I can also report the exact same symptoms on an AMD 850 in an
ABIT KT7 motherboard using 512MB DIMMS. My system was pretty
stable under 1.5.3 although it wasn't used for much more than
sup, building a few packages, and some light NFS serving.

After upgrading to 1.6, I could crash it reliably every time
I built a kernel or package. After one crash the filesystems
were all dirty and the system could not finish the parallel
fsck without a panic - I had to boot single and check each
one individually.

Most, but not all, of the panics were pagefaults during disk
activity. Many of the panics were in the "random" hooks in
the ufs code.

I've tried swapping 4 DIMMS and setting the BIOS to "fail-safe"
mode with no luck yet.


Frank van der Linden wrote:
>>Peter> A kernel trace in the debugger shows different things
>>Peter> everytime. (ie everytime different things cause it)
> 
> 
>>Also what I've seen.  Most, if not all, of the time, the processes
>>involved with compiling the kernel aren't involved.
> 
> 
> I'm sorry to say that this still looks like a hardware problem..
> It's certainly not a problem with AMD chips in general; I have
> 5 systems with AMD CPUs in them, varying from a K6-2 500 (which has
> been up for 416 days running 1.5.x) to a dual Athlon 1.2Ghz (which
> I use as my desktop and compile machine, doing many recompiles of
> kernels, and sometimes whole builds, a day).
> 
> Maybe you're being bitten by the old VIA chipset DMA bug. Try going
> into the BIOS setup, and set everything to conservative values (try
> "fail-safe defaults" if your BIOS has that option). Try disabling
> DMA on pciide altogether. Try DIMMs of smaller memory size. See if
> you can get a flash update for your motherboard's BIOS. Boot
> with a minimal configuration (i.e. all PCI cards out, except
> video and IDE), and see if that's stable.
> 
> - Frank