x86/PCI: don't use native Broadcom CNB20LE driver when ACPI is available

The broadcom_bus.c quirk was written (without benefit of documentation)
to support PCI hotplug on an old system that doesn't have ACPI.  As
such, we should only use it when the system doesn't have ACPI.

If the system does have ACPI and we need the host bridge description, we
should get it from the ACPI _CRS method.  On machines older than 2008,
we currently ignore _CRS, but that doesn't mean we should use
broadcom_bus.c.  It means we should either (a) do what we've done in the
past and assume everything in the PCI gap is routed to bus 0 (so hotplug
may not work), or (b) arrange to use _CRS.  This patch does (a).

Reference: https://bugzilla.redhat.com/show_bug.cgi?id=665109
Acked-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

diff --git a/arch/x86/pci/broadcom_bus.c b/arch/x86/pci/broadcom_bus.c
index 0846a5b..ab8269b 100644 (file)
--- a/arch/x86/pci/broadcom_bus.c
+++ b/arch/x86/pci/broadcom_bus.c
@@ -9,6 +9,7 @@
  * option) any later version.
  */
 
+#include <linux/acpi.h>
 #include <linux/delay.h>
 #include <linux/dmi.h>
 #include <linux/pci.h>
@@ -25,12 +26,14 @@ static void __devinit cnb20le_res(struct pci_dev *dev)
        u8 fbus, lbus;
        int i;
 
+#ifdef CONFIG_ACPI
        /*
-        * The x86_pci_root_bus_res_quirks() function already refuses to use
-        * this information if ACPI _CRS was used. Therefore, we don't bother
-        * checking if ACPI is enabled, and just generate the information
-        * for both the ACPI _CRS and no ACPI cases.
+        * We should get host bridge information from ACPI unless the BIOS
+        * doesn't support it.
         */
+       if (acpi_os_get_root_pointer())
+               return;
+#endif
 
        info = &pci_root_info[pci_root_num];
        pci_root_num++;