From f3fe820c20a1a36c790545184e734e78d61cd68d Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Mon, 7 Jan 2013 17:03:21 -0500 Subject: Btrfs: add orphan before truncating pagecache Running xfstests 83 in a loop would sometimes fail the fsck. This happens because if we invalidate a page that already has an ordered extent setup for it we will complete the ordered extent ourselves, assuming that the truncate will clean everything up. The problem with this is there is plenty of time for the truncate to fail after we've done this work. So to fix this we need to add the orphan item first to make sure the cleanup gets done properly, and then we can truncate the pagecache and all that stuff and be safe. This fixes the btrfsck failures I was seeing while running 83 in a loop. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 53 ++++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 38 insertions(+), 15 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 67ed24ae86b..4ddcf79e789 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2478,6 +2478,18 @@ int btrfs_orphan_cleanup(struct btrfs_root *root) continue; } nr_truncate++; + + /* 1 for the orphan item deletion. */ + trans = btrfs_start_transaction(root, 1); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + goto out; + } + ret = btrfs_orphan_add(trans, inode); + btrfs_end_transaction(trans, root); + if (ret) + goto out; + ret = btrfs_truncate(inode); } else { nr_unlink++; @@ -3783,9 +3795,34 @@ static int btrfs_setsize(struct inode *inode, loff_t newsize) set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE, &BTRFS_I(inode)->runtime_flags); + /* + * 1 for the orphan item we're going to add + * 1 for the orphan item deletion. + */ + trans = btrfs_start_transaction(root, 2); + if (IS_ERR(trans)) + return PTR_ERR(trans); + + /* + * We need to do this in case we fail at _any_ point during the + * actual truncate. Once we do the truncate_setsize we could + * invalidate pages which forces any outstanding ordered io to + * be instantly completed which will give us extents that need + * to be truncated. If we fail to get an orphan inode down we + * could have left over extents that were never meant to live, + * so we need to garuntee from this point on that everything + * will be consistent. + */ + ret = btrfs_orphan_add(trans, inode); + btrfs_end_transaction(trans, root); + if (ret) + return ret; + /* we don't support swapfiles, so vmtruncate shouldn't fail */ truncate_setsize(inode, newsize); ret = btrfs_truncate(inode); + if (ret && inode->i_nlink) + btrfs_orphan_del(NULL, inode); } return ret; @@ -6929,11 +6966,9 @@ static int btrfs_truncate(struct inode *inode) /* * 1 for the truncate slack space - * 1 for the orphan item we're going to add - * 1 for the orphan item deletion * 1 for updating the inode. */ - trans = btrfs_start_transaction(root, 4); + trans = btrfs_start_transaction(root, 2); if (IS_ERR(trans)) { err = PTR_ERR(trans); goto out; @@ -6944,12 +6979,6 @@ static int btrfs_truncate(struct inode *inode) min_size); BUG_ON(ret); - ret = btrfs_orphan_add(trans, inode); - if (ret) { - btrfs_end_transaction(trans, root); - goto out; - } - /* * setattr is responsible for setting the ordered_data_close flag, * but that is only tested during the last file release. That @@ -7018,12 +7047,6 @@ static int btrfs_truncate(struct inode *inode) ret = btrfs_orphan_del(trans, inode); if (ret) err = ret; - } else if (ret && inode->i_nlink > 0) { - /* - * Failed to do the truncate, remove us from the in memory - * orphan list. - */ - ret = btrfs_orphan_del(NULL, inode); } if (trans) { -- cgit v1.2.3-70-g09d2 From f9e4fb53938de5db01950c9dfe479703b2f5c964 Mon Sep 17 00:00:00 2001 From: Liu Bo Date: Mon, 7 Jan 2013 10:10:12 +0000 Subject: Btrfs: fix a bug when llseek for delalloc bytes behind prealloc extents xfstests case 285 complains. It it because btrfs did not try to find unwritten delalloc bytes(only dirty pages, not yet writeback) behind prealloc extents, it ends up finding nothing while we're with SEEK_DATA. Signed-off-by: Liu Bo Signed-off-by: Josef Bacik --- fs/btrfs/file.c | 9 ++++++--- fs/btrfs/inode.c | 11 ++++++++--- 