Re: [PATCH 3/7] arm64: dts: qcom: sc8180x: create common zap-shader node

[Konrad Dybcio]

On 10/30/25 7:45 PM, Bjorn Andersson wrote:
> On Thu, Oct 30, 2025 at 07:58:46PM +0200, Dmitry Baryshkov wrote:
>> On Thu, Oct 30, 2025 at 12:28:31PM -0500, Bjorn Andersson wrote:
>>> On Thu, Oct 30, 2025 at 11:59:00AM +0100, Konrad Dybcio wrote:
>>>> On 10/30/25 11:58 AM, Konrad Dybcio wrote:
>>>>> On 10/28/25 10:00 PM, Dmitry Baryshkov wrote:
>>>>>> In order to reduce duplication, move common GPU memory configuration
>>>>>> from individual board files to sc8180x.dtsi.
>>>>>>
>>>>>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxxxxxxxx>
>>>>>> ---
>>>>>
>>>>> [...]
>>>>>
>>>>>> diff --git a/arch/arm64/boot/dts/qcom/sc8180x-primus.dts b/arch/arm64/boot/dts/qcom/sc8180x-primus.dts
>>>>>> index 93de9fe918ebdadf239832db647b84ac9d5a33f6..069953dcad378448800d45e14931efe1fe1a69fc 100644
>>>>>> --- a/arch/arm64/boot/dts/qcom/sc8180x-primus.dts
>>>>>> +++ b/arch/arm64/boot/dts/qcom/sc8180x-primus.dts
>>>>>> @@ -14,6 +14,8 @@
>>>>>>  #include "sc8180x.dtsi"
>>>>>>  #include "sc8180x-pmics.dtsi"
>>>>>>  
>>>>>> +/delete-node/ &gpu_mem;
>>>
>>> I agree with your hmm, seems this line should be dropped(?)
>>>
>>> Dmitry, please confirm and I can fix it up as I'm applying the series.
>>
>> Oh, and this line is necessary, because sc8180x-primus has its own
>> location for the gpu_mem.
>>
> 
> Hmm, you're right, not sure why the memory map differs... But this looks
> good then.

The latest memory map for 8180 says:

GPU micro-code @ 0x98715000, 0x2000-long
(matching the Flex 5G)

Across all the revisions, Primus's range is either part of a
"reserved" region, or the CDSP's memory.

So it's either wrong or OEM-modified, with the latter being rather
an odd thing given it's a CRD..

Konrad