IRQ assign for some PCIe devices

Hello all,

Fallowing a Benjamin Herrenschmidt request, I sending you a fix for IRQ assign for some PCIe devices. This bug affects multiple PCIe devices including Cadet-E, Squib-E, CISCO 4X SDR IB, and Knox adapters.

The problem lays in the fact OF does not create an "interrupt" property for some PCIe device (for instance "pci\@800000020000204/pci1014\,0339\@0/", an IBM Raid Controller) and the kernel code fails returning a IRQ 0 (invalid one) if this property is not present.

This patch changes the way to map interrupts to if the code can not get "interrupts" property from PCI OF node, it falls back to standard OF parsing. I verified and it worked fine with a pair of Squib-E SAS adapter on a P6-570.

index bc1fb27..a11d689 100644
@@ -250,8 +250,11 @@ int of_irq_map_pci(struct pci_dev *pdev, struct of_irq *out_irq)
 	 * parsing
 	 */
 	dn = pci_device_to_OF_node(pdev);
-	if (dn)
-		return of_irq_map_one(dn, 0, out_irq);
+	if (dn) {
+		rc = of_irq_map_one(dn, 0, out_irq);
+		if (!rc)
+			return rc;
+	}
 
 	/* Ok, we don't, time to have fun. Let's start by building up an
 	 * interrupt spec.  we assume #interrupt-cells is 1, which is standard

Thanks for tracking this down !

We'll pick it up. In the future, it would be nice if you could post the patch inline rather than as an attachment (you may want to train via private email to make sure you get the right settings in your mailer to avoid damaging it since some mailers love to mangle patches) and with the appropriate "Signed-off-by:..." line.

The patch is correct, Paul, can you pick it up ?

Cheers,
Ben.