[SPARC64]: Fix 2 bugs in PCI Sabre bus scanning.
authorDavid S. Miller <davem@ultra5.davemloft.net>
Wed, 6 Jun 2007 21:03:08 +0000 (14:03 -0700)
committerDavid S. Miller <davem@sunset.davemloft.net>
Thu, 7 Jun 2007 23:59:46 +0000 (16:59 -0700)
If we are on hummingbird, bus runs at 66MHZ.

pbm->pci_bus should be setup with the result of pci_scan_one_pbm()
or else we deref NULL pointers in the error interrupt handlers.

Signed-off-by: David S. Miller <davem@davemloft.net>
arch/sparc64/kernel/pci_sabre.c

index 323d6c2..22e1be5 100644 (file)
@@ -636,13 +636,18 @@ static void apb_init(struct pci_bus *sabre_bus)
 static void sabre_scan_bus(struct pci_pbm_info *pbm)
 {
        static int once;
-       struct pci_bus *pbus;
 
        /* The APB bridge speaks to the Sabre host PCI bridge
         * at 66Mhz, but the front side of APB runs at 33Mhz
         * for both segments.
+        *
+        * Hummingbird systems do not use APB, so they run
+        * at 66MHZ.
         */
-       pbm->is_66mhz_capable = 0;
+       if (hummingbird_p)
+               pbm->is_66mhz_capable = 1;
+       else
+               pbm->is_66mhz_capable = 0;
 
        /* This driver has not been verified to handle
         * multiple SABREs yet, so trap this.
@@ -656,13 +661,13 @@ static void sabre_scan_bus(struct pci_pbm_info *pbm)
        }
        once++;
 
-       pbus = pci_scan_one_pbm(pbm);
-       if (!pbus)
+       pbm->pci_bus = pci_scan_one_pbm(pbm);
+       if (!pbm->pci_bus)
                return;
 
-       sabre_root_bus = pbus;
+       sabre_root_bus = pbm->pci_bus;
 
-       apb_init(pbus);
+       apb_init(pbm->pci_bus);
 
        sabre_register_error_handlers(pbm);
 }