In a few days time I intend to commit a fairly invasive set of changes
to many files in arch/sparc.  The changes designed to get a grip on
the multitude of sun4m processor variants, but affect the sun4 and sun4c
code as well. Since I cannot test on every type of machine on which
NetBSD/sparc is currently known to work, this may cause some lossage
on a few machines. However, I'm confident this can be sorted out
quickly.

Here's a birds-eye view of what's going to happen:

	- CPU detection: converted to hybrid table-driven form allowing
	  the major players (sun4, sun4c, microsparc,supersparc, etc.)
	  to sorted into broad classes. Fine tuning of attributes can
	  be done by individual routines attached to each "class".

	- Collect a bunch of CPU/module specific routines (like cache
	  handling) in function pointers which replace many
	  "if (footype == BAR)"-like constructs.  Improves effiency
	  in some cases (not all); definitely improves clarity.

	- Location of the above per-CPU information is located at a fixed
	  virtual address. The idea is to have direct access to from any
	  CPU in a SMP system, without costly de-referencing through
	  pointers.

	- pmap_bootstrap: simplified in some spots + some preparations
	  for SMP (e.g. to implement the previous item). Still lots of
	  work left.

-pk