X-Git-Url: http://git.lttng.org./?a=blobdiff_plain;f=src%2Flib%2Flttng-ust%2Flttng-ust-comm.c;h=a46ca4e40bb4b79bf0002fc1a2f7ea0cd35e9627;hb=362ea43ac96394dc68dc22677cd9b84f5fa29b76;hp=e63ee835de458c748f7c6675d12995cd84f3ef01;hpb=edea8013ea262aa31b40368a09edce121c845088;p=lttng-ust.git

diff --git a/src/lib/lttng-ust/lttng-ust-comm.c b/src/lib/lttng-ust/lttng-ust-comm.c
index e63ee835..a46ca4e4 100644
--- a/src/lib/lttng-ust/lttng-ust-comm.c
+++ b/src/lib/lttng-ust/lttng-ust-comm.c
@@ -118,6 +118,28 @@ static int lttng_ust_comm_should_quit;
  */
 int lttng_ust_loaded __attribute__((weak));
 
+/*
+ * Notes on async-signal-safety of ust lock: a few libc functions are used
+ * which are not strictly async-signal-safe:
+ *
+ * - pthread_setcancelstate
+ * - pthread_mutex_lock
+ * - pthread_mutex_unlock
+ *
+ * As of glibc 2.35, the implementation of pthread_setcancelstate only
+ * touches TLS data, and it appears to be safe to use from signal
+ * handlers. If the libc implementation changes, this will need to be
+ * revisited, and we may ask glibc to provide an async-signal-safe
+ * pthread_setcancelstate.
+ *
+ * As of glibc 2.35, the implementation of pthread_mutex_lock/unlock
+ * for fast mutexes only relies on the pthread_mutex_t structure.
+ * Disabling signals around all uses of this mutex ensures
+ * signal-safety. If the libc implementation changes and eventually uses
+ * other global resources, this will need to be revisited and we may
+ * need to implement our own mutex.
+ */
+
 /*
  * Return 0 on success, -1 if should quit.
  * The lock is taken in both cases.
@@ -1735,18 +1757,25 @@ void wait_for_sessiond(struct sock_info *sock_info)
 
 	DBG("Waiting for %s apps sessiond", sock_info->name);
 	/* Wait for futex wakeup */
-	if (uatomic_read((int32_t *) sock_info->wait_shm_mmap))
-		goto end_wait;
-
-	while (lttng_ust_futex_async((int32_t *) sock_info->wait_shm_mmap,
-			FUTEX_WAIT, 0, NULL, NULL, 0)) {
+	while (!uatomic_read((int32_t *) sock_info->wait_shm_mmap)) {
+		if (!lttng_ust_futex_async((int32_t *) sock_info->wait_shm_mmap, FUTEX_WAIT, 0, NULL, NULL, 0)) {
+			/*
+			 * Prior queued wakeups queued by unrelated code
+			 * using the same address can cause futex wait to
+			 * return 0 even through the futex value is still
+			 * 0 (spurious wakeups). Check the value again
+			 * in user-space to validate whether it really
+			 * differs from 0.
+			 */
+			continue;
+		}
 		switch (errno) {
-		case EWOULDBLOCK:
+		case EAGAIN:
 			/* Value already changed. */
 			goto end_wait;
 		case EINTR:
 			/* Retry if interrupted by signal. */
-			break;	/* Get out of switch. */
+			break;	/* Get out of switch. Check again. */
 		case EFAULT:
 			wait_poll_fallback = 1;
 			DBG(