From 786fc12457268cc9b555dde6c22ae7300d4b40e1 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:03 -0700 Subject: mptcp: fix connect timeout handling Ondrej reported a functional issue WRT timeout handling on connect with a nice reproducer. The problem is that the current mptcp connect waits for both the MPTCP socket level timeout, and the first subflow socket timeout. The latter is not influenced/touched by the exposed setsockopt(). Overall the above makes the SO_SNDTIMEO a no-op on connect. Since mptcp_connect is invoked via inet_stream_connect and the latter properly handle the MPTCP level timeout, we can address the issue making the nested subflow level connect always unblocking. This also allow simplifying a bit the code, dropping an ugly hack to handle the fastopen and custom proto_ops connect. The issues predates the blamed commit below, but the current resolution requires the infrastructure introduced there. Fixes: 54f1944ed6d2 ("mptcp: factor out mptcp_connect()") Reported-by: Ondrej Mosnacek Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/399 Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 29 +++++++---------------------- 1 file changed, 7 insertions(+), 22 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 08dc53f56bc2..9cafd3b89908 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -1702,7 +1702,6 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg, lock_sock(ssk); msg->msg_flags |= MSG_DONTWAIT; - msk->connect_flags = O_NONBLOCK; msk->fastopening = 1; ret = tcp_sendmsg_fastopen(ssk, msg, copied_syn, len, NULL); msk->fastopening = 0; @@ -3617,9 +3616,9 @@ static int mptcp_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) * acquired the subflow socket lock, too. */ if (msk->fastopening) - err = __inet_stream_connect(ssock, uaddr, addr_len, msk->connect_flags, 1); + err = __inet_stream_connect(ssock, uaddr, addr_len, O_NONBLOCK, 1); else - err = inet_stream_connect(ssock, uaddr, addr_len, msk->connect_flags); + err = inet_stream_connect(ssock, uaddr, addr_len, O_NONBLOCK); inet_sk(sk)->defer_connect = inet_sk(ssock->sk)->defer_connect; /* on successful connect, the msk state will be moved to established by @@ -3632,12 +3631,10 @@ static int mptcp_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len) mptcp_copy_inaddrs(sk, ssock->sk); - /* unblocking connect, mptcp-level inet_stream_connect will error out - * without changing the socket state, update it here. + /* silence EINPROGRESS and let the caller inet_stream_connect + * handle the connection in progress */ - if (err == -EINPROGRESS) - sk->sk_socket->state = ssock->state; - return err; + return 0; } static struct proto mