Re: [PATCH] cpufreq: intel_pstate: Synchronize sysfs limits

[Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>]

On Mon, 2016-11-14 at 02:04 +0100, Rafael J. Wysocki wrote:
> On Sat, Nov 12, 2016 at 1:11 AM, Srinivas Pandruvada
> 
[...]
> +       get_online_cpus();
> > +       for_each_online_cpu(cpu) {
> > +               if (all_cpu_data[cpu])
> > +                       cpufreq_update_policy(cpu);
> 
> cpufreq_update_policy() calls cpufreq_cpu_get() to get the policy
> anyway which does the requisite policy existence check (although it
> is
> a bit racy now, but that's a bug in there that we should not have to
> work around here), so it should be sufficient to do this
> for_each_possible_cpu() without additional locking.
> 
I will change in the next patch set.

> > 
[...]

> > +       cpufreq_register_notifier(&intel_pstate_cpufreq_notifier_bl
> > ock,
> > +                                 CPUFREQ_POLICY_NOTIFIER);
> > +
> 
> cpufreq_set_policy() will call our ->verify() and ->set_policy()
> things, so why do we need the notifier?
> 
I was simply replicating what is done for _PPC notifiers. But we can do
in verify() callback here. In the next patch, I will change this.

Thanks,
Srinivas