Subject: Re: PPC601 support, and macs without openfirmware
To: None <port-macppc@netbsd.org>
From: Jeff Walther <trag@io.com>
List: port-macppc
Date: 11/23/2002 21:29:42
At 09:30 -0800 11/23/2002, David Gatwood wrote:
>On Fri, 22 Nov 2002, Jeff Walther wrote:

>>  Hmmm.  Is there a listing of interrupts and their respective devices
>>  for the 9500 and the 7500/8500?
>
>There's a mostly-complete list in the MkLinux interrupt handling code, but
>even we depend on Open Firmware to flesh it out.  The best thing to do is
>to just walk through the OF tree doing a ".properties" on every node and
>write down all the interrupt numbers you see and what device the interrupt
>is associated with.
>
>
>>  Is the mapping info from Open Firmware something which is coded into
>>  the OF version.  That is, is it hard coded into the ROM?  Or is it
>>  something that OF detects from the motherboard?
>
>If I were gueswsing... probably hard-coded into the ROM... but I'm not
>certain.

Thanks for the information, David.  I appreciate it.

What I'm interested in doing is adding a Bandit to the Umax S900 to 
create three more PCI slots.   This would ordinarily be too 
challenging because of the difficulty of plugging into the data and 
address busses, but the S900 has that Secondary CPU slot, which has 
all but three of the necessary lines.  With that present, it's simple 
enough to design a circuit board for the CPU slot, have a few 
prototypes whipped up for about $140 (Sierra Proto Express No-Touch 
Special on four layer boards) and get the Bandit and Arbiter off of a 
7200 board.   That part is fairly straight forward, because it just 
duplicates the architecture one would find on a 9500.

The other thing I would like to try is to put two Bandits on the 
add-in board, creating a three Bandit machine.  This presents 
uncertainties and difficulties,   one of which is finding interrupts 
for the extra slots.  Another issue is determining which are the 
unused bus arbitration pins on Hammerhead.

Bandit is also capable of driving more than three PCI slots/devices. 
But again, more interrupts and probably some kind of remapping in OF 
would be needed.  So I'm trying to collect what information I can.  I 
think the simple, add a second Bandit with three slots will be 
straight-forward.  Adding more than three slots or a third Bandit is 
where the real hacking will begin.

No, there isn't much practicle use for this project.  It's just fun.


>>  I could move the Grand Central chip from one model machine to another
>>  and see if the recipient board still works properly.
>
>No need to do any major surgery to check that.

It's not major surgery to me and I really enjoy soldering.  :-)  But 
I agree, your way is much simpler and easier.

Jeff Walther