[PATCH 2/2] nvme-multipath: do not fall back to __nvme_find_path() for non-optimized paths

[Christoph Hellwig]

On Tue, Jul 28, 2020 at 10:14:14AM +0200, Hannes Reinecke wrote:
> On 7/28/20 10:02 AM, Christoph Hellwig wrote:
> > On Tue, Jul 28, 2020 at 08:25:18AM +0200, Hannes Reinecke wrote:
> >>> Maybe instead of the early return put the following condition on else?
> >>
> >> That depends on whether we want to have a fallback to __nvme_find_path() or 
> >> not. With your suggestion we would lose that; I'd rather keep it.
> > 
> > Why would we want to fall back?  nvme_round_robin_path only
> > returns NULL in case there is a single disabled path, in which
> > case __nvme_find_path won't find anything either.  I'd go for something
> > like this:
> > 
> > ---
> > From 4f578364a3d15898c7d715315a3371b2f71db416 Mon Sep 17 00:00:00 2001
> > From: Hannes Reinecke <hare at suse.de>
> > Date: Mon, 27 Jul 2020 18:08:03 +0200
> > Subject: nvme-multipath: do not fall back to __nvme_find_path() for
> >  non-optimized paths
> > 
> > When nvme_round_robin_path() finds a valid namespace we should be using it;
> > falling back to __nvme_find_path() for non-optimized paths will cause the
> > result from nvme_round_robin_path() to be ignored for non-optimized paths.
> > 
> > Signed-off-by: Martin Wilck <mwilck at suse.com>
> > Signed-off-by: Hannes Reinecke <hare at suse.de>
> > Reviewed-by: Sagi Grimberg <sagi at grimberg.me>
> > Signed-off-by: Christoph Hellwig <hch at lst.de>
> > ---
> >  drivers/nvme/host/multipath.c | 11 +++++++----
> >  1 file changed, 7 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
> > index 93c70e1591de8f..3ded54d2c9c6ad 100644
> > --- a/drivers/nvme/host/multipath.c
> > +++ b/drivers/nvme/host/multipath.c
> > @@ -281,10 +281,13 @@ inline struct nvme_ns *nvme_find_path(struct nvme_ns_head *head)
> >  	struct nvme_ns *ns;
> >  
> >  	ns = srcu_dereference(head->current_path[node], &head->srcu);
> > -	if (READ_ONCE(head->subsys->iopolicy) == NVME_IOPOLICY_RR && ns)
> > -		ns = nvme_round_robin_path(head, node, ns);
> > -	if (unlikely(!ns || !nvme_path_is_optimized(ns)))
> > -		ns = __nvme_find_path(head, node);
> > +	if (unlikely(!ns))
> > +		return __nvme_find_path(head, node);
> > +
> > +	if (READ_ONCE(head->subsys->iopolicy) == NVME_IOPOLICY_RR)
> > +		return nvme_round_robin_path(head, node, ns);
> > +	if (unlikely(!nvme_path_is_optimized(ns)))
> > +		return __nvme_find_path(head, node);
> >  	return ns;
> >  }
> >  
> > 
> 
> Yes, that would work, too.
> 
> Should I resend?

No need, I can apply it with the modification.   But if you have a
RR test setup it would be great to run it through that again, just in
case.