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/ordered-data.h | 39 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) create mode 100644 fs/btrfs/ordered-data.h (limited to 'fs/btrfs/ordered-data.h') 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 -- 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/ordered-data.h') 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 4d5e74bc0aec3f54b7e429d77b7c35de042c507d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 16 Jan 2008 16:09:22 -0500 Subject: Btrfs: Fix data=ordered vs wait_on_inode deadlock on older kernels Using ilookup5 during data=ordered writeback could deadlock on I_LOCK. This saves a pointer to the inode instead. Signed-off-by: Chris Mason --- fs/btrfs/ordered-data.c | 12 ++++++++++-- fs/btrfs/ordered-data.h | 6 ++++-- fs/btrfs/transaction.c | 30 +++++++++++++----------------- 3 files changed, 27 insertions(+), 21 deletions(-) (limited to 'fs/btrfs/ordered-data.h') diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index cba2b623d02..3ee51e10c18 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -25,6 +25,7 @@ struct tree_entry { u64 root_objectid; u64 objectid; + struct inode *inode; struct rb_node rb_node; }; @@ -144,6 +145,7 @@ int btrfs_add_ordered_inode(struct inode *inode) write_lock(&tree->lock); entry->objectid = inode->i_ino; entry->root_objectid = root_objectid; + entry->inode = inode; node = tree_insert(&tree->tree, root_objectid, inode->i_ino, &entry->rb_node); @@ -159,7 +161,8 @@ int btrfs_add_ordered_inode(struct inode *inode) } int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, - u64 *root_objectid, u64 *objectid) + u64 *root_objectid, u64 *objectid, + struct inode **inode) { struct tree_entry *entry; struct rb_node *node; @@ -184,13 +187,16 @@ int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, } *root_objectid = entry->root_objectid; + *inode = entry->inode; + atomic_inc(&entry->inode->i_count); *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) + u64 *root_objectid, u64 *objectid, + struct inode **inode) { struct tree_entry *entry; struct rb_node *node; @@ -216,6 +222,8 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree, *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); diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 26b26212865..f25c6771ec6 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -33,8 +33,10 @@ btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t) 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); + 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); + u64 *root_objectid, u64 *objectid, + struct inode **inode); int btrfs_del_ordered_inode(struct inode *inode); #endif diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 08f7a188dc3..b6bbfc179c2 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -490,19 +490,17 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, while(1) { ret = btrfs_find_first_ordered_inode( &cur_trans->ordered_inode_tree, - &root_objectid, &objectid); + &root_objectid, &objectid, &inode); 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); - } + + 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); } @@ -511,19 +509,17 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, objectid = 0; ret = btrfs_find_del_first_ordered_inode( &cur_trans->ordered_inode_tree, - &root_objectid, &objectid); + &root_objectid, &objectid, &inode); 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); - atomic_dec(&inode->i_count); - iput(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); mutex_lock(&root->fs_info->trans_mutex); } -- 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/ordered-data.h') 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 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/ordered-data.h') 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 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/ordered-data.h') 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 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/ordered-data.h') 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 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/ordered-data.h') 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 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/ordered-data.h') 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 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/ordered-data.h') 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/ordered-data.h') 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 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/ordered-data.h') 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/ordered-data.h') 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 9ba4611a3a7902c6bad70c5c205de5161fcfc17b Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 23 Jul 2008 09:26:26 -0400 Subject: Btrfs: Fix 32 bit compiles by using an unsigned long byte count in the ordered extent The ordered extents have to fit in memory, so an unsigned long is sufficient. Signed-off-by: Chris Mason --- fs/btrfs/ordered-data.