perf: Do poll_wait() before checking condition in perf_poll()

One should first enqueue to the waitqueue and then check for the
condition. If the condition gets true after mutex_unlock() but before
poll_wait() then we lose it and would have wait for another wakeup.

This has been like this since v2.6.31-rc1 commit c7138f3 ("perf_counter:
fix perf_poll()"). Before that it was slightly worse. I guess we get enough
wakeups so if we miss here one it doesn't really matter. It is still a
bad example.

Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
Signed-off-by: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Linus Torvalds <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>

Author: Sebastian Andrzej Siewior

kernel/events/core.c

@@ -3629,6 +3629,7 @@ static unsigned int perf_poll(struct file *file, poll_table *wait)
 	struct ring_buffer *rb;
 	unsigned int events = POLL_HUP;
 
+	poll_wait(file, &event->waitq, wait);
 	/*
 	 * Pin the event->rb by taking event->mmap_mutex; otherwise
 	 * perf_event_set_output() can swizzle our rb and make us miss wakeups.
@@ -3638,9 +3639,6 @@ static unsigned int perf_poll(struct file *file, poll_table *wait)
 	if (rb)
 		events = atomic_xchg(&rb->poll, 0);
 	mutex_unlock(&event->mmap_mutex);
-
-	poll_wait(file, &event->waitq, wait);
-
 	return events;
 }