Re: [PATCH v2 3/6] dt-bindings: soc: starfive: syscon: Add optional patternProperties

From: Xingyu Wu
Date: Mon Mar 20 2023 - 04:26:26 EST


On 2023/3/20 15:40, Krzysztof Kozlowski wrote:
> On 20/03/2023 08:29, Xingyu Wu wrote:
>> On 2023/3/20 14:37, Krzysztof Kozlowski wrote:
>>> On 20/03/2023 04:54, Xingyu Wu wrote:
>>>> On 2023/3/19 20:28, Krzysztof Kozlowski wrote:
>>>>> On 16/03/2023 04:05, Xingyu Wu wrote:
>>>>>> Add optional compatible and patternProperties.
>>>>>>
>>>>>> Signed-off-by: Xingyu Wu <xingyu.wu@xxxxxxxxxxxxxxxx>
>>>>>> ---
>>>>>> .../soc/starfive/starfive,jh7110-syscon.yaml | 39 ++++++++++++++++---
>>>>>> 1 file changed, 33 insertions(+), 6 deletions(-)
>>>>>>
>>>>>> diff --git a/Documentation/devicetree/bindings/soc/starfive/starfive,jh7110-syscon.yaml b/Documentation/devicetree/bindings/soc/starfive/starfive,jh7110-syscon.yaml
>>>>>> index ae7f1d6916af..b61d8921ef42 100644
>>>>>> --- a/Documentation/devicetree/bindings/soc/starfive/starfive,jh7110-syscon.yaml
>>>>>> +++ b/Documentation/devicetree/bindings/soc/starfive/starfive,jh7110-syscon.yaml
>>>>>> @@ -15,16 +15,31 @@ description: |
>>>>>>
>>>>>> properties:
>>>>>> compatible:
>>>>>> - items:
>>>>>> - - enum:
>>>>>> - - starfive,jh7110-aon-syscon
>>>>>> - - starfive,jh7110-stg-syscon
>>>>>> - - starfive,jh7110-sys-syscon
>>>>>> - - const: syscon
>>>>>> + oneOf:
>>>>>> + - items:
>>>>>> + - enum:
>>>>>> + - starfive,jh7110-aon-syscon
>>>>>> + - starfive,jh7110-stg-syscon
>>>>>> + - starfive,jh7110-sys-syscon
>>>>>> + - const: syscon
>>>>>> + - items:
>>>>>> + - enum:
>>>>>> + - starfive,jh7110-aon-syscon
>>>>>> + - starfive,jh7110-stg-syscon
>>>>>> + - starfive,jh7110-sys-syscon
>>>>>> + - const: syscon
>>>>>> + - const: simple-mfd
>
> BTW, this also looks wrong. You just said that clock controller exists
> only in few variants. Also, why sometimes the same device goes with
> simple-mfd and sometimies without? It's the same device.

Oh yes, If modified to:

oneOf:
- items:
- enum:
- starfive,jh7110-aon-syscon
- starfive,jh7110-stg-syscon
- const: syscon
- items:
- const: starfive,jh7110-sys-syscon
- const: syscon
- const: simple-mfd

Or:

- minItems: 2
items:
- enum:
- starfive,jh7110-aon-syscon
- starfive,jh7110-stg-syscon
- starfive,jh7110-sys-syscon
- const: syscon
- const: simple-mfd


Which one is better?

>
>>>>>>
>>>>>> reg:
>>>>>> maxItems: 1
>>>>>>
>>>>>> +patternProperties:
>>>>>> + # Optional children
>>>>>> + "pll-clock-controller":
>>>>>
>>>>> It's not a pattern.
>>>>
>>>> Does it use 'properties' instead of 'patternProperties'?
>>>
>>> Yes.
>>>
>>>>
>>>>>
>>>>> Anyway should be clock-controller
>>>>
>>>> Will fix.
>>>>
>>>>>
>>>>>> + type: object
>>>>>> + $ref: /schemas/clock/starfive,jh7110-pll.yaml#
>>>>>> + description: Clock provider for PLL.
>>>>>> +
>>>>>
>>>>> You just added these bindings! So the initial submission was incomplete
>>>>> on purpose?
>>>>>
>>>>> No, add complete bindings.
>>>>
>>>> Does you mean that it should drop the 'description', or add complete 'description',
>>>> or add 'compatible', 'clocks' and 'clock-cells' of complete clock-controller bindings?
>>>
>>> It means it should be squashed with the patch which adds it.
>>
>> Should I drop the 'decription' here and keep the 'decription' in patch1?
>
> There should be no this patch at all. However I do not understand what
> you want to do with description. What's wrong with description?

I thought you were commenting under description, saying a conflict with pll dtbindings' description.
Is that mean I should add it in the syscon patch fo william not this patchset?

>>
>>>
>>>>
>>>>>
>>>>>> required:
>>>>>> - compatible
>>>>>> - reg
>>>>>> @@ -38,4 +53,16 @@ examples:
>>>>>> reg = <0x10240000 0x1000>;
>>>>>> };
>>>>>>
>>>>>> + - |
>>>>>> + syscon@13030000 {
>>>>>
>>>>> No need for new example... Just put it in existing one.
>>>>>
>>>>
>>>> Actually, the PLL clock-controller are just set in sys-syscon resgisters. The stg-syscon and
>>>> aon-syscon don't need it. So PLL clock-controller node only is added in sys-syscon node.
>>>
>>> So why having other examples if they are included here? Drop them.
>>>
>>
>> Should I drop the old example of stg-syscon and add a new example of sys-syscon which
>> include clock-controller child node?
>
> No, there should be no stg-syscon example, it's useless.
>

Thanks. I will remind william to use sys-syscon example instead.

Best regards,
Xingyu Wu