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/ordered-data.h') diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 36e63f1f79b..199cb0b4f1d 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -95,7 +95,8 @@ struct btrfs_ordered_extent { * 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) +static inline int btrfs_ordered_sum_size(struct btrfs_root *root, + unsigned long bytes) { unsigned long num_sectors = (bytes + root->sectorsize - 1) / root->sectorsize; -- cgit v1.2.3-70-g09d2 From 3eaa2885276fd6dac7b076a793932428b7168e74 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 24 Jul 2008 11:57:52 -0400 Subject: Btrfs: Fix the defragmention code and the block relocation code for data=ordered Before setting an extent to delalloc, the code needs to wait for pending ordered extents. Also, the relocation code needs to wait for ordered IO before scanning the block group again. This is because the extents are not removed until the IO for the new extents is finished Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 7 +++++++ fs/btrfs/disk-io.c | 3 +++ fs/btrfs/extent-tree.c | 39 ++++++++++++++++++++++++---------- fs/btrfs/ioctl.c | 21 ++++++++++++------- fs/btrfs/ordered-data.c | 56 +++++++++++++++++++++++++++++++++++++++++++++++-- fs/btrfs/ordered-data.h | 7 +++++++ 6 files changed, 113 insertions(+), 20 deletions(-) (limited to 'fs/btrfs/ordered-data.h') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8ecac2e77a4..6675e916ebc 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -538,6 +538,13 @@ struct btrfs_fs_info { struct list_head dead_roots; atomic_t nr_async_submits; + /* + * this is used by the balancing code to wait for all the pending + * ordered extents + */ + spinlock_t ordered_extent_lock; + struct list_head ordered_extents; + /* * there is a pool of worker threads for checksumming during writes * and a pool for checksumming after reads. This is because readers diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 7ce3f83c5dd..ec01062eb41 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1252,6 +1252,9 @@ struct btrfs_root *open_ctree(struct super_block *sb, 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); + spin_lock_init(&fs_info->ordered_extent_lock); + sb->s_blocksize = 4096; sb->s_blocksize_bits = blksize_bits(4096); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index febc6295c7a..f92b297e7da 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2640,6 +2640,7 @@ 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_ordered_extent *ordered; struct btrfs_trans_handle *trans; ra = kzalloc(sizeof(*ra), GFP_NOFS); @@ -2658,9 +2659,9 @@ 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) +again: + if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode)) goto truncate_racing; - page = grab_cache_page(inode->i_mapping, i); if (!page) { goto out_unlock; @@ -2674,18 +2675,24 @@ 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); + 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_start_ordered_extent(inode, ordered, 1); + btrfs_put_ordered_extent(ordered); + goto again; + } + set_page_extent_mapped(page); + + set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); set_page_dirty(page); @@ -2694,10 +2701,18 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start, unlock_page(page); page_cache_release(page); } - balance_dirty_pages_ratelimited_nr(inode->i_mapping, - total_read); out_unlock: + /* we have to start the IO in order to get the ordered extents + * instantiated. This allows the relocation to code to wait + * for all the ordered extents to hit the disk. + * + * Otherwise, it would constantly loop over the same extents + * because the old ones don't get deleted until the IO is + * started + */ + btrfs_fdatawrite_range(inode->i_mapping, start, start + len - 1, + WB_SYNC_NONE); kfree(ra); trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1); if (trans) { @@ -3238,6 +3253,8 @@ next: btrfs_clean_old_snapshots(tree_root); + btrfs_wait_ordered_extents(tree_root); + trans = btrfs_start_transaction(tree_root, 1); btrfs_commit_transaction(trans, tree_root); mutex_lock(&root->fs_info->alloc_mutex); diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 83f17a5cbd6..a61f2e7e2db 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -213,6 +213,7 @@ 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 btrfs_ordered_extent *ordered; struct page *page; unsigned long last_index; unsigned long ra_pages = root->fs_info->bdi.ra_pages; @@ -234,6 +235,7 @@ int btrfs_defrag_file(struct file *file) min(last_index, i + ra_pages - 1)); } total_read++; +again: page = grab_cache_page(inode->i_mapping, i); if (!page) goto out_unlock; @@ -247,18 +249,23 @@ int btrfs_defrag_file(struct file *file) } } -#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); + + 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_start_ordered_extent(inode, ordered, 1); + btrfs_put_ordered_extent(ordered); + goto again; + } + set_page_extent_mapped(page); + set_extent_delalloc(io_tree, page_start, page_end, GFP_NOFS); diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index e42fd233e04..676e4bd65c5 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -167,20 +167,28 @@ 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); INIT_LIST_HEAD(&entry->list); + INIT_LIST_HEAD(&entry->root_extent_list); 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); + printk("warning dup entry from add_ordered_extent\n"); + BUG(); } set_extent_ordered(&BTRFS_I(inode)->io_tree, file_offset, entry_end(entry) - 1, GFP_NOFS); + spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock); + list_add_tail(&entry->root_extent_list, + &BTRFS_I(inode)->root->fs_info->ordered_extents); + spin_unlock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock); + mutex_unlock(&tree->mutex); BUG_ON(node); return 0; @@ -285,11 +293,55 @@ int btrfs_remove_ordered_extent(struct inode *inode, rb_erase(node, &tree->tree); tree->last = NULL; set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags); + + spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock); + list_del_init(&entry->root_extent_list); + spin_unlock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock); + mutex_unlock(&tree->mutex); wake_up(&entry->wait); return 