Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220427143304.3950488-1-vincent.donnefort@arm.com>
 <20220427143304.3950488-3-vincent.donnefort@arm.com> <Yml/icTe26CfweCd@geo.homenetwork>
In-Reply-To: <Yml/icTe26CfweCd@geo.homenetwork>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Thu, 28 Apr 2022 15:38:39 +0200
Message-ID: <CAKfTPtCsZTm_jx-BS00UkFUqW66x--6T8Bb2LRsD1S0a3rN0iQ@mail.gmail.com>
Subject: Re: [PATCH v7 2/7] sched/fair: Decay task PELT values during wakeup migration
To:     Tao Zhou <tao.zhou@linux.dev>
Cc:     Vincent Donnefort <vincent.donnefort@arm.com>,
        peterz@infradead.org, mingo@redhat.com,
        linux-kernel@vger.kernel.org, dietmar.eggemann@arm.com,
        morten.rasmussen@arm.com, chris.redpath@arm.com, qperret@google.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, 27 Apr 2022 at 19:37, Tao Zhou <tao.zhou@linux.dev> wrote:
>
> On Wed, Apr 27, 2022 at 03:32:59PM +0100, Vincent Donnefort wrote:
> > Before being migrated to a new CPU, a task sees its PELT values
> > synchronized with rq last_update_time. Once done, that same task will also
> > have its sched_avg last_update_time reset. This means the time between
> > the migration and the last clock update (B) will not be accounted for in
> > util_avg and a discontinuity will appear. This issue is amplified by the
> > PELT clock scaling. If the clock hasn't been updated while the CPU is
> > idle, clock_pelt will not be aligned with clock_task and that time (A)
> > will be also lost.
> >
> >    ---------|----- A -----|-----------|------- B -----|>
> >         clock_pelt   clock_task     clock            now
> >
> > This is especially problematic for asymmetric CPU capacity systems which
> > need stable util_avg signals for task placement and energy estimation.
> >
> > Ideally, this problem would be solved by updating the runqueue clocks
> > before the migration. But that would require taking the runqueue lock
> > which is quite expensive [1]. Instead estimate the missing time and update
> > the task util_avg with that value:
> >
> >   A + B = clock_task - clock_pelt + sched_clock_cpu() - clock
> >
> > sched_clock_cpu() is a costly function. Limit the usage to the case where
> > the source CPU is idle as we know this is when the clock is having the
> > biggest risk of being outdated.
> >
> > Neither clock_task, clock_pelt nor clock can be accessed without the
> > runqueue lock. We then need to store those values in a timestamp variable
> > which can be accessed during the migration. rq's enter_idle will give the
> > wall-clock time when the rq went idle. We have then:
> >
> >   B = sched_clock_cpu() - rq->enter_idle.
> >
> > Then, to catch-up the PELT clock scaling (A), two cases:
> >
> >   * !CFS_BANDWIDTH: We can simply use clock_task(). This value is stored
> >     in rq's clock_pelt_idle, before the rq enters idle. The estimated time
> >     is then:
> >
> >       rq->clock_pelt_idle + sched_clock_cpu() - rq->enter_idle.
> >
> >   * CFS_BANDWIDTH: We can't catch-up with clock_task because of the
> >     throttled_clock_task_time offset. cfs_rq's clock_pelt_idle is then
> >     giving the PELT clock when the cfs_rq becomes idle. This gives:
> >
> >       A = rq->clock_pelt_idle - cfs_rq->clock_pelt_idle
> >
> >     And gives the following estimated time:
> >
> >       cfs_rq->last_update_time +
> >       rq->clock_pelt_idle - cfs_rq->clock_pelt_idle + (A)
> >       sched_clock_cpu() - rq->enter_idle (B)
> >
> > The (B) part of the missing time is however an estimation that doesn't
> > take into account IRQ and Paravirt time.
> >
> > [1] https://lore.kernel.org/all/20190709115759.10451-1-chris.redpath@arm.com/
> >
> > Signed-off-by: Vincent Donnefort <vincent.donnefort@arm.com>
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index abd1feeec0c2..9cd506dc682c 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -3694,6 +3694,57 @@ static inline void add_tg_cfs_propagate(struct cfs_rq *cfs_rq, long runnable_sum
> >
> >  #endif /* CONFIG_FAIR_GROUP_SCHED */
> >
> > +#ifdef CONFIG_NO_HZ_COMMON
> > +static inline void migrate_se_pelt_lag(struct sched_entity *se)
> > +{
> > +     struct cfs_rq *cfs_rq;
> > +     struct rq *rq;
> > +     bool is_idle;
> > +     u64 now;
> > +

would it make sense to check if pelt value of the task are not fully
decayed before starting the below : ie after syncing with
last_update_time of the cfs

