Re: CLOSE/OPEN race

[Jeff Layton <jlayton@redhat.com>]

On Sun, 2016-11-13 at 08:34 -0500, Benjamin Coddington wrote:
> 
> On 12 Nov 2016, at 21:56, Jeff Layton wrote:
> 
> > On Sat, 2016-11-12 at 16:16 -0500, Jeff Layton wrote:
> > > On Sat, 2016-11-12 at 13:03 -0500, Benjamin Coddington wrote:
> > > > 
> > > > On 12 Nov 2016, at 11:52, Jeff Layton wrote:
> > > > 
> > > > > 
> > > > > On Sat, 2016-11-12 at 10:31 -0500, Benjamin Coddington wrote:
> > > > > > 
> > > > > > On 12 Nov 2016, at 7:54, Jeff Layton wrote:
> > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > > > On Sat, 2016-11-12 at 06:08 -0500, Benjamin Coddington wrote:
> > > > > > > > 
> > > > > > > > 
> > > > > > > > I've been seeing the following on a modified version of 
> > > > > > > > generic/089
> > > > > > > > that gets the client stuck sending LOCK with 
> > > > > > > > NFS4ERR_OLD_STATEID.
> > > > > > > > 
> > > > > > > > 1. Client has open stateid A, sends a CLOSE
> > > > > > > > 2. Client sends OPEN with same owner
> > > > > > > > 3. Client sends another OPEN with same owner
> > > > > > > > 4. Client gets a reply to OPEN in 3, stateid is B.2 (stateid B
> > > > > > > > sequence 2)
> > > > > > > > 5. Client does LOCK,LOCKU,FREE_STATEID from B.2
> > > > > > > > 6. Client gets a reply to CLOSE in 1
> > > > > > > > 7. Client gets reply to OPEN in 2, stateid is B.1
> > > > > > > > 8. Client sends LOCK with B.1 - OLD_STATEID, now stuck in a loop
> > > > > > > > 
> > > > > > > > The CLOSE response in 6 causes us to clear NFS_OPEN_STATE, so 
> > > > > > > > that
> > > > > > > > the OPEN
> > > > > > > > response in 7 is able to update the open_stateid even though it 
> > > > > > > > has a
> > > > > > > > lower
> > > > > > > > sequence number.
> > > > > > > > 
> > > > > > > > I think this case could be handled by never updating the 
> > > > > > > > open_stateid
> > > > > > > > if the
> > > > > > > > stateids match but the sequence number of the new state is less 
> > > > > > > > than
> > > > > > > > the
> > > > > > > > current open_state.
> > > > > > > > 
> > > > > > > 
> > > > > > > What kernel is this on?
> > > > > > 
> > > > > > On v4.9-rc2 with a couple fixups.  Without them, I can't test long
> > > > > > enough to
> > > > > > reproduce this race.  I don't think any of those are involved in 
> > > > > > this
> > > > > > problem, though.
> > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > > > Yes, that seems wrong. The client should be picking B.2 for the 
> > > > > > > open
> > > > > > > stateid to use. I think that decision of whether to take a seqid 
> > > > > > > is
> > > > > > > made
> > > > > > > in nfs_need_update_open_stateid. The logic in there looks 
> > > > > > > correct to
> > > > > > > me
> > > > > > > at first glance though.
> > > > > > 
> > > > > > nfs_need_update_open_stateid() will return true if NFS_OPEN_STATE 
> > > > > > is
> > > > > > unset.
> > > > > > That's the precondition set up by steps 1-6.  Perhaps it should 
> > > > > > not
> > > > > > update
> > > > > > the stateid if they match but the sequence number is less, and 
> > > > > > still set
> > > > > > NFS_OPEN_STATE once more.  That will fix _this_ case.  Are there 
> > > > > > other
> > > > > > cases
> > > > > > where that would be a problem?
> > > > > > 
> > > > > > Ben
> > > > > 
> > > > > That seems wrong.
> > > > 
> > > > I'm not sure what you mean: what seems wrong?
> > > > 
> > > 
> > > Sorry, it seems wrong that the client would issue the LOCK with B.1
> > > there.
> > > 
> > > > > 
> > > > > The only close was sent in step 1, and that was for a
> > > > > completely different stateid (A rather than B). It seems likely 
> > > > > that
> > > > > that is where the bug is.
> > > > 
> > > > I'm still not sure what point you're trying to make..
> > > > 
> > > > Even though the close was sent in step 1, the response wasn't 
> > > > processed
> > > > until step 6..
> > > 
> > > Not really a point per-se, I was just saying where I think the bug 
> > > might
> > > be...
> > > 
> > > When you issue a CLOSE, you issue it vs. a particular stateid 
> > > (stateid
> > > "A" in this case). Once the open stateid has been superseded by "B", 
> > > the
> > > closing of "A" should have no effect.
> > > 
> > > Perhaps nfs_clear_open_stateid needs to check and see whether the 
> > > open
> > > stateid has been superseded before doing its thing?
> > > 
> > 
> > Ok, I see something that might be a problem in this call in
> > nfs4_close_done:
> > 
> >        nfs_clear_open_stateid(state, &calldata->arg.stateid,
> >                         res_stateid, 
> > calldata->arg.fmode);
> > 
> > Note that we pass two nfs4_stateids to this call. The first is the
> > stateid that got sent in the CLOSE call, and the second is the stateid
> > that came back in the CLOSE response.
> > 
> > RFC5661 and RFC7530 both indicate that the stateid in a CLOSE response
> > should be ignored.
> > 
> > So, I think a patch like this may be in order. As to whether it will
> > fix this bug, I sort of doubt it, but it might not hurt to test it 
> > out?
> 
> Doesn't this undo the fix in 4a1e2feb9d24 ("NFSv4.1: Fix a protocol 
> issue
> with CLOSE stateids")?
> 

I don't think so. The (updated) specs are pretty clear that the stateid
in a CLOSE response is not to be trusted in any way. I think we're
better off just relying on the stateid that was sent in the request.

