Re: [PATCH 6/6] per-mount noatime and nodiratime

[Christoph Hellwig <hch@lst.de>]

On Tue, Dec 13, 2005 at 02:36:38PM -0800, Andrew Morton wrote:
> Where'd this hunk come from?
> 
> > Index: linux-2.6.15-rc5/fs/super.c
> > ===================================================================
> > --- linux-2.6.15-rc5.orig/fs/super.c	2005-12-13 11:27:14.000000000 +0100
> > +++ linux-2.6.15-rc5/fs/super.c	2005-12-13 12:06:00.000000000 +0100
> > @@ -830,9 +830,9 @@
> >  	mnt->mnt_parent = mnt;
> >  
> >  	if (type->fs_flags & FS_NOATIME)
> > -		sb->s_flags |= MS_NOATIME;
> > +		mnt->mnt_flags |= MNT_NOATIME;
> >  	if (type->fs_flags & FS_NODIRATIME)
> > -		sb->s_flags |= MS_NODIRATIME;
> > +		mnt->mnt_flags |= MNT_NODIRATIME;
> >  
> >  	up_write(&sb->s_umount);
> >  	free_secdata(secdata);
> 
> I just dropped it, but it's a worry...

This is totally intentional.  The FS_* flags were introduced in the
previous patch to mark filesystems that always are NOATIME/NODIRATIME,
and with this patch we need to set them in the vfsmount instead of the
superblock.