perf: Optimize perf_output_begin()

There's no point in re-doing the memory-barrier when we fail the
cmpxchg(). Also placing it after the space reservation loop makes it
clearer it only separates the userpage->tail read from the data
stores.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Michael Ellerman <michael@ellerman.id.au>
Cc: Michael Neuling <mikey@neuling.org>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: james.hogan@imgtec.com
Cc: Vince Weaver <vince@deater.net>
Cc: Victor Kaplansky <VICTORK@il.ibm.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Anton Blanchard <anton@samba.org>
Link: http://lkml.kernel.org/n/tip-c19u6egfldyx86tpyc3zgkw9@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

kernel/events/ring_buffer.c

@@ -141,15 +141,7 @@ int perf_output_begin(struct perf_output_handle *handle,
 	perf_output_get_handle(handle);
 
 	do {
-		/*
-		 * Userspace could choose to issue a mb() before updating the
-		 * tail pointer. So that all reads will be completed before the
-		 * write is issued.
-		 *
-		 * See perf_output_put_handle().
-		 */
 		tail = ACCESS_ONCE(rb->user_page->data_tail);
-		smp_mb();
 		offset = head = local_read(&rb->head);
 		if (!rb->overwrite &&
 		    unlikely(CIRC_SPACE(head, tail, perf_data_size(rb)) < size))
@@ -157,6 +149,15 @@ int perf_output_begin(struct perf_output_handle *handle,
 		head += size;
 	} while (local_cmpxchg(&rb->head, offset, head) != offset);
 
+	/*
+	 * Separate the userpage->tail read from the data stores below.
+	 * Matches the MB userspace SHOULD issue after reading the data
+	 * and before storing the new tail position.
+	 *
+	 * See perf_output_put_handle().
+	 */
+	smp_mb();
+
 	if (unlikely(head - local_read(&rb->wakeup) > rb->watermark))
 		local_add(rb->watermark, &rb->wakeup);