Towards PowerMac SMP G4

Hi all,

The attached patch to prom.c (against 2.2.18pre2-benh)
allows SMP kernels to boot on the SMP PowerMac G4.
But, of course, the second CPU isn't started. The
change is to breakout the CHRP SMP code in
prom_init() as a seperate function and to change
the CHRP test from the simple test for the presence
of a device_type property in the root node, which
the new-world Macs pass, to a test similar to that
in setup.c. This needs to be tested on an SMP CHRP.

I've noticed that /proc/device-tree/cpus has duplicate
directory names for the two CPU's of "PowerPC,G4", rather
then something like "PowerPC,G4@1" for the second.
Printing out the node name and full_name in finish_node()
shows them to have values of "PowerPC,G4" and
"/cpus/PowerPC,G4", respectively, for both CPU's.
Should this be fixed-up during the OF devtree copy
or in procfs?

Is anyone else working on these basic boot and devtree
init cleanups yet (fewer things todo when someone
figures out the startup sequences, or they show up
in Darwin)? If not, I'll continue looking into them
as I have time.

The only functional problem I've run into so
far are is that VT-switching doesn't work once
xpmac is started in anything other than 8-bit depth.
That may have been a result of trying to trim out some
of the fbcon chipset drivers from .config (I'll be
trying a clean build later), but perhaps this is a
general SMP kernel fbcon problem? There are several
messages like the following in dmesg after X startup:

   fbcon_setup: No support for fontwidth 8
   fbcon_setup: type 0 (aux 0, depth 32) not supported


Henry