Re: [PATCH 10/10] arm64: dts: allwinner: h5: Add CPU Operating Performance Points table

[Chen-Yu Tsai]

On Wed, Jan 30, 2019 at 5:41 PM Chen-Yu Tsai <wens@xxxxxxxx> wrote:
>
> On Wed, Jan 30, 2019 at 5:29 PM Maxime Ripard <maxime.ripard@xxxxxxxxxxx> wrote:
> >
> > On Wed, Jan 30, 2019 at 04:42:03PM +0800, Chen-Yu Tsai wrote:
> > >                       enable-method = "psci";
> > >                       clocks = <&ccu CLK_CPUX>;
> > >                       clock-latency-ns = <244144>; /* 8 32k periods */
> > > +                     operating-points-v2 = <&cpu_opp_table>;
> > > +                     #cooling-cells = <2>;
> > > +             };
> > > +     };
> > > +
> > > +     cpu_opp_table: opp_table {
> > > +             compatible = "operating-points-v2";
> > > +             opp-shared;
> > > +
> > > +             opp@408000000 {
> > > +                     opp-hz = /bits/ 64 <408000000>;
> > > +                     opp-microvolt = <1000000 1000000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@648000000 {
> > > +                     opp-hz = /bits/ 64 <648000000>;
> > > +                     opp-microvolt = <1040000 1040000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@816000000 {
> > > +                     opp-hz = /bits/ 64 <816000000>;
> > > +                     opp-microvolt = <1080000 1080000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@912000000 {
> > > +                     opp-hz = /bits/ 64 <912000000>;
> > > +                     opp-microvolt = <1120000 1120000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@960000000 {
> > > +                     opp-hz = /bits/ 64 <960000000>;
> > > +                     opp-microvolt = <1160000 1160000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@1008000000 {
> > > +                     opp-hz = /bits/ 64 <1008000000>;
> > > +                     opp-microvolt = <1200000 1200000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@1056000000 {
> > > +                     opp-hz = /bits/ 64 <1056000000>;
> > > +                     opp-microvolt = <1240000 1240000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@1104000000 {
> > > +                     opp-hz = /bits/ 64 <1104000000>;
> > > +                     opp-microvolt = <1260000 1260000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> > > +             };
> > > +
> > > +             opp@1152000000 {
> > > +                     opp-hz = /bits/ 64 <1152000000>;
> > > +                     opp-microvolt = <1300000 1300000 1310000>;
> > > +                     clock-latency-ns = <244144>; /* 8 32k periods */
> >
> > What is the frequency and voltage that U-Boot sets up?
>
> 1008 MHz, and whatever voltage the board design defaults to (typically
> the higher setting).

FYI I got this wrong. U-boot's default is 816 MHz. However it seems even
this is misleading, as cpufreq in Linux reports:

    cpufreq: cpufreq_online: CPU0: Running at unlisted freq: 792000 KHz
    cpufreq: cpufreq_online: CPU0: Unlisted initial frequency changed
to: 816000 KHz

So there seems tp be an off-by-1 error in some multiplier calculation.

>
> > We've had the issue with the A33 that it's started at 1008MHz, with
> > the matching voltage, and ramping up the frequency to 1.2GHz on boards
> > without PMIC support would increase the frequency but not the voltage,
> > resulting in a brownout.
>
> Which is why I added the regulator to all boards before this patch. At
> least for Linux, once the regulator supply is described in the device
> tree, if the driver is missing, regulator_get_* and thus cpufreq should
> fail with -EPROBE_DEFER.

I believe despite the above, this still stands. So we should be OK.