RE: [PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

RE: [PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

[Brown, Len]

>The system has multiple PCI segments and we don't handle that properly
>yet in PCI and ACPI. Short term before this is fixed blacklist it to
>pci=noacpi.

I'm okay with the patch, but it makes me wonder...

Is this the 1st/only system Linux has run on with multiple PCI segments?
What are your expectations for where "short-term" ends and "long-term"
begins?

thanks,
-Len

Re: [PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

On Monday 15 May 2006 18:47, Brown, Len wrote:
> 
> >The system has multiple PCI segments and we don't handle that properly
> >yet in PCI and ACPI. Short term before this is fixed blacklist it to
> >pci=noacpi.
> 
> I'm okay with the patch, but it makes me wonder...
> 
> Is this the 1st/only system 

x86-

IA64/PA-RISC support subdomains successfully

> Linux has run on with multiple PCI segments? 

I think IBM summit somehow uses it too.

And there is a patch from Jeff Garzik I think to make the xw9300 subdomains 
work (or rather implement subdomain support in arch/i386/pci/*), but it 
breaks the Summit and possibly other non x86 systems. Greg 
should know details about that.

As usual the systems usually boot even without this patch, but you can't 
reach all PCI devices.

> What are your expectations for where "short-term" ends and "long-term"
> begins?

I think Greg has Jeff's patch still queued somewhere, but it needs
to be debugged to work everywhere. After that is done we can drop the blacklist 
entry. Hopefully for 2.6.19?

For 2.6.17 I don't see any alternative to blacklisting.

-Andi

Re: [PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

[Jeff Garzik]

Andi Kleen wrote:
> On Monday 15 May 2006 18:47, Brown, Len wrote:
>>> The system has multiple PCI segments and we don't handle that properly
>>> yet in PCI and ACPI. Short term before this is fixed blacklist it to
>>> pci=noacpi.
>> I'm okay with the patch, but it makes me wonder...
>>
>> Is this the 1st/only system 
> 
> x86-
> 
> IA64/PA-RISC support subdomains successfully
> 
>> Linux has run on with multiple PCI segments? 
> 
> I think IBM summit somehow uses it too.
> 
> And there is a patch from Jeff Garzik I think to make the xw9300 subdomains 
> work (or rather implement subdomain support in arch/i386/pci/*), but it 
> breaks the Summit and possibly other non x86 systems. Greg 
> should know details about that.
> 
> As usual the systems usually boot even without this patch, but you can't 
> reach all PCI devices.
> 
>> What are your expectations for where "short-term" ends and "long-term"
>> begins?
> 
> I think Greg has Jeff's patch still queued somewhere, but it needs
> to be debugged to work everywhere. After that is done we can drop the blacklist 
> entry. Hopefully for 2.6.19?
> 
> For 2.6.17 I don't see any alternative to blacklisting.
> 
> -Andi
> 
> 


FWIW I also maintain this "PCI domain support for x86" stuff in a 
separate branch, which I keep up to date:

'pciseg' branch of
  git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/misc-2.6.git

Haven't done anything substantive since the original implementation, 
just kept up with recent kernels.

The xw9300 continues to be my primary workstation (2x2 Opteron), but 
haven't found time to mess with PCI domains in a while.  The BIOS gives 
me the option to disable PCI domains, so I took the slack way out :)

	Jeff

Re: [PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

[Greg KH]

On Mon, May 15, 2006 at 12:47:31PM -0400, Brown, Len wrote:
> 
> >The system has multiple PCI segments and we don't handle that properly
> >yet in PCI and ACPI. Short term before this is fixed blacklist it to
> >pci=noacpi.
> 
> I'm okay with the patch, but it makes me wonder...
> 
> Is this the 1st/only system Linux has run on with multiple PCI segments?
> What are your expectations for where "short-term" ends and "long-term"
> begins?

Jeff's older patches caused some bad problems on some odd NUMA machines,
so I dropped them.  Someone needs to take the time and clean them all up
and get them working properly and included into the tree.

Right now, it's very low on my todo list as I don't have such a machine.

thanks,

greg k-h

[PATCH for 2.6.17] [3/5] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

This is needed to see all devices.

The system has multiple PCI segments and we don't handle that properly
yet in PCI and ACPI. Short term before this is fixed blacklist it to
pci=noacpi.

Cc: len.brown@intel.com
Cc: gregkh@suse.de

Signed-off-by: Andi Kleen <ak@suse.de>

---
 arch/i386/kernel/acpi/boot.c |    8 ++++++++
 1 files changed, 8 insertions(+)

Index: linux/arch/i386/kernel/acpi/boot.c
===================================================================
--- linux.orig/arch/i386/kernel/acpi/boot.c
+++ linux/arch/i386/kernel/acpi/boot.c
@@ -1066,6 +1066,14 @@ static struct dmi_system_id __initdata a
 		     DMI_MATCH(DMI_PRODUCT_NAME, "TravelMate 360"),
 		     },
 	 },
+	{
+	 .callback = disable_acpi_pci,
+	 .ident = "HP xw9300",
+	 .matches = {
+		    DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
+		    DMI_MATCH(DMI_PRODUCT_NAME, "HP xw9300 Workstation"),
+		    },
+	},
 	{}
 };