Re: [PATCH v3 1/8] dt-bindings: marvell: a38x: convert the soc compatibles description to yaml

[Josua Mayer <josua@solid-run.com>]

Am 27.12.23 um 12:49 schrieb Krzysztof Kozlowski:
> On 26/12/2023 19:11, Josua Mayer wrote:
>> Convert the existing txt binding for armada-38x socs to DT schema
>> format.
> ...
>
>> +---
>> +$id: http://devicetree.org/schemas/arm/marvell/armada-38x.yaml#
>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
>> +
>> +title: Marvell Armada 38x Platforms
>> +
>> +maintainers:
>> +  - Gregory CLEMENT <gregory.clement@bootlin.com>
>> +
>> +properties:
>> +  $nodename:
>> +    const: '/'
>> +  compatible:
>> +    oneOf:
>> +
>> +      - description: Armada 380 SoC
>> +        items:
>> +          - const: marvell,armada380
>> +
> Thanks for the conversion. I think this is still not what we want. The
> main problem is of course in the original binding, not your conversion.
> Compatible armada380 above alone is not correct. This is SoC and SoC
> compatible cannot be used alone, because it is not specific enough to
> differentiate boards. Same applies for further lists, which I removed
> from the quote, for the same reasons.
>
> What we could do here is to fix this instead of pure conversion:
> 1. Add this binding file with new compatibles for existing boards and
> drop the TXT file, with explanation that you actually document properly
> all existing board compatibles.
So - we could add a boilerplate yaml with zero boards in conversion commit,
and then add new or existing boards through separate commits?

I feel that collecting / fixing the compatibles for all a38x based 
boards might be a lot of effort.
Would it be acceptable to start with the few that I know / use myself?
In addition to those already submitted I could include 
armada-388-helios4.dts.

> 2. Add compatibles to existing DTS.
For boards I do not know / using codenames - is there a template / 
generic description string?
> 3. Add your new board compatibles.
>
> At least this is what I would do for typical cases, but as I said
> before, Marvell here is in some surprising limbo state.

Thank you for the constructive notes!

> Best regards,
> Krzysztof
>