ppc platform for AMCC-440EPx - ELDK to the rescue

I finally resolved my nightmare of getting linux-2.6.24/5... to work on my AMCC-440Epx - Sequoia spinoff board. Following a clue I derived from a reply from Wolfgang Denx, I went to the DENX Engineering site where I discoved and downloaded the ELDK for the ppc_4xxFP. the ELDK uses the ppc platform instead of the powerpc platform.  FINALLY the early debugging console output worked, which doesn't work at all using the powerpc platform (and yes, I had the correct port address configured).  After that, I was able to see that it was the "setup_hose" function called from /arch/ppc/platforms/4xx/sequoia.c that caused my board to hang. This Sequoia spinoff board has the PCI Host Bridge disabled which I didn't realize until that point.

Looking back at the powerpc platform, I don't see how to disable the Host Bridge initialization without disabling the entire PCI bus.  This board does have 2 devices hooked on the PCI bus, just no need for a Host Bridge.

Also, I fail to see what the DTS virtualization layer of the hardware buys us. When debugging, it's so much harder to use than the ppc platform's straight forward and standard methodologies.  I can't understand to goal to eliminate the ppc platform, especially given the fact that DENX's ELDK is still using it with Linux 2.6.24.2 !!

Best Regards to all,
Special Thanks to Wolfgang Denx,
Steve Heflin

Steve Heflin wrote:

After that, I was able to see that it was the "setup_hose" function called from /arch/ppc/platforms/4xx/sequoia.c that caused my board to hang. This Sequoia spinoff board has the PCI Host Bridge disabled which I didn't realize until that point.

Looking back at the powerpc platform, I don't see how to disable the Host Bridge initialization without disabling the entire PCI bus.  This board does have 2 devices hooked on the PCI bus, just no need for a Host Bridge.

And how do you propose to talk to said PCI devices without configuring the host bridge?

Also, I fail to see what the DTS virtualization layer of the hardware buys us.

Maintainability.  Ease of porting to new hardware.

When debugging, it's so much harder to use than the ppc platform's straight forward and standard methodologies.  I can't understand to goal to eliminate the ppc platform,

Are you volunteering to maintain it, including the rest of us not having to care when we change something in common code that breaks it?

especially given the fact that DENX's ELDK is still using it with Linux 2.6.24.2 !!

Take that up with Denx.

-Scott

On Wed, 26 Mar 2008 09:08:03 -0400
Steve Heflin [off-list ref] wrote:

 FINALLY the 
early debugging console output worked, which doesn't work at all 
using the powerpc platform (and yes, I had the correct port address 
configured). 

It should work.  It did when the board support was first brought in to
arch/powerpc. If it doesn't now, I consider that a regression that needs
fixing, at least for Sequoia itself.
 

Looking back at the powerpc platform, I don't see how to disable the 
Host Bridge initialization without disabling the entire PCI 
bus.  This board does have 2 devices hooked on the PCI bus, just no 
need for a Host Bridge.

As Scott said, how would you talk to any PCI device without the host
bridge being configured and enabled?

Also, I fail to see what the DTS virtualization layer of the hardware 
buys us. When debugging, it's so much harder to use than the ppc 
platform's straight forward and standard methodologies.  I can't 

There are tradeoffs, yes.

understand to goal to eliminate the ppc platform, especially given 
the fact that DENX's ELDK is still using it with Linux 2.6.24.2 !!

The ELDK uses both arch/ppc and arch/powerpc.  I consider them quite
reasonable people, and foresee them eventually switching to
arch/powerpc over time.  In fact, their efforts are a large part of the
continued progress of 4xx on arch/powerpc.

josh