Me and Maki have hit a dead end in our work.  Basically, we are seeing clock
interrupts (INT4) *before* the clock is being enabled by initclock().
This is what's causing _splnone() to hang in cpu_configure().  When it enables
all interrupts, so many clock interrupts have been queued up as to overflow
and hang the kernel.  We know machdep.c is initially disabling all interrupts,
and we know that arc_trap.c:set_intr() is enabling all interrupts except INT4,
so we shouldn't be seeing this.  But we are.  We've gone over the boot process
with a fine-toothed comb, and manually traced through the whole clock attach
and initialization process.  At this point, we're guessing that the "R4030"
in the Magnum 4000SC somehow works differently than the 4000PC and compatibles.
Our suspicion is with R4030_SYS_EXT_IMASK, since it doesn't seem to be working
the way it claims to be.  We're supposed to be able to mask external interrupts
with this register; in our case INT4; but they are still occuring despite
masking.
Does anyone on the list have access to the Magnum 4000SC tech manual, so we
can verify the register settings of the R4030/R4230?  I can provide more 
details if necessary.

Again, any help is much appreciated.

Cheers,
-Mark