Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200801023248.90104-1-benbjiang@gmail.com> <5ed0fd46-3a3d-3c1a-5d75-03a74864e640@arm.com>
 <592F24A7-BF43-457D-AC40-DC5E35279730@tencent.com> <8bef1f94-f9bf-08a5-2ff3-3485d7796a96@arm.com>
 <8629CB9F-AFC8-43D6-BD14-B60A0B85ADB3@tencent.com> <5f870781-1648-b4ac-6026-557dfc347109@arm.com>
 <CCA1D942-3669-4216-92BD-768967B1ECE5@tencent.com> <4964e359-afc5-a256-4950-853a9485eeff@arm.com>
 <70236E62-AA36-48C1-9382-86353649253C@tencent.com> <3a1047fc-a86a-014a-b17a-eae71f669da1@arm.com>
 <643B0ECE-D758-4D08-8B1B-C70F34DD9943@tencent.com> <55f04582-69d6-aeb4-85be-3e46a3b15beb@arm.com>
 <755BFAD0-9072-4D73-9CD7-AF4F74A79D21@tencent.com> <729675a2-b083-4211-62c0-f7ed7f483ae2@arm.com>
 <CAKfTPtATf94n4t6EnB5_76ygox26xd9EwvEHiv5cGN=E9BdLgg@mail.gmail.com>
 <bb7456af-6a4d-874c-e635-8d21e1ab3efc@arm.com> <A8E52241-0206-447D-9EA0-699F8F82AFAB@tencent.com>
 <CAKfTPtBiZHwkhCkMQs9iOF=eBLgubsJkARN5zOKZzmfo_jTJzg@mail.gmail.com>
 <3A85DD77-2A4B-466F-A1F4-1BF2AF02CF58@tencent.com> <CAKfTPtATpDB+ihpMf854nTdDfy4Bw_BU=21PRHVczEcE42271Q@mail.gmail.com>
 <D80F1584-F569-4CEE-8DCC-7841CF7E159F@tencent.com>
In-Reply-To: <D80F1584-F569-4CEE-8DCC-7841CF7E159F@tencent.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Thu, 20 Aug 2020 14:45:59 +0200
Message-ID: <CAKfTPtA_QtEbYT7iBxAZWBq_w=PGAuvg7LsxsccOSRBk19bW8Q@mail.gmail.com>
Subject: Re: [PATCH] sched/fair: reduce preemption with IDLE tasks
 runable(Internet mail)
To:     =?UTF-8?B?YmVuYmppYW5nKOiSi+W9qik=?= <benbjiang@tencent.com>
Cc:     Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Jiang Biao <benbjiang@gmail.com>,
        "mingo@redhat.com" <mingo@redhat.com>,
        "peterz@infradead.org" <peterz@infradead.org>,
        "juri.lelli@redhat.com" <juri.lelli@redhat.com>,
        "rostedt@goodmis.org" <rostedt@goodmis.org>,
        "bsegall@google.com" <bsegall@google.com>,
        "mgorman@suse.de" <mgorman@suse.de>,
        "linux-kernel@vger.kernel.org" <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 Thu, 20 Aug 2020 at 13:28, benbjiang(=E8=92=8B=E5=BD=AA) <benbjiang@tenc=
ent.com> wrote:
>
>
>
> > On Aug 20, 2020, at 3:35 PM, Vincent Guittot <vincent.guittot@linaro.or=
g> wrote:
> >
> > On Thu, 20 Aug 2020 at 02:13, benbjiang(=E8=92=8B=E5=BD=AA) <benbjiang@=
tencent.com> wrote:
> >>
> >>
> >>
> >>> On Aug 19, 2020, at 10:55 PM, Vincent Guittot <vincent.guittot@linaro=
.org> wrote:
> >>>
> >>> On Wed, 19 Aug 2020 at 16:27, benbjiang(=E8=92=8B=E5=BD=AA) <benbjian=
