t's a bit embarrasing that some of this was undone until recently, but we now

t's a bit embarrasing that some of this was undone until recently, but we now
correctly communicate the boot-time memory map to kernel land.  In question is
still how the ppcmmu idea of free pages and the ntoskrnl one will coincide.

Lots of things were adapted in ofwldr as ntoskrnl and freeldr changed a bit
since this was forked.  Also, some hacks were in here to allow ofwldr to be
ignorant of how some things were mapped.  That's fixed.

Many stupid and obvious bugs have been fixed.

I've basically gutted a lot of HAL stuff, so it isn't committed here.

I had mistakenly used the same array size as x86 did in the ports array in
kdcom.  This is corrected.

I added in some support for boot time device detection (sadly, I haven't yet
decided what will become of openfirmware properties yet).

Much has been done to simplify the mess that mboot became.

svn path=/trunk/; revision=31625