perf: Generate EXIT event only once per task context
authorNamhyung Kim <namhyung.kim@lge.com>
Fri, 15 Mar 2013 07:27:13 +0000 (16:27 +0900)
committerIngo Molnar <mingo@kernel.org>
Mon, 18 Mar 2013 08:47:33 +0000 (09:47 +0100)
perf_event_task_event() iterates pmu list and generate events
for each eligible pmu context.  But if task_event has task_ctx
like in EXIT it'll generate events even though the pmu doesn't
have an eligible one. Fix it by moving the code to proper
places.

Before this patch:

  $ perf record -n true
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.006 MB perf.data (~248 samples) ]

  $ perf report -D | tail
  Aggregated stats:
             TOTAL events:         73
              MMAP events:         67
              COMM events:          2
              EXIT events:          4
  cycles stats:
             TOTAL events:         73
              MMAP events:         67
              COMM events:          2
              EXIT events:          4

After this patch:

  $ perf report -D | tail
  Aggregated stats:
             TOTAL events:         70
              MMAP events:         67
              COMM events:          2
              EXIT events:          1
  cycles stats:
             TOTAL events:         70
              MMAP events:         67
              COMM events:          2
              EXIT events:          1

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1363332433-7637-1-git-send-email-namhyung@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/events/core.c

index fa79c377d65dd027946d2bb141fad0893bbc4ee2..59412d037eed2a81163483b89fb38d9feb30623a 100644 (file)
@@ -4434,12 +4434,15 @@ static void perf_event_task_event(struct perf_task_event *task_event)
                        if (ctxn < 0)
                                goto next;
                        ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
+                       if (ctx)
+                               perf_event_task_ctx(ctx, task_event);
                }
-               if (ctx)
-                       perf_event_task_ctx(ctx, task_event);
 next:
                put_cpu_ptr(pmu->pmu_cpu_context);
        }
+       if (task_event->task_ctx)
+               perf_event_task_ctx(task_event->task_ctx, task_event);
+
        rcu_read_unlock();
 }