[Pkg-fglrx-devel] Bug#576942: Bug#576942: /boot/vmlinuz-2.6.32-3-amd64: kernel fix for wrong mtrr mask from MS-7252 bios may be wrong and cause problems with fglrx-driver

[Michael Gilbert]

On Sat, 10 Apr 2010 15:57:03 +0100 Ben Hutchings wrote:

> On Sat, 2010-04-10 at 10:45 -0400, Michael Gilbert wrote:
> > On Sat, 10 Apr 2010 15:18:43 +0200 Friedrich Delgado Friedrichs wrote:
> > 
> > > 
> > > Ben Hutchings schrieb:
> > > > On Thu, 2010-04-08 at 14:26 +0200, Friedrich Delgado Friedrichs wrote:
> > > > > [    0.000000] WARNING: at /build/mattems-linux-2.6_2.6.32-9-amd64-NYTFdD/linux-2.6-2.6.32-9/debian/build/source_amd64_none/arch/x86/kernel/cpu/mtrr/generic.c:467 generic_get_mtrr+0xbf/0xf9()
> > > > > [    0.000000] Hardware name: MS-7252
> > > > > [    0.000000] mtrr: your BIOS has set up an incorrect mask, fixing it up.
> > > 
> > > > > are supposed to indicate a crash in the kernel.
> > > > It's not a crash, it's a warning about a BIOS bug.
> > > 
> > > In fact, that's what I suspected as well.
> > > 
> > > Is there any way this bios bug could cause me further trouble?
> > 
> > the key here is that the submitter's MTRR is crashing irrespective of fglrx.
> 
> Nothing is crashing.

"crash" may not have been the best choice of terminology, since yes it
did not bring the whole system down.  it is an "oops" (in kernel
speak), and as such ill-defined behavior should be expected from
applications that expect the MTRR code to be functioning properly.

> [...]
> > there is probably some incompatibility with MTRR and his hardware.
> 
> You clearly have no idea what an MTRR is.

i'll readily admit to that; although i don't see why that matters.
from a high-level perspective, i know that it's just a
memory/communication interface between the processor and video card.
i've never had cause to concern myself with MTRR before, and i have no
desire to expend much cognitive effort since it's not something i have
much interest in anyway.

but even so, you're calling me out on my choice of verbiage again, which
isn't very productive. maybe a better sentance would have been: my best
guess is that there is a problem in linux's MTRR code due to the
submitter's specific hardware.

so i guess the real question is: should the MTRR code not oops for his
hardware, or is it the responsibility of downstream applications to be
more robust when the MTRR code does oops?

mike