Re: [PATCH v2 16/21] dt-bindings: spi: document support for SA8255p

[Krzysztof Kozlowski <krzk@xxxxxxxxxx>]

On 05/09/2024 18:56, Krzysztof Kozlowski wrote:
> On 05/09/2024 18:08, Nikunj Kela wrote:
>>
>> On 9/5/2024 7:39 AM, Krzysztof Kozlowski wrote:
>>> On 05/09/2024 16:15, Nikunj Kela wrote:
>>>> On 9/5/2024 7:09 AM, Krzysztof Kozlowski wrote:
>>>>> On 05/09/2024 16:03, Nikunj Kela wrote:
>>>>>> On 9/5/2024 1:04 AM, Krzysztof Kozlowski wrote:
>>>>>>> On 04/09/2024 23:06, Nikunj Kela wrote:
>>>>>>>> On 9/4/2024 9:58 AM, Andrew Lunn wrote:
>>>>>>>>>> Sorry, didn't realize SPI uses different subject format than other
>>>>>>>>>> subsystems. Will fix in v3. Thanks
>>>>>>>>> Each subsystem is free to use its own form. e.g for netdev you will
>>>>>>>>> want the prefix [PATCH net-next v42] net: stmmac: dwmac-qcom-ethqos:
>>>>>>>> of course they are! No one is disputing that.
>>>>>>>>> This is another reason why you should be splitting these patches per
>>>>>>>>> subsystem, and submitting both the DT bindings and the code changes as
>>>>>>>>> a two patch patchset. You can then learn how each subsystem names its
>>>>>>>>> patches.
>>>>>>>> Qualcomm QUPs chips have serial engines that can be configured as
>>>>>>>> UART/I2C/SPI so QUPs changes require to be pushed in one series for all
>>>>>>>> 3 subsystems as they all are dependent.
>>>>>>> No, they are not dependent. They have never been. Look how all other
>>>>>>> upstreaming process worked in the past.
>>>>>> Top level QUP node(patch#18) includes i2c,spi,uart nodes.
>>>>>> soc/qcom/qcom,geni-se.yaml validate those subnodes against respective
>>>>>> yaml. The example that is added in YAML file for QUP node will not find
>>>>>> sa8255p compatibles if all 4 yaml(qup, i2c, spi, serial nodes) are not
>>>>>> included in the same series.
>>>>>>
>>>>> So where is the dependency? I don't see it. 
>>>> Ok, what is your suggestion on dt-schema check failure in that case as I
>>>> mentioned above? Shall we remove examples from yaml that we added?
>>> I don't understand what sort of failure you want to fix and why examples
>>> have any problem here. 
>>
>> If the QUPs yaml changes are not included in the same series with
> 
> They cannot be included in the same series. You just think that
> including here solves the problem so go ahead, simulate the merging:
> 1. Bjorn applies soc/qcom/qcom,geni-se.yaml patch and tests. His tree
> MUST build, so it also must pass dt_binding_check.
> Does it pass? No.
> 
> 2. SPI maintainer... ah, no point even going there.
> 
>> i2c,serial yaml changes, you see these errors:
>>
>> /builds/robherring/dt-review-ci/linux/Documentation/devicetree/bindings/soc/qcom/qcom,geni-se.example.dtb: geniqup@9c0000: serial@990000:compatible:0: 'qcom,sa8255p-geni-uart' is not one of ['qcom,geni-uart', 'qcom,geni-debug-uart']
>> /builds/robherring/dt-review-ci/linux/Documentation/devicetree/bindings/soc/qcom/qcom,geni-se.example.dtb: geniqup@9c0000: i2c@984000:compatible:0: 'qcom,sa8255p-geni-i2c' is not one of ['qcom,geni-i2c', 'qcom,geni-i2c-master-hub']
> 
> Don't grow examples if not needed. Or create dependencies and ask
> maintainers to cross-merge.

Or soc/geni-se binding could be also converted to just list compatibles
instead of referencing other schema, just like MDSS.

Best regards,
Krzysztof