630 is for _correctable_ memory errors, but you don't see those
in the panic case because the architecture-level register has
identified it and we've returned. Errors in cache memories are
also lumped in here.

While it would be interesting to decode the platform stuff (on the a12 we
print out the pcb locator for the simm, among other things) NetBSD has
really made the crucial distinction here, between recoverable and non-
recoverable.

The only important refinement I really see here is that some non-correctable
errors effect only the currently running image...so in some cases (but not
this one, apparently) we could just blow away the individual process. There
are also some rare non-correctable i-cache errors that are in fact retryable.

The short version of the story is:

	"I think it's a hardware problem."