[PATCH v3] sched,fair: skip newidle_balance if a wakeup is pending

[PATCH v3] sched,fair: skip newidle_balance if a wakeup is pending

[Rik van Riel]

The try_to_wake_up function has an optimization where it can queue
a task for wakeup on its previous CPU, if the task is still in the
middle of going to sleep inside schedule().

Once schedule() re-enables IRQs, the task will be woken up with an
IPI, and placed back on the runqueue.

If we have such a wakeup pending, there is no need to search other
CPUs for runnable tasks. Just skip (or bail out early from) newidle
balancing, and run the just woken up task.

For a memcache like workload test, this reduces total CPU use by
about 2%, proportionally split between user and system time,
and p99 and p95 application response time by 10% on average.
The schedstats run_delay number shows a similar improvement.

Signed-off-by: Rik van Riel <riel@surriel.com>
---
 kernel/sched/fair.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 69680158963f..fd80175c3b3e 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -10594,6 +10594,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 	u64 curr_cost = 0;
 
 	update_misfit_status(NULL, this_rq);
+
+	/*
+	 * There is a task waiting to run. No need to search for one.
+	 * Return 0; the task will be enqueued when switching to idle.
+	 */
+	if (this_rq->ttwu_pending)
+		return 0;
+
 	/*
 	 * We must set idle_stamp _before_ calling idle_balance(), such that we
 	 * measure the duration of idle_balance() as idle time.
@@ -10661,7 +10669,8 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 		 * Stop searching for tasks to pull if there are
 		 * now runnable tasks on this rq.
 		 */
-		if (pulled_task || this_rq->nr_running > 0)
+		if (pulled_task || this_rq->nr_running > 0 ||
+		    this_rq->ttwu_pending)
 			break;
 	}
 	rcu_read_unlock();
@@ -10688,7 +10697,12 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 	if (this_rq->nr_running != this_rq->cfs.h_nr_running)
 		pulled_task = -1;
 
-	if (pulled_task)
+	/*
+	 * If we are no longer idle, do not let the time spent here pull
+	 * down this_rq->avg_idle. That could lead to newidle_balance not
+	 * doing enough work, and the CPU actually going idle.
+	 */
+	if (pulled_task || this_rq->ttwu_pending)
 		this_rq->idle_stamp = 0;
 
 	rq_repin_lock(this_rq, rf);
-- 
2.25.4

Re: [PATCH v3] sched,fair: skip newidle_balance if a wakeup is pending

[Vincent Guittot]

Hi Rik,

On Tue, 20 Apr 2021 at 18:07, Rik van Riel <riel@surriel.com> wrote:
>
> The try_to_wake_up function has an optimization where it can queue
> a task for wakeup on its previous CPU, if the task is still in the
> middle of going to sleep inside schedule().
>
> Once schedule() re-enables IRQs, the task will be woken up with an
> IPI, and placed back on the runqueue.
>
> If we have such a wakeup pending, there is no need to search other
> CPUs for runnable tasks. Just skip (or bail out early from) newidle
> balancing, and run the just woken up task.
>
> For a memcache like workload test, this reduces total CPU use by
> about 2%, proportionally split between user and system time,
> and p99 and p95 application response time by 10% on average.
> The schedstats run_delay number shows a similar improvement.
>
> Signed-off-by: Rik van Riel <riel@surriel.com>
> ---
>  kernel/sched/fair.c | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 69680158963f..fd80175c3b3e 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -10594,6 +10594,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
>         u64 curr_cost = 0;
>
>         update_misfit_status(NULL, this_rq);
> +
> +       /*
> +        * There is a task waiting to run. No need to search for one.
> +        * Return 0; the task will be enqueued when switching to idle.
> +        */
> +       if (this_rq->ttwu_pending)
> +               return 0;
> +
>         /*
>          * We must set idle_stamp _before_ calling idle_balance(), such that we
>          * measure the duration of idle_balance() as idle time.
> @@ -10661,7 +10669,8 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
>                  * Stop searching for tasks to pull if there are
>                  * now runnable tasks on this rq.
>                  */
> -               if (pulled_task || this_rq->nr_running > 0)
> +               if (pulled_task || this_rq->nr_running > 0 ||
> +                   this_rq->ttwu_pending)
>                         break;
>         }
>         rcu_read_unlock();
> @@ -10688,7 +10697,12 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
>         if (this_rq->nr_running != this_rq->cfs.h_nr_running)
>                 pulled_task = -1;
>
> -       if (pulled_task)
> +       /*
> +        * If we are no longer idle, do not let the time spent here pull
> +        * down this_rq->avg_idle. That could lead to newidle_balance not
> +        * doing enough work, and the CPU actually going idle.
> +        */
> +       if (pulled_task || this_rq->ttwu_pending)

