On 9/3/2025 6:12 PM, Konrad Dybcio wrote: > On 8/28/25 11:59 PM, Georgi Djakov wrote: >> On 8/28/25 9:20 PM, Krzysztof Kozlowski wrote: >>> On 28/08/2025 20:16, Odelu Kukatla wrote: >>>> >>>>>> QoS configuration is essential for managing latency and bandwidth across >>>>>> subsystems such as CPU, GPU, and multimedia engines. Without it, the >>>>>> system may experience performance degradation, especially under >>>>> >>>>> So how was it working for the last 2 years? >>>>> >>>> The system may function normally without this feature. However, enabling >>> >>> >>> Huh? So you agree but keep continuing the discussion? >>> >>> I don't understand what we are discussing in such case, but just to >>> close the topic from my side and be explicit: based on above you cannot >>> break the ABI. >> >> To be even more specific, if we already have some DT binding without any >> clocks and reg properties, we can't just suddenly change them from now >> on to be "required". But they can still be "optional" and this will not >> break the ABI, right? The old DT is still valid and the QoS will be >> active when the new properties are present and this is handled properly >> by the driver. > > Correct and this very approach was used to retrofit QoS onto an even older > sc7280 icc driver. > > The icc-rpmh core already ignores QoS configuration if the clocks are not > provided. > > Konrad Thanks Krzysztof, Georgi, Konrad for the detailed discussion. To summarize: we can’t make reg and clocks properties required now for sa8775p, and making these properties optional ensures backward compatibility. I will address this in the next revision. Thanks again for the comments. Best regards, Odelu
