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

Message ID 161907696062.29796.108437696048031441.tip-bot2@tip-bot2
State New, archived
Headers show
Series
  • [tip:,sched/core] sched,fair: skip newidle_balance if a wakeup is pending
Related show

Commit Message

tip-bot2 for Peter Zijlstra April 22, 2021, 7:36 a.m. UTC
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(-)

Comments

Peter Zijlstra April 22, 2021, 8:15 a.m. UTC | #1
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 :)

Patch
diff mbox series

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);