_entry(head->next, struct btrfs_inode, delalloc_inodes); inode = igrab(&binode->vfs_inode); if (!inode) list_del_init(&binode->delalloc_inodes); - spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + spin_unlock(&root->fs_info->delalloc_lock); if (inode) { filemap_flush(inode->i_mapping); iput(inode); } cond_resched(); - spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); + spin_lock(&root->fs_info->delalloc_lock); } - spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); + spin_unlock(&root->fs_info->delalloc_lock); /* the filemap_flush will queue IO into the worker threads, but * we have to make sure the IO is actually started and that -- cgit v1.2.3-70-g09d2 From cad321ad529400c6ab24c501a67c3be720a0744c Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 17 Dec 2008 14:51:42 -0500 Subject: Btrfs: shift all end_io work to thread pools bio_end_io for reads without checksumming on and btree writes were happening without using async thread pools. This means the extent_io.c code had to use spin_lock_irq and friends on the rb tree locks for extent state. There were some irq safe vs unsafe lock inversions between the delallock lock and the extent state locks. This patch gets rid of them by moving all end_io code into the thread pools. To avoid contention and deadlocks between the data end_io processing and the metadata end_io processing yet another thread pool is added to finish off metadata writes. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 39 ++++++++++++++++++++++++++------------- fs/btrfs/extent_io.c | 51 ++++++++++++++++++++++++--------------------------- fs/btrfs/inode.c | 12 ++++++------ 4 files changed, 57 insertions(+), 46 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b89999de456..ccea0648e10 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -758,6 +758,7 @@ struct btrfs_fs_info { struct btrfs_workers delalloc_workers; struct btrfs_workers endio_workers; struct btrfs_workers endio_meta_workers; + struct btrfs_workers endio_meta_write_workers; struct btrfs_workers endio_write_workers; struct btrfs_workers submit_workers; /* diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 541a8279ac7..04f8d7080b1 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -447,8 +447,12 @@ static void end_workqueue_bio(struct bio *bio, int err) end_io_wq->work.flags = 0; if (bio->bi_rw & (1 << BIO_RW)) { - btrfs_queue_worker(&fs_info->endio_write_workers, - &end_io_wq->work); + if (end_io_wq->metadata) + btrfs_queue_worker(&fs_info->endio_meta_write_workers, + &end_io_wq->work); + else + btrfs_queue_worker(&fs_info->endio_write_workers, + &end_io_wq->work); } else { if (end_io_wq->metadata) btrfs_queue_worker(&fs_info->endio_meta_workers, @@ -624,23 +628,24 @@ static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) { - /* - * kthread helpers are used to submit writes so that checksumming - * can happen in parallel across all CPUs - */ + int ret; + + ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, + bio, 1); + BUG_ON(ret); + if (!(rw & (1 << BIO_RW))) { - int ret; /* * called for a read, do the setup so that checksum validation * can happen in the async kernel threads */ - ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, - bio, 1); - BUG_ON(ret); - return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 0); } + /* + * kthread helpers are used to submit writes so that checksumming + * can happen in parallel across all CPUs + */ return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, inode, rw, bio, mirror_num, 0, __btree_submit_bio_start, @@ -1350,12 +1355,13 @@ static void end_workqueue_fn(struct btrfs_work *work) bio = end_io_wq->bio; fs_info = end_io_wq->info; - /* metadata bios are special because the whole tree block must + /* metadata bio reads are special because the whole tree block must * be checksummed at once. This makes sure the entire block is in * ram and up to date before trying to verify things. For * blocksize <= pagesize, it is basically a noop */ - if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { + if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata && + !bio_ready_for_csum(bio)) { btrfs_queue_worker(&fs_info->endio_meta_workers, &end_io_wq->work); return; @@ -1668,6 +1674,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", fs_info->thread_pool_size); + btrfs_init_workers(&fs_info->endio_meta_write_workers, + "endio-meta-write", fs_info->thread_pool_size); btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", fs_info->thread_pool_size); @@ -1677,6 +1685,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, */ fs_info->endio_workers.idle_thresh = 4; fs_info->endio_write_workers.idle_thresh = 64; + fs_info->endio_meta_write_workers.idle_thresh = 64; btrfs_start_workers(&fs_info->workers, 1); btrfs_start_workers(&fs_info->submit_workers, 1); @@ -1685,6 +1694,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); btrfs_start_workers(&fs_info->endio_meta_workers, fs_info->thread_pool_size); + btrfs_start_workers(&fs_info->endio_meta_write_workers, + fs_info->thread_pool_size); btrfs_start_workers(&fs_info->endio_write_workers, fs_info->thread_pool_size); @@ -1866,6 +1877,7 @@ fail_sb_buffer: btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_meta_workers); + btrfs_stop_workers(&fs_info->endio_meta_write_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); fail_iput: @@ -2253,6 +2265,7 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->workers); btrfs_stop_workers(&fs_info->endio_workers); btrfs_stop_workers(&fs_info->endio_meta_workers); + btrfs_stop_workers(&fs_info->endio_meta_write_workers); btrfs_stop_workers(&fs_info->endio_write_workers); btrfs_stop_workers(&fs_info->submit_workers); diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 607f5ff2791..25ce2d18e5b 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -477,7 +477,6 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, struct extent_state *state; struct extent_state *prealloc = NULL; struct rb_node *node; - unsigned long flags; int err; int set = 0; @@ -488,7 +487,7 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); /* * this search will find the extents that end after * our range starts @@ -559,7 +558,7 @@ again: goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -568,7 +567,7 @@ out: search_again: if (start > end) goto out; - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -582,9 +581,9 @@ static int wait_on_state(struct extent_io_tree *tree, { DEFINE_WAIT(wait); prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); schedule(); - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); finish_wait(&state->wq, &wait); return 0; } @@ -599,7 +598,7 @@ int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) struct extent_state *state; struct rb_node *node; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); again: while (1) { /* @@ -628,13 +627,13 @@ again: break; if (need_resched()) { - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); cond_resched(); - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); } } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return 0; } EXPORT_SYMBOL(wait_extent_bit); @@ -668,7 +667,6 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int b struct extent_state *state; struct extent_state *prealloc = NULL; struct rb_node *node; - unsigned long flags; int err = 0; int set; u64 last_start; @@ -680,7 +678,7 @@ again: return -ENOMEM; } - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -800,7 +798,7 @@ again: goto search_again; out: - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (prealloc) free_extent_state(prealloc); @@ -809,7 +807,7 @@ out: search_again: if (start > end) goto out; - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); if (mask & __GFP_WAIT) cond_resched(); goto again; @@ -1021,7 +1019,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, struct extent_state *state; int ret = 1; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1044,7 +1042,7 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } EXPORT_SYMBOL(find_first_extent_bit); @@ -1097,7 +1095,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, u64 found = 0; u64 total_bytes = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after @@ -1134,7 +1132,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return found; } @@ -1391,7 +1389,7 @@ u64 count_range_bits(struct extent_io_tree *tree, return 0; } - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); if (cur_start == 0 && bits == EXTENT_DIRTY) { total_bytes = tree->dirty_bytes; goto out; @@ -1424,7 +1422,7 @@ u64 count_range_bits(struct extent_io_tree *tree, break; } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return total_bytes; } @@ -1501,7 +1499,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) struct extent_state *state; int ret = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1518,7 +1516,7 @@ int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) } state->private = private; out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } @@ -1528,7 +1526,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) struct extent_state *state; int ret = 0; - spin_lock_irq(&tree->lock); + spin_lock(&tree->lock); /* * this search will find all the extents that end after * our range starts. @@ -1545,7 +1543,7 @@ int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) } *private = state->private; out: - spin_unlock_irq(&tree->lock); + spin_unlock(&tree->lock); return ret; } @@ -1561,9 +1559,8 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, struct extent_state *state = NULL; struct rb_node *node; int bitset = 0; - unsigned long flags; - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&tree->lock); node = tree_search(tree, start); while (node && start <= end) { state = rb_entry(node, struct extent_state, rb_node); @@ -1594,7 +1591,7 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, break; } } - spin_unlock_irqrestore(&tree->lock, flags); + spin_unlock(&tree->lock); return bitset; } EXPORT_SYMBOL(test_range_bit); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0577e77e661..068bad46338 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1282,8 +1282,8 @@ static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, } /* - * extent_io.c submission hook. This does the right thing for csum calculation on write, - * or reading the csums from the tree before a read + * extent_io.c submission hook. This does the right thing for csum calculation + * on write, or reading the csums from the tree before a read */ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int mirror_num, unsigned long bio_flags) @@ -1292,11 +1292,11 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, int ret = 0; int skip_sum; + skip_sum = btrfs_test_flag(inode, NODATASUM); + ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); BUG_ON(ret); - skip_sum = btrfs_test_flag(inode, NODATASUM); - if (!(rw & (1 << BIO_RW))) { if (bio_flags & EXTENT_BIO_COMPRESSED) { return btrfs_submit_compressed_read(inode, bio, @@ -1648,13 +1648,13 @@ static int btrfs_io_failed_hook(struct bio *failed_bio, failrec->logical, failrec->len); failrec->last_mirror++; if (!state) { - spin_lock_irq(&BTRFS_I(inode)->io_tree.lock); + spin_lock(&BTRFS_I(inode)->io_tree.lock); state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree, failrec->start, EXTENT_LOCKED); if (state && state->start != failrec->start) state = NULL; - spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock); + spin_unlock(&BTRFS_I(inode)->io_tree.lock); } if (!state || failrec->last_mirror > num_copies) { set_state_private(failure_tree, failrec->start, 0); -- cgit v1.2.3-70-g09d2 From d397712bcc6a759a560fd247e6053ecae091f958 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Mon, 5 Jan 2009 21:25:51 -0500 Subject: Btrfs: Fix checkpatch.pl warnings There were many, most are fixed now. struct-funcs.c generates some warnings but these are bogus. Signed-off-by: Chris Mason --- fs/btrfs/acl.c | 5 +- fs/btrfs/async-thread.c | 6 +- fs/btrfs/compat.h | 4 +- fs/btrfs/compression.c | 12 +-- fs/btrfs/ctree.c | 217 ++++++++++++++++-------------------------- fs/btrfs/ctree.h | 30 +++--- fs/btrfs/dir-item.c | 2 +- fs/btrfs/disk-io.c | 187 +++++++++++++------------------------ fs/btrfs/export.c | 8 +- fs/btrfs/extent-tree.c | 223 +++++++++++++++++++++++--------------------- fs/btrfs/extent_io.c | 213 ++++++++++++++++++++---------------------- fs/btrfs/extent_map.c | 14 +-- fs/btrfs/file-item.c | 18 ++-- fs/btrfs/file.c | 49 +++++----- fs/btrfs/free-space-cache.c | 37 ++++---- fs/btrfs/inode-map.c | 1 - fs/btrfs/inode.c | 173 +++++++++++++++++----------------- fs/btrfs/ioctl.c | 37 ++++---- fs/btrfs/locking.c | 5 +- fs/btrfs/ordered-data.c | 34 +++---- fs/btrfs/print-tree.c | 73 +++++++++------ fs/btrfs/ref-cache.c | 12 +-- fs/btrfs/root-tree.c | 17 ++-- fs/btrfs/struct-funcs.c | 4 +- fs/btrfs/super.c | 25 +++-- fs/btrfs/sysfs.c | 6 +- fs/btrfs/transaction.c | 45 ++++----- fs/btrfs/transaction.h | 6 +- fs/btrfs/tree-defrag.c | 9 +- fs/btrfs/tree-log.c | 70 +++++++------- fs/btrfs/volumes.c | 78 ++++++++-------- fs/btrfs/xattr.c | 3 +- fs/btrfs/zlib.c | 45 ++++----- 33 files changed, 770 insertions(+), 898 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c index 867eaf1f8ef..1d53b62dbba 100644 --- a/fs/btrfs/acl.c +++ b/fs/btrfs/acl.c @@ -161,8 +161,7 @@ static int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type) ret = __btrfs_setxattr(inode, name, value, size, 0); out: - if (value) - kfree(value); + kfree(value); if (!ret) btrfs_update_cached_acl(inode, p_acl, acl); @@ -213,7 +212,7 @@ static int btrfs_xattr_acl_default_get(struct inode *inode, const char *name, } static int btrfs_xattr_acl_default_set(struct inode *inode, const char *name, - const void *value, size_t size, int flags) + const void *value, size_t size, int flags) { return btrfs_xattr_set_acl(inode, ACL_TYPE_DEFAULT, value, size); } diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c index 4229450b759..8e2fec05dbe 100644 --- a/fs/btrfs/async-thread.c +++ b/fs/btrfs/async-thread.c @@ -104,7 +104,7 @@ static noinline int run_ordered_completions(struct btrfs_workers *workers, spin_lock_irqsave(&workers->lock, flags); - while(!list_empty(&workers->order_list)) { + while (!list_empty(&workers->order_list)) { work = list_entry(workers->order_list.next, struct btrfs_work, order_list); @@ -143,7 +143,7 @@ static int worker_loop(void *arg) struct btrfs_work *work; do { spin_lock_irq(&worker->lock); - while(!list_empty(&worker->pending)) { + while (!list_empty(&worker->pending)) { cur = worker->pending.next; work = list_entry(cur, struct btrfs_work, list); list_del(&work->list); @@ -188,7 +188,7 @@ int btrfs_stop_workers(struct btrfs_workers *workers) struct btrfs_worker_thread *worker; list_splice_init(&workers->idle_list, &workers->worker_list); - while(!list_empty(&workers->worker_list)) { + while (!list_empty(&workers->worker_list)) { cur = workers->worker_list.next; worker = list_entry(cur, struct btrfs_worker_thread, worker_list); diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h index 75e4426d6fb..594d60bdd3c 100644 --- a/fs/btrfs/compat.h +++ b/fs/btrfs/compat.h @@ -4,7 +4,7 @@ #define btrfs_drop_nlink(inode) drop_nlink(inode) #define btrfs_inc_nlink(inode) inc_nlink(inode) -#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,27) +#if LINUX_VERSION_CODE <= KERNEL_VERSION(2, 6, 27) static inline struct dentry *d_obtain_alias(struct inode *inode) { struct dentry *d; @@ -21,7 +21,7 @@ static inline struct dentry *d_obtain_alias(struct inode *inode) } #endif -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) +#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 28) # define __pagevec_lru_add_file __pagevec_lru_add # define open_bdev_exclusive open_bdev_excl # define close_bdev_exclusive(bdev, mode) close_bdev_excl(bdev) diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c index 2436163d543..ee848d8585d 100644 --- a/fs/btrfs/compression.c +++ b/fs/btrfs/compression.c @@ -137,7 +137,8 @@ static int check_compressed_csum(struct inode *inode, kunmap_atomic(kaddr, KM_USER0); if (csum != *cb_sum) { - printk("btrfs csum failed ino %lu extent %llu csum %u " + printk(KERN_INFO "btrfs csum failed ino %lu " + "extent %llu csum %u " "wanted %u mirror %d\n", inode->i_ino, (unsigned long long)disk_start, csum, *cb_sum, cb->mirror_num); @@ -217,7 +218,7 @@ csum_failed: * we have verified the checksum already, set page * checked so the end_io handlers know about it */ - while(bio_index < cb->orig_bio->bi_vcnt) { + while (bio_index < cb->orig_bio->bi_vcnt) { SetPageChecked(bvec->bv_page); bvec++; bio_index++; @@ -246,7 +247,7 @@ static noinline int end_compressed_writeback(struct inode *inode, u64 start, int i; int ret; - while(nr_pages > 0) { + while (nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nr_pages, ARRAY_SIZE(pages)), pages); @@ -463,7 +464,7 @@ static noinline int add_ra_bio_pages(struct inode *inode, end_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT; pagevec_init(&pvec, 0); - while(last_offset < compressed_end) { + while (last_offset < compressed_end) { page_index = last_offset >> PAGE_CACHE_SHIFT; if (page_index > end_index) @@ -697,9 +698,8 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio, ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); BUG_ON(ret); - if (!btrfs_test_flag(inode, NODATASUM)) { + if (!btrfs_test_flag(inode, NODATASUM)) btrfs_lookup_bio_sums(root, inode, comp_bio, sums); - } ret = btrfs_map_bio(root, READ, comp_bio, mirror_num, 0); BUG_ON(ret); diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 7fad2e3ad6f..9e46c077681 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -67,7 +67,7 @@ void btrfs_free_path(struct btrfs_path *p) * * It is safe to call this on paths that no locks or extent buffers held. */ -void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) +noinline void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p) { int i; @@ -112,7 +112,7 @@ struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root) { struct extent_buffer *eb; - while(1) { + while (1) { eb = btrfs_root_node(root); btrfs_tree_lock(eb); @@ -202,22 +202,22 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans, } /* - * does the dirty work in cow of a single block. The parent block - * (if supplied) is updated to point to the new cow copy. The new - * buffer is marked dirty and returned locked. If you modify the block - * it needs to be marked dirty again. + * does the dirty work in cow of a single block. The parent block (if + * supplied) is updated to point to the new cow copy. The new buffer is marked + * dirty and returned locked. If you modify the block it needs to be marked + * dirty again. * * search_start -- an allocation hint for the new block * - * empty_size -- a hint that you plan on doing more cow. This is the size in bytes - * the allocator should try to find free next to the block it returns. This is - * just a hint and may be ignored by the allocator. + * empty_size -- a hint that you plan on doing more cow. This is the size in + * bytes the allocator should try to find free next to the block it returns. + * This is just a hint and may be ignored by the allocator. * * prealloc_dest -- if you have already reserved a destination for the cow, - * this uses that block instead of allocating a new one. btrfs_alloc_reserved_extent - * is used to finish the allocation. + * this uses that block instead of allocating a new one. + * btrfs_alloc_reserved_extent is used to finish the allocation. */ -static int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, +static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, @@ -366,7 +366,7 @@ static int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans, * This version of it has extra checks so that a block isn't cow'd more than * once per transaction, as long as it hasn't been written yet */ -int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, +noinline int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *buf, struct extent_buffer *parent, int parent_slot, struct extent_buffer **cow_ret, u64 prealloc_dest) @@ -375,13 +375,16 @@ int noinline btrfs_cow_block(struct btrfs_trans_handle *trans, int ret; if (trans->transaction != root->fs_info->running_transaction) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, + printk(KERN_CRIT "trans %llu running %llu\n", + (unsigned long long)trans->transid, + (unsigned long long) root->fs_info->running_transaction->transid); WARN_ON(1); } if (trans->transid != root->fs_info->generation) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, - root->fs_info->generation); + printk(KERN_CRIT "trans %llu running %llu\n", + (unsigned long long)trans->transid, + (unsigned long long)root->fs_info->generation); WARN_ON(1); } @@ -489,16 +492,10 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, if (cache_only && parent_level != 1) return 0; - if (trans->transaction != root->fs_info->running_transaction) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, - root->fs_info->running_transaction->transid); + if (trans->transaction != root->fs_info->running_transaction) WARN_ON(1); - } - if (trans->transid != root->fs_info->generation) { - printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid, - root->fs_info->generation); + if (trans->transid != root->fs_info->generation) WARN_ON(1); - } parent_nritems = btrfs_header_nritems(parent); blocksize = btrfs_level_size(root, parent_level - 1); @@ -681,51 +678,18 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, BUG_ON(btrfs_node_blockptr(parent, parent_slot) != btrfs_header_bytenr(leaf)); } -#if 0 - for (i = 0; nritems > 1 && i < nritems - 2; i++) { - btrfs_item_key_to_cpu(leaf, &cpukey, i + 1); - btrfs_item_key(leaf, &leaf_key, i); - if (comp_keys(&leaf_key, &cpukey) >= 0) { - btrfs_print_leaf(root, leaf); - printk("slot %d offset bad key\n", i); - BUG_ON(1); - } - if (btrfs_item_offset_nr(leaf, i) != - btrfs_item_end_nr(leaf, i + 1)) { - btrfs_print_leaf(root, leaf); - printk("slot %d offset bad\n", i); - BUG_ON(1); - } - if (i == 0) { - if (btrfs_item_offset_nr(leaf, i) + - btrfs_item_size_nr(leaf, i) != - BTRFS_LEAF_DATA_SIZE(root)) { - btrfs_print_leaf(root, leaf); - printk("slot %d first offset bad\n", i); - BUG_ON(1); - } - } - } - if (nritems > 0) { - if (btrfs_item_size_nr(leaf, nritems - 1) > 4096) { - btrfs_print_leaf(root, leaf); - printk("slot %d bad size \n", nritems - 1); - BUG_ON(1); - } - } -#endif if (slot != 0 && slot < nritems - 1) { btrfs_item_key(leaf, &leaf_key, slot); btrfs_item_key_to_cpu(leaf, &cpukey, slot - 1); if (comp_keys(&leaf_key, &cpukey) <= 0) { btrfs_print_leaf(root, leaf); - printk("slot %d offset bad key\n", slot); + printk(KERN_CRIT "slot %d offset bad key\n", slot); BUG_ON(1); } if (btrfs_item_offset_nr(leaf, slot - 1) != btrfs_item_end_nr(leaf, slot)) { btrfs_print_leaf(root, leaf); - printk("slot %d offset bad\n", slot); + printk(KERN_CRIT "slot %d offset bad\n", slot); BUG_ON(1); } } @@ -736,7 +700,7 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, if (btrfs_item_offset_nr(leaf, slot) != btrfs_item_end_nr(leaf, slot + 1)) { btrfs_print_leaf(root, leaf); - printk("slot %d offset bad\n", slot); + printk(KERN_CRIT "slot %d offset bad\n", slot); BUG_ON(1); } } @@ -745,30 +709,10 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path, return 0; } -static int noinline check_block(struct btrfs_root *root, +static noinline int check_block(struct btrfs_root *root, struct btrfs_path *path, int level) { - u64 found_start; return 0; - if (btrfs_header_level(path->nodes[level]) != level) - printk("warning: bad level %Lu wanted %d found %d\n", - path->nodes[level]->start, level, - btrfs_header_level(path->nodes[level])); - found_start = btrfs_header_bytenr(path->nodes[level]); - if (found_start != path->nodes[level]->start) { - printk("warning: bad bytentr %Lu found %Lu\n", - path->nodes[level]->start, found_start); - } -#if 0 - struct extent_buffer *buf = path->nodes[level]; - - if (memcmp_extent_buffer(buf, root->fs_info->fsid, - (unsigned long)btrfs_header_fsid(buf), - BTRFS_FSID_SIZE)) { - printk("warning bad block %Lu\n", buf->start); - return 1; - } -#endif if (level == 0) return check_leaf(root, path, level); return check_node(root, path, level); @@ -802,7 +746,7 @@ static noinline int generic_bin_search(struct extent_buffer *eb, unsigned long map_len = 0; int err; - while(low < high) { + while (low < high) { mid = (low + high) / 2; offset = p + mid * item_size; @@ -1130,7 +1074,7 @@ enospc: * when they are completely full. This is also done top down, so we * have to be pessimistic. */ -static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans, +static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { @@ -1296,7 +1240,7 @@ static noinline void reada_for_search(struct btrfs_root *root, nritems = btrfs_header_nritems(node); nr = slot; - while(1) { + while (1) { if (direction < 0) { if (nr == 0) break; @@ -1322,7 +1266,8 @@ static noinline void reada_for_search(struct btrfs_root *root, nscan++; if (path->reada < 2 && (nread > (64 * 1024) || nscan > 32)) break; - if(nread > (256 * 1024) || nscan > 128) + + if (nread > (256 * 1024) || nscan > 128) break; if (search < lowest_read) @@ -1333,17 +1278,17 @@ static noinline void reada_for_search(struct btrfs_root *root, } /* - * when we walk down the tree, it is usually safe to unlock the higher layers in - * the tree. The exceptions are when our path goes through slot 0, because operations - * on the tree might require changing key pointers higher up in the tree. + * when we walk down the tree, it is usually safe to unlock the higher layers + * in the tree. The exceptions are when our path goes through slot 0, because + * operations on the tree might require changing key pointers higher up in the + * tree. * - * callers might also have set path->keep_locks, which tells this code to - * keep the lock if the path points to the last slot in the block. This is - * part of walking through the tree, and selecting the next slot in the higher - * block. + * callers might also have set path->keep_locks, which tells this code to keep + * the lock if the path points to the last slot in the block. This is part of + * walking through the tree, and selecting the next slot in the higher block. * - * lowest_unlock sets the lowest level in the tree we're allowed to unlock. - * so if lowest_unlock is 1, level 0 won't be unlocked + * lowest_unlock sets the lowest level in the tree we're allowed to unlock. so + * if lowest_unlock is 1, level 0 won't be unlocked */ static noinline void unlock_up(struct btrfs_path *path, int level, int lowest_unlock) @@ -1832,9 +1777,8 @@ static int push_node_left(struct btrfs_trans_handle *trans, if (!empty && src_nritems <= 8) return 1; - if (push_items <= 0) { + if (push_items <= 0) return 1; - } if (empty) { push_items = min(src_nritems, push_items); @@ -1854,7 +1798,7 @@ static int push_node_left(struct btrfs_trans_handle *trans, copy_extent_buffer(dst, src, btrfs_node_key_ptr_offset(dst_nritems), btrfs_node_key_ptr_offset(0), - push_items * sizeof(struct btrfs_key_ptr)); + push_items * sizeof(struct btrfs_key_ptr)); if (push_items < src_nritems) { memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0), @@ -1899,19 +1843,16 @@ static int balance_node_right(struct btrfs_trans_handle *trans, src_nritems = btrfs_header_nritems(src); dst_nritems = btrfs_header_nritems(dst); push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems; - if (push_items <= 0) { + if (push_items <= 0) return 1; - } - if (src_nritems < 4) { + if (src_nritems < 4) return 1; - } max_push = src_nritems / 2 + 1; /* don't try to empty the node */ - if (max_push >= src_nritems) { + if (max_push >= src_nritems) return 1; - } if (max_push < push_items) push_items = max_push; @@ -1924,7 +1865,7 @@ static int balance_node_right(struct btrfs_trans_handle *trans, copy_extent_buffer(dst, src, btrfs_node_key_ptr_offset(0), btrfs_node_key_ptr_offset(src_nritems - push_items), - push_items * sizeof(struct btrfs_key_ptr)); + push_items * sizeof(struct btrfs_key_ptr)); btrfs_set_header_nritems(src, src_nritems - push_items); btrfs_set_header_nritems(dst, dst_nritems + push_items); @@ -1945,7 +1886,7 @@ static int balance_node_right(struct btrfs_trans_handle *trans, * * returns zero on success or < 0 on failure. */ -static int noinline insert_new_root(struct btrfs_trans_handle *trans, +static noinline int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level) { @@ -2176,14 +2117,15 @@ static int leaf_space_used(struct extent_buffer *l, int start, int nr) * the start of the leaf data. IOW, how much room * the leaf has left for both items and data */ -int noinline btrfs_leaf_free_space(struct btrfs_root *root, +noinline int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf) { int nritems = btrfs_header_nritems(leaf); int ret; ret = BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems); if (ret < 0) { - printk("leaf free space ret %d, leaf data size %lu, used %d nritems %d\n", + printk(KERN_CRIT "leaf free space ret %d, leaf data size %lu, " + "used %d nritems %d\n", ret, (unsigned long) BTRFS_LEAF_DATA_SIZE(root), leaf_space_used(leaf, 0, nritems), nritems); } @@ -2219,9 +2161,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root int ret; slot = path->slots[1]; - if (!path->nodes[1]) { + if (!path->nodes[1]) return 1; - } + upper = path->nodes[1]; if (slot >= btrfs_header_nritems(upper) - 1) return 1; @@ -2418,9 +2360,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root return 1; right_nritems = btrfs_header_nritems(right); - if (right_nritems == 0) { + if (right_nritems == 0) return 1; - } WARN_ON(!btrfs_tree_locked(path->nodes[1])); @@ -2502,7 +2443,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root push_items * sizeof(struct btrfs_item)); push_space = BTRFS_LEAF_DATA_SIZE(root) - - btrfs_item_offset_nr(right, push_items -1); + btrfs_item_offset_nr(right, push_items - 1); copy_extent_buffer(left, right, btrfs_leaf_data(left) + leaf_data_end(root, left) - push_space, @@ -2537,7 +2478,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root /* fixup right node */ if (push_items > right_nritems) { - printk("push items %d nr %u\n", push_items, right_nritems); + printk(KERN_CRIT "push items %d nr %u\n", push_items, + right_nritems); WARN_ON(1); } @@ -2640,9 +2582,8 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans, /* first try to make some room by pushing left and right */ if (data_size && ins_key->type != BTRFS_DIR_ITEM_KEY) { wret = push_leaf_right(trans, root, path, data_size, 0); - if (wret < 0) { + if (wret < 0) return wret; - } if (wret) { wret = push_leaf_left(trans, root, path, data_size, 0); if (wret < 0) @@ -2665,7 +2606,7 @@ again: l = path->nodes[0]; slot = path->slots[0]; nritems = btrfs_header_nritems(l); - mid = (nritems + 1)/ 2; + mid = (nritems + 1) / 2; right = btrfs_alloc_free_block(trans, root, root->leafsize, path->nodes[1]->start, @@ -2734,7 +2675,7 @@ again: path->slots[0] = 0; if (path->slots[1] == 0) { wret = fixup_low_keys(trans, root, - path, &disk_key, 1); + path, &disk_key, 1); if (wret) ret = wret; } @@ -3033,8 +2974,8 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans, BTRFS_FILE_EXTENT_INLINE) { ptr = btrfs_item_ptr_offset(leaf, slot); memmove_extent_buffer(leaf, ptr, - (unsigned long)fi, - offsetof(struct btrfs_file_extent_item, + (unsigned long)fi, + offsetof(struct btrfs_file_extent_item, disk_bytenr)); } } @@ -3096,7 +3037,8 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, BUG_ON(slot < 0); if (slot >= nritems) { btrfs_print_leaf(root, leaf); - printk("slot %d too large, nritems %d\n", slot, nritems); + printk(KERN_CRIT "slot %d too large, nritems %d\n", + slot, nritems); BUG_ON(1); } @@ -3218,7 +3160,7 @@ int btrfs_insert_some_items(struct btrfs_trans_handle *trans, if (old_data < data_end) { btrfs_print_leaf(root, leaf); - printk("slot %d old_data %d data_end %d\n", + printk(KERN_CRIT "slot %d old_data %d data_end %d\n", slot, old_data, data_end); BUG_ON(1); } @@ -3317,9 +3259,8 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, unsigned int data_end; struct btrfs_disk_key disk_key; - for (i = 0; i < nr; i++) { + for (i = 0; i < nr; i++) total_data += data_size[i]; - } total_size = total_data + (nr * sizeof(struct btrfs_item)); ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); @@ -3336,7 +3277,7 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, if (btrfs_leaf_free_space(root, leaf) < total_size) { btrfs_print_leaf(root, leaf); - printk("not enough freespace need %u have %d\n", + printk(KERN_CRIT "not enough freespace need %u have %d\n", total_size, btrfs_leaf_free_space(root, leaf)); BUG(); } @@ -3349,7 +3290,7 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, if (old_data < data_end) { btrfs_print_leaf(root, leaf); - printk("slot %d old_data %d data_end %d\n", + printk(KERN_CRIT "slot %d old_data %d data_end %d\n", slot, old_data, data_end); BUG_ON(1); } @@ -3457,7 +3398,7 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, int wret; nritems = btrfs_header_nritems(parent); - if (slot != nritems -1) { + if (slot != nritems - 1) { memmove_extent_buffer(parent, btrfs_node_key_ptr_offset(slot), btrfs_node_key_ptr_offset(slot + 1), @@ -3614,7 +3555,8 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, if (btrfs_header_nritems(leaf) == 0) { path->slots[1] = slot; - ret = btrfs_del_leaf(trans, root, path, leaf->start); + ret = btrfs_del_leaf(trans, root, path, + leaf->start); BUG_ON(ret); free_extent_buffer(leaf); } else { @@ -3717,7 +3659,7 @@ again: ret = 1; goto out; } - while(1) { + while (1) { nritems = btrfs_header_nritems(cur); level = btrfs_header_level(cur); sret = bin_search(cur, min_key, level, &slot); @@ -3738,7 +3680,7 @@ again: * min_trans parameters. If it isn't in cache or is too * old, skip to the next one. */ - while(slot < nritems) { + while (slot < nritems) { u64 blockptr; u64 gen; struct extent_buffer *tmp; @@ -3830,7 +3772,7 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, struct extent_buffer *c; WARN_ON(!path->keep_locks); - while(level < BTRFS_MAX_LEVEL) { + while (level < BTRFS_MAX_LEVEL) { if (!path->nodes[level]) return 1; @@ -3839,9 +3781,8 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, next: if (slot >= btrfs_header_nritems(c)) { level++; - if (level == BTRFS_MAX_LEVEL) { + if (level == BTRFS_MAX_LEVEL) return 1; - } continue; } if (level == 0) @@ -3889,9 +3830,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) int ret; nritems = btrfs_header_nritems(path->nodes[0]); - if (nritems == 0) { + if (nritems == 0) return 1; - } btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1); @@ -3915,7 +3855,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) goto done; } - while(level < BTRFS_MAX_LEVEL) { + while (level < BTRFS_MAX_LEVEL) { if (!path->nodes[level]) return 1; @@ -3923,9 +3863,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) c = path->nodes[level]; if (slot >= btrfs_header_nritems(c)) { level++; - if (level == BTRFS_MAX_LEVEL) { + if (level == BTRFS_MAX_LEVEL) return 1; - } continue; } @@ -3946,7 +3885,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) break; } path->slots[level] = slot; - while(1) { + while (1) { level--; c = path->nodes[level]; if (path->locks[level]) @@ -3986,7 +3925,7 @@ int btrfs_previous_item(struct btrfs_root *root, u32 nritems; int ret; - while(1) { + while (1) { if (path->slots[0] == 0) { ret = btrfs_prev_leaf(root, path); if (ret != 0) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ccea0648e10..eee060f8811 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -126,7 +126,6 @@ struct btrfs_ordered_sum; static int btrfs_csum_sizes[] = { 4, 0 }; /* four bytes for CRC32 */ -//#define BTRFS_CRC32_SIZE 4 #define BTRFS_EMPTY_DIR_SIZE 0 #define BTRFS_FT_UNKNOWN 0 @@ -283,8 +282,8 @@ struct btrfs_header { } __attribute__ ((__packed__)); #define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->nodesize - \ - sizeof(struct btrfs_header)) / \ - sizeof(struct btrfs_key_ptr)) + sizeof(struct btrfs_header)) / \ + sizeof(struct btrfs_key_ptr)) #define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header)) #define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->leafsize)) #define BTRFS_MAX_INLINE_DATA_SIZE(r) (BTRFS_LEAF_DATA_SIZE(r) - \ @@ -1512,7 +1511,7 @@ static inline struct btrfs_header *btrfs_buffer_header(struct extent_buffer *eb) static inline int btrfs_is_leaf(struct extent_buffer *eb) { - return (btrfs_header_level(eb) == 0); + return btrfs_header_level(eb) == 0; } /* struct btrfs_root_item */ @@ -1597,8 +1596,8 @@ static inline unsigned long btrfs_leaf_data(struct extent_buffer *l) /* struct btrfs_file_extent_item */ BTRFS_SETGET_FUNCS(file_extent_type, struct btrfs_file_extent_item, type, 8); -static inline unsigned long btrfs_file_extent_inline_start(struct - btrfs_file_extent_item *e) +static inline unsigned long +btrfs_file_extent_inline_start(struct btrfs_file_extent_item *e) { unsigned long offset = (unsigned long)e; offset += offsetof(struct btrfs_file_extent_item, disk_bytenr); @@ -1660,20 +1659,20 @@ static inline int btrfs_set_root_name(struct btrfs_root *root, const char *name, int len) { /* if we already have a name just free it */ - if (root->name) - kfree(root->name); + kfree(root->name); root->name = kmalloc(len+1, GFP_KERNEL); if (!root->name) return -ENOMEM; memcpy(root->name, name, len); - root->name[len] ='\0'; + root->name[len] = '\0'; return 0; } -static inline u32 btrfs_level_size(struct btrfs_root *root, int level) { +static inline u32 btrfs_level_size(struct btrfs_root *root, int level) +{ if (level == 0) return root->leafsize; return root->nodesize; @@ -1707,9 +1706,9 @@ int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans, int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy); -struct btrfs_block_group_cache *btrfs_lookup_block_group(struct - btrfs_fs_info *info, - u64 bytenr); +struct btrfs_block_group_cache *btrfs_lookup_block_group( + struct btrfs_fs_info *info, + u64 bytenr); u64 btrfs_find_block_group(struct btrfs_root *root, u64 search_start, u64 search_hint, int owner); struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, @@ -1908,8 +1907,9 @@ int btrfs_search_root(struct btrfs_root *root, u64 search_start, int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, struct btrfs_root *latest_root); /* dir-item.c */ -int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root - *root, const char *name, int name_len, u64 dir, +int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, + struct btrfs_root *root, const char *name, + int name_len, u64 dir, struct btrfs_key *location, u8 type, u64 index); struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c index 5040b71f190..926a0b287a7 100644 --- a/fs/btrfs/dir-item.c +++ b/fs/btrfs/dir-item.c @@ -333,7 +333,7 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root, leaf = path->nodes[0]; dir_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dir_item); total_len = btrfs_item_size_nr(leaf, path->slots[0]); - while(cur < total_len) { + while (cur < total_len) { this_len = sizeof(*dir_item) + btrfs_dir_name_len(leaf, dir_item) + btrfs_dir_data_len(leaf, dir_item); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index dae25e78a6b..81a313874ae 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -23,7 +23,7 @@ #include #include #include -#include // for block_sync_page +#include #include #include #include @@ -40,19 +40,6 @@ #include "ref-cache.h" #include "tree-log.h" -#if 0 -static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) -{ - if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { - printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", - (unsigned long long)extent_buffer_blocknr(buf), - (unsigned long long)btrfs_header_blocknr(buf)); - return 1; - } - return 0; -} -#endif - static struct extent_io_ops btree_extent_io_ops; static void end_workqueue_fn(struct btrfs_work *work); @@ -128,23 +115,13 @@ static struct extent_map *btree_get_extent(struct inode *inode, u64 failed_start = em->start; u64 failed_len = em->len; - printk("failed to insert %Lu %Lu -> %Lu into tree\n", - em->start, em->len, em->block_start); free_extent_map(em); em = lookup_extent_mapping(em_tree, start, len); if (em) { - printk("after failing, found %Lu %Lu %Lu\n", - em->start, em->len, em->block_start); ret = 0; } else { em = lookup_extent_mapping(em_tree, failed_start, failed_len); - if (em) { - printk("double failure lookup gives us " - "%Lu %Lu -> %Lu\n", em->start, - em->len, em->block_start); - free_extent_map(em); - } ret = -EIO; } } else if (ret) { @@ -191,15 +168,12 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, unsigned long inline_result; len = buf->len - offset; - while(len > 0) { + while (len > 0) { err = map_private_extent_buffer(buf, offset, 32, &map_token, &kaddr, &map_start, &map_len, KM_USER0); - if (err) { - printk("failed to map extent buffer! %lu\n", - offset); + if (err) return 1; - } cur_len = min(len, map_len - (offset - map_start)); crc = btrfs_csum_data(root, kaddr + offset - map_start, crc, cur_len); @@ -218,15 +192,14 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, btrfs_csum_final(crc, result); if (verify) { - /* FIXME, this is not good */ if (memcmp_extent_buffer(buf, result, 0, csum_size)) { u32 val; u32 found = 0; memcpy(&found, result, csum_size); read_extent_buffer(buf, &val, 0, csum_size); - printk("btrfs: %s checksum verify failed on %llu " - "wanted %X found %X level %d\n", + printk(KERN_INFO "btrfs: %s checksum verify failed " + "on %llu wanted %X found %X level %d\n", root->fs_info->sb->s_id, buf->start, val, found, btrfs_header_level(buf)); if (result != (char *)&inline_result) @@ -293,7 +266,7 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root, if (!ret && !verify_parent_transid(io_tree, eb, parent_transid)) return ret; -printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num); + num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, eb->start, eb->len); if (num_copies == 1) @@ -307,9 +280,10 @@ printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror } /* - * checksum a dirty tree block before IO. This has extra checks to make - * sure we only fill in the checksum field in the first page of a multi-page block + * checksum a dirty tree block before IO. This has extra checks to make sure + * we only fill in the checksum field in the first page of a multi-page block */ + static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) { struct extent_io_tree *tree; @@ -327,28 +301,22 @@ static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) if (!page->private) goto out; len = page->private >> 2; - if (len == 0) { - WARN_ON(1); - } + WARN_ON(len == 0); + eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, btrfs_header_generation(eb)); BUG_ON(ret); found_start = btrfs_header_bytenr(eb); if (found_start != start) { - printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", - start, found_start, len); WARN_ON(1); goto err; } if (eb->first_page != page) { - printk("bad first page %lu %lu\n", eb->first_page->index, - page->index); WARN_ON(1); goto err; } if (!PageUptodate(page)) { - printk("csum not up to date page %lu\n", page->index); WARN_ON(1); goto err; } @@ -396,29 +364,30 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, goto out; if (!page->private) goto out; + len = page->private >> 2; - if (len == 0) { - WARN_ON(1); - } + WARN_ON(len == 0); + eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); found_start = btrfs_header_bytenr(eb); if (found_start != start) { - printk("bad tree block start %llu %llu\n", + printk(KERN_INFO "btrfs bad tree block start %llu %llu\n", (unsigned long long)found_start, (unsigned long long)eb->start); ret = -EIO; goto err; } if (eb->first_page != page) { - printk("bad first page %lu %lu\n", eb->first_page->index, - page->index); + printk(KERN_INFO "btrfs bad first page %lu %lu\n", + eb->first_page->index, page->index); WARN_ON(1); ret = -EIO; goto err; } if (check_tree_block_fsid(root, eb)) { - printk("bad fsid on block %Lu\n", eb->start); + printk(KERN_INFO "btrfs bad fsid on block %llu\n", + (unsigned long long)eb->start); ret = -EIO; goto err; } @@ -578,7 +547,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, HZ/10); } #endif - while(atomic_read(&fs_info->async_submit_draining) && + while (atomic_read(&fs_info->async_submit_draining) && atomic_read(&fs_info->nr_async_submits)) { wait_event(fs_info->async_submit_wait, (atomic_read(&fs_info->nr_async_submits) == 0)); @@ -594,7 +563,7 @@ static int btree_csum_one_bio(struct bio *bio) struct btrfs_root *root; WARN_ON(bio->bi_vcnt <= 0); - while(bio_index < bio->bi_vcnt) { + while (bio_index < bio->bi_vcnt) { root = BTRFS_I(bvec->bv_page->mapping->host)->root; csum_dirty_buffer(root, bvec->bv_page); bio_index++; @@ -680,9 +649,8 @@ static int btree_writepages(struct address_space *mapping, num_dirty = count_range_bits(tree, &start, (u64)-1, thresh, EXTENT_DIRTY); - if (num_dirty < thresh) { + if (num_dirty < thresh) return 0; - } } return extent_writepages(tree, mapping, btree_get_extent, wbc); } @@ -701,15 +669,14 @@ static int btree_releasepage(struct page *page, gfp_t gfp_flags) int ret; if (PageWriteback(page) || PageDirty(page)) - return 0; + return 0; tree = &BTRFS_I(page->mapping->host)->io_tree; map = &BTRFS_I(page->mapping->host)->extent_tree; ret = try_release_extent_state(map, tree, page, gfp_flags); - if (!ret) { + if (!ret) return 0; - } ret = try_release_extent_buffer(tree, page); if (ret == 1) { @@ -728,8 +695,8 @@ static void btree_invalidatepage(struct page *page, unsigned long offset) extent_invalidatepage(tree, page, offset); btree_releasepage(page, GFP_NOFS); if (PagePrivate(page)) { - printk("warning page private not zero on page %Lu\n", - page_offset(page)); + printk(KERN_WARNING "btrfs warning page private not zero " + "on page %llu\n", (unsigned long long)page_offset(page)); ClearPagePrivate(page); set_page_private(page, 0); page_cache_release(page); @@ -813,7 +780,7 @@ int btrfs_write_tree_block(struct extent_buffer *buf) int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) { return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, - buf->start, buf->start + buf->len -1); + buf->start, buf->start + buf->len - 1); } struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, @@ -832,11 +799,10 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); - if (ret == 0) { + if (ret == 0) buf->flags |= EXTENT_UPTODATE; - } else { + else WARN_ON(1); - } return buf; } @@ -944,7 +910,7 @@ int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, if (!log_root_tree) return 0; - while(1) { + while (1) { ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, 0, &start, &end, EXTENT_DIRTY); if (ret) @@ -1165,24 +1131,6 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, root->in_sysfs = 1; return root; } -#if 0 -static int add_hasher(struct btrfs_fs_info *info, char *type) { - struct btrfs_hasher *hasher; - - hasher = kmalloc(sizeof(*hasher), GFP_NOFS); - if (!hasher) - return -ENOMEM; - hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); - if (!hasher->hash_tfm) { - kfree(hasher); - return -EINVAL; - } - spin_lock(&info->hash_lock); - list_add(&hasher->list, &info->hashers); - spin_unlock(&info->hash_lock); - return 0; -} -#endif static int btrfs_congested_fn(void *congested_data, int bdi_bits) { @@ -1226,9 +1174,8 @@ static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) continue; bdi = blk_get_backing_dev_info(device->bdev); - if (bdi->unplug_io_fn) { + if (bdi->unplug_io_fn) bdi->unplug_io_fn(bdi, page); - } } } @@ -1420,8 +1367,9 @@ static int transaction_kthread(void *arg) mutex_lock(&root->fs_info->transaction_kthread_mutex); if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { - printk("btrfs: total reference cache size %Lu\n", - root->fs_info->total_ref_cache_size); + printk(KERN_INFO "btrfs: total reference cache " + "size %llu\n", + root->fs_info->total_ref_cache_size); } mutex_lock(&root->fs_info->trans_mutex); @@ -1592,14 +1540,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, atomic_set(&fs_info->tree_log_writers, 0); fs_info->tree_log_transid = 0; -#if 0 - ret = add_hasher(fs_info, "crc32c"); - if (ret) { - printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); - err = -ENOMEM; - goto fail_iput; - } -#endif __setup_root(4096, 4096, 4096, 4096, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); @@ -1720,7 +1660,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, sizeof(disk_super->magic))) { - printk("btrfs: valid FS not found on %s\n", sb->s_id); + printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id); goto fail_sb_buffer; } @@ -1728,8 +1668,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, ret = btrfs_read_sys_array(tree_root); mutex_unlock(&fs_info->chunk_mutex); if (ret) { - printk("btrfs: failed to read the system array on %s\n", - sb->s_id); + printk(KERN_WARNING "btrfs: failed to read the system " + "array on %s\n", sb->s_id); goto fail_sys_array; } @@ -1746,14 +1686,15 @@ struct btrfs_root *open_ctree(struct super_block *sb, BUG_ON(!chunk_root->node); read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, - (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), - BTRFS_UUID_SIZE); + (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), + BTRFS_UUID_SIZE); mutex_lock(&fs_info->chunk_mutex); ret = btrfs_read_chunk_tree(chunk_root); mutex_unlock(&fs_info->chunk_mutex); if (ret) { - printk("btrfs: failed to read chunk tree on %s\n", sb->s_id); + printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n", + sb->s_id); goto fail_chunk_root; } @@ -1812,7 +1753,8 @@ struct btrfs_root *open_ctree(struct super_block *sb, u64 bytenr = btrfs_super_log_root(disk_super); if (fs_devices->rw_devices == 0) { - printk("Btrfs log replay required on RO media\n"); + printk(KERN_WARNING "Btrfs log replay required " + "on RO media\n"); err = -EIO; goto fail_trans_kthread; } @@ -2097,7 +2039,8 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors) total_errors++; } if (total_errors > max_errors) { - printk("btrfs: %d errors while writing supers\n", total_errors); + printk(KERN_ERR "btrfs: %d errors while writing supers\n", + total_errors); BUG(); } @@ -2114,7 +2057,8 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors) total_errors++; } if (total_errors > max_errors) { - printk("btrfs: %d errors while writing supers\n", total_errors); + printk(KERN_ERR "btrfs: %d errors while writing supers\n", + total_errors); BUG(); } return 0; @@ -2137,16 +2081,11 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) down_write(&root->anon_super.s_umount); kill_anon_super(&root->anon_super); } -#if 0 - if (root->in_sysfs) - btrfs_sysfs_del_root(root); -#endif if (root->node) free_extent_buffer(root->node); if (root->commit_root) free_extent_buffer(root->commit_root); - if (root->name) - kfree(root->name); + kfree(root->name); kfree(root); return 0; } @@ -2157,7 +2096,7 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info) struct btrfs_root *gang[8]; int i; - while(1) { + while (1) { ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, (void **)gang, 0, ARRAY_SIZE(gang)); @@ -2228,18 +2167,17 @@ int close_ctree(struct btrfs_root *root) if (!(fs_info->sb->s_flags & MS_RDONLY)) { ret = btrfs_commit_super(root); - if (ret) { - printk("btrfs: commit super returns %d\n", ret); - } + if (ret) + printk(KERN_ERR "btrfs: commit super ret %d\n", ret); } if (fs_info->delalloc_bytes) { - printk("btrfs: at unmount delalloc count %Lu\n", + printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n", fs_info->delalloc_bytes); } if (fs_info->total_ref_cache_size) { - printk("btrfs: at umount reference cache size %Lu\n", - fs_info->total_ref_cache_size); + printk(KERN_INFO "btrfs: at umount reference cache size %llu\n", + (unsigned long long)fs_info->total_ref_cache_size); } if (fs_info->extent_root->node) @@ -2248,13 +2186,13 @@ int close_ctree(struct btrfs_root *root) if (fs_info->tree_root->node) free_extent_buffer(fs_info->tree_root->node); - if (root->fs_info->chunk_root->node); + if (root->fs_info->chunk_root->node) free_extent_buffer(root->fs_info->chunk_root->node); - if (root->fs_info->dev_root->node); + if (root->fs_info->dev_root->node) free_extent_buffer(root->fs_info->dev_root->node); - if (root->fs_info->csum_root->node); + if (root->fs_info->csum_root->node) free_extent_buffer(root->fs_info->csum_root->node); btrfs_free_block_groups(root->fs_info); @@ -2273,7 +2211,7 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->submit_workers); #if 0 - while(!list_empty(&fs_info->hashers)) { + while (!list_empty(&fs_info->hashers)) { struct btrfs_hasher *hasher; hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, hashers); @@ -2324,9 +2262,11 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf) WARN_ON(!btrfs_tree_locked(buf)); if (transid != root->fs_info->generation) { - printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", + printk(KERN_CRIT "btrfs transid mismatch buffer %llu, " + "found %llu running %llu\n", (unsigned long long)buf->start, - transid, root->fs_info->generation); + (unsigned long long)transid, + (unsigned long long)root->fs_info->generation); WARN_ON(1); } set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); @@ -2361,9 +2301,8 @@ int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; int ret; ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); - if (ret == 0) { + if (ret == 0) buf->flags |= EXTENT_UPTODATE; - } return ret; } diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c index 48b82cd7583..85315d2c90d 100644 --- a/fs/btrfs/export.c +++ b/fs/btrfs/export.c @@ -7,9 +7,11 @@ #include "export.h" #include "compat.h" -#define BTRFS_FID_SIZE_NON_CONNECTABLE (offsetof(struct btrfs_fid, parent_objectid)/4) -#define BTRFS_FID_SIZE_CONNECTABLE (offsetof(struct btrfs_fid, parent_root_objectid)/4) -#define BTRFS_FID_SIZE_CONNECTABLE_ROOT (sizeof(struct btrfs_fid)/4) +#define BTRFS_FID_SIZE_NON_CONNECTABLE (offsetof(struct btrfs_fid, \ + parent_objectid) / 4) +#define BTRFS_FID_SIZE_CONNECTABLE (offsetof(struct btrfs_fid, \ + parent_root_objectid) / 4) +#define BTRFS_FID_SIZE_CONNECTABLE_ROOT (sizeof(struct btrfs_fid) / 4) static int btrfs_encode_fh(struct dentry *dentry, u32 *fh, int *max_len, int connectable) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 780c1eeb829..ec43fa526d7 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -49,10 +49,10 @@ struct pending_extent_op { int del; }; -static int finish_current_insert(struct btrfs_trans_handle *trans, struct - btrfs_root *extent_root, int all); -static int del_pending_extents(struct btrfs_trans_handle *trans, struct - btrfs_root *extent_root, int all); +static int finish_current_insert(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, int all); +static int del_pending_extents(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, int all); static int pin_down_bytes(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 bytenr, u64 num_bytes, int is_data); @@ -247,7 +247,7 @@ static int cache_block_group(struct btrfs_root *root, if (ret < 0) goto err; - while(1) { + while (1) { leaf = path->nodes[0]; slot = path->slots[0]; if (slot >= btrfs_header_nritems(leaf)) { @@ -292,9 +292,8 @@ err: /* * return the block group that starts at or after bytenr */ -static struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct - btrfs_fs_info *info, - u64 bytenr) +static struct btrfs_block_group_cache * +btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr) { struct btrfs_block_group_cache *cache; @@ -306,9 +305,9 @@ static struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct /* * return the block group that contains teh given bytenr */ -struct btrfs_block_group_cache *btrfs_lookup_block_group(struct - btrfs_fs_info *info, - u64 bytenr) +struct btrfs_block_group_cache *btrfs_lookup_block_group( + struct btrfs_fs_info *info, + u64 bytenr) { struct btrfs_block_group_cache *cache; @@ -492,7 +491,7 @@ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len) * to the key objectid. */ -static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, +static noinline int lookup_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 bytenr, u64 parent, @@ -537,7 +536,7 @@ out: * updates all the backrefs that are pending on update_list for the * extent_root */ -static int noinline update_backrefs(struct btrfs_trans_handle *trans, +static noinline int update_backrefs(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct btrfs_path *path, struct list_head *update_list) @@ -573,9 +572,11 @@ loop: btrfs_ref_generation(leaf, ref) != op->orig_generation || (ref_objectid != op->level && ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) { - printk(KERN_ERR "couldn't find %Lu, parent %Lu, root %Lu, " - "owner %u\n", op->bytenr, op->orig_parent, - ref_root, op->level); + printk(KERN_ERR "btrfs couldn't find %llu, parent %llu, " + "root %llu, owner %u\n", + (unsigned long long)op->bytenr, + (unsigned long long)op->orig_parent, + (unsigned long long)ref_root, op->level); btrfs_print_leaf(extent_root, leaf); BUG(); } @@ -620,7 +621,7 @@ out: return 0; } -static int noinline insert_extents(struct btrfs_trans_handle *trans, +static noinline int insert_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct btrfs_path *path, struct list_head *insert_list, int nr) @@ -781,7 +782,7 @@ static int noinline insert_extents(struct btrfs_trans_handle *trans, return ret; } -static int noinline insert_extent_backref(struct btrfs_trans_handle *trans, +static noinline int insert_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, u64 bytenr, u64 parent, @@ -840,7 +841,7 @@ out: return ret; } -static int noinline remove_extent_backref(struct btrfs_trans_handle *trans, +static noinline int remove_extent_backref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path) { @@ -868,7 +869,7 @@ static int noinline remove_extent_backref(struct btrfs_trans_handle *trans, static void btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len) { -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,28) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 28) blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL); #else blkdev_issue_discard(bdev, start >> 9, len >> 9); @@ -908,7 +909,7 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, #endif } -static int noinline free_extents(struct btrfs_trans_handle *trans, +static noinline int free_extents(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct list_head *del_list) { @@ -937,10 +938,11 @@ search: extent_root->root_key.objectid, op->orig_generation, op->level, 1); if (ret) { - printk("Unable to find backref byte nr %Lu root %Lu gen %Lu " - "owner %u\n", op->bytenr, - extent_root->root_key.objectid, op->orig_generation, - op->level); + printk(KERN_ERR "btrfs unable to find backref byte nr %llu " + "root %llu gen %llu owner %u\n", + (unsigned long long)op->bytenr, + (unsigned long long)extent_root->root_key.objectid, + (unsigned long long)op->orig_generation, op->level); btrfs_print_leaf(extent_root, path->nodes[0]); WARN_ON(1); goto out; @@ -1282,7 +1284,9 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, btrfs_item_key_to_cpu(l, &key, path->slots[0]); if (key.objectid != bytenr) { btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]); - printk("wanted %Lu found %Lu\n", bytenr, key.objectid); + printk(KERN_ERR "btrfs wanted %llu found %llu\n", + (unsigned long long)bytenr, + (unsigned long long)key.objectid); BUG(); } BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY); @@ -1353,7 +1357,8 @@ int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans, goto out; if (ret != 0) { btrfs_print_leaf(root, path->nodes[0]); - printk("failed to find block number %Lu\n", bytenr); + printk(KERN_INFO "btrfs failed to find block number %llu\n", + (unsigned long long)bytenr); BUG(); } l = path->nodes[0]; @@ -1738,7 +1743,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, if (!path) return -ENOMEM; - while(1) { + while (1) { cache = NULL; spin_lock(&root->fs_info->block_group_cache_lock); for (n = rb_first(&root->fs_info->block_group_cache_tree); @@ -1921,10 +1926,8 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, spin_unlock(&space_info->lock); ret = btrfs_alloc_chunk(trans, extent_root, flags); - if (ret) { -printk("space info full %Lu\n", flags); + if (ret) space_info->full = 1; - } out: mutex_unlock(&extent_root->fs_info->chunk_mutex); return ret; @@ -1941,7 +1944,7 @@ static int update_block_group(struct btrfs_trans_handle *trans, u64 old_val; u64 byte_in_group; - while(total) { + while (total) { cache = btrfs_lookup_block_group(info, bytenr); if (!cache) return -1; @@ -2089,7 +2092,7 @@ int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) int ret; mutex_lock(&root->fs_info->pinned_mutex); - while(1) { + while (1) { ret = find_first_extent_bit(pinned_extents, last, &start, &end, EXTENT_DIRTY); if (ret) @@ -2110,7 +2113,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, int ret; mutex_lock(&root->fs_info->pinned_mutex); - while(1) { + while (1) { ret = find_first_extent_bit(unpin, 0, &start, &end, EXTENT_DIRTY); if (ret) @@ -2400,7 +2403,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, if (ret == 0) { struct btrfs_key found_key; extent_slot = path->slots[0]; - while(extent_slot > 0) { + while (extent_slot > 0) { extent_slot--; btrfs_item_key_to_cpu(path->nodes[0], &found_key, extent_slot); @@ -2422,8 +2425,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, &key, path, -1, 1); if (ret) { printk(KERN_ERR "umm, got %d back from search" - ", was looking for %Lu\n", ret, - bytenr); + ", was looking for %llu\n", ret, + (unsigned long long)bytenr); btrfs_print_leaf(extent_root, path->nodes[0]); } BUG_ON(ret); @@ -2432,9 +2435,12 @@ static int __free_extent(struct btrfs_trans_handle *trans, } else { btrfs_print_leaf(extent_root, path->nodes[0]); WARN_ON(1); - printk("Unable to find ref byte nr %Lu root %Lu " - "gen %Lu owner %Lu\n", bytenr, - root_objectid, ref_generation, owner_objectid); + printk(KERN_ERR "btrfs unable to find ref byte nr %llu " + "root %llu gen %llu owner %llu\n", + (unsigned long long)bytenr, + (unsigned long long)root_objectid, + (unsigned long long)ref_generation, + (unsigned long long)owner_objectid); } leaf = path->nodes[0]; @@ -2517,8 +2523,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, * find all the blocks marked as pending in the radix tree and remove * them from the extent map */ -static int del_pending_extents(struct btrfs_trans_handle *trans, struct - btrfs_root *extent_root, int all) +static int del_pending_extents(struct btrfs_trans_handle *trans, + struct btrfs_root *extent_root, int all) { int ret; int err = 0; @@ -2539,7 +2545,7 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct again: mutex_lock(&info->extent_ins_mutex); - while(1) { + while (1) { ret = find_first_extent_bit(pending_del, search, &start, &end, EXTENT_WRITEBACK); if (ret) { @@ -2753,7 +2759,7 @@ static u64 stripe_align(struct btrfs_root *root, u64 val) * ins->offset == number of blocks * Any available blocks before search_start are skipped. */ -static int noinline find_free_extent(struct btrfs_trans_handle *trans, +static noinline int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *orig_root, u64 num_bytes, u64 empty_size, u64 search_start, u64 search_end, @@ -2762,7 +2768,7 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans, int data) { int ret = 0; - struct btrfs_root * root = orig_root->fs_info->extent_root; + struct btrfs_root *root = orig_root->fs_info->extent_root; u64 total_needed = num_bytes; u64 *last_ptr = NULL; u64 last_wanted = 0; @@ -2995,8 +3001,10 @@ loop_check: *last_ptr = ins->objectid + ins->offset; ret = 0; } else if (!ret) { - printk(KERN_ERR "we were searching for %Lu bytes, num_bytes %Lu," - " loop %d, allowed_alloc %d\n", total_needed, num_bytes, + printk(KERN_ERR "btrfs searching for %llu bytes, " + "num_bytes %llu, loop %d, allowed_alloc %d\n", + (unsigned long long)total_needed, + (unsigned long long)num_bytes, loop, allowed_chunk_alloc); ret = -ENOSPC; } @@ -3012,19 +3020,22 @@ static void dump_space_info(struct btrfs_space_info *info, u64 bytes) struct btrfs_block_group_cache *cache; struct list_head *l; - printk(KERN_INFO "space_info has %Lu free, is %sfull\n", - info->total_bytes - info->bytes_used - info->bytes_pinned - - info->bytes_reserved, (info->full) ? "" : "not "); + printk(KERN_INFO "space_info has %llu free, is %sfull\n", + (unsigned long long)(info->total_bytes - info->bytes_used - + info->bytes_pinned - info->bytes_reserved), + (info->full) ? "" : "not "); down_read(&info->groups_sem); list_for_each(l, &info->block_groups) { cache = list_entry(l, struct btrfs_block_group_cache, list); spin_lock(&cache->lock); - printk(KERN_INFO "block group %Lu has %Lu bytes, %Lu used " - "%Lu pinned %Lu reserved\n", - cache->key.objectid, cache->key.offset, - btrfs_block_group_used(&cache->item), - cache->pinned, cache->reserved); + printk(KERN_INFO "block group %llu has %llu bytes, %llu used " + "%llu pinned %llu reserved\n", + (unsigned long long)cache->key.objectid, + (unsigned long long)cache->key.offset, + (unsigned long long)btrfs_block_group_used(&cache->item), + (unsigned long long)cache->pinned, + (unsigned long long)cache->reserved); btrfs_dump_free_space(cache, bytes); spin_unlock(&cache->lock); } @@ -3045,15 +3056,15 @@ static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans, if (data) { alloc_profile = info->avail_data_alloc_bits & - info->data_alloc_profile; + info->data_alloc_profile; data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; } else if (root == root->fs_info->chunk_root) { alloc_profile = info->avail_system_alloc_bits & - info->system_alloc_profile; + info->system_alloc_profile; data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; } else { alloc_profile = info->avail_metadata_alloc_bits & - info->metadata_alloc_profile; + info->metadata_alloc_profile; data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; } again: @@ -3092,8 +3103,9 @@ again: struct btrfs_space_info *sinfo; sinfo = __find_space_info(root->fs_info, data); - printk("allocation failed flags %Lu, wanted %Lu\n", - data, num_bytes); + printk(KERN_ERR "btrfs allocation failed flags %llu, " + "wanted %llu\n", (unsigned long long)data, + (unsigned long long)num_bytes); dump_space_info(sinfo, num_bytes); BUG(); } @@ -3108,7 +3120,8 @@ int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len) cache = btrfs_lookup_block_group(root->fs_info, start); if (!cache) { - printk(KERN_ERR "Unable to find block group for %Lu\n", start); + printk(KERN_ERR "Unable to find block group for %llu\n", + (unsigned long long)start); return -ENOSPC; } @@ -3235,10 +3248,12 @@ static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans, } update_block: - ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); + ret = update_block_group(trans, root, ins->objectid, + ins->offset, 1, 0); if (ret) { - printk("update block group failed for %Lu %Lu\n", - ins->objectid, ins->offset); + printk(KERN_ERR "btrfs update block group failed for %llu " + "%llu\n", (unsigned long long)ins->objectid, + (unsigned long long)ins->offset); BUG(); } out: @@ -3420,7 +3435,7 @@ int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans, return 0; } -static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, +static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_leaf_ref *ref) { @@ -3445,15 +3460,15 @@ static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans, return 0; } -static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len, - u32 *refs) +static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, + u64 len, u32 *refs) { int ret; ret = btrfs_lookup_extent_ref(NULL, root, start, len, refs); BUG_ON(ret); -#if 0 // some debugging code in case we see problems here +#if 0 /* some debugging code in case we see problems here */ /* if the refs count is one, it won't get increased again. But * if the ref count is > 1, someone may be decreasing it at * the same time we are. @@ -3474,8 +3489,8 @@ static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len free_extent_buffer(eb); } if (*refs == 1) { - printk("block %llu went down to one during drop_snap\n", - (unsigned long long)start); + printk(KERN_ERR "btrfs block %llu went down to one " + "during drop_snap\n", (unsigned long long)start); } } @@ -3489,7 +3504,7 @@ static int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len * helper function for drop_snapshot, this walks down the tree dropping ref * counts as it goes. */ -static int noinline walk_down_tree(struct btrfs_trans_handle *trans, +static noinline int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level) { @@ -3516,7 +3531,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, /* * walk down to the last node level and free all the leaves */ - while(*level >= 0) { + while (*level >= 0) { WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); cur = path->nodes[*level]; @@ -3576,10 +3591,6 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans, *level = 0; break; } - if (printk_ratelimit()) { - printk("leaf ref miss for bytenr %llu\n", - (unsigned long long)bytenr); - } } next = btrfs_find_tree_block(root, bytenr, blocksize); if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) { @@ -3641,7 +3652,7 @@ out: * walk_down_tree. The main difference is that it checks reference * counts while tree blocks are locked. */ -static int noinline walk_down_subtree(struct btrfs_trans_handle *trans, +static noinline int walk_down_subtree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level) { @@ -3730,7 +3741,7 @@ out: * to find the first node higher up where we haven't yet gone through * all the slots */ -static int noinline walk_up_tree(struct btrfs_trans_handle *trans, +static noinline int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level, int max_level) @@ -3839,7 +3850,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root } } } - while(1) { + while (1) { wret = walk_down_tree(trans, root, path, &level); if (wret > 0) break; @@ -3920,7 +3931,7 @@ static unsigned long calc_ra(unsigned long start, unsigned long last, return min(last, start + nr - 1); } -static int noinline relocate_inode_pages(struct inode *inode, u64 start, +static noinline int relocate_inode_pages(struct inode *inode, u64 start, u64 len) { u64 page_start; @@ -4011,7 +4022,7 @@ out_unlock: return ret; } -static int noinline relocate_data_extent(struct inode *reloc_inode, +static noinline int relocate_data_extent(struct inode *reloc_inode, struct btrfs_key *extent_key, u64 offset) { @@ -4087,7 +4098,7 @@ static int is_cowonly_root(u64 root_objectid) return 0; } -static int noinline __next_ref_path(struct btrfs_trans_handle *trans, +static noinline int __next_ref_path(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct btrfs_ref_path *ref_path, int first_time) @@ -4119,11 +4130,10 @@ walk_down: if (level < ref_path->lowest_level) break; - if (level >= 0) { + if (level >= 0) bytenr = ref_path->nodes[level]; - } else { + else bytenr = ref_path->extent_start; - } BUG_ON(bytenr == 0); parent = ref_path->nodes[level + 1]; @@ -4170,11 +4180,12 @@ walk_up: level = ref_path->current_level; while (level < BTRFS_MAX_LEVEL - 1) { u64 ref_objectid; - if (level >= 0) { + + if (level >= 0) bytenr = ref_path->nodes[level]; - } else { + else bytenr = ref_path->extent_start; - } + BUG_ON(bytenr == 0); key.objectid = bytenr; @@ -4299,7 +4310,7 @@ static int btrfs_next_ref_path(struct btrfs_trans_handle *trans, return __next_ref_path(trans, extent_root, ref_path, 0); } -static int noinline get_new_locations(struct inode *reloc_inode, +static noinline int get_new_locations(struct inode *reloc_inode, struct btrfs_key *extent_key, u64 offset, int no_fragment, struct disk_extent **extents, @@ -4420,7 +4431,7 @@ out: return ret; } -static int noinline replace_one_extent(struct btrfs_trans_handle *trans, +static noinline int replace_one_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *extent_key, @@ -4778,7 +4789,7 @@ int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans, return 0; } -static int noinline invalidate_extent_cache(struct btrfs_root *root, +static noinline int invalidate_extent_cache(struct btrfs_root *root, struct extent_buffer *leaf, struct btrfs_block_group_cache *group, struct btrfs_root *target_root) @@ -4826,7 +4837,7 @@ static int noinline invalidate_extent_cache(struct btrfs_root *root, return 0; } -static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans, +static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *leaf, struct btrfs_block_group_cache *group, @@ -5035,7 +5046,7 @@ int btrfs_cleanup_reloc_trees(struct btrfs_root *root) return 0; } -static int noinline init_reloc_tree(struct btrfs_trans_handle *trans, +static noinline int init_reloc_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root) { struct btrfs_root *reloc_root; @@ -5102,7 +5113,7 @@ static int noinline init_reloc_tree(struct btrfs_trans_handle *trans, * tree blocks are shared between reloc trees, so they are also shared * between subvols. */ -static int noinline relocate_one_path(struct btrfs_trans_handle *trans, +static noinline int relocate_one_path(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *first_key, @@ -5199,7 +5210,7 @@ static int noinline relocate_one_path(struct btrfs_trans_handle *trans, return 0; } -static int noinline relocate_tree_block(struct btrfs_trans_handle *trans, +static noinline int relocate_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, struct btrfs_key *first_key, @@ -5217,7 +5228,7 @@ static int noinline relocate_tree_block(struct btrfs_trans_handle *trans, return 0; } -static int noinline del_extent_zero(struct btrfs_trans_handle *trans, +static noinline int del_extent_zero(struct btrfs_trans_handle *trans, struct btrfs_root *extent_root, struct btrfs_path *path, struct btrfs_key *extent_key) @@ -5233,7 +5244,7 @@ out: return ret; } -static struct btrfs_root noinline *read_ref_root(struct btrfs_fs_info *fs_info, +static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info, struct btrfs_ref_path *ref_path) { struct btrfs_key root_key; @@ -5248,7 +5259,7 @@ static struct btrfs_root noinline *read_ref_root(struct btrfs_fs_info *fs_info, return btrfs_read_fs_root_no_name(fs_info, &root_key); } -static int noinline relocate_one_extent(struct btrfs_root *extent_root, +static noinline int relocate_one_extent(struct btrfs_root *extent_root, struct btrfs_path *path, struct btrfs_key *extent_key, struct btrfs_block_group_cache *group, @@ -5276,8 +5287,8 @@ static int noinline relocate_one_extent(struct btrfs_root *extent_root, ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS); if (!ref_path) { - ret = -ENOMEM; - goto out; + ret = -ENOMEM; + goto out; } for (loops = 0; ; loops++) { @@ -5497,7 +5508,7 @@ out: return ret; } -static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info, +static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info, struct btrfs_block_group_cache *group) { struct inode *inode = NULL; @@ -5617,7 +5628,7 @@ int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start) block_group = btrfs_lookup_block_group(info, group_start); BUG_ON(!block_group); - printk("btrfs relocating block group %llu flags %llu\n", + printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n", (unsigned long long)block_group->key.objectid, (unsigned long long)block_group->flags); @@ -5649,7 +5660,7 @@ again: btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1); mutex_unlock(&root->fs_info->cleaner_mutex); - while(1) { + while (1) { ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto out; @@ -5712,7 +5723,7 @@ next: } if (total_found > 0) { - printk("btrfs found %llu extents in pass %d\n", + printk(KERN_INFO "btrfs found %llu extents in pass %d\n", (unsigned long long)total_found, pass); pass++; if (total_found == skipped && pass > 2) { @@ -5754,7 +5765,7 @@ static int find_first_block_group(struct btrfs_root *root, if (ret < 0) goto out; - while(1) { + while (1) { slot = path->slots[0]; leaf = path->nodes[0]; if (slot >= btrfs_header_nritems(leaf)) { @@ -5825,7 +5836,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) if (!path) return -ENOMEM; - while(1) { + while (1) { ret = find_first_block_group(root, path, &key); if (ret > 0) { ret = 0; diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 0bf7684207a..39edb551dca 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -32,7 +32,7 @@ static LIST_HEAD(states); #define LEAK_DEBUG 0 #ifdef LEAK_DEBUG -static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED; +static DEFINE_SPINLOCK(leak_lock); #endif #define BUFFER_LRU_MAX 64 @@ -81,7 +81,11 @@ void extent_io_exit(void) while (!list_empty(&states)) { state = list_entry(states.next, struct extent_state, leak_list); - printk("state leak: start %Lu end %Lu state %lu in tree %p refs %d\n", state->start, state->end, state->state, state->tree, atomic_read(&state->refs)); + printk(KERN_ERR "btrfs state leak: start %llu end %llu " + "state %lu in tree %p refs %d\n", + (unsigned long long)state->start, + (unsigned long long)state->end, + state->state, state->tree, atomic_read(&state->refs)); list_del(&state->leak_list); kmem_cache_free(extent_state_cache, state); @@ -89,7 +93,9 @@ void extent_io_exit(void) while (!list_empty(&buffers)) { eb = list_entry(buffers.next, struct extent_buffer, leak_list); - printk("buffer leak start %Lu len %lu refs %d\n", eb->start, eb->len, atomic_read(&eb->refs)); + printk(KERN_ERR "btrfs buffer leak start %llu len %lu " + "refs %d\n", (unsigned long long)eb->start, + eb->len, atomic_read(&eb->refs)); list_del(&eb->leak_list); kmem_cache_free(extent_buffer_cache, eb); } @@ -158,11 +164,11 @@ EXPORT_SYMBOL(free_extent_state); static struct rb_node *tree_insert(struct rb_root *root, u64 offset, struct rb_node *node) { - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct tree_entry *entry; - while(*p) { + while (*p) { parent = *p; entry = rb_entry(parent, struct tree_entry, rb_node); @@ -185,13 +191,13 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, struct rb_node **next_ret) { struct rb_root *root = &tree->state; - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct rb_node *prev = NULL; struct rb_node *orig_prev = NULL; struct tree_entry *entry; struct tree_entry *prev_entry = NULL; - while(n) { + while (n) { entry = rb_entry(n, struct tree_entry, rb_node); prev = n; prev_entry = entry; @@ -200,14 +206,13 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, n = n->rb_left; else if (offset > entry->end) n = n->rb_right; - else { + else return n; - } } if (prev_ret) { orig_prev = prev; - while(prev && offset > prev_entry->end) { + while (prev && offset > prev_entry->end) { prev = rb_next(prev); prev_entry = rb_entry(prev, struct tree_entry, rb_node); } @@ -217,7 +222,7 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, if (next_ret) { prev_entry = rb_entry(prev, struct tree_entry, rb_node); - while(prev && offset < prev_entry->start) { + while (prev && offset < prev_entry->start) { prev = rb_prev(prev); prev_entry = rb_entry(prev, struct tree_entry, rb_node); } @@ -233,9 +238,8 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree, struct rb_node *ret; ret = __etree_search(tree, offset, &prev, NULL); - if (!ret) { + if (!ret) return prev; - } return ret; } @@ -243,11 +247,11 @@ static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree, u64 offset, struct rb_node *node) { struct rb_root *root = &tree->buffer; - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct extent_buffer *eb; - while(*p) { + while (*p) { parent = *p; eb = rb_entry(parent, struct extent_buffer, rb_node); @@ -268,10 +272,10 @@ static struct extent_buffer *buffer_search(struct extent_io_tree *tree, u64 offset) { struct rb_root *root = &tree->buffer; - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct extent_buffer *eb; - while(n) { + while (n) { eb = rb_entry(n, struct extent_buffer, rb_node); if (offset < eb->start) n = n->rb_left; @@ -363,7 +367,9 @@ static int insert_state(struct extent_io_tree *tree, struct rb_node *node; if (end < start) { - printk("end < start %Lu %Lu\n", end, start); + printk(KERN_ERR "btrfs end < start %llu %llu\n", + (unsigned long long)end, + (unsigned long long)start); WARN_ON(1); } if (bits & EXTENT_DIRTY) @@ -376,7 +382,10 @@ static int insert_state(struct extent_io_tree *tree, if (node) { struct extent_state *found; found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end); + printk(KERN_ERR "btrfs found node %llu %llu on insert of " + "%llu %llu\n", (unsigned long long)found->start, + (unsigned long long)found->end, + (unsigned long long)start, (unsigned long long)end); free_extent_state(state); return -EEXIST; } @@ -412,7 +421,6 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, if (node) { struct extent_state *found; found = rb_entry(node, struct extent_state, rb_node); - printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end); free_extent_state(prealloc); return -EEXIST; } @@ -661,8 +669,9 @@ static void set_state_bits(struct extent_io_tree *tree, * [start, end] is inclusive * This takes the tree lock. */ -static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits, - int exclusive, u64 *failed_start, gfp_t mask) +static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, + int bits, int exclusive, u64 *failed_start, + gfp_t mask) { struct extent_state *state; struct extent_state *prealloc = NULL; @@ -763,7 +772,7 @@ again: if (end < last_start) this_end = end; else - this_end = last_start -1; + this_end = last_start - 1; err = insert_state(tree, prealloc, start, this_end, bits); prealloc = NULL; @@ -891,8 +900,8 @@ int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, } EXPORT_SYMBOL(set_extent_uptodate); -static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) +static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, + u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask); } @@ -904,8 +913,8 @@ static int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, 0, NULL, mask); } -static int clear_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end, - gfp_t mask) +static int clear_extent_writeback(struct extent_io_tree *tree, u64 start, + u64 end, gfp_t mask) { return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask); } @@ -1025,11 +1034,10 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start, * our range starts. */ node = tree_search(tree, start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (state->end >= start && (state->state & bits)) { *start_ret = state->start; @@ -1062,15 +1070,14 @@ struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); - if (state->end >= start && (state->state & bits)) { + if (state->end >= start && (state->state & bits)) return state; - } + node = rb_next(node); if (!node) break; @@ -1108,7 +1115,7 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree, goto out; } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (found && (state->start != cur_start || (state->state & EXTENT_BOUNDARY))) { @@ -1150,7 +1157,7 @@ static noinline int __unlock_for_delalloc(struct inode *inode, if (index == locked_page->index && end_index == index) return 0; - while(nr_pages > 0) { + while (nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nr_pages, ARRAY_SIZE(pages)), pages); @@ -1186,7 +1193,7 @@ static noinline int lock_delalloc_pages(struct inode *inode, /* skip the page at the start index */ nrpages = end_index - index + 1; - while(nrpages > 0) { + while (nrpages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nrpages, ARRAY_SIZE(pages)), pages); @@ -1263,17 +1270,16 @@ again: * pages in order, so we can't process delalloc bytes before * locked_page */ - if (delalloc_start < *start) { + if (delalloc_start < *start) delalloc_start = *start; - } /* * make sure to limit the number of pages we try to lock down * if we're looping. */ - if (delalloc_end + 1 - delalloc_start > max_bytes && loops) { + if (delalloc_end + 1 - delalloc_start > max_bytes && loops) delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1; - } + /* step two, lock all the pages after the page that has start */ ret = lock_delalloc_pages(inode, locked_page, delalloc_start, delalloc_end); @@ -1341,7 +1347,7 @@ int extent_clear_unlock_delalloc(struct inode *inode, if (!(unlock_pages || clear_dirty || set_writeback || end_writeback)) return 0; - while(nr_pages > 0) { + while (nr_pages > 0) { ret = find_get_pages_contig(inode->i_mapping, index, min_t(unsigned long, nr_pages, ARRAY_SIZE(pages)), pages); @@ -1384,7 +1390,6 @@ u64 count_range_bits(struct extent_io_tree *tree, int found = 0; if (search_end <= cur_start) { - printk("search_end %Lu start %Lu\n", search_end, cur_start); WARN_ON(1); return 0; } @@ -1399,11 +1404,10 @@ u64 count_range_bits(struct extent_io_tree *tree, * our range starts. */ node = tree_search(tree, cur_start); - if (!node) { + if (!node) goto out; - } - while(1) { + while (1) { state = rb_entry(node, struct extent_state, rb_node); if (state->start > search_end) break; @@ -1927,19 +1931,15 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree, nr = bio_get_nr_vecs(bdev); bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); - if (!bio) { - printk("failed to allocate bio nr %d\n", nr); - } bio_add_page(bio, page, page_size, offset); bio->bi_end_io = end_io_func; bio->bi_private = tree; - if (bio_ret) { + if (bio_ret) *bio_ret = bio; - } else { + else ret = submit_one_bio(rw, bio, mirror_num, bio_flags); - } return ret; } @@ -2028,13 +2028,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree, break; } extent_offset = cur - em->start; - if (extent_map_end(em) <= cur) { -printk("bad mapping em [%Lu %Lu] cur %Lu\n", em->start, extent_map_end(em), cur); - } BUG_ON(extent_map_end(em) <= cur); - if (end < cur) { -printk("2bad mapping end %Lu cur %Lu\n", end, cur); - } BUG_ON(end < cur); if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) @@ -2199,7 +2193,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, delalloc_end = 0; page_started = 0; if (!epd->extent_locked) { - while(delalloc_end < page_end) { + while (delalloc_end < page_end) { nr_delalloc = find_lock_delalloc_range(inode, tree, page, &delalloc_start, @@ -2242,9 +2236,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, nr_written++; end = page_end; - if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) { - printk("found delalloc bits after lock_extent\n"); - } + if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) + printk(KERN_ERR "btrfs delalloc bits after lock_extent\n"); if (last_byte <= start) { clear_extent_dirty(tree, start, page_end, GFP_NOFS); @@ -2297,7 +2290,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS); - unlock_extent(tree, unlock_start, cur + iosize -1, + unlock_extent(tree, unlock_start, cur + iosize - 1, GFP_NOFS); /* @@ -2344,9 +2337,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc, set_range_writeback(tree, cur, cur + iosize - 1); if (!PageWriteback(page)) { - printk("warning page %lu not writeback, " - "cur %llu end %llu\n", page->index, - (unsigned long long)cur, + printk(KERN_ERR "btrfs warning page %lu not " + "writeback, cur %llu end %llu\n", + page->index, (unsigned long long)cur, (unsigned long long)end); } @@ -2430,8 +2423,8 @@ static int extent_write_cache_pages(struct extent_io_tree *tree, retry: while (!done && (index <= end) && (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, - PAGECACHE_TAG_DIRTY, - min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) { + PAGECACHE_TAG_DIRTY, min(end - index, + (pgoff_t)PAGEVEC_SIZE-1) + 1))) { unsigned i; scanned = 1; @@ -2536,9 +2529,8 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page, extent_write_cache_pages(tree, mapping, &wbc_writepages, __extent_writepage, &epd, flush_write_bio); - if (epd.bio) { + if (epd.bio) submit_one_bio(WRITE, epd.bio, 0, 0); - } return ret; } EXPORT_SYMBOL(extent_write_full_page); @@ -2568,7 +2560,7 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, .range_end = end + 1, }; - while(start <= end) { + while (start <= end) { page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT); if (clear_page_dirty_for_io(page)) ret = __extent_writepage(page, &wbc_writepages, &epd); @@ -2606,9 +2598,8 @@ int extent_writepages(struct extent_io_tree *tree, ret = extent_write_cache_pages(tree, mapping, wbc, __extent_writepage, &epd, flush_write_bio); - if (epd.bio) { + if (epd.bio) submit_one_bio(WRITE, epd.bio, 0, 0); - } return ret; } EXPORT_SYMBOL(extent_writepages); @@ -2666,7 +2657,7 @@ int extent_invalidatepage(struct extent_io_tree *tree, u64 end = start + PAGE_CACHE_SIZE - 1; size_t blocksize = page->mapping->host->i_sb->s_blocksize; - start += (offset + blocksize -1) & ~(blocksize - 1); + start += (offset + blocksize - 1) & ~(blocksize - 1); if (start > end) return 0; @@ -2727,12 +2718,12 @@ int extent_prepare_write(struct extent_io_tree *tree, orig_block_start = block_start; lock_extent(tree, page_start, page_end, GFP_NOFS); - while(block_start <= block_end) { + while (block_start <= block_end) { em = get_extent(inode, page, page_offset, block_start, block_end - block_start + 1, 1); - if (IS_ERR(em) || !em) { + if (IS_ERR(em) || !em) goto err; - } + cur_end = min(block_end, extent_map_end(em) - 1); block_off_start = block_start & (PAGE_CACHE_SIZE - 1); block_off_end = block_off_start + blocksize; @@ -3170,7 +3161,7 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree, } __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); set_extent_dirty(tree, page_offset(page), - page_offset(page) + PAGE_CACHE_SIZE -1, + page_offset(page) + PAGE_CACHE_SIZE - 1, GFP_NOFS); unlock_page(page); } @@ -3235,7 +3226,7 @@ int extent_range_uptodate(struct extent_io_tree *tree, ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1); if (ret) return 1; - while(start <= end) { + while (start <= end) { index = start >> PAGE_CACHE_SHIFT; page = find_get_page(tree->mapping, index); uptodate = PageUptodate(page); @@ -3321,16 +3312,12 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, lock_page(page); } locked_pages++; - if (!PageUptodate(page)) { + if (!PageUptodate(page)) all_uptodate = 0; - } } if (all_uptodate) { if (start_i == 0) eb->flags |= EXTENT_UPTODATE; - if (ret) { - printk("all up to date but ret is %d\n", ret); - } goto unlock_exit; } @@ -3345,10 +3332,8 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, err = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num, &bio_flags); - if (err) { + if (err) ret = err; - printk("err %d from __extent_read_full_page\n", ret); - } } else { unlock_page(page); } @@ -3357,26 +3342,23 @@ int read_extent_buffer_pages(struct extent_io_tree *tree, if (bio) submit_one_bio(READ, bio, mirror_num, bio_flags); - if (ret || !wait) { - if (ret) - printk("ret %d wait %d returning\n", ret, wait); + if (ret || !wait) return ret; - } + for (i = start_i; i < num_pages; i++) { page = extent_buffer_page(eb, i); wait_on_page_locked(page); - if (!PageUptodate(page)) { - printk("page not uptodate after wait_on_page_locked\n"); + if (!PageUptodate(page)) ret = -EIO; - } } + if (!ret) eb->flags |= EXTENT_UPTODATE; return ret; unlock_exit: i = start_i; - while(locked_pages > 0) { + while (locked_pages > 0) { page = extent_buffer_page(eb, i); i++; unlock_page(page); @@ -3403,7 +3385,7 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); cur = min(len, (PAGE_CACHE_SIZE - offset)); @@ -3442,8 +3424,11 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, offset = 0; *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; } + if (start + min_len > eb->len) { -printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len); + printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, " + "wanted %lu %lu\n", (unsigned long long)eb->start, + eb->len, start, min_len); WARN_ON(1); } @@ -3506,7 +3491,7 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); cur = min(len, (PAGE_CACHE_SIZE - offset)); @@ -3542,7 +3527,7 @@ void write_extent_buffer(struct extent_buffer *eb, const void *srcv, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); WARN_ON(!PageUptodate(page)); @@ -3574,7 +3559,7 @@ void memset_extent_buffer(struct extent_buffer *eb, char c, offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(eb, i); WARN_ON(!PageUptodate(page)); @@ -3607,7 +3592,7 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, offset = (start_offset + dst_offset) & ((unsigned long)PAGE_CACHE_SIZE - 1); - while(len > 0) { + while (len > 0) { page = extent_buffer_page(dst, i); WARN_ON(!PageUptodate(page)); @@ -3674,17 +3659,17 @@ void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, unsigned long src_i; if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " + "len %lu dst len %lu\n", src_offset, len, dst->len); BUG_ON(1); } if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " + "len %lu dst len %lu\n", dst_offset, len, dst->len); BUG_ON(1); } - while(len > 0) { + while (len > 0) { dst_off_in_page = (start_offset + dst_offset) & ((unsigned long)PAGE_CACHE_SIZE - 1); src_off_in_page = (start_offset + src_offset) & @@ -3722,20 +3707,20 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, unsigned long src_i; if (src_offset + len > dst->len) { - printk("memmove bogus src_offset %lu move len %lu len %lu\n", - src_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " + "len %lu len %lu\n", src_offset, len, dst->len); BUG_ON(1); } if (dst_offset + len > dst->len) { - printk("memmove bogus dst_offset %lu move len %lu len %lu\n", - dst_offset, len, dst->len); + printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " + "len %lu len %lu\n", dst_offset, len, dst->len); BUG_ON(1); } if (dst_offset < src_offset) { memcpy_extent_buffer(dst, dst_offset, src_offset, len); return; } - while(len > 0) { + while (len > 0) { dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c index fd3ebfb8c3c..4a83e33ada3 100644 --- a/fs/btrfs/extent_map.c +++ b/fs/btrfs/extent_map.c @@ -89,11 +89,11 @@ EXPORT_SYMBOL(free_extent_map); static struct rb_node *tree_insert(struct rb_root *root, u64 offset, struct rb_node *node) { - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct extent_map *entry; - while(*p) { + while (*p) { parent = *p; entry = rb_entry(parent, struct extent_map, rb_node); @@ -122,13 +122,13 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, struct rb_node **prev_ret, struct rb_node **next_ret) { - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct rb_node *prev = NULL; struct rb_node *orig_prev = NULL; struct extent_map *entry; struct extent_map *prev_entry = NULL; - while(n) { + while (n) { entry = rb_entry(n, struct extent_map, rb_node); prev = n; prev_entry = entry; @@ -145,7 +145,7 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, if (prev_ret) { orig_prev = prev; - while(prev && offset >= extent_map_end(prev_entry)) { + while (prev && offset >= extent_map_end(prev_entry)) { prev = rb_next(prev); prev_entry = rb_entry(prev, struct extent_map, rb_node); } @@ -155,7 +155,7 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset, if (next_ret) { prev_entry = rb_entry(prev, struct extent_map, rb_node); - while(prev && offset < prev_entry->start) { + while (prev && offset < prev_entry->start) { prev = rb_prev(prev); prev_entry = rb_entry(prev, struct extent_map, rb_node); } diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index cc6e0b6de94..b11abfad81a 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -24,7 +24,7 @@ #include "transaction.h" #include "print-tree.h" -#define MAX_CSUM_ITEMS(r,size) ((((BTRFS_LEAF_DATA_SIZE(r) - \ +#define MAX_CSUM_ITEMS(r, size) ((((BTRFS_LEAF_DATA_SIZE(r) - \ sizeof(struct btrfs_item) * 2) / \ size) - 1)) int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, @@ -166,7 +166,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, WARN_ON(bio->bi_vcnt <= 0); disk_bytenr = (u64)bio->bi_sector << 9; - while(bio_index < bio->bi_vcnt) { + while (bio_index < bio->bi_vcnt) { offset = page_offset(bvec->bv_page) + bvec->bv_offset; ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum); if (ret == 0) @@ -192,8 +192,9 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, offset + bvec->bv_len - 1, EXTENT_NODATASUM, GFP_NOFS); } else { - printk("no csum found for inode %lu " - "start %llu\n", inode->i_ino, + printk(KERN_INFO "btrfs no csum found " + "for inode %lu start %llu\n", + inode->i_ino, (unsigned long long)offset); } item = NULL; @@ -373,7 +374,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, BUG_ON(!ordered); sums->bytenr = ordered->start; - while(bio_index < bio->bi_vcnt) { + while (bio_index < bio->bi_vcnt) { if (!contig) offset = page_offset(bvec->bv_page) + bvec->bv_offset; @@ -507,7 +508,7 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); - while(1) { + while (1) { key.objectid = BTRFS_EXTENT_CSUM_OBJECTID; key.offset = end_byte - 1; key.type = BTRFS_EXTENT_CSUM_KEY; @@ -715,9 +716,8 @@ again: goto csum; diff = diff - btrfs_item_size_nr(leaf, path->slots[0]); - if (diff != csum_size) { + if (diff != csum_size) goto insert; - } ret = btrfs_extend_item(trans, root, path, diff); BUG_ON(ret); @@ -732,7 +732,7 @@ insert: u64 next_sector = sector_sum->bytenr; struct btrfs_sector_sum *next = sector_sum + 1; - while(tmp < sums->len) { + while (tmp < sums->len) { if (next_sector + root->sectorsize != next->bytenr) break; tmp += root->sectorsize; diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 5908521922f..0e3a13a4565 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -44,10 +44,10 @@ /* simple helper to fault in pages and copy. This should go away * and be replaced with calls into generic code. */ -static int noinline btrfs_copy_from_user(loff_t pos, int num_pages, +static noinline int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes, struct page **prepared_pages, - const char __user * buf) + const char __user *buf) { long page_fault = 0; int i; @@ -78,7 +78,7 @@ static int noinline btrfs_copy_from_user(loff_t pos, int num_pages, /* * unlocks pages after btrfs_file_write is done with them */ -static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) +static noinline void btrfs_drop_pages(struct page **pages, size_t num_pages) { size_t i; for (i = 0; i < num_pages; i++) { @@ -103,7 +103,7 @@ static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages) * this also makes the decision about creating an inline extent vs * doing real data extents, marking pages dirty and delalloc as required. */ -static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, +static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct file *file, struct page **pages, @@ -137,9 +137,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans, btrfs_set_trans_block_group(trans, inode); hint_byte = 0; - if ((end_of_last_block & 4095) == 0) { - printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block); - } set_extent_uptodate(io_tree, start_pos, end_of_last_block, GFP_NOFS); /* check for reserved extents on each page, we don't want @@ -185,7 +182,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end, len = (u64)-1; testend = 0; } - while(1) { + while (1) { if (!split) split = alloc_extent_map(GFP_NOFS); if (!split2) @@ -295,7 +292,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) path = btrfs_alloc_path(); ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino, last_offset, 0); - while(1) { + while (1) { nritems = btrfs_header_nritems(path->nodes[0]); if (path->slots[0] >= nritems) { ret = btrfs_next_leaf(root, path); @@ -314,8 +311,10 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) if (found_key.offset < last_offset) { WARN_ON(1); btrfs_print_leaf(root, leaf); - printk("inode %lu found offset %Lu expected %Lu\n", - inode->i_ino, found_key.offset, last_offset); + printk(KERN_ERR "inode %lu found offset %llu " + "expected %llu\n", inode->i_ino, + (unsigned long long)found_key.offset, + (unsigned long long)last_offset); err = 1; goto out; } @@ -331,7 +330,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) extent_end = found_key.offset + btrfs_file_extent_inline_len(leaf, extent); extent_end = (extent_end + root->sectorsize - 1) & - ~((u64)root->sectorsize -1 ); + ~((u64)root->sectorsize - 1); } last_offset = extent_end; path->slots[0]++; @@ -339,8 +338,9 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode) if (0 && last_offset < inode->i_size) { WARN_ON(1); btrfs_print_leaf(root, leaf); - printk("inode %lu found offset %Lu size %Lu\n", inode->i_ino, - last_offset, inode->i_size); + printk(KERN_ERR "inode %lu found offset %llu size %llu\n", + inode->i_ino, (unsigned long long)last_offset, + (unsigned long long)inode->i_size); err = 1; } @@ -362,7 +362,7 @@ out: * inline_limit is used to tell this code which offsets in the file to keep * if they contain inline extents. */ -int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, +noinline int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 start, u64 end, u64 inline_limit, u64 *hint_byte) { @@ -398,7 +398,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); if (!path) return -ENOMEM; - while(1) { + while (1) { recow = 0; btrfs_release_path(root, path); ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino, @@ -649,16 +649,15 @@ next_slot: if (disk_bytenr != 0) { ret = btrfs_update_extent_ref(trans, root, disk_bytenr, orig_parent, - leaf->start, + leaf->start, root->root_key.objectid, trans->transid, ins.objectid); BUG_ON(ret); } btrfs_release_path(root, path); - if (disk_bytenr != 0) { + if (disk_bytenr != 0) inode_add_bytes(inode, extent_end - end); - } } if (found_extent && !keep) { @@ -944,7 +943,7 @@ done: * waits for data=ordered extents to finish before allowing the pages to be * modified. */ -static int noinline prepare_pages(struct btrfs_root *root, struct file *file, +static noinline int prepare_pages(struct btrfs_root *root, struct file *file, struct page **pages, size_t num_pages, loff_t pos, unsigned long first_index, unsigned long last_index, size_t write_bytes) @@ -979,7 +978,8 @@ again: struct btrfs_ordered_extent *ordered; lock_extent(&BTRFS_I(inode)->io_tree, start_pos, last_pos - 1, GFP_NOFS); - ordered = btrfs_lookup_first_ordered_extent(inode, last_pos -1); + ordered = btrfs_lookup_first_ordered_extent(inode, + last_pos - 1); if (ordered && ordered->file_offset + ordered->len > start_pos && ordered->file_offset < last_pos) { @@ -1085,7 +1085,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, } } - while(count > 0) { + while (count > 0) { size_t offset = pos & (PAGE_CACHE_SIZE - 1); size_t write_bytes = min(count, nrptrs * (size_t)PAGE_CACHE_SIZE - @@ -1178,7 +1178,7 @@ out_nolock: return num_written ? num_written : err; } -int btrfs_release_file(struct inode * inode, struct file * filp) +int btrfs_release_file(struct inode *inode, struct file *filp) { if (filp->private_data) btrfs_ioctl_trans_end(filp); @@ -1237,9 +1237,8 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) } ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); - if (ret < 0) { + if (ret < 0) goto out; - } /* we've logged all the items and now have a consistent * version of the file in the log. It is possible that diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 2e69b9c3043..d1e5f0e84c5 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -213,10 +213,13 @@ static int __btrfs_add_free_space(struct btrfs_block_group_cache *block_group, info->offset = offset; info->bytes += bytes; } else if (right_info && right_info->offset != offset+bytes) { - printk(KERN_ERR "adding space in the middle of an existing " - "free space area. existing: offset=%Lu, bytes=%Lu. " - "new: offset=%Lu, bytes=%Lu\n", right_info->offset, - right_info->bytes, offset, bytes); + printk(KERN_ERR "btrfs adding space in the middle of an " + "existing free space area. existing: " + "offset=%llu, bytes=%llu. new: offset=%llu, " + "bytes=%llu\n", (unsigned long long)right_info->offset, + (unsigned long long)right_info->bytes, + (unsigned long long)offset, + (unsigned long long)bytes); BUG(); } @@ -225,11 +228,14 @@ static int __btrfs_add_free_space(struct btrfs_block_group_cache *block_group, if (unlikely((left_info->offset + left_info->bytes) != offset)) { - printk(KERN_ERR "free space to the left of new free " - "space isn't quite right. existing: offset=%Lu," - " bytes=%Lu. new: offset=%Lu, bytes=%Lu\n", - left_info->offset, left_info->bytes, offset, - bytes); + printk(KERN_ERR "btrfs free space to the left " + "of new free space isn't " + "quite right. existing: offset=%llu, " + "bytes=%llu. new: offset=%llu, bytes=%llu\n", + (unsigned long long)left_info->offset, + (unsigned long long)left_info->bytes, + (unsigned long long)offset, + (unsigned long long)bytes); BUG(); } @@ -265,8 +271,7 @@ out: BUG(); } - if (alloc_info) - kfree(alloc_info); + kfree(alloc_info); return ret; } @@ -283,9 +288,11 @@ __btrfs_remove_free_space(struct btrfs_block_group_cache *block_group, if (info && info->offset == offset) { if (info->bytes < bytes) { - printk(KERN_ERR "Found free space at %Lu, size %Lu," - "trying to use %Lu\n", - info->offset, info->bytes, bytes); + printk(KERN_ERR "Found free space at %llu, size %llu," + "trying to use %llu\n", + (unsigned long long)info->offset, + (unsigned long long)info->bytes, + (unsigned long long)bytes); WARN_ON(1); ret = -EINVAL; goto out; @@ -401,8 +408,6 @@ void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group, info = rb_entry(n, struct btrfs_free_space, offset_index); if (info->bytes >= bytes) count++; - //printk(KERN_INFO "offset=%Lu, bytes=%Lu\n", info->offset, - // info->bytes); } printk(KERN_INFO "%d blocks of free space at or bigger than bytes is" "\n", count); diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 80038c5ef7c..2aa79873eb4 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -129,7 +129,6 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, last_ino = key.objectid + 1; path->slots[0]++; } - // FIXME -ENOSPC BUG_ON(1); found: btrfs_release_path(root, path); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 068bad46338..1b35ea63b6c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -124,7 +124,7 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, * the btree. The caller should have done a btrfs_drop_extents so that * no overlapping inline items exist in the btree */ -static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, +static noinline int insert_inline_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode, u64 start, size_t size, size_t compressed_size, struct page **compressed_pages) @@ -148,7 +148,8 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, cur_size = compressed_size; } - path = btrfs_alloc_path(); if (!path) + path = btrfs_alloc_path(); + if (!path) return -ENOMEM; btrfs_set_trans_block_group(trans, inode); @@ -165,7 +166,6 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, BUG_ON(ret); if (ret) { err = ret; - printk("got bad ret %d\n", ret); goto fail; } leaf = path->nodes[0]; @@ -181,7 +181,7 @@ static int noinline insert_inline_extent(struct btrfs_trans_handle *trans, if (use_compress) { struct page *cpage; int i = 0; - while(compressed_size > 0) { + while (compressed_size > 0) { cpage = compressed_pages[i]; cur_size = min_t(unsigned long, compressed_size, PAGE_CACHE_SIZE); @@ -519,8 +519,7 @@ free_pages_out: WARN_ON(pages[i]->mapping); page_cache_release(pages[i]); } - if (pages) - kfree(pages); + kfree(pages); goto out; } @@ -549,7 +548,7 @@ static noinline int submit_compressed_extents(struct inode *inode, trans = btrfs_join_transaction(root, 1); - while(!list_empty(&async_cow->extents)) { + while (!list_empty(&async_cow->extents)) { async_extent = list_entry(async_cow->extents.next, struct async_extent, list); list_del(&async_extent->list); @@ -562,8 +561,8 @@ static noinline int submit_compressed_extents(struct inode *inode, unsigned long nr_written = 0; lock_extent(io_tree, async_extent->start, - async_extent->start + async_extent->ram_size - 1, - GFP_NOFS); + async_extent->start + + async_extent->ram_size - 1, GFP_NOFS); /* allocate blocks */ cow_file_range(inode, async_cow->locked_page, @@ -581,7 +580,7 @@ static noinline int submit_compressed_extents(struct inode *inode, if (!page_started) extent_write_locked_range(io_tree, inode, async_extent->start, - async_extent->start + + async_extent->start + async_extent->ram_size - 1, btrfs_get_extent, WB_SYNC_ALL); @@ -618,7 +617,7 @@ static noinline int submit_compressed_extents(struct inode *inode, set_bit(EXTENT_FLAG_PINNED, &em->flags); set_bit(EXTENT_FLAG_COMPRESSED, &em->flags); - while(1) { + while (1) { spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); spin_unlock(&em_tree->lock); @@ -651,11 +650,11 @@ static noinline int submit_compressed_extents(struct inode *inode, NULL, 1, 1, 0, 1, 1, 0); ret = btrfs_submit_compressed_write(inode, - async_extent->start, - async_extent->ram_size, - ins.objectid, - ins.offset, async_extent->pages, - async_extent->nr_pages); + async_extent->start, + async_extent->ram_size, + ins.objectid, + ins.offset, async_extent->pages, + async_extent->nr_pages); BUG_ON(ret); trans = btrfs_join_transaction(root, 1); @@ -735,14 +734,13 @@ static noinline int cow_file_range(struct inode *inode, btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); - while(disk_num_bytes > 0) { + while (disk_num_bytes > 0) { cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent); ret = btrfs_reserve_extent(trans, root, cur_alloc_size, root->sectorsize, 0, alloc_hint, (u64)-1, &ins, 1); - if (ret) { - BUG(); - } + BUG_ON(ret); + em = alloc_extent_map(GFP_NOFS); em->start = start; em->orig_start = em->start; @@ -755,7 +753,7 @@ static noinline int cow_file_range(struct inode *inode, em->bdev = root->fs_info->fs_devices->latest_bdev; set_bit(EXTENT_FLAG_PINNED, &em->flags); - while(1) { + while (1) { spin_lock(&em_tree->lock); ret = add_extent_mapping(em_tree, em); spin_unlock(&em_tree->lock); @@ -779,11 +777,9 @@ static noinline int cow_file_range(struct inode *inode, BUG_ON(ret); } - if (disk_num_bytes < cur_alloc_size) { - printk("num_bytes %Lu cur_alloc %Lu\n", disk_num_bytes, - cur_alloc_size); + if (disk_num_bytes < cur_alloc_size) break; - } + /* we're not doing compressed IO, don't unlock the first * page (which the caller expects to stay locked), don't * clear any dirty bits and don't set any writeback bits @@ -842,9 +838,8 @@ static noinline void async_cow_submit(struct btrfs_work *work) waitqueue_active(&root->fs_info->async_submit_wait)) wake_up(&root->fs_info->async_submit_wait); - if (async_cow->inode) { + if (async_cow->inode) submit_compressed_extents(async_cow->inode, async_cow); - } } static noinline void async_cow_free(struct btrfs_work *work) @@ -871,7 +866,7 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED | EXTENT_DELALLOC, 1, 0, GFP_NOFS); - while(start < end) { + while (start < end) { async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS); async_cow->inode = inode; async_cow->root = root; @@ -904,7 +899,7 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, limit)); } - while(atomic_read(&root->fs_info->async_submit_draining) && + while (atomic_read(&root->fs_info->async_submit_draining) && atomic_read(&root->fs_info->async_delalloc_pages)) { wait_event(root->fs_info->async_submit_wait, (atomic_read(&root->fs_info->async_delalloc_pages) == @@ -918,7 +913,7 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, return 0; } -static int noinline csum_exist_in_range(struct btrfs_root *root, +static noinline int csum_exist_in_range(struct btrfs_root *root, u64 bytenr, u64 num_bytes) { int ret; @@ -1146,13 +1141,13 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, if (btrfs_test_flag(inode, NODATACOW)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 1, nr_written); + page_started, 1, nr_written); else if (btrfs_test_flag(inode, PREALLOC)) ret = run_delalloc_nocow(inode, locked_page, start, end, - page_started, 0, nr_written); + page_started, 0, nr_written); else ret = cow_file_range_async(inode, locked_page, start, end, - page_started, nr_written); + page_started, nr_written); return ret; } @@ -1200,8 +1195,11 @@ static int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, spin_lock(&root->fs_info->delalloc_lock); if (end - start + 1 > root->fs_info->delalloc_bytes) { - printk("warning: delalloc account %Lu %Lu\n", - end - start + 1, root->fs_info->delalloc_bytes); + printk(KERN_INFO "btrfs warning: delalloc account " + "%llu %llu\n", + (unsigned long long)end - start + 1, + (unsigned long long) + root->fs_info->delalloc_bytes); root->fs_info->delalloc_bytes = 0; BTRFS_I(inode)->delalloc_bytes = 0; } else { @@ -1241,9 +1239,8 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, ret = btrfs_map_block(map_tree, READ, logical, &map_length, NULL, 0); - if (map_length < length + size) { + if (map_length < length + size) return 1; - } return 0; } @@ -1255,8 +1252,9 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset, * At IO completion time the cums attached on the ordered extent record * are inserted into the btree */ -static int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio, - int mirror_num, unsigned long bio_flags) +static int __btrfs_submit_bio_start(struct inode *inode, int rw, + struct bio *bio, int mirror_num, + unsigned long bio_flags) { struct btrfs_root *root = BTRFS_I(inode)->root; int ret = 0; @@ -1341,9 +1339,8 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end) { - if ((end & (PAGE_CACHE_SIZE - 1)) == 0) { + if ((end & (PAGE_CACHE_SIZE - 1)) == 0) WARN_ON(1); - } return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS); } @@ -1755,14 +1752,14 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, } local_irq_save(flags); kaddr = kmap_atomic(page, KM_IRQ0); - if (ret) { + if (ret) goto zeroit; - } + csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1); btrfs_csum_final(csum, (char *)&csum); - if (csum != private) { + if (csum != private) goto zeroit; - } + kunmap_atomic(kaddr, KM_IRQ0); local_irq_restore(flags); good: @@ -1773,9 +1770,10 @@ good: return 0; zeroit: - printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n", - page->mapping->host->i_ino, (unsigned long long)start, csum, - private); + printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u " + "private %llu\n", page->mapping->host->i_ino, + (unsigned long long)start, csum, + (unsigned long long)private); memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); kunmap_atomic(kaddr, KM_IRQ0); @@ -2097,9 +2095,8 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, /* * copy everything in the in-memory inode into the btree. */ -int noinline btrfs_update_inode(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct inode *inode) +noinline int btrfs_update_inode(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct inode *inode) { struct btrfs_inode_item *inode_item; struct btrfs_path *path; @@ -2174,7 +2171,7 @@ int btrfs_unlink_inode(struct btrfs_trans_handle *trans, inode->i_ino, dir->i_ino, &index); if (ret) { - printk("failed to delete reference to %.*s, " + printk(KERN_INFO "btrfs failed to delete reference to %.