thanks heaps, Michael.

The <mips/profile.h> MCOUNT macro I'd been using had been pushing 32
bytes, to see if that helped; it didn't. I also found that zeroing out
all 32 bytes was a really good way to trigger bugs.

My hunch is the problem area is somehwere in the
cpu_switch()/idle()/switch_exit()/sw1 machinery, since I consistently
saw traps immediately after the asm() label in tsleep() -- when
running on proc0, so it's probably setting a process up to be reaped.

Also, I haven't done anything for a day or so since my apt is being
painted, and I wont get things back together until tomorrow evening at
best.  i will get the changes to <mips/profile.h> and to use C structs
for trapframe in genassym into the tree then.