Kernel: Panic problem with TCPDUMP

[scion+fbsdq at webrelay.net]

From: "Guthemberg Silvestre" <gutho80 at hotmail.com>
Date: Tue, 30 Mar 2004 13:46:06 -0300

>Hi everyone,
>I'm using the FreeBSD 5.2(FreeBSD 5.2-RELEASE #0: Sun Jan 11 04:21:45 GMT 
>2004),default installation, to collect packets with TCPDUMP. The computer 
>has:
>-> CPU: AMD Athlon(tm) XP 2400+ (2004.56-MHz 686-class CPU)
>-> real memory  = 536805376 (511 MB)
>-> rl0: <RealTek 8139 10/100BaseTX> port 0xd400-0xd4ff mem 
>0xdfffbf00-0xdfffbfff irq 18 at device 8.0 on pci0, etc.
>The problem is when I start the program TCPDUMP to collect the packets from 
>a 32Mb/s link (at the rl0 interface). After some time, the following log 
>error messages appear:
>"Mar 30 11:29:22 p2p kernel: panic: vm_fault: fault on nofault entry, addr: 
>e0aae000
>Mar 30 11:29:22 p2p kernel: cpuid = 0;
>Mar 30 11:29:22 p2p kernel:
>Mar 30 11:29:22 p2p kernel: syncing disks, buffers remaining... 3842 3842 
>3842 3842 3842 3842 3842 3842 3842 3842 3842 3842 3842 3842 3842 38
>42 3842 3842 3842 3842

You don't by any chance have a Shuttle Xpc do you?

Those are very similar dmesg lines to my Xpc.  I have
found that the memory boards are woefully undercooled
in this box.  I experimented with many coverings of the 
vent holes to increase the pressures near the front
where the memory resides.  Eventually I wedged a fan
inside and aimed it along the two dimms.  Runs fine now.

Before that I had the *same* results as you whenever
I ran software that exercised dma to disk, net, whatever.
Which meant that the system often crashed the first 
time I read mail...