[PATCH v2 0/6] call _OSC support during root bridge discovery

[PATCH v2 0/6] call _OSC support during root bridge discovery

[Andrew Patterson]

This is v2 of the "call _OSC support during root bridge discovery"
patchset.  At Bjorn Helgaas's suggestion, I have merged all the MSI
handling patches into one patch. Kenji Kaneshige also noted that I was
not checking aspm_disabled before setting ASPM _OSC capabilities. This
has been fixed.

Kenji also noted that the pci_msi_enable can be changed in various
quirks after root bridge discovery, resulting in the incorrect setting
of the MSI _OSC capability. I do not yet have a solution for this problem.

I still need Matthew Wilcox to sign off on the first two patches.


I recently reported a problem where a machine with close to a 100 PCIe
root bridges was taking half an hour to just call _OSC.  The root
cause is the AER code calling:

        pcie_osc_support_set(OSC_EXT_PCI_CONFIG_SUPPORT);

for each bridge.  The pcie_osc_support_set() function also iterates
over each bridge, so _OSC is called a quadratic number of times.

One solution to this problem would be to just move the call to
pcie_osc_support_set() to aer_service_init(), so _OSC support is only
called on each bridge once.

Matthew Wilcox came up with a better solution.  He says "Why should a
driver be calling pcie_osc_support_set() anyway?  This is something
the PCI core should be taking care of for it."

Matthew provided a patch for this solution that I massaged into the
following patch series.

It creates a new function (pci_acpi_osc_support()) which is called
from pci_root.c before we call any other methods on the device (as
recommended by the ACPI spec).  Since we know what capabilities the
system supports, individual modules do not now need to inform the core
of their support for various capabilities.

Some work that still needs to be done (provided by Matthew): 

o All the existing _OSC code should be moved from pci-acpi.c to
  pci_root.c.  I also think the acpi_osc_data should be made part of
  the acpi_pci_root struct, eliminating a separate allocation.

o We could also use an interface that iterates over all existing
  busses calling _OSC with new flags.  Shouldn't be hard, once it's
  integrated into pci_root.c.

o Further ahead, we don't actually check that the bits we asked for
  (in 'control') were actually granted to us.  The PCI firmware spec
  is quite explicit about interdependencies amongst the bits.

o We also need to re-evaluate _OSC when coming out of S4.

This patch series duplicates currently functionality while removing
our quadratic problem.  I believe that it can be applied now while the
above new functionality can be implemented some time in the future.

This patch series applies to the linux-next branch of pci-2.6 git
repository.  I expect it will also work with linux-next.

Diff stats:

 drivers/acpi/pci_root.c            |   12 ++++++++++++
 drivers/pci/msi.c                  |   31 +++++++++++-------------------
 drivers/pci/pci-acpi.c             |   37 +++++++++++++-----------------------
 drivers/pci/pci.c                  |    2 --
 drivers/pci/pci.h                  |    2 --
 drivers/pci/pcie/aer/aerdrv_acpi.c |    1 -
 drivers/pci/pcie/aspm.c            |   27 +++++++++-----------------
 include/linux/pci-acpi.h           |   14 +++-----------
 include/linux/pci.h                |   14 ++++++++++++++
 9 files changed, 62 insertions(+), 78 deletions(-)

Commits:

  - Include missing acpi.h file in pci-acpi.h.
  - Call _OSC support during root bridge discovery.
  - PCIe ASPM _OSC support capabilities called when root bridge added.
  - PCIe AER _OSC support capabilities called when root bridge added.
  - PCI MSI _OSC support capabilities called when root bridge added.
  - Remove obsolete _OSC capability support functions.

-- 
Andrew Patterson

[PATCH 1/6] PCI, ACPI: include missing acpi.h file in pci-acpi.h.

[Andrew Patterson]

PCI, ACPI: include missing acpi.h file in pci-acpi.h.

