Skip to contentSkip to conversation

Intel pci device quirk remove D3 delay

PATCH In review Todd E Brandt 4 participants 6 messages Sep 10, 2013 – Sep 25, 2013 PCI

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Intel pci device quirk remove D3 delay

The latest Ivy Bridge Intel chipsets have a hardware optimization which allows on-chip PCI devices to ignore the 10ms delay before entering or exitting D3 suspend.

This patch implements the optimization as a pci quirk, since we want tight control over which devices use it. This way we can test each device individually to be sure there are no issues before we enable the quirk. The first set of devices are from the Haswell platform, which includes every PCI device that is on the northbridge and southbridge.

After testing this patch reduces the haswell suspend time from 93 ms to 47 ms and resume time from 160 ms to 64 ms.

 drivers/pci/quirks.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/pci/quirks.c +23 -0
index e85d230..98ea960 100644
@@ -2955,6 +2955,29 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0102, disable_igfx_irq);
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x010a, disable_igfx_irq);
 
 /*
+ * PCI devices which are on Intel chips can skip the 10ms delay
+ * before entering D3 mode.
+ */
+static void quirk_remove_d3_delay(struct pci_dev *dev)
+{
+	dev->d3_delay = 0;
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c00, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0412, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c0c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c31, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3a, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c2d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c20, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c18, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c1c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c26, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c4e, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c02, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c22, quirk_remove_d3_delay);
+
+/*
  * Some devices may pass our check in pci_intx_mask_supported if
  * PCI_COMMAND_INTX_DISABLE works though they actually do not properly
  * support this feature.
Acked-by on v1 1/1
#1 lore raw
Quoted 10 lines · Todd E Brandt

The latest Ivy Bridge Intel chipsets have a hardware optimization which allows on-chip PCI devices to ignore the 10ms delay before entering or exitting D3 suspend.

This patch implements the optimization as a pci quirk, since we want tight control over which devices use it. This way we can test each device individually to be sure there are no issues before we enable the quirk. The first set of devices are from the Haswell platform, which includes every PCI device that is on the northbridge and southbridge.

Quoted 4 lines

After testing this patch reduces the haswell suspend time from 93 ms to 47 ms and resume time from 160 ms to 64 ms.

Quoted 42 lines
 drivers/pci/quirks.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/pci/quirks.c +23 -1
index e85d230..98ea960 100644
@@ -2955,6 +2955,29 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0102, disable_igfx_irq);
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x010a, disable_igfx_irq);
 
 /*
+ * PCI devices which are on Intel chips can skip the 10ms delay
+ * before entering D3 mode.
+ */
+static void quirk_remove_d3_delay(struct pci_dev *dev)
+{
+	dev->d3_delay = 0;
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c00, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0412, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c0c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c31, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3a, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c2d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c20, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c18, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c1c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c26, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c4e, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c02, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c22, quirk_remove_d3_delay);
+
+/*
  * Some devices may pass our check in pci_intx_mask_supported if
  * PCI_COMMAND_INTX_DISABLE works though they actually do not properly
  * support this feature.
 
--

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

Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
maintainer
#2 lore raw
In reply to Todd E Brandt
Quoted 1 line · Todd E Brandt

The latest Ivy Bridge Intel chipsets have a hardware optimization which

You said "Ivy Bridge" above but use "Haswell" below. Let me know if anything needs to be corrected here.

Quoted 9 lines

allows on-chip PCI devices to ignore the 10ms delay before entering or exitting D3 suspend.

This patch implements the optimization as a pci quirk, since we want tight control over which devices use it. This way we can test each device individually to be sure there are no issues before we enable the quirk. The first set of devices are from the Haswell platform, which includes every PCI device that is on the northbridge and southbridge.

Quoted 7 lines

After testing this patch reduces the haswell suspend time from 93 ms to 47 ms and resume time from 160 ms to 64 ms.

 drivers/pci/quirks.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

I added Rafael's ack and applied this for v3.13 (but I can still make corrections if required).

Let me know if you want this in v3.12 instead.

Bjorn

Quoted 36 lines
drivers/pci/quirks.c +23 -0
index e85d230..98ea960 100644
@@ -2955,6 +2955,29 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0102, disable_igfx_irq);
 DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x010a, disable_igfx_irq);
 
 /*
+ * PCI devices which are on Intel chips can skip the 10ms delay
+ * before entering D3 mode.
+ */
+static void quirk_remove_d3_delay(struct pci_dev *dev)
+{
+	dev->d3_delay = 0;
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c00, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0412, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0c0c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c31, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3a, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c3d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c2d, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c20, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c18, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c1c, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c26, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c4e, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c02, quirk_remove_d3_delay);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x8c22, quirk_remove_d3_delay);
+
+/*
  * Some devices may pass our check in pci_intx_mask_supported if
  * PCI_COMMAND_INTX_DISABLE works though they actually do not properly
  * support this feature.
maintainer
#3 lore raw
Quoted 5 lines · Bjorn Helgaas
Quoted 1 line · Todd E Brandt

The latest Ivy Bridge Intel chipsets have a hardware optimization which

You said "Ivy Bridge" above but use "Haswell" below. Let me know if anything needs to be corrected here.

Reading this again, the text really doesn't make sense as-is. So please confirm what this should say.

It sure looks like it should say "The latest *Haswell* Intel chipsets ..."

If you really mean "Ivy Bridge," then you should say something about how Haswell is related to Ivy Bridge. No doubt this is all obvious to Intel folks, but it's not to me :)

Bjorn

Quoted 11 lines
Quoted 11 lines

allows on-chip PCI devices to ignore the 10ms delay before entering or exitting D3 suspend.

This patch implements the optimization as a pci quirk, since we want tight control over which devices use it. This way we can test each device individually to be sure there are no issues before we enable the quirk. The first set of devices are from the Haswell platform, which includes every PCI device that is on the northbridge and southbridge.

After testing this patch reduces the haswell suspend time from 93 ms to 47 ms and resume time from 160 ms to 64 ms.

#4 lore raw
Quoted 7 lines · Bjorn Helgaas
Quoted 5 lines · Bjorn Helgaas
Quoted 1 line · Todd E Brandt

The latest Ivy Bridge Intel chipsets have a hardware optimization which

You said "Ivy Bridge" above but use "Haswell" below. Let me know if anything needs to be corrected here.

Quoted 8 lines

Reading this again, the text really doesn't make sense as-is. So please confirm what this should say.

It sure looks like it should say "The latest *Haswell* Intel chipsets ..."

If you really mean "Ivy Bridge," then you should say something about how Haswell is related to Ivy Bridge. No doubt this is all obvious to Intel folks, but it's not to me :)

it should read "Haswell".

maintainer
#5 lore raw
Quoted 17 lines · Arjan van de Ven
Quoted 16 lines · Bjorn Helgaas
Quoted 8 lines · Bjorn Helgaas
Quoted 2 lines · Todd E Brandt

The latest Ivy Bridge Intel chipsets have a hardware optimization which

You said "Ivy Bridge" above but use "Haswell" below. Let me know if anything needs to be corrected here.

Reading this again, the text really doesn't make sense as-is. So please confirm what this should say.

Quoted 8 lines
Quoted 5 lines

It sure looks like it should say "The latest *Haswell* Intel chipsets ..."

If you really mean "Ivy Bridge," then you should say something about how Haswell is related to Ivy Bridge. No doubt this is all obvious to Intel folks, but it's not to me :)

it should read "Haswell".

Fixed, thanks!

Bjorn