Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Masami Hiramatsu <mhiramat@kernel.org>
To:     Steven Rostedt <rostedt@goodmis.org>
Cc:     linux-kernel@vger.kernel.org, Ingo Molnar <mingo@kernel.org>,
        Tom Zanussi <tom.zanussi@linux.intel.com>,
        Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Subject: [PATCH v2 02/12] tracing: Lock event_mutex before synth_event_mutex
Date:   Mon,  5 Nov 2018 18:00:43 +0900
Message-Id: <154140844377.17322.13781091165954002713.stgit@devbox>
In-Reply-To: <154140838606.17322.15294184388075458777.stgit@devbox>
References: <154140838606.17322.15294184388075458777.stgit@devbox>
User-Agent: StGit/0.17.1-dirty
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

synthetic event is using synth_event_mutex for protecting
synth_event_list, and event_trigger_write() path acquires
locks as below order.

event_trigger_write(event_mutex)
  ->trigger_process_regex(trigger_cmd_mutex)
    ->event_hist_trigger_func(synth_event_mutex)

On the other hand, synthetic event creation and deletion paths
finally calls trace_add_event_call() and trace_remove_event_call()
which acquires event_mutex. In that case, if we keep the
synth_event_mutex locked while registering/unregistering synthetic
events, its dependency will be inversed.

To avoid this issue, current synthetic event is using 2 phases
process to create/delete events. For example, it searches existing
event under synth_event_mutex to checking event-name conflict, and
unlock synth_event_mutex, then register new event under event_mutex
locked. Finally, it locks synth_event_mutex and tries to add the
new event to the list. But it can introduce complexity and a chance
of name conflict.

To solve this simpler, this introduces trace_add_event_call_nolock()
and trace_remove_event_call_nolock() which don't acquires
event_mutex inside. synthetic event can lock event_mutex before
synth_event_mutex for solving lock dependency issue simpler.

Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
---
 include/linux/trace_events.h     |    2 ++
 kernel/trace/trace_events.c      |   34 ++++++++++++++++++++++++++++------
 kernel/trace/trace_events_hist.c |   24 ++++++++++--------------
 3 files changed, 40 insertions(+), 20 deletions(-)

diff --git a/include/linux/trace_events.h b/include/linux/trace_events.h
index 4130a5497d40..3aa05593a53f 100644
--- a/include/linux/trace_events.h
+++ b/include/linux/trace_events.h
@@ -529,6 +529,8 @@ extern int trace_event_raw_init(struct trace_event_call *call);
 extern int trace_define_field(struct trace_event_call *call, const char *type,
 			      const char *name, int offset, int size,
 			      int is_signed, int filter_type);
+extern int trace_add_event_call_nolock(struct trace_event_call *call);
+extern int trace_remove_event_call_nolock(struct trace_event_call *call);
 extern int trace_add_event_call(struct trace_event_call *call);
 extern int trace_remove_event_call(struct trace_event_call *call);
 extern int trace_event_get_offsets(struct trace_event_call *call);
diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index f94be0c2827b..a3b157f689ee 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -2305,11 +2305,11 @@ __trace_early_add_new_event(struct trace_event_call *call,
 struct ftrace_module_file_ops;
 static void __add_event_to_tracers(struct trace_event_call *call);
 
-/* Add an additional event_call dynamically */
-int trace_add_event_call(struct trace_event_call *call)
+int trace_add_event_call_nolock(struct trace_event_call *call)
 {
 	int ret;
-	mutex_lock(&event_mutex);
+	lockdep_assert_held(&event_mutex);
+
 	mutex_lock(&trace_types_lock);
 
 	ret = __register_event(call, NULL);
@@ -2317,6 +2317,16 @@ int trace_add_event_call(struct trace_event_call *call)
 		__add_event_to_tracers(call);
 
 	mutex_unlock(&trace_types_lock);
+	return ret;
+}
+
+/* Add an additional event_call dynamically */
+int trace_add_event_call(struct trace_event_call *call)
+{
+	int ret;
+
+	mutex_lock(&event_mutex);
+	ret = trace_add_event_call_nolock(call);
 	mutex_unlock(&event_mutex);
 	return ret;
 }
@@ -2366,17 +2376,29 @@ static int probe_remove_event_call(struct trace_event_call *call)
 	return 0;
 }
 
