Re: [PATCH v23 5/9] sched: Add logic to zap balance callbacks if we pick again

[K Prateek Nayak]

Hello John,

On 10/31/2025 8:45 AM, John Stultz wrote:
>>> -             if (next == rq->idle)
>>> +             }
>>> +             if (next == rq->idle) {
>>> +                     zap_balance_callbacks(rq);
>>
>> Also I would have preferred to have that zap_balance_callbacks() in
>> proxy_resched_idle() but this is okay too.
> 
> So my initial hesitation here is just we call proxy_resched_idle() in
> other situations where we might return NULL from find_proxy_task() as
> well. So this avoids calling zap_balance_callbacks() twice.
> 
> But thinking some more, later in the full series we often call
> proxy_resched_idle() in those paths where we are briefly dropping the
> rq lock and we often call zap_balance_callbacks as well there. I'll
> take a closer look at the full patch series and see if that doesn't
> make sense to consolidate then.  Not 100% sure it will work out, but
> worth looking into.

I don't have any strong feelings btw. What you have currently works
well. thank you for the additional background. I should go back and
take a look at the full tree again to get a full picture.

-- 
Thanks and Regards,
Prateek