01.09.2021 19:59, Ulf Hansson пишет:
>> +static int genpd_dev_initialize_performance_state(struct device *dev,
>> + struct device *base_dev,
>> + unsigned int index)
>> +{
>> + struct generic_pm_domain *pd = dev_to_genpd(dev);
>> + bool state_default = false, suspended = false;
>> + int pstate, err;
>> +
>> + pstate = genpd_dev_get_current_performance_state(dev, base_dev, index,
>> + &state_default,
>> + &suspended);
>> + if (pstate <= 0)
>> + return pstate;
>> +
>> + /*
>> + * If device is suspended, then performance state will be applied
>> + * on RPM-resume. This prevents potential extra power consumption
>> + * if device won't be resumed soon.
>> + *
>> + * If device is known to be active at the moment, then the performance
>> + * state should be updated immediately to sync state with hardware.
>> + */
>> + if (suspended) {
>> + dev_gpd_data(dev)->rpm_pstate = pstate;
>> + } else {
>> + err = dev_pm_genpd_set_performance_state(dev, pstate);
>> + if (err) {
>> + dev_err(dev, "failed to set performance state for power-domain %s: %d\n",
>> + pd->name, err);
>> + return err;
>> + }
>> +
>> + if (state_default)
>> + dev_gpd_data(dev)->default_pstate = pstate;
>> + }
>> +
>> + return 0;
>> +}
> As I kind of indicated in my previous reply on the earlier version, I
> think the above code can be made a lot less complicated. Although,
> perhaps I may be missing some points.
>
> Anyway, I will post a couple patches, later this evening or tomorrow,
> to show more exactly what I had in mind. Let's see if that can work
> for you.
It's not obvious what you're wanting to improve, I'm probably missing
yours point. So indeed will be good to see the code sample, thanks.
