From: Steven Rostedtrostedt@goodmis.org <rostedt@goodmis.org>
CPU0 expires a posix-timer and runs the callback function.  The signal is
queued.
After releasing the posix-timer lock and before returning to hrtimer_run_queue
CPU0 gets interrupted.  CPU1 delivers the queued signal and rearms the timer.
CPU0 comes back to hrtimer_run_queue and sets the timer state to expired.
The next modification of the timer can result in an oops, because the state
information is wrong.
Keep track of state = RUNNING and check if the state has been in the return
path of hrtimer_run_queue.  In case the state has been changed, ignore a
restart request and do not touch the state variable.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
 enum hrtimer_state {
        HRTIMER_INACTIVE,       /* Timer is inactive */
        HRTIMER_EXPIRED,                /* Timer is expired */
+       HRTIMER_RUNNING,                /* Timer is running the callback function */
        HRTIMER_PENDING,                /* Timer is pending */
 };
 
 
                fn = timer->function;
                data = timer->data;
                set_curr_timer(base, timer);
+               timer->state = HRTIMER_RUNNING;
                __remove_hrtimer(timer, base);
                spin_unlock_irq(&base->lock);
 
 
                spin_lock_irq(&base->lock);
 
+               /* Another CPU has added back the timer */
+               if (timer->state != HRTIMER_RUNNING)
+                       continue;
+
                if (restart == HRTIMER_RESTART)
                        enqueue_hrtimer(timer, base);
                else