Re: Access to non-RAM pages

[Linus Torvalds]

On Fri, Aug 31, 2018 at 2:18 PM Jiri Kosina <jikos@xxxxxxxxxx> wrote:
>
> If noone has any clever idea how to work this around (I don't), I am
> afraid we'd have to ditch the whole DCACHE_WORD_ACCESS optimization, as
> it's silently dangerous.

No way in hell will I apply such a stupid patch.

It is NOT dangerous.

If you have a machine with RAM that touches IO, you need to disable
the last page, exactly the same way we disable and marked reserved the
first page at zero.

I thought we already did that.

I suspect this is a Xen bug, where the fake BIOS sets up a garbage
description of the hardware that is simply not realistic. I don't
think I've ever seen a machine that didn't have some reserved memory
at the top, but hey, if we don't expressly mark the last page reserved
already, doing so should be trivial.

No way do we disable the word accesses just because of some crazy
corner case that doesn't matter and doesn't happen in reality.

               Linus