> That sounds like a tolerable idea, though perhaps said file should
> actually be port-specific (I've not looked at the code).

Right, so we have basically four types of system to support at this time:

- Machines that use vidc

- Machines that are based around ofw, which use dev/cninit.c

- Machines based on footbridge, which already provide their own consinit.

- Machines based on HPCARM, which also provide their own consinit (and 
machdep, though I think this is substantially the same as arm32/machdep.c).

I think that is it 

The machines based on vidc are: RISCPC, A7000, NC and RC7500 plus all the 
derivatives of these.

Machines based on OFW are SHARK and OFWGENCFG (plus the DNARD in a 
separate arch subtree).

Machines based on Footbridge are CATS, EBSA285 and NETWINDER (some of 
which are in separate directories).

So, I propose two new files, arm32/vidc/console/consinit.c and 
arm32/ofw/consinit.c, providing the consinit function for VIDC- and 
OFW-based machines respectively.  Footbridge can continue to do its own 
thing for the time being.

Once this structure is in place, the actual source file changes are 
trivial, so I think it's only the principles that need agreement.

I've built kernels for CATS (separate arch subtree), RISCPC & SHARK 
(though not booted, since I don't have hardware for all these).  I've also 
tried to build kernels for NETWINDER, DNARD and RC7500; but none of these 
compile at this time for unrelated reasons.

Is everyone happy with this?

R.