Re: [PATCH v2 1/9] dt-bindings: arm-smmu: Allow up to 3 power-domains

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 15/11/2022 14:06, Konrad Dybcio wrote:>> diff --git
a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
>>> b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
>>> index 28f5720824cd..55759aebc4a0 100644
>>> --- a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
>>> +++ b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
>>> @@ -200,7 +200,7 @@ properties:
>>>        maxItems: 7
>>>
>>>      power-domains:
>>
>> As I mentioned before - minItems: 1.
> But not all SMMUs require a power domain :/

It does not matter. This does not require power-domains.

> 
>>
>> Just like the link I gave you.
>>
>>> -    maxItems: 1
>>> +    maxItems: 3
>>>
>>>      nvidia,memory-controller:
>>>        description: |
>>> @@ -364,6 +364,26 @@ allOf:
>>>                - description: interface clock required to access smmu's
>>> registers
>>>                    through the TCU's programming interface.
>>>
>>> +  - if:
>>> +      properties:
>>> +        compatible:
>>> +          contains:
>>> +            const: qcom,sm6375-smmu-500
>>> +    then:
>>> +      properties:
>>> +        power-domains:
>>> +          items:
>>> +            - description: SNoC MMU TBU RT GDSC
>>> +            - description: SNoC MMU TBU NRT GDSC
>>> +            - description: SNoC TURING MMU TBU0 GDSC
>>> +
>>> +      required:
>>> +        - power-domains
>>> +    else:
>>> +      properties:
>>> +        power-domains:
>>> +          maxItems: 1
>>> +
>>>    examples:
>>>      - |+
>>>        /* SMMU with stream matching or stream indexing */
>>>
>>>
>>> In my eyes, this should work, but I still get errors like:
>>>
>>> /home/konrad/linux/arch/arm64/boot/dts/qcom/sm8250-hdk.dtb:
>>> iommu@3da0000: power-domains: [[108, 0]] is too short
>>>
>>> as if the else: path was never taken..
>>
>> It was, but the top-level property said that minItems=3 (implicitly), so
>> it is too short.
> So the top-level properties take precedence over the ones that come from 
> the if-then-else?? Ugh.

It's a sum of them. Top level is expected to define the widest
constraints and if-then-else narrows them per variants.

Best regards,
Krzysztof