On 3/11/07, Jordan Crouse <[EMAIL PROTECTED]> wrote:
On 10/03/07 20:31 -0500, Albert Cahalan wrote:
> When faced with this, I'd just provide config space accessor functions
> that make things look normal. See arch/i386/pci for where you set that.
> Call something that does:
>
> raw_pci_ops = pci_geode_access;
You are way off. You are suggesting that the Geode cannot access
the config space, which is incorrect - it can either access it through
CF8/CFC when the VSA is included, or through Mitch's wrapper when it is
Nope. I'm not suggesting that, though I did think VSA was
the only way. Actually, I take that back, because VSA and
Mitch's wrapper are not hardware.
not. The real problem is that we have four devices with distinct drivers
that are all enumerated under one device - which doesn't fit into the
Linux pci driver model.
As soon as you change the config space accessor functions,
this need not be true. You can make hide devices from the rest
of the kernel. You can make the rest of the kernel see devices
that don't really exist. You can make the rest of the kernel see
smaller BARs, reordered BARs, altered capabilities, etc.
Most importantly, you can split 1 device into 4.
_______________________________________________
Devel mailing list
[email protected]
http://mailman.laptop.org/mailman/listinfo/devel
