Re: x86: move the LAPIC va

[Maxime Villard <max%m00nbsd.net@localhost>]

Le 30/12/2016 à 18:33, Maxime Villard a écrit :
> Which more appropriate magic address? On i386 for example, the VA bounds are
> variable-sized, and there is no place where you could hard-code the lapic. We
> would have to split the kernel VA space in two around that page, etc.

Just a precision on this: on i386 the kernel is embedded into kernel_map, and
the end of the bootstrap memory is the beginning of the kernel heap. Because of
that, you can't use a magic virtual address that is above the kernel, since it
will point to somewhere in kernel_map. You can't use a va below the kernel
either, since there is the PTE space (and then, userland).

On amd64 however, kernel_map is below the kernel, and we actually already have
a set of magic addresses available (and mapped) at kernel + static_offset. But
they have to do with tmp mappings, which I think are currently bug-friendly.

Therefore, in either case, we need a dynamic va.