On 4.01.2024 10:59, Rajendra Nayak wrote: > > > On 1/2/2024 11:59 PM, Konrad Dybcio wrote: >> The downstream kernel has infrastructure for passing votes from different >> interconnect nodes onto different RPMh RSCs. This neither implemented, not >> is going to be implemented upstream (in favor of a different solution >> using ICC tags through the same node). >> >> Unfortunately, as it happens, meaningless (in the upstream context) parts >> of the vendor driver were copied, ending up causing havoc - since all >> "per-RSC" (in quotes because they all point to the main APPS one) BCMs >> defined within the driver overwrite the value in RPMh on every >> aggregation. >> >> To both avoid keeping bogus code around and possibly introducing >> impossible-to-track-down bugs (busses shutting down for no reason), get >> rid of the duplicated BCMs and their associated ICC nodes. > > Thanks Konrad for catching this, I do see these nodes in other Qualcomm SoCs upstream (atleast sm8350/sm8450 and sm8550), perhaps they need to be cleaned up as well? Yes, I cleaned up 8550 last week, I'll look into the rest. Konrad
