Re: [PATCH 4/4] NFSD: Add support for dynamic slot changes

["J. Bruce Fields" <bfields@fieldses.org>]

On Thu, Dec 06, 2012 at 11:20:15PM +0000, Myklebust, Trond wrote:
> > -----Original Message-----
> > From: J. Bruce Fields [mailto:bfields@fieldses.org]
> > Sent: Thursday, December 06, 2012 10:29 PM
> > To: Myklebust, Trond
> > Cc: linux-nfs@vger.kernel.org
> > Subject: Re: [PATCH 4/4] NFSD: Add support for dynamic slot changes
> > 
> > On Wed, Nov 28, 2012 at 10:17:43PM -0500, Trond Myklebust wrote:
> > > @@ -2159,11 +2223,9 @@ nfsd4_sequence(struct svc_rqst *rqstp,
> > >  	else
> > >  		slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
> > >
> > > -	/* Retrieve new target/highest slotid values */
> > > -	spin_lock(&session->se_slots.slt_lock);
> > > -	seq->target_highest_slotid = session-
> > >se_slots.slt_target_highest_slotid;
> > > -	seq->highest_slotid = session->se_slots.slt_highest_slotid;
> > > -	spin_unlock(&session->se_slots.slt_lock);
> > > +	/* Adjust slot table, and retrieve new target/highest slotid values */
> > > +	nfsd4_sequence_adjust_slot_table(session, slot,
> > > +			seq->highest_slotid, seq);
> > 
> > This is allocating under a spin lock (see the client_lock that most of
> > nfsd4_sequence is under.)
> > 
> > (But I'm not sure what we should be doing instead.)
> 
> I've already fixed this issue in the latest patches: the right thing
> is simply to do all this after we drop the client lock.

Oh, sounds reasonable.

> I'll send the
> updated patches when I get a moment: probably Saturday...

Thanks!

--b.