RE: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

RE: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Deguara, Joachim]

> -----Original Message-----
> From: Jeff Garzik [mailto:jeff@garzik.org] 
> Sent: 16 May 2006 19:25
> To: Andi Kleen
> Cc: Deguara, Joachim; Linux Kernel Mailing List; Linus 
> Torvalds; Andrew Morton
> Subject: Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300
> 
> Andi Kleen wrote:
> > On Tuesday 16 May 2006 19:12, Jeff Garzik wrote:
> >> Andi Kleen wrote:
> >>> Did you test that? I had two persons with that 
> workstation test all 
> >>> combinations and it worked for them.
> >> Not yet, it's queued for my next test run.
> > 
> > You complained without testing anything? 
> 
> When I first got the box, pci=noacpi made mmconfig space go 
> away, or some other breakage.  If your patch forces that, 
> then logically that condition should reappear by default.
> 
> 	Jeff
> 

The fix worked for me.  But as I noted, the lspci output changed and
going back I see the PCI-X bridge went AWOL, though the the SCSI
controller (part of that bug) is PCI-X non-bridge anyway.  Also I tested
this with the SLES kernel which IIRC has mmconfig off by default.  So
Jeff, I am interested to hear how your testing goes.

-joachim

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

On Tuesday 16 May 2006 20:33, Deguara, Joachim wrote:
> > -----Original Message-----
> > From: Jeff Garzik [mailto:jeff@garzik.org] 
> > Sent: 16 May 2006 19:25
> > To: Andi Kleen
> > Cc: Deguara, Joachim; Linux Kernel Mailing List; Linus 
> > Torvalds; Andrew Morton
> > Subject: Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300
> > 
> > Andi Kleen wrote:
> > > On Tuesday 16 May 2006 19:12, Jeff Garzik wrote:
> > >> Andi Kleen wrote:
> > >>> Did you test that? I had two persons with that 
> > workstation test all 
> > >>> combinations and it worked for them.
> > >> Not yet, it's queued for my next test run.
> > > 
> > > You complained without testing anything? 
> > 
> > When I first got the box, pci=noacpi made mmconfig space go 
> > away, or some other breakage.  If your patch forces that, 
> > then logically that condition should reappear by default.
> > 
> > 	Jeff
> > 
> 
> The fix worked for me.  But as I noted, the lspci output changed and
> going back I see the PCI-X bridge went AWOL, though the the SCSI
> controller (part of that bug) is PCI-X non-bridge anyway.  Also I tested
> this with the SLES kernel which IIRC has mmconfig off by default.  So
> Jeff, I am interested to hear how your testing goes.

But PCI-X reappears when you disable the PCI segmentation in the BIOS
right?

-Andi

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

> The fix worked for me.  But as I noted, the lspci output changed and
> going back I see the PCI-X bridge went AWOL, though the the SCSI
> controller (part of that bug) is PCI-X non-bridge anyway.  Also I tested
> this with the SLES kernel which IIRC has mmconfig off by default.  So
> Jeff, I am interested to hear how your testing goes.

Didn't you guys tell me earlier the machine doesn't boot without
pci=noacpi and segmentation enabled? And now it suddenly does again?

If it works without that option in all cases the patch can be of course 
dropped.

-Andi (confused) 

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Jeff Garzik]

Linux Kernel Mailing List wrote:
> commit 5491d0f3e206beb95eeb506510d62a1dab462df1
> tree 5c4aadcfb4a93535e2f6e0f5977e930ccacec0e9
> parent f0fdabf8bf187c9aafeb139a828c530ef45cf022
> author Andi Kleen <ak@suse.de> Mon, 15 May 2006 18:19:41 +0200
> committer Linus Torvalds <torvalds@g5.osdl.org> Tue, 16 May 2006 21:59:31 -0700
> 
> [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300
> 
> 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.
> 
> Acked-by: len.brown@intel.com
> Cc: gregkh@suse.de
> Signed-off-by: Andi Kleen <ak@suse.de>
> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
> 
>  arch/i386/kernel/acpi/boot.c |    8 ++++++++
>  1 files changed, 8 insertions(+)
> 
> diff --git a/arch/i386/kernel/acpi/boot.c b/arch/i386/kernel/acpi/boot.c
> index 40e5aba..daee695 100644
> --- a/arch/i386/kernel/acpi/boot.c
> +++ b/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"),

Strong NAK.  Please revert.  This majorly screws my primary workstation, 
and many other users with this workstation.

