Re: [Question]: about 'cpuinfo_cur_freq' shown in sysfs when the CPU is in idle state

[Sudeep Holla]

On Wed, Jun 03, 2020 at 03:51:59PM +0530, Viresh Kumar wrote:
> On 03-06-20, 11:17, Sudeep Holla wrote:
> > On Wed, Jun 03, 2020 at 03:40:10PM +0530, Viresh Kumar wrote:
> > > On 03-06-20, 11:07, Sudeep Holla wrote:
> > > > But I have another question. If we can detect that CPPC on some platforms
> > > > rely on CPU registers(I assume FFH registers here and not system/io/...
> > > > type of GAS registers), can we set dvfs_on_any_cpu(can't recall exact
> > > > flag name) to false if not already done to prevent such issues. Or I am
> > > > talking non-sense as it may be applicable only for _set operation and
> > >
> > >           Yes, non-sense :)
> > >
> >
> > Thanks for confirming ð.
>
> Hehe.
>
> So, do you agree that we better do the read from the CPUs themselves ?
>

Yes if that is fine. I thought waking up the core was not a good solution
in terms of power, but I have no objection for that as well as return 0
as IMO it aligns with the AMU counters when CPU is idle.

--
Regards,
Sudeep