[PATCH 1/1] Fix: trace sched switch start/stop refcount racy updates

[Mathieu Desnoyers]

Reading the sched_cmdline_ref and sched_tgid_ref initial state within
tracing_start_sched_switch without holding the sched_register_mutex is
racy against concurrent updates, which can lead to tracepoint probes
being registered more than once (and thus trigger warnings within
tracepoint.c).

[ Compile-tested only. I suspect it might fix the following syzbot
  report:

  syzbot+774fddf07b7ab29a1e55@xxxxxxxxxxxxxxxxxxxxxxxxx ]

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx>
CC: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
CC: Joel Fernandes (Google) <joel@xxxxxxxxxxxxxxxxx>
CC: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
CC: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
CC: Paul E. McKenney <paulmck@xxxxxxxxxxxxx>
---
 kernel/trace/trace_sched_switch.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c
index e288168661e1..e304196d7c28 100644
--- a/kernel/trace/trace_sched_switch.c
+++ b/kernel/trace/trace_sched_switch.c
@@ -89,8 +89,10 @@ static void tracing_sched_unregister(void)
 
 static void tracing_start_sched_switch(int ops)
 {
-	bool sched_register = (!sched_cmdline_ref && !sched_tgid_ref);
+	bool sched_register;
+
 	mutex_lock(&sched_register_mutex);
+	sched_register = (!sched_cmdline_ref && !sched_tgid_ref);
 
 	switch (ops) {
 	case RECORD_CMDLINE:
-- 
2.11.0