perf: Fix interrupt handler timing harness
authorStephane Eranian <eranian@google.com>
Thu, 4 Jul 2013 22:30:11 +0000 (00:30 +0200)
committerIngo Molnar <mingo@kernel.org>
Fri, 5 Jul 2013 06:54:43 +0000 (08:54 +0200)
This patch fixes a serious bug in:

  14c63f17b1fd perf: Drop sample rate when sampling is too slow

There was an misunderstanding on the API of the do_div()
macro. It returns the remainder of the division and this
was not what the function expected leading to disabling the
interrupt latency watchdog.

This patch also remove a duplicate assignment in
perf_sample_event_took().

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: peterz@infradead.org
Cc: dave.hansen@linux.intel.com
Cc: ak@linux.intel.com
Cc: jolsa@redhat.com
Link: http://lkml.kernel.org/r/20130704223010.GA30625@quad
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/events/core.c

index 1db3af93370410ed8538e04b785f7cb2a3e4b33c..1833bc5a84a73049b6e1f27caed87adfbdf44b2c 100644 (file)
@@ -182,7 +182,7 @@ void update_perf_cpu_limits(void)
        u64 tmp = perf_sample_period_ns;
 
        tmp *= sysctl_perf_cpu_time_max_percent;
-       tmp = do_div(tmp, 100);
+       do_div(tmp, 100);
        atomic_set(&perf_sample_allowed_ns, tmp);
 }
 
@@ -232,7 +232,7 @@ DEFINE_PER_CPU(u64, running_sample_length);
 void perf_sample_event_took(u64 sample_len_ns)
 {
        u64 avg_local_sample_len;
-       u64 local_samples_len = __get_cpu_var(running_sample_length);
+       u64 local_samples_len;
 
        if (atomic_read(&perf_sample_allowed_ns) == 0)
                return;