Felix Blyakher wrote:
> I've been chasing this as well, and I found the double xfs iolock
> unlock which is causing xfs_fsr deadlock itself. I have a fix tested.
> The patch is coming.
>
> Felix
>
Oh :) I just sent this one to the list ... :)
XFS: Fix double unlock of inodes in xfs_swap_extents()
commit ef8f7fc549bf345d92f396f5aa7b152b4969cbf7 had an error
where we would try to re-unlock the inodes after they had been
committed in the transaction; this double unlock caused a
=====================================
[ BUG: bad unlock balance detected! ]
-------------------------------------
xfs_fsr/1459 is trying to release lock (&(&ip->i_iolock)->mr_lock) at:
[<e248dedb>] xfs_iunlock+0x2c/0x92 [xfs]
but there are no more locks to release!
Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxxx>
---
Index: linux-2.6/fs/xfs/xfs_dfrag.c
===================================================================
--- linux-2.6.orig/fs/xfs/xfs_dfrag.c
+++ linux-2.6/fs/xfs/xfs_dfrag.c
@@ -347,13 +347,15 @@ xfs_swap_extents(
error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
-out_unlock:
- xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
- xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
out:
kmem_free(tempifp);
return error;
+out_unlock:
+ xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
+ xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
+ goto out;
+
out_trans_cancel:
xfs_trans_cancel(tp, 0);
goto out_unlock;
|