09.12.2019 23:47, Mark Brown пишет: > On Mon, Dec 09, 2019 at 11:31:59PM +0300, Dmitry Osipenko wrote: >> 09.12.2019 19:40, Mark Brown пишет: > >>> Why would this need to be a stable fix? Presumably people with stable >>> kernels are using the old device tree anyway? > >> At least Rob Herring is asking to maintain backwards compatibility >> because some ditros are using newer device-trees with stable kernels. > > You're talking about forwards compatibility not backwards here. Are > those distros actually using LTS kernels? I think openSUSE Leap could be one of those distros that use LTS kernel with newer device-trees, but that's not 100%. Maybe Rob could help clarifying that. >> I'm personally also tending to use the newer DTB with older kernel >> version whenever there is a need to check something using stable kernel. >> Perhaps losing sound is not very important, but will be nicer if that >> doesn't happen. > > That really does sound like a "you broke it, you get all the pieces" > situation TBH... I'd be a lot more comfortable if the stable kernels > were sticking to bugfix only though I do appreciate that they're not > really that any more. In some cases it could be painful to maintain device-tree compatibility for platforms like NVIDIA Tegra SoCs because hardware wasn't modeled correctly from the start. I agree that people should use relevant device-trees. It's quite a lot of hassle to care about compatibility for platforms that are permanently in a development state. It could be more reasonable to go through the pain if kernel required a full-featured device tree for every SoC from the start. But maybe Tegra / device-tree maintainers have a different opinion. IIUC, device-trees are meant to be stable and software-agnostic, at least that was the case not so long time ago and I don't think that this premise changed.