I'm still running some benchmarks to evaluate the impact of your patch
and more especially the line above which clears this_rq->idle_stamp
and skips the time spent in newidle_balance from being accounted for
in avg_idle. I have some results which  show some regression because
of this test especially with hackbench.
On large system, the time spent in newidle_balance can be significant
and we can't ignore it just because this_rq->ttwu_pending is set while
looping the domains because without newidle_balance the idle time
would have been large and we end up screwing up the metric

>                 this_rq->idle_stamp = 0;
>
>         rq_repin_lock(this_rq, rf);
> --
> 2.25.4
>
>

[tip: sched/core] sched,fair: skip newidle_balance if a wakeup is pending

[tip-bot2 for Rik van Riel]

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     9c9f520a14670ad59da2f700660f7601ec9e0b07
Gitweb:        https://git.kernel.org/tip/9c9f520a14670ad59da2f700660f7601ec9e0b07
Author:        Rik van Riel <riel@surriel.com>
AuthorDate:    Tue, 20 Apr 2021 12:07:05 -04:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 21 Apr 2021 13:55:43 +02:00

sched,fair: skip newidle_balance if a wakeup is pending

The try_to_wake_up function has an optimization where it can queue
a task for wakeup on its previous CPU, if the task is still in the
middle of going to sleep inside schedule().

Once schedule() re-enables IRQs, the task will be woken up with an
IPI, and placed back on the runqueue.

If we have such a wakeup pending, there is no need to search other
CPUs for runnable tasks. Just skip (or bail out early from) newidle
balancing, and run the just woken up task.

For a memcache like workload test, this reduces total CPU use by
about 2%, proportionally split between user and system time,
and p99 and p95 application response time by 10% on average.
The schedstats run_delay number shows a similar improvement.

Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20210420120705.5c705d4b@imladris.surriel.com
---
 kernel/sched/fair.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 1d75af1..83cd2bd 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -10592,6 +10592,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 	u64 curr_cost = 0;
 
 	update_misfit_status(NULL, this_rq);
+
+	/*
+	 * There is a task waiting to run. No need to search for one.
+	 * Return 0; the task will be enqueued when switching to idle.
+	 */
+	if (this_rq->ttwu_pending)
+		return 0;
+
 	/*
 	 * We must set idle_stamp _before_ calling idle_balance(), such that we
 	 * measure the duration of idle_balance() as idle time.
@@ -10657,7 +10665,8 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 		 * Stop searching for tasks to pull if there are
 		 * now runnable tasks on this rq.
 		 */
-		if (pulled_task || this_rq->nr_running > 0)
+		if (pulled_task || this_rq->nr_running > 0 ||
+		    this_rq->ttwu_pending)
 			break;
 	}
 	rcu_read_unlock();
@@ -10684,7 +10693,12 @@ out:
 	if (time_after(this_rq->next_balance, next_balance))
 		this_rq->next_balance = next_balance;
 
-	if (pulled_task)
+	/*
+	 * If we are no longer idle, do not let the time spent here pull
+	 * down this_rq->avg_idle. That could lead to newidle_balance not
+	 * doing enough work, and the CPU actually going idle.
+	 */
+	if (pulled_task || this_rq->ttwu_pending)
 		this_rq->idle_stamp = 0;
 	else
 		nohz_newidle_balance(this_rq);

Re: [tip: sched/core] sched,fair: skip newidle_balance if a wakeup is pending

[Peter Zijlstra]

On Thu, Apr 22, 2021 at 07:36:00AM -0000, tip-bot2 for Rik van Riel wrote:
> @@ -10684,7 +10693,12 @@ out:
>  	if (time_after(this_rq->next_balance, next_balance))
>  		this_rq->next_balance = next_balance;
>  
> -	if (pulled_task)
> +	/*
> +	 * If we are no longer idle, do not let the time spent here pull
> +	 * down this_rq->avg_idle. That could lead to newidle_balance not
> +	 * doing enough work, and the CPU actually going idle.
> +	 */
> +	if (pulled_task || this_rq->ttwu_pending)
>  		this_rq->idle_stamp = 0;

I've un-committed this patch, because vingu was reporting increased idle
time because of this hunk.  I had mistakenly assumed that was sorted
with v3, sorry for not keeping better track of things.

