Re: [PATCH 2/2] OPP/pmdomain: Fix the assignment of the required-devs

[Viresh Kumar]

FYI, I am on holidays now :)

On Fri, 6 Sept 2024 at 14:19, Ulf Hansson <ulf.hansson@xxxxxxxxxx> wrote:
> > How do we differentiate between two cases where the required-opps can
> > be defined as either of these:
> >
> > required-opps = <&opp_pd_50, &opp_pd_51>; //corresponds to pd_perf1 and pd_perf0 (in reverse order)
> >
> > OR
> >
> > required-opps = <&opp_pd_51, &opp_pd_50>; //corresponds to pd_perf0 and pd_perf1
> >
> > I thought this can't be fixed without some platform code telling how
> > the DT is really configured, i.e. order of the power domains in the
> > required-opps.
>
> I don't think we need platform code for this.
>
> When registering a genpd provider, an OPP table gets assigned to it.

So we will create a real OPP table in code, which will point to the common
OPP table in DT. Fine.

> When hooking up a device to one of its genpd providers, that virtual
> device then also gets a handle to its genpd's OPP table.

Right.

If there are two genpds required for a device from the same genpd provider, the
picture isn't very clear at this point. i.e. which required OPP
belongs to which genpd,
as both have same table in DT.

> Each of the phandles in the required-opps points to another OPP table,
> which OPP table should be associated with a specific genpd.

Yes, but a simple order reversal in DT (which I sent in my last
email), will not be picked
by code at all. i.e. DT doesn't give the order in which required OPPs
are present.

> In other words, the information is there, we should not need anything
> additional in DT.