fix lost signals in cond vars

due to moving waiters from the cond var to the mutex in bcast, these
waiters upon wakeup would steal slots in the count from newer waiters
that had not yet been signaled, preventing the signal function from
taking any action.

to solve the problem, we simply use two separate waiter counts, and so
that the original "total" waiters count is undisturbed by broadcast
and still available for signal.

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]

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. */

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) {
+	if (c->_c_mutex != (void *)-1) c->_c_mutex = m;
+
 	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);
-	}
+	a_inc(&c->_c_waiters2);
 
 	seq = c->_c_seq;