Re: [PATCH 2/2] cpufreq: simplify cpufreq_set_policy() interface

[Zihuan Zhang]

在 2025/8/19 18:59, Viresh Kumar 写道:
> On 19-08-25, 18:39, Zihuan Zhang wrote:
> >   static int cpufreq_set_policy(struct cpufreq_policy *policy,
> > -			      struct cpufreq_governor *new_gov,
> > -			      unsigned int new_pol);
> > +			      struct cpufreq_governor *new_gov);
> A driver will either support the policy or the governor. If we are
> keeping `new_gov` around, I don't see why `new_pol` should be dropped.

Thanks for the reminder.

If we remove new_pol, then new_gov should indeed be removed as well.

> And changing the policy for a `setpolicy` driver should happen from
> within cpufreq_set_policy() instead of the caller. Also there is at
> least one case (verify()) where we may end up returning early, before
> changing the policy.
You’re right, we didn’t really consider that case before.

The interface of cpufreq_set_policy() does look a bit odd:

- drivers using governors don’t really need the new_pol parameter

- while drivers using the setpolicy method don’t need the new_gov one.


I guess this might be due to some historical reasons.

The question is whether it’s worth modifying this function, or if we 
should just keep it as it is.