At a minimum, you should test to see if the BIOS has activated PCI 
domain support first!

	Jeff

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

On Tuesday 16 May 2006 18:17, Jeff Garzik wrote:
> Linux Kernel Mailing List wrote:
> > commit 5491d0f3e206beb95eeb506510d62a1dab462df1
> > tree 5c4aadcfb4a93535e2f6e0f5977e930ccacec0e9
> > parent f0fdabf8bf187c9aafeb139a828c530ef45cf022
> > author Andi Kleen <ak@suse.de> Mon, 15 May 2006 18:19:41 +0200
> > committer Linus Torvalds <torvalds@g5.osdl.org> Tue, 16 May 2006 21:59:31 -0700
> > 
> > [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300
> > 
> > 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.
> > 
> > Acked-by: len.brown@intel.com
> > Cc: gregkh@suse.de
> > Signed-off-by: Andi Kleen <ak@suse.de>
> > Signed-off-by: Linus Torvalds <torvalds@osdl.org>
> > 
> >  arch/i386/kernel/acpi/boot.c |    8 ++++++++
> >  1 files changed, 8 insertions(+)
> > 
> > diff --git a/arch/i386/kernel/acpi/boot.c b/arch/i386/kernel/acpi/boot.c
> > index 40e5aba..daee695 100644
> > --- a/arch/i386/kernel/acpi/boot.c
> > +++ b/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"),
> 
> Strong NAK.  Please revert.  This majorly screws my primary workstation, 
> and many other users with this workstation.

Did you test that? I had two persons with that workstation test all combinations
and it worked for them. 

Wait - you have a engineering sample haven't you? I don't think i care about
those. Maybe they behave differently. Use pci=acpi

[Don't try to update the BIOS either - it can be deadly]

> At a minimum, you should test to see if the BIOS has activated PCI 
> domain support first!

It worked in both cases on the production system. Without pci=noacpi
the PCI-X bus couldn't be seen.

-Andi
 

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Jeff Garzik]

Andi Kleen wrote:
> Did you test that? I had two persons with that workstation test all combinations
> and it worked for them. 

Not yet, it's queued for my next test run.


> Wait - you have a engineering sample haven't you? I don't think i care about
> those. Maybe they behave differently. Use pci=acpi
> 
> [Don't try to update the BIOS either - it can be deadly]

Well, its a free sample of a production machine, not a pre-production 
machine.  And the BIOS updates just fine :)

	Jeff

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Andi Kleen]

On Tuesday 16 May 2006 19:12, Jeff Garzik wrote:
> Andi Kleen wrote:
> > Did you test that? I had two persons with that workstation test all combinations
> > and it worked for them. 
> 
> Not yet, it's queued for my next test run.

You complained without testing anything? 

-Andi

Re: [PATCH] i386/x86_64: Force pci=noacpi on HP XW9300

[Jeff Garzik]

Andi Kleen wrote:
> On Tuesday 16 May 2006 19:12, Jeff Garzik wrote:
>> Andi Kleen wrote:
>>> Did you test that? I had two persons with that workstation test all combinations
>>> and it worked for them. 
>> Not yet, it's queued for my next test run.
> 
> You complained without testing anything? 

When I first got the box, pci=noacpi made mmconfig space go away, or 
some other breakage.  If your patch forces that, then logically that 
condition should reappear by default.

	Jeff