linux-kernel - Re: [PATCH 1/2] sched/fair: skip newidle update stats

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAKfTPtBmGCQKnjP4SBMskBLWXrYKkWRpjrdNpJua6aERa85Z_w@mail.gmail.com>
Date:   Fri, 12 Nov 2021 17:00:04 +0100
From:   Vincent Guittot <vincent.guittot@...aro.org>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     mingo@...hat.com, juri.lelli@...hat.com, dietmar.eggemann@....com,
        rostedt@...dmis.org, bsegall@...gle.com, mgorman@...e.de,
        bristot@...hat.com, linux-kernel@...r.kernel.org,
        tim.c.chen@...ux.intel.com, joel@...lfernandes.org
Subject: Re: [PATCH 1/2] sched/fair: skip newidle update stats

On Fri, 12 Nov 2021 at 16:29, Peter Zijlstra <peterz@...radead.org> wrote:
>
> On Fri, Nov 12, 2021 at 03:47:21PM +0100, Vincent Guittot wrote:
> > On Fri, 12 Nov 2021 at 15:29, Peter Zijlstra <peterz@...radead.org> wrote:
> > >
> > > On Fri, Nov 12, 2021 at 10:58:56AM +0100, Vincent Guittot wrote:
> > > > In case we skip the newly idle LB entirely or we abort it because we are
> > > > going to exceed the avg_idle, we have to make sure to not start an update
> > > > of the blocked load when entering idle
> > > >
> > > > Signed-off-by: Vincent Guittot <vincent.guittot@...aro.org>
> > > > ---
> > > >  kernel/sched/fair.c | 18 ++++++++++++++----
> > > >  1 file changed, 14 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > > > index 13950beb01a2..a162b0ec8963 100644
> > > > --- a/kernel/sched/fair.c
> > > > +++ b/kernel/sched/fair.c
> > > > @@ -10861,7 +10861,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> > > >       int this_cpu = this_rq->cpu;
> > > >       u64 t0, t1, curr_cost = 0;
> > > >       struct sched_domain *sd;
> > > > -     int pulled_task = 0;
> > > > +     int pulled_task = 0, early_stop = 0;
> > > >
> > > >       update_misfit_status(NULL, this_rq);
> > > >
> > > > @@ -10898,8 +10898,16 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> > > >       if (!READ_ONCE(this_rq->rd->overload) ||
> > > >           (sd && this_rq->avg_idle < sd->max_newidle_lb_cost)) {
> > > >
> > > > -             if (sd)
> > > > +             if (sd) {
> > > >                       update_next_balance(sd, &next_balance);
> > > > +
> > > > +                     /*
> > > > +                      * We skip new idle LB because there is not enough
> > > > +                      * time before next wake up. Make sure that we will
> > > > +                      * not kick NOHZ_NEWILB_KICK
> > > > +                      */
> > > > +                     early_stop = 1;
> > > > +             }
> > > >               rcu_read_unlock();
> > > >
> > > >               goto out;
>
> > > Anyway, does nohz_newidle_balance() want to loose it's ->avg_idle test
> > > with this on?
> >
> > This test still covers cases with short newly idle balance. Being
> > conservative, people never complained that the update of blocked load
> > average of idle CPUs doesn't happen often enough. It's most often the
> > opposite
>
> Well, per commit c5b0a7eefc70 ("sched/fair: Remove
> sysctl_sched_migration_cost condition") combined with the above change,
> we no longer call nohz_newidle_balance() in exactly that condition,
> right?
>
> Or are we worried about that !overload case?

we can do a complete newly idle LB but have this_rq->avg_idle <
sysctl_sched_migration_cost. In this case, the condition will continue
to skip update of other idle CPUs