Re: [RFC PATCH 5/5] SUNRPC: Reduce the priority of the xprtiod queue

[Olga Kornievskaia <aglo@umich.edu>]

On Tue, May 28, 2019 at 4:10 PM Olga Kornievskaia <aglo@umich.edu> wrote:
>
> On Fri, May 3, 2019 at 7:24 AM Trond Myklebust <trondmy@gmail.com> wrote:
> >
> > Allow more time for softirqd
> >
> > Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
> > ---
> >  net/sunrpc/sched.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
> > index c7e81336620c..6b37c9a4b48f 100644
> > --- a/net/sunrpc/sched.c
> > +++ b/net/sunrpc/sched.c
> > @@ -1253,7 +1253,7 @@ static int rpciod_start(void)
> >                 goto out_failed;
> >         rpciod_workqueue = wq;
> >         /* Note: highpri because network receive is latency sensitive */
> > -       wq = alloc_workqueue("xprtiod", WQ_UNBOUND|WQ_MEM_RECLAIM|WQ_HIGHPRI, 0);
>
> I thought we needed UNBOUND otherwise there was performance
> degradation for read IO.

I remove my objection as this is for the xprtiod queue and not the
rpciod queue. The latter is the one when removing WQ_UNBOUND would
only use a single rpciod thread for doing all the crypto and thus
impact performance.

>
> > +       wq = alloc_workqueue("xprtiod", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
> >         if (!wq)
> >                 goto free_rpciod;
> >         xprtiod_workqueue = wq;
> > --
> > 2.21.0
> >