On Fri, Jun 21, 2002 at 10:17:25AM -0700, Jason R Thorpe wrote:
> On Sat, Jun 22, 2002 at 01:54:20AM +0900, MOCHIDA Shuji wrote:
> 
>  >   FreeBSD:
>  >     motherboard -+- pcib0 - pci0 -+- fxp0
>  >                  |               (+- etc.)
>  >                  |
>  >                  +- pcib1 - pci1 -+- ahc0
>  >                                   +- ahc1
>  >                                  (+- etc.)
>  > 
>  > 
>  >   NetBSD(with Masanori Kanaoka's patch):
>  >     mainbus0 - pci0 -+- pchb0 - pci1 -+- pchb1 (- nothing)
>  >                      +- fxp0          +- ahc0
>  >                     (+- etc.)         +- ahc1
>  >                                      (+- etc.)
>  > 
>  >  ...strange.
> 
> Not that strange -- it's an artifact of how FreeBSD does PCI
> configuration.  The way we do it is actually more logical,
> since we follow the way the PCI devices actually present
> themselves.

And if they are the motherboards I'm thinking of, the they
physically match the 'freebsd' picture (ie data for bus 1
doesn't go anywhere near bus o) but have some odd logic
signal lurking so that they logically look like the
'netbsd' picture.

OTOH I don't know havethe bridge 'bus-number' registers have
to be set in order to get configuration cycles passed down to
subsiduary busses....

	David

-- 
David Laight: david@l8s.co.uk