From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Date: Wed, 17 May 2017 22:36:54 +0000 (-0400)
Subject: Fix: thread exit vs futex wait/wakeup race
X-Git-Tag: v2.8.7~10
X-Git-Url: https://git.liburcu.org/?p=lttng-tools.git;a=commitdiff_plain;h=212a7c5ffa425113159ddf0190c2b65d75c03615

Fix: thread exit vs futex wait/wakeup race

relayd_live_stop performs, in this order:

        CMM_STORE_SHARED(live_dispatch_thread_exit, 1);   [A]
        futex_nto1_wake(&viewer_conn_queue.futex);        [B]

whereas thread_dispatcher does:

   while (!CMM_LOAD_SHARED(live_dispatch_thread_exit)) {  [1]

     [...]
     futex_nto1_prepare(&viewer_conn_queue.futex);        [2]
     [...]
     futex_nto1_wait(&viewer_conn_queue.futex);           [3]

Unfortunately, on the following sequence:

[1] [A] [B] [2] [3]

thread_dispatcher will end up hanging.

We need to move the live_dispatch_thread_exit load between "prepare" and
"wait" to fix this.

There are similar scenarios with relay_thread_dispatcher, and the
session daemon thread_dispatch_ust_registration, which are also fixed
here.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
---

diff --git a/src/bin/lttng-relayd/live.c b/src/bin/lttng-relayd/live.c
index cc88032dc..ffec5977e 100644
--- a/src/bin/lttng-relayd/live.c
+++ b/src/bin/lttng-relayd/live.c
@@ -654,12 +654,16 @@ void *thread_dispatcher(void *data)
 
 	health_code_update();
 
-	while (!CMM_LOAD_SHARED(live_dispatch_thread_exit)) {
+	for (;;) {
 		health_code_update();
 
 		/* Atomically prepare the queue futex */
 		futex_nto1_prepare(&viewer_conn_queue.futex);
 
+		if (CMM_LOAD_SHARED(live_dispatch_thread_exit)) {
+			break;
+		}
+
 		do {
 			health_code_update();
 
diff --git a/src/bin/lttng-relayd/main.c b/src/bin/lttng-relayd/main.c
index 56efaae34..9c96af127 100644
--- a/src/bin/lttng-relayd/main.c
+++ b/src/bin/lttng-relayd/main.c
@@ -969,12 +969,16 @@ static void *relay_thread_dispatcher(void *data)
 
 	health_code_update();
 
-	while (!CMM_LOAD_SHARED(dispatch_thread_exit)) {
+	for (;;) {
 		health_code_update();
 
 		/* Atomically prepare the queue futex */
 		futex_nto1_prepare(&relay_conn_queue.futex);
 
+		if (CMM_LOAD_SHARED(dispatch_thread_exit)) {
+			break;
+		}
+
 		do {
 			health_code_update();
 
diff --git a/src/bin/lttng-sessiond/main.c b/src/bin/lttng-sessiond/main.c
index 8527ad04c..50eead010 100644
--- a/src/bin/lttng-sessiond/main.c
+++ b/src/bin/lttng-sessiond/main.c
@@ -1916,12 +1916,16 @@ static void *thread_dispatch_ust_registration(void *data)
 
 	DBG("[thread] Dispatch UST command started");
 
-	while (!CMM_LOAD_SHARED(dispatch_thread_exit)) {
+	for (;;) {
 		health_code_update();
 
 		/* Atomically prepare the queue futex */
 		futex_nto1_prepare(&ust_cmd_queue.futex);
 
+		if (CMM_LOAD_SHARED(dispatch_thread_exit)) {
+			break;
+		}
+
 		do {
 			struct ust_app *app = NULL;
 			ust_cmd = NULL;