On 08/09/2022 18:13, Dmitry Baryshkov wrote:
>>>>
>>>> + mdss: mdss@1a00000 {
>>>> + compatible = "qcom,mdss";
>>>> +
>>>> + reg = <0x1a00000 0x1000>,
>>>> + <0x1ab0000 0x1040>;
>>>> + reg-names = "mdss_phys",
>>>> + "vbif_phys";
>>>> +
>>>> + power-domains = <&gcc MDSS_GDSC>;
>>>> + interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_HIGH>;
>>>> +
>>>> + interrupt-controller;
>>>> + #interrupt-cells = <1>;
>>>> +
>>>> + clocks = <&gcc GCC_MDSS_AHB_CLK>,
>>>> + <&gcc GCC_MDSS_AXI_CLK>,
>>>> + <&gcc GCC_MDSS_VSYNC_CLK>,
>>>> + <&gcc GCC_MDSS_MDP_CLK>;
>>>> + clock-names = "iface",
>>>> + "bus",
>>>> + "vsync",
>>>> + "core";
>>>> +
>>>> + #address-cells = <1>;
>>>> + #size-cells = <1>;
>>>> + ranges;
>>>> +
>>>> + status = "disabled";
>>>> +
>>>> + mdp: mdp@1a01000 {
>>>> + compatible = "qcom,mdp5";
>>>
>>> Could you please change this to "qcom,msm8953-mdp5", "qcom,mdp5".
>>
>> This would be the first dtsi using the two compatibles then, correct? Are there
>> any plans to adjust other SoCs?
>
> Yes, this is a long-going plan. Having just "qcom,mdp5" doesn't allow
> switching between mdp5 and dpu1 drivers. Thus I'd ask to add per-SoC
> compat strings.
>
> It's up to you (and Rob/Krzysztof) whether to leave just one compat
> string or have both of them: a per-soc one and a generic one.
If device can bind to generic fallback ("qcom,mdp5") and still work
somehow, then the fallback is OK. However if generic "qcom,mdp5" does
not work at all, let's just choose something which is matching current
patterns.
Best regards,
Krzysztof
