Re: [PATCH v5 1/2] dt-bindings: cpufreq: Document support for Airoha EN7581 CPUFreq

[Christian Marangi]

On Wed, Dec 04, 2024 at 02:30:17PM -0600, Rob Herring wrote:
> On Wed, Dec 4, 2024 at 12:51 PM Christian Marangi <ansuelsmth@xxxxxxxxx> wrote:
> >
> > On Wed, Dec 04, 2024 at 12:42:53PM -0600, Rob Herring wrote:
> > > On Tue, Dec 03, 2024 at 05:31:49PM +0100, Christian Marangi wrote:
> > > > Document required property for Airoha EN7581 CPUFreq .
> > > >
> > > > On newer Airoha SoC, CPU Frequency is scaled indirectly with SMCCC commands
> > > > to ATF and no clocks are exposed to the OS.
> > > >
> > > > The SoC have performance state described by ID for each OPP, for this a
> > > > Power Domain is used that sets the performance state ID according to the
> > > > required OPPs defined in the CPU OPP tables.
> > > >
> > > > Signed-off-by: Christian Marangi <ansuelsmth@xxxxxxxxx>
> > > > Reviewed-by: Ulf Hansson <ulf.hansson@xxxxxxxxxx>
> > > > ---
> > > > Changes v5:
> > > > - Add Reviewed-by tag
> > > > - Fix OPP node name error
> > > > - Rename cpufreq node name to power-domain
> > > > - Rename CPU node power domain name to perf
> > > > - Add model and compatible to example
> > > > Changes v4:
> > > > - Add this patch
> > > >
> > > >  .../cpufreq/airoha,en7581-cpufreq.yaml        | 262 ++++++++++++++++++
> > > >  1 file changed, 262 insertions(+)
> > > >  create mode 100644 Documentation/devicetree/bindings/cpufreq/airoha,en7581-cpufreq.yaml
> > > >
> > > > diff --git a/Documentation/devicetree/bindings/cpufreq/airoha,en7581-cpufreq.yaml b/Documentation/devicetree/bindings/cpufreq/airoha,en7581-cpufreq.yaml
> > > > new file mode 100644
> > > > index 000000000000..7e36fa037e4b
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/cpufreq/airoha,en7581-cpufreq.yaml
> > > > @@ -0,0 +1,262 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/cpufreq/airoha,en7581-cpufreq.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Airoha EN7581 CPUFreq
> > > > +
> > > > +maintainers:
> > > > +  - Christian Marangi <ansuelsmth@xxxxxxxxx>
> > > > +
> > > > +description: |
> > > > +  On newer Airoha SoC, CPU Frequency is scaled indirectly with SMCCC commands
> > > > +  to ATF and no clocks are exposed to the OS.
> > > > +
> > > > +  The SoC have performance state described by ID for each OPP, for this a
> > > > +  Power Domain is used that sets the performance state ID according to the
> > > > +  required OPPs defined in the CPU OPP tables.
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +    const: airoha,en7581-cpufreq
> > > > +
> > > > +  '#clock-cells':
> > > > +    const: 0
> > >
> > > You just said no clocks are exposed to the OS.
> > >
> >
> > Well we now simulate one due to request from cpufreq reviewers.
> >
> > Everything is still handled by SMC that only report the current
> > frequency of the CPU.
> >
> > > > +
> > > > +  '#power-domain-cells':
> > > > +    const: 0
> > > > +
> > > > +  operating-points-v2: true
> > > > +
> > > > +required:
> > > > +  - compatible
> > > > +  - '#clock-cells'
> > > > +  - '#power-domain-cells'
> > > > +  - operating-points-v2
> > > > +
> > > > +additionalProperties: false
> > > > +
> > > > +examples:
> > > > +  - |
> > > > +    / {
> > > > +        model = "Airoha EN7581 Evaluation Board";
> > > > +        compatible = "airoha,en7581-evb", "airoha,en7581";
> > > > +
> > > > +        #address-cells = <2>;
> > > > +           #size-cells = <2>;
> > >
> > > mixed tab and spaces.
> > >
> > > Can't I just go read the actual .dts files if I want to see
> > > *everything*? Examples should generally be just what the schema covers.
> > >
> >
> > Idea here is to give example as both clock and power-domain property are
> > needed in the CPU nodes for the CPUFreq driver to correctly work.
> 
> If we want to do that, then we really should have a schema defining
> that. But since there's only 1 for cpus that doesn't really work.
> 
> > Should I drop and just define the CPUFreq node?
> 
> Yes.
> 
> > > > +
> > > > +        cpus {
> > > > +            #address-cells = <1>;
> > > > +            #size-cells = <0>;
> > > > +
> > > > +            cpu0: cpu@0 {
> > > > +                device_type = "cpu";
> > > > +                compatible = "arm,cortex-a53";
> > > > +                reg = <0x0>;
> > > > +                operating-points-v2 = <&cpu_opp_table>;
> > > > +                enable-method = "psci";
> > > > +                clocks = <&cpu_pd>;
> > > > +                clock-names = "cpu";
> > > > +                power-domains = <&cpu_pd>;
> > > > +                power-domain-names = "perf";
> > > > +                next-level-cache = <&l2>;
> > > > +                #cooling-cells = <2>;
> > >
> > > I don't understand why you have clocks, power-domains and OPP?
> > > Certainly that's conceivable, but not with how you're abusing
> > > power-domains for performance points and you said clocks are not exposed
> > > to the OS.
> > >
> >
> > SMC scale based on index values not frequency. That really resembles a
> > power-domain.
> 
> So what is the point of the OPP table with frequency? You can set an
> OPP and read the frequency, right? So a table of frequencies is
> redundant.
>

The OPP for CPU node is to describe the supported frequency and then
each OPP have a required-opp property to describe the level to configure
the power-domain. It's really to make a connection between the 2. I need
to check but from my test the separate OPP table for the power domain is
needed or it does refuse to probe.

This is a common pattern also used by Qcom and Mediatek. Example qcs404 [0]

As you notice the very same pattern is used here.

> > SMC provide frequency in MHz tho so we model that as a
> > get-only clock.
> >
> > At times with no clocks are exposed I intend that they SoC doesn't
> > provide any raw control on them in the normal way with a register, bits
> > to change and logic to apply for mux and divisor, this thing is very
> > special and works only with 2 command and nothing else so I'm trying my
> > best to model this in the most descriptive and complete way possible.
> 
> Fair enough for the clock. Please clarify the description with what
> clock is provided. Just to make sure, all CPUs run at the same
> frequency?
>

Ok, yes it's all global also signaled by the opp-shared property.

[0] https://elixir.bootlin.com/linux/v6.12.1/source/arch/arm64/boot/dts/qcom/qcs404.dtsi

-- 
	Ansuel