sched, tracing: Stop/start critical timings around the idle=poll idle loop

When using idle=poll, the preemptoff tracer is always showing
the idle task as the culprit for long latencies. That happens
because critical timings are not stopped before idle loop. This
patch stops critical timings before entering the idle loop,
starting it again after the idle loop.

This problem does not affect the irqsoff tracer because
interruptions are enabled before entering the idle loop.

Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Reviewed-by: Luis Claudio R. Goncalves <lgoncalv@redhat.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/10fc3705874aef11dbe152a068b591a7be1899b4.1444314899.git.bristot@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

diff --git a/kernel/sched/idle.c b/kernel/sched/idle.c
index 8f177c7..4a2ef5a 100644 (file)
--- a/kernel/sched/idle.c
+++ b/kernel/sched/idle.c
@@ -57,9 +57,11 @@ static inline int cpu_idle_poll(void)
        rcu_idle_enter();
        trace_cpu_idle_rcuidle(0, smp_processor_id());
        local_irq_enable();
+       stop_critical_timings();
        while (!tif_need_resched() &&
                (cpu_idle_force_poll || tick_check_broadcast_expired()))
                cpu_relax();
+       start_critical_timings();
        trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
        rcu_idle_exit();
        return 1;