*s, " "inode %lu parent %lu\n", name_len, name, inode->i_ino, dir->i_ino); goto err; @@ -2280,9 +2277,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) /* now the directory is empty */ err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, dentry->d_name.name, dentry->d_name.len); - if (!err) { + if (!err) btrfs_i_size_write(inode, 0); - } fail_trans: nr = trans->blocks_used; @@ -2516,9 +2512,9 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, search_again: ret = btrfs_search_slot(trans, root, &key, path, -1, 1); - if (ret < 0) { + if (ret < 0) goto error; - } + if (ret > 0) { /* there are no items in the tree for us to truncate, we're * done @@ -2530,7 +2526,7 @@ search_again: path->slots[0]--; } - while(1) { + while (1) { fi = NULL; leaf = path->nodes[0]; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); @@ -2562,19 +2558,18 @@ search_again: item_end--; } if (item_end < new_size) { - if (found_type == BTRFS_DIR_ITEM_KEY) { + if (found_type == BTRFS_DIR_ITEM_KEY) found_type = BTRFS_INODE_ITEM_KEY; - } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { + else if (found_type == BTRFS_EXTENT_ITEM_KEY) found_type = BTRFS_EXTENT_DATA_KEY; - } else if (found_type == BTRFS_EXTENT_DATA_KEY) { + else if (found_type == BTRFS_EXTENT_DATA_KEY) found_type = BTRFS_XATTR_ITEM_KEY; - } else if (found_type == BTRFS_XATTR_ITEM_KEY) { + else if (found_type == BTRFS_XATTR_ITEM_KEY) found_type = BTRFS_INODE_REF_KEY; - } else if (found_type) { + else if (found_type) found_type--; - } else { + else break; - } btrfs_set_key_type(&key, found_type); goto next; } @@ -2656,7 +2651,7 @@ delete: pending_del_nr++; pending_del_slot = path->slots[0]; } else { - printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot); + BUG(); } } else { break; @@ -2938,9 +2933,10 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, namelen, 0); if (IS_ERR(di)) ret = PTR_ERR(di); - if (!di || IS_ERR(di)) { + + if (!di || IS_ERR(di)) goto out_err; - } + btrfs_dir_item_key_to_cpu(path->nodes[0], di, location); out: btrfs_free_path(path); @@ -3020,8 +3016,8 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p) static int btrfs_find_actor(struct inode *inode, void *opaque) { struct btrfs_iget_args *args = opaque; - return (args->ino == inode->i_ino && - args->root == BTRFS_I(inode)->root); + return args->ino == inode->i_ino && + args->root == BTRFS_I(inode)->root; } struct inode *btrfs_ilookup(struct super_block *s, u64 objectid, @@ -3085,7 +3081,7 @@ struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry) { - struct inode * inode; + struct inode *inode; struct btrfs_inode *bi = BTRFS_I(dir); struct btrfs_root *root = bi->root; struct btrfs_root *sub_root = root; @@ -3385,9 +3381,8 @@ int btrfs_set_inode_index(struct inode *dir, u64 *index) if (BTRFS_I(dir)->index_cnt == (u64)-1) { ret = btrfs_set_inode_index_count(dir); - if (ret) { + if (ret) return ret; - } } *index = BTRFS_I(dir)->index_cnt; @@ -3879,12 +3874,13 @@ static noinline int uncompress_inline(struct btrfs_path *path, /* * a bit scary, this does extent mapping from logical file offset to the disk. - * the ugly parts come from merging extents from the disk with the - * in-ram representation. This gets more complex because of the data=ordered code, + * the ugly parts come from merging extents from the disk with the in-ram + * representation. This gets more complex because of the data=ordered code, * where the in-ram extents might be locked pending data=ordered completion. * * This also copies inline extents directly into the page. */ + struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, size_t pg_offset, u64 start, u64 len, int create) @@ -4081,7 +4077,7 @@ again: extent_map_end(em) - 1, GFP_NOFS); goto insert; } else { - printk("unkknown found_type %d\n", found_type); + printk(KERN_ERR "btrfs unknown found_type %d\n", found_type); WARN_ON(1); } not_found: @@ -4093,7 +4089,11 @@ not_found_em: insert: btrfs_release_path(root, path); if (em->start > start || extent_map_end(em) <= start) { - printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len); + printk(KERN_ERR "Btrfs: bad extent! em: [%llu %llu] passed " + "[%llu %llu]\n", (unsigned long long)em->start, + (unsigned long long)em->len, + (unsigned long long)start, + (unsigned long long)len); err = -EIO; goto out; } @@ -4130,8 +4130,6 @@ insert: } } else { err = -EIO; - printk("failing to insert %Lu %Lu\n", - start, len); free_extent_map(em); em = NULL; } @@ -4147,9 +4145,8 @@ out: btrfs_free_path(path); if (trans) { ret = btrfs_end_transaction(trans, root); - if (!err) { + if (!err) err = ret; - } } if (err) { free_extent_map(em); @@ -4482,13 +4479,15 @@ void btrfs_destroy_inode(struct inode *inode) } spin_unlock(&BTRFS_I(inode)->root->list_lock); - while(1) { + while (1) { ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); if (!ordered) break; else { - printk("found ordered extent %Lu %Lu\n", - ordered->file_offset, ordered->len); + printk(KERN_ERR "btrfs found ordered " + "extent %llu %llu on inode cleanup\n", + (unsigned long long)ordered->file_offset, + (unsigned long long)ordered->len); btrfs_remove_ordered_extent(inode, ordered); btrfs_put_ordered_extent(ordered); btrfs_put_ordered_extent(ordered); @@ -4572,8 +4571,8 @@ static int btrfs_getattr(struct vfsmount *mnt, return 0; } -static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, - struct inode * new_dir,struct dentry *new_dentry) +static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry, + struct inode *new_dir, struct dentry *new_dentry) { struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(old_dir)->root; @@ -4663,7 +4662,7 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) return -EROFS; spin_lock(&root->fs_info->delalloc_lock); - while(!list_empty(head)) { + while (!list_empty(head)) { binode = list_entry(head->next, struct btrfs_inode, delalloc_inodes); inode = igrab(&binode->vfs_inode); @@ -4684,7 +4683,7 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root) * ordered extents get created before we return */ atomic_inc(&root->fs_info->async_submit_draining); - while(atomic_read(&root->fs_info->nr_async_submits) || + while (atomic_read(&root->fs_info->nr_async_submits) || atomic_read(&root->fs_info->async_delalloc_pages)) { wait_event(root->fs_info->async_submit_wait, (atomic_read(&root->fs_info->nr_async_submits) == 0 && diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index ba484aac1b9..c2aa33e3feb 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -311,7 +311,7 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name, * to see if is references the subvolume where we are * placing this new snapshot. */ - while(1) { + while (1) { if (!test || dir == snap_src->fs_info->sb->s_root || test == snap_src->fs_info->sb->s_root || @@ -319,7 +319,8 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name, break; } if (S_ISLNK(test->d_inode->i_mode)) { - printk("Symlink in snapshot path, failed\n"); + printk(KERN_INFO "Btrfs symlink in snapshot " + "path, failed\n"); error = -EMLINK; btrfs_free_path(path); goto out_drop_write; @@ -329,7 +330,8 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name, ret = btrfs_find_root_ref(snap_src->fs_info->tree_root, path, test_oid, parent_oid); if (ret == 0) { - printk("Snapshot creation failed, looping\n"); + printk(KERN_INFO "Btrfs snapshot creation " + "failed, looping\n"); error = -EMLINK; btrfs_free_path(path); goto out_drop_write; @@ -617,7 +619,8 @@ static noinline int btrfs_ioctl_snap_create(struct file *file, src_inode = src_file->f_path.dentry->d_inode; if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) { - printk("btrfs: Snapshot src from another FS\n"); + printk(KERN_INFO "btrfs: Snapshot src from " + "another FS\n"); ret = -EINVAL; fput(src_file); goto out; @@ -810,9 +813,6 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, ((off + len) & (bs-1))) goto out_unlock; - printk("final src extent is %llu~%llu\n", off, len); - printk("final dst extent is %llu~%llu\n", destoff, len); - /* do any pending delalloc/csum calc on src, one way or another, and lock file content */ while (1) { @@ -883,10 +883,13 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, comp = btrfs_file_extent_compression(leaf, extent); type = btrfs_file_extent_type(leaf, extent); if (type == BTRFS_FILE_EXTENT_REG) { - disko = btrfs_file_extent_disk_bytenr(leaf, extent); - diskl = btrfs_file_extent_disk_num_bytes(leaf, extent); + disko = btrfs_file_extent_disk_bytenr(leaf, + extent); + diskl = btrfs_file_extent_disk_num_bytes(leaf, + extent); datao = btrfs_file_extent_offset(leaf, extent); - datal = btrfs_file_extent_num_bytes(leaf, extent); + datal = btrfs_file_extent_num_bytes(leaf, + extent); } else if (type == BTRFS_FILE_EXTENT_INLINE) { /* take upper bound, may be compressed */ datal = btrfs_file_extent_ram_bytes(leaf, @@ -916,8 +919,6 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, extent = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); - printk(" orig disk %llu~%llu data %llu~%llu\n", - disko, diskl, datao, datal); if (off > key.offset) { datao += off - key.offset; @@ -929,8 +930,6 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, /* disko == 0 means it's a hole */ if (!disko) datao = 0; - printk(" final disk %llu~%llu data %llu~%llu\n", - disko, diskl, datao, datal); btrfs_set_file_extent_offset(leaf, extent, datao); @@ -952,12 +951,11 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, skip = off - key.offset; new_key.offset += skip; } + if (key.offset + datal > off+len) trim = key.offset + datal - (off+len); - printk("len %lld skip %lld trim %lld\n", - datal, skip, trim); + if (comp && (skip || trim)) { - printk("btrfs clone_range can't split compressed inline extents yet\n"); ret = -EINVAL; goto out; } @@ -969,7 +967,8 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, goto out; if (skip) { - u32 start = btrfs_file_extent_calc_inline_size(0); + u32 start = + btrfs_file_extent_calc_inline_size(0); memmove(buf+start, buf+start+skip, datal); } @@ -985,7 +984,7 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, btrfs_mark_buffer_dirty(leaf); } - next: +next: btrfs_release_path(root, path); key.offset++; } diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c index e30aa6e2958..39bae7761db 100644 --- a/fs/btrfs/locking.c +++ b/fs/btrfs/locking.c @@ -31,9 +31,10 @@ * difference in almost every workload, but spinning for the right amount of * time needs some help. * - * In general, we want to spin as long as the lock holder is doing btree searches, - * and we should give up if they are in more expensive code. + * In general, we want to spin as long as the lock holder is doing btree + * searches, and we should give up if they are in more expensive code. */ + int btrfs_tree_lock(struct extent_buffer *eb) { int i; diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index d9e232227da..a2094017027 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -39,11 +39,11 @@ static u64 entry_end(struct btrfs_ordered_extent *entry) static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, struct rb_node *node) { - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct btrfs_ordered_extent *entry; - while(*p) { + while (*p) { parent = *p; entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node); @@ -67,13 +67,13 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, struct rb_node **prev_ret) { - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct rb_node *prev = NULL; struct rb_node *test; struct btrfs_ordered_extent *entry; struct btrfs_ordered_extent *prev_entry = NULL; - while(n) { + while (n) { entry = rb_entry(n, struct btrfs_ordered_extent, rb_node); prev = n; prev_entry = entry; @@ -88,7 +88,7 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, if (!prev_ret) return NULL; - while(prev && file_offset >= entry_end(prev_entry)) { + while (prev && file_offset >= entry_end(prev_entry)) { test = rb_next(prev); if (!test) break; @@ -102,7 +102,7 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, if (prev) prev_entry = rb_entry(prev, struct btrfs_ordered_extent, rb_node); - while(prev && file_offset < entry_end(prev_entry)) { + while (prev && file_offset < entry_end(prev_entry)) { test = rb_prev(prev); if (!test) break; @@ -193,10 +193,8 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, node = tree_insert(&tree->tree, file_offset, &entry->rb_node); - if (node) { - printk("warning dup entry from add_ordered_extent\n"); - BUG(); - } + BUG_ON(node); + set_extent_ordered(&BTRFS_I(inode)->io_tree, file_offset, entry_end(entry) - 1, GFP_NOFS); @@ -282,7 +280,7 @@ int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) struct btrfs_ordered_sum *sum; if (atomic_dec_and_test(&entry->refs)) { - while(!list_empty(&entry->list)) { + while (!list_empty(&entry->list)) { cur = entry->list.next; sum = list_entry(cur, struct btrfs_ordered_sum, list); list_del(&sum->list); @@ -432,11 +430,10 @@ again: orig_end >> PAGE_CACHE_SHIFT); end = orig_end; - while(1) { + while (1) { ordered = btrfs_lookup_first_ordered_extent(inode, end); - if (!ordered) { + if (!ordered) break; - } if (ordered->file_offset > orig_end) { btrfs_put_ordered_extent(ordered); break; @@ -492,7 +489,7 @@ out: * if none is found */ struct btrfs_ordered_extent * -btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset) +btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset) { struct btrfs_ordered_inode_tree *tree; struct rb_node *node; @@ -553,7 +550,7 @@ int btrfs_ordered_update_i_size(struct inode *inode, * yet */ node = &ordered->rb_node; - while(1) { + while (1) { node = rb_prev(node); if (!node) break; @@ -581,9 +578,8 @@ int btrfs_ordered_update_i_size(struct inode *inode, * between our ordered extent and the next one. */ test = rb_entry(node, struct btrfs_ordered_extent, rb_node); - if (test->file_offset > entry_end(ordered)) { + if (test->file_offset > entry_end(ordered)) i_size_test = test->file_offset; - } } else { i_size_test = i_size_read(inode); } diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c index 64725c13aa1..5f8f218c100 100644 --- a/fs/btrfs/print-tree.c +++ b/fs/btrfs/print-tree.c @@ -24,13 +24,14 @@ static void print_chunk(struct extent_buffer *eb, struct btrfs_chunk *chunk) { int num_stripes = btrfs_chunk_num_stripes(eb, chunk); int i; - printk("\t\tchunk length %llu owner %llu type %llu num_stripes %d\n", + printk(KERN_INFO "\t\tchunk length %llu owner %llu type %llu " + "num_stripes %d\n", (unsigned long long)btrfs_chunk_length(eb, chunk), (unsigned long long)btrfs_chunk_owner(eb, chunk), (unsigned long long)btrfs_chunk_type(eb, chunk), num_stripes); for (i = 0 ; i < num_stripes ; i++) { - printk("\t\t\tstripe %d devid %llu offset %llu\n", i, + printk(KERN_INFO "\t\t\tstripe %d devid %llu offset %llu\n", i, (unsigned long long)btrfs_stripe_devid_nr(eb, chunk, i), (unsigned long long)btrfs_stripe_offset_nr(eb, chunk, i)); } @@ -38,8 +39,8 @@ static void print_chunk(struct extent_buffer *eb, struct btrfs_chunk *chunk) static void print_dev_item(struct extent_buffer *eb, struct btrfs_dev_item *dev_item) { - printk("\t\tdev item devid %llu " - "total_bytes %llu bytes used %Lu\n", + printk(KERN_INFO "\t\tdev item devid %llu " + "total_bytes %llu bytes used %llu\n", (unsigned long long)btrfs_device_id(eb, dev_item), (unsigned long long)btrfs_device_total_bytes(eb, dev_item), (unsigned long long)btrfs_device_bytes_used(eb, dev_item)); @@ -61,14 +62,15 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) struct btrfs_dev_extent *dev_extent; u32 type; - printk("leaf %llu total ptrs %d free space %d\n", + printk(KERN_INFO "leaf %llu total ptrs %d free space %d\n", (unsigned long long)btrfs_header_bytenr(l), nr, btrfs_leaf_free_space(root, l)); for (i = 0 ; i < nr ; i++) { item = btrfs_item_nr(l, i); btrfs_item_key_to_cpu(l, &key, i); type = btrfs_key_type(&key); - printk("\titem %d key (%llu %x %llu) itemoff %d itemsize %d\n", + printk(KERN_INFO "\titem %d key (%llu %x %llu) itemoff %d " + "itemsize %d\n", i, (unsigned long long)key.objectid, type, (unsigned long long)key.offset, @@ -76,33 +78,36 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) switch (type) { case BTRFS_INODE_ITEM_KEY: ii = btrfs_item_ptr(l, i, struct btrfs_inode_item); - printk("\t\tinode generation %llu size %llu mode %o\n", - (unsigned long long)btrfs_inode_generation(l, ii), + printk(KERN_INFO "\t\tinode generation %llu size %llu " + "mode %o\n", + (unsigned long long) + btrfs_inode_generation(l, ii), (unsigned long long)btrfs_inode_size(l, ii), btrfs_inode_mode(l, ii)); break; case BTRFS_DIR_ITEM_KEY: di = btrfs_item_ptr(l, i, struct btrfs_dir_item); btrfs_dir_item_key_to_cpu(l, di, &found_key); - printk("\t\tdir oid %llu type %u\n", + printk(KERN_INFO "\t\tdir oid %llu type %u\n", (unsigned long long)found_key.objectid, btrfs_dir_type(l, di)); break; case BTRFS_ROOT_ITEM_KEY: ri = btrfs_item_ptr(l, i, struct btrfs_root_item); - printk("\t\troot data bytenr %llu refs %u\n", - (unsigned long long)btrfs_disk_root_bytenr(l, ri), + printk(KERN_INFO "\t\troot data bytenr %llu refs %u\n", + (unsigned long long) + btrfs_disk_root_bytenr(l, ri), btrfs_disk_root_refs(l, ri)); break; case BTRFS_EXTENT_ITEM_KEY: ei = btrfs_item_ptr(l, i, struct btrfs_extent_item); - printk("\t\textent data refs %u\n", + printk(KERN_INFO "\t\textent data refs %u\n", btrfs_extent_refs(l, ei)); break; case BTRFS_EXTENT_REF_KEY: ref = btrfs_item_ptr(l, i, struct btrfs_extent_ref); - printk("\t\textent back ref root %llu gen %llu " - "owner %llu num_refs %lu\n", + printk(KERN_INFO "\t\textent back ref root %llu " + "gen %llu owner %llu num_refs %lu\n", (unsigned long long)btrfs_ref_root(l, ref), (unsigned long long)btrfs_ref_generation(l, ref), (unsigned long long)btrfs_ref_objectid(l, ref), @@ -114,26 +119,36 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) struct btrfs_file_extent_item); if (btrfs_file_extent_type(l, fi) == BTRFS_FILE_EXTENT_INLINE) { - printk("\t\tinline extent data size %u\n", - btrfs_file_extent_inline_len(l, fi)); + printk(KERN_INFO "\t\tinline extent data " + "size %u\n", + btrfs_file_extent_inline_len(l, fi)); break; } - printk("\t\textent data disk bytenr %llu nr %llu\n", - (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi), - (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi)); - printk("\t\textent data offset %llu nr %llu ram %llu\n", - (unsigned long long)btrfs_file_extent_offset(l, fi), - (unsigned long long)btrfs_file_extent_num_bytes(l, fi), - (unsigned long long)btrfs_file_extent_ram_bytes(l, fi)); + printk(KERN_INFO "\t\textent data disk bytenr %llu " + "nr %llu\n", + (unsigned long long) + btrfs_file_extent_disk_bytenr(l, fi), + (unsigned long long) + btrfs_file_extent_disk_num_bytes(l, fi)); + printk(KERN_INFO "\t\textent data offset %llu " + "nr %llu ram %llu\n", + (unsigned long long) + btrfs_file_extent_offset(l, fi), + (unsigned long long) + btrfs_file_extent_num_bytes(l, fi), + (unsigned long long) + btrfs_file_extent_ram_bytes(l, fi)); break; case BTRFS_BLOCK_GROUP_ITEM_KEY: bi = btrfs_item_ptr(l, i, struct btrfs_block_group_item); - printk("\t\tblock group used %llu\n", - (unsigned long long)btrfs_disk_block_group_used(l, bi)); + printk(KERN_INFO "\t\tblock group used %llu\n", + (unsigned long long) + btrfs_disk_block_group_used(l, bi)); break; case BTRFS_CHUNK_ITEM_KEY: - print_chunk(l, btrfs_item_ptr(l, i, struct btrfs_chunk)); + print_chunk(l, btrfs_item_ptr(l, i, + struct btrfs_chunk)); break; case BTRFS_DEV_ITEM_KEY: print_dev_item(l, btrfs_item_ptr(l, i, @@ -142,7 +157,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l) case BTRFS_DEV_EXTENT_KEY: dev_extent = btrfs_item_ptr(l, i, struct btrfs_dev_extent); - printk("\t\tdev extent chunk_tree %llu\n" + printk(KERN_INFO "\t\tdev extent chunk_tree %llu\n" "\t\tchunk objectid %llu chunk offset %llu " "length %llu\n", (unsigned long long) @@ -171,13 +186,13 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c) btrfs_print_leaf(root, c); return; } - printk("node %llu level %d total ptrs %d free spc %u\n", + printk(KERN_INFO "node %llu level %d total ptrs %d free spc %u\n", (unsigned long long)btrfs_header_bytenr(c), btrfs_header_level(c), nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr); for (i = 0; i < nr; i++) { btrfs_node_key_to_cpu(c, &key, i); - printk("\tkey %d (%llu %u %llu) block %llu\n", + printk(KERN_INFO "\tkey %d (%llu %u %llu) block %llu\n", i, (unsigned long long)key.objectid, key.type, diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c index a50ebb67055..6f0acc4c9ea 100644 --- a/fs/btrfs/ref-cache.c +++ b/fs/btrfs/ref-cache.c @@ -74,11 +74,11 @@ void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref) static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr, struct rb_node *node) { - struct rb_node ** p = &root->rb_node; - struct rb_node * parent = NULL; + struct rb_node **p = &root->rb_node; + struct rb_node *parent = NULL; struct btrfs_leaf_ref *entry; - while(*p) { + while (*p) { parent = *p; entry = rb_entry(parent, struct btrfs_leaf_ref, rb_node); @@ -98,10 +98,10 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr, static struct rb_node *tree_search(struct rb_root *root, u64 bytenr) { - struct rb_node * n = root->rb_node; + struct rb_node *n = root->rb_node; struct btrfs_leaf_ref *entry; - while(n) { + while (n) { entry = rb_entry(n, struct btrfs_leaf_ref, rb_node); WARN_ON(!entry->in_tree); @@ -127,7 +127,7 @@ int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen, return 0; spin_lock(&tree->lock); - while(!list_empty(&tree->list)) { + while (!list_empty(&tree->list)) { ref = list_entry(tree->list.next, struct btrfs_leaf_ref, list); BUG_ON(ref->tree != tree); if (ref->root_gen > max_root_gen) diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c index f99335a999d..b48650de447 100644 --- a/fs/btrfs/root-tree.c +++ b/fs/btrfs/root-tree.c @@ -132,8 +132,9 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root if (ret != 0) { btrfs_print_leaf(root, path->nodes[0]); - printk("unable to update root key %Lu %u %Lu\n", - key->objectid, key->type, key->offset); + printk(KERN_CRIT "unable to update root key %llu %u %llu\n", + (unsigned long long)key->objectid, key->type, + (unsigned long long)key->offset); BUG_ON(1); } @@ -159,9 +160,9 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root /* * at mount time we want to find all the old transaction snapshots that were in - * the process of being deleted if we crashed. This is any root item with an offset - * lower than the latest root. They need to be queued for deletion to finish - * what was happening when we crashed. + * the process of being deleted if we crashed. This is any root item with an + * offset lower than the latest root. They need to be queued for deletion to + * finish what was happening when we crashed. */ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid, struct btrfs_root *latest) @@ -188,7 +189,7 @@ again: ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto err; - while(1) { + while (1) { leaf = path->nodes[0]; nritems = btrfs_header_nritems(leaf); slot = path->slots[0]; @@ -258,11 +259,7 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, ret = btrfs_search_slot(trans, root, key, path, -1, 1); if (ret < 0) goto out; - if (ret) { -btrfs_print_leaf(root, path->nodes[0]); -printk("failed to del %Lu %u %Lu\n", key->objectid, key->type, key->offset); - } BUG_ON(ret != 0); leaf = path->nodes[0]; ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item); diff --git a/fs/btrfs/struct-funcs.c b/fs/btrfs/struct-funcs.c index 8d7f568009c..c0f7ecaf1e7 100644 --- a/fs/btrfs/struct-funcs.c +++ b/fs/btrfs/struct-funcs.c @@ -66,7 +66,7 @@ u##bits btrfs_##name(struct extent_buffer *eb, \ unsigned long map_len; \ u##bits res; \ err = map_extent_buffer(eb, offset, \ - sizeof(((type *)0)->member), \ + sizeof(((type *)0)->member), \ &map_token, &kaddr, \ &map_start, &map_len, KM_USER1); \ if (err) { \ @@ -103,7 +103,7 @@ void btrfs_set_##name(struct extent_buffer *eb, \ unsigned long map_start; \ unsigned long map_len; \ err = map_extent_buffer(eb, offset, \ - sizeof(((type *)0)->member), \ + sizeof(((type *)0)->member), \ &map_token, &kaddr, \ &map_start, &map_len, KM_USER1); \ if (err) { \ diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index ccdcb7bb7ad..b4c101d9322 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -55,18 +55,12 @@ static struct super_operations btrfs_super_ops; -static void btrfs_put_super (struct super_block * sb) +static void btrfs_put_super(struct super_block *sb) { struct btrfs_root *root = btrfs_sb(sb); int ret; ret = close_ctree(root); - if (ret) { - printk("close ctree returns %d\n", ret); - } -#if 0 - btrfs_sysfs_del_super(root->fs_info); -#endif sb->s_fs_info = NULL; } @@ -299,12 +293,12 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags, return error; } -static int btrfs_fill_super(struct super_block * sb, +static int btrfs_fill_super(struct super_block *sb, struct btrfs_fs_devices *fs_devices, - void * data, int silent) + void *data, int silent) { - struct inode * inode; - struct dentry * root_dentry; + struct inode *inode; + struct dentry *root_dentry; struct btrfs_super_block *disk_super; struct btrfs_root *tree_root; struct btrfs_inode *bi; @@ -479,8 +473,10 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags, root = dget(s->s_root); else { mutex_lock(&s->s_root->d_inode->i_mutex); - root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name)); + root = lookup_one_len(subvol_name, s->s_root, + strlen(subvol_name)); mutex_unlock(&s->s_root->d_inode->i_mutex); + if (IS_ERR(root)) { up_write(&s->s_umount); deactivate_super(s); @@ -557,8 +553,9 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) buf->f_bavail = buf->f_bfree; buf->f_bsize = dentry->d_sb->s_blocksize; buf->f_type = BTRFS_SUPER_MAGIC; + /* We treat it as constant endianness (it doesn't matter _which_) - because we want the fsid to come out the same whether mounted + because we want the fsid to come out the same whether mounted on a big-endian or little-endian host */ buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); @@ -658,7 +655,7 @@ static int btrfs_interface_init(void) static void btrfs_interface_exit(void) { if (misc_deregister(&btrfs_misc) < 0) - printk("misc_deregister failed for control device"); + printk(KERN_INFO "misc_deregister failed for control device"); } static int __init init_btrfs_fs(void) diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c index 04087c02084..a240b6fa81d 100644 --- a/fs/btrfs/sysfs.c +++ b/fs/btrfs/sysfs.c @@ -67,7 +67,8 @@ struct btrfs_root_attr { }; #define ROOT_ATTR(name, mode, show, store) \ -static struct btrfs_root_attr btrfs_root_attr_##name = __ATTR(name, mode, show, store) +static struct btrfs_root_attr btrfs_root_attr_##name = __ATTR(name, mode, \ + show, store) ROOT_ATTR(blocks_used, 0444, root_blocks_used_show, NULL); ROOT_ATTR(block_limit, 0644, root_block_limit_show, NULL); @@ -86,7 +87,8 @@ struct btrfs_super_attr { }; #define SUPER_ATTR(name, mode, show, store) \ -static struct btrfs_super_attr btrfs_super_attr_##name = __ATTR(name, mode, show, store) +static struct btrfs_super_attr btrfs_super_attr_##name = __ATTR(name, mode, \ + show, store) SUPER_ATTR(blocks_used, 