The pci-acpi.h file will not compile without including linux/acpi.h.
---

 include/linux/pci-acpi.h |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)


diff --git a/include/linux/pci-acpi.h b/include/linux/pci-acpi.h
index 8837928..a9e4c34 100644
--- a/include/linux/pci-acpi.h
+++ b/include/linux/pci-acpi.h
@@ -8,6 +8,8 @@
 #ifndef _PCI_ACPI_H_
 #define _PCI_ACPI_H_
 
+#include <linux/acpi.h>
+
 #define OSC_QUERY_TYPE			0
 #define OSC_SUPPORT_TYPE 		1
 #define OSC_CONTROL_TYPE		2

[PATCH 2/6] ACPI, PCI: call _OSC support during root bridge discovery

[Andrew Patterson]

ACPI, PCI: call _OSC support during root bridge discovery

Added pci_acpi_osc_support() which is called when a PCI bridge is
added, so individual PCI root bridge drivers do not have to call _OSC
support for every root bridge in their probe functions.
---

 drivers/acpi/pci_root.c  |    7 +++++++
 drivers/pci/pci-acpi.c   |   24 +++++++++++++++++++-----
 include/linux/pci-acpi.h |    1 +
 3 files changed, 27 insertions(+), 5 deletions(-)


diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index 1b8f67d..b4d55f2 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -31,6 +31,7 @@
 #include <linux/spinlock.h>
 #include <linux/pm.h>
 #include <linux/pci.h>
+#include <linux/pci-acpi.h>
 #include <linux/acpi.h>
 #include <acpi/acpi_bus.h>
 #include <acpi/acpi_drivers.h>
@@ -193,6 +194,7 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	unsigned long long value = 0;
 	acpi_handle handle = NULL;
 	struct acpi_device *child;
+	u32 flags;
 
 
 	if (!device)
@@ -210,6 +212,11 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 
 	device->ops.bind = acpi_pci_bind;
 
+	flags = (OSC_EXT_PCI_CONFIG_SUPPORT |
+		 OSC_PCI_SEGMENT_GROUPS_SUPPORT |
+		 0);
+	pci_acpi_osc_support(device->handle, flags);
+
 	/* 
 	 * Segment
 	 * -------
diff --git a/drivers/pci/pci-acpi.c b/drivers/pci/pci-acpi.c
index dfe7c8e..f457387 100644
--- a/drivers/pci/pci-acpi.c
+++ b/drivers/pci/pci-acpi.c
@@ -139,28 +139,42 @@ static acpi_status __acpi_query_osc(u32 flags, struct acpi_osc_data *osc_data,
 	return status;
 }
 
-static acpi_status acpi_query_osc(acpi_handle handle,
-				  u32 level, void *context, void **retval)
+/*
+ * pci_acpi_osc_support: Invoke _OSC indicating support for the given feature
+ * @flags: Bitmask of flags to support
+ *
+ * See the ACPI spec for the definition of the flags
+ */
+int pci_acpi_osc_support(acpi_handle handle, u32 flags)
 {
+	u32 dummy;
 	acpi_status status;
-	struct acpi_osc_data *osc_data;
-	u32 flags = (unsigned long)context, dummy;
 	acpi_handle tmp;
+	struct acpi_osc_data *osc_data;
+	int rc = 0;
 
 	status = acpi_get_handle(handle, "_OSC", &tmp);
 	if (ACPI_FAILURE(status))
-		return AE_OK;
+		return -ENOTTY;
 
 	mutex_lock(&pci_acpi_lock);
 	osc_data = acpi_get_osc_data(handle);
 	if (!osc_data) {
 		printk(KERN_ERR "acpi osc data array is full\n");
+		rc = -ENOMEM;
 		goto out;
 	}
 
