[PATCH 02/10] nohz: Kick full dynticks timer targets with an empty IPI

[Frederic Weisbecker]

While we enqueue a new timer on a dynticks target, we must send it an
IPI so that it reschedules the next tick accordingly.

Now all we need for that is to run irq_exit() on the target. Hence
an empty IRQ is way enough. We don't need to run tick_nohz_full_check()
and all the overhead that comes with the several cmpxchg() necessary for
irq work list and state progress.

So lets use a void irq work instead.

Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Nicolas Pitre <nicolas.pitre@xxxxxxxxxx>
Cc: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
Signed-off-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>
---
 kernel/sched/core.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 7f3063c..171f9ce 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -685,15 +685,13 @@ static void wake_up_idle_cpu(int cpu)
 static bool wake_up_full_nohz_cpu(int cpu)
 {
 	/*
-	 * We just need the target to call irq_exit() and re-evaluate
-	 * the next tick. The nohz full kick at least implies that.
-	 * If needed we can still optimize that later with an
-	 * empty IRQ.
+	 * Kick the target to make it call irq_exit() and re-evaluate
+	 * the next tick.
 	 */
 	if (tick_nohz_full_cpu(cpu)) {
 		if (cpu != smp_processor_id() ||
 		    tick_nohz_tick_stopped())
-			tick_nohz_full_kick_cpu(cpu);
+			kick_cpu_async(cpu);
 		return true;
 	}
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/