On Fri, Sep 20, 2024 at 12:34 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
>
> On Fri, 2024-09-20 at 16:30 +0000, Chuck Lever III wrote:
> >
> > > On Sep 20, 2024, at 12:24 PM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > >
> > > On Fri, 2024-09-20 at 12:05 -0400, Olga Kornievskaia wrote:
> > > > Current, the server returns that it supports NFS4_SHARE_WANT_DELEG_TIMESTAMPS
> > > > but when the client sends that on the open, knfsd returns back with
> > > > bad_xdr error.
> > > >
> > > > Fixes: d0eab73d48a0 ("nfsd: add support for delegated timestamps")
> > > > Signed-off-by: Olga Kornievskaia <okorniev@xxxxxxxxxx>
> > > > ---
> > > > fs/nfsd/nfs4xdr.c | 1 +
> > > > 1 file changed, 1 insertion(+)
> > > >
> > > > diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> > > > index c0bad580ab6d..adda8b489175 100644
> > > > --- a/fs/nfsd/nfs4xdr.c
> > > > +++ b/fs/nfsd/nfs4xdr.c
> > > > @@ -1109,6 +1109,7 @@ static __be32 nfsd4_decode_share_access(struct nfsd4_compoundargs *argp, u32 *sh
> > > > case NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED:
> > > > case (NFS4_SHARE_SIGNAL_DELEG_WHEN_RESRC_AVAIL |
> > > > NFS4_SHARE_PUSH_DELEG_WHEN_UNCONTENDED):
> > > > + case NFS4_SHARE_WANT_DELEG_TIMESTAMPS:
> > > > return nfs_ok;
> > > > }
> > > > return nfserr_bad_xdr;
> > >
> > > Ouch.
> > >
> > > The problem here is that we had to drop the patch that added
> > > OPEN_XOR_DELEGATION support. That patch reworked the flag handling in
> > > this function in a way that allowed the new delstid flags to be
> > > properly supported.
> > >
> > > I think we probably want to resurrect the parts of this patch that
> > > alter nfsd4_decode_share_access:
> > >
> > > https://lore.kernel.org/linux-nfs/20240905-delstid-v4-8-d3e5fd34d107@xxxxxxxxxx/
> > >
> > > Olga, would you be OK with that approach instead?
> >
> > At this point, I'd like to consider postponing the delstid patches
> > until v6.13. They haven't gotten enough testing in their current
> > form...
> >
>
> It's your call, but that seems like an extreme reaction to a flag
> handling fix, given that we have several weeks of -rc's ahead of us.
I'm OK with whatever fix/approach it's going to go (Jeff if you have a
patch to test that would go on top of what's in Chuck's nfsd-next and
fixes the issue do post it) . I'm not ok with the code that's at the
moment in nfsd-next :).
> --
> Jeff Layton <jlayton@xxxxxxxxxx>
>
