linux-kernel - Re: [PATCH] [resend] intel_agp: Don't oops with zero stolen memory

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <87zky9iou7.fsf@pollan.anholt.net>
Date:	Fri, 02 Jul 2010 13:57:36 -0700
From:	Eric Anholt <eric@...olt.net>
To:	Ondrej Zary <linux@...nbow-software.org>
Cc:	David Airlie <airlied@...ux.ie>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] [resend] intel_agp: Don't oops with zero stolen memory

On Fri, 2 Jul 2010 21:59:20 +0200, Ondrej Zary <linux@...nbow-software.org> wrote:
> On Friday 02 July 2010 00:44:39 Eric Anholt wrote:
> > On Wed, 16 Jun 2010 10:13:52 +0200, Ondrej Zary <linux@...nbow-software.org> 
> wrote:
> > > When "onboard video memory" is set do "disabled" in BIOS on Asus
> > > P4P800-VM board (i865G), kernel oopses with memory corruption:
> > > https://bugs.freedesktop.org/show_bug.cgi?id=28430
> > >
> > > Fix that by cleanly aborting the initialization.
> > >
> > > Signed-off-by: Ondrej Zary <linux@...nbow-software.org>
> >
> > I haven't seen a system with that option before. Is the integrated
> > graphics still supposed to work at that point, or is it "graphics is
> > disabled."?  What do other OSes do here?
> 
> Testing it right now with Windows XP. Integrated graphics works as secondary 
> only when the "onboard video memory" is set to "disabled"! If anything other 
> is set (e.g. 1MB or 32MB), the driver does not load (code 10).

OK, I was thinking in that case that we should fix the AGP driver's math
so that it correctly set up a chip with no stolen memory allocation.  On
the other hand, I don't think we know what the correct math is, so maybe
we should just go ahead with disabling AGP in that case.


Content of type "application/pgp-signature" skipped