2 files changed, 14 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index fa48051484b..841cfe3be0e 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -2309,9 +2309,12 @@ static int find_desired_extent(struct inode *inode, loff_t *offset, int origin) } } - *offset = start; - free_extent_map(em); - break; + if (!test_bit(EXTENT_FLAG_PREALLOC, + &em->flags)) { + *offset = start; + free_extent_map(em); + break; + } } } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 4ddcf79e789..ac98384b174 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -5623,10 +5623,13 @@ struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *pag return em; if (em) { /* - * if our em maps to a hole, there might - * actually be delalloc bytes behind it + * if our em maps to + * - a hole or + * - a pre-alloc extent, + * there might actually be delalloc bytes behind it. */ - if (em->block_start != EXTENT_MAP_HOLE) + if (em->block_start != EXTENT_MAP_HOLE && + !test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) return em; else hole_em = em; @@ -5708,6 +5711,8 @@ struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *pag */ em->block_start = hole_em->block_start; em->block_len = hole_len; + if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags)) + set_bit(EXTENT_FLAG_PREALLOC, &em->flags); } else { em->start = range_start; em->len = found; -- cgit v1.2.3-70-g09d2 From f276795627045a3c599a60b476767861e4318c7d Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Tue, 8 Jan 2013 19:37:58 +0000 Subject: btrfs: fix btrfs_cont_expand() freeing IS_ERR em btrfs_cont_expand() tries to free an IS_ERR em as it gets an error from btrfs_get_extent() and breaks out of its loop. An instance of -EEXIST was reported in the wild: https://bugzilla.redhat.com/show_bug.cgi?id=874407 I have no idea if that -EEXIST is surprising, or not. Regardless, this error handling should be cleaned up to handle other reasonable errors (ENOMEM, EIO; whatever). This seemed to be the only buggy freeing of the relatively rare IS_ERR em so I opted to fix the caller rather than teach free_extent_map() to use IS_ERR_OR_NULL(). Signed-off-by: Zach Brown Reviewed-by: Eric Sandeen Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index ac98384b174..3d2c64d4734 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3677,6 +3677,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size) block_end - cur_offset, 0); if (IS_ERR(em)) { err = PTR_ERR(em); + em = NULL; break; } last_byte = min(extent_map_end(em), block_end); -- cgit v1.2.3-70-g09d2 From 3972f2603d8570effaf633cea52b12c7c2773c11 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Sat, 12 Jan 2013 02:57:22 +0000 Subject: btrfs: update timestamps on truncate() truncate() vs. ftruncate() differ in the VFS; truncate() doesn't set (ATTR_CTIME | ATTR_MTIME), and it's up to the fs to do the timestamp updates if the size changes. Signed-off-by: Eric Sandeen Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3d2c64d4734..9bc6c40b182 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -88,7 +88,7 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, }; -static int btrfs_setsize(struct inode *inode, loff_t newsize); +static int btrfs_setsize(struct inode *inode, struct iattr *attr); static int btrfs_truncate(struct inode *inode); static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent); static noinline int cow_file_range(struct inode *inode, @@ -3761,16 +3761,27 @@ next: return err; } -static int btrfs_setsize(struct inode *inode, loff_t newsize) +static int btrfs_setsize(struct inode *inode, struct iattr *attr) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_trans_handle *trans; loff_t oldsize = i_size_read(inode); + loff_t newsize = attr->ia_size; + int mask = attr->ia_valid; int ret; if (newsize == oldsize) return 0; + /* + * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a + * special case where we need to update the times despite not having + * these flags set. For all other operations the VFS set these flags + * explicitly if it wants a timestamp update. + */ + if (newsize != oldsize && (!