From: Paolo Bonzini <pbonzini@redhat.com>
Date: Mon, 22 Feb 2010 19:02:00 -0500
Subject: [x86_64] k8: do not mark early_is_k8_nb as __init
Message-id: <1266865320-10213-1-git-send-email-pbonzini@redhat.com>
Patchwork-id: 23401
O-Subject: [RHEL5.5 PATCH] Do not mark early_is_k8_nb as __init
Bugzilla: 567275
RH-Acked-by: Pete Zaitcev <zaitcev@redhat.com>
RH-Acked-by: Juan Quintela <quintela@redhat.com>
RH-Acked-by: Andrew Jones <drjones@redhat.com>

Bugzilla: http://bugzilla.redhat.com/show_bug.cgi?id=567275

Upstream status: no way without a usage of early_is_k8_nb outside __init

Brew build: http://brewweb.devel.redhat.com/brew/taskinfo?taskID=2274638

AMD QA team is reporting that the patch in bug 560013 (for RHEL5.5,
bug 547518 for RHEL5.4.z) breaks CPU offlining and onlining for M-C bare
metal systems.

The cause is a function marked as __init called from a __cpuinit function.

On an affected machine,

  echo 0 > /sys/devices/system/cpu/cpu1/online    # works
  echo 1 > /sys/devices/system/cpu/cpu1/online    # hangs

without the patch, while both work with the patch (tested by me).
This bug is not present in RHEL6, because the aforementioned BZs were
RHEL5 material only.

Please review and ack for both 5.5 and 5.4.z.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: Bhavna Sarathy <bhavna.sarathy@amd.com>

diff --git a/arch/x86_64/kernel/k8.c b/arch/x86_64/kernel/k8.c
index bc11b32..f9d9770 100644
--- a/arch/x86_64/kernel/k8.c
+++ b/arch/x86_64/kernel/k8.c
@@ -71,7 +71,7 @@ EXPORT_SYMBOL_GPL(cache_k8_northbridges);
 
 /* Ignores subdevice/subvendor but as far as I can figure out
    they're useless anyways */
-int __init early_is_k8_nb(u32 device)
+int early_is_k8_nb(u32 device)
 {
 	struct pci_device_id *id;
 	u32 vendor = device & 0xffff;