0; } +int btrfs_wait_ordered_extents(struct btrfs_root *root) +{ + struct list_head splice; + struct list_head *cur; + struct btrfs_ordered_extent *ordered; + struct inode *inode; + + INIT_LIST_HEAD(&splice); + + spin_lock(&root->fs_info->ordered_extent_lock); + list_splice_init(&root->fs_info->ordered_extents, &splice); + while(!list_empty(&splice)) { + cur = splice.next; + ordered = list_entry(cur, struct btrfs_ordered_extent, + root_extent_list); + 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 + */ + atomic_inc(&inode->i_count); + spin_unlock(&root->fs_info->ordered_extent_lock); + + btrfs_start_ordered_extent(inode, ordered, 1); + btrfs_put_ordered_extent(ordered); + iput(inode); + + spin_lock(&root->fs_info->ordered_extent_lock); + } + spin_unlock(&root->fs_info->ordered_extent_lock); + return 0; +} + /* * Used to start IO or wait for a given ordered extent to finish. * diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index 199cb0b4f1d..5efe6b63c74 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -80,6 +80,9 @@ struct btrfs_ordered_extent { /* reference count */ atomic_t refs; + /* the inode we belong to */ + struct inode *inode; + /* list of checksums for insertion when the extent io is done */ struct list_head list; @@ -88,6 +91,9 @@ struct btrfs_ordered_extent { /* our friendly rbtree entry */ struct rb_node rb_node; + + /* a per root list of all the pending ordered extents */ + struct list_head root_extent_list; }; @@ -137,4 +143,5 @@ 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); #endif -- 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/ordered-data.h') 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 cb843a6f513a1a91c54951005e60bd9b95bdf973 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Oct 2008 12:30:02 -0400 Subject: Btrfs: O_DIRECT writes via buffered writes + invaldiate This reworks the btrfs O_DIRECT write code a bit. It had always fallen back to buffered IO and done an invalidate, but needed to be updated for the data=ordered code. The invalidate wasn't actually removing pages because they were still inside an ordered extent. This also combines the O_DIRECT/O_SYNC paths where possible, and kicks off IO in the main btrfs_file_write loop to keep the pipe down the the disk full as we process long writes. Signed-off-by: Chris Mason --- fs/btrfs/file.c | 64 +++++++++++++++++++++++++++---------------------- fs/btrfs/ordered-data.c | 3 ++- fs/btrfs/ordered-data.h | 2 +- 3 files changed, 38 insertions(+), 31 deletions(-) (limited to 'fs/btrfs/ordered-data.h') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3088a118448..a03d1bbb19a 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -905,6 +905,10 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, struct page *pinned[2]; unsigned long first_index; unsigned long last_index; + int will_write; + + will_write = ((file->f_flags & O_SYNC) || IS_SYNC(inode) || + (file->f_flags & O_DIRECT)); nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE, PAGE_CACHE_SIZE / (sizeof(struct page *))); @@ -1001,15 +1005,24 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf, if (ret) goto out; + if (will_write) { + btrfs_fdatawrite_range(inode->i_mapping, pos, + pos + write_bytes - 1, + WB_SYNC_NONE); + } else { + 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); + } + buf += write_bytes; count -= write_bytes; pos += write_bytes; num_written += write_bytes; - 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: @@ -1023,36 +1036,29 @@ out_nolock: page_cache_release(pinned[1]); *ppos = pos; - if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) { + if (num_written > 0 && will_write) { 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) + err = btrfs_wait_ordered_range(inode, start_pos, num_written); + if (err) 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); + if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) { + 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); + } + } + if (file->f_flags & O_DIRECT) { + invalidate_mapping_pages(inode->i_mapping, + start_pos >> PAGE_CACHE_SHIFT, + (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); } - } 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/ordered-data.c b/fs/btrfs/ordered-data.c index dcc1730dd83..2eb6caba57c 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -397,7 +397,7 @@ void btrfs_start_ordered_extent(struct inode *inode, /* * Used to wait on ordered extents across a large range of bytes. */ -void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) +int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) { u64 end; u64 orig_end; @@ -451,6 +451,7 @@ again: (unsigned long long)orig_end); goto again; } + return 0; } /* diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h index fd45519f30a..f50f8870a14 100644 --- a/fs/btrfs/ordered-data.h +++ b/fs/btrfs/ordered-data.h @@ -135,7 +135,7 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, u64 file_offset); 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); +int 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_ordered_update_i_size(struct inode *inode, -- 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/ordered-data.h') 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 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/ordered-data.h') 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/ordered-data.h') 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 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/ordered-data.h') 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