Re: [PATCH] xfs: remove unneeded ASSERT from xfs_itruncate_extents

[Dave Chinner <david@fromorbit.com>]

On Mon, Jan 28, 2013 at 09:04:30AM -0500, Carlos Maiolino wrote:
> There is no reason to ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); twice, so,
> remove one of these ASSERT calls

Second assert is for the IOLOCK, not the ILOCK....

> Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> ---
>  fs/xfs/xfs_inode.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> index 66282dc..25226ea 100644
> --- a/fs/xfs/xfs_inode.c
> +++ b/fs/xfs/xfs_inode.c
> @@ -1396,8 +1396,7 @@ xfs_itruncate_extents(
>  	int			done = 0;
>  
>  	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
> -	ASSERT(!atomic_read(&VFS_I(ip)->i_count) ||
> -	       xfs_isilocked(ip, XFS_IOLOCK_EXCL));
> +	ASSERT(!atomic_read(&VFS_I(ip)->i_count));

The code is correct. The ASSERT is testing the locking constraints on
the XFS_IOLOCK_EXCL. That is, if xfs_itruncate_extents() is called
in the VFS inode reclaim path (i.e. via xfs_inactive()), the IO lock
is not used (throws lockdep warnings). Hence the ASSERT is checking
that if we hold an inode reference, we are also holding the IO lock.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs