Re: [PATCH] cpufreq: schedutil: Move max CPU capacity to sugov_policy

[Lukasz Luba <lukasz.luba@arm.com>]

Hi Rafael,

On 7/25/22 09:07, Lukasz Luba wrote:
> Hi Rafael,
> 
> On 7/15/22 18:29, Rafael J. Wysocki wrote:
>> On Fri, Jul 15, 2022 at 1:47 PM Lukasz Luba <lukasz.luba@arm.com> wrote:
>>>
>>>
>>>
>>> On 7/15/22 12:44, Rafael J. Wysocki wrote:
>>>> On Fri, Jul 15, 2022 at 10:47 AM Lukasz Luba <lukasz.luba@arm.com> 
>>>> wrote:
>>>>>
>>>>> Hi Rafael,
>>>>>
>>>>> gentle ping.
>>>>>
>>>>> On 7/11/22 13:42, Lukasz Luba wrote:
>>>>>> There is no need to keep the max CPU capacity in the per_cpu 
>>>>>> instance.
>>>>>> Furthermore, there is no need to check and update that variable
>>>>>> (sg_cpu->max) everytime in the frequency change request, which is 
>>>>>> part
>>>>>> of hot path. Instead use struct sugov_policy to store that 
>>>>>> information.
>>>>>> Initialize the max CPU capacity during the setup and start callback.
>>>>>> We can do that since all CPUs in the same frequency domain have 
>>>>>> the same
>>>>>> max capacity (capacity setup and thermal pressure are based on that).
>>>>>>
>>>>>> Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
>>>>>> ---
>>>>>>     kernel/sched/cpufreq_schedutil.c | 30 
>>>>>> +++++++++++++++---------------
>>>>>>     1 file changed, 15 insertions(+), 15 deletions(-)
>>>>>
>>>>> The patch got Ack from Viresh.
>>>>> Could you take it?
>>>>
>>>> Yes, it's there in my queue.  Same for the EM changes.
>>>
>>> Thank you Rafael!
>>
>> Well, the patch doesn't apply on top of 5.19-rc6, because
>> sugov_get_util() is somewhat different.
>>
>> Please rebase it and resend.
> 
> My apologies for the delay, I was on holidays.
> 
> I'll do that today and resend it.
> 

I have found the reason why it doesn't apply
on your tree. I have used next-20220711
to base this work on. It contains Peter's
branch sched/core, where there is this Dietmar's patch:

https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=sched/core&id=bb4479994945e9170534389a7762eb56149320ac

That causes the issue. I thing it might collide when I re-base my patch
on top of 5.19-rc6 and you apply it into pm tree...

What do you think about this?