sched: sync wakeups preempt too
authorIngo Molnar <mingo@elte.hu>
Mon, 15 Oct 2007 15:00:20 +0000 (17:00 +0200)
committerIngo Molnar <mingo@elte.hu>
Mon, 15 Oct 2007 15:00:20 +0000 (17:00 +0200)
make sure sync wakeups preempt too - the scheduler will not
overschedule as we've got various throttles against that.
As a result, sync wakeups can be used more widely in the kernel
(to signal wakeup affinity between tasks), and no arbitrary
latencies will be introduced either.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/sched.c

index 7fd343462597089def2cbb251e4702bd3709b854..bba57adb95044ba215caa58ee7bdd514ccd71c1c 100644 (file)
@@ -1596,16 +1596,7 @@ out_activate:
                schedstat_inc(p, se.nr_wakeups_remote);
        update_rq_clock(rq);
        activate_task(rq, p, 1);
-       /*
-        * Sync wakeups (i.e. those types of wakeups where the waker
-        * has indicated that it will leave the CPU in short order)
-        * don't trigger a preemption, if the woken up task will run on
-        * this cpu. (in this case the 'I will reschedule' promise of
-        * the waker guarantees that the freshly woken up task is going
-        * to be considered on this CPU.)
-        */
-       if (!sync || rq->curr == rq->idle)
-               check_preempt_curr(rq, p);
+       check_preempt_curr(rq, p);
        success = 1;
 
 out_running: