Re: [PATCH 1/2] OPP: Use _set_opp_level() for single genpd case

[Ulf Hansson]

On Mon, 30 Oct 2023 at 11:29, Viresh Kumar <viresh.kumar@xxxxxxxxxx> wrote:
>
> On 26-10-23, 11:53, Ulf Hansson wrote:
> > On Wed, 25 Oct 2023 at 15:49, Stephan Gerhold <stephan@xxxxxxxxxxx> wrote:
> > >  2. The OPP WARNing triggers with both variants because it just checks
> > >     if "required-opps" has a single entry. I guess we need extra checks
> > >     to exclude the "parent genpd" case compared to the "OPP" case.
> > >
> > >         [    1.116244] WARNING: CPU: 2 PID: 36 at drivers/opp/of.c:331 _link_required_opps+0x180/0x1cc
> > >         [    1.125897] Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
> > >         [    1.146887] pc : _link_required_opps+0x180/0x1cc
> > >         [    1.146902] lr : _link_required_opps+0xdc/0x1cc
> > >         [    1.276408] Call trace:
> > >         [    1.283519]  _link_required_opps+0x180/0x1cc
> > >         [    1.285779]  _of_add_table_indexed+0x61c/0xd40
> > >         [    1.290292]  dev_pm_opp_of_add_table+0x10/0x18
> > >         [    1.294546]  of_genpd_add_provider_simple+0x80/0x160
> > >         [    1.298974]  cpr_probe+0x6a0/0x97c
> > >         [    1.304092]  platform_probe+0x64/0xbc
> > >
> > > It does seem to work correctly, with and without this patch. So I guess
> > > another option might be to simply silence this WARN_ON(). :')
> >
> > Oh, thanks for pointing this out! This case haven't crossed my mind yet!
> >
> > Allow me to think a bit more about it. I will get back to you again
> > with a suggestion soon, unless Viresh comes back first. :-)
>
> I have resent the series now.
>
> Stephan, please give it a try again. Thanks.
>
> Regarding this case where a genpd's table points to a parent genpd's table via
> the required-opps, it is a bit tricky to solve and the only way around that I
> could think of is that someone needs to call dev_pm_opp_set_config() with the
> right device pointer, with that we won't hit the warning anymore and things will
> work as expected.
>
> In this case the OPP core needs to call dev_pm_domain_set_performance_state()
> for device and then its genpd. We need the right device pointers :(
>
> Ulf, also another important thing here is that maybe we would want the genpd
> core to not propagate the voting anymore to the parent genpd's ? The
> dev_pm_opp_set_opp() call is better placed at handling all things and not just
> the performance state, like clk, regulator, bandwidth and so the recursion
> should happen at OPP level only.

Are you saying that the OPP library should be capable of managing the
parent-clock-rates too, when there is a new rate being requested for a
clock that belongs to an OPP? To me, that sounds like replicating
framework specific knowledge into the OPP library, no? Why do we want
this?

Unless I totally misunderstood your suggestion, I think it would be
better if the OPP library remained simple and didn't run recursive
calls, but instead relied on each framework to manage the aggregation
and propagation to parents.

> For now my series shouldn't break anything,
> just that we will try to set performance state twice for the parent genpd, the
> second call should silently return as the target state should be equal to
> current state.
>
> --
> viresh

Kind regards
Uffe