> > +     cfs_rq = cfs_rq_of(se);
> > +     rq = rq_of(cfs_rq);
> > +
> > +     rcu_read_lock();
> > +     is_idle = is_idle_task(rcu_dereference(rq->curr));
> > +     rcu_read_unlock();
> > +
> > +     /*
> > +      * The lag estimation comes with a cost we don't want to pay all the
> > +      * time. Hence, limiting to the case where the source CPU is idle and
> > +      * we know we are at the greatest risk to have an outdated clock.
> > +      */
> > +     if (!is_idle)
> > +             return;
> > +
> > +     /*
> > +      * estimated "now" is:
> > +      *   last_update_time +
> > +      *   PELT scaling (rq->clock_pelt_idle - cfs_rq->clock_pelt_idle) +

PELT scaling is in fact the time between cfs becoming idle and rq
becoming idle. Naming it PELT scaling is misleading because even at
max frequency (ie without pelt scaling) we can have this delta.

> > +      *   rq clock lag (sched_clock_cpu() - rq->enter_idle)

and this is the time between rq becoming idle and current time

> > +      *
> > +      *   The PELT scaling contribution is always 0 when !CFS_BANDWIDTH.
> > +      *   (see clock_pelt = clock_task in _update_idle_rq_clock_pelt())

The contribution becomes 0 because we use the same clock reference

last_update_time (cfs_clock_pelt when cfs became idle) +
rq->clock_pelt_idle (rq_clock_pelt when rq became idle) -
cfs_rq->clock_pelt_idle (rq_clock_pelt when cfs became idle)

when !CFS_BANDWIDTH, cfs_clock_pelt == rq_clock_pelt because there is
no throttling offset (which can dynamically change)
so we have:

last_update_time (rq_clock_pelt when cfs became idle) +
rq->clock_pelt_idle (rq_clock_pelt when rq became idle) -
cfs_rq->clock_pelt_idle (rq_clock_pelt when cfs became idle)

which is equals to rq->clock_pelt_idle (rq_clock_pelt when rq became idle)

This also means that we only need a snapshot of the
cfs_rq->throttled_clock_pelt_time when cfs became idle and the
equation becomes like below for CFS_BANDWIDTH

 rq->clock_pelt_idle - snapshot of cfs_rq->throttled_clock_pelt_time
when entering idle

