x86: C1E late detection fix. Really switch off lapic timer

Doh, I completely missed that devices marked DUMMY are not running
the set_mode function. So we force broadcasting, but we keep the
local APIC timer running.

Let the clock event layer mark the device _after_ switching it off.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 4 insertions, 14 deletions

diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c
index fab9dd8bbd6b..8cfb8b2ce773 100644
--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -222,20 +222,8 @@ static void tick_do_broadcast_on_off(void *why)
         if (!dev || !(dev->features & CLOCK_EVT_FEAT_C3STOP))
                 goto out;
 
-        /*
-         * Defect device ?
-         */
-        if (!tick_device_is_functional(dev)) {
-                /*
-                 * AMD C1E wreckage fixup:
-                 *
-                 * Device was registered functional in the first
-                 * place. Now the secondary CPU detected the C1E
-                 * misfeature and notifies us to fix it up
-                 */
-                if (*reason != CLOCK_EVT_NOTIFY_BROADCAST_FORCE)
-                        goto out;
-        }
+        if (!tick_device_is_functional(dev))
+                goto out;
 
         switch (*reason) {
         case CLOCK_EVT_NOTIFY_BROADCAST_ON:
@@ -246,6 +234,8 @@ static void tick_do_broadcast_on_off(void *why)
                                 clockevents_set_mode(dev,
                                                      CLOCK_EVT_MODE_SHUTDOWN);
                 }
+                if (*reason == CLOCK_EVT_NOTIFY_BROADCAST_FORCE)
+                        dev->features |= CLOCK_EVT_FEAT_DUMMY;
                 break;
         case CLOCK_EVT_NOTIFY_BROADCAST_OFF:
                 if (cpu_isset(cpu, tick_broadcast_mask)) {