Re: [PATCH 2/8] dt-bindings: soc: socionext: Add UniPhier SoC-glue logic

[Krzysztof Kozlowski]

On 30/11/2022 09:59, Kunihiko Hayashi wrote:
> Hi Krzysztof,
> 
> On 2022/11/29 23:43, Krzysztof Kozlowski wrote:
>> On 29/11/2022 11:35, Kunihiko Hayashi wrote:
>>> Add devicetree binding schema for the SoC-glue logic implemented on
>>> Socionext Uniphier SoCs.
>>>
>>> This SoC-glue logic is a set of miscellaneous function registers
>>> handling signals for specific devices outside system components,
>>> and also has multiple functions such as I/O pinmux, usb-phy, debug,
>>> clock-mux for a specific SoC, and so on.
>>>
>>> Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@xxxxxxxxxxxxx>
>>> ---
>>>   .../socionext,uniphier-soc-glue.yaml          | 94 +++++++++++++++++++
>>>   1 file changed, 94 insertions(+)
>>>   create mode 100644
>>> Documentation/devicetree/bindings/soc/socionext/socionext,uniphier-soc-glue.yaml
>>>
>>> diff --git
>>> a/Documentation/devicetree/bindings/soc/socionext/socionext,uniphier-soc-glue.yaml
>>> b/Documentation/devicetree/bindings/soc/socionext/socionext,uniphier-soc-glue.yaml
>>> new file mode 100644
>>> index 000000000000..3f571e3e1339
>>> --- /dev/null
>>> +++
>>> b/Documentation/devicetree/bindings/soc/socionext/socionext,uniphier-soc-glue.yaml
>>> @@ -0,0 +1,94 @@
>>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
>>> +%YAML 1.2
>>> +---
>>> +$id:
>>> http://devicetree.org/schemas/soc/socionext/socionext,uniphier-soc-glue.yaml#
>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>>> +
>>> +title: Socionext UniPhier SoC-glue logic
>>> +
>>> +maintainers:
>>> +  - Kunihiko Hayashi <hayashi.kunihiko@xxxxxxxxxxxxx>
>>> +
>>> +description: |+
>>> +  SoC-glue logic implemented on Socionext UniPhier SoCs is a collection
>>> of
>>> +  miscellaneous function registers handling signals outside system
>>> components.
>>> +
>>> +properties:
>>> +  compatible:
>>> +    items:
>>> +      - enum:
>>> +          - socionext,uniphier-ld4-soc-glue
>>> +          - socionext,uniphier-pro4-soc-glue
>>> +          - socionext,uniphier-pro5-soc-glue
>>> +          - socionext,uniphier-pxs2-soc-glue
>>> +          - socionext,uniphier-ld6b-soc-glue
>>> +          - socionext,uniphier-sld8-soc-glue
>>> +          - socionext,uniphier-ld11-soc-glue
>>> +          - socionext,uniphier-ld20-soc-glue
>>> +          - socionext,uniphier-pxs3-soc-glue
>>> +          - socionext,uniphier-nx1-soc-glue
>>> +          - socionext,uniphier-soc-glue
>>
>> This one looks generic - why having it next to specific ones?
> 
> SoC-glue has the same register set, but different implementations
> for each SoC.

Sure, but you did not model it as a compatible fallback, but like one of
variants. It is not tied to specific SoC, thus too generic.

> I thought of defining the same register set as a common specs,
> but each compatibles are sufficient. I'll remove it.
> 
>> Same question for your previous patch - socionext,uniphier-sysctrl.
>>
>> And similarly to previous patch, do you expect child nodes everywhere?
> 
> In case of this SoC-glue logic, all SoCs has pinctrl, however,
> only SoCs with USB2 host has usb-controller (phy-hub).
> And only legacy SoCs implement clock-controller (clk-mux) here.
> 
> Should child nodes that exist only in a specific "compatible" be defined
> conditionally?

No, rather define them in top level but disallow for specific compatibles:

allOf:
 - if:
  ....
   then:
     patternProperties:
       ...: false

Assuming that this does not over-complicate schema.


Best regards,
Krzysztof