Re: [PATCH v2 1/3] cpufreq: CPPC: Keep the policy across CPU hotplug

[Sumit Gupta <sumitg@nvidia.com>]

On 22/07/26 15:48, Christian Loehle wrote:
> External email: Use caution opening links or attachments
>
>
> On 7/16/26 16:38, Sumit Gupta wrote:
>> Without online()/offline() callbacks, the cpufreq core fully tears
>> down a policy during exit() when its last online CPU is offlined,
>> and rebuilds it during init() when it comes back.
>>
>> Add lightweight online()/offline() callbacks so the core instead
>> keeps the policy live and reuses the driver's cpu_data across
>> CPU hotplug. This avoids re-reading the CPPC capabilities on every
>> offline/online, making CPU hotplug faster.
>>
>> Re-enable CPPC from online(), as it may have been disabled while
>> the CPU was offline.
>>
>> Signed-off-by: Sumit Gupta <sumitg@nvidia.com>
>> ---
>>   drivers/cpufreq/cppc_cpufreq.c | 27 +++++++++++++++++++++++++++
>>   1 file changed, 27 insertions(+)
>>
>> diff --git a/drivers/cpufreq/cppc_cpufreq.c b/drivers/cpufreq/cppc_cpufreq.c
>> index f6cea0c54dd9..432c6a6288a7 100644
>> --- a/drivers/cpufreq/cppc_cpufreq.c
>> +++ b/drivers/cpufreq/cppc_cpufreq.c
>> @@ -722,6 +722,31 @@ static int cppc_cpufreq_cpu_init(struct cpufreq_policy *policy)
>>        return ret;
>>   }
>>
>> +/*
>> + * With offline() defined, the cpufreq core keeps the policy alive when
>> + * a CPU is hotplugged out.
>> + */
>> +static int cppc_cpufreq_cpu_offline(struct cpufreq_policy *policy)
>> +{
>> +     return 0;
>> +}
>> +
>> +/*
>> + * Re-enable CPPC when the policy's CPU comes back online, since the platform
>> + * may have disabled it while the CPU was offline.
>> + */
>> +static int cppc_cpufreq_cpu_online(struct cpufreq_policy *policy)
>> +{
>> +     unsigned int cpu = policy->cpu;
>> +     int ret;
>> +
>> +     ret = cppc_set_enable(cpu, true);
>> +     if (ret && ret != -EOPNOTSUPP)
>> +             pr_warn("Failed to re-enable CPPC for CPU%d (%d)\n", cpu, ret);
> Keeping the policy alive means cpu_data->perf_ctrls survives hotplug, but the callback does not
> reapply it. A governor restart does not guarantee an immediate target callback, so the controls
> may remain reset indefinitely.
>
> I think the online path needs to restore a complete, current performance-control request before
> returning. MIN/MAX should be recomputed from the policy and the last requested DESIRED value
> clamped into that range.

Agreed.
Will change online() to recompute MIN/MAX from the policy and clamp the
retained DESIRED into that range, as described in my reply on patch 3/3.

>
> Also, what about suspend/resume?
>

Good point. On the platform I test on these registers are retained
across both hotplug and suspend, so I left suspend/resume out initially.
But the same mechanism extends to platforms that do reset them.
So I will add it as its own patch in v3 where suspend() saves the
OSPM set values and resume() restores them, reusing the online() path.

Thanks,
Sumit

....