From: Filipe Manana fdmanana@suse.com
commit ba0b084ac309283db6e329785c1dc4f45fdbd379 upstream.
We were checking for the full fsync flag in the inode before locking the inode, which is racy, since at that that time it might not be set but after we acquire the inode lock some other task set it. One case where this can happen is on a system low on memory and some concurrent task failed to allocate an extent map and therefore set the full sync flag on the inode, to force the next fsync to work in full mode.
A consequence of missing the full fsync flag set is hitting the problems fixed by commit 0c713cbab620 ("Btrfs: fix race between ranged fsync and writeback of adjacent ranges"), BUG_ON() when dropping extents from a log tree, hitting assertion failures at tree-log.c:copy_items() or all sorts of weird inconsistencies after replaying a log due to file extents items representing ranges that overlap.
So just move the check such that it's done after locking the inode and before starting writeback again.
Fixes: 0c713cbab620 ("Btrfs: fix race between ranged fsync and writeback of adjacent ranges") CC: stable@vger.kernel.org # 5.2+ Signed-off-by: Filipe Manana fdmanana@suse.com Signed-off-by: David Sterba dsterba@suse.com Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- fs/btrfs/file.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-)
--- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -2056,25 +2056,7 @@ int btrfs_sync_file(struct file *file, l struct btrfs_trans_handle *trans; struct btrfs_log_ctx ctx; int ret = 0, err; - u64 len;
- /* - * If the inode needs a full sync, make sure we use a full range to - * avoid log tree corruption, due to hole detection racing with ordered - * extent completion for adjacent ranges, and assertion failures during - * hole detection. - */ - if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, - &BTRFS_I(inode)->runtime_flags)) { - start = 0; - end = LLONG_MAX; - } - - /* - * The range length can be represented by u64, we have to do the typecasts - * to avoid signed overflow if it's [0, LLONG_MAX] eg. from fsync() - */ - len = (u64)end - (u64)start + 1; trace_btrfs_sync_file(file, datasync);
btrfs_init_log_ctx(&ctx, inode); @@ -2101,6 +2083,19 @@ int btrfs_sync_file(struct file *file, l atomic_inc(&root->log_batch);
/* + * If the inode needs a full sync, make sure we use a full range to + * avoid log tree corruption, due to hole detection racing with ordered + * extent completion for adjacent ranges, and assertion failures during + * hole detection. Do this while holding the inode lock, to avoid races + * with other tasks. + */ + if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, + &BTRFS_I(inode)->runtime_flags)) { + start = 0; + end = LLONG_MAX; + } + + /* * Before we acquired the inode's lock, someone may have dirtied more * pages in the target range. We need to make sure that writeback for * any such pages does not start while we are logging the inode, because @@ -2127,8 +2122,11 @@ int btrfs_sync_file(struct file *file, l /* * We have to do this here to avoid the priority inversion of waiting on * IO of a lower priority task while holding a transaciton open. + * + * Also, the range length can be represented by u64, we have to do the + * typecasts to avoid signed overflow if it's [0, LLONG_MAX]. */ - ret = btrfs_wait_ordered_range(inode, start, len); + ret = btrfs_wait_ordered_range(inode, start, (u64)end - (u64)start + 1); if (ret) { up_write(&BTRFS_I(inode)->dio_sem); inode_unlock(inode);