xfs
[Top] [All Lists]

Re: [PATCH 03/17] xfs: cleanup handling in xfs_swap_extents

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 03/17] xfs: cleanup handling in xfs_swap_extents
From: Eric Sandeen <sandeen@xxxxxxxxxxx>
Date: Thu, 07 May 2009 19:45:36 -0500
Cc: xfs@xxxxxxxxxxx, Alexander Beregalov <a.beregalov@xxxxxxxxx>
In-reply-to: <20090126073200.459094000@xxxxxxxxxxxxxxxxxxxxxx>
References: <20090126073136.384490000@xxxxxxxxxxxxxxxxxxxxxx> <20090126073200.459094000@xxxxxxxxxxxxxxxxxxxxxx>
User-agent: Thunderbird 2.0.0.21 (Macintosh/20090302)
Christoph Hellwig wrote:

> Use multiple lables for proper error unwinding and get rid of some now
> superflous variables.
>
>
> Signed-off-by: Josef 'Jeff' Sipek <jeffpc@xxxxxxxxxxxxxx>
> Reviewed-by: Christoph Hellwig <hch@xxxxxx>

Problem in this patch, I think, getting hangs on x86 fsr...

> Index: xfs/fs/xfs/xfs_dfrag.c
> ===================================================================
> --- xfs.orig/fs/xfs/xfs_dfrag.c       2008-12-19 15:02:54.003908425 +0100
> +++ xfs/fs/xfs/xfs_dfrag.c    2008-12-22 15:59:55.013247371 +0100
<snip>
> @@ -352,19 +344,19 @@ xfs_swap_extents(
>        * If this is a synchronous mount, make sure that the
>        * transaction goes to disk before returning to the user.
>        */
> -     if (mp->m_flags & XFS_MOUNT_WSYNC) {
> +     if (mp->m_flags & XFS_MOUNT_WSYNC)
>               xfs_trans_set_sync(tp);
> -     }
>  
>       error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
> -     locked = 0;

old code said "unlocked" here thanks to the trans commit ...

> - error0:
> -     if (locked) {
> -             xfs_iunlock(ip,  lock_flags);
> -             xfs_iunlock(tip, lock_flags);
> -     }
and so we wouldn't unlock again ...
> -     if (tempifp != NULL)
> -             kmem_free(tempifp);
> +out_unlock:
> +     xfs_iunlock(ip,  XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
> +     xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);

But now we do it unconditionally, and ruh-roh.

> +out:
> +     kmem_free(tempifp);
>       return error;
> +
> +out_trans_cancel:
> +     xfs_trans_cancel(tp, 0);
> +     goto out_unlock;
>  }

Is this too ugly a fix?

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;


<Prev in Thread] Current Thread [Next in Thread>
  • Re: [PATCH 03/17] xfs: cleanup handling in xfs_swap_extents, Eric Sandeen <=