tcp: allow again tcp_disconnect() when threads are waiting

As reported by Tom, .NET and applications build on top of it rely
on connect(AF_UNSPEC) to async cancel pending I/O operations on TCP
socket.

The blamed commit below caused a regression, as such cancellation
can now fail.

As suggested by Eric, this change addresses the problem explicitly
causing blocking I/O operation to terminate immediately (with an error)
when a concurrent disconnect() is executed.

Instead of tracking the number of threads blocked on a given socket,
track the number of disconnect() issued on such socket. If such counter
changes after a blocking operation releasing and re-acquiring the socket
lock, error out the current operation.

Fixes: 4faeee0cf8a5 ("tcp: deny tcp_disconnect() when threads are waiting")
Reported-by: Tom Deseyn <tdeseyn@redhat.com>
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1886305
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/f3b95e47e3dbed840960548aebaa8d954372db41.1697008693.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

8 files changed, 47 insertions, 32 deletions

diff --git a/net/core/stream.c b/net/core/stream.c
index f5c4e47df165..96fbcb9bbb30 100644
--- a/net/core/stream.c
+++ b/net/core/stream.c
@@ -117,7 +117,7 @@ EXPORT_SYMBOL(sk_stream_wait_close);
  */
 int sk_stream_wait_memory(struct sock *sk, long *timeo_p)
 {
-	int err = 0;
+	int ret, err = 0;
 	long vm_wait = 0;
 	long current_timeo = *timeo_p;
 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
@@ -142,11 +142,13 @@ int sk_stream_wait_memory(struct sock *sk, long *timeo_p)
 
 		set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
 		sk->sk_write_pending++;
-		sk_wait_event(sk, &current_timeo, READ_ONCE(sk->sk_err) ||
-						  (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN) ||
-						  (sk_stream_memory_free(sk) &&
-						  !vm_wait), &wait);
+		ret = sk_wait_event(sk, &current_timeo, READ_ONCE(sk->sk_err) ||
+				    (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN) ||
+				    (sk_stream_memory_free(sk) && !vm_wait),
+				    &wait);
 		sk->sk_write_pending--;
+		if (ret < 0)
+			goto do_error;
 
 		if (vm_wait) {
 			vm_wait -= current_timeo;
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 3d2e30e20473..2713c9b06c4c 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -597,7 +597,6 @@ static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
 
 	add_wait_queue(sk_sleep(sk), &wait);
 	sk->sk_write_pending += writebias;
-	sk->sk_wait_pending++;
 
 	/* Basic assumption: if someone sets sk->sk_err, he _must_
 	 * change state of the socket from TCP_SYN_*.
@@ -613,7 +612,6 @@ static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
 	}
 	remove_wait_queue(sk_sleep(sk), &wait);
 	sk->sk_write_pending -= writebias;
-	sk->sk_wait_pending--;
 	return timeo;
 }
 
@@ -642,6 +640,7 @@ int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
 			return -EINVAL;
 
 		if (uaddr->sa_family == AF_UNSPEC) {
+			sk->sk_disconnects++;
 			err = sk->sk_prot->disconnect(sk, flags);
 			sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
 			goto out;
@@ -696,6 +695,7 @@ int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
 		int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
 				tcp_sk(sk)->fastopen_req &&
 				tcp_sk(sk)->fastopen_req->data ? 1 : 0;
+		int dis = sk->sk_disconnects;
 
 		/* Error code is set above */
 		if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
@@ -704,6 +704,11 @@ int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
 		err = sock_intr_errno(timeo);
 		if (signal_pending(current))
 			goto out;
+
+		if (dis != sk->sk_disconnects) {
+			err = -EPIPE;
+			goto out;
+		}
 	}
 
 	/* Connection was closed by RST, timeout, ICMP error
@@ -725,6 +730,7 @@ out:
 sock_error:
 	err = sock_error(sk) ? : -ECONNABORTED;
 	sock->state = SS_UNCONNECTED;
+	sk->sk_disconnects++;
 	if (sk->sk_prot->disconnect(sk, flags))
 		sock->state = SS_DISCONNECTING;
 	goto out;
diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index aeebe8816689..394a498c2823 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -1145,7 +1145,6 @@ struct sock *inet_csk_clone_lock(const struct sock *sk,
 	if (newsk) {
 		struct inet_connection_sock *newicsk = inet_csk(newsk);
 
-		newsk->sk_wait_pending = 0;
 		inet_sk_set_state(newsk, TCP_SYN_RECV);
 		newicsk->icsk_bind_hash = NULL;
 		newicsk->icsk_bind2_hash = NULL;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 3f66cdeef7de..d3456cf840de 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -831,7 +831,9 @@ ssize_t tcp_splice_read(struct socket *sock, loff_t *ppos,
 			 */
 			if (!skb_queue_empty(&sk->sk_receive_queue))
 				break;
-			sk_wait_data(sk, &timeo, NULL);
+			ret = sk_wait_data(sk, &timeo, NULL);
+			if (ret < 0)
+				break;
 			if (signal_pending(current)) {
 				ret = sock_intr_errno(timeo);
 				break;
@@ -2442,7 +2444,11 @@ static int tcp_recvmsg_locked(struct sock *sk, struct msghdr *msg, size_t len,
 			__sk_flush_backlog(sk);
 		} else {
 			tcp_cleanup_rbuf(sk, copied);
-			sk_wait_data(sk, &timeo, last);
+			err = sk_wait_data(sk, &timeo, last);
+			if (err < 0) {
+				err = copied ? : err;
+				goto out;
+			}
 		}
 
 		if ((flags & MSG_PEEK) &&
@@ -2966,12 +2972,6 @@ int tcp_disconnect(struct sock *sk, int flags)
 	int old_state = sk->sk_state;
 	u32 seq;
 
