Re: [PATCH 2/2] PM / devfreq: Use exclusively PM QoS to determine frequency limits

[Matthias Kaehlcke]

On Wed, Jan 15, 2020 at 02:35:48AM +0900, Chanwoo Choi wrote:
> On Wed, Jan 15, 2020 at 1:08 AM Leonard Crestez <leonard.crestez@xxxxxxx> wrote:
> >
> > On 13.01.2020 09:24, Chanwoo Choi wrote:
> > > Hi,
> > >
> > > Any device driver except for devfreq_cooling.c might
> > > use dev_pm_opp_enable/disable interface.
> > > So, don't need to remove the devfreq->scaling_max_freq
> > > and devfreq->scaling_min_freq for supporting OPP interface.
> >
> > It seems that devfreq_cooling was the only upstream user of
> > dev_pm_opp_enable and the remaining callers of dev_pm_opp_disable are
> > probe-time checks.
> 
> OPP interface has still dev_pm_opp_enable and dev_pm_opp_disable
> function. As long as remains them, any device driver related to devfreq
> could call them at some time. The devfreq supports the OPP interface,
> not just for only devfreq_cooling.

I would like to remove the disabled OPP handling since no devfreq device
makes use of dev_pm_opp_enable/disable, but I fear you are right that
we have to keep it as long as the API is available.