Re: [PATCH] 2.6 SGI Altix I/O code reorganization

[Colin Ngam]

Matthew Wilcox wrote:

> On Wed, Oct 06, 2004 at 01:48:32PM -0700, Grant Grundler wrote:
> > Agreed. I'm not real clear on why drivers/acpi didn't do that.
> > But apperently ACPI supports many methods to PCI or PCI-Like (can you
> > guess I'm not clear on this?) config space. raw_pci_ops supports
> > multiple methods in i386. ia64 only happens to use one so far.
> > It seems right for SN2 to use this mechanism if it needs a different
> > method.
> >
> > Willy tried to explain this to me yesterday and I thought I understood
> > most of it...apperently that was a transient moment of clarity. :^/
>
> Let's try it again, by email this time.
>
> Fundamentally, there is a huge impedence mismatch between how the ACPI
> interpreter wants to access PCI configuration space, and how Linux wants
> to access PCI configuration space.  Linux always has at least a pci_bus
> around, if not a pci_dev.  So we can use dev->bus->ops to abstract the
> architecture-specific implementation of "how do I get to configuration
> space for this bus?"
>
> ACPI doesn't have a pci_bus.  It just passes around a struct of { domain,
> bus, dev, function } and expects the OS-specific code to determine what
> to do with it.  The original hacky code constructed a fake pci_dev on the
> stack and called the regular methods.  This broke ia64 because we needed
> something else to be valid (I forget what), so as part of the grand "get
> ia64 fully merged upstream" effort, I redesigned the OS-specific code.
>
> Fortunately, neither i386 nor ia64 actually need the feature Linux has
> to have a per-bus pci_ops -- it's always the same.  I think powerpc is
> the only architecture that needs it.  So I introduced a pci_raw_ops that
> both ACPI and a generic pci_root_ops could call.
>
> The part I didn't seem to be able to get across to you yesterday was
> that pci_root_ops is not just used for the PCI root bridge, it's used
> for accessing every PCI device underneath that root bridge.

Hi Guys,

Therefore, would it be perfectly fine if we remove the static from pci_root_ops
so that we can use it outside of pci/pci.c??  I can include this in a follow-on
patch.

Thanks.

colin

>
>
> --
> "Next the statesmen will invent cheap lies, putting the blame upon
> the nation that is attacked, and every man will be glad of those
> conscience-soothing falsities, and will diligently study them, and refuse
> to examine any refutations of them; and thus he will by and by convince
> himself that the war is just, and will thank God for the better sleep
> he enjoys after this process of grotesque self-deception." -- Mark Twain

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/