(also, now that I look again, please also fix the Subject to have a
capital after the :)

Re: [PATCH v3] sched,fair: skip newidle_balance if a wakeup is pending

[Vincent Guittot]

On Wed, 21 Apr 2021 at 19:27, Vincent Guittot
<vincent.guittot@linaro.org> wrote:
>
> Hi Rik,
>
> On Tue, 20 Apr 2021 at 18:07, Rik van Riel <riel@surriel.com> wrote:
> >
> > The try_to_wake_up function has an optimization where it can queue
> > a task for wakeup on its previous CPU, if the task is still in the
> > middle of going to sleep inside schedule().
> >
> > Once schedule() re-enables IRQs, the task will be woken up with an
> > IPI, and placed back on the runqueue.
> >
> > If we have such a wakeup pending, there is no need to search other
> > CPUs for runnable tasks. Just skip (or bail out early from) newidle
> > balancing, and run the just woken up task.
> >
> > For a memcache like workload test, this reduces total CPU use by
> > about 2%, proportionally split between user and system time,
> > and p99 and p95 application response time by 10% on average.
> > The schedstats run_delay number shows a similar improvement.
> >
> > Signed-off-by: Rik van Riel <riel@surriel.com>
> > ---
> >  kernel/sched/fair.c | 18 ++++++++++++++++--
> >  1 file changed, 16 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index 69680158963f..fd80175c3b3e 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -10594,6 +10594,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> >         u64 curr_cost = 0;
> >
> >         update_misfit_status(NULL, this_rq);
> > +
> > +       /*
> > +        * There is a task waiting to run. No need to search for one.
> > +        * Return 0; the task will be enqueued when switching to idle.
> > +        */
> > +       if (this_rq->ttwu_pending)
> > +               return 0;
> > +
> >         /*
> >          * We must set idle_stamp _before_ calling idle_balance(), such that we
> >          * measure the duration of idle_balance() as idle time.
> > @@ -10661,7 +10669,8 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> >                  * Stop searching for tasks to pull if there are
> >                  * now runnable tasks on this rq.
> >                  */
> > -               if (pulled_task || this_rq->nr_running > 0)
> > +               if (pulled_task || this_rq->nr_running > 0 ||
> > +                   this_rq->ttwu_pending)
> >                         break;
> >         }
> >         rcu_read_unlock();
> > @@ -10688,7 +10697,12 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> >         if (this_rq->nr_running != this_rq->cfs.h_nr_running)
> >                 pulled_task = -1;
> >
> > -       if (pulled_task)
> > +       /*
> > +        * If we are no longer idle, do not let the time spent here pull
> > +        * down this_rq->avg_idle. That could lead to newidle_balance not
> > +        * doing enough work, and the CPU actually going idle.
> > +        */
> > +       if (pulled_task || this_rq->ttwu_pending)
>
> I'm still running some benchmarks to evaluate the impact of your patch
> and more especially the line above which clears this_rq->idle_stamp
> and skips the time spent in newidle_balance from being accounted for
> in avg_idle. I have some results which  show some regression because
> of this test especially with hackbench.
> On large system, the time spent in newidle_balance can be significant
> and we can't ignore it just because this_rq->ttwu_pending is set while
> looping the domains because without newidle_balance the idle time
> would have been large and we end up screwing up the metric

I confirmed that the line above generate hackbench regression on my
large arm64 system (2 * 112 CPUs)
I'm testing hackbench with various number of group : 1, 2, 4, 16, 32,
64, 128, 256 but I have only put the 2 results which significantly
regress. The other ones are in the +/-1% variation range

hackbench -g $group

group    v5.12-rc8+tip    w/ this patch          w/ this patch without
the line above
64       2.862(+/- 9%)    2.952(+/-11%) -3%      2.807(+/- 7%) +2%
128      3.334(+/-10%)    3.561-+/-13%) -7%      3.181(+/- 6%) +4%



