On Sat, Jun 25, 2022 at 4:42 AM Guo Ren <guoren@xxxxxxxxxx> wrote: > > In that case, you probably need a boot-time check for this feature bit > > to refuse booting a kernel with qspinlock enabled when it has more than > > one active CPU but does not support the random backoff, > Do you mean we should combine ticket-lock into qspinlock, and let the > machine choose during boot-time? That is not what I meant here. It would be great if it could be done, but I doubt this is possible without adding excessive complexity. > From Peter's comment, seems the arch is broken without a strong fwd guarantee. > https://lore.kernel.org/linux-riscv/YGwKpmPkn5xIxIyx@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/ > > I'm not sure qspinlock guys would welcome ticket-lock getting into the > qspinlock data structure (Although ticket-lock is also made by peter). What I meant is that a kernel that is not safe to run on hardware without the necessary guarantees should just refuse to boot, the same way that we tend to warn when the CPU instruction set version is too old for what the kernel was built against. Arnd
