On 22/05/2024 20:39, Péter Ujfalusi wrote: > > > On 22/05/2024 20:07, Krzysztof Kozlowski wrote: >> On 22/05/2024 19:02, Mithil wrote: >>>> Yep. And testing DTS should clearly show that conversion leads to >>>> incomplete binding. >>>> >>>>> >>>>>> I assume the DTS was validated with the binding. Isn't the case here? >>>> >>>> Mithil Bavishi, >>>> Are you sure you tested the DTS? >>> >>> dt_binding_check did not give me any errors. Yeah the example is >>> different from how it is implemented in the kernel ie board specific >>> (omap4, omap5 etc). Should the example be changed according to that >>> dtsi then? >> >> Binding needs to be adapted to match DTS or DTS has to be fixed to match >> binding, depending which one is correct. > > Normally the DTS is written based on the binding document and the driver > is written also to follow the binding document. > However in this case we have a broken/inaccurate binding document and > the existing DTS files and binaries in wild have deviated (there are > boards out there using qnx or BSD and use this binding), or to be > precise the binding document was not updated. > > The existing DTS files are the ABI, so we cannot deviate from them, > unfortunately. > > In this case the DTS / driver needs to be reverse engineered to create a > binding document. Ah, yes, the third option - ABI should not be broken and sometimes binding and DTS needs fixes. Best regards, Krzysztof
