Re: [PATCH v6 05/10] dt-bindings: soc: mediatek: convert pwrap documentation

From: Alexandre Mergnat
Date: Mon Nov 28 2022 - 09:04:09 EST


Le dim. 27 nov. 2022 à 14:11, Krzysztof Kozlowski
<krzysztof.kozlowski@xxxxxxxxxx> a écrit :
>
> >> + pmic:
> >> + type: object
> >
> > What's here? Other schema? If not then maybe compatible? What about
> > description?
>
> I guess this was comment from Rob, so it's fine.

Yes it is.

> >> +allOf:
> >> + - if:
> >> + properties:
> >> + compatible:
> >> + contains:
> >> + const: mediatek,mt8365-pwrap
> >> + then:
> >> + properties:
> >> + clocks:
> >> + minItems: 4
> >> +
> >> + clock-names:
> >> + minItems: 4
> >
> > else:
> > ???
>
> Actually this looks less complete than your previous patch.
>
> else:
> clocks:
> maxItems: 2
> same for clock-names
>

I think I’ve followed the feedback done here [1]
I’ve declared `minItems: 2` globally and override it to 4 if
mediatek,mt8365-pwrap is used. Isn’t it the right way to implement it
?

> >> + compatible = "mediatek,mt8135-pwrap";
> >> + reg = <0 0x1000f000 0 0x1000>,
> >
> > This does not match your unit address. No warnings when compile testing?
> >

There are no warnings when compile testing. I will fix the unit
address anyway, sorry.

> >> + <0 0x11017000 0 0x1000>;
> >> + reg-names = "pwrap", "pwrap-bridge";
> >> + interrupts = <GIC_SPI 128 IRQ_TYPE_LEVEL_HIGH>;
> >> + clocks = <&clk26m>, <&clk26m>;
> >> + clock-names = "spi", "wrap";
> >> + resets = <&infracfg MT8135_INFRA_PMIC_WRAP_RST>,
> >> + <&pericfg MT8135_PERI_PWRAP_BRIDGE_SW_RST>;
> >> + reset-names = "pwrap", "pwrap-bridge";
> >
> > Missing pmic. Make your example complete.
>
> Probably pmic should be skipped, I understand it is described in MFD
> binding.
>

Put the pmic in the example have 2 constraints:
- The original pmic "mediatek,mt6397" isn’t supported by a yaml
schema, so I’ve a dt_binding_check fail: `failed to match any schema
with compatible: ['mediatek,mt6397']`
- If I put another pmic that supports a yaml schema, I need to put all
required properties for the pmic, which I thought was unnecessary
since it’s already done in its own schema and can change for another
pmic, so less consistent.

Then yes, IMHO, PMIC should be skipped in the example.

[1] https://lore.kernel.org/all/fe898d24-54fa-56bb-8067-b422a3a52ff5@xxxxxxxxxxxxx/

Alex