[Cluster-devel] Re: [PATCH 00/25] move handling of setuid/gid bits from VFS into individual setattr functions (RESEND)

[Jeff Layton <jlayton@redhat.com>]

On Tue, 07 Aug 2007 20:45:34 -0400
Trond Myklebust <trond.myklebust@fys.uio.no> wrote:
> > - rename something so that unconverted filesystems will reliably fail to
> >   compile?
> > 
> > - leave existing filesystems alone, but add a new
> >   inode_operations.setattr_jeff, which the networked filesytems can
> >   implement, and teach core vfs to call setattr_jeff in preference to
> >   setattr?
> 
> If you really need to know that the filesystem is handling the flags,
> then how about instead having ->setattr() return something which
> indicates which flags it actually handled? That is likely to be a far
> more intrusive change, but it is one which is future-proof.
> 

One thing that we could do here is have notify_change check
attr->ia_valid after the setattr operation returns. If either ATTR_KILL_*
bit is set then BUG(). The helper function already clears those bits
so anything using it should automatically be ok. We'd have to fix
up NFS and a few others that don't implement suid/sgid.

This is not as certain as changing the name of the inode operation. It
would only pop when someone is attempting to change a setuid/setgid
file on these filesystems. Still, it should conceivably catch most if
not all offenders. Would that be sufficient to take care of everyone's
concerns?

-- 
Jeff Layton <jlayton@redhat.com>