Created attachment 103171 [details]
dmesg bootlog #0 - Seems OK

I opened this ticket on request by Björn Helgaas because of a mail thread where I reported PCI bus assignment issues on PA-RISC. 

The mail thread was titled 
"[PATCH] parisc/PCI: lba: fix: convert to pci_create_root_bus() for correct root bus resources" in May 2013.
Read: http://www.spinics.net/lists/linux-pci/msg22680.html:

Björn:
> Good, glad that works.
> We *ought* to be able to do better, e.g., by just ignoring cards that we
> can't use.
>
> If you have a chance, you might collect the complete dmesg log from
> both the working configuration and the broken one, and attach them to
> a new kernel.org PCI bugzilla.  Maybe we can ponder them and figure
> something out.
> 
> I suspect the problem is related to the fact that we don't really
> enforce the host bridge apertures reported by the platform.  But we
> might be able to tighten that up, or do something specific to parisc
> that would help this situation.

Short summary for the attached dmeg logs:
- This is a parisc machine (C3000)
- In the PCI slots there are one "HP Visualize EG PCI card" and a "Vodoo gfx".
- In the logs, "stifb" is the driver for the Visualize card.
- In the logs, "sstfb" is the driver for the Vodoo card.
- It would be nice if both cards gets activated (fb0 and fb1)
- I kept the stifb in SLOT4, while plugging the Vodoo in PCI1 or PCI5 slots.

- dmesg.0: seems OK. 
- dmesg.3 has an "no compatible bridge window for [mem 0xfa000000-0xfbffffff]" which leads to "sti 0000:01:04.0: device not available (can't reserve [mem 0xfa000000-0xfbffffff])"

- I think dmesg.1,2 and .4 are not important, but maybe you see something?

I'm willing to do some more testing if required.