On Sun, 27 Oct 2024 at 22:41, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>
> It has caught real problems and as long as we have highmem support, it
> should stay IMO to provide test coverage.
Yeah. I'd love to get rid of highmem support entirely, and that day
*will* come. Old 32-bit architectures that do stupid things can just
deal with old kernels, we need to leave that braindamage behind some
day.
But as long as we support it, we should at least also have the debug
support for it on sane hardware.
Of course, maybe we should just make PageHighMem() always return true
for CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP, but I suspect that would cause
more pain than is worth it.
But yeah, I do think we should seriously start thinking about just
getting rid of HIGHMEM.
Linus
