On Tue, Jul 30, 2019 at 12:03 PM Olga Kornievskaia
<olga.kornievskaia@xxxxxxxxx> wrote:
>
> On Tue, Jul 23, 2019 at 4:59 PM J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
> >
> > On Mon, Jul 22, 2019 at 04:24:08PM -0400, Olga Kornievskaia wrote:
> > > On Fri, Jul 19, 2019 at 6:01 PM J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
> > > >
> > > > On Mon, Jul 08, 2019 at 03:23:49PM -0400, Olga Kornievskaia wrote:
> > > > > Incoming stateid (used by a READ) could be a saved copy stateid.
> > > > > On first use make it active and check that the copy has started
> > > > > within the allowable lease time.
> > > > >
> > > > > Signed-off-by: Olga Kornievskaia <kolga@xxxxxxxxxx>
> > > > > ---
> > > > > fs/nfsd/nfs4state.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
> > > > > 1 file changed, 45 insertions(+)
> > > > >
> > > > > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > > > > index 2555eb9..b786625 100644
> > > > > --- a/fs/nfsd/nfs4state.c
> > > > > +++ b/fs/nfsd/nfs4state.c
> > > > > @@ -5232,6 +5232,49 @@ static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
> > > > >
> > > > > return 0;
> > > > > }
> > > > > +/*
> > > > > + * A READ from an inter server to server COPY will have a
> > > > > + * copy stateid. Return the parent nfs4_stid.
> > > > > + */
> > > > > +static __be32 _find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
> > > > > + struct nfs4_cpntf_state **cps)
> > > > > +{
> > > > > + struct nfs4_cpntf_state *state = NULL;
> > > > > +
> > > > > + if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
> > > > > + return nfserr_bad_stateid;
> > > > > + spin_lock(&nn->s2s_cp_lock);
> > > > > + state = idr_find(&nn->s2s_cp_stateids, st->si_opaque.so_id);
> > > > > + if (state)
> > > > > + refcount_inc(&state->cp_p_stid->sc_count);
> > > > > + spin_unlock(&nn->s2s_cp_lock);
> > > > > + if (!state)
> > > > > + return nfserr_bad_stateid;
> > > > > + *cps = state;
> > > > > + return 0;
> > > > > +}
> > > > > +
> > > > > +static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
> > > > > + struct nfs4_stid **stid)
> > > > > +{
> > > > > + __be32 status;
> > > > > + struct nfs4_cpntf_state *cps = NULL;
> > > > > +
> > > > > + status = _find_cpntf_state(nn, st, &cps);
> > > > > + if (status)
> > > > > + return status;
> > > > > +
> > > > > + /* Did the inter server to server copy start in time? */
> > > > > + if (cps->cp_active == false && !time_after(cps->cp_timeout, jiffies)) {
> > > > > + nfs4_put_stid(cps->cp_p_stid);
> > > > > + return nfserr_partner_no_auth;
> > > >
> > > > I wonder whether instead of checking the time we should instead be
> > > > destroying copy stateid's as they expire, so the fact that you were
> > > > still able to look up the stateid suggests that it's good. Or would
> > > > that result in returning the wrong error here? Just curious.
> > >
> > > In order to destroy copy stateid as they expire we need some thread
> > > monitoring the copies and then remove the expired one.
> >
> > It would be just another thing to do in the laundromat thread.
>
> This still seems simpler. You'd need to traverse the list and do more
> work? What's the advantage of laundry vs this? Given that laundry
> thread doesn't run all the time, there might still be a gap with it
> was last run and stateid expiring before the next run.
>
> >
> > So when do we free these things? The only free_cpntf_state() caller I
> > can find is in nfsd4_offload_cancel,
>
> There is a caller in the nfs4_put_stid. Copy notify state is freed
> when the associated stateid going away.
>
> > but I think the client only calls
> > those in case of interrupts or other unusual events. What about a copy
> > that terminates normally?
>
> At this point, are you asking about a copy state or a copy_notify
> state? When the copy is done, then the destination server will free
> the copy state. However, source server doesn't keep track of when the
> source server is done with the copy (I don't think we want to do that
> to store how much is read and state of the file seems like
> unnecessary).
>
> >
> > > That seems like
> > > a lot more work than what's currently there. The spec says that the
> > > use of the copy has to start without a certain timeout and that's what
> > > this is suppose to enforce. If the client took too long start the
> > > copy, it'll get an error. I don't think it matters what error code is
> > > returned BAD_STATEID or PARTNER_NO_AUTH both imply the stateid is bad.
> > >
> > > >
> > > > > + } else
> > > > > + cps->cp_active = true;
> > > > > +
> > > > > + *stid = cps->cp_p_stid;
> > > >
> > > > What guarantees that cp_p_stid still points to a valid stateid? (E.g.
> > > > if this is an open stateid that has since been closed.)
> > >
> > > A copy (or copy_notify) stateid takes a reference on the parent, thus
> > > we guaranteed that pointer is still a valid stateid.
> >
> > I only see a reference count taken when one is looked up, in
> > find_internal_cpntf_state. That's too late.
>
> Hm, right so this is tricky. With copy_notify, if I were to take a
> reference on the parent when copy_notify is processed, there is no way
> to free this reference because the source server never knows when the
> copy was done.
I'm having difficulty with this patch because there is no good way to
know when the copy_notify stateid can be freed. What I can propose is
to have the linux client send a FREE_STATEID with the copy_notify
stateid and use that as the trigger to free the state. In that case,
I'll keep a reference on the parent until the FREE_STATEID is
received.
This is not in the spec (though seems like a good idea to tell the
source server it's ok to clean up) so other implementations might not
choose this approach so we'll have problems with stateids sticking
around.
Thoughts?
>
>
>
> >
> > --b.
> >
> > >
> > > >
> > > > --b.
> > > >
> > > > > +
> > > > > + return nfs_ok;
> > > > > +}
> > > > >
> > > > > /*
> > > > > * Checks for stateid operations
> > > > > @@ -5264,6 +5307,8 @@ static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
> > > > > status = nfsd4_lookup_stateid(cstate, stateid,
> > > > > NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
> > > > > &s, nn);
> > > > > + if (status == nfserr_bad_stateid)
> > > > > + status = find_cpntf_state(nn, stateid, &s);
> > > > > if (status)
> > > > > return status;
> > > > > status = nfsd4_stid_check_stateid_generation(stateid, s,
> > > > > --
> > > > > 1.8.3.1