-	/* Deny disconnect if other threads are blocked in sk_wait_event()
-	 * or inet_wait_for_connect().
-	 */
-	if (sk->sk_wait_pending)
-		return -EBUSY;
-
 	if (old_state != TCP_CLOSE)
 		tcp_set_state(sk, TCP_CLOSE);
 
diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
index 327268203001..ba2e92188124 100644
--- a/net/ipv4/tcp_bpf.c
+++ b/net/ipv4/tcp_bpf.c
@@ -307,6 +307,8 @@ msg_bytes_ready:
 		}
 
 		data = tcp_msg_wait_data(sk, psock, timeo);
+		if (data < 0)
+			return data;
 		if (data && !sk_psock_queue_empty(psock))
 			goto msg_bytes_ready;
 		copied = -EAGAIN;
@@ -351,6 +353,8 @@ msg_bytes_ready:
 
 		timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 		data = tcp_msg_wait_data(sk, psock, timeo);
+		if (data < 0)
+			return data;
 		if (data) {
 			if (!sk_psock_queue_empty(psock))
 				goto msg_bytes_ready;
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index c3b83cb390d9..d1902373c974 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -3098,12 +3098,6 @@ static int mptcp_disconnect(struct sock *sk, int flags)
 {
 	struct mptcp_sock *msk = mptcp_sk(sk);
 
-	/* Deny disconnect if other threads are blocked in sk_wait_event()
-	 * or inet_wait_for_connect().
-	 */
-	if (sk->sk_wait_pending)
-		return -EBUSY;
-
 	/* We are on the fastopen error path. We can't call straight into the
 	 * subflows cleanup code due to lock nesting (we are already under
 	 * msk->firstsocket lock).
@@ -3173,7 +3167,6 @@ struct sock *mptcp_sk_clone_init(const struct sock *sk,
 		inet_sk(nsk)->pinet6 = mptcp_inet6_sk(nsk);
 #endif
 
-	nsk->sk_wait_pending = 0;
 	__mptcp_init_sock(nsk);
 
 	msk = mptcp_sk(nsk);
diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c
index 02f583ff9239..002483e60c19 100644
--- a/net/tls/tls_main.c
+++ b/net/tls/tls_main.c
@@ -139,8 +139,8 @@ void update_sk_prot(struct sock *sk, struct tls_context *ctx)
 
 int wait_on_pending_writer(struct sock *sk, long *timeo)
 {
-	int rc = 0;
 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
+	int ret, rc = 0;
 
 	add_wait_queue(sk_sleep(sk), &wait);
 	while (1) {
@@ -154,9 +154,13 @@ int wait_on_pending_writer(struct sock *sk, long *timeo)
 			break;
 		}
 
-		if (sk_wait_event(sk, timeo,
-				  !READ_ONCE(sk->sk_write_pending), &wait))
+		ret = sk_wait_event(sk, timeo,
+				    !READ_ONCE(sk->sk_write_pending), &wait);
+		if (ret) {
+			if (ret < 0)
+				rc = ret;
 			break;
+		}
 	}
 	remove_wait_queue(sk_sleep(sk), &wait);
 	return rc;
diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index d1fc295b83b5..e9d1e83a859d 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -1291,6 +1291,7 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *psock, bool nonblock,
 	struct tls_context *tls_ctx = tls_get_ctx(sk);
 	struct tls_sw_context_rx *ctx = tls_sw_ctx_rx(tls_ctx);
 	DEFINE_WAIT_FUNC(wait, woken_wake_function);
+	int ret = 0;
 	long timeo;
 
 	timeo = sock_rcvtimeo(sk, nonblock);
@@ -1302,6 +1303,9 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *psock, bool nonblock,
 		if (sk->sk_err)
 			return sock_error(sk);
 
+		if (ret < 0)
+			return ret;
+
 		if (!skb_queue_empty(&sk->sk_receive_queue)) {
 			tls_strp_check_rcv(&ctx->strp);
 			if (tls_strp_msg_ready(ctx))
@@ -1320,10 +1324,10 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *psock, bool nonblock,
 		released = true;
 		add_wait_queue(sk_sleep(sk), &wait);
 		sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
-		sk_wait_event(sk, &timeo,
-			      tls_strp_msg_ready(ctx) ||
-			      !sk_psock_queue_empty(psock),
-			      &wait);
+		ret = sk_wait_event(sk, &timeo,
+				    tls_strp_msg_ready(ctx) ||
+				    !sk_psock_queue_empty(psock),
+				    &wait);
 		sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
 		remove_wait_queue(sk_sleep(sk), &wait);
 
@@ -1852,6 +1856,7 @@ static int tls_rx_reader_acquire(struct sock *sk, struct tls_sw_context_rx *ctx,
 				 bool nonblock)
 {
 	long timeo;
+	int ret;
 
 	timeo = sock_rcvtimeo(sk, nonblock);
 
@@ -1861,14 +1866,16 @@ static int tls_rx_reader_acquire(struct sock *sk, struct tls_sw_context_rx *ctx,
 		ctx->reader_contended = 1;
 
 		add_wait_queue(&ctx->wq, &wait);
-		sk_wait_event(sk, &timeo,
-			      !READ_ONCE(ctx->reader_present), &wait);
+		ret = sk_wait_event(sk, &timeo,
+				    !READ_ONCE(ctx->reader_present), &wait);
 		remove_wait_queue(&ctx->wq, &wait);
 
 		if (timeo <= 0)
 			return -EAGAIN;
 		if (signal_pending(current))
 			return sock_intr_errno(timeo);
+		if (ret < 0)
+			return ret;
 	}
 
 	WRITE_ONCE(ctx->reader_present, 1);