Re: [External] Re: [PATCH v5 2/4] sched/core: Avoid double calling update_rq_clock() in __balance_push_cpu_stop()

[Hao Jia]

On 2023/6/15 Peter Zijlstra wrote:
> On Tue, Jun 13, 2023 at 04:20:10PM +0800, Hao Jia wrote:
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index a8be5415daba..1eca36299d8b 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -2398,7 +2398,6 @@ static struct rq *__migrate_task(struct rq *rq, struct rq_flags *rf,
> >   	if (!is_cpu_allowed(p, dest_cpu))
> >   		return rq;
> >   
> > -	update_rq_clock(rq);
> >   	rq = move_queued_task(rq, rf, p, dest_cpu);
> >   
> >   	return rq;
> > @@ -2456,10 +2455,12 @@ static int migration_cpu_stop(void *data)
> >   				goto out;
> >   		}
> >   
> > -		if (task_on_rq_queued(p))
> > +		if (task_on_rq_queued(p)) {
> > +			update_rq_clock(rq);
> >   			rq = __migrate_task(rq, &rf, p, arg->dest_cpu);
> > -		else
> > +		} else {
> >   			p->wake_cpu = arg->dest_cpu;
> > +		}
> >   
> >   		/*
> >   		 * XXX __migrate_task() can fail, at which point we might end
>
> So now you've got update_rq_clock() in both callers, why not remove it
> from __balance_push_cpu_stop() ?
>
> Afaict nothing actually needs it before __migrate_task().

Thanks for your review.
I'm afraid not, the rq clock also needs to be updated before
select_fallback_rq() is called.


> ---
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -9435,8 +9435,6 @@ static int __balance_push_cpu_stop(void
>   	raw_spin_lock_irq(&p->pi_lock);
>   	rq_lock(rq, &rf);
>   
> -	update_rq_clock(rq);
> -
>   	if (task_rq(p) == rq && task_on_rq_queued(p)) {
>   		cpu = select_fallback_rq(rq->cpu, p);
>   		rq = __migrate_task(rq, &rf, p, cpu);

If we just remove update_rq_clock() from __balance_push_cpu_stop(), we 
will get this warning.


[ 1260.960166] rq->clock_update_flags < RQCF_ACT_SKIP
[ 1260.960170] WARNING: CPU: 25 PID: 196 at kernel/sched/sched.h:1496 
update_curr+0xf6/0x1f0

[ 1260.960318] Call Trace:
[ 1260.960320]  <TASK>
[ 1260.960323]  ? show_regs+0x5b/0x60
[ 1260.960327]  ? __warn+0x89/0x140
[ 1260.960331]  ? update_curr+0xf6/0x1f0
[ 1260.960334]  ? report_bug+0x1b7/0x1e0
[ 1260.960338]  ? __wake_up_klogd.part.25+0x5a/0x80
[ 1260.960342]  ? handle_bug+0x45/0x80
[ 1260.960346]  ? exc_invalid_op+0x18/0x70
[ 1260.960348]  ? asm_exc_invalid_op+0x1b/0x20
[ 1260.960354]  ? update_curr+0xf6/0x1f0
[ 1260.960356]  ? update_curr+0xf6/0x1f0
[ 1260.960359]  dequeue_entity+0x3b/0x410
[ 1260.960361]  dequeue_task_fair+0xc7/0x3c0
[ 1260.960363]  dequeue_task+0x30/0xf0
[ 1260.960365]  __do_set_cpus_allowed+0x94/0x130
[ 1260.960366]  do_set_cpus_allowed+0x38/0x60
[ 1260.960368]  cpuset_cpus_allowed_fallback+0x70/0x80
[ 1260.960372]  select_fallback_rq+0x20f/0x250            <----
[ 1260.960374]  __balance_push_cpu_stop+0x13f/0x1a0
[ 1260.960377]  ? migration_cpu_stop+0x2b0/0x2b0
[ 1260.960379]  cpu_stopper_thread+0xaf/0x140
[ 1260.960382]  smpboot_thread_fn+0x158/0x220
[ 1260.960387]  ? sort_range+0x30/0x30
[ 1260.960390]  kthread+0xfe/0x130
[ 1260.960392]  ? kthread_complete_and_exit+0x20/0x20
[ 1260.960394]  ret_from_fork+0x1f/0x30
[ 1260.960399]  </TASK>

Thanks,
Hao