re: basic support for (software concept) "pci domains" in the MI pci code

[matthew green <mrg%eterna.com.au@localhost>]

   .. which, for lack of something better, I use in libpciaccess. If DRM  
   could do the same the problem would go away.

OK, this works.

it's very ugly but it's only a few lines of code, and works with your
libpciaccess code as-is:

        parent_dev = kdev;
        do {
                parent_dev = device_parent(parent_dev);
        } while (parent_dev && !device_is_a(parent_dev, "pci"));
        parent_unit = device_unit(parent_dev);
        [ ... ]
        dev->pci_domain = parent_unit;

i withdraw my previous patch.


there's still a story to solve wrt /dev/pci* nodes.  we only create
upto /dev/pci7 and that isn't enough, and this is probably what is
driving cegger's problem.

but this doesn't require any more kernel changes...


.mrg.