On Mon, 2017-02-13 at 22:08 -0800, Christoph Hellwig wrote:
> On Tue, Feb 14, 2017 at 08:34:16AM +1100, Dave Chinner wrote:
> > >
> > > - if (pip && XFS_INHERIT_GID(pip)) {
> > > - ip->i_d.di_gid = pip->i_d.di_gid;
> > > - if ((VFS_I(pip)->i_mode & S_ISGID) &&
> > > S_ISDIR(mode))
> > > - inode->i_mode |= S_ISGID;
> > > - }
> > > -
> >
> > Doesn't this hunk break the "nogrpid" mount option?
>
> It does.
OK, so I'll fix up the s_user_mount problems and I'll let you sort out
the internals of removing the di_uid/gid if you wish to. I checked the
quota code and I think there are only a couple of places you're using
the kernel view of the ids where you should be using the filesystem
view. They're all identified by current_fsuid/fsgid(), so I think
(with the helper in patch 1) that this is the fix.
James
