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

From: Jeff Garzik
Date: Thu Sep 28 2006 - 19:04:05 EST

Muli Ben-Yehuda wrote:
On Thu, Sep 28, 2006 at 05:45:41AM -0400, Jeff Garzik wrote:
Muli Ben-Yehuda wrote:
On Tue, Sep 26, 2006 at 03:15:08PM -0400, Jeff Garzik wrote:
The x86[-64] PCI domain effort needs to be restarted, because we've got
machines out in the field that need this in order for some devices to

This breaks the Calgary IOMMU, since it uses sysdata for other
purposes (going back from a bus to its IO address space). I'm looking
into it.
You'll need to modify struct pci_sysdata in include/asm-{i386,x86_64}/pci.h to include the data that you previously stored directly into the sysdata pointer.

Something like this should do the trick. Note - this should not be
applied yet - after several gigabytes of network and disk activity it
takes aic94xx down. More investigation required.

hmmmm. What kernels did you test?

I would suggest testing

jgarzik/misc-2.6.git#master -> vanilla Linux kernel
jgarzik/misc-2.6.git#pciseg -> #master + PCI domain support
#pciseg + your patch

That should narrow down the problems. A problem with aic94xx sorta sounds like something unrelated.

diff -Naurp -X /home/muli/w/dontdiff pci-domains/arch/x86_64/kernel/pci-calgary.c linux/arch/x86_64/kernel/pci-calgary.c
--- pci-domains/arch/x86_64/kernel/pci-calgary.c 2006-09-28 13:31:14.000000000 +0300
+++ linux/arch/x86_64/kernel/pci-calgary.c 2006-09-28 13:14:38.000000000 +0300

ACK patch

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at