Subject: Re: kernel panics on heavy disk I/O
To: Bill Russell Nicolas <nicolas_bill_os@yahoo.co.uk>
From: Richard Rauch <rkr@olib.org>
List: port-amd64
Date: 10/06/2004 12:32:58
On Wed, Oct 06, 2004 at 05:49:34PM +0100, Bill Russell Nicolas wrote:
> Hi,
> 
> My system constantly crashes and reboot whenever I
> perform heavy disk I/O processes like extracting
> pkgsrc.tgz. My system is an AMD64 2800+ on MSI Neo
> Platinum board (nForce3) and 512 DDR 400. I'm using
> NetBSD_RC1, NetBSD_RC2 and the 20040715 kernels but
> all panicked in the same manner. I've already disabled
> ACPI so now I'm hoping it's not a memory problem.

Do you get any additional information about the panic?

Did you disable IOAPIC (not, I think, related to ACPI)?
Or did you mean IOAPIC when you said ACPI?

I never had a panic caused by IOAPIC, but had some strange
problems caused by this.  I have been told that running
without IOAPIC is "not really supoprted", but if I don't
disable it, I would lose devices such as my ethernet
or USB.  Some would never work, others would only work
for a while, then would appear to lose track of their
interrupts and stop working.

I do not know if this could cause a panic, but exactly
*which* devices I lost depended upon my PCI card
arrangements.  I assume that with the "right" setup,
you could lose almost any device.  I never noticed any
association to intensive use of any resources.

The LINUX kernels just disable IOAPIC (or did around
the beginning of 2004) if they find this chipset.
(You might try dropping a LINUX or FreeBSD or OpenBSD
install on your box to see how they get along with the
system.)


Also, re. the possibility of a memory problem, there's
a new memtest86 (called memtest86+, based upon but
independant from the original memtest86).  The new
memtest86+ works on my AMD64.  You might download the
ISO image and burn a CD to give it a try...


> Does anybody have similar experience with this kind of
> problem?

The closest I have to match is my computer's memory.
If I did anything intensive for more than a few minutes,
I'd get weird things happening.  That turned out to be
the memory module that I replaced.

(P.S.: Any suggestions about what to do with an old
memory module?  (^&  I tried applying a heatsink which
had adhesive, so it looks like a harmonica now...  It
is nominally DDR400.  Can it be dropped into a slower
speed memory system, or is it basically a trinket?)


-- 
  "I probably don't know what I'm talking about."  http://www.olib.org/~rkr/