perf: Add PERF_EVENT_STATE_EXIT state for events with exited task

Adding new perf event state to indicate that the monitored task has
exited.  In this case the event stays alive until the owner task exits
or close the event fd while providing the last data through the read
syscall and ring buffer.

Instead it needs to propagate the error info (monitored task has died)
via poll and read  syscalls by  returning POLLHUP and 0 respectively.

Signed-off-by: Jiri Olsa <[email protected]>
Acked-by: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Cc: Adrian Hunter <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Corey Ashford <[email protected]>
Cc: David Ahern <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jean Pihet <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/n/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>

Author: olsajiri

include/linux/perf_event.h

@@ -269,6 +269,7 @@ struct pmu {
  * enum perf_event_active_state - the states of a event
  */
 enum perf_event_active_state {
+	PERF_EVENT_STATE_EXIT		= -3,
 	PERF_EVENT_STATE_ERROR		= -2,
 	PERF_EVENT_STATE_OFF		= -1,
 	PERF_EVENT_STATE_INACTIVE	=  0,

kernel/events/core.c

@@ -3600,7 +3600,8 @@ perf_read_hw(struct perf_event *event, char __user *buf, size_t count)
 	 * error state (i.e. because it was pinned but it couldn't be
 	 * scheduled on to the CPU at some point).
 	 */
-	if (event->state == PERF_EVENT_STATE_ERROR)
+	if ((event->state == PERF_EVENT_STATE_ERROR) ||
+	    (event->state == PERF_EVENT_STATE_EXIT))
 		return 0;
 
 	if (count < event->read_size)
@@ -3630,6 +3631,10 @@ static unsigned int perf_poll(struct file *file, poll_table *wait)
 	unsigned int events = POLLHUP;
 
 	poll_wait(file, &event->waitq, wait);
+
+	if (event->state == PERF_EVENT_STATE_EXIT)
+		return events;
+
 	/*
 	 * Pin the event->rb by taking event->mmap_mutex; otherwise
 	 * perf_event_set_output() can swizzle our rb and make us miss wakeups.
@@ -7588,6 +7593,9 @@ __perf_event_exit_task(struct perf_event *child_event,
 	if (child_event->parent) {
 		sync_child_event(child_event, child);
 		free_event(child_event);
+	} else {
+		child_event->state = PERF_EVENT_STATE_EXIT;
+		perf_event_wakeup(child_event);
 	}
 }