 	__acpi_query_osc(flags, osc_data, &dummy);
 out:
 	mutex_unlock(&pci_acpi_lock);
+	return rc;
+}
+
+static acpi_status acpi_query_osc(acpi_handle handle, u32 level,
+				  void *context, void **retval)
+{
+	pci_acpi_osc_support(handle, (unsigned long)context);
 	return AE_OK;
 }
 
diff --git a/include/linux/pci-acpi.h b/include/linux/pci-acpi.h
index a9e4c34..424f06f 100644
--- a/include/linux/pci-acpi.h
+++ b/include/linux/pci-acpi.h
@@ -51,6 +51,7 @@
 #ifdef CONFIG_ACPI
 extern acpi_status pci_osc_control_set(acpi_handle handle, u32 flags);
 extern acpi_status __pci_osc_support_set(u32 flags, const char *hid);
+int pci_acpi_osc_support(acpi_handle handle, u32 flags);
 static inline acpi_status pci_osc_support_set(u32 flags)
 {
 	return __pci_osc_support_set(flags, PCI_ROOT_HID_STRING);

[PATCH 3/6] ACPI, PCI: PCIe ASPM _OSC support capabilities called when root bridge added

[Andrew Patterson]

ACPI, PCI: PCIe ASPM _OSC support capabilities called when root bridge added

The _OSC capabilities OSC_ACTIVE_STATE_PWR_SUPPORT and
OSC_CLOCK_PWR_CAPABILITY_SUPPORT are set when the root bridge is added
with pci_acpi_osc_support(), so we no longer need to do it in the
ASPM driver.

Added the function pcie_aspm_enabled, which returns true if pcie_aspm=off
is not on the kernel command-line.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
---

 drivers/acpi/pci_root.c |    3 +++
 drivers/pci/pcie/aspm.c |   27 +++++++++------------------
 include/linux/pci.h     |    9 +++++++++
 3 files changed, 21 insertions(+), 18 deletions(-)


diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index b4d55f2..c36dc1d 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -215,6 +215,9 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	flags = (OSC_EXT_PCI_CONFIG_SUPPORT |
 		 OSC_PCI_SEGMENT_GROUPS_SUPPORT |
 		 0);
+	if (pcie_aspm_enabled())
+		flags |= OSC_ACTIVE_STATE_PWR_SUPPORT |
+			OSC_CLOCK_PWR_CAPABILITY_SUPPORT;
 	pci_acpi_osc_support(device->handle, flags);
 
 	/* 
diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
index 8f63f4c..1089f86 100644
--- a/drivers/pci/pcie/aspm.c
+++ b/drivers/pci/pcie/aspm.c
@@ -834,24 +834,15 @@ void pcie_no_aspm(void)
 		aspm_disabled = 1;
 }
 
-#ifdef CONFIG_ACPI
-#include <acpi/acpi_bus.h>
-#include <linux/pci-acpi.h>
-static void pcie_aspm_platform_init(void)
-{
-	pcie_osc_support_set(OSC_ACTIVE_STATE_PWR_SUPPORT|
-		OSC_CLOCK_PWR_CAPABILITY_SUPPORT);
-}
-#else
-static inline void pcie_aspm_platform_init(void) { }
-#endif
-
-static int __init pcie_aspm_init(void)
+/**
+ * pcie_aspm_enabled - is PCIe ASPM enabled?
+ *
+ * Returns true if ASPM has not been disabled by the command-line option
+ * pcie_aspm=off.
+ **/
+int pcie_aspm_enabled(void)
 {
-	if (aspm_disabled)
-		return 0;
-	pcie_aspm_platform_init();
-	return 0;
+       return !aspm_disabled;
 }
+EXPORT_SYMBOL(pcie_aspm_enabled);
 
-fs_initcall(pcie_aspm_init);
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 752def8..682033c 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -779,6 +779,15 @@ extern void msi_remove_pci_irq_vectors(struct pci_dev *dev);
 extern void pci_restore_msi_state(struct pci_dev *dev);
 #endif
 
+#ifndef CONFIG_PCIEASPM
+static inline int pcie_aspm_enabled(void)
+{
+	return 0;
+}
+#else
+extern int pcie_aspm_enabled(void);
+#endif
+
 #ifdef CONFIG_HT_IRQ
 /* The functions a driver should call */
 int  ht_create_irq(struct pci_dev *dev, int idx);

