x86: AMD Northbridge: Verify NB's node is online

Fix panic seen on some IBM and HP systems on 2.6.32-rc6:

 BUG: unable to handle kernel NULL pointer dereference at (null)
 IP: [<ffffffff8120bf3f>] find_next_bit+0x77/0x9c
  [...]
  [<ffffffff8120bbde>] cpumask_next_and+0x2e/0x3b
  [<ffffffff81225c62>] pci_device_probe+0x8e/0xf5
  [<ffffffff812b9be6>] ? driver_sysfs_add+0x47/0x6c
  [<ffffffff812b9da5>] driver_probe_device+0xd9/0x1f9
  [<ffffffff812b9f1d>] __driver_attach+0x58/0x7c
  [<ffffffff812b9ec5>] ? __driver_attach+0x0/0x7c
  [<ffffffff812b9298>] bus_for_each_dev+0x54/0x89
  [<ffffffff812b9b4f>] driver_attach+0x19/0x1b
  [<ffffffff812b97ae>] bus_add_driver+0xd3/0x23d
  [<ffffffff812ba1e7>] driver_register+0x98/0x109
  [<ffffffff81225ed0>] __pci_register_driver+0x63/0xd3
  [<ffffffff81072776>] ? up_read+0x26/0x2a
  [<ffffffffa0081000>] ? k8temp_init+0x0/0x20 [k8temp]
  [<ffffffffa008101e>] k8temp_init+0x1e/0x20 [k8temp]
  [<ffffffff8100a073>] do_one_initcall+0x6d/0x185
  [<ffffffff8108d765>] sys_init_module+0xd3/0x236
  [<ffffffff81011ac2>] system_call_fastpath+0x16/0x1b

I put in a printk and commented out the set_dev_node()
call when and got this output:

 quirk_amd_nb_node: current numa_node = 0x0, would set to val & 7 = 0x0
 quirk_amd_nb_node: current numa_node = 0x0, would set to val & 7 = 0x1
 quirk_amd_nb_node: current numa_node = 0x0, would set to val & 7 = 0x2
 quirk_amd_nb_node: current numa_node = 0x0, would set to val & 7 = 0x3

I.e. the issue appears to be that the HW has set val to a valid
value, however, the system is only configured for a single
node -- 0, the others are offline.

Check to see if the node is actually online before setting
the numa node for an AMD northbridge in quirk_amd_nb_node().

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Cc: bhavna.sarathy@amd.com
Cc: jbarnes@virtuousgeek.org
Cc: andreas.herrmann3@amd.com
LKML-Reference: <20091112180933.12532.98685.sendpatchset@prarit.bos.redhat.com>
[ v2: clean up the code and add comments ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>

diff --git a/arch/x86/kernel/quirks.c b/arch/x86/kernel/quirks.c
index 6c3b2c6fd77257d643e3eb666a3a1f36a9c29efa..18093d7498f0d70144c19cbc954a644decda597d 100644 (file)
--- a/arch/x86/kernel/quirks.c
+++ b/arch/x86/kernel/quirks.c
@@ -499,6 +499,7 @@ static void __init quirk_amd_nb_node(struct pci_dev *dev)
 {
        struct pci_dev *nb_ht;
        unsigned int devfn;
+       u32 node;
        u32 val;
 
        devfn = PCI_DEVFN(PCI_SLOT(dev->devfn), 0);
@@ -507,7 +508,13 @@ static void __init quirk_amd_nb_node(struct pci_dev *dev)
                return;
 
        pci_read_config_dword(nb_ht, 0x60, &val);
-       set_dev_node(&dev->dev, val & 7);
+       node = val & 7;
+       /*
+        * Some hardware may return an invalid node ID,
+        * so check it first:
+        */
+       if (node_online(node))
+               set_dev_node(&dev->dev, node);
        pci_dev_put(nb_ht);
 }