tracing/ftrace: fix a race condition in sched_switch tracer

Impact: fix race condition in sched_switch tracer

This patch fixes a race condition in the sched_switch tracer. If
several tasks (IE: concurrent initcalls) are playing with
tracing_start_cmdline_record() and tracing_stop_cmdline_record(), the
following situation could happen:

_ Task A and B are using the same tracepoint probe. Task A holds it.
  Task B is sleeping and doesn't hold it.

_ Task A frees the sched tracer, then sched_ref is decremented to 0.

_ Task A is preempted and hadn't yet unregistered its tracepoint
  probe, then B runs.

_ B increments sched_ref, sees it's 1 and then guess it has to
  register its probe. But it has not been freed by task A.

_ A lot of bad things can happen after that...

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 5 insertions, 0 deletions

diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c
index b8f56beb1a62..59de5141207c 100644
--- a/kernel/trace/trace_sched_switch.c
+++ b/kernel/trace/trace_sched_switch.c
@@ -17,6 +17,7 @@
 static struct trace_array       *ctx_trace;
 static int __read_mostly        tracer_enabled;
 static atomic_t                 sched_ref;
+static DEFINE_MUTEX(tracepoint_mutex);
 
 static void
 probe_sched_switch(struct rq *__rq, struct task_struct *prev,
@@ -125,18 +126,22 @@ static void tracing_start_sched_switch(void)
 {
         long ref;
 
+        mutex_lock(&tracepoint_mutex);
         ref = atomic_inc_return(&sched_ref);
         if (ref == 1)
                 tracing_sched_register();
+        mutex_unlock(&tracepoint_mutex);
 }
 
 static void tracing_stop_sched_switch(void)
 {
         long ref;
 
+        mutex_lock(&tracepoint_mutex);
         ref = atomic_dec_and_test(&sched_ref);
         if (ref)
                 tracing_sched_unregister();
+        mutex_unlock(&tracepoint_mutex);
 }
 
 void tracing_start_cmdline_record(void)