Re: [PATCH v2 1/2] dt-bindings: display/msm: dsi-controller-main: Fix deprecated QCM2290 compatible

[Konrad Dybcio <konrad.dybcio@linaro.org>]

On 18.02.2023 15:49, Krzysztof Kozlowski wrote:
> On 18/02/2023 12:23, Konrad Dybcio wrote:
>>
>>
>> On 18.02.2023 11:14, Krzysztof Kozlowski wrote:
>>> On 17/02/2023 22:13, Bryan O'Donoghue wrote:
>>>> On 17/02/2023 12:24, Krzysztof Kozlowski wrote:
>>>>> First, it would be nice to know what was the intention of Bryan's commit?
>>>>
>>>> Sorry I've been grazing this thread but, not responding.
>>>>
>>>> - qcom,dsi-ctrl-6g-qcm2290
>>>>
>>>> is non-compliant with qcom,socid-dsi-ctrl which is our desired naming 
>>>> convention, so that's what the deprecation is about i.e. moving this 
>>>> compat to "qcom,qcm2290-dsi-ctrl"
>>>
>>> OK, then there was no intention to deprecate qcom,mdss-dsi-ctrl and it
>>> should be left as allowed compatible.
>> Not sure if we're on the same page.
> 
> We are.
> 
>>
>> It wasn't intended to deprecate [1] "qcom,qcm2290-dsi-ctrl", "qcom-mdss-dsi-ctrl";
>> (newly-introduced in Bryan's cleanup patchset) but it was intended to deprecate
>> [2] "qcom,dsi-ctrl-6g-qcm2290"; which was introduced long before that *and* used in
>> the 6115 dt (and it still is in linux-next today, as my cleanup hasn't landed yet).
>>
>> [3] "qcom,dsi-ctrl-6g-qcm2290", "qcom,mdss-dsi-ctrl" was never used (and should never
>> be, considering there's a proper compatible [1] now) so adding it to bindings
>> didn't solve the undocumented-ness issue. Plus the fallback would have never
>> worked back then, as the DSI hw revision check would spit out 2.4.1 or 2.4.
>> which is SC7180 or SDM845 and then it would never match the base register, as
>> they're waay different.
> 
> All these were known. I was asking about "qcom,mdss-dsi-ctrl", because
> the original intention also affects the way we want to keep it now
> (unless there are other reasons).
Okay, so we want to deprecate:

"qcom,dsi-ctrl-6g-qcm2290", "qcom,mdss-dsi-ctrl"

because it is:

1) non-compliant with the qcom,socname-hwblock formula
2) replaceable since we rely on the fallback compatible
3) "qcom,dsi-ctrl-6g-qcm2290" alone would have been expected to
   be fixed in the DTSI similar to other SoCs

Is that correct?

Because 2) doesn't hold, as - at the time of the introduction
of Bryan's patchset - the fallback compatible would not have
been sufficient from the Linux POV [1], though it would have been
sufficient from the hardware description POV, as the hardware
on the SoC *is* essentially what qcom,mdss-dsi-ctrl refers to.

[1] The driver would simply not probe. It *would be* Linux-correct
after my code-fixing series was applied, but I think I'm just failing
to comprehend what sort of ABI we're trying to preserve here :/

Konrad

> 
> Best regards,
> Krzysztof
>