>> The issue was spotted by Ethan Solomita, after NetBSD kernel was changed
>> to have kthreads for various purposes, and fixed by me.
>> 
>> From CVS commit log of NetBSD/pmax machdep.c;
>
> Ah, I didn't realize it was done in the port-specific code... I kept 
> looking in generic MIPS code.  This explains why I couldn't find it.
>
> My second question is 'why?'.  Why was this done in port-specific code
> and not in the general mips code?

Well, first of all, it's a sub-optimal solution to fix "kthread left
in splhigh condition" by mucking cpu_fork().  The true source resides
in pcb0.  Some CPU architectures require to have a set of bit pattern
1 to make processes spl0.

It was natural to add a single line initialize proc0 pcb in init() path.

I could argue here about how to _implement_ and _organize_ common-ness,
but the arguement is easily diversified into ruin, so I avoid.

Tohru Nishimura