On a related note, we probably need to fix knfsd not to send anything 
in there as well, but we probably need to think about older clients
there as well.

Thanks for testing the patch. I'm not too surprised this didn't help
here though. It didn't really change the logic appreciably, unless you
have a server sending weird stateids in the CLOSE response.

> I don't think it will help with the race either, since I think the issue 
> is
> that the race above unsets NFS_OPEN_STATE, then that is reason enough to
> overwrite the existing open_state, even if it is going to overwrite with 
> an
> older stateid.  What close does with its stateid is unimportant to that
> problem.
> 
> In the interest of science, I'll kick off a test with this patch.  It 
> should
> take a few hours to reproduce, but I'm also fairly busy today, so I'll
> report back this evening most likely.   Thanks for the patch.
> 
> Ben
> 
> > ----------------------8<--------------------------
> > 
> > [RFC PATCH] nfs: properly ignore the stateid in a CLOSE response
> > 
> > Signed-off-by: Jeff Layton
> > ---
> >  fs/nfs/nfs4proc.c | 14 +++-----------
> >  1 file changed, 3 insertions(+), 11 deletions(-)
> > 
> > diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
> > index 7897826d7c51..58413bd0aae2 100644
> > --- a/fs/nfs/nfs4proc.c
> > +++ b/fs/nfs/nfs4proc.c
> > @@ -1451,7 +1451,6 @@ static void 
> > nfs_resync_open_stateid_locked(struct nfs4_state *state)
> >  }
> > 
> >  static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
> > > > -		nfs4_stateid *arg_stateid,
> > > >  		nfs4_stateid *stateid, fmode_t fmode)
> >  {
> > > >  	clear_bit(NFS_O_RDWR_STATE, &state->flags);
> > @@ -1467,12 +1466,8 @@ static void 
> > nfs_clear_open_stateid_locked(struct nfs4_state *state,
> > > >  		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
> > > >  		clear_bit(NFS_OPEN_STATE, &state->flags);
> > > >  	}
> > > > -	if (stateid == NULL)
> > > > -		return;
> > > >  	/* Handle races with OPEN */
> > > > -	if (!nfs4_stateid_match_other(arg_stateid, &state->open_stateid) ||
> > > > -	    (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
> > > > -	    !nfs4_stateid_is_newer(stateid, &state->open_stateid))) {
> > > > +	if (!nfs4_stateid_match_other(stateid, &state->open_stateid)) {
> > > >  		nfs_resync_open_stateid_locked(state);
> > > >  		return;
> > > >  	}
> > @@ -1482,11 +1477,10 @@ static void 
> > nfs_clear_open_stateid_locked(struct nfs4_state *state,
> >  }
> > 
> >  static void nfs_clear_open_stateid(struct nfs4_state *state,
> > > > -	nfs4_stateid *arg_stateid,
> > > >  	nfs4_stateid *stateid, fmode_t fmode)
> >  {
> > > >  	write_seqlock(&state->seqlock);
> > > > -	nfs_clear_open_stateid_locked(state, arg_stateid, stateid, fmode);
> > > > +	nfs_clear_open_stateid_locked(state, stateid, fmode);
> > > >  	write_sequnlock(&state->seqlock);
> > > >  	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
> > > >  		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
> > @@ -3042,7 +3036,6 @@ static void nfs4_close_done(struct rpc_task 
> > *task, void *data)
> > > >  	struct nfs4_closedata *calldata = data;
> > > >  	struct nfs4_state *state = calldata->state;
> > > >  	struct nfs_server *server = NFS_SERVER(calldata->inode);
> > > > -	nfs4_stateid *res_stateid = NULL;
> > 
> > > >  	dprintk("%s: begin!\n", __func__);
> > > >  	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
> > @@ -3053,7 +3046,6 @@ static void nfs4_close_done(struct rpc_task 
> > *task, void *data)
> > > >  	 */
> > > >  	switch (task->tk_status) {
> > > >  		case 0:
> > > > -			res_stateid = &calldata->res.stateid;
> > > >  			if (calldata->roc)
> > > >  				pnfs_roc_set_barrier(state->inode,
> > > >  						     calldata->roc_barrier);
> > @@ -3081,7 +3073,7 @@ static void nfs4_close_done(struct rpc_task 
> > *task, void *data)
> > > >  			}
> > > >  	}
> > > >  	nfs_clear_open_stateid(state, &calldata->arg.stateid,
> > > > -			res_stateid, calldata->arg.fmode);
> > > > +				calldata->arg.fmode);
> >  out_release:
> > > >  	nfs_release_seqid(calldata->arg.seqid);
> > > >  	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
> > -- 
> > 2.9.3
> > 

-- 
Jeff Layton <jlayton@redhat.com>