-/* Remove an event_call */
-int trace_remove_event_call(struct trace_event_call *call)
+/* no event_mutex version */
+int trace_remove_event_call_nolock(struct trace_event_call *call)
 {
 	int ret;
 
-	mutex_lock(&event_mutex);
+	lockdep_assert_held(&event_mutex);
+
 	mutex_lock(&trace_types_lock);
 	down_write(&trace_event_sem);
 	ret = probe_remove_event_call(call);
 	up_write(&trace_event_sem);
 	mutex_unlock(&trace_types_lock);
+
+	return ret;
+}
+
+/* Remove an event_call */
+int trace_remove_event_call(struct trace_event_call *call)
+{
+	int ret;
+
+	mutex_lock(&event_mutex);
+	ret = trace_remove_event_call_nolock(call);
 	mutex_unlock(&event_mutex);
 
 	return ret;
diff --git a/kernel/trace/trace_events_hist.c b/kernel/trace/trace_events_hist.c
index eb908ef2ecec..1670c65389fe 100644
--- a/kernel/trace/trace_events_hist.c
+++ b/kernel/trace/trace_events_hist.c
@@ -912,7 +912,7 @@ static int register_synth_event(struct synth_event *event)
 	call->data = event;
 	call->tp = event->tp;
 
-	ret = trace_add_event_call(call);
+	ret = trace_add_event_call_nolock(call);
 	if (ret) {
 		pr_warn("Failed to register synthetic event: %s\n",
 			trace_event_name(call));
@@ -936,7 +936,7 @@ static int unregister_synth_event(struct synth_event *event)
 	struct trace_event_call *call = &event->call;
 	int ret;
 
-	ret = trace_remove_event_call(call);
+	ret = trace_remove_event_call_nolock(call);
 
 	return ret;
 }
@@ -1013,12 +1013,10 @@ static void add_or_delete_synth_event(struct synth_event *event, int delete)
 	if (delete)
 		free_synth_event(event);
 	else {
-		mutex_lock(&synth_event_mutex);
 		if (!find_synth_event(event->name))
 			list_add(&event->list, &synth_event_list);
 		else
 			free_synth_event(event);
-		mutex_unlock(&synth_event_mutex);
 	}
 }
 
@@ -1030,6 +1028,7 @@ static int create_synth_event(int argc, char **argv)
 	int i, consumed = 0, n_fields = 0, ret = 0;
 	char *name;
 
+	mutex_lock(&event_mutex);
 	mutex_lock(&synth_event_mutex);
 
 	/*
@@ -1102,8 +1101,6 @@ static int create_synth_event(int argc, char **argv)
 		goto err;
 	}
  out:
-	mutex_unlock(&synth_event_mutex);
-
 	if (event) {
 		if (delete_event) {
 			ret = unregister_synth_event(event);
@@ -1113,10 +1110,13 @@ static int create_synth_event(int argc, char **argv)
 			add_or_delete_synth_event(event, ret);
 		}
 	}
+	mutex_unlock(&synth_event_mutex);
+	mutex_unlock(&event_mutex);
 
 	return ret;
  err:
 	mutex_unlock(&synth_event_mutex);
+	mutex_unlock(&event_mutex);
 
 	for (i = 0; i < n_fields; i++)
 		free_synth_field(fields[i]);
@@ -1127,12 +1127,10 @@ static int create_synth_event(int argc, char **argv)
 
 static int release_all_synth_events(void)
 {
-	struct list_head release_events;
 	struct synth_event *event, *e;
 	int ret = 0;
 
-	INIT_LIST_HEAD(&release_events);
-
+	mutex_lock(&event_mutex);
 	mutex_lock(&synth_event_mutex);
 
 	list_for_each_entry(event, &synth_event_list, list) {
@@ -1142,16 +1140,14 @@ static int release_all_synth_events(void)
 		}
 	}
 
-	list_splice_init(&event->list, &release_events);
-
-	mutex_unlock(&synth_event_mutex);
-
-	list_for_each_entry_safe(event, e, &release_events, list) {
+	list_for_each_entry_safe(event, e, &synth_event_list, list) {
 		list_del(&event->list);
 
 		ret = unregister_synth_event(event);
 		add_or_delete_synth_event(event, !ret);
 	}
+	mutex_unlock(&synth_event_mutex);
+	mutex_unlock(&event_mutex);
 
 	return ret;
 }