diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
index 63639ec250dd39bd0799dc24f35446e25f6cde2e..bbb4502f089874c1da1be26ae526b0a46168d895 100644
--- a/src/internal/pthread_impl.h
+++ b/src/internal/pthread_impl.h
@@ -70,6 +70,7 @@ struct __timer {
 #define _c_clock __u.__i[4]
 #define _c_lock __u.__i[5]
 #define _c_lockwait __u.__i[6]
+#define _c_waiters2 __u.__i[7]
 #define _rw_lock __u.__i[0]
 #define _rw_waiters __u.__i[1]
 #define _b_inst __u.__p[0]
diff --git a/src/thread/pthread_cond_broadcast.c b/src/thread/pthread_cond_broadcast.c
index bf6de048c1f83a7ef708964deb42a1cc9d9c6419..9c6a462b781bd5bb6435df47f6890c02c41b2f4a 100644
--- a/src/thread/pthread_cond_broadcast.c
+++ b/src/thread/pthread_cond_broadcast.c
@@ -22,8 +22,12 @@ int pthread_cond_broadcast(pthread_cond_t *c)
 	m = c->_c_mutex;
 
 	/* Move waiter count to the mutex */
-	a_fetch_add(&m->_m_waiters, c->_c_waiters);
-	a_store(&c->_c_waiters, 0);
+	for (;;) {
+		int w = c->_c_waiters2;
+		a_fetch_add(&m->_m_waiters, w);
+		if (a_cas(&c->_c_waiters2, w, 0) == w) break;
+		a_fetch_add(&m->_m_waiters, -w);
+	}
 
 	/* Perform the futex requeue, waking one waiter unless we know
 	 * that the calling thread holds the mutex. */
diff --git a/src/thread/pthread_cond_timedwait.c b/src/thread/pthread_cond_timedwait.c
index e9b5e2fcb82ffa4f76dd1efee77e5b2ca9d287db..e3dc8147c8b3ae89569a6f4d4a2751147ed5d9e4 100644
--- a/src/thread/pthread_cond_timedwait.c
+++ b/src/thread/pthread_cond_timedwait.c
@@ -7,6 +7,8 @@ struct cm {
 
 static void unwait(pthread_cond_t *c, pthread_mutex_t *m)
 {
+	int w;
+
 	/* Removing a waiter is non-trivial if we could be using requeue
 	 * based broadcast signals, due to mutex access issues, etc. */
 
@@ -18,8 +20,10 @@ static void unwait(pthread_cond_t *c, pthread_mutex_t *m)
 	while (a_swap(&c->_c_lock, 1))
 		__wait(&c->_c_lock, &c->_c_lockwait, 1, 1);
 
-	if (c->_c_waiters) c->_c_waiters--;
-	else a_dec(&m->_m_waiters);
+	/* Atomically decrement waiters2 if positive, else mutex waiters. */
+	do w = c->_c_waiters2;
+	while (w && a_cas(&c->_c_waiters2, w, w-1)!=w);
+	if (!w) a_dec(&m->_m_waiters);
 
 	a_store(&c->_c_lock, 0);
 	if (c->_c_lockwait) __wake(&c->_c_lock, 1, 1);
@@ -42,16 +46,10 @@ int pthread_cond_timedwait(pthread_cond_t *c, pthread_mutex_t *m, const struct t
 
 	pthread_testcancel();
 
-	if (c->_c_mutex == (void *)-1) {
-		a_inc(&c->_c_waiters);
-	} else {
-		c->_c_mutex = m;
-		while (a_swap(&c->_c_lock, 1))
-			__wait(&c->_c_lock, &c->_c_lockwait, 1, 1);
-		c->_c_waiters++;
-		a_store(&c->_c_lock, 0);
-		if (c->_c_lockwait) __wake(&c->_c_lock, 1, 1);
-	}
+	if (c->_c_mutex != (void *)-1) c->_c_mutex = m;
+
+	a_inc(&c->_c_waiters);
+	a_inc(&c->_c_waiters2);
 
 	seq = c->_c_seq;