Re: [RFC PATCH v15 4/7] sched: Fix runtime accounting w/ split exec & sched contexts

[John Stultz]

On Thu, Mar 13, 2025 at 3:26 AM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Wed, 12 Mar 2025 15:11:34 -0700
> John Stultz <jstultz@xxxxxxxxxx> wrote:
>
> > The idea here is we want to charge the scheduler-context task's
> > vruntime but charge the execution-context task's sum_exec_runtime.
>
> The "but" is confusing me. Do you mean, "and also"? The sentence
> doesn't make sense with "but" unless it was:
>
>   "The idea here is we DON'T want to charge the scheduler-context
>   task's vruntime but charge the execution-context task's
>   sum_exec_runtime INSTEAD."
>
> >
> > This way cputime accounting goes against the task actually running
> > but vruntime accounting goes against the rq->donor task so we get
> > proper fairness.
>
> But this shows that you want to do both, although, I would remove the
> "but" here too. Replace it with "while".
>
> Or maybe I'm just confused.

Apologies for beingconfusing. I know I can tangle my words sometimes. :)

Hrmm. I think maybe it's a bit more clear if I switch the order. ie:

Without proxy-exec, we normally charge the "current" task for both its
vruntime as well as its sum_exec_runtime.

With proxy, we want to charge the rq->curr (proxy/lock holder) time to
its sum_exec_runtime (so it's clear to userland the rq->curr task *is*
running).

*But*, instead of charging rq->curr for the vruntime, that is charged
against the rq->donor(lock waiter) task, because that is what it is
donating when it is used as the scheduler-context.

If the donor and curr tasks are the same, then it's the same as without proxy.

Your suggestion of "while" is good as well. I'll try to reword this.


> > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > index c798d27952431..f8ad3a44b3771 100644
> > --- a/kernel/sched/fair.c
> > +++ b/kernel/sched/fair.c
> > @@ -1129,22 +1129,33 @@ static void update_tg_load_avg(struct cfs_rq *cfs_rq)
> >  }
> >  #endif /* CONFIG_SMP */
> >
> > -static s64 update_curr_se(struct rq *rq, struct sched_entity *curr)
> > +static s64 update_curr_se(struct rq *rq, struct sched_entity *se)
>
> Should this be renamed to "update_se()" as it no longer appears to be
> updating "curr_se".
>
> >  {
> >       u64 now = rq_clock_task(rq);
> >       s64 delta_exec;
> >
> > -     delta_exec = now - curr->exec_start;
> > +     delta_exec = now - se->exec_start;
> >       if (unlikely(delta_exec <= 0))
> >               return delta_exec;
> >
> > -     curr->exec_start = now;
> > -     curr->sum_exec_runtime += delta_exec;
> > +     se->exec_start = now;
> > +     if (entity_is_task(se)) {
> > +             struct task_struct *running = rq->curr;
> > +             /*
> > +              * If se is a task, we account the time against the running
> > +              * task, as w/ proxy-exec they may not be the same.
> > +              */
> > +             running->se.exec_start = now;
> > +             running->se.sum_exec_runtime += delta_exec;
> > +     } else {
> > +             /* If not task, account the time against se */
> > +             se->sum_exec_runtime += delta_exec;
> > +     }
>
> Or maybe: update_proxy_se() ?

I'm hesitant to call it update_proxy_se() since it's still used even
when proxy-exec isn't enabled.

update_se() could work, but also feels a little generic.

Maybe update_se_times()? or account_time_se()?


Appreciate the review and feedback!

thanks
-john