From: Andy Gospodarek <gospo@redhat.com>
Date: Fri, 10 Jul 2009 09:53:21 -0400
Subject: [net] igb: fix panic when assigning device to guest
Message-id: 20090710135321.GQ20245@gospo.rdu.redhat.com
O-Subject: [RHEL5.4 PATCH] igb: fix panic when assigning device to guest
Bugzilla: 507173
RH-Acked-by: Dean Nelson <dnelson@redhat.com>
RH-Acked-by: Jiri Pirko <jpirko@redhat.com>
RH-Acked-by: Stefan Assmann <sassmann@redhat.com>
RH-Acked-by: David Miller <davem@redhat.com>

Under certain system configurations when using SR-IOV, igb will panic
during init because dev->bus->self will be NULL for some pci devices.
This seems to be the case when using the Stoakley platform and is
expected based on the device layout.

This code was recently removed from the upstream driver and a pci quirk
was added instead.  That quirk calls pci_disable_link_state (which is
not available in RHEL5), but this patch emulates some of the work done
there by checking for NULL.

For the next update we should consider taking the two upstream patches
that remove this in igb and add the pci-quirk (in addition to
backporting the full functionality of pci_disable_link_state), but this
is not the best time to make changes that large.

This patch was verified by Intel.  The igb device can now be assigned to
the guest and run successfully.

This will resolve RHBZ 507173.

diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index 405906d..bfa80a2 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -1100,6 +1100,8 @@ static int __devinit igb_probe(struct pci_dev *pdev,
 	case E1000_DEV_ID_82575EB_FIBER_SERDES:
 	case E1000_DEV_ID_82575GB_QUAD_COPPER:
 		us_dev = pdev->bus->self;
+		if (!us_dev)
+			break;
 		pos = pci_find_capability(us_dev, PCI_CAP_ID_EXP);
 		if (pos) {
 			pci_read_config_word(us_dev, pos + PCI_EXP_LNKCTL,