0444, super_blocks_used_show, NULL); SUPER_ATTR(total_blocks, 0444, super_total_blocks_show, NULL); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 4e7b56e9d3a..56ab1f5ea11 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -28,9 +28,6 @@ #include "ref-cache.h" #include "tree-log.h" -extern struct kmem_cache *btrfs_trans_handle_cachep; -extern struct kmem_cache *btrfs_transaction_cachep; - #define BTRFS_ROOT_TRANS_TAG 0 static noinline void put_transaction(struct btrfs_transaction *transaction) @@ -85,10 +82,10 @@ static noinline int join_transaction(struct btrfs_root *root) } /* - * this does all the record keeping required to make sure that a - * reference counted root is properly recorded in a given transaction. - * This is required to make sure the old root from before we joined the transaction - * is deleted when the transaction commits + * this does all the record keeping required to make sure that a reference + * counted root is properly recorded in a given transaction. This is required + * to make sure the old root from before we joined the transaction is deleted + * when the transaction commits */ noinline int btrfs_record_root_in_trans(struct btrfs_root *root) { @@ -144,7 +141,7 @@ static void wait_current_trans(struct btrfs_root *root) if (cur_trans && cur_trans->blocked) { DEFINE_WAIT(wait); cur_trans->use_count++; - while(1) { + while (1) { prepare_to_wait(&root->fs_info->transaction_wait, &wait, TASK_UNINTERRUPTIBLE); if (cur_trans->blocked) { @@ -213,7 +210,7 @@ static noinline int wait_for_commit(struct btrfs_root *root, { DEFINE_WAIT(wait); mutex_lock(&root->fs_info->trans_mutex); - while(!commit->commit_done) { + while (!commit->commit_done) { prepare_to_wait(&commit->commit_wait, &wait, TASK_UNINTERRUPTIBLE); if (commit->commit_done) @@ -228,8 +225,8 @@ static noinline int wait_for_commit(struct btrfs_root *root, } /* - * rate limit against the drop_snapshot code. This helps to slow down new operations - * if the drop_snapshot code isn't able to keep up. + * rate limit against the drop_snapshot code. This helps to slow down new + * operations if the drop_snapshot code isn't able to keep up. */ static void throttle_on_drops(struct btrfs_root *root) { @@ -332,12 +329,12 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, u64 end; unsigned long index; - while(1) { + while (1) { ret = find_first_extent_bit(dirty_pages, start, &start, &end, EXTENT_DIRTY); if (ret) break; - while(start <= end) { + while (start <= end) { cond_resched(); index = start >> PAGE_CACHE_SHIFT; @@ -368,14 +365,14 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, page_cache_release(page); } } - while(1) { + while (1) { ret = find_first_extent_bit(dirty_pages, 0, &start, &end, EXTENT_DIRTY); if (ret) break; clear_extent_dirty(dirty_pages, start, end, GFP_NOFS); - while(start <= end) { + while (start <= end) { index = start >> PAGE_CACHE_SHIFT; start = (u64)(index + 1) << PAGE_CACHE_SHIFT; page = find_get_page(btree_inode->i_mapping, index); @@ -431,7 +428,7 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans, btrfs_write_dirty_block_groups(trans, root); btrfs_extent_post_op(trans, root); - while(1) { + while (1) { old_root_bytenr = btrfs_root_bytenr(&root->root_item); if (old_root_bytenr == root->node->start) break; @@ -472,7 +469,7 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, btrfs_extent_post_op(trans, fs_info->tree_root); - while(!list_empty(&fs_info->dirty_cowonly_roots)) { + while (!list_empty(&fs_info->dirty_cowonly_roots)) { next = fs_info->dirty_cowonly_roots.next; list_del_init(next); root = list_entry(next, struct btrfs_root, dirty_list); @@ -521,7 +518,7 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans, int err = 0; u32 refs; - while(1) { + while (1) { ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, ARRAY_SIZE(gang), BTRFS_ROOT_TRANS_TAG); @@ -653,7 +650,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, int ret = 0; int err; - while(!list_empty(list)) { + while (!list_empty(list)) { struct btrfs_root *root; dirty = list_entry(list->prev, struct btrfs_dirty_root, list); @@ -663,13 +660,12 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, root = dirty->latest_root; atomic_inc(&root->fs_info->throttles); - while(1) { + while (1) { trans = btrfs_start_transaction(tree_root, 1); mutex_lock(&root->fs_info->drop_mutex); ret = btrfs_drop_snapshot(trans, dirty->root); - if (ret != -EAGAIN) { + if (ret != -EAGAIN) break; - } mutex_unlock(&root->fs_info->drop_mutex); err = btrfs_update_root(trans, @@ -874,7 +870,7 @@ static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans, struct list_head *head = &trans->transaction->pending_snapshots; int ret; - while(!list_empty(head)) { + while (!list_empty(head)) { pending = list_entry(head->next, struct btrfs_pending_snapshot, list); ret = finish_pending_snapshot(fs_info, pending); @@ -1076,9 +1072,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, kmem_cache_free(btrfs_trans_handle_cachep, trans); - if (root->fs_info->closing) { + if (root->fs_info->closing) drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots); - } return ret; } diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index ffe7f639732..ea292117f88 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -66,9 +66,9 @@ static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans, trans->block_group = BTRFS_I(inode)->block_group; } -static inline void btrfs_update_inode_block_group(struct - btrfs_trans_handle *trans, - struct inode *inode) +static inline void btrfs_update_inode_block_group( + struct btrfs_trans_handle *trans, + struct inode *inode) { BTRFS_I(inode)->block_group = trans->block_group; } diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c index a6a3956cedf..3e8358c3616 100644 --- a/fs/btrfs/tree-defrag.c +++ b/fs/btrfs/tree-defrag.c @@ -23,10 +23,11 @@ #include "transaction.h" #include "locking.h" -/* defrag all the leaves in a given btree. If cache_only == 1, don't read things - * from disk, otherwise read all the leaves and try to get key order to +/* defrag all the leaves in a given btree. If cache_only == 1, don't read + * things from disk, otherwise read all the leaves and try to get key order to * better reflect disk order */ + int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, struct btrfs_root *root, int cache_only) { @@ -65,9 +66,9 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans, level = btrfs_header_level(root->node); orig_level = level; - if (level == 0) { + if (level == 0) goto out; - } + if (root->defrag_progress.objectid == 0) { struct extent_buffer *root_node; u32 nritems; diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index b1c2921f5be..3a72a1b6c24 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -829,7 +829,7 @@ conflict_again: */ ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); ptr_end = ptr + btrfs_item_size_nr(leaf, path->slots[0]); - while(ptr < ptr_end) { + while (ptr < ptr_end) { victim_ref = (struct btrfs_inode_ref *)ptr; victim_name_len = btrfs_inode_ref_name_len(leaf, victim_ref); @@ -938,9 +938,8 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans, file_bytes = (item_size / csum_size) * root->sectorsize; sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS); - if (!sums) { + if (!sums) return -ENOMEM; - } INIT_LIST_HEAD(&sums->list); sums->len = file_bytes; @@ -952,7 +951,7 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans, sector_sum = sums->sums; cur_offset = key->offset; ptr = btrfs_item_ptr_offset(eb, slot); - while(item_size > 0) { + while (item_size > 0) { sector_sum->bytenr = cur_offset; read_extent_buffer(eb, §or_sum->sum, ptr, csum_size); sector_sum++; @@ -995,7 +994,7 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans, path = btrfs_alloc_path(); - while(1) { + while (1) { ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) break; @@ -1012,7 +1011,7 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans, ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]); ptr_end = ptr + btrfs_item_size_nr(path->nodes[0], path->slots[0]); - while(ptr < ptr_end) { + while (ptr < ptr_end) { struct btrfs_inode_ref *ref; ref = (struct btrfs_inode_ref *)ptr; @@ -1048,7 +1047,7 @@ static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans, key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID; key.type = BTRFS_ORPHAN_ITEM_KEY; key.offset = (u64)-1; - while(1) { + while (1) { ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) break; @@ -1206,8 +1205,7 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans, if (key->type == BTRFS_DIR_ITEM_KEY) { dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid, name, name_len, 1); - } - else if (key->type == BTRFS_DIR_INDEX_KEY) { + } else if (key->type == BTRFS_DIR_INDEX_KEY) { dst_di = btrfs_lookup_dir_index_item(trans, root, path, key->objectid, key->offset, name, @@ -1282,7 +1280,7 @@ static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans, ptr = btrfs_item_ptr_offset(eb, slot); ptr_end = ptr + item_size; - while(ptr < ptr_end) { + while (ptr < ptr_end) { di = (struct btrfs_dir_item *)ptr; name_len = btrfs_dir_name_len(eb, di); ret = replay_one_name(trans, root, path, eb, di, key); @@ -1408,7 +1406,7 @@ again: item_size = btrfs_item_size_nr(eb, slot); ptr = btrfs_item_ptr_offset(eb, slot); ptr_end = ptr + item_size; - while(ptr < ptr_end) { + while (ptr < ptr_end) { di = (struct btrfs_dir_item *)ptr; name_len = btrfs_dir_name_len(eb, di); name = kmalloc(name_len, GFP_NOFS); @@ -1513,14 +1511,14 @@ static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans, again: range_start = 0; range_end = 0; - while(1) { + while (1) { ret = find_dir_range(log, path, dirid, key_type, &range_start, &range_end); if (ret != 0) break; dir_key.offset = range_start; - while(1) { + while (1) { int nritems; ret = btrfs_search_slot(NULL, root, &dir_key, path, 0, 0); @@ -1676,7 +1674,7 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb, return 0; } -static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans, +static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level, struct walk_control *wc) @@ -1694,7 +1692,7 @@ static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans, WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); - while(*level > 0) { + while (*level > 0) { WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); cur = path->nodes[*level]; @@ -1753,11 +1751,11 @@ static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans, WARN_ON(*level < 0); WARN_ON(*level >= BTRFS_MAX_LEVEL); - if (path->nodes[*level] == root->node) { + if (path->nodes[*level] == root->node) parent = path->nodes[*level]; - } else { + else parent = path->nodes[*level + 1]; - } + bytenr = path->nodes[*level]->start; blocksize = btrfs_level_size(root, *level); @@ -1790,7 +1788,7 @@ static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans, return 0; } -static int noinline walk_up_log_tree(struct btrfs_trans_handle *trans, +static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int *level, struct walk_control *wc) @@ -1801,7 +1799,7 @@ static int noinline walk_up_log_tree(struct btrfs_trans_handle *trans, int slot; int ret; - for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { + for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { slot = path->slots[i]; if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { struct extent_buffer *node; @@ -1875,7 +1873,7 @@ static int walk_log_tree(struct btrfs_trans_handle *trans, extent_buffer_get(log->node); path->slots[level] = 0; - while(1) { + while (1) { wret = walk_down_log_tree(trans, log, path, &level, wc); if (wret > 0) break; @@ -1941,7 +1939,7 @@ static int wait_log_commit(struct btrfs_root *log) schedule(); finish_wait(&log->fs_info->tree_log_wait, &wait); mutex_lock(&log->fs_info->tree_log_mutex); - } while(transid == log->fs_info->tree_log_transid && + } while (transid == log->fs_info->tree_log_transid && atomic_read(&log->fs_info->tree_log_commit)); return 0; } @@ -1965,13 +1963,13 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans, } atomic_set(&log->fs_info->tree_log_commit, 1); - while(1) { + while (1) { batch = log->fs_info->tree_log_batch; mutex_unlock(&log->fs_info->tree_log_mutex); schedule_timeout_uninterruptible(1); mutex_lock(&log->fs_info->tree_log_mutex); - while(atomic_read(&log->fs_info->tree_log_writers)) { + while (atomic_read(&log->fs_info->tree_log_writers)) { DEFINE_WAIT(wait); prepare_to_wait(&log->fs_info->tree_log_wait, &wait, TASK_UNINTERRUPTIBLE); @@ -2030,7 +2028,7 @@ int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root) ret = walk_log_tree(trans, log, &wc); BUG_ON(ret); - while(1) { + while (1) { ret = find_first_extent_bit(&log->dirty_log_pages, 0, &start, &end, EXTENT_DIRTY); if (ret) @@ -2287,9 +2285,8 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans, struct btrfs_key tmp; btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]); - if (key_type == tmp.type) { + if (key_type == tmp.type) first_offset = max(min_offset, tmp.offset) + 1; - } } goto done; } @@ -2319,7 +2316,7 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans, * we have a block from this transaction, log every item in it * from our directory */ - while(1) { + while (1) { struct btrfs_key tmp; src = path->nodes[0]; nritems = btrfs_header_nritems(src); @@ -2396,7 +2393,7 @@ static noinline int log_directory_changes(struct btrfs_trans_handle *trans, again: min_key = 0; max_key = 0; - while(1) { + while (1) { ret = log_dir_items(trans, root, inode, path, dst_path, key_type, min_key, &max_key); @@ -2432,7 +2429,7 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans, key.type = max_key_type; key.offset = (u64)-1; - while(1) { + while (1) { ret = btrfs_search_slot(trans, log, &key, path, -1, 1); if (ret != 1) @@ -2481,7 +2478,7 @@ static noinline int copy_extent_csums(struct btrfs_trans_handle *trans, list_add_tail(&sums->list, list); path = btrfs_alloc_path(); - while(disk_bytenr < end) { + while (disk_bytenr < end) { if (!item || disk_bytenr < item_start_offset || disk_bytenr >= item_last_offset) { struct btrfs_key found_key; @@ -2496,7 +2493,8 @@ static noinline int copy_extent_csums(struct btrfs_trans_handle *trans, if (ret == -ENOENT || ret == -EFBIG) ret = 0; sum = 0; - printk("log no csum found for byte %llu\n", + printk(KERN_INFO "log no csum found for " + "byte %llu\n", (unsigned long long)disk_bytenr); item = NULL; btrfs_release_path(root, path); @@ -2643,7 +2641,7 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, * we have to do this after the loop above to avoid changing the * log tree while trying to change the log tree. */ - while(!list_empty(&ordered_sums)) { + while (!list_empty(&ordered_sums)) { struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next, struct btrfs_ordered_sum, list); @@ -2736,7 +2734,7 @@ static int __btrfs_log_inode(struct btrfs_trans_handle *trans, BUG_ON(ret); path->keep_locks = 1; - while(1) { + while (1) { ins_nr = 0; ret = btrfs_search_forward(root, &min_key, &max_key, path, 0, trans->transid); @@ -2848,7 +2846,7 @@ int btrfs_log_dentry(struct btrfs_trans_handle *trans, start_log_trans(trans, root); sb = dentry->d_inode->i_sb; - while(1) { + while (1) { ret = __btrfs_log_inode(trans, root, dentry->d_inode, inode_only); BUG_ON(ret); @@ -2919,7 +2917,7 @@ again: key.offset = (u64)-1; btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); - while(1) { + while (1) { ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0); if (ret < 0) break; diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 6672adcec9f..b187b537888 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -140,7 +140,7 @@ static noinline struct btrfs_fs_devices *find_fsid(u8 *fsid) * the list if the block device is congested. This way, multiple devices * can make progress from a single worker thread. */ -static int noinline run_scheduled_bios(struct btrfs_device *device) +static noinline int run_scheduled_bios(struct btrfs_device *device) { struct bio *pending; struct backing_dev_info *bdi; @@ -187,7 +187,7 @@ loop: } spin_unlock(&device->io_lock); - while(pending) { + while (pending) { cur = pending; pending = pending->bi_next; cur->bi_next = NULL; @@ -458,7 +458,7 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices, bdev = open_bdev_exclusive(device->name, flags, holder); if (IS_ERR(bdev)) { - printk("open %s failed\n", device->name); + printk(KERN_INFO "open %s failed\n", device->name); goto error; } set_blocksize(bdev, 4096); @@ -570,14 +570,15 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder, devid = le64_to_cpu(disk_super->dev_item.devid); transid = btrfs_super_generation(disk_super); if (disk_super->label[0]) - printk("device label %s ", disk_super->label); + printk(KERN_INFO "device label %s ", disk_super->label); else { /* FIXME, make a readl uuid parser */ - printk("device fsid %llx-%llx ", + printk(KERN_INFO "device fsid %llx-%llx ", *(unsigned long long *)disk_super->fsid, *(unsigned long long *)(disk_super->fsid + 8)); } - printk("devid %Lu transid %Lu %s\n", devid, transid, path); + printk(KERN_INFO "devid %llu transid %llu %s\n", + (unsigned long long)devid, (unsigned long long)transid, path); ret = device_list_add(path, disk_super, devid, fs_devices_ret); brelse(bh); @@ -683,9 +684,8 @@ no_more_items: goto check_pending; } } - if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { + if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) goto next; - } start_found = 1; dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); @@ -1001,14 +1001,16 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) && root->fs_info->fs_devices->rw_devices <= 4) { - printk("btrfs: unable to go below four devices on raid10\n"); + printk(KERN_ERR "btrfs: unable to go below four devices " + "on raid10\n"); ret = -EINVAL; goto out; } if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) && root->fs_info->fs_devices->rw_devices <= 2) { - printk("btrfs: unable to go below two devices on raid1\n"); + printk(KERN_ERR "btrfs: unable to go below two " + "devices on raid1\n"); ret = -EINVAL; goto out; } @@ -1031,7 +1033,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) bh = NULL; disk_super = NULL; if (!device) { - printk("btrfs: no missing devices found to remove\n"); + printk(KERN_ERR "btrfs: no missing devices found to " + "remove\n"); goto out; } } else { @@ -1060,7 +1063,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path) } if (device->writeable && root->fs_info->fs_devices->rw_devices == 1) { - printk("btrfs: unable to remove the only writeable device\n"); + printk(KERN_ERR "btrfs: unable to remove the only writeable " + "device\n"); ret = -EINVAL; goto error_brelse; } @@ -1286,9 +1290,8 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) return -EINVAL; bdev = open_bdev_exclusive(device_path, 0, root->fs_info->bdev_holder); - if (!bdev) { + if (!bdev) return -EIO; - } if (root->fs_info->fs_devices->seeding) { seeding_dev = 1; @@ -1401,8 +1404,8 @@ error: goto out; } -static int noinline btrfs_update_device(struct btrfs_trans_handle *trans, - struct btrfs_device *device) +static noinline int btrfs_update_device(struct btrfs_trans_handle *trans, + struct btrfs_device *device) { int ret; struct btrfs_path *path; @@ -1563,7 +1566,7 @@ static int btrfs_relocate_chunk(struct btrfs_root *root, int ret; int i; - printk("btrfs relocating chunk %llu\n", + printk(KERN_INFO "btrfs relocating chunk %llu\n", (unsigned long long)chunk_offset); root = root->fs_info->chunk_root; extent_root = root->fs_info->extent_root; @@ -1748,7 +1751,7 @@ int btrfs_balance(struct btrfs_root *dev_root) key.offset = (u64)-1; key.type = BTRFS_CHUNK_ITEM_KEY; - while(1) { + while (1) { ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); if (ret < 0) goto error; @@ -1916,7 +1919,7 @@ static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, return 0; } -static u64 noinline chunk_bytes_by_type(u64 type, u64 calc_size, +static noinline u64 chunk_bytes_by_type(u64 type, u64 calc_size, int num_stripes, int sub_stripes) { if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) @@ -2041,7 +2044,7 @@ again: min_free += 1024 * 1024; INIT_LIST_HEAD(&private_devs); - while(index < num_stripes) { + while (index < num_stripes) { device = list_entry(cur, struct btrfs_device, dev_alloc_list); BUG_ON(!device->writeable); if (device->total_bytes > device->bytes_used) @@ -2242,7 +2245,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, return 0; } -static int noinline init_first_rw_device(struct btrfs_trans_handle *trans, +static noinline int init_first_rw_device(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_device *device) { @@ -2338,7 +2341,7 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) { struct extent_map *em; - while(1) { + while (1) { spin_lock(&tree->map_tree.lock); em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); if (em) @@ -2413,9 +2416,8 @@ static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, int max_errors = 0; struct btrfs_multi_bio *multi = NULL; - if (multi_ret && !(rw & (1 << BIO_RW))) { + if (multi_ret && !(rw & (1 << BIO_RW))) stripes_allocated = 1; - } again: if (multi_ret) { multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), @@ -2434,7 +2436,9 @@ again: return 0; if (!em) { - printk("unable to find logical %Lu len %Lu\n", logical, *length); + printk(KERN_CRIT "unable to find logical %llu len %llu\n", + (unsigned long long)logical, + (unsigned long long)*length); BUG(); } @@ -2541,9 +2545,8 @@ again: device = map->stripes[stripe_index].dev; if (device->bdev) { bdi = blk_get_backing_dev_info(device->bdev); - if (bdi->unplug_io_fn) { + if (bdi->unplug_io_fn) bdi->unplug_io_fn(bdi, unplug_page); - } } } else { multi->stripes[i].physical = @@ -2717,7 +2720,7 @@ struct async_sched { * This will add one bio to the pending list for a device and make sure * the work struct is scheduled. */ -static int noinline schedule_bio(struct btrfs_root *root, +static noinline int schedule_bio(struct btrfs_root *root, struct btrfs_device *device, int rw, struct bio *bio) { @@ -2785,8 +2788,10 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, total_devs = multi->num_stripes; if (map_length < length) { - printk("mapping failed logical %Lu bio len %Lu " - "len %Lu\n", logical, length, map_length); + printk(KERN_CRIT "mapping failed logical %llu bio len %llu " + "len %llu\n", (unsigned long long)logical, + (unsigned long long)length, + (unsigned long long)map_length); BUG(); } multi->end_io = first_bio->bi_end_io; @@ -2794,7 +2799,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, multi->orig_bio = first_bio; atomic_set(&multi->stripes_pending, multi->num_stripes); - while(dev_nr < total_devs) { + while (dev_nr < total_devs) { if (total_devs > 1) { if (dev_nr < total_devs - 1) { bio = bio_clone(first_bio, GFP_NOFS); @@ -3058,7 +3063,8 @@ static int read_one_dev(struct btrfs_root *root, return -EIO; if (!device) { - printk("warning devid %Lu missing\n", devid); + printk(KERN_WARNING "warning devid %llu missing\n", + (unsigned long long)devid); device = add_missing_dev(root, devid, dev_uuid); if (!device) return -ENOMEM; @@ -3078,12 +3084,6 @@ static int read_one_dev(struct btrfs_root *root, if (device->writeable) device->fs_devices->total_rw_bytes += device->total_bytes; ret = 0; -#if 0 - ret = btrfs_open_device(device); - if (ret) { - kfree(device); - } -#endif return ret; } @@ -3174,7 +3174,7 @@ int btrfs_read_chunk_tree(struct btrfs_root *root) key.type = 0; again: ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); - while(1) { + while (1) { leaf = path->nodes[0]; slot = path->slots[0]; if (slot >= btrfs_header_nritems(leaf)) { diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c index 4146f0710e6..7f332e27089 100644 --- a/fs/btrfs/xattr.c +++ b/fs/btrfs/xattr.c @@ -264,7 +264,8 @@ struct xattr_handler *btrfs_xattr_handlers[] = { */ static bool btrfs_is_valid_xattr(const char *name) { - return !