On 2/25/2025 9:07 AM, Krzysztof Kozlowski wrote: > On 25/02/2025 17:44, Jeff Johnson wrote: >>>>> >>>>> But nothing parses such string as 'qcom,ath12k-calibration-variant' (see >>>>> git grep), so how would driver use it? >>>> >>>> That's what I'm asking: is the property redundant or is it correct and >>>> it is a driver that needs to be fixed? >>> >>> I assume driver will need something like that property, but that's not a >>> reason to accept incorrect one in DTS. One cannot add properties to DTS >>> without bindings, so bypassing bindings review, and then claim "but my >>> driver needs them". Send proper patches for driver first which will get >>> a review. >> >> We definitely need a calibration variant entry. >> I've pinged the development team to get the driver patch. > > > The patches were on the lists but were not accepted. Therefore DTS > property cannot get into the kernel. I am sorry, but this is not somehow > fluid or flexible that internal team can squeeze something into the kernel. I see bindings and DTS patches but no driver patch, even in my internal queue. > > Also post factum reasoning is not correct, because this would open the > gate to bypass any sort of review. Just squeeze your stuff into the DTS > and then you can bypass all DT maintainers :/ > > All properties must be documented and bindings must be accepted *before* > DTS patch is applied. There is no intention to bypass DT maintainers. We are just trying to upstream a large amount of downstream code, and in the process some pieces are coming out of order. And there is also confusion if binding, driver, and DTS changes should be in one series or three separate series. We are moving towards an upstream-first model, but we still have to address the existing technical debt. /jeff