[PATCH 4/6] ACPI, PCI: PCIe AER _OSC support capabilities called when root bridge added

[Andrew Patterson]

ACPI, PCI: PCIe AER _OSC support capabilities called when root bridge added

The _OSC capability OSC_EXT_PCI_CONFIG_SUPPORT is set when the root
bridge is added with pci_acpi_osc_support(), so we no longer need to do
it in the PCIe AER driver.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
---

 drivers/pci/pcie/aer/aerdrv_acpi.c |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)


diff --git a/drivers/pci/pcie/aer/aerdrv_acpi.c b/drivers/pci/pcie/aer/aerdrv_acpi.c
index 6dd7b13..ebce26c 100644
--- a/drivers/pci/pcie/aer/aerdrv_acpi.c
+++ b/drivers/pci/pcie/aer/aerdrv_acpi.c
@@ -38,7 +38,6 @@ int aer_osc_setup(struct pcie_device *pciedev)
 
 	handle = acpi_find_root_bridge_handle(pdev);
 	if (handle) {
-		pcie_osc_support_set(OSC_EXT_PCI_CONFIG_SUPPORT);
 		status = pci_osc_control_set(handle,
 					OSC_PCI_EXPRESS_AER_CONTROL |
 					OSC_PCI_EXPRESS_CAP_STRUCTURE_CONTROL);

[PATCH 5/6] ACPI, PCI: PCI MSI _OSC support capabilities called when root bridge added

[Andrew Patterson]

ACPI, PCI: PCI MSI _OSC support capabilities called when root bridge added

The _OSC capabilityy OSC_MSI_SUPPORT is set when the root
bridge is added with pci_acpi_osc_support(), so we no longer
need to do it in the PCI MSI driver.

Added the function pci_msi_enabled, which returns true if pci=nomsi is not
on the kernel command-line.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
---

 drivers/acpi/pci_root.c |    2 ++
 drivers/pci/msi.c       |   31 +++++++++++--------------------
 drivers/pci/pci.c       |    2 --
 drivers/pci/pci.h       |    2 --
 include/linux/pci.h     |    5 +++++
 5 files changed, 18 insertions(+), 24 deletions(-)


diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
index c36dc1d..69d637a 100644
--- a/drivers/acpi/pci_root.c
+++ b/drivers/acpi/pci_root.c
@@ -218,6 +218,8 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device)
 	if (pcie_aspm_enabled())
 		flags |= OSC_ACTIVE_STATE_PWR_SUPPORT |
 			OSC_CLOCK_PWR_CAPABILITY_SUPPORT;
+	if (pci_msi_enabled())
+		flags |= OSC_MSI_SUPPORT;
 	pci_acpi_osc_support(device->handle, flags);
 
 	/* 
diff --git a/drivers/pci/msi.c b/drivers/pci/msi.c
index 74801f7..0e8dae1 100644
--- a/drivers/pci/msi.c
+++ b/drivers/pci/msi.c
@@ -755,28 +755,19 @@ void pci_no_msi(void)
 	pci_msi_enable = 0;
 }
 
-void pci_msi_init_pci_dev(struct pci_dev *dev)
-{
-	INIT_LIST_HEAD(&dev->msi_list);
-}
-
-#ifdef CONFIG_ACPI
-#include <linux/acpi.h>
-#include <linux/pci-acpi.h>
-static void __devinit msi_acpi_init(void)
+/**
+ * pci_msi_enabled - is MSI enabled?
+ *
+ * Returns true if MSI has not been disabled by the command-line option
+ * pci=nomsi.
+ **/
+int pci_msi_enabled(void)
 {
-	if (acpi_pci_disabled)
-		return;
-	pci_osc_support_set(OSC_MSI_SUPPORT);
-	pcie_osc_support_set(OSC_MSI_SUPPORT);
+	return pci_msi_enable;
 }
-#else
-static inline void msi_acpi_init(void) { }
-#endif /* CONFIG_ACPI */
+EXPORT_SYMBOL(pci_msi_enabled);
 
