mptcp: ensure listener is unhashed before updating the sk status

commit 57fc0f1ceaa4016354cf6f88533e20b56190e41a upstream.

The MPTCP protocol access the listener subflow in a lockless
manner in a couple of places (poll, diag). That works only if
the msk itself leaves the listener status only after that the
subflow itself has been closed/disconnected. Otherwise we risk
deadlock in diag, as reported by Christoph.

Address the issue ensuring that the first subflow (the listener
one) is always disconnected before updating the msk socket status.

Reported-by: Christoph Paasch <cpaasch@apple.com>
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/407
Fixes: b29fcfb54cd7 ("mptcp: full disconnect implementation")
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/net/mptcp/pm_netlink.c b/net/mptcp/pm_netlink.c
index 5396907cc596371195e92f341b15212c55215574..7b185a064acca504b7bc1f3df407fd760ce63beb 100644 (file)
--- a/net/mptcp/pm_netlink.c
+++ b/net/mptcp/pm_netlink.c
@@ -1047,6 +1047,7 @@ static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
        if (err)
                return err;
 
+       inet_sk_state_store(newsk, TCP_LISTEN);
        err = kernel_listen(ssock, backlog);
        if (err)
                return err;

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index 59c8be9a3a5cee4c1fa6ecd3f570e6f3c8344bb4..a08b7419062c2eea64f1ebbabd4bd73e00d54a64 100644 (file)
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -2385,13 +2385,6 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk,
                kfree_rcu(subflow, rcu);
        } else {
                /* otherwise tcp will dispose of the ssk and subflow ctx */
-               if (ssk->sk_state == TCP_LISTEN) {
-                       tcp_set_state(ssk, TCP_CLOSE);
-                       mptcp_subflow_queue_clean(sk, ssk);
-                       inet_csk_listen_stop(ssk);
-                       mptcp_event_pm_listener(ssk, MPTCP_EVENT_LISTENER_CLOSED);
-               }
-
                __tcp_close(ssk, 0);
 
                /* close acquired an extra ref */
@@ -2926,10 +2919,24 @@ static __poll_t mptcp_check_readable(struct mptcp_sock *msk)
        return EPOLLIN | EPOLLRDNORM;
 }
 
-static void mptcp_listen_inuse_dec(struct sock *sk)
+static void mptcp_check_listen_stop(struct sock *sk)
 {
-       if (inet_sk_state_load(sk) == TCP_LISTEN)
-               sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
+       struct sock *ssk;
+
+       if (inet_sk_state_load(sk) != TCP_LISTEN)
+               return;
+
+       sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
+       ssk = mptcp_sk(sk)->first;
+       if (WARN_ON_ONCE(!ssk || inet_sk_state_load(ssk) != TCP_LISTEN))
+               return;
+
+       lock_sock_nested(ssk, SINGLE_DEPTH_NESTING);
+       mptcp_subflow_queue_clean(sk, ssk);
+       inet_csk_listen_stop(ssk);
+       mptcp_event_pm_listener(ssk, MPTCP_EVENT_LISTENER_CLOSED);
+       tcp_set_state(ssk, TCP_CLOSE);
+       release_sock(ssk);
 }
 
 bool __mptcp_close(struct sock *sk, long timeout)
@@ -2942,7 +2949,7 @@ bool __mptcp_close(struct sock *sk, long timeout)
        WRITE_ONCE(sk->sk_shutdown, SHUTDOWN_MASK);
 
        if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) {
-               mptcp_listen_inuse_dec(sk);
+               mptcp_check_listen_stop(sk);
                inet_sk_state_store(sk, TCP_CLOSE);
                goto cleanup;
        }
@@ -3056,7 +3063,7 @@ static int mptcp_disconnect(struct sock *sk, int flags)
        if (msk->fastopening)
                return -EBUSY;
 
-       mptcp_listen_inuse_dec(sk);
+       mptcp_check_listen_stop(sk);
        inet_sk_state_store(sk, TCP_CLOSE);
 
        mptcp_stop_timer(sk);