g@tencent.com> wrote:
> >>>>
> >>>>
> >>>>
> >>>>> On Aug 19, 2020, at 7:55 PM, Dietmar Eggemann <dietmar.eggemann@arm=
.com> wrote:
> >>>>>
> >>>>> On 19/08/2020 13:05, Vincent Guittot wrote:
> >>>>>> On Wed, 19 Aug 2020 at 12:46, Dietmar Eggemann <dietmar.eggemann@a=
rm.com> wrote:
> >>>>>>>
> >>>>>>> On 17/08/2020 14:05, benbjiang(=E8=92=8B=E5=BD=AA) wrote:
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>> On Aug 17, 2020, at 4:57 PM, Dietmar Eggemann <dietmar.eggemann=
@arm.com> wrote:
> >>>>>>>>>
> >>>>>>>>> On 14/08/2020 01:55, benbjiang(=E8=92=8B=E5=BD=AA) wrote:
> >>>>>>>>>> Hi,
> >>>>>>>>>>
> >>>>>>>>>>> On Aug 13, 2020, at 2:39 AM, Dietmar Eggemann <dietmar.eggema=
nn@arm.com> wrote:
> >>>>>>>>>>>
> >>>>>>>>>>> On 12/08/2020 05:19, benbjiang(=E8=92=8B=E5=BD=AA) wrote:
> >>>>>>>>>>>> Hi,
> >>>>>>>>>>>>
> >>>>>>>>>>>>> On Aug 11, 2020, at 11:54 PM, Dietmar Eggemann <dietmar.egg=
emann@arm.com> wrote:
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> On 11/08/2020 02:41, benbjiang(=E8=92=8B=E5=BD=AA) wrote:
> >>>>>>>>>>>>>> Hi,
> >>>>>>>>>>>>>>
> >>>>>>>>>>>>>>> On Aug 10, 2020, at 9:24 PM, Dietmar Eggemann <dietmar.eg=
gemann@arm.com> wrote:
> >>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>> On 06/08/2020 17:52, benbjiang(=E8=92=8B=E5=BD=AA) wrote:
> >>>>>>>>>>>>>>>> Hi,
> >>>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>>>> On Aug 6, 2020, at 9:29 PM, Dietmar Eggemann <dietmar.e=
ggemann@arm.com> wrote:
> >>>>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>>>> On 03/08/2020 13:26, benbjiang(=E8=92=8B=E5=BD=AA) wrot=
e:
> >>>>>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>>>>>> On Aug 3, 2020, at 4:16 PM, Dietmar Eggemann <dietmar=
.eggemann@arm.com> wrote:
> >>>>>>>>>>>>>>>>>>>
> >>>>>>>>>>>>>>>>>>> On 01/08/2020 04:32, Jiang Biao wrote:
> >>>>>>>>>>>>>>>>>>>> From: Jiang Biao <benbjiang@tencent.com>
> >>>>>>>
> >>>>>>> [...]
> >>>>>>>
> >>>>>>>>> Are you sure about this?
> >>>>>>>> Yes. :)
> >>>>>>>>>
> >>>>>>>>> The math is telling me for the:
> >>>>>>>>>
> >>>>>>>>> idle task:      (3 / (1024 + 1024 + 3))^(-1) * 4ms =3D 2735ms
> >>>>>>>>>
> >>>>>>>>> normal task: (1024 / (1024 + 1024 + 3))^(-1) * 4ms =3D    8ms
> >>>>>>>>>
> >>>>>>>>> (4ms - 250 Hz)
> >>>>>>>> My tick is 1ms - 1000HZ, which seems reasonable for 600ms? :)
> >>>>>>>
> >>>>>>> OK, I see.
> >>>>>>>
> >>>>>>> But here the different sched slices (check_preempt_tick()->
> >>>>>>> sched_slice()) between normal tasks and the idle task play a role=
 to.
> >>>>>>>
> >>>>>>> Normal tasks get ~3ms whereas the idle task gets <0.01ms.
> >>>>>>
> >>>>>> In fact that depends on the number of CPUs on the system
> >>>>>> :sysctl_sched_latency =3D 6ms * (1 + ilog(ncpus)) . On a 8 cores s=
ystem,
> >>>>>> normal task will run around 12ms in one shoot and the idle task st=
ill
> >>>>>> one tick period
> >>>>>
> >>>>> True. This is on a single CPU.
> >>>> Agree. :)
> >>>>
> >>>>>
> >>>>>> Also, you can increase even more the period between 2 runs of idle
> >>>>>> task by using cgroups and min shares value : 2
> >>>>>
> >>>>> Ah yes, maybe this is what Jiang wants to do then? If his runtime d=
oes
> >>>>> not have other requirements preventing this.
> >>>> That could work for increasing the period between 2 runs. But could =
not
> >>>> reduce the single runtime of idle task I guess, which means normal t=
ask
> >>>> could have 1-tick schedule latency because of idle task.
> >>>
> >>> Yes.  An idle task will preempt an always running task during 1 tick
> >>> every 680ms. But also you should keep in mind that a waking normal
> >>> task will preempt the idle task immediately which means that it will
> >>> not add scheduling latency to a normal task but "steal" 0.14% of
> >>> normal task throughput (1/680) at most
> >> That=E2=80=99s true. But in the VM case, when VM are busy(MWAIT passth=
rough
> >> or running cpu eating works), the 1-tick scheduling latency could be
> >> detected by cyclictest running in the VM.
> >>
> >> OTOH, we compensate vruntime in place_entity() to boot waking
> >> without distinguish SCHED_IDLE task, do you think it=E2=80=99s necessa=
ry to
> >> do that? like
> >>
> >> --- a/kernel/sched/fair.c
> >> +++ b/kernel/sched/fair.c
> >> @@ -4115,7 +4115,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched=
_entity *se, int initial)
> >>                vruntime +=3D sched_vslice(cfs_rq, se);
> >>
> >>        /* sleeps up to a single latency don't count. */
> >> -       if (!initial) {
> >> +       if (!initial && likely(!task_has_idle_policy(task_of(se)))) {
> >>                unsigned long thresh =3D sysctl_sched_latency;
> >
> > Yeah, this is a good improvement.
> Thanks, I=E2=80=99ll send a patch for that. :)
>
> > Does it solve your problem ?
> >
> Not exactly. :)  I wonder if we can make SCHED_IDLE more pure(harmless)?

We can't prevent it from running time to time. Proxy execution feature
could be a step for considering to relax this constraint

> Or introduce a switch(or flag) to control it, and make it available for c=
ases like us.
>
> Thanks a lot.
> Regards,
> Jiang
>
> >>
> >>>
> >>>> OTOH, cgroups(shares) could introduce extra complexity. :)
> >>>>
> >>>> I wonder if there=E2=80=99s any possibility to make SCHED_IDLEs=E2=
=80=99 priorities absolutely
> >>>> lower than SCHED_NORMAL(OTHER), which means no weights/shares
> >>>> for them, and they run only when no other task=E2=80=99s runnable.
> >>>> I guess there may be priority inversion issue if we do that. But may=
be we
> >>>
> >>> Exactly, that's why we must ensure a minimum running time for sched_i=
dle task
> >>
> >> Still for VM case, different VMs have been much isolated from each oth=
er,
> >> priority inversion issue could be very rare, we=E2=80=99re trying to m=
ake offline tasks
> >> absoultly harmless to online tasks. :)
> >>
> >> Thanks a lot for your time.
> >> Regards,
> >> Jiang
> >>
> >>>
> >>>> could avoid it by load-balance more aggressively, or it(priority inv=
ersion)
> >>>> could be ignored in some special case.
>