Subject: Re: Kernel panics with Asus P4R800V Celeron 2400 256 MB ram
To: None <port-i386@NetBSD.org>
From: David Lord <netbsd@lordynet.org>
List: port-i386
Date: 07/08/2007 12:52:24
On 5 Jul 2007, at 15:51, David Lord wrote:
> For several weeks I've been trying to get NetBSD 3.1 to run without
> panics on this system, Asus P4R800V m/b, Celeron 2.4GHz and 256 MB
> ram. It's was previously running FreeBSD as fileserver and build
> system without problem (like 6mth+ uptime no problem). I think NetBSD
> 3.0 was installed August 2006 and only/big problem then was with
> auixp driver codec error preventing bootup but a kernel without that
> driver allowed boot but with kernel panics if loaded. I've gone
> through following:
>
> swapped ram and memtest
Despite replacement of ram and no errors from hours running memtest
between attempts at getting this working it appears to be a memory
related fault. From single user I was getting a memory error and core
dump when running 'tar -xzf src.tgz' with src.tgz being about 216MB
and extracted files using nearly 1GB. I've reduced ram speed setting
from 400 to 266 and extracted ok. I've since re-enabled softdeps and
booted with full set of services running and again file extracted ok.
Moving same src over nfs - seemed ok and tools+kernel build on copied
src directory completed in 19 minutes.
David
> seatools check of hdd
>
> updated m/b bios
>
> installed NetBSD-daily/netbsd-3-1/200707020002Z after wget and mkiso
> on this suspect system and ftp of iso to desktop to burn install cd.
>
> Problem is now much worse.
>
> eg. Remote nfs mount and attempt to copy src.tgz (216MB) gives
> uvm_fault (0xc08ae4e0,0x966000,0,1) -> 0xe
> kernel: page fault trap, code=0
> Stopped in pid 895.1 (nfsio) at netbsd: sched_wakeup+0x24:
> cmpl %edi,0x34(%ebx)
>
> What appears to happen from 'top' is that files increases and free
> decreases then when free is bouncing around 300K - 1MB there will be
> a panic/lockup/segmentation fault or whatever.
>
> Bios update didn't help as original minimum ram for vga was 8MB but
> now is 16MB so max total ram is now 238MB rather than 246MB. Problem
> seems worse with this 238MB total vs previous 190MB (16MB for vga vs
> 64MB).
>
> Any hints as to how best to proceed with sorting this. To me it just
> appears as if kernel has overestimate of free ram as I've not seen
> any swap space in use.
>
> dmesg attached
>
> David