Axp-List Archive
Re: PCI-PCI bridges mess in 2.4.x

Subject: Re: PCI-PCI bridges mess in 2.4.x
From: Ivan Kokshaysky (ink@jurassic.park.msu.ru)
Date: Mon Nov 6 08:29:30 2000


Cc'ed to axp-list in a hope that people could test this
on alphas with pci-pci bridges - ruffians, noritakes etc.

On Fri, Nov 03, 2000 at 01:16:40AM -0800, Richard Henderson wrote:
> On Fri, Nov 03, 2000 at 11:16:47AM +0300, Ivan Kokshaysky wrote:
> > I'll be gone for two or three days, so hopefully I'll have
> > something for public review next week.
>
> I'll look forward to trying it on the rawhide.

Quite likely there are some bugs in the patch - changes are large...

Some details of this implementation:
allocation of io/mem ranges starts from the root bus to child
buses - reversed to the order you supposed, but technically
it makes no difference. Almost everything is done in a single
pass walking the bus tree - allocating device's resources,
setting up bus ranges and initializing bridge's hardware.

It turned out that except alpha, only arm and mips/ddb5074
are using support functions from drivers/pci/setup-xxx.c.
This patch shouldn't break these platforms in a single-bus
configuration, but to work with bridges they need additional
fixes in the pcibios_fixup_bus()...

Also cy82c693 code cleaned up a bit - moved to quirks and better
documented. And compile fix for CLOCKS_PER_SEC.

Ivan.




This archive was generated by hypermail version 2a22 on Fri Dec 1 08:00:06 2000 PST
Send any problems or questions about this archive to webmaster@alphalinux.org.