Re: [RFC PATCH 2/6] sched/dl: Capacity-aware migrations

[Dietmar Eggemann]

On 7/8/19 9:41 AM, luca abeni wrote:
> Hi Dietmar,
> 
> On Thu, 4 Jul 2019 14:05:22 +0200
> Dietmar Eggemann <dietmar.eggemann@xxxxxxx> wrote:
> 
>> On 5/6/19 6:48 AM, Luca Abeni wrote:
>>
>> [...]
>>
>>> diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
>>> index 5b981eeeb944..3436f3d8fa8f 100644
>>> --- a/kernel/sched/deadline.c
>>> +++ b/kernel/sched/deadline.c
>>> @@ -1584,6 +1584,9 @@ select_task_rq_dl(struct task_struct *p, int
>>> cpu, int sd_flag, int flags) if (sd_flag != SD_BALANCE_WAKE)
>>>  		goto out;
>>>  
>>> +	if (dl_entity_is_special(&p->dl))
>>> +		goto out;  
>>
>> I wonder if this is really required. The if condition
>>
>> 1591         if (unlikely(dl_task(curr)) &&
>> 1592             (curr->nr_cpus_allowed < 2 ||
>> 1593              !dl_entity_preempt(&p->dl, &curr->dl)) &&
>> 1594             (p->nr_cpus_allowed > 1)) {
>>
>> further below uses '!dl_entity_preempt(&p->dl, &curr->dl))' which
>> returns 'dl_entity_is_special(a) || ...'
> 
> Uhm... I do not remember the details; I remember that the check fixed
> something during the development of the patchset, but I did not check
> if it was still needed when I forward-ported the patches...
> 
> So, maybe it worked around some bugs in previous versions of the
> kernel, but is not needed now.

I figured it out in the meantime ... you added a ' ... ||
static_branch_unlikely(&sched_asym_cpucapacity)' in this patch to this
if condition which let's special dl tasks (like sugov:X) enter this
condition e.g. on big.LITTLE.