Re: [patch] pci: revert "PCI: remove transparent bridge sizing"

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Wed, 2008-03-26 at 15:47 -0700, Linus Torvalds wrote:
> > I was talking about IO not memory mostly here.
> 
> Yeah, low IO is also reserved on PC's (the low 256 IO ports are 
> motherboard resources and contain stuff like legacy DMA channel setup 
> etc)

Sure but can't that be in a kind of southbridge ? Like HT or PCIe
segment out of the CPU gets through a virtual P2P wich then hits the
"legacy" combo blob masquerading as a PCI device ? I remember seeing
that sort of thing in the past and I -think- it was some kind of x86
chipset hijacked on powerpc... 

> You could imagine having it behind a PCI bridge, but in practice it's 
> always on the NB/SB (and if you want to support some of the odder
> things 
> like the NMI reason and the i387 error ports, they pretty much have
> to 
> be - it would be insane to make a special PCI chips on a separate bus 
> that does things like that).

It's often all virtual inside a single chip.

Anyway, doesn't matter much at this stage I suppose, but it would be
nice to not use 0 as meaning invalid when sizing bridge windows and I'm
not sure at all about using "start" as an alignment indicator neither...
It will be much over-aligned in some cases, adding constraints to the
allocator where we didn't have any before no ?

Ben.