On 12/09/2025 10:27, Chen-Yu Tsai wrote: >>> +properties: >>> + compatible: >>> + oneOf: >>> + - enum: >>> + - mediatek,mt8173-mdp-rdma >>> + - mediatek,mt8173-mdp-rsz >>> + - mediatek,mt8173-mdp-wdma >>> + - mediatek,mt8173-mdp-wrot >> >> Why there is no mediatek,mt8173-mdp here? What does this compatible >> represent? >> >>> + - items: >>> + - const: mediatek,mt8173-mdp-rdma >> >> Still suspicious. Device cannot be simulatanously: compatible and not >> compatible. This is not a well known cat that has superposition of two >> states, whenenver you look the other way. >> >> Maybe the old binding was incorrect, maybe the in-tree DTS is incorrect. >> Whichever the reason, this must be investigated and documented, because >> by standard rules this is wrong. Each wrong code needs very clear >> explanations (and "someone did it" is not a good enough explanation). > > My guess is that "mediatek,mt8173-mdp" is meant to serve as a single > entry point for the implementation to bind the driver to. The MDP is
I am speaking about hardware. What piece of hardware it implements and why sometimes it is RDMA MDP and sometimes it is not RDMA MDP, but only MDP. > a Data Pipeline and there could be multiple instances of the same > IP block, as seen in the original example. > > The datasheet I have doesn't cover the "RDMA" block specifically, so > I can't say whether there is an actual difference between the two RDMA > blocks. > > Best regards, Krzysztof
