MVME5500 status, and on to PMC interrupts

[Peter Dufault]

On Mar 1, 2005, at 7:00 AM, Peter Dufault wrote:

>  I'll send a summary in a little while after I go over the code to see 
> what it is doing and what it isn't.

Here's an ASCII art sketch.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: draw.txt
URL: <http://lists.rtems.org/pipermail/users/attachments/20050301/4749d82a/attachment-0001.txt>
-------------- next part --------------


The numbers in parenthesis are the PCI bus/slot/function codes, beneath 
them are the assignments for PCI IRQA,B,C,D.  Note that the VME 
interface and the PMCSPAN slots do share interrupts.

Here's my take on what's up.  I'm not a PCI expert, so don't be 
surprised if I'm wrong.

The current code in mvme5500/pci doesn't do any bridge discovery or 
setup, and neither does the MOTBoot boot code.  I plan to change the 
code in mvme5500/pci to setup the PMCSPAN bridge, but not to do any 
general fix.  Since a PMCSLOT could have a device with a PCI bridge in 
it this is only a stopgap.

I think what has to happen long-term is to setup the two busses using 
the code in powerpc/shared/pci, only keeping the "PCIX" stuff in 
mvme5500 (the code that needs to know about the two separate busses) 
and calling into the shared PCI code for the bus discovery and setup.  
That code will also need to twiddle the INT numbers returned from the 
shared code to go from PCI interrupts to the way they feed into the 
mvme5500 board.

Peter

Peter Dufault
HD Associates, Inc.