Re: [PATCH v3 07/20] sched/fair: Select preferred CPU at wakeup when possible

[Shrikanth Hegde <sshegde@linux.ibm.com>]

On 6/4/26 10:44 AM, K Prateek Nayak wrote:
> Hello Shrikanth,
> 
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 29fbb5287cfc..a704285ac55a 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -7743,7 +7743,8 @@ static inline int sched_balance_find_dst_cpu(struct sched_domain *sd, struct tas
>>   {
>>   	int new_cpu = cpu;
>>   
>> -	if (!cpumask_intersects(sched_domain_span(sd), p->cpus_ptr))
>> +	if (!cpumask_intersects(sched_domain_span(sd), p->cpus_ptr) &&
> 
> Doesn't this mean we have no CPUs in this sd that fits the affinity
> requirements?
> 
>> +	    cpu_preferred(prev_cpu))
> 
> There is no point in going down the function if the affinity constraints
> aren't met since we are traversing the hierarchy top-down and none of
> the groups / child domains will satisfy the affinity requirements.
> 
> Since the CPU returned here is again checked for is_cpu_allowed(), we'll
> eventually hit the select_fallback_rq() path anyways so the check for
> cpu_preferred(prev_cpu) seems unnecessary unless I'm missing something.
> 

Ok. You are right, we call is_cpu_allowed() further. that will catch it.
I will keep only available_idle_cpu then.

>>   		return prev_cpu;
>>   
>>   	/*
>