Merge branch 'mptcp-stalled-connections-fix'

Matthieu Baerts says:

====================
mptcp: fix stalled connections

Daire reported a few issues with MPTCP where some connections were
stalled in different states. Paolo did a great job fixing them.

Patch 1 fixes bogus receive window shrinkage with multiple subflows. Due
to a race condition and unlucky circumstances, that may lead to
TCP-level window shrinkage, and the connection being stalled on the
sender end.

Patch 2 is a preparation for patch 3 which processes pending subflow
errors on close. Without that and under specific circumstances, the
MPTCP-level socket might not switch to the CLOSE state and stall.

Patch 4 is also a preparation patch for the next one. Patch 5 fixes
MPTCP connections not switching to the CLOSE state when all subflows
have been closed but no DATA_FIN have been exchanged to explicitly close
the MPTCP connection. Now connections in such state will switch to the
CLOSE state after a timeout, still allowing the "make-after-break"
feature but making sure connections don't stall forever. It will be
possible to modify this timeout -- currently matching TCP TIMEWAIT value
(60 seconds) -- in a future version.
====================
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>

net/mptcp/options.c

@@ -1269,12 +1269,13 @@ static void mptcp_set_rwin(struct tcp_sock *tp, struct tcphdr *th)
 
 			if (rcv_wnd == rcv_wnd_old)
 				break;
-			if (before64(rcv_wnd_new, rcv_wnd)) {
+
+			rcv_wnd_old = rcv_wnd;
+			if (before64(rcv_wnd_new, rcv_wnd_old)) {
 				MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_RCVWNDCONFLICTUPDATE);
 				goto raise_win;
 			}
 			MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_RCVWNDCONFLICT);
-			rcv_wnd_old = rcv_wnd;
 		}
 		return;
 	}

net/mptcp/protocol.h

@@ -718,7 +718,29 @@ void mptcp_get_options(const struct sk_buff *skb,
 
 void mptcp_finish_connect(struct sock *sk);
 void __mptcp_set_connected(struct sock *sk);
-void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
+void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
+
+static inline void mptcp_stop_tout_timer(struct sock *sk)
+{
+	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
+		return;
+
+	sk_stop_timer(sk, &sk->sk_timer);
+	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
+}
+
+static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
+{
+	/* avoid 0 timestamp, as that means no close timeout */
+	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
+}
+
+static inline void mptcp_start_tout_timer(struct sock *sk)
+{
+	mptcp_set_close_tout(sk, tcp_jiffies32);
+	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
+}
+
 static inline bool mptcp_is_fully_established(struct sock *sk)
 {
 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&

net/mptcp/subflow.c

@@ -1226,7 +1226,7 @@ static void mptcp_subflow_fail(struct mptcp_sock *msk, struct sock *ssk)
 	WRITE_ONCE(subflow->fail_tout, fail_tout);
 	tcp_send_ack(ssk);
 
-	mptcp_reset_timeout(msk, subflow->fail_tout);
+	mptcp_reset_tout_timer(msk, subflow->fail_tout);
 }
 
 static bool subflow_check_data_avail(struct sock *ssk)
@@ -1362,42 +1362,6 @@ void mptcp_space(const struct sock *ssk, int *space, int *full_space)
 	*full_space = mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf));
 }
 
-void __mptcp_error_report(struct sock *sk)
-{
-	struct mptcp_subflow_context *subflow;
-	struct mptcp_sock *msk = mptcp_sk(sk);
-
-	mptcp_for_each_subflow(msk, subflow) {
-		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
-		int err = sock_error(ssk);
-		int ssk_state;
-
-		if (!err)
-			continue;
-
-		/* only propagate errors on fallen-back sockets or
-		 * on MPC connect
-		 */
-		if (sk->sk_state != TCP_SYN_SENT && !__mptcp_check_fallback(msk))
-			continue;
-
-		/* We need to propagate only transition to CLOSE state.
-		 * Orphaned socket will see such state change via
-		 * subflow_sched_work_if_closed() and that path will properly
-		 * destroy the msk as needed.
-		 */
-		ssk_state = inet_sk_state_load(ssk);
-		if (ssk_state == TCP_CLOSE && !sock_flag(sk, SOCK_DEAD))
-			inet_sk_state_store(sk, ssk_state);
-		WRITE_ONCE(sk->sk_err, -err);
-
-		/* This barrier is coupled with smp_rmb() in mptcp_poll() */
-		smp_wmb();
-		sk_error_report(sk);
-		break;
-	}
-}
-
 static void subflow_error_report(struct sock *ssk)
 {
 	struct sock *sk = mptcp_subflow_ctx(ssk)->conn;
@@ -1588,6 +1552,7 @@ int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
 	mptcp_sock_graft(ssk, sk->sk_socket);
 	iput(SOCK_INODE(sf));
 	WRITE_ONCE(msk->allow_infinite_fallback, false);
+	mptcp_stop_tout_timer(sk);
 	return 0;
 
 failed_unlink: