* Kees Cook <keesc...@chromium.org> wrote:
> On Wed, May 18, 2016 at 4:29 AM, Ingo Molnar <mi...@kernel.org> wrote:
> >
> > * Kees Cook <keesc...@chromium.org> wrote:
> >
> >> > I think there is something way more subtle going on here, and it bothers
> >> > me
> >> > exactly because it is subtle. It may be that it is OK right now, but
> >> > there
> >> > are alarm bells going on all over my brain on this. I'm going to stare
> >> > at
> >> > this for a bit and see if I can make sense of it; but if it turns out
> >> > that
> >> > what we have is something really problematic it might be better to apply
> >> > a big
> >> > hammer and avoid future breakage once and for all.
> >>
> >> Sounds good. I would just like to decouple this from the KASLR
> >> improvements.
> >> This fragility hasn't changed as a result of that work, but I'd really
> >> like to
> >> have that series put to bed -- I've spent a lot of time already cleaning
> >> up it
> >> and other areas of the compressed kernel code. :)
> >
> > So I disagree on that: while technically kASLR is independent of
> > relocations, your
> > series already introduced such a relocation bug and I don't want to further
> > increase complexity via kASLR without first increasing robustness.
>
> Well, in my defense, the bug was never actually reachable.
Hm, the changelog says a crash/reboot might happen:
commit 434a6c9f90f7ab5ade619455df01ef5ebea533ee
Author: Kees Cook <keesc...@chromium.org>
Date: Mon May 9 13:22:04 2016 -0700
x86/KASLR: Initialize mapping_info every time
As it turns out, mapping_info DOES need to be initialized every
time, because pgt_data address could be changed during kernel
relocation. So it can not be build time assigned.
Without this, page tables were not being corrected updated, which
could cause reboots when a physical address beyond 2G was chosen.
is the changelog wrong?
> > So could we try something to either detect or avoid such subtle and hard to
> > debug relocation bugs in very early boot code?
>
> I've sent this (the readelf patch which detects the bug from the KASLR
> series),
> but hpa wants to do a more comprehensive version. Could we temporarily use my
> version of this, since it appears to accomplish at least a subset of the new
> goal?
Yeah, that's fine with me.
> And on a related topic, how would you like me to send Thomas Garnier's memory
> base randomization series? Pull request, or as a series like I've done with
> the
> other KASLR improvements?
A series (size limited if necessary) would be nice!
Thanks,
Ingo
