Re: [PATCH 1/2] cpufreq,arm,arm64: restructure definitions of arch_set_freq_scale()

[Ionela Voinescu]

Hi Viresh,

On Tuesday 06 Oct 2020 at 12:35:51 (+0530), Viresh Kumar wrote:
> On 24-09-20, 13:30, Ionela Voinescu wrote:
> > diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> > index 2ea245a6c0c0..f34d3a3d5ba6 100644
> > --- a/drivers/cpufreq/cpufreq.c
> > +++ b/drivers/cpufreq/cpufreq.c
> > @@ -160,13 +160,6 @@ u64 get_cpu_idle_time(unsigned int cpu, u64 *wall, int io_busy)
> >  }
> >  EXPORT_SYMBOL_GPL(get_cpu_idle_time);
> >  
> > -__weak void arch_set_freq_scale(const struct cpumask *cpus,
> > -				unsigned long cur_freq,
> > -				unsigned long max_freq)
> > -{
> > -}
> > -EXPORT_SYMBOL_GPL(arch_set_freq_scale);
> > -
> 
> Why can't we have this anymore ? Because it is a macro now instead of a routine
> for ARM ?

Yes, instead of having a strong counterpart in arch_topology.c to be used
for both arm64 and arm, I separated the definitions of the
arch_set_freq_scale symbol in each architecture's topology.h in order to
be filtered if needed (as is the case for arm's BL_SWITCHER).

I could have kept either this weak stub definition or the stub that is
now in cpufreq.h under "#ifndef arch_set_freq_scale", but I chose the
latter as it's more aligned with the other architectural functions
referenced in the commit message.

So this patch has both the purpose of enabling a nicer filtering of
frequency invariance for BL_SWITCHER but also follows the definition
pattern of the other functions so we don't keep having two mindsets when
working with them.

Thanks,
Ionela.