]> err.no Git - linux-2.6/commitdiff
Fix assertion failure with MSI on sparc64
authorDavid Miller <davem@davemloft.net>
Fri, 11 May 2007 20:26:44 +0000 (13:26 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Fri, 11 May 2007 23:01:18 +0000 (16:01 -0700)
Today's find is a triggered assertion in msi_free_irqs() when the system
doesn't support MSI, in which case arch_setup_msi_irqs() always returns
an error.

The problem is that when this happens we branch into msi_free_irqs(), to
which you added the following assertion loop:

list_for_each_entry(entry, &dev->msi_list, list)
BUG_ON(irq_has_action(entry->irq));

Well, if arch_setup_msi_irqs() fails, entry->irq will be zero and
although that's never assigned to any normal devices we use that IRQ
number for the timer interrupt on sparc64 so this assertion triggers.

Better to test for zero before doing the irq_has_action() assertion
thing.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/pci/msi.c

index e6740d1a0824d36540b415018814ab6dc3b82344..d9cbd586ae4bc282815a6b7aad16ddb3ecb523df 100644 (file)
@@ -549,8 +549,10 @@ static int msi_free_irqs(struct pci_dev* dev)
 {
        struct msi_desc *entry, *tmp;
 
-       list_for_each_entry(entry, &dev->msi_list, list)
-               BUG_ON(irq_has_action(entry->irq));
+       list_for_each_entry(entry, &dev->msi_list, list) {
+               if (entry->irq)
+                       BUG_ON(irq_has_action(entry->irq));
+       }
 
        arch_teardown_msi_irqs(dev);