Re: [PATCH v2 4/4] arm64: dts: mediatek: mt8167: Add DRM nodes
From: Krzysztof Kozlowski
Date: Tue Nov 17 2026 - 08:12:20 EST
On 17/11/2026 14:00, AngeloGioacchino Del Regno wrote:
> On 11/17/26 13:55, Krzysztof Kozlowski wrote:
>> On 17/11/2026 13:53, AngeloGioacchino Del Regno wrote:
>>> On 11/17/26 13:35, Krzysztof Kozlowski wrote:
>>>> On 21/09/2026 13:38, Luca Leonardo Scorcia wrote:
>>>>> Add all the DRM nodes required to get DSI and DPI to work on MT8167 SoC.
>>>>>
>>>>> Signed-off-by: Luca Leonardo Scorcia <l.scorcia@xxxxxxxxx>
>>>>> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@xxxxxxxxxxxxx>
>>>>> ---
>>>>> arch/arm64/boot/dts/mediatek/mt8167.dtsi | 369 +++++++++++++++++++++++
>>>>> 1 file changed, 369 insertions(+)
>>>>>
>>>>> diff --git a/arch/arm64/boot/dts/mediatek/mt8167.dtsi b/arch/arm64/boot/dts/mediatek/mt8167.dtsi
>>>>> index 27cf32d7ae35..a50bae4307c1 100644
>>>>> --- a/arch/arm64/boot/dts/mediatek/mt8167.dtsi
>>>>> +++ b/arch/arm64/boot/dts/mediatek/mt8167.dtsi
>>>>> @@ -16,6 +16,22 @@
>>>>> / {
>>>>> compatible = "mediatek,mt8167";
>>>>>
>>>>> + aliases {
>>>>> + aal0 = &aal;
>>>>> + ccorr0 = &ccorr;
>>>>> + color0 = &color;
>>>>> + dither0 = &dither;
>>>>> + dpi0 = &dpi0;
>>>>> + dpi1 = &dpi1;
>>>>> + dsi0 = &dsi;
>>>>> + gamma0 = γ
>>>>> + ovl0 = &ovl0;
>>>>> + pwm0 = &disp_pwm;
>>>>> + rdma0 = &rdma0;
>>>>> + rdma1 = &rdma1;
>>>>> + wdma0 = &wdma;
>>>>
>>>> Folks, what is happening here? Why Mediatek is adding bunch of
>>>> undocumented aliases? This should have never been accepted to Mediatek SoC.
>>>>
>>>
>>> It's always been like this on MediaTek from day 1, what's up with that!?
>>
>> I wrote what is wrong - they are undocumented. You were never allowed to
>> add undocumented ABI, never. Even 10 years ago.
>>
>
> Well, that's right.
>
>> I don't know what is day 1 in this context, but for a few (5-8 years?)
>> we do not allow more custom aliases. Neither new DTS users of these.
>>
>
> That actually poses a severe issue, as mediatek-drm relies on aliases to find out
> the instance number of the display controller components, on all SoCs.
If someone sent these for DT maintainers review, we would provide review
telling they are not allowed and you need to use other methods. The
problem is that someone sneaked undocumented ABI.
Instance IDs are not allowed in DT - neither as aliases nor as properties.
If the graph topology or phandles do not point to proper relation
between components, then the recommended way I know (happy to learn more
myself) is to map instances to IDs in the drivers. For example
Qualcomm/MSM is doing that.
>
> Any advice on how to possibly resolve this without rewriting the entire thing?
>
> Mapping that by mmio address in the driver seems too stupid/dirty/bad/etc.
Driver needs that, so why exactly stupid? You do not represent any
actual hardware information in DT with these aliases. You want to tell
drivers how drivers should behave - that's not a job for DT, that's job
for driver data. Of course if there is smarter way than MMIO->ID, I am
all in.
Best regards,
Krzysztof