amd64/87348: amd64+smp+startkde always crashing

[Vasil Dimov]

The following reply was made to PR amd64/87348; it has been noted by GNATS.

From: Vasil Dimov <vd at datamax.bg>
To: "D\. J\. Bernstein" <freebsd-d8749b97554608222e6dfa2fbc915a0b at box.cr.yp.to>
Cc: freebsd-gnats-submit at FreeBSD.org
Subject: Re: amd64/87348: amd64+smp+startkde always crashing
Date: Wed, 19 Oct 2005 08:43:17 +0300

 On Wed, Oct 12, 2005 at 11:02:44PM +0000, D. J. Bernstein wrote:
 > Built an Athlon 64 X2 box; Asus A8V, 250GB SATA, Radeon VE (7000). Installed FreeBSD 5.4-RELEASE from the amd64 CDs. Modified kernel from GENERIC as follows: changed ident, add ``device sound'', add ``device "snd_via8233"'', add ``options SMP''.
 ...
 > Major problem: startkde always freezes the machine. Caps Lock doesn't toggle keyboard lights; Ctrl-Alt-Backspace has no effect; only way to affect the machine, apparently, is to press the power button.
 ...
 > Removing SMP appears to eliminate the freeze, but I'd rather run Linux than throw away half of the CPU power. Haven't done serious stress tests, so I don't know whether this is tied to KDE, or to X, or to overlal load. Have reproduced on an identical Athlon 64 X2 box; could be a problem with the A8V etc., but clearly isn't a problem with assembly of this particular box.
 > 
 
 I have the same motherboard (A8V) and since I got it, I started
 experiencing problems, including (but not limited to the following):
 
 * random reboots (just as if I had pressed the reset button)
 * kernel panics
 * machine freezings (like yours - kbd dead, must press reset btn)
 * random programs crashes (signal 11) with programs that are otherwise
   rock stable like: sed, ispell, perl, gcc
 
 All of these occured when there was some load on the machine - like
 make index in the ports dir.
 
 The peak moment was 2 days ago when I wasn't able to compile the world,
 because gcc failed with signal 11 two subsequent times on different
 pieces of code.
 
 Then I rebooted and changed the Memclock setting in the BIOS from "Auto"
 to "5:3 DDR333", although the memory is DDR400, and this fixed the
 problem - /usr/coredumps is empty since then :) and I compiled the world
 with the first attempt, running make index in parallel.
 
 I hope that helps.