Re: [PATCH v2 1/2] regulator: dt-bindings: qcom,rpmh: document supplies per variant

[Rob Herring <robh@kernel.org>]

On Tue, Apr 19, 2022 at 08:44:48AM +0200, Krzysztof Kozlowski wrote:
> On 14/04/2022 17:51, Rob Herring wrote:
> > On Tue, Apr 12, 2022 at 09:31:22AM +0200, Krzysztof Kozlowski wrote:
> >> The RPMH regulator binding covers several devices with different
> >> regulator supplies, so it uses patterns matching broad range of these
> >> supplies.  This works fine but is not specific and might miss actual
> >> mistakes when a wrong supply property is used for given variant.
> >>
> >> Describe the supplies depending on the compatible, using a defs-allOf
> >> method.
> >>
> >> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
> >> ---
> >>  .../regulator/qcom,rpmh-regulator.yaml        | 290 +++++++++++++++++-
> >>  1 file changed, 276 insertions(+), 14 deletions(-)
> >>
> >> diff --git a/Documentation/devicetree/bindings/regulator/qcom,rpmh-regulator.yaml b/Documentation/devicetree/bindings/regulator/qcom,rpmh-regulator.yaml
> >> index 842ccef691b8..773536fe37c7 100644
> >> --- a/Documentation/devicetree/bindings/regulator/qcom,rpmh-regulator.yaml
> >> +++ b/Documentation/devicetree/bindings/regulator/qcom,rpmh-regulator.yaml
> >> @@ -95,35 +95,297 @@ properties:
> >>    vdd-rgb-supply:
> >>      description: Input supply phandle of rgb.
> >>  
> >> -  vin-lvs-1-2-supply:
> >> -    description: Input supply phandle of one or more regulators.
> >> -
> >> -  vdd-bob-supply:
> >> -    description: BOB regulator parent supply phandle.
> >> -
> >>    bob:
> >>      type: object
> >>      $ref: "regulator.yaml#"
> >>      description: BOB regulator node.
> >>  
> >>  patternProperties:
> >> -  "^vdd-s([0-9]+)-supply$":
> >> -    description: Input supply phandle(s) of one or more regulators.
> >> -
> >> -  "^vdd-(l[0-9]+[-]){1,5}supply$":
> >> -    description: Input supply phandle(s) of one or more regulators.
> >> -
> >>    "^(smps|ldo|lvs)[0-9]+$":
> >>      type: object
> >>      $ref: "regulator.yaml#"
> >>      description: smps/ldo regulator nodes(s).
> >>  
> >> -additionalProperties: false
> >> -
> >>  required:
> >>    - compatible
> >>    - qcom,pmic-id
> >>  
> >> +allOf:
> >> +  - $ref: "#/$defs/pm6150"
> >> +  - $ref: "#/$defs/pm6150l"
> >> +  - $ref: "#/$defs/pm7325"
> >> +  - $ref: "#/$defs/pm8005"
> >> +  - $ref: "#/$defs/pm8009"
> >> +  - $ref: "#/$defs/pm8150"
> >> +  - $ref: "#/$defs/pm8150l"
> >> +  - $ref: "#/$defs/pm8350"
> >> +  - $ref: "#/$defs/pm8350c"
> >> +  - $ref: "#/$defs/pm8450"
> >> +  - $ref: "#/$defs/pm8998"
> >> +  - $ref: "#/$defs/pmg1110"
> >> +  - $ref: "#/$defs/pmi8998"
> >> +  - $ref: "#/$defs/pmr735a"
> >> +  - $ref: "#/$defs/pmx55"
> >> +  - $ref: "#/$defs/pmx65"
> >> +
> >> +unevaluatedProperties: false
> >> +
> >> +$defs:
> > 
> > I'm not following on why you need $defs here rather than putting the 
> > if/then schemas under the 'allOf'. $defs should primarily only be used 
> > where it saves duplicating a schema 2 or more times. That could be the 
> > case here if there's a case that's a subset of another case.
> 
> The allOf+defs is a workaround for schema behavior. The entire approach
> if defining properties in "if:then:" works only with
> unevaluatedProperties, not with additionalProperties. However
> unevaluatedProperties require to reference other schema, which I do not
> do here. I don't have other schema.
> 
> allOf+def has references tricking schema to accept unevaluatedProperties.

I'm going to relax the meta-schema instead.

Rob