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) || + return !strncmp(name, XATTR_SECURITY_PREFIX, + XATTR_SECURITY_PREFIX_LEN) || !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN) || !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) || !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN); diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c index c4617cde6c7..ecfbce836d3 100644 --- a/fs/btrfs/zlib.c +++ b/fs/btrfs/zlib.c @@ -152,7 +152,7 @@ static int free_workspace(struct workspace *workspace) static void free_workspaces(void) { struct workspace *workspace; - while(!list_empty(&idle_workspace)) { + while (!list_empty(&idle_workspace)) { workspace = list_entry(idle_workspace.next, struct workspace, list); list_del(&workspace->list); @@ -397,12 +397,10 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, ret = -1; goto out; } - while(workspace->inf_strm.total_in < srclen) { + while (workspace->inf_strm.total_in < srclen) { ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH); - if (ret != Z_OK && ret != Z_STREAM_END) { + if (ret != Z_OK && ret != Z_STREAM_END) break; - } - /* * buf start is the byte offset we're of the start of * our workspace buffer @@ -424,16 +422,14 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, /* we didn't make progress in this inflate * call, we're done */ - if (ret != Z_STREAM_END) { + if (ret != Z_STREAM_END) ret = -1; - } break; } /* we haven't yet hit data corresponding to this page */ - if (total_out <= start_byte) { + if (total_out <= start_byte) goto next; - } /* * the start of the data we care about is offset into @@ -448,7 +444,7 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, current_buf_start = buf_start; /* copy bytes from the working buffer into the pages */ - while(working_bytes > 0) { + while (working_bytes > 0) { bytes = min(PAGE_CACHE_SIZE - pg_offset, PAGE_CACHE_SIZE - buf_offset); bytes = min(bytes, working_bytes); @@ -471,6 +467,7 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, ret = 0; goto done; } + page_out = bvec[page_out_index].bv_page; pg_offset = 0; page_bytes_left = PAGE_CACHE_SIZE; @@ -480,9 +477,8 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in, * make sure our new page is covered by this * working buffer */ - if (total_out <= start_byte) { + if (total_out <= start_byte) goto next; - } /* the next page in the biovec might not * be adjacent to the last page, but it @@ -517,11 +513,10 @@ next: PAGE_CACHE_SIZE); } } - if (ret != Z_STREAM_END) { + if (ret != Z_STREAM_END) ret = -1; - } else { + else ret = 0; - } done: zlib_inflateEnd(&workspace->inf_strm); if (data_in) @@ -579,16 +574,15 @@ int btrfs_zlib_decompress(unsigned char *data_in, goto out; } - while(bytes_left > 0) { + while (bytes_left > 0) { unsigned long buf_start; unsigned long buf_offset; unsigned long bytes; unsigned long pg_offset = 0; ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH); - if (ret != Z_OK && ret != Z_STREAM_END) { + if (ret != Z_OK && ret != Z_STREAM_END) break; - } buf_start = total_out; total_out = workspace->inf_strm.total_out; @@ -598,15 +592,13 @@ int btrfs_zlib_decompress(unsigned char *data_in, break; } - if (total_out <= start_byte) { + if (total_out <= start_byte) goto next; - } - if (total_out > start_byte && buf_start < start_byte) { + if (total_out > start_byte && buf_start < start_byte) buf_offset = start_byte - buf_start; - } else { + else buf_offset = 0; - } bytes = min(PAGE_CACHE_SIZE - pg_offset, PAGE_CACHE_SIZE - buf_offset); @@ -622,11 +614,12 @@ next: workspace->inf_strm.next_out = workspace->buf; workspace->inf_strm.avail_out = PAGE_CACHE_SIZE; } - if (ret != Z_STREAM_END && bytes_left != 0) { + + if (ret != Z_STREAM_END && bytes_left != 0) ret = -1; - } else { + else ret = 0; - } + zlib_inflateEnd(&workspace->inf_strm); out: free_workspace(workspace); -- cgit v1.2.3-70-g09d2 From 180591bcfed1a2cec048abb21d3dab840625caab Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Tue, 6 Jan 2009 09:58:06 -0500 Subject: Btrfs: Use btrfs_join_transaction to avoid deadlocks during snapshot creation Snapshot creation happens at a specific time during transaction commit. We need to make sure the code called by snapshot creation doesn't wait for the running transaction to commit. This changes btrfs_delete_inode and finish_pending_snaps to use btrfs_join_transaction instead of btrfs_start_transaction to avoid deadlocks. It would be better if btrfs_delete_inode didn't use the join, but the call path that triggers it is: btrfs_commit_transaction->create_pending_snapshots-> create_pending_snapshot->btrfs_lookup_dentry-> fixup_tree_root_location->btrfs_read_fs_root-> btrfs_read_fs_root_no_name->btrfs_orphan_cleanup->iput This will be fixed in a later patch by moving the orphan cleanup to the cleaner thread. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 2 +- fs/btrfs/transaction.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 1b35ea63b6c..c0ca9c3723c 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2886,7 +2886,7 @@ void btrfs_delete_inode(struct inode *inode) btrfs_wait_ordered_range(inode, 0, (u64)-1); btrfs_i_size_write(inode, 0); - trans = btrfs_start_transaction(root, 1); + trans = btrfs_join_transaction(root, 1); btrfs_set_trans_block_group(trans, inode); ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, 0); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 56ab1f5ea11..8a08f944334 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -800,7 +800,7 @@ static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info, parent_inode = pending->dentry->d_parent->d_inode; parent_root = BTRFS_I(parent_inode)->root; - trans = btrfs_start_transaction(parent_root, 1); + trans = btrfs_join_transaction(parent_root, 1); /* * insert the directory item -- cgit v1.2.3-70-g09d2 From 07d400a6df4767a90d49a153fdb7f4cfa1e3f23e Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Tue, 6 Jan 2009 11:42:00 -0500 Subject: Btrfs: tree logging checksum fixes This patch contains following things. 1) Limit the max size of btrfs_ordered_sum structure to PAGE_SIZE. This struct is kmalloced so we want to keep it reasonable. 2) Replace copy_extent_csums by btrfs_lookup_csums_range. This was duplicated code in tree-log.c 3) Remove replay_one_csum. csum items are replayed at the same time as replaying file extents. This guarantees we only replay useful csums. 4) nbytes accounting fix. Signed-off-by: Yan Zheng --- fs/btrfs/extent-tree.c | 2 +- fs/btrfs/file-item.c | 62 ++++++----- fs/btrfs/inode.c | 5 +- fs/btrfs/tree-log.c | 293 +++++++++++++++---------------------------------- 4 files changed, 130 insertions(+), 232 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 171ca30a375..293da650873 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5579,7 +5579,7 @@ int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len) BUG_ON(ordered->file_offset != file_pos || ordered->len != len); disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt; - ret = btrfs_lookup_csums_range(root, disk_bytenr, + ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr, disk_bytenr + len - 1, &list); while (!list_empty(&list)) { diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c index b11abfad81a..964652435fd 100644 --- a/fs/btrfs/file-item.c +++ b/fs/btrfs/file-item.c @@ -27,6 +27,12 @@ #define MAX_CSUM_ITEMS(r, size) ((((BTRFS_LEAF_DATA_SIZE(r) - \ sizeof(struct btrfs_item) * 2) / \ size) - 1)) + +#define MAX_ORDERED_SUM_BYTES(r) ((PAGE_SIZE - \ + sizeof(struct btrfs_ordered_sum)) / \ + sizeof(struct btrfs_sector_sum) * \ + (r)->sectorsize - (r)->sectorsize) + int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 objectid, u64 pos, @@ -259,8 +265,7 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end, key.offset = start; key.type = BTRFS_EXTENT_CSUM_KEY; - ret = btrfs_search_slot(NULL, root->fs_info->csum_root, - &key, path, 0, 0); + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto fail; if (ret > 0 && path->slots[0] > 0) { @@ -279,7 +284,7 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end, while (start <= end) { leaf = path->nodes[0]; if (path->slots[0] >= btrfs_header_nritems(leaf)) { - ret = btrfs_next_leaf(root->fs_info->csum_root, path); + ret = btrfs_next_leaf(root, path); if (ret < 0) goto fail; if (ret > 0) @@ -306,33 +311,38 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end, continue; } - size = min(csum_end, end + 1) - start; - sums = kzalloc(btrfs_ordered_sum_size(root, size), GFP_NOFS); - BUG_ON(!sums); + csum_end = min(csum_end, end + 1); + item = btrfs_item_ptr(path->nodes[0], path->slots[0], + struct btrfs_csum_item); + while (start < csum_end) { + size = min_t(size_t, csum_end - start, + MAX_ORDERED_SUM_BYTES(root)); + sums = kzalloc(btrfs_ordered_sum_size(root, size), + GFP_NOFS); + BUG_ON(!sums); - sector_sum = sums->sums; - sums->bytenr = start; - sums->len = size; + sector_sum = sums->sums; + sums->bytenr = start; + sums->len = size; - offset = (start - key.offset) >> - root->fs_info->sb->s_blocksize_bits; - offset *= csum_size; + offset = (start - key.offset) >> + root->fs_info->sb->s_blocksize_bits; + offset *= csum_size; - item = btrfs_item_ptr(path->nodes[0], path->slots[0], - struct btrfs_csum_item); - while (size > 0) { - read_extent_buffer(path->nodes[0], §or_sum->sum, - ((unsigned long)item) + offset, - csum_size); - sector_sum->bytenr = start; - - size -= root->sectorsize; - start += root->sectorsize; - offset += csum_size; - sector_sum++; + while (size > 0) { + read_extent_buffer(path->nodes[0], + §or_sum->sum, + ((unsigned long)item) + + offset, csum_size); + sector_sum->bytenr = start; + + size -= root->sectorsize; + start += root->sectorsize; + offset += csum_size; + sector_sum++; + } + list_add_tail(&sums->list, list); } - list_add_tail(&sums->list, list); - path->slots[0]++; } ret = 0; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index c0ca9c3723c..4e57fe68e4b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -157,7 +157,6 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans, key.objectid = inode->i_ino; key.offset = start; btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); - inode_add_bytes(inode, size); datasize = btrfs_file_extent_calc_inline_size(cur_size); inode_add_bytes(inode, size); @@ -920,8 +919,8 @@ static noinline int csum_exist_in_range(struct btrfs_root *root, struct btrfs_ordered_sum *sums; LIST_HEAD(list); - ret = btrfs_lookup_csums_range(root, bytenr, bytenr + num_bytes - 1, - &list); + ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr, + bytenr + num_bytes - 1, &list); if (ret == 0 && list_empty(&list)) return 0; diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 3a72a1b6c24..332ec35d2c0 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -433,49 +433,6 @@ insert: trans->transid); } } - - if (overwrite_root && - key->type == BTRFS_EXTENT_DATA_KEY) { - int extent_type; - struct btrfs_file_extent_item *fi; - - fi = (struct btrfs_file_extent_item *)dst_ptr; - extent_type = btrfs_file_extent_type(path->nodes[0], fi); - if (extent_type == BTRFS_FILE_EXTENT_REG || - extent_type == BTRFS_FILE_EXTENT_PREALLOC) { - struct btrfs_key ins; - ins.objectid = btrfs_file_extent_disk_bytenr( - path->nodes[0], fi); - ins.offset = btrfs_file_extent_disk_num_bytes( - path->nodes[0], fi); - ins.type = BTRFS_EXTENT_ITEM_KEY; - - /* - * is this extent already allocated in the extent - * allocation tree? If so, just add a reference - */ - ret = btrfs_lookup_extent(root, ins.objectid, - ins.offset); - if (ret == 0) { - ret = btrfs_inc_extent_ref(trans, root, - ins.objectid, ins.offset, - path->nodes[0]->start, - root->root_key.objectid, - trans->transid, key->objectid); - } else { - /* - * insert the extent pointer in the extent - * allocation tree - */ - ret = btrfs_alloc_logged_extent(trans, root, - path->nodes[0]->start, - root->root_key.objectid, - trans->transid, key->objectid, - &ins); - BUG_ON(ret); - } - } - } no_copy: btrfs_mark_buffer_dirty(path->nodes[0]); btrfs_release_path(root, path); @@ -530,6 +487,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, u64 extent_end; u64 alloc_hint; u64 start = key->offset; + u64 saved_nbytes; struct btrfs_file_extent_item *item; struct inode *inode = NULL; unsigned long size; @@ -591,17 +549,95 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans, } btrfs_release_path(root, path); + saved_nbytes = inode_get_bytes(inode); /* drop any overlapping extents */ ret = btrfs_drop_extents(trans, root, inode, start, extent_end, start, &alloc_hint); BUG_ON(ret); - /* insert the extent */ - ret = overwrite_item(trans, root, path, eb, slot, key); - BUG_ON(ret); + if (found_type == BTRFS_FILE_EXTENT_REG || + found_type == BTRFS_FILE_EXTENT_PREALLOC) { + unsigned long dest_offset; + struct btrfs_key ins; + + ret = btrfs_insert_empty_item(trans, root, path, key, + sizeof(*item)); + BUG_ON(ret); + dest_offset = btrfs_item_ptr_offset(path->nodes[0], + path->slots[0]); + copy_extent_buffer(path->nodes[0], eb, dest_offset, + (unsigned long)item, sizeof(*item)); + + ins.objectid = btrfs_file_extent_disk_bytenr(eb, item); + ins.offset = btrfs_file_extent_disk_num_bytes(eb, item); + ins.type = BTRFS_EXTENT_ITEM_KEY; + + if (ins.objectid > 0) { + u64 csum_start; + u64 csum_end; + LIST_HEAD(ordered_sums); + /* + * is this extent already allocated in the extent + * allocation tree? If so, just add a reference + */ + ret = btrfs_lookup_extent(root, ins.objectid, + ins.offset); + if (ret == 0) { + ret = btrfs_inc_extent_ref(trans, root, + ins.objectid, ins.offset, + path->nodes[0]->start, + root->root_key.objectid, + trans->transid, key->objectid); + } else { + /* + * insert the extent pointer in the extent + * allocation tree + */ + ret = btrfs_alloc_logged_extent(trans, root, + path->nodes[0]->start, + root->root_key.objectid, + trans->transid, key->objectid, + &ins); + BUG_ON(ret); + } + btrfs_release_path(root, path); + + if (btrfs_file_extent_compression(eb, item)) { + csum_start = ins.objectid; + csum_end = csum_start + ins.offset; + } else { + csum_start = ins.objectid + + btrfs_file_extent_offset(eb, item); + csum_end = csum_start + + btrfs_file_extent_num_bytes(eb, item); + } + + ret = btrfs_lookup_csums_range(root->log_root, + csum_start, csum_end - 1, + &ordered_sums); + BUG_ON(ret); + while (!list_empty(&ordered_sums)) { + struct btrfs_ordered_sum *sums; + sums = list_entry(ordered_sums.next, + struct btrfs_ordered_sum, + list); + ret = btrfs_csum_file_blocks(trans, + root->fs_info->csum_root, + sums); + BUG_ON(ret); + list_del(&sums->list); + kfree(sums); + } + } else { + btrfs_release_path(root, path); + } + } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { + /* inline extents are easy, we just overwrite them */ + ret = overwrite_item(trans, root, path, eb, slot, key); + BUG_ON(ret); + } - /* btrfs_drop_extents changes i_bytes & i_blocks, update it here */ - inode_add_bytes(inode, extent_end - start); + inode_set_bytes(inode, saved_nbytes); btrfs_update_inode(trans, root, inode); out: if (inode) @@ -902,70 +938,6 @@ out_nowrite: return 0; } -/* - * replay one csum item from the log tree into the subvolume 'root' - * eb, slot and key all refer to the log tree - * path is for temp use by this function and should be released on return - * - * This copies the checksums out of the log tree and inserts them into - * the subvolume. Any existing checksums for this range in the file - * are overwritten, and new items are added where required. - * - * We keep this simple by reusing the btrfs_ordered_sum code from - * the data=ordered mode. This basically means making a copy - * of all the checksums in ram, which we have to do anyway for kmap - * rules. - * - * The copy is then sent down to btrfs_csum_file_blocks, which - * does all the hard work of finding existing items in the file - * or adding new ones. - */ -static noinline int replay_one_csum(struct btrfs_trans_handle *trans, - struct btrfs_root *root, - struct btrfs_path *path, - struct extent_buffer *eb, int slot, - struct btrfs_key *key) -{ - int ret; - u32 item_size = btrfs_item_size_nr(eb, slot); - u64 cur_offset; - u16 csum_size = - btrfs_super_csum_size(&root->fs_info->super_copy); - unsigned long file_bytes; - struct btrfs_ordered_sum *sums; - struct btrfs_sector_sum *sector_sum; - unsigned long ptr; - - file_bytes = (item_size / csum_size) * root->sectorsize; - sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS); - if (!sums) - return -ENOMEM; - - INIT_LIST_HEAD(&sums->list); - sums->len = file_bytes; - sums->bytenr = key->offset; - - /* - * copy all the sums into the ordered sum struct - */ - sector_sum = sums->sums; - cur_offset = key->offset; - ptr = btrfs_item_ptr_offset(eb, slot); - while (item_size > 0) { - sector_sum->bytenr = cur_offset; - read_extent_buffer(eb, §or_sum->sum, ptr, csum_size); - sector_sum++; - item_size -= csum_size; - ptr += csum_size; - cur_offset += root->sectorsize; - } - - /* let btrfs_csum_file_blocks add them into the file */ - ret = btrfs_csum_file_blocks(trans, root->fs_info->csum_root, sums); - BUG_ON(ret); - kfree(sums); - return 0; -} /* * There are a few corners where the link count of the file can't * be properly maintained during replay. So, instead of adding @@ -1659,10 +1631,6 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb, ret = replay_one_extent(wc->trans, root, path, eb, i, &key); BUG_ON(ret); - } else if (key.type == BTRFS_EXTENT_CSUM_KEY) { - ret = replay_one_csum(wc->trans, root, path, - eb, i, &key); - BUG_ON(ret); } else if (key.type == BTRFS_DIR_ITEM_KEY || key.type == BTRFS_DIR_INDEX_KEY) { ret = replay_one_dir_item(wc->trans, root, path, @@ -2021,7 +1989,7 @@ int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root) .process_func = process_one_buffer }; - if (!root->log_root) + if (!root->log_root || root->fs_info->log_root_recovering) return 0; log = root->log_root; @@ -2453,86 +2421,6 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans, return 0; } -static noinline int copy_extent_csums(struct btrfs_trans_handle *trans, - struct list_head *list, - struct btrfs_root *root, - u64 disk_bytenr, u64 len) -{ - struct btrfs_ordered_sum *sums; - struct btrfs_sector_sum *sector_sum; - int ret; - struct btrfs_path *path; - struct btrfs_csum_item *item = NULL; - u64 end = disk_bytenr + len; - u64 item_start_offset = 0; - u64 item_last_offset = 0; - u32 diff; - u32 sum; - u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy); - - sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS); - - sector_sum = sums->sums; - sums->bytenr = disk_bytenr; - sums->len = len; - list_add_tail(&sums->list, list); - - path = btrfs_alloc_path(); - while (disk_bytenr < end) { - if (!item || disk_bytenr < item_start_offset || - disk_bytenr >= item_last_offset) { - struct btrfs_key found_key; - u32 item_size; - - if (item) - btrfs_release_path(root, path); - item = btrfs_lookup_csum(NULL, root, path, - disk_bytenr, 0); - if (IS_ERR(item)) { - ret = PTR_ERR(item); - if (ret == -ENOENT || ret == -EFBIG) - ret = 0; - sum = 0; - printk(KERN_INFO "log no csum found for " - "byte %llu\n", - (unsigned long long)disk_bytenr); - item = NULL; - btrfs_release_path(root, path); - goto found; - } - btrfs_item_key_to_cpu(path->nodes[0], &found_key, - path->slots[0]); - - item_start_offset = found_key.offset; - item_size = btrfs_item_size_nr(path->nodes[0], - path->slots[0]); - item_last_offset = item_start_offset + - (item_size / csum_size) * - root->sectorsize; - item = btrfs_item_ptr(path->nodes[0], path->slots[0], - struct btrfs_csum_item); - } - /* - * this byte range must be able to fit inside - * a single leaf so it will also fit inside a u32 - */ - diff = disk_bytenr - item_start_offset; - diff = diff / root->sectorsize; - diff = diff * csum_size; - - read_extent_buffer(path->nodes[0], &sum, - ((unsigned long)item) + diff, - csum_size); -found: - sector_sum->bytenr = disk_bytenr; - sector_sum->sum = sum; - disk_bytenr += root->sectorsize; - sector_sum++; - } - btrfs_free_path(path); - return 0; -} - static noinline int copy_items(struct btrfs_trans_handle *trans, struct btrfs_root *log, struct btrfs_path *dst_path, @@ -2622,10 +2510,10 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, trans->transid, ins_keys[i].objectid); BUG_ON(ret); - ret = copy_extent_csums(trans, - &ordered_sums, - log->fs_info->csum_root, - ds + cs, cl); + ret = btrfs_lookup_csums_range( + log->fs_info->csum_root, + ds + cs, ds + cs + cl - 1, + &ordered_sums); BUG_ON(ret); } } @@ -2942,9 +2830,9 @@ again: tmp_key.offset = (u64)-1; wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key); - BUG_ON(!wc.replay_dest); + wc.replay_dest->log_root = log; btrfs_record_root_in_trans(wc.replay_dest); ret = walk_log_tree(trans, log, &wc); BUG_ON(ret); @@ -2961,6 +2849,7 @@ again: } key.offset = found_key.offset - 1; + wc.replay_dest->log_root = NULL; free_extent_buffer(log->node); kfree(log); -- cgit v1.2.3-70-g09d2 From cc7172defcf253335b16cf703fe4ac1ade15e1b1 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 6 Jan 2009 13:26:40 -0500 Subject: Btrfs: Don't use kmap_atomic(..., KM_IRQ0) during checksum verifies Checksum verification happens in a helper thread, and there is no need to mess with interrupts. This switches to kmap() instead. Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4e57fe68e4b..cdb701165a0 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1727,7 +1727,6 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, int ret; struct btrfs_root *root = BTRFS_I(inode)->root; u32 csum = ~(u32)0; - unsigned long flags; if (PageChecked(page)) { ClearPageChecked(page); @@ -1749,8 +1748,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, } else { ret = get_state_private(io_tree, start, &private); } - local_irq_save(flags); - kaddr = kmap_atomic(page, KM_IRQ0); + kaddr = kmap(page); if (ret) goto zeroit; @@ -1759,8 +1757,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, if (csum != private) goto zeroit; - kunmap_atomic(kaddr, KM_IRQ0); - local_irq_restore(flags); + kunmap(page); good: /* if the io failure tree for this inode is non-empty, * check to see if we've recovered from a failed IO @@ -1775,8 +1772,7 @@ zeroit: (unsigned long long)private); memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); - kunmap_atomic(kaddr, KM_IRQ0); - local_irq_restore(flags); + kunmap(page); if (private == 0) return 0; return -EIO; -- cgit v1.2.3-70-g09d2 From 9ab86c8e01c3f298dba0cbf2502c635b7f6fc6f9 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 7 Jan 2009 09:48:51 -0500 Subject: Btrfs: kmap_atomic(KM_USER0) is safe for btrfs_readpage_end_io_hook None of the checksum verification code schedules, so we can use the faster kmap_atomic Signed-off-by: Chris Mason --- fs/btrfs/inode.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index cdb701165a0..8adfe059ab4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1748,7 +1748,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, } else { ret = get_state_private(io_tree, start, &private); } - kaddr = kmap(page); + kaddr = kmap_atomic(page, KM_USER0); if (ret) goto zeroit; @@ -1757,7 +1757,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, if (csum != private) goto zeroit; - kunmap(page); + kunmap_atomic(kaddr, KM_USER0); good: /* if the io failure tree for this inode is non-empty, * check to see if we've recovered from a failed IO @@ -1772,7 +1772,7 @@ zeroit: (unsigned long long)private); memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page); - kunmap(page); + kunmap_atomic(kaddr, KM_USER0); if (private == 0) return 0; return -EIO; -- cgit v1.2.3-70-g09d2