Skip to contentSkip to conversation

sched: don't check rq after newidle_balance return positive

PATCH In review [email protected] 3 participants 3 messages Dec 15, 2020 – Dec 16, 2020 Scheduler

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sched: don't check rq after newidle_balance return positive

From: Chen Xiaoguang <[email protected]>

In pick_next_task_fair, if CPU is going to idle newidle_balance is called first trying to pull some tasks.
When newidle_balance returns positive which means it does pulls tasks or some tasks enqueued then there is no need to check sched_fair_runnable again.

---
 kernel/sched/fair.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
kernel/sched/fair.c +1 -2
index ae7ceba..c2f7eac 100644
@@ -7004,10 +7004,9 @@ struct task_struct *
 	struct task_struct *p;
 	int new_tasks;
 
-again:
 	if (!sched_fair_runnable(rq))
 		goto idle;
-
+again:
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (!prev || prev->sched_class != &fair_sched_class)
 		goto simple;
Signature · 1 line
1.8.3.1
maintainer
#1 lore raw
Quoted 7 lines · On Tue, Dec 15, 2020 at 02:48:50PM +0800, [email protected] wrote:

From: Chen Xiaoguang <[email protected]>

In pick_next_task_fair, if CPU is going to idle newidle_balance is called first trying to pull some tasks.
When newidle_balance returns positive which means it does pulls tasks or some tasks enqueued then there is no need to check sched_fair_runnable again.

No, I think it actually does need to, because while it counts the number of tasks it pulled, it doesn't verify it still has them after it re-acquires rq->lock. That is, someone could've come along and stolen them right from under our noses.

Quoted 3 lines

This SoB chain is broken. The first SoB should be the author, but From does not match.

#2 lore raw

答复: [PATCH] sched: don't check rq after newidle_balance return positive(Internet mail)

-----邮件原件-----
发件人: Peter Zijlstra <[email protected]>
发送时间: 2020年12月15日 16:33
收件人: [email protected]
抄送: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; heddchen(陈贺) <[email protected]>; xiaoggchen(陈小光) <[email protected]> 主题: Re: [PATCH] sched: don't check rq after newidle_balance return positive(Internet mail)

Quoted 10 lines · On Tue, Dec 15, 2020 at 02:48:50PM +0800, [email protected] wrote:
Quoted 7 lines

From: Chen Xiaoguang <[email protected]>

In pick_next_task_fair, if CPU is going to idle newidle_balance is called first trying to pull some tasks.
When newidle_balance returns positive which means it does pulls tasks or some tasks enqueued then there is no need to check sched_fair_runnable again.

No, I think it actually does need to, because while it counts the number of tasks it pulled, it doesn't verify it still has them after it re-acquires rq->lock. That is, someone could've come along and stolen them right from under our noses.

Ah, yes, our change only make sense when pulling nothing in load_balance but some tasks enqueued this rq during the lock of this rq is released.

Thanks.

Quoted 5 lines
Quoted 3 lines

This SoB chain is broken. The first SoB should be the author, but From does not match.

We will fix this next time.