(mask & (ATTR_CTIME | ATTR_MTIME)))) + inode->i_ctime = inode->i_mtime = current_fs_time(inode->i_sb); + if (newsize > oldsize) { truncate_pagecache(inode, oldsize, newsize); ret = btrfs_cont_expand(inode, oldsize, newsize); @@ -3843,7 +3854,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) return err; if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) { - err = btrfs_setsize(inode, attr->ia_size); + err = btrfs_setsize(inode, attr); if (err) return err; } -- cgit v1.2.3-70-g09d2 From 1eafa6c73791e4f312324ddad9cbcaf6a1b6052b Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 22 Jan 2013 10:49:00 +0000 Subject: Btrfs: fix repeated delalloc work allocation btrfs_start_delalloc_inodes() locks the delalloc_inodes list, fetches the first inode, unlocks the list, triggers btrfs_alloc_delalloc_work/ btrfs_queue_worker for this inode, and then it locks the list, checks the head of the list again. But because we don't delete the first inode that it deals with before, it will fetch the same inode. As a result, this function allocates a huge amount of btrfs_delalloc_work structures, and OOM happens. Fix this problem by splice this delalloc list. Reported-by: Alex Lyakas Signed-off-by: Miao Xie Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 55 +++++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 41 insertions(+), 14 deletions(-) (limited to 'fs/btrfs/inode.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 9bc6c40b182..ca7ace7b7b5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -7585,41 +7585,61 @@ void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work) */ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput) { - struct list_head *head = &root->fs_info->delalloc_inodes; struct btrfs_inode *binode; struct inode *inode; struct btrfs_delalloc_work *work, *next; struct list_head works; + struct list_head splice; int ret = 0; if (root->fs_info->sb->s_flags & MS_RDONLY) return -EROFS; INIT_LIST_HEAD(&works); - + INIT_LIST_HEAD(&splice); +again: spin_lock(&root->fs_info->delalloc_lock); - while (!list_empty(head)) { - binode = list_entry(head->next, struct btrfs_inode, + list_splice_init(&root->fs_info->delalloc_inodes, &splice); + while (!list_empty(&splice)) { + binode = list_entry(splice.next, struct btrfs_inode, delalloc_inodes); + + list_del_init(&binode->delalloc_inodes); + inode = igrab(&binode->vfs_inode); if (!inode) - list_del_init(&binode->delalloc_inodes); + continue; + + list_add_tail(&binode->delalloc_inodes, + &root->fs_info->delalloc_inodes); spin_unlock(&root->fs_info->delalloc_lock); - if (inode) { - work = btrfs_alloc_delalloc_work(inode, 0, delay_iput); - if (!work) { - ret = -ENOMEM; - goto out; - } - list_add_tail(&work->list, &works); - btrfs_queue_worker(&root->fs_info->flush_workers, - &work->work); + + work = btrfs_alloc_delalloc_work(inode, 0, delay_iput); + if (unlikely(!work)) { + ret = -ENOMEM; + goto out; } + list_add_tail(&work->list, &works); + btrfs_queue_worker(&root->fs_info->flush_workers, + &work->work); + cond_resched(); spin_lock(&root->fs_info->delalloc_lock); } spin_unlock(&root->fs_info->delalloc_lock); + list_for_each_entry_safe(work, next, &works, list) { + list_del_init(&work->list); + btrfs_wait_and_free_delalloc_work(work); + } + + spin_lock(&root->fs_info->delalloc_lock); + if (!list_empty(&root->fs_info->delalloc_inodes)) { + spin_unlock(&root->fs_info->delalloc_lock); + goto again; + } + spin_unlock(&root->fs_info->delalloc_lock); + /* the filemap_flush will queue IO into the worker threads, but * we have to make sure the IO is actually started and that * ordered extents get created before we return @@ -7632,11 +7652,18 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput) atomic_read(&root->fs_info->async_delalloc_pages) == 0)); } atomic_dec(&root->fs_info->async_submit_draining); + return 0; out: list_for_each_entry_safe(work, next, &works, list) { list_del_init(&work->list); btrfs_wait_and_free_delalloc_work(work); } + + if (!list_empty_careful(&splice)) { + spin_lock(&root->fs_info->delalloc_lock); + list_splice_tail(&splice, &root->fs_info->delalloc_inodes); + spin_unlock(&root->fs_info->delalloc_lock); + } return ret; } -- cgit v1.2.3-70-g09d2