x86_64: Calgary - Fix mis-handled PCI topology
authorMurillo Fernandes Bernardes <bernarde@br.ibm.com>
Fri, 10 Aug 2007 20:31:00 +0000 (22:31 +0200)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Sat, 11 Aug 2007 22:58:12 +0000 (15:58 -0700)
Current code assumed that devices were directly connected to a Calgary
bridge, as it tried to get the iommu table directly from the parent bus
controller.

When we have another bridge between the Calgary/CalIOC2 bridge and the
device we should look upwards until we get to the top (Calgary/CalIOC2
bridge), where the iommu table resides.

Signed-off-by: Murillo Fernandes Bernardes <mfb@br.ibm.com>
Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/x86_64/kernel/pci-calgary.c

index ba16c96..71da01e 100644 (file)
@@ -367,16 +367,15 @@ static inline struct iommu_table *find_iommu_table(struct device *dev)
 
        pdev = to_pci_dev(dev);
 
-       /* is the device behind a bridge? */
-       if (unlikely(pdev->bus->parent))
-               pbus = pdev->bus->parent;
-       else
-               pbus = pdev->bus;
+       pbus = pdev->bus;
+
+       /* is the device behind a bridge? Look for the root bus */
+       while (pbus->parent)
+               pbus = pbus->parent;
 
        tbl = pci_iommu(pbus);
 
-       BUG_ON(pdev->bus->parent &&
-              (tbl->it_busno != pdev->bus->parent->number));
+       BUG_ON(tbl && (tbl->it_busno != pbus->number));
 
        return tbl;
 }