Re: [PATCH v3 1/4] dt-bindings: qcom,qcs615-venus: add support for video hardware

[Krzysztof Kozlowski <krzk@kernel.org>]

On 26/11/2024 10:39, Renjiang Han wrote:
>>>>> If so, do I also need to remove these two nodes from the dtsi file and add
>>>> Yes
>>>>
>>>>> them in the qcs615-ride.dts file?
>>>> Well, no, how would it pass dtbs_check?
>>>>
>>>> Don't add nodes purely for Linux driver instantiation.
>>> OK, I got it. I'll update like this. If video-decoder and video-encoder are
>>>
>>> removed from dtsi file and not added to qcs615-ride.dts file, then the
>>>
>>> video decoder and encoder functions will not be available on the qcs615
>>>
>>> platform. So I think these two nodes should be added to the
>>>
>>> qcs615-ride.dts file to ensure that the qcs615 platform can enable the
>>>
>>> video decoder and encoder functions.
>> You just repeated the same sentences. Address my comment instead - empty
>> device nodes should not be used just to instantiate Linux device drivers.
> 
> Thanks for your reply. I agree with your comment. The two nodes 
> video-decoder and
> 
> video-encoder should not be placed in the devicetree. But this is 
> affected by the venus
> 
> driver. On the old platform, some only need to enable the video-decoder 
> function or
> 
> only enable the video-encoder function. So these two nodes were added to the
> 
> devicetree at that time. For new platforms, the iris driver will be used 
> in the future,
> 
> and this situation will not occur.
These are new bindings, for new device, so please fix your driver. We
had similar talk long time ago and answer was that it's a legacy driver
which won't be developed. This means also no new devices. If you bring
new devices to old driver, instead of to new iris, then it means you
still develop old driver. Fix the old driver.


Best regards,
Krzysztof