perf: Fix PERF_EVENT_IOC_PERIOD to force-reset the period
authorPeter Zijlstra <peterz@infradead.org>
Wed, 27 Nov 2013 13:54:38 +0000 (13:54 +0000)
committerIngo Molnar <mingo@kernel.org>
Tue, 17 Dec 2013 14:21:33 +0000 (15:21 +0100)
Vince Weaver reports that, on all architectures apart from ARM,
PERF_EVENT_IOC_PERIOD doesn't actually update the period until the next
event fires. This is counter-intuitive behaviour and is better dealt
with in the core code.

This patch ensures that the period is forcefully reset when dealing with
such a request in the core code. A subsequent patch removes the
equivalent hack from the ARM back-end.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Link: http://lkml.kernel.org/r/1385560479-11014-1-git-send-email-will.deacon@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/events/core.c

index 403b781daafb694ef9baa94e5a176212a23678e1..89d34f9bb8cb7b325b91c5314e90d58ab99d29ba 100644 (file)
@@ -3527,7 +3527,7 @@ static void perf_event_for_each(struct perf_event *event,
 static int perf_event_period(struct perf_event *event, u64 __user *arg)
 {
        struct perf_event_context *ctx = event->ctx;
-       int ret = 0;
+       int ret = 0, active;
        u64 value;
 
        if (!is_sampling_event(event))
@@ -3551,6 +3551,20 @@ static int perf_event_period(struct perf_event *event, u64 __user *arg)
                event->attr.sample_period = value;
                event->hw.sample_period = value;
        }
+
+       active = (event->state == PERF_EVENT_STATE_ACTIVE);
+       if (active) {
+               perf_pmu_disable(ctx->pmu);
+               event->pmu->stop(event, PERF_EF_UPDATE);
+       }
+
+       local64_set(&event->hw.period_left, 0);
+
+       if (active) {
+               event->pmu->start(event, PERF_EF_RELOAD);
+               perf_pmu_enable(ctx->pmu);
+       }
+
 unlock:
        raw_spin_unlock_irq(&ctx->lock);