tracing/perf: Fix double put of trace event when init fails
authorSteven Rostedt (Google) <rostedt@goodmis.org>
Tue, 16 Aug 2022 23:28:17 +0000 (19:28 -0400)
committerSteven Rostedt (Google) <rostedt@goodmis.org>
Sun, 21 Aug 2022 19:56:07 +0000 (15:56 -0400)
If in perf_trace_event_init(), the perf_trace_event_open() fails, then it
will call perf_trace_event_unreg() which will not only unregister the perf
trace event, but will also call the put() function of the tp_event.

The problem here is that the trace_event_try_get_ref() is called by the
caller of perf_trace_event_init() and if perf_trace_event_init() returns a
failure, it will then call trace_event_put(). But since the
perf_trace_event_unreg() already called the trace_event_put() function, it
triggers a WARN_ON().

 WARNING: CPU: 1 PID: 30309 at kernel/trace/trace_dynevent.c:46 trace_event_dyn_put_ref+0x15/0x20

If perf_trace_event_reg() does not call the trace_event_try_get_ref() then
the perf_trace_event_unreg() should not be calling trace_event_put(). This
breaks symmetry and causes bugs like these.

Pull out the trace_event_put() from perf_trace_event_unreg() and call it
in the locations that perf_trace_event_unreg() is called. This not only
fixes this bug, but also brings back the proper symmetry of the reg/unreg
vs get/put logic.

Link: https://lore.kernel.org/all/cover.1660347763.git.kjlx@templeofstupid.com/
Link: https://lkml.kernel.org/r/20220816192817.43d5e17f@gandalf.local.home
Cc: stable@vger.kernel.org
Fixes: 1d18538e6a092 ("tracing: Have dynamic events have a ref counter")
Reported-by: Krister Johansen <kjlx@templeofstupid.com>
Reviewed-by: Krister Johansen <kjlx@templeofstupid.com>
Tested-by: Krister Johansen <kjlx@templeofstupid.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
kernel/trace/trace_event_perf.c

index a114549..61e3a26 100644 (file)
@@ -157,7 +157,7 @@ static void perf_trace_event_unreg(struct perf_event *p_event)
        int i;
 
        if (--tp_event->perf_refcount > 0)
-               goto out;
+               return;
 
        tp_event->class->reg(tp_event, TRACE_REG_PERF_UNREGISTER, NULL);
 
@@ -176,8 +176,6 @@ static void perf_trace_event_unreg(struct perf_event *p_event)
                        perf_trace_buf[i] = NULL;
                }
        }
-out:
-       trace_event_put_ref(tp_event);
 }
 
 static int perf_trace_event_open(struct perf_event *p_event)
@@ -241,6 +239,7 @@ void perf_trace_destroy(struct perf_event *p_event)
        mutex_lock(&event_mutex);
        perf_trace_event_close(p_event);
        perf_trace_event_unreg(p_event);
+       trace_event_put_ref(p_event->tp_event);
        mutex_unlock(&event_mutex);
 }
 
@@ -292,6 +291,7 @@ void perf_kprobe_destroy(struct perf_event *p_event)
        mutex_lock(&event_mutex);
        perf_trace_event_close(p_event);
        perf_trace_event_unreg(p_event);
+       trace_event_put_ref(p_event->tp_event);
        mutex_unlock(&event_mutex);
 
        destroy_local_trace_kprobe(p_event->tp_event);
@@ -347,6 +347,7 @@ void perf_uprobe_destroy(struct perf_event *p_event)
        mutex_lock(&event_mutex);
        perf_trace_event_close(p_event);
        perf_trace_event_unreg(p_event);
+       trace_event_put_ref(p_event->tp_event);
        mutex_unlock(&event_mutex);
        destroy_local_trace_uprobe(p_event->tp_event);
 }