>
> >                 this_rq->idle_stamp = 0;
> >
> >         rq_repin_lock(this_rq, rf);
> > --
> > 2.25.4
> >
> >

Re: [PATCH v3] sched,fair: skip newidle_balance if a wakeup is pending

[Rik van Riel]

[-- Attachment #1: Type: text/plain, Size: 1286 bytes --]

On Thu, 2021-04-22 at 10:37 +0200, Vincent Guittot wrote:
> On Wed, 21 Apr 2021 at 19:27, Vincent Guittot
> <vincent.guittot@linaro.org> wrote:
> > 
> > > -       if (pulled_task)
> > > +       /*
> > > +        * If we are no longer idle, do not let the time spent
> > > here pull
> > > +        * down this_rq->avg_idle. That could lead to
> > > newidle_balance not
> > > +        * doing enough work, and the CPU actually going idle.
> > > +        */
> > > +       if (pulled_task || this_rq->ttwu_pending)
> > 
> I confirmed that the line above generate hackbench regression on my
> large arm64 system (2 * 112 CPUs)
> I'm testing hackbench with various number of group : 1, 2, 4, 16, 32,
> 64, 128, 256 but I have only put the 2 results which significantly
> regress. The other ones are in the +/-1% variation range
> 
> hackbench -g $group
> 
> group    v5.12-rc8+tip    w/ this patch          w/ this patch
> without
> the line above
> 64       2.862(+/- 9%)    2.952(+/-11%) -3%      2.807(+/- 7%) +2%
> 128      3.334(+/-10%)    3.561-+/-13%) -7%      3.181(+/- 6%) +4%

OK, I guess this part of the patch needs additional work.

I'll send a v4 with just the first two changes.

Thank you for running those tests.

-- 
All Rights Reversed.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[tip: sched/core] sched,fair: Skip newidle_balance if a wakeup is pending

[tip-bot2 for Rik van Riel]

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     e5e678e4fea26d73444f4427cbbaeab4fa79ecee
Gitweb:        https://git.kernel.org/tip/e5e678e4fea26d73444f4427cbbaeab4fa79ecee
Author:        Rik van Riel <riel@surriel.com>
AuthorDate:    Thu, 22 Apr 2021 13:02:36 -04:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Wed, 12 May 2021 11:43:23 +02:00

sched,fair: Skip newidle_balance if a wakeup is pending

The try_to_wake_up function has an optimization where it can queue
a task for wakeup on its previous CPU, if the task is still in the
middle of going to sleep inside schedule().

Once schedule() re-enables IRQs, the task will be woken up with an
IPI, and placed back on the runqueue.

If we have such a wakeup pending, there is no need to search other
CPUs for runnable tasks. Just skip (or bail out early from) newidle
balancing, and run the just woken up task.

For a memcache like workload test, this reduces total CPU use by
about 2%, proportionally split between user and system time,
and p99 and p95 application response time by 10% on average.
The schedstats run_delay number shows a similar improvement.

Signed-off-by: Rik van Riel <riel@surriel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Link: https://lkml.kernel.org/r/20210422130236.0bb353df@imladris.surriel.com
---
 kernel/sched/fair.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 3248e24..d10c6cc 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -10592,6 +10592,14 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 	u64 curr_cost = 0;
 
 	update_misfit_status(NULL, this_rq);
+
+	/*
+	 * There is a task waiting to run. No need to search for one.
+	 * Return 0; the task will be enqueued when switching to idle.
+	 */
+	if (this_rq->ttwu_pending)
+		return 0;
+
 	/*
 	 * We must set idle_stamp _before_ calling idle_balance(), such that we
 	 * measure the duration of idle_balance() as idle time.
@@ -10657,7 +10665,8 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 		 * Stop searching for tasks to pull if there are
 		 * now runnable tasks on this rq.
 		 */
-		if (pulled_task || this_rq->nr_running > 0)
+		if (pulled_task || this_rq->nr_running > 0 ||
+		    this_rq->ttwu_pending)
 			break;
 	}
 	rcu_read_unlock();