-void __devinit msi_init(void)
+void pci_msi_init_pci_dev(struct pci_dev *dev)
 {
-	if (!pci_msi_enable)
-		return;
-	msi_acpi_init();
+	INIT_LIST_HEAD(&dev->msi_list);
 }
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 533aeb5..d77e477 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -2032,8 +2032,6 @@ static int __devinit pci_init(void)
 		pci_fixup_device(pci_fixup_final, dev);
 	}
 
-	msi_init();
-
 	return 0;
 }
 
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 9de87e9..b205ab8 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -98,11 +98,9 @@ extern unsigned int pci_pm_d3_delay;
 #ifdef CONFIG_PCI_MSI
 void pci_no_msi(void);
 extern void pci_msi_init_pci_dev(struct pci_dev *dev);
-extern void __devinit msi_init(void);
 #else
 static inline void pci_no_msi(void) { }
 static inline void pci_msi_init_pci_dev(struct pci_dev *dev) { }
-static inline void msi_init(void) { }
 #endif
 
 #ifdef CONFIG_PCIEAER
diff --git a/include/linux/pci.h b/include/linux/pci.h
index 682033c..9ea6bdc 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -767,6 +767,10 @@ static inline void msi_remove_pci_irq_vectors(struct pci_dev *dev)
 
 static inline void pci_restore_msi_state(struct pci_dev *dev)
 { }
+static inline int pci_msi_enabled(void)
+{
+	return 0;
+}
 #else
 extern int pci_enable_msi(struct pci_dev *dev);
 extern void pci_msi_shutdown(struct pci_dev *dev);
@@ -777,6 +781,7 @@ extern void pci_msix_shutdown(struct pci_dev *dev);
 extern void pci_disable_msix(struct pci_dev *dev);
 extern void msi_remove_pci_irq_vectors(struct pci_dev *dev);
 extern void pci_restore_msi_state(struct pci_dev *dev);
+extern int pci_msi_enabled(void);
 #endif
 
 #ifndef CONFIG_PCIEASPM

[PATCH 6/6] PCI, ACPI: remove obsolete _OSC capability support functions

[Andrew Patterson]

PCI, ACPI: remove obsolete _OSC capability support functions

The acpi_query_osc, __pci_osc_support_set, pci_osc_support_set, and
pcie_osc_support_set functions have been obsoleted in favor of setting
these capabilities during root bridge discovery with pci_acpi_osc_support.
There are no longer any callers of these functions, so they are removed.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
---

 drivers/pci/pci-acpi.c   |   25 -------------------------
 include/linux/pci-acpi.h |   11 -----------
 2 files changed, 0 insertions(+), 36 deletions(-)


diff --git a/drivers/pci/pci-acpi.c b/drivers/pci/pci-acpi.c
index f457387..128ce7d 100644
--- a/drivers/pci/pci-acpi.c
+++ b/drivers/pci/pci-acpi.c
@@ -171,31 +171,6 @@ out:
 	return rc;
 }
 
-static acpi_status acpi_query_osc(acpi_handle handle, u32 level,
-				  void *context, void **retval)
-{
-	pci_acpi_osc_support(handle, (unsigned long)context);
-	return AE_OK;
-}
-
-/**
- * __pci_osc_support_set - register OS support to Firmware
- * @flags: OS support bits
- * @hid: hardware ID
- *
- * Update OS support fields and doing a _OSC Query to obtain an update
- * from Firmware on supported control bits.
- **/
-acpi_status __pci_osc_support_set(u32 flags, const char *hid)
-{
-	if (!(flags & OSC_SUPPORT_MASKS))
-		return AE_TYPE;
-
-	acpi_get_devices(hid, acpi_query_osc,
-			 (void *)(unsigned long)flags, NULL);
-	return AE_OK;
-}
-
 /**
  * pci_osc_control_set - commit requested control to Firmware
  * @handle: acpi_handle for the target ACPI object
diff --git a/include/linux/pci-acpi.h b/include/linux/pci-acpi.h
index 424f06f..871e096 100644
--- a/include/linux/pci-acpi.h
+++ b/include/linux/pci-acpi.h
@@ -50,16 +50,7 @@
 
