hrtimer: fix timerqueue conversion flub
authorJohn Stultz <john.stultz@linaro.org>
Sat, 11 Dec 2010 06:19:53 +0000 (22:19 -0800)
committerJohn Stultz <john.stultz@linaro.org>
Sat, 11 Dec 2010 06:19:53 +0000 (22:19 -0800)
In converting the hrtimers to timerqueue, I missed
a spot in hrtimer_run_queues where we loop running
timers. We end up not pulling the new next value out
and instead just use the last next value, causing
boot time hangs in some cases.

The proper fix is to pull timerqueue_getnext each iteration
instead of using a local next value.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: John Stultz <john.stultz@linaro.org>
kernel/hrtimer.c

index 93976ad42f5aada175a540f953e1d1119b9f7062..7a7a2061c24daea1d7a4a09507df92f3a78fba56 100644 (file)
@@ -1412,11 +1412,8 @@ void hrtimer_run_queues(void)
                return;
 
        for (index = 0; index < HRTIMER_MAX_CLOCK_BASES; index++) {
-               struct timerqueue_node *next;
-
                base = &cpu_base->clock_base[index];
-               next = timerqueue_getnext(&base->active);
-               if (!next)
+               if (!timerqueue_getnext(&base->active))
                        continue;
 
                if (gettime) {
@@ -1426,7 +1423,7 @@ void hrtimer_run_queues(void)
 
                raw_spin_lock(&cpu_base->lock);
 
-               while ((node = next)) {
+               while ((node = timerqueue_getnext(&base->active))) {
                        struct hrtimer *timer;
 
                        timer = container_of(node, struct hrtimer, node);