On 30/10/2024 13:14, Nemanov, Michael wrote:
> On 10/30/2024 1:09 PM, Krzysztof Kozlowski wrote:
>> On 30/10/2024 11:59, Nemanov, Michael wrote:
>>>>
>>>> Your changelog does not explain these three. "Fixed compatibility" is
>>>> way too vague, especially that you do not fix anything here.
>>>>
>>>
>>> I was trying to address the feedback from previous patch. You said:
>>>
>>>>>>> +static const struct of_device_id cc33xx_sdio_of_match_table[] = {
>>>>>>> + { .compatible = "ti,cc3300", .data = &cc33xx_data },
>>>>>>> + { .compatible = "ti,cc3301", .data = &cc33xx_data },
>>>>>>> + { .compatible = "ti,cc3350", .data = &cc33xx_data },
>>>>>>> + { .compatible = "ti,cc3351", .data = &cc33xx_data },
>>>>>>> + { }
>>>>>>> +};
>>>>>>
>>>>>>
>>>>>> Eh? What happened here? So devices are compatibles thus make them
>>>>>> compatible in the bindings.
>>>>>>
>>>>>
>>>>> I thought this is the right way to do it (originally taken from [1]).
>>>>> How can I solve it via DT bindings?
>>>>
>>>> It's all over the bindings (also example-schema). Use fallback and oneOf.
>>>>
>>>
>>> Looking at [2] and [3] as an example I tried to do the same (make cc33xx
>>> driver compatible with all chip variants).
>>> How should have I done it?
>>
>> qcom-wdt is quite a different device. It's true you should have here
>> oneOf, but for a purpose. oneOf without purpose does not make sense, right?
>>
>> I think other TI bindings would serve you as an example. Or this one:
>>
>> https://elixir.bootlin.com/linux/v6.3-rc6/source/Documentation/devicetree/bindings/sound/nvidia,tegra210-ope.yaml#L31
>>
>>
>> Best regards,
>> Krzysztof
>>
>
> OK.
> So I should make one of the variants the base and declare others as
> compatible? i.e:
>
Yes
Best regards,
Krzysztof