 #ifdef CONFIG_ACPI
 extern acpi_status pci_osc_control_set(acpi_handle handle, u32 flags);
-extern acpi_status __pci_osc_support_set(u32 flags, const char *hid);
 int pci_acpi_osc_support(acpi_handle handle, u32 flags);
-static inline acpi_status pci_osc_support_set(u32 flags)
-{
-	return __pci_osc_support_set(flags, PCI_ROOT_HID_STRING);
-}
-static inline acpi_status pcie_osc_support_set(u32 flags)
-{
-	return __pci_osc_support_set(flags, PCI_EXPRESS_ROOT_HID_STRING);
-}
 static inline acpi_handle acpi_find_root_bridge_handle(struct pci_dev *pdev)
 {
 	/* Find root host bridge */
@@ -76,8 +67,6 @@ typedef u32 		acpi_status;
 #endif    
 static inline acpi_status pci_osc_control_set(acpi_handle handle, u32 flags)
 {return AE_ERROR;}
-static inline acpi_status pci_osc_support_set(u32 flags) {return AE_ERROR;} 
-static inline acpi_status pcie_osc_support_set(u32 flags) {return AE_ERROR;}
 static inline acpi_handle acpi_find_root_bridge_handle(struct pci_dev *pdev)
 { return NULL; }
 #endif

Re: [PATCH v2 0/6] call _OSC support during root bridge discovery

[Kenji Kaneshige]

Andrew Patterson wrote:
> This is v2 of the "call _OSC support during root bridge discovery"
> patchset.  At Bjorn Helgaas's suggestion, I have merged all the MSI
> handling patches into one patch. Kenji Kaneshige also noted that I was
> not checking aspm_disabled before setting ASPM _OSC capabilities. This
> has been fixed.
> 
> Kenji also noted that the pci_msi_enable can be changed in various
> quirks after root bridge discovery, resulting in the incorrect setting
> of the MSI _OSC capability. I do not yet have a solution for this problem.
> 

Unfortunately, I don't have good idea about it yet too. In addition,
I don't understand how to handle the case if a feature is supported
but disabled by OS, from the PCI firmware spec.

If we should not set _OSC capabilities when it is supported but
disabled by OS, maybe we need also consider what we should do for
OSC_EXT_PCI_CONFIG_SUPPORT and OSC_PCI_SEGMENT_GROUPS_SUPPORT
if mmconfig is disabled, as well as aspm_disabled and pci_msi_enable.

By the way, Taku Izumi plan to post some patches to reduce _OSC
evaluation for _OSC control soon. I'm appreciate it if you will try
it on your big server.

