Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0
From:   "NeilBrown" <neilb@suse.de>
To:     "J. Bruce Fields" <bfields@fieldses.org>
Cc:     "Chuck Lever" <chuck.lever@oracle.com>, linux-nfs@vger.kernel.org
Subject: Re: [PATCH 00/14] SUNRPC: clean up server thread management.
In-reply-to: <20211122151823.GA22802@fieldses.org>
References: <163710954700.5485.5622638225352156964.stgit@noble.brown>,
 <20211117141231.GA24762@fieldses.org>,
 <163753863448.13692.4142092237119935826@noble.neil.brown.name>,
 <20211122005639.GA12035@fieldses.org>, <20211122005901.GB12035@fieldses.org>,
 <163754358887.13692.5665882865660886756@noble.neil.brown.name>,
 <20211122023753.GC12035@fieldses.org>, <20211122033840.GD12035@fieldses.org>,
 <20211122151823.GA22802@fieldses.org>
Date:   Tue, 23 Nov 2021 10:25:40 +1100
Message-id: <163762354003.7248.5982295057216919621@noble.neil.brown.name>
Precedence: bulk

On Tue, 23 Nov 2021, J. Bruce Fields wrote:
> > 
> > Hm, playing with reproducer; it takes more than one mount.  My simplest
> > reproducer is:
> > 
> > 	mount -overs=3 server:/path /mnt/
> > 	umount /mnt/
> > 	mount -overs=4.0 server:/path /mnt/
> > 
> > ... and the client crashes here.
> 
> Also: the problem starts with the last patch ("lockd: use
> svc_set_num_threads() for thread start and stop").

That helps.  It points to lockd, and so to lockd stopping.
Lockd wasn't stopping for me because I had nfsd running.
I disabled nfsd and tried again - and got the crash.

The problem is 
#define svc_serv_is_pooled(serv)    ((serv)->sv_ops->svo_function)

Why does the presence of a svo_function mark the serv as being
'pooled'???

That last patch gave lockd an svo_function, so when threats were
stopped, the pool was released.  But as lockd doesn't create a pooled
service, the pool was never claimed.  So the pool_map->npools was set
to zero even though it should have still been active.
I'll probably change svc_serv_is_pooled() to test serv->sv_nrpools, and
allow that to be zero for non-pooled services.

Thanks for your testing help,
NeilBrown