Patchwork PCI / ACPI: Do not request control of ASPM if the BIOS has disabled it

login
register
mail settings
Submitter Rafael J. Wysocki
Date May 24, 2012, 8:45 p.m.
Message ID <201205242245.45888.rjw@sisk.pl>
Download mbox | patch
Permalink /patch/161201/
State Not Applicable
Headers show

Comments

Matthew Garrett - May 24, 2012, 8:45 p.m.
On Thu, May 24, 2012 at 10:45:45PM +0200, Rafael J. Wysocki wrote:

> However, knowing that the BIOS doesn't support ASPM, we shouldn't
> request the control of it, so we should remove
> (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> from the flags before calling acpi_pci_osc_control_set() in those
> cases.  Failing to do so causes the evaluation of _OSC for the PCI
> root bridge to return error codes on some systems where it should
> work correctly.

Is there an example of an affected machine's DSDT?
Rafael J. Wysocki - May 24, 2012, 8:45 p.m.
From: Rafael J. Wysocki <rjw@sisk.pl>

We want to report that the kernel supports ASPM to the BIOS even if
the BIOS signals us that it doesn't.  So, we need the flags to include
(OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
before calling acpi_pci_osc_support(root, flags) in
acpi_pci_root_add() (unless there's a command line telling us that
ASPM is not to be supported).

However, knowing that the BIOS doesn't support ASPM, we shouldn't
request the control of it, so we should remove
(OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
from the flags before calling acpi_pci_osc_control_set() in those
cases.  Failing to do so causes the evaluation of _OSC for the PCI
root bridge to return error codes on some systems where it should
work correctly.

Reported-and-tested-by: Wei Kong <wkong@novell.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---

-stable material, I think.

Thanks,
Rafael

---
 drivers/acpi/pci_root.c |    4 ++++
 1 file changed, 4 insertions(+)

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jean Delvare - May 25, 2012, 11:53 a.m.
Hi Matthew,

Le jeudi 24 mai 2012 à 21:45 +0100, Matthew Garrett a écrit :
> On Thu, May 24, 2012 at 10:45:45PM +0200, Rafael J. Wysocki wrote:
> 
> > However, knowing that the BIOS doesn't support ASPM, we shouldn't
> > request the control of it, so we should remove
> > (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> > from the flags before calling acpi_pci_osc_control_set() in those
> > cases.  Failing to do so causes the evaluation of _OSC for the PCI
> > root bridge to return error codes on some systems where it should
> > work correctly.
> 
> Is there an example of an affected machine's DSDT?

Attached.

Jean
Bjorn Helgaas - Aug. 15, 2012, 7:47 p.m.
On Thu, May 24, 2012 at 2:45 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> From: Rafael J. Wysocki <rjw@sisk.pl>
>
> We want to report that the kernel supports ASPM to the BIOS even if
> the BIOS signals us that it doesn't.  So, we need the flags to include
> (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> before calling acpi_pci_osc_support(root, flags) in
> acpi_pci_root_add() (unless there's a command line telling us that
> ASPM is not to be supported).
>
> However, knowing that the BIOS doesn't support ASPM, we shouldn't
> request the control of it, so we should remove
> (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> from the flags before calling acpi_pci_osc_control_set() in those
> cases.  Failing to do so causes the evaluation of _OSC for the PCI
> root bridge to return error codes on some systems where it should
> work correctly.
>
> Reported-and-tested-by: Wei Kong <wkong@novell.com>
> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
> ---
>
> -stable material, I think.

Len, since you've applied recent patches in this area, can you take
care of this, too?

> ---
>  drivers/acpi/pci_root.c |    4 ++++
>  1 file changed, 4 insertions(+)
>
> Index: linux/drivers/acpi/pci_root.c
> ===================================================================
> --- linux.orig/drivers/acpi/pci_root.c
> +++ linux/drivers/acpi/pci_root.c
> @@ -574,6 +574,10 @@ static int __devinit acpi_pci_root_add(s
>         if (flags != base_flags)
>                 acpi_pci_osc_support(root, flags);
>
> +       if (!pcie_aspm_enabled())
> +               flags &= ~(OSC_ACTIVE_STATE_PWR_SUPPORT |
> +                       OSC_CLOCK_PWR_CAPABILITY_SUPPORT);
> +
>         if (!pcie_ports_disabled
>             && (flags & ACPI_PCIE_REQ_SUPPORT) == ACPI_PCIE_REQ_SUPPORT) {
>                 flags = OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Rafael J. Wysocki - Aug. 15, 2012, 8:03 p.m.
On Wednesday, August 15, 2012, Bjorn Helgaas wrote:
> On Thu, May 24, 2012 at 2:45 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > From: Rafael J. Wysocki <rjw@sisk.pl>
> >
> > We want to report that the kernel supports ASPM to the BIOS even if
> > the BIOS signals us that it doesn't.  So, we need the flags to include
> > (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> > before calling acpi_pci_osc_support(root, flags) in
> > acpi_pci_root_add() (unless there's a command line telling us that
> > ASPM is not to be supported).
> >
> > However, knowing that the BIOS doesn't support ASPM, we shouldn't
> > request the control of it, so we should remove
> > (OSC_ACTIVE_STATE_PWR_SUPPORT | OSC_CLOCK_PWR_CAPABILITY_SUPPORT)
> > from the flags before calling acpi_pci_osc_control_set() in those
> > cases.  Failing to do so causes the evaluation of _OSC for the PCI
> > root bridge to return error codes on some systems where it should
> > work correctly.
> >
> > Reported-and-tested-by: Wei Kong <wkong@novell.com>
> > Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
> > ---
> >
> > -stable material, I think.
> 
> Len, since you've applied recent patches in this area, can you take
> care of this, too?

No, please don't.

I'd discussed that with Matthew and we concluded that it would break things
sometimes.  Please disregard it (and sorry for not saying that earlier).

Thanks,
Rafael


> > ---
> >  drivers/acpi/pci_root.c |    4 ++++
> >  1 file changed, 4 insertions(+)
> >
> > Index: linux/drivers/acpi/pci_root.c
> > ===================================================================
> > --- linux.orig/drivers/acpi/pci_root.c
> > +++ linux/drivers/acpi/pci_root.c
> > @@ -574,6 +574,10 @@ static int __devinit acpi_pci_root_add(s
> >         if (flags != base_flags)
> >                 acpi_pci_osc_support(root, flags);
> >
> > +       if (!pcie_aspm_enabled())
> > +               flags &= ~(OSC_ACTIVE_STATE_PWR_SUPPORT |
> > +                       OSC_CLOCK_PWR_CAPABILITY_SUPPORT);
> > +
> >         if (!pcie_ports_disabled
> >             && (flags & ACPI_PCIE_REQ_SUPPORT) == ACPI_PCIE_REQ_SUPPORT) {
> >                 flags = OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL
> 
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

Index: linux/drivers/acpi/pci_root.c
===================================================================
--- linux.orig/drivers/acpi/pci_root.c
+++ linux/drivers/acpi/pci_root.c
@@ -574,6 +574,10 @@  static int __devinit acpi_pci_root_add(s
 	if (flags != base_flags)
 		acpi_pci_osc_support(root, flags);
 
+	if (!pcie_aspm_enabled())
+		flags &= ~(OSC_ACTIVE_STATE_PWR_SUPPORT |
+			OSC_CLOCK_PWR_CAPABILITY_SUPPORT);
+
 	if (!pcie_ports_disabled
 	    && (flags & ACPI_PCIE_REQ_SUPPORT) == ACPI_PCIE_REQ_SUPPORT) {
 		flags = OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL