On Wed, 30 Aug 2023 16:25:48 +0100, Jiaxun Yang <jiaxun.yang@xxxxxxxxxxx> wrote: > > > > 在 2023/8/30 21:44, Marc Zyngier 写道: > [...] > >> What's the best way, in your opinion, to overhaul this property? As we don't > >> really care backward compatibility of DTBs on those systems we can > >> just redesign it. > > You may not care about backward compatibility, but I do. We don't > > break existing systems, full stop. > Ah it won't break any existing system. Sorry for not giving enough insight > into the platform in previous reply. As for Loongson64 all DTBs are built > into kernel binary. So as long as binding are changed together with all DTS > in tree we won't break any system. This is factually wrong. QEMU produces a DT for Loongarch at runtime. So no, you're not allowed to just drop bindings on the floor. They stay forever. > > As for the offending property, it has no place here either. DT is not > > the place where you put "performance knobs". > Hmm, I can see various bindings with vendor prefix exposing device > configurations. If we seen this interrupt routing as a device configuration > I don't think it's against devicetree design philosophy. Just because we have tons of crap in the device trees doesn't give you a license to be just as bad. M. -- Without deviation from the norm, progress is not possible.