Thanks,
Kenji Kaneshige




> I still need Matthew Wilcox to sign off on the first two patches.
> 
> 
> I recently reported a problem where a machine with close to a 100 PCIe
> root bridges was taking half an hour to just call _OSC.  The root
> cause is the AER code calling:
> 
>         pcie_osc_support_set(OSC_EXT_PCI_CONFIG_SUPPORT);
> 
> for each bridge.  The pcie_osc_support_set() function also iterates
> over each bridge, so _OSC is called a quadratic number of times.
> 
> One solution to this problem would be to just move the call to
> pcie_osc_support_set() to aer_service_init(), so _OSC support is only
> called on each bridge once.
> 
> Matthew Wilcox came up with a better solution.  He says "Why should a
> driver be calling pcie_osc_support_set() anyway?  This is something
> the PCI core should be taking care of for it."
> 
> Matthew provided a patch for this solution that I massaged into the
> following patch series.
> 
> It creates a new function (pci_acpi_osc_support()) which is called
> from pci_root.c before we call any other methods on the device (as
> recommended by the ACPI spec).  Since we know what capabilities the
> system supports, individual modules do not now need to inform the core
> of their support for various capabilities.
> 
> Some work that still needs to be done (provided by Matthew): 
> 
> o All the existing _OSC code should be moved from pci-acpi.c to
>   pci_root.c.  I also think the acpi_osc_data should be made part of
>   the acpi_pci_root struct, eliminating a separate allocation.
> 
> o We could also use an interface that iterates over all existing
>   busses calling _OSC with new flags.  Shouldn't be hard, once it's
>   integrated into pci_root.c.
> 
> o Further ahead, we don't actually check that the bits we asked for
>   (in 'control') were actually granted to us.  The PCI firmware spec
>   is quite explicit about interdependencies amongst the bits.
> 
> o We also need to re-evaluate _OSC when coming out of S4.
> 
> This patch series duplicates currently functionality while removing
> our quadratic problem.  I believe that it can be applied now while the
> above new functionality can be implemented some time in the future.
> 
> This patch series applies to the linux-next branch of pci-2.6 git
> repository.  I expect it will also work with linux-next.
> 
> Diff stats:
> 
>  drivers/acpi/pci_root.c            |   12 ++++++++++++
>  drivers/pci/msi.c                  |   31 +++++++++++-------------------
>  drivers/pci/pci-acpi.c             |   37 +++++++++++++-----------------------
>  drivers/pci/pci.c                  |    2 --
>  drivers/pci/pci.h                  |    2 --
>  drivers/pci/pcie/aer/aerdrv_acpi.c |    1 -
>  drivers/pci/pcie/aspm.c            |   27 +++++++++-----------------
>  include/linux/pci-acpi.h           |   14 +++-----------
>  include/linux/pci.h                |   14 ++++++++++++++
>  9 files changed, 62 insertions(+), 78 deletions(-)
> 
> Commits:
> 
>   - Include missing acpi.h file in pci-acpi.h.
>   - Call _OSC support during root bridge discovery.
>   - PCIe ASPM _OSC support capabilities called when root bridge added.
>   - PCIe AER _OSC support capabilities called when root bridge added.
>   - PCI MSI _OSC support capabilities called when root bridge added.
>   - Remove obsolete _OSC capability support functions.
> 

Re: [PATCH v2 0/6] call _OSC support during root bridge discovery

[Andrew Patterson]

On Thu, 2008-10-30 at 18:08 +0900, Kenji Kaneshige wrote:
> Andrew Patterson wrote:
> > This is v2 of the "call _OSC support during root bridge discovery"
> > patchset.  At Bjorn Helgaas's suggestion, I have merged all the MSI
> > handling patches into one patch. Kenji Kaneshige also noted that I was
> > not checking aspm_disabled before setting ASPM _OSC capabilities. This
> > has been fixed.
> > 
> > Kenji also noted that the pci_msi_enable can be changed in various
> > quirks after root bridge discovery, resulting in the incorrect setting
> > of the MSI _OSC capability. I do not yet have a solution for this problem.
> > 
> 
> Unfortunately, I don't have good idea about it yet too. In addition,
> I don't understand how to handle the case if a feature is supported
> but disabled by OS, from the PCI firmware spec.


> If we should not set _OSC capabilities when it is supported but
> disabled by OS, maybe we need also consider what we should do for
> OSC_EXT_PCI_CONFIG_SUPPORT and OSC_PCI_SEGMENT_GROUPS_SUPPORT
> if mmconfig is disabled, as well as aspm_disabled and pci_msi_enable.
> 

A good idea. I try and come up with something.

> By the way, Taku Izumi plan to post some patches to reduce _OSC
> evaluation for _OSC control soon. I'm appreciate it if you will try
> it on your big server.
> 

I would be happy to help.  Getting time on this server is sometimes
difficult, so it may take a while before I can run a test.

Andrew

> Thanks,
> Kenji Kaneshige
> 
> 

>