Opteron errata / nVidia issues - links and stuff [was Re: [Pc_Support] It's here!]

[Austin Denyer (Ozz)]

On Sat, 05 Nov 2005 10:45:50 -0600, "Bryan J. Smith"
<b.j.smith at ieee.org> wrote:
>
> On Sat, 2005-11-05 at 01:56 -0500, Phil Barnett wrote:
> > nvidia drivers working here... (compiled)
> > $ uname -a
> > Linux localhost.localdomain 2.6.13-1.1532_FC4 #1 Thu Oct 20 01:30:08 EDT 2005 
> > i686 athlon i386 GNU/Linux
> 
> I think he was referring to ...
>   1.  Opteron 2xx (or 8xx) series (dual to 8-way)
>   2.  Linux/x86-64 kernel

That's correct.  These are Opteron 246 CPUs, Debian amd64 distro.
 
> AMD will always have to deal with the fact that it's architecture
> doesn't have full OS support yet (except for maybe in Solaris), so it
> has to ensure that it's "front-side bottleneck" compatibility for
> assumptions in the OS work flawlessly.  Unfortunately, there are always
> issues with that -- because of the extra coherency placed in hardware.
> 
> So the fact that the Opteron has compatibility issues with video and
> other I/O designed for Intel platforms doesn't surprise me.  Intel is
> just now getting into multiple FSBs with its new Socket-771 platform,
> and will run into much of what AMD already learned on the Athlon MP.
> But even then, they don't have the issue that AMD has -- multiple I/O
> channels.  ;->

The new Intel dual-cores are downright pathetic from the reviews I've
seen so far.
 
> I'd be interest in any links he has to those issues mentioned.  I'd like
> to know more about them, because I want to know where AMD's backward
> compatibility on the interconnect is still lacking.

See below.

Regards,
Ozz.


Here's a link to the kernel patch issue.  It's long, so I've
reproduced the clip below.

http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.14-rc2

<quote>
 commit bc5e8fdfc622b03acf5ac974a1b8b26da6511c99
 Author: Linus Torvalds <torvalds at g5.osdl.org>
 Date:   Sat Sep 17 15:41:04 2005 -0700

     x86-64/smp: fix random SIGSEGV issues
    
     They seem to have been due to AMD errata 63/122; the fix is to
     disable TLB flush filtering in SMP configurations.
    
     Confirmed to fix the problem by Andrew Walrond <andrew at walrond.org>
    
     [ Let's see if we'll have a better fix eventually, this is the Q&D
       "let's get this fixed and out there" version ]
    
     Signed-off-by: Linus Torvalds <torvalds at osdl.org>
</quote>

And here's another link:

http://lkml.org/lkml/2005/9/20/207

<quote>
Date  Tue, 20 Sep 2005 10:30:48 -0700 (PDT)
  From  Linus Torvalds <>
  Subject  Re: x86-64 bad pmds in 2.6.11.6
  
On Tue, 20 Sep 2005, Charles McCreary wrote:
>
> Another datapoint for this thread. The box spewing the bad pmds messages is a 
> dual opteron 246 on a TYAN S2885 Thunder K8W motherboard. Kernel is 
> 2.6.11.4-20a-smp.

This is quite possibly the result of an Opteron errata (tlb flush
filtering is broken on SMP) that we worked around as of 2.6.14-rc4.

So either just try 2.6.14-rc2, or try the appended patch (it has since 
been confirmed by many more people).

                Linus
</quote>


And finally, details of the nVidia kernel issue.  This is from the
Debian-amd64 mailing list.

From: Hans <x at y.z>
To: debian-amd64 at lists.debian.org
Cc: Gilles <x at y.z>
Subject: Re: nvidia (graphics) driver pain
Date: Thu, 3 Nov 2005 15:07:13 +0100
User-Agent: KMail/1.6.2

Am Donnerstag, 3. November 2005 15:11 schrieb Gilles:
> > I had a similar problem. Please remember, that nvidia-kernel and
> > nvidai-glx must be the same version.
>
> Both are from version 1.0.7676-1.  So that's not the problem...
>
> But I didn't install "nvidia-glx-ia32".  Is it necessary?
>
>
> Gilles
No, this is only necessary, if you want to use 32-bit programs in
64-bit environment (I use this i.e. for X-Plane (this is a flight
simulator)).

BTW: In Kernel 2.6.14, the kernel-module can be built, but not loaded,
due to an obsolete parameter in the sources ! I recommend to use
2.6.12, this work fine. But use 7174, 7176 and higher will not work (so
it is for me).