which remove one u64_u32_load

> > +      */
> > +#ifdef CONFIG_CFS_BANDWIDTH
> > +     now = u64_u32_load(cfs_rq->clock_pelt_idle);
> > +     /* The clock has been stopped for throttling */
> > +     if (now == U64_MAX)
> > +             return;
> > +
> > +     now = u64_u32_load(rq->clock_pelt_idle) - now;
> > +     now += cfs_rq_last_update_time(cfs_rq);
> > +#else
> > +     now = u64_u32_load(rq->clock_pelt_idle);
> > +#endif
> > +     now += sched_clock_cpu(cpu_of(rq)) - u64_u32_load(rq->enter_idle);
> > +
> > +     __update_load_avg_blocked_se(now, se);
> > +}
> > +#else
> > +static void migrate_se_pelt_lag(struct sched_entity *se) {}
> > +#endif
> > +
> >  /**
> >   * update_cfs_rq_load_avg - update the cfs_rq's load/util averages
> >   * @now: current time, as per cfs_rq_clock_pelt()
> > @@ -4429,6 +4480,9 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
> >        */
> >       if ((flags & (DEQUEUE_SAVE | DEQUEUE_MOVE)) != DEQUEUE_SAVE)
> >               update_min_vruntime(cfs_rq);
> > +
> > +     if (cfs_rq->nr_running == 0)
> > +             update_idle_cfs_rq_clock_pelt(cfs_rq);
> >  }
> >
> >  /*
> > @@ -6946,6 +7000,8 @@ static void detach_entity_cfs_rq(struct sched_entity *se);
> >   */
> >  static void migrate_task_rq_fair(struct task_struct *p, int new_cpu)
> >  {
> > +     struct sched_entity *se = &p->se;
> > +
> >       /*
> >        * As blocked tasks retain absolute vruntime the migration needs to
> >        * deal with this by subtracting the old and adding the new
> > @@ -6953,7 +7009,6 @@ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu)
> >        * the task on the new runqueue.
> >        */
> >       if (READ_ONCE(p->__state) == TASK_WAKING) {
> > -             struct sched_entity *se = &p->se;
> >               struct cfs_rq *cfs_rq = cfs_rq_of(se);
> >
> >               se->vruntime -= u64_u32_load(cfs_rq->min_vruntime);
> > @@ -6965,25 +7020,29 @@ static void migrate_task_rq_fair(struct task_struct *p, int new_cpu)
> >                * rq->lock and can modify state directly.
> >                */
> >               lockdep_assert_rq_held(task_rq(p));
> > -             detach_entity_cfs_rq(&p->se);
> > +             detach_entity_cfs_rq(se);
> >
> >       } else {
> > +             remove_entity_load_avg(se);
> > +
> >               /*
> > -              * We are supposed to update the task to "current" time, then
> > -              * its up to date and ready to go to new CPU/cfs_rq. But we
> > -              * have difficulty in getting what current time is, so simply
> > -              * throw away the out-of-date time. This will result in the
> > -              * wakee task is less decayed, but giving the wakee more load
> > -              * sounds not bad.
> > +              * Here, the task's PELT values have been updated according to
> > +              * the current rq's clock. But if that clock hasn't been
> > +              * updated in a while, a substantial idle time will be missed,
> > +              * leading to an inflation after wake-up on the new rq.
> > +              *
> > +              * Estimate the missing time from the cfs_rq last_update_time
> > +              * and update sched_avg to improve the PELT continuity after
> > +              * migration.
> >                */
> > -             remove_entity_load_avg(&p->se);
> > +             migrate_se_pelt_lag(se);
> >       }
> >
> >       /* Tell new CPU we are migrated */
> > -     p->se.avg.last_update_time = 0;
> > +     se->avg.last_update_time = 0;
> >
> >       /* We have migrated, no longer consider this task hot */
> > -     p->se.exec_start = 0;
> > +     se->exec_start = 0;
> >
> >       update_scan_period(p, new_cpu);
> >  }
> > @@ -8149,6 +8208,10 @@ static bool __update_blocked_fair(struct rq *rq, bool *done)
> >               if (update_cfs_rq_load_avg(cfs_rq_clock_pelt(cfs_rq), cfs_rq)) {
> >                       update_tg_load_avg(cfs_rq);
> >
> > +                     /* sync clock_pelt_idle with last update */
> > +                     if (cfs_rq->nr_running == 0)
> > +                             update_idle_cfs_rq_clock_pelt(cfs_rq);
>
> I think that if cfs_rq->nr_running == 0 then use cfs rq pelt_idle to update
> idle cfs rq.

update_blocked_averages() updates all cfs rq to be aligned with now so
we don't need to calculate an estimated now. update_rq_clock(rq) is
called 1st to update the rq->clock and childs

With only need to save when happened the last update which is done in
update_rq_clock_pelt(rq) for rq->clock_pelt and with
update_idle_cfs_rq_clock_pelt(cfs) for the cfs_rq_clock_pelt


