diff options
| author | Liu Bo <bo.li.liu@oracle.com> | 2012-11-05 12:42:08 (GMT) |
|---|---|---|
| committer | Josef Bacik <jbacik@fusionio.com> | 2012-12-12 22:15:25 (GMT) |
| commit | 37c4146d2208ba7e4463e8dd95a1bf9e3d865280 (patch) | |
| tree | dcd62a0b6b30338be75cd2acfe5cbbcbfe726c3a /fs/btrfs | |
| parent | 6c1500f22a7be3a24ad3dffcdbf04be3f676521b (diff) | |
| download | linux-37c4146d2208ba7e4463e8dd95a1bf9e3d865280.tar.xz | |
Btrfs: fix a deadlock in aborting transaction due to ENOSPC
When committing a transaction, we may bail out of running delayed refs
due to ENOSPC, and then abort the current transaction to flip into readonly.
But we'll hit a deadlock on ref head's lock since we forget to release
its lock and other cleanup stuff.
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Diffstat (limited to 'fs/btrfs')
| -rw-r--r-- | fs/btrfs/extent-tree.c | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 0bcb954..f8a358a 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2297,6 +2297,9 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans, kfree(extent_op); if (ret) { + list_del_init(&locked_ref->cluster); + mutex_unlock(&locked_ref->mutex); + printk(KERN_DEBUG "btrfs: run_delayed_extent_op returned %d\n", ret); spin_lock(&delayed_refs->lock); return ret; @@ -2339,6 +2342,10 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans, count++; if (ret) { + if (locked_ref) { + list_del_init(&locked_ref->cluster); + mutex_unlock(&locked_ref->mutex); + } printk(KERN_DEBUG "btrfs: run_one_delayed_ref returned %d\n", ret); spin_lock(&delayed_refs->lock); return ret; |