Skip to contentSkip to conversation

improve INTx toggle for PCI MSI

PATCH In review Jeff Garzik 5 participants 5 messages Dec 7, 2006 – Dec 8, 2006 PCI

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 improve INTx toggle for PCI MSI

"it boots" on ICH7 at least.

Please pull from 'intx' branch of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/misc-2.6.git intx

to receive the following updates:

 drivers/pci/msi.c |   12 ++++--------
 1 files changed, 4 insertions(+), 8 deletions(-)

Jeff Garzik:
PCI MSI: always toggle legacy-INTx-enable bit upon MSI entry/exit

drivers/pci/msi.c +4 -8
index 9fc9a34..c2828a3 100644
@@ -255,10 +255,8 @@ static void enable_msi_mode(struct pci_d
 		pci_write_config_word(dev, msi_control_reg(pos), control);
 		dev->msix_enabled = 1;
 	}
-    	if (pci_find_capability(dev, PCI_CAP_ID_EXP)) {
-		/* PCI Express Endpoint device detected */
-		pci_intx(dev, 0);  /* disable intx */
-	}
+
+	pci_intx(dev, 0);  /* disable intx */
 }
 
 void disable_msi_mode(struct pci_dev *dev, int pos, int type)
@@ -276,10 +274,8 @@ void disable_msi_mode(struct pci_dev *de
 		pci_write_config_word(dev, msi_control_reg(pos), control);
 		dev->msix_enabled = 0;
 	}
-    	if (pci_find_capability(dev, PCI_CAP_ID_EXP)) {
-		/* PCI Express Endpoint device detected */
-		pci_intx(dev, 1);  /* enable intx */
-	}
+
+	pci_intx(dev, 1);  /* enable intx */
 }
 
 static int msi_lookup_irq(struct pci_dev *dev, int type)
#1 lore raw
Quoted 2 lines · On Thu, 7 Dec 2006 17:58:12 -0500 Jeff Garzik <[email protected]> wrote:

"it boots" on ICH7 at least.

That's a rave review ;)

Quoted 10 lines

Please pull from 'intx' branch of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik/misc-2.6.git intx

to receive the following updates:

 drivers/pci/msi.c |   12 ++++--------
 1 files changed, 4 insertions(+), 8 deletions(-)

Jeff Garzik:
PCI MSI: always toggle legacy-INTx-enable bit upon MSI entry/exit

Why?

Quoted 30 lines
drivers/pci/msi.c +4 -8
index 9fc9a34..c2828a3 100644
@@ -255,10 +255,8 @@ static void enable_msi_mode(struct pci_d
 		pci_write_config_word(dev, msi_control_reg(pos), control);
 		dev->msix_enabled = 1;
 	}
-    	if (pci_find_capability(dev, PCI_CAP_ID_EXP)) {
-		/* PCI Express Endpoint device detected */
-		pci_intx(dev, 0);  /* disable intx */
-	}
+
+	pci_intx(dev, 0);  /* disable intx */
 }
 
 void disable_msi_mode(struct pci_dev *dev, int pos, int type)
@@ -276,10 +274,8 @@ void disable_msi_mode(struct pci_dev *de
 		pci_write_config_word(dev, msi_control_reg(pos), control);
 		dev->msix_enabled = 0;
 	}
-    	if (pci_find_capability(dev, PCI_CAP_ID_EXP)) {
-		/* PCI Express Endpoint device detected */
-		pci_intx(dev, 1);  /* enable intx */
-	}
+
+	pci_intx(dev, 1);  /* enable intx */
 }
 
 static int msi_lookup_irq(struct pci_dev *dev, int type)
#2 lore raw
In reply to Jeff Garzik
Quoted 2 lines · Jeff Garzik

"it boots" on ICH7 at least.

Ok. Pulled, pushed out.

There was some noise saying that this may actually fix the problems with the NVidia "Intel HDA" sound situation? Can people who saw that issue try it out whether this just makes MSI works for them?

Takashi added to the To: field, because he hopefully remembers and has a clue about the proper identities in question.. Iirc, you needed to have not only a NVidia chipset, but also have the legacy interrupt shared with some other device to see the problem.

Linus

#3 lore raw

2006/12/8, Linus Torvalds <[email protected]>:

Quoted 8 lines · Jeff Garzik
Quoted 2 lines

"it boots" on ICH7 at least.

Ok. Pulled, pushed out.

Quoted 8 lines

There was some noise saying that this may actually fix the problems with the NVidia "Intel HDA" sound situation? Can people who saw that issue try it out whether this just makes MSI works for them?

Takashi added to the To: field, because he hopefully remembers and has a clue about the proper identities in question.. Iirc, you needed to have not only a NVidia chipset, but also have the legacy interrupt shared with some other device to see the problem.

Well, I'm on vacation now, so cannot answer much, too :)

IIRC, the problem was with HD-audio and network devices on Nvidia. The explicit call of pci_intx() (currently implemented in hd-audio driver locally) helped avoiding the orphan irq problem, but doesn't fix the MSI problem itself. (Some might have been fixed indeed, but I don't have reports in hands.)

Takashi