]> git.ipfire.org Git - thirdparty/linux.git/commitdiff
btrfs: do not output error message if a qgroup has been already cleaned up
authorQu Wenruo <wqu@suse.com>
Sun, 19 Jan 2025 23:10:43 +0000 (09:40 +1030)
committerDavid Sterba <dsterba@suse.com>
Thu, 23 Jan 2025 21:34:11 +0000 (22:34 +0100)
[BUG]
There is a bug report that btrfs outputs the following error message:

  BTRFS info (device nvme0n1p2): qgroup scan completed (inconsistency flag cleared)
  BTRFS warning (device nvme0n1p2): failed to cleanup qgroup 0/1179: -2

[CAUSE]
The error itself is pretty harmless, and the end user should ignore it.

When a subvolume is fully dropped, btrfs will call
btrfs_qgroup_cleanup_dropped_subvolume() to delete the qgroup.

However if a qgroup rescan happened before a subvolume fully dropped,
qgroup for that subvolume will not be re-created, as rescan will only
create new qgroup if there is a BTRFS_ROOT_REF_KEY found.

But before we drop a subvolume, the subvolume is unlinked thus there is no
BTRFS_ROOT_REF_KEY.

In that case, btrfs_remove_qgroup() will fail with -ENOENT and trigger
the above error message.

[FIX]
Just ignore -ENOENT error from btrfs_remove_qgroup() inside
btrfs_qgroup_cleanup_dropped_subvolume().

Reported-by: John Shand <jshand2013@gmail.com>
Link: https://bugzilla.suse.com/show_bug.cgi?id=1236056
Fixes: 839d6ea4f86d ("btrfs: automatically remove the subvolume qgroup")
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/qgroup.c

index b90fabe302e6153e8729f9aa03d11981c7eb5d6b..aaf16019d829abefefbfa5169667b8b7295e8f94 100644 (file)
@@ -1897,8 +1897,11 @@ int btrfs_qgroup_cleanup_dropped_subvolume(struct btrfs_fs_info *fs_info, u64 su
        /*
         * It's squota and the subvolume still has numbers needed for future
         * accounting, in this case we can not delete it.  Just skip it.
+        *
+        * Or the qgroup is already removed by a qgroup rescan. For both cases we're
+        * safe to ignore them.
         */
-       if (ret == -EBUSY)
+       if (ret == -EBUSY || ret == -ENOENT)
                ret = 0;
        return ret;
 }