On Tue, 9 Aug 2016, Mark Kettenis wrote: > Early ARM CPUs had the exception vector table at address 0. That's a > terrible idea, since that overlaps with the address space most kernels > use for userland. This was fixed in later versions of the architecture > which allows for putting them somewhere high up in the address space > (where the kernel lives) or even remap them to an arbitrary address. > All ARMv7 CPUs have this feature so we use it unconditionally. And > therefore the code in cpuswitch() that tries to switch around the vector > page is useless. > > ok?
Makes sense to me, looks correct, and works on cubox. ok guenther@
