Skip to contentSkip to conversation

PCI: Fix regression in powerpc MSI-X

PATCH In review Andre Detsch 3 participants 4 messages Nov 4, 2009 – Nov 5, 2009 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PCI: Fix regression in powerpc MSI-X

Patch f598282f5145036312d90875d0ed5c14b49fd8a7 exposed a problem in powerpc MSI-X functionality, making network interfaces such as ixgbe and cxgb3 stop to work when MSI-X is enabled. RX interrupts were not being generated.

The problem was caused because MSI irq was not being effectively unmasked after device initialization.

Index: linux-2.6/arch/powerpc/platforms/pseries/msi.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/pseries/msi.c 2009-11-04
06:35:39.000000000 -0700
+++ linux-2.6/arch/powerpc/platforms/pseries/msi.c 2009-11-04
07:23:27.000000000 -0700
@@ -432,8 +432,6 @@ static int rtas_setup_msi_irqs(struct pc
/* Read config space back so we can restore after reset */ read_msi_msg(virq, &msg);
entry->msg = msg;
-
- unmask_msi_irq(virq);
}

return 0;
Index: linux-2.6/arch/powerpc/platforms/pseries/xics.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/pseries/xics.c 2009-11-04
06:35:39.000000000 -0700
+++ linux-2.6/arch/powerpc/platforms/pseries/xics.c 2009-11-04
07:23:27.000000000 -0700
@@ -18,6 +18,7 @@
#include <linux/init.h>
#include <linux/radix-tree.h>
#include <linux/cpu.h>
+#include <linux/msi.h>
#include <linux/of.h>

#include <asm/firmware.h>
@@ -219,6 +220,14 @@ static void xics_unmask_irq(unsigned int

static unsigned int xics_startup(unsigned int virq) {
+ /*
+ * The generic MSI code returns with the interrupt disabled on the
+ * card, using the MSI mask bits. Firmware doesn't appear to unmask
+ * at that level, so we do it here by hand.
+ */
+ if (irq_to_desc(virq)->msi_desc)
+ unmask_msi_irq(virq);
+
/* unmask it */ xics_unmask_irq(virq);
return 0;

#1 lore raw
Quoted 4 lines · Andre Detsch

Patch f598282f5145036312d90875d0ed5c14b49fd8a7 exposed a problem in powerpc MSI-X functionality, making network interfaces such as ixgbe and cxgb3 stop to work when MSI-X is enabled. RX interrupts were not being generated.

Which means we'd like it in 32 please Ben.

Also Andre, once Ben has merged this you should forward it to [email protected] for the 31 stable series.

cheers

#2 lore raw
In reply to Andre Detsch
Quoted 7 lines · Andre Detsch

Patch f598282f5145036312d90875d0ed5c14b49fd8a7 exposed a problem in powerpc MSI-X functionality, making network interfaces such as ixgbe and cxgb3 stop to work when MSI-X is enabled. RX interrupts were not being generated.

The problem was caused because MSI irq was not being effectively unmasked after device initialization.

Patch is wrapped...

Ben.

Quoted 51 lines

Index: linux-2.6/arch/powerpc/platforms/pseries/msi.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/pseries/msi.c 2009-11-04
06:35:39.000000000 -0700
+++ linux-2.6/arch/powerpc/platforms/pseries/msi.c 2009-11-04
07:23:27.000000000 -0700

(quoted diff) +0 -2
@@ -432,8 +432,6 @@ static int rtas_setup_msi_irqs(struct pc
 		/* Read config space back so we can restore after reset */
 		read_msi_msg(virq, &msg);
 		entry->msg = msg;
-
-		unmask_msi_irq(virq);
 	}
 
 	return 0;

Index: linux-2.6/arch/powerpc/platforms/pseries/xics.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/pseries/xics.c 2009-11-04
06:35:39.000000000 -0700
+++ linux-2.6/arch/powerpc/platforms/pseries/xics.c 2009-11-04
07:23:27.000000000 -0700

(quoted diff) +9 -0
@@ -18,6 +18,7 @@
 #include <linux/init.h>
 #include <linux/radix-tree.h>
 #include <linux/cpu.h>
+#include <linux/msi.h>
 #include <linux/of.h>
 
 #include <asm/firmware.h>
@@ -219,6 +220,14 @@ static void xics_unmask_irq(unsigned int
 
 static unsigned int xics_startup(unsigned int virq)
 {
+	/*
+	 * The generic MSI code returns with the interrupt disabled on the
+	 * card, using the MSI mask bits. Firmware doesn't appear to unmask
+	 * at that level, so we do it here by hand.
+	 */
+	if (irq_to_desc(virq)->msi_desc)
+		unmask_msi_irq(virq);
+
 	/* unmask it */
 	xics_unmask_irq(virq);
 	return 0;

_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://lists.ozlabs.org/listinfo/linuxppc-dev

#3 lore raw

PATCH 1/1 PCI: Fix regression in powerpc MSI-X

Patch f598282f5145036312d90875d0ed5c14b49fd8a7 exposed a problem in powerpc MSI-X functionality, making network interfaces such as ixgbe and cxgb3 stop to work when MSI-X is enabled. RX interrupts were not being generated.

The problem was caused because MSI irq was not being effectively unmasked after device initialization.

arch/powerpc/platforms/pseries/msi.c +0 -2
Index: linux-2.6/arch/powerpc/platforms/pseries/msi.c
===================================================================
@@ -432,8 +432,6 @@ static int rtas_setup_msi_irqs(struct pc
 		/* Read config space back so we can restore after reset */
 		read_msi_msg(virq, &msg);
 		entry->msg = msg;
-
-		unmask_msi_irq(virq);
 	}
 
 	return 0;
arch/powerpc/platforms/pseries/xics.c +9 -0
Index: linux-2.6/arch/powerpc/platforms/pseries/xics.c
===================================================================
@@ -18,6 +18,7 @@
 #include <linux/init.h>
 #include <linux/radix-tree.h>
 #include <linux/cpu.h>
+#include <linux/msi.h>
 #include <linux/of.h>
 
 #include <asm/firmware.h>
@@ -219,6 +220,14 @@ static void xics_unmask_irq(unsigned int
 
 static unsigned int xics_startup(unsigned int virq)
 {
+	/*
+	 * The generic MSI code returns with the interrupt disabled on the
+	 * card, using the MSI mask bits. Firmware doesn't appear to unmask
+	 * at that level, so we do it here by hand.
+	 */
+	if (irq_to_desc(virq)->msi_desc)
+		unmask_msi_irq(virq);
+
 	/* unmask it */
 	xics_unmask_irq(virq);
 	return 0;