af_unix: Annotate data-races around sk->sk_state in unix_write_space() and poll().

[ Upstream commit eb0718f ]

unix_poll() and unix_dgram_poll() read sk->sk_state locklessly and
calls unix_writable() which also reads sk->sk_state without holding
unix_state_lock().

Let's use READ_ONCE() in unix_poll() and unix_dgram_poll() and pass
it to unix_writable().

While at it, we remove TCP_SYN_SENT check in unix_dgram_poll() as
that state does not exist for AF_UNIX socket since the code was added.

Fixes: 1586a58 ("af_unix: do not report POLLOUT on listeners")
Fixes: 3c73419 ("af_unix: fix 'poll for write'/ connected DGRAM sockets")
Fixes: 1da177e ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Author: q2ven

net/unix/af_unix.c

@@ -518,9 +518,9 @@ static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other)
 	return 0;
 }
 
-static int unix_writable(const struct sock *sk)
+static int unix_writable(const struct sock *sk, unsigned char state)
 {
-	return sk->sk_state != TCP_LISTEN &&
+	return state != TCP_LISTEN &&
 	       (refcount_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
 }
 
@@ -529,7 +529,7 @@ static void unix_write_space(struct sock *sk)
 	struct socket_wq *wq;
 
 	rcu_read_lock();
-	if (unix_writable(sk)) {
+	if (unix_writable(sk, READ_ONCE(sk->sk_state))) {
 		wq = rcu_dereference(sk->sk_wq);
 		if (skwq_has_sleeper(wq))
 			wake_up_interruptible_sync_poll(&wq->wait,
@@ -3180,12 +3180,14 @@ static int unix_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned lon
 static __poll_t unix_poll(struct file *file, struct socket *sock, poll_table *wait)
 {
 	struct sock *sk = sock->sk;
+	unsigned char state;
 	__poll_t mask;
 	u8 shutdown;
 
 	sock_poll_wait(file, sock, wait);
 	mask = 0;
 	shutdown = READ_ONCE(sk->sk_shutdown);
+	state = READ_ONCE(sk->sk_state);
 
 	/* exceptional events? */
 	if (sk->sk_err)
@@ -3207,14 +3209,14 @@ static __poll_t unix_poll(struct file *file, struct socket *sock, poll_table *wa
 
 	/* Connection-based need to check for termination and startup */
 	if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
-	    sk->sk_state == TCP_CLOSE)
+	    state == TCP_CLOSE)
 		mask |= EPOLLHUP;
 
 	/*
 	 * we set writable also when the other side has shut down the
 	 * connection. This prevents stuck sockets.
 	 */
-	if (unix_writable(sk))
+	if (unix_writable(sk, state))
 		mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
 
 	return mask;
@@ -3225,12 +3227,14 @@ static __poll_t unix_dgram_poll(struct file *file, struct socket *sock,
 {
 	struct sock *sk = sock->sk, *other;
 	unsigned int writable;
+	unsigned char state;
 	__poll_t mask;
 	u8 shutdown;
 
 	sock_poll_wait(file, sock, wait);
 	mask = 0;
 	shutdown = READ_ONCE(sk->sk_shutdown);
+	state = READ_ONCE(sk->sk_state);
 
 	/* exceptional events? */
 	if (sk->sk_err || !skb_queue_empty_lockless(&sk->sk_error_queue))
@@ -3249,19 +3253,14 @@ static __poll_t unix_dgram_poll(struct file *file, struct socket *sock,
 		mask |= EPOLLIN | EPOLLRDNORM;
 
 	/* Connection-based need to check for termination and startup */
-	if (sk->sk_type == SOCK_SEQPACKET) {
-		if (sk->sk_state == TCP_CLOSE)
-			mask |= EPOLLHUP;
-		/* connection hasn't started yet? */
-		if (sk->sk_state == TCP_SYN_SENT)
-			return mask;
-	}
+	if (sk->sk_type == SOCK_SEQPACKET && state == TCP_CLOSE)
+		mask |= EPOLLHUP;
 
 	/* No write status requested, avoid expensive OUT tests. */
 	if (!(poll_requested_events(wait) & (EPOLLWRBAND|EPOLLWRNORM|EPOLLOUT)))
 		return mask;
 
-	writable = unix_writable(sk);
+	writable = unix_writable(sk, state);
 	if (writable) {
 		unix_state_lock(sk);