>
> if (!cfs_rq->nr_running) {
>     /* A part. calculation of idle cfs rq */
>     calculate now like in migrate_se_pelt_lag().
>     decay = update_cfs_rq_load_avg(now, cfs_rq);
> } else {
>     decay = update_cfs_rq_load_avg(cfs_rq_clock_pelt(cfs_rq), cfs_rq))
> }
>
> if (decay) {
>     update_tg_load_avg(cfs_rq);
>     if (cfs_rq == &rq->cfs)
>         decayed == ture;
> }
>
> Thanks,
> Tao
> >                       if (cfs_rq == &rq->cfs)
> >                               decayed = true;
> >               }
> > diff --git a/kernel/sched/pelt.h b/kernel/sched/pelt.h
> > index 4ff2ed4f8fa1..6b39e07b2919 100644
> > --- a/kernel/sched/pelt.h
> > +++ b/kernel/sched/pelt.h
> > @@ -61,6 +61,23 @@ static inline void cfs_se_util_change(struct sched_avg *avg)
> >       WRITE_ONCE(avg->util_est.enqueued, enqueued);
> >  }
> >
> > +static inline u64 rq_clock_pelt(struct rq *rq)
> > +{
> > +     lockdep_assert_rq_held(rq);
> > +     assert_clock_updated(rq);
> > +
> > +     return rq->clock_pelt - rq->lost_idle_time;
> > +}
> > +
> > +/* The rq is idle, we can sync to clock_task */
> > +static inline void _update_idle_rq_clock_pelt(struct rq *rq)
> > +{
> > +     rq->clock_pelt  = rq_clock_task(rq);
> > +
> > +     u64_u32_store(rq->enter_idle, rq_clock(rq));
> > +     u64_u32_store(rq->clock_pelt_idle, rq_clock_pelt(rq));
> > +}
> > +
> >  /*
> >   * The clock_pelt scales the time to reflect the effective amount of
> >   * computation done during the running delta time but then sync back to
> > @@ -76,8 +93,7 @@ static inline void cfs_se_util_change(struct sched_avg *avg)
> >  static inline void update_rq_clock_pelt(struct rq *rq, s64 delta)
> >  {
> >       if (unlikely(is_idle_task(rq->curr))) {
> > -             /* The rq is idle, we can sync to clock_task */
> > -             rq->clock_pelt  = rq_clock_task(rq);
> > +             _update_idle_rq_clock_pelt(rq);
> >               return;
> >       }
> >
> > @@ -130,17 +146,20 @@ static inline void update_idle_rq_clock_pelt(struct rq *rq)
> >        */
> >       if (util_sum >= divider)
> >               rq->lost_idle_time += rq_clock_task(rq) - rq->clock_pelt;
> > +
> > +     _update_idle_rq_clock_pelt(rq);
> >  }
> >
> > -static inline u64 rq_clock_pelt(struct rq *rq)
> > +#ifdef CONFIG_CFS_BANDWIDTH
> > +static inline void update_idle_cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
> >  {
> > -     lockdep_assert_rq_held(rq);
> > -     assert_clock_updated(rq);
> > -
> > -     return rq->clock_pelt - rq->lost_idle_time;
> > +     if (unlikely(cfs_rq->throttle_count))
> > +             u64_u32_store(cfs_rq->clock_pelt_idle, U64_MAX);
> > +     else
> > +             u64_u32_store(cfs_rq->clock_pelt_idle,
> > +                           rq_clock_pelt(rq_of(cfs_rq)));
> >  }
> >
> > -#ifdef CONFIG_CFS_BANDWIDTH
> >  /* rq->task_clock normalized against any time this cfs_rq has spent throttled */
> >  static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
> >  {
> > @@ -150,6 +169,7 @@ static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
> >       return rq_clock_pelt(rq_of(cfs_rq)) - cfs_rq->throttled_clock_pelt_time;
> >  }
> >  #else
> > +static inline void update_idle_cfs_rq_clock_pelt(struct cfs_rq *cfs_rq) { }
> >  static inline u64 cfs_rq_clock_pelt(struct cfs_rq *cfs_rq)
> >  {
> >       return rq_clock_pelt(rq_of(cfs_rq));
> > @@ -204,6 +224,7 @@ update_rq_clock_pelt(struct rq *rq, s64 delta) { }
> >  static inline void
> >  update_idle_rq_clock_pelt(struct rq *rq) { }
> >
> > +static inline void update_idle_cfs_rq_clock_pelt(struct cfs_rq *cfs_rq) { }
> >  #endif
> >
> >
> > diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> > index e2cf6e48b165..07014e8cbae2 100644
> > --- a/kernel/sched/sched.h
> > +++ b/kernel/sched/sched.h
> > @@ -641,6 +641,10 @@ struct cfs_rq {
> >       int                     runtime_enabled;
> >       s64                     runtime_remaining;
> >
> > +     u64                     clock_pelt_idle;
> > +#ifndef CONFIG_64BIT
> > +     u64                     clock_pelt_idle_copy;
> > +#endif
> >       u64                     throttled_clock;
> >       u64                     throttled_clock_pelt;
> >       u64                     throttled_clock_pelt_time;
> > @@ -1013,6 +1017,12 @@ struct rq {
> >       u64                     clock_task ____cacheline_aligned;
> >       u64                     clock_pelt;
> >       unsigned long           lost_idle_time;
> > +     u64                     clock_pelt_idle;
> > +     u64                     enter_idle;
> > +#ifndef CONFIG_64BIT
> > +     u64                     clock_pelt_idle_copy;
> > +     u64                     enter_idle_copy;
> > +#endif
> >
> >       atomic_t                nr_iowait;
> >
> > --
> > 2.25.1
> >