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: <44fa1cee-08db-e4ab-e5ab-08d6fbd421d7@linux.alibaba.com>
 <20200303195245.GF2596@hirez.programming.kicks-ass.net> <xm26o8tc3qkv.fsf@bsegall-linux.svl.corp.google.com>
 <1180c6cd-ff61-2c9f-d689-ffe58f8c5a68@linux.alibaba.com> <xm267dzx47k9.fsf@bsegall-linux.svl.corp.google.com>
In-Reply-To: <xm267dzx47k9.fsf@bsegall-linux.svl.corp.google.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Mon, 9 Mar 2020 12:15:36 +0100
Message-ID: <CAKfTPtDKTp_G1VNgAXnh=_yLS_T6YkipOsQQ52tBRp-m612JEw@mail.gmail.com>
Subject: Re: [RFC PATCH] sched: fix the nonsense shares when load of cfs_rq is
 too, small
To:     Ben Segall <bsegall@google.com>
Cc:     =?UTF-8?B?546L6LSH?= <yun.wang@linux.alibaba.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Mel Gorman <mgorman@suse.de>,
        "open list:SCHEDULER" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, 6 Mar 2020 at 20:17, <bsegall@google.com> wrote:
>
> =E7=8E=8B=E8=B4=87 <yun.wang@linux.alibaba.com> writes:
>
> > On 2020/3/5 =E4=B8=8A=E5=8D=882:47, bsegall@google.com wrote:
> > [snip]
> >>> Argh, because A->cfs_rq.load.weight is B->se.load.weight which is
> >>> B->shares/nr_cpus.
> >>>
> >>>> While the se of D on root cfs_rq is far more bigger than 2, so it
> >>>> wins the battle.
> >>>>
> >>>> This patch add a check on the zero load and make it as MIN_SHARES
> >>>> to fix the nonsense shares, after applied the group C wins as
> >>>> expected.
> >>>>
> >>>> Signed-off-by: Michael Wang <yun.wang@linux.alibaba.com>
> >>>> ---
> >>>>  kernel/sched/fair.c | 2 ++
> >>>>  1 file changed, 2 insertions(+)
> >>>>
> >>>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> >>>> index 84594f8aeaf8..53d705f75fa4 100644
> >>>> --- a/kernel/sched/fair.c
> >>>> +++ b/kernel/sched/fair.c
> >>>> @@ -3182,6 +3182,8 @@ static long calc_group_shares(struct cfs_rq *c=
fs_rq)
> >>>>    tg_shares =3D READ_ONCE(tg->shares);
> >>>>
> >>>>    load =3D max(scale_load_down(cfs_rq->load.weight), cfs_rq->avg.lo=
ad_avg);
> >>>> +  if (!load && cfs_rq->load.weight)
> >>>> +          load =3D MIN_SHARES;
> >>>>
> >>>>    tg_weight =3D atomic_long_read(&tg->load_avg);
> >>>
> >>> Yeah, I suppose that'll do. Hurmph, wants a comment though.
> >>>
> >>> But that has me looking at other users of scale_load_down(), and does=
n't
> >>> at least update_tg_cfs_load() suffer the same problem?
> >>
> >> I think instead we should probably scale_load_down(tg_shares) and
> >> scale_load(load_avg). tg_shares is always a scaled integer, so just
> >> moving the source of the scaling in the multiply should do the job.
> >>
> >> ie
> >>
> >> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> >> index fcc968669aea..6d7a9d72d742 100644
> >> --- a/kernel/sched/fair.c
> >> +++ b/kernel/sched/fair.c
> >> @@ -3179,9 +3179,9 @@ static long calc_group_shares(struct cfs_rq *cfs=
_rq)
> >>         long tg_weight, tg_shares, load, shares;
> >>         struct task_group *tg =3D cfs_rq->tg;
> >>
> >> -       tg_shares =3D READ_ONCE(tg->shares);
> >> +       tg_shares =3D scale_load_down(READ_ONCE(tg->shares));
> >>
> >> -       load =3D max(scale_load_down(cfs_rq->load.weight), cfs_rq->avg=
.load_avg);
> >> +       load =3D max(cfs_rq->load.weight, scale_load(cfs_rq->avg.load_=
avg));
> >>
> >>         tg_weight =3D atomic_long_read(&tg->load_avg);
> >
> > Get the point, but IMHO fix scale_load_down() sounds better, to
> > cover all the similar cases, let's first try that way see if it's
> > working :-)
>
> Yeah, that might not be a bad idea as well; it's just that doing this
> fix would keep you from losing all your precision (and I'd have to think
> if that would result in fairness issues like having all the group ses
> having the full tg shares, or something like that).

AFAICT, we already have a fairness problem case because
scale_load_down is used in calc_delta_fair() so all sched groups that
have a weight lower than 1024 will end up with the same increase of
their vruntime when running.
Then the load_avg is used to balance between rq so load_balance will
ensure at least 1 task per CPU but not more because the load_avg which
is then used will stay null.

That being said, having a min of 2 for scale_load_down will enable us
to have the tg->load_avg !=3D 0 so a tg_weight !=3D 0 and each sched group
will not have the full shares. But it will make those group completely
fair anyway.
The best solution would be not to scale down the weight but that's a
bigger change