Re: [PATCH 1/4] PM/Domains: Add support for retrieving genpd performance states information

[Ulf Hansson <ulf.hansson@linaro.org>]

On Sat, 7 Sep 2019 at 00:24, Thara Gopinath <thara.gopinath@linaro.org> wrote:
>
> On 08/22/2019 11:03 AM, Ulf Hansson wrote:
> > On Sat, 10 Aug 2019 at 02:58, Thara Gopinath <thara.gopinath@linaro.org> wrote:
> >>
> >> Add two new APIs in the genpd framework,
> >> dev_pm_genpd_get_performance_state to return the current performance
> >> state of a power domain and dev_pm_genpd_performance_state_count to
> >> return the total number of performance states supported by a
> >> power domain. Since the genpd framework does not maintain
> >> a count of number of performance states supported by a power domain,
> >> introduce a new callback(.get_performance_state_count) that can be used
> >> to retrieve this information from power domain drivers.
> >
> > I think some brief background to *why* this is useful needs to be
> > squeezed into the changelog. Or at least state that following changes
> > makes use of it, somehow.
> >
> >>
> >> Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>
> >> ---
> >>  drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++
> >>  include/linux/pm_domain.h   | 18 ++++++++++++++++++
> >>  2 files changed, 56 insertions(+)
> >>
> >> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> >> index b063bc4..17e0375 100644
> >> --- a/drivers/base/power/domain.c
> >> +++ b/drivers/base/power/domain.c
> >> @@ -413,6 +413,44 @@ int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)
> >>  }
> >>  EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);
> >>
> >> +int dev_pm_genpd_get_performance_state(struct device *dev,
> >> +                                      unsigned int *state)
> >> +{
> >> +       struct generic_pm_domain *genpd;
> >> +
> >> +       genpd = dev_to_genpd(dev);
> >
> > We need to verify that the there is a genpd attached before doing this
> > cast. Let me post a patch in a day or so, it will give you a helper
> > function that covers this.
> >
> >> +       if (IS_ERR(genpd))
> >> +               return -ENODEV;
> >> +
> >> +       genpd_lock(genpd);
> >> +       *state = genpd->performance_state;
> >
> > Why not return the state, rather than assigning an out-parameter?
> >
> >> +       genpd_unlock(genpd);
> >> +
> >> +       return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);
> >> +
> >> +int dev_pm_genpd_performance_state_count(struct device *dev,
> >> +                                        unsigned int *count)
> >> +{
> >> +       struct generic_pm_domain *genpd;
> >> +       int ret;
> >> +
> >> +       genpd = dev_to_genpd(dev);
> >> +       if (IS_ERR(genpd))
> >> +               return -ENODEV;
> >> +
> >> +       if (unlikely(!genpd->get_performance_state_count))
> >> +               return -EINVAL;
> >> +
> >> +       genpd_lock(genpd);
> >> +       ret = genpd->get_performance_state_count(genpd, count);
> >
> > Why not having the callback to return the state, rather than using an
> > out-parameter?
> Hi Ulf,
> I just realized that returning the state instead of using a parameter
> will prevent me from access under lock. Is that okay ?

Not sure I understand. Why can't you just assign a local variable and
return that?

Like this:

genpd_lock();
count = genpd->get_performance_state_count();
genpd_unlock();

return count;

[...]

Kind regards
Uffe