The system boots successfully.  DMESG output is normal until it tries to
	load the hme0 driver.  At this point, the machine locks solid.  STOP-A is
	unresponsive.

	After a cold-boot of the machine, the machine has lost it's MAC address
	and the HOSTID PROM contents are invalid.

	We reset them by:

	set-defaults
	setenv diag-switch? false
	8 0 30 ?? ?? ?? c0ffee mkpl [control-d] [control-r]

	?? being the 3 last octets of my MAC address (which I luckily had written
	down during rarpd configs)

	I can understand a kernel not booting.  I can understand it panic'ing out.  
	But resetting my NVRAM values?  That's pretty rough ;-)

This seems to be a problem with certain machines.  My theory is that it has to do
with how the PROM allocates its mappings.  Seems that it maps in pages contiguously    
regardless of that they are, so data and device mappings (and worse, the stack) are    
all in the same area.  The kernel will also allocate I/O mappings in the same
general area.

Newer kernels have some code changes to hopefully prevent this from happening:

	The NVRAM is write-protected unless it's being updated

	The entire kernel has been relocated so as not to conflict with the PROM.

	(These changes have not made it to 1.5 yet).

In the mean time, the NVRAM FAQ has instructions about recovering from a zapped
NVRAM.

Eduardo