This is a heads-up...

In preparation for 1.4, and with hopes of supporting the DEcsystem
5100, I've just ifdef'ed out the locore_mips1.S code which implicitly
waited for a CPU write-buffer drain every time the Icache or Dcache
was flushed.

Other mips ports wire up the writebuffer lines differently; this code
was already only used on pmaxes. If the pmax device drivers have
wbflush() in all the right places, this should be a no-op.

This *should* be innocuous, since there are few places where the
I-cache or D-cache is flushed, and I haven't spotted any where there's
a possibly-missing wbflush() call.

Still, if people running current on mips1 CPUs could bang on things
over the next couple of weeks, that'd be great.