This is a familiar problem now. I have the same thing with an IDE
drive.

Specificly, I have:
Pentium -90 (Intel Zappa board with PCI IDE controllers)
16 Meg of RAM
2 Conner 420 Meg drives
NetBSD-1.1 with Patch 1 installed in the kernel.

Questions:

Has everyone who seen this tested their hardware? I did, and the
software found nothing wrong in repeated tests of the memory, cache
and system board.

If it is the case that everyone who has seen it can prove that it's
not the hardware, this would seem it the VM-System or interconnecting
components.

The second question: has anyone seen this on non i386 systems?

In a related question, is it possible to prime the kernel to drop into
debug when recieving signal 11 or any other VM fault.

Another question: Has it happened to anyone outside of cc1? I think I
have had this happen, but I can't verify that it wasn't just a local
memory leak.

For those going through it, it seems that just recontinuing your make
seems to be O.K. I built one kernel through this stuff, just kept
typing make, and It worked fine, running X-Clients, emacs, building
small and large programs, and using the network. Just smile and type
make again. :-)
--
Hacksaw