Re: [PATCH] x86[-64] PCI domain support

[Muli Ben-Yehuda <muli@il.ibm.com>]

On Sat, Sep 30, 2006 at 06:03:11AM -0400, Jeff Garzik wrote:

> Would you also make sure that Andrew has the necessary bits to keep 
> Calgary going under PCI domains?  If it's a patch that sits on top of 
> linux-2.6.git + my patch, I can merge it into misc-2.6.git#pciseg (which 
> automatically goes into -mm).  Otherwise, make sure -mm has the stack of 
> patches necessary.

The patch I posted earlier is all that's needed, if you could merge it
into #pciseg that would be fine. I'm pondering making one small
change though: in your pci domains patch, you have this snippet:

+#ifdef CONFIG_PCI_DOMAINS
+static inline int pci_domain_nr(struct pci_bus *bus)
+{
+	struct pci_sysdata *sd = bus->sysdata;
+	return sd->domain;
+}
+
+static inline int pci_proc_domain(struct pci_bus *bus)
+{
+	return pci_domain_nr(bus);
+}
+#endif /* CONFIG_PCI_DOMAINS */

So pci_domain_nr and pci_proc_domain are only available if
CONFIG_PCI_DOMAINS is defined. I followed suit and make pci_iommu only
available if CONFIG_CALGARY_IOMMU is defined, but perhaps it would be
better to make it unconditional, since ->sysdata will always be
available anyway. Was there a specific reason why pci_domain_nr is
only available if CONFIG_PCI_DOMAINS?

Cheers,
Muli