>I believe I've just fixed it (mips/mips_machdep.c rev 1.37).
>It seems that this was caused by use of uninitialized memory
>for the u-area, so it depends on previous use of the machine's
>physical memory.


Ack.  Thanks for fixing this.

Was this a problem for 1.3.2? If it is, could you fix it (looks like
its in pmax/pmax/machdep.c) and submit a pullup request for 1.3.x too,
please?

Hmm... couldn't we just put the initial u-area in the kernel bss,
and forget about it?