On Wed, Mar 03, 2021 at 05:49:01PM +0100, Krzysztof Kozlowski wrote: > On 03/03/2021 17:43, Greg Kroah-Hartman wrote: > > > > > > I don't think that will work in practice. Many ARCH_ symbols for various > > > > > > architectures contradict with each other. Almost all watchdog drivers > > > > > > only _build_ for specific platforms/architectures. > > > > > > > > > > Great, that's horrible to hear, so much for a "generic arm64 kernel > > > > > binary" which I _thought_ was the goal. > > > > > > > > > > ugh, you would have thought we would have learned our lesson with > > > > > arm32... > > > > > > I have no idea what you are talking about here. arm64 kernels have > > > always been generic, but you still need drivers for each piece of > > > hardware, we unfortunately can't stop SoC vendors from reinventing > > > the wheel with each new platform and then having to add yet another > > > driver for each subsystems. > > > > That's fine, drivers are easy, but when I see comments like "ARCH_ > > symbols contradict" that means that we can not make a generic kernel > > image. Otherwise there's no contradiction :) > > No, they don't contradict. > > > > > And "new drivers" are almost always not really "new" as everyone uses > > much the same IP blocks. As proof of this patch where the DWC3 IP block > > is being used by multiple SoC vendors. To handle that, you split out > > the SoC-specific portions into sub-drivers, so that you can build a > > single image of the driver that works on multiple platforms. Nothing > > new, we've been doing this for years, it's just that out-of-mainline SoC > > trees that think they can touch "core IP block code" break this all the > > time, which is what I am pushing back on. > > I am perfectly fine with (and like it!) putting dwc3 exynos back into > base/main dwc3 and getting rid of USB_DWC3_EXYNOS entirely. But this was > not part of this patch... I doubt that will happen, and it's not something that I expect. It's ok to have platform-specific "sub-drivers" for common IP blocks, we do it all the time. But making it separate is good, much like has been done for xhci as well. > > Anyway, this is just me as a driver subsystem maintainer being grumpy to > > see ARCH_ dependancies on tiny little things like SoC-portions for > > generic IP drivers. Or on individual drivers (i.e. Samsung serial port > > driver), where they don't belong at all. > > At least with Samsung serial driver we see adding new SoC - Apple M1. > > Here, the guys in Samsung want to tweak several kernel parts to work with > their out-of-tree code without contributing this code back. It's not a > community-friendly approach. The upstream kernel should be tweaked to the > out-of-tree unknown, hidden and uncontrollable code. Totally agreed, that's not ok. But a different issue than what is happening here. thanks, greg k-h
