On 08/04/2024 08:04, Abel Vesa wrote: > On 24-04-07 19:07:03, Bjorn Andersson wrote: >> On Sun, Apr 07, 2024 at 07:23:21PM +0300, Abel Vesa wrote: >>> Add dedicated schema for X1E80100 PMIC ARB (v7) as it allows multiple >>> buses by declaring them as child nodes. >>> >> >> But is this really a "dedicated schema for X1E80100"? Isn't it "the >> schema for all multi-bus controllers"? >> >> I.e. isn't this a "dedicated schema for all platforms starting with >> SM8450"? > > Suggestion was from Krzysztof to add platform specific comaptible (and > therefore schema). Since the first platform that will support in > upstream proper multi bus is the x1e80100, the schema needs to bear the > same name as the compatible. When support for multi bus will be added to > the other platforms (including the SM8450), they will use the fallback > compatible of the x1e80100 and will be documented in this newly added > schema. We did the same thing with some PHYs drivers, IIRC. > >> >> Can you please use the commit message to document the actual reason why >> you choose to create a dedicated schema for this? Is it simply to avoid >> having to schema with either pmics or multiple buses as children? > > I can re-send the patchset with such a phrase in commit message. > > One of the early versions of this patchset was actually submitting a > generic compatible for multi bus, but I remember that there was a > request for following the platform dedicated approach. > > Krzysztof, can you please provide here the argument for why that is > preferred? I could not find such suggestions from my side in the archives, except: https://lore.kernel.org/all/dd86117e-0196-499b-b8b3-efe4013cbc07@xxxxxxxxxx/ where I want SoC specific compatibles to be used, not versions. Now about this binding, it is not a schema for all platforms starting with sm8450, but only for x1e. I do not understand why this would be a problem? If you ask why this is not a schema for all platforms, then because: 1. maybe no one tested other SoCs? 2. maybe no one cares? 3. maybe other boards need some quirks, so this would be applicable but not fully? I don't know... since when do we add "generic schemas"? However maybe the question is different: why other devices are not described here, while they should? Then probably Abel can answer what he wants and what he does not want to describe. There is no requirement to model all possible hardware in a binding, but instead describe one hardware, so x1e, fully. Best regards, Krzysztof
