[PATCH 07/27] xfs: always log timestamp updates in xfs_setattr_size

[Dave Chinner]

On Wed, Jun 29, 2011 at 05:13:23PM -0500, Alex Elder wrote:
> On Wed, 2011-06-29 at 10:01 -0400, Christoph Hellwig wrote:
> > plain text document attachment (xfs-simplify-setattr)
> > Get rid of the special case where we use unlogged timestamp updates for
> > a truncate to the current inode size, and just call xfs_setattr_nonsize
> > for it to treat it like a utimes calls.
> 
> I'm probably just missing something, but I have a
> question below.  If it does in fact call out a
> problem, it will be easily fixed.
> 
> Reviewed-by: Alex Elder <aelder at sgi.com>
> 
> > Signed-off-by: Christoph Hellwig <hch at lst.de>
> > Reviewed-by: Dave Chinner <dchinner at redhat.com>
> > 
> > Index: xfs/fs/xfs/linux-2.6/xfs_iops.c
> > ===================================================================
> > --- xfs.orig/fs/xfs/linux-2.6/xfs_iops.c	2011-06-21 11:12:11.008794493 +0200
> > +++ xfs/fs/xfs/linux-2.6/xfs_iops.c	2011-06-21 11:12:22.398793917 +0200
> > @@ -773,14 +773,16 @@ xfs_setattr_size(
> >  	 */
> >  	if (iattr->ia_size == 0 &&
> >  	    ip->i_size == 0 && ip->i_d.di_nextents == 0) {
> > -		xfs_iunlock(ip, XFS_ILOCK_EXCL);
> > -		lock_flags &= ~XFS_ILOCK_EXCL;
> > -		if (mask & ATTR_CTIME) {
> > -			inode->i_mtime = inode->i_ctime =
> > -					current_fs_time(inode->i_sb);
> > -			xfs_mark_inode_dirty_sync(ip);
> > -		}
> > -		goto out_unlock;
> > +		if (!(mask & ATTR_CTIME))
> > +			goto out_unlock;
> > +
> > +		/*
> > +		 * Use the regular setattr path to update the timestamps.
> > +		 */
> > +		xfs_iunlock(ip, lock_flags);
> > +		iattr->ia_valid &= ~ATTR_SIZE;
> > +		iattr->ia_valid |= ATTR_MTIME;
> 
> How/where does iattr->ia_mtime get initialized to the current
> value (I believe it would just match iattr->ia_ctime)?

notify_change() always sets both the ia_ctime and ia_mtime to the
current time unless the caller has specified ATTR_MTIME_SET, in
which case they've set a specific ia_mtime themselves. IOWs, I'm
pretty sure we can rely on both ia_mtime and ia_ctime being correct
through the ATTR_SIZE truncate path without having to touch it
here...

Cheers,

Dave.
-- 
Dave Chinner
david at fromorbit.com