Re: [PATCH 2/2] sched: adjust SCHED_IDLE interactions

[Josh Don <joshdon@google.com>]

Hi Peter,

On Mon, Aug 16, 2021 at 5:32 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Thu, Aug 12, 2021 at 02:09:15PM -0700, Josh Don wrote:
> > > > @@ -697,8 +699,18 @@ static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
> > > >                 slice = __calc_delta(slice, se->load.weight, load);
> > > >         }
> > > >
> > > > -       if (sched_feat(BASE_SLICE))
> > > > -               slice = max(slice, (u64)w);
> > > > +       if (sched_feat(BASE_SLICE)) {
> > > > +               /*
> > > > +                * SCHED_IDLE entities are not subject to min_granularity if
> > > > +                * they are competing with non SCHED_IDLE entities. As a result,
> > > > +                * non SCHED_IDLE entities will have reduced latency to get back
> > > > +                * on cpu, at the cost of increased context switch frequency of
> > > > +                * SCHED_IDLE entities.
> > > > +                */
> > >
> > > Ensuring that the entity will have a minimum runtime has been added to
> > > ensure that we let enough time to move forward.
> > > If you exclude sched_idle entities from this min runtime, the
> > > sched_slice of an idle_entity will be really small.
> > > I don't have details of your example above but I can imagine that it's
> > > a 16 cpus system which means a sysctl_sched_min_granularity=3.75ms
> > > which explains the 4ms running time of an idle entity
> > > For a 16 cpus system, the sched_slice of an idle_entity in your
> > > example in the cover letter is: 6*(1+log2(16))*3/1027=87us. Of course
> > > this become even worse with more threads and cgroups or thread with
> > > ncie prio -19
> > >
> > > This value is then used to set the next hrtimer event in SCHED_HRTICK
> > > and 87us is too small to make any progress
> > >
> > > The 1ms of your test comes from the tick which could be a good
> > > candidate for a min value or the
> > > normalized_sysctl_sched_min_granularity which has the advantage of not
> > > increasing with number of CPU
> >
> > Fair point, this shouldn't completely ignore min granularity. Something like
> >
> > unsigned int sysctl_sched_idle_min_granularity = NSEC_PER_MSEC;
> >
> > (and still only using this value instead of the default
> > min_granularity when the SCHED_IDLE entity is competing with normal
> > entities)
>
> TICK_NSEC ?

Based on discussion with Vincent, added a separate
sched_idle_min_granularity. Didn't make sense to me to define in terms
of TICK_NSEC, since e.g. we wouldn't necessarily want a smaller slice
on a higher HZ machine.