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 01d34ee..9127a7f 100644 (file)
--- a/net/mptcp/pm_netlink.c
+++ b/net/mptcp/pm_netlink.c
@@ -1039,6 +1039,7 @@ static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
                return err;
        }
 
+       inet_sk_state_store(newsk, TCP_LISTEN);
        err = kernel_listen(ssock, backlog);
        if (err) {
                pr_warn("kernel_listen error, err=%d", err);

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index a2c6ce4..4ca61e8 100644 (file)
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -2400,12 +2400,6 @@ 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);
-               }
-
                __tcp_close(ssk, 0);
 
                /* close acquired an extra ref */
@@ -2939,6 +2933,24 @@ static __poll_t mptcp_check_readable(struct mptcp_sock *msk)
        return EPOLLIN | EPOLLRDNORM;
 }
 
+static void mptcp_check_listen_stop(struct sock *sk)
+{
+       struct sock *ssk;
+
+       if (inet_sk_state_load(sk) != TCP_LISTEN)
+               return;
+
+       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);
+       tcp_set_state(ssk, TCP_CLOSE);
+       release_sock(ssk);
+}
+
 bool __mptcp_close(struct sock *sk, long timeout)
 {
        struct mptcp_subflow_context *subflow;
@@ -2949,6 +2961,7 @@ 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_check_listen_stop(sk);
                inet_sk_state_store(sk, TCP_CLOSE);
                goto cleanup;
        }
@@ -3062,6 +3075,7 @@ static int mptcp_disconnect(struct sock *sk, int flags)
        if (msk->fastopening)
                return -EBUSY;
 
+       mptcp_check_listen_stop(sk);
        inet_sk_state_store(sk, TCP_CLOSE);
 
        mptcp_stop_timer(sk);