Re: [0/4] NUMA-Q: remove PCI bus number mangling

[William Lee Irwin III <wli@holomorphy.com>]

At some point in the past, I wrote:
>> Also, every PCI bridge in my box has a bus number of 3 so the lookup
>> table will produce wrong answers every time.

On Tue, Nov 12, 2002 at 02:46:29PM -0800, Martin J. Bligh wrote:
> Isn't that the local bus number though? The topology functions take
> global bus numbers, which should be unique ...
> M.

That bus number mangling scheme is an instance of an approach vetoed
over a year ago by Dave Miller and others, and does not work for bridges
because arch code does not get the opportunity to mangle the bus number
during bridge discovery.

The inheritance of ->sysdata is the proper method, used by Alpha, PPC,
ARM, and others, and has been generally acknowledged as the acceptable
solution to the PCI domain/segment problem.


Bill