Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200226181640.21664-1-vincent.guittot@linaro.org>
 <xm26r1yhtbjr.fsf@bsegall-linux.svl.corp.google.com> <CAKfTPtBm9Gt16gqQgxoErOOmpbUHit6bNf4CVLvDzf04SjWtEg@mail.gmail.com>
 <8f72ea72-f36d-2611-e026-62ddff5c3422@arm.com> <CAKfTPtC9bkMQJsWw6Z2QD0RrV=qN7yMFviVnSeTpDp=-vLBL0g@mail.gmail.com>
In-Reply-To: <CAKfTPtC9bkMQJsWw6Z2QD0RrV=qN7yMFviVnSeTpDp=-vLBL0g@mail.gmail.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Thu, 27 Feb 2020 15:58:02 +0100
Message-ID: <CAKfTPtD4iVQmxWgNDDVhKPbu+rYEf=_1xKoPVOy343qo51pD_A@mail.gmail.com>
Subject: Re: [PATCH] sched/fair: fix runnable_avg for throttled cfs
To:     Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc:     Ben Segall <bsegall@google.com>, Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Juri Lelli <juri.lelli@redhat.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Mel Gorman <mgorman@suse.de>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Phil Auld <pauld@redhat.com>, Parth Shah <parth@linux.ibm.com>,
        Valentin Schneider <valentin.schneider@arm.com>,
        Hillf Danton <hdanton@sina.com>, zhout@vivaldi.net
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 27 Feb 2020 at 14:10, Vincent Guittot
<vincent.guittot@linaro.org> wrote:
>
> On Thu, 27 Feb 2020 at 12:20, Dietmar Eggemann <dietmar.eggemann@arm.com> wrote:
> >
> > On 26.02.20 21:01, Vincent Guittot wrote:
> > > On Wed, 26 Feb 2020 at 20:04, <bsegall@google.com> wrote:
> > >>
> > >> Vincent Guittot <vincent.guittot@linaro.org> writes:
> > >>
> > >>> When a cfs_rq is throttled, its group entity is dequeued and its running
> > >>> tasks are removed. We must update runnable_avg with current h_nr_running
> > >>> and update group_se->runnable_weight with new h_nr_running at each level
> >
> >                                               ^^^
> >
> > Shouldn't this be 'current' rather 'new' h_nr_running for
> > group_se->runnable_weight? IMHO, you want to cache the current value
> > before you add/subtract task_delta.
>
> hmm... it can't be current in both places. In my explanation,
> "current" means the current situation when we started to throttle cfs
> and "new" means the new situation after we finished to throttle the
> cfs. I should probably use old and new to prevent any
> misunderstanding.

I'm about to send a new version to fix some minor changes: The if
statement should have some  { }   as there are some on the else part

Would it be better for you if i use old and new instead of current and
new in the commit message ?

>
> That being said, we need to update runnable_avg with the old
> h_nr_running: the one before we started to throttle the cfs which is
> the value saved in group_se->runnable_weight. Once we have updated
> runnable_avg, we save the new h_nr_running in
> group_se->runnable_weight that will be used for next updates.
>
> >
> > >>> of the hierarchy.
> > >>
> > >> You'll also need to do this for task enqueue/dequeue inside of a
> > >> throttled hierarchy, I'm pretty sure.
> > >
> > > AFAICT, this is already done with patch "sched/pelt: Add a new
> > > runnable average signal" when task is enqueued/dequeued inside a
> > > throttled hierarchy
> > >
> > >>
> > >>>
> > >>> Fixes: 9f68395333ad ("sched/pelt: Add a new runnable average signal")
> > >>> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
> > >>> ---
> > >>> This patch applies on top of tip/sched/core
> > >>>
> > >>>  kernel/sched/fair.c | 10 ++++++++++
> > >>>  1 file changed, 10 insertions(+)
> > >>>
> > >>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > >>> index fcc968669aea..6d46974e9be7 100644
> > >>> --- a/kernel/sched/fair.c
> > >>> +++ b/kernel/sched/fair.c
> > >>> @@ -4703,6 +4703,11 @@ static void throttle_cfs_rq(struct cfs_rq *cfs_rq)
> > >>>
> > >>>               if (dequeue)
> > >>>                       dequeue_entity(qcfs_rq, se, DEQUEUE_SLEEP);
> > >>> +             else {
> > >>> +                     update_load_avg(qcfs_rq, se, 0);
> > >>> +                     se_update_runnable(se);
> > >>> +             }
> > >>> +
> > >>>               qcfs_rq->h_nr_running -= task_delta;
> > >>>               qcfs_rq->idle_h_nr_running -= idle_task_delta;
> > >>>
> > >>> @@ -4772,6 +4777,11 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
> > >>>               cfs_rq = cfs_rq_of(se);
> > >>>               if (enqueue)
> > >>>                       enqueue_entity(cfs_rq, se, ENQUEUE_WAKEUP);
> > >>> +             else {
> > >>> +                     update_load_avg(cfs_rq, se, 0);
> > >>
> > >>
> > >>> +                     se_update_runnable(se);
> > >>> +             }
> > >>> +
> > >>>               cfs_rq->h_nr_running += task_delta;
> > >>>               cfs_rq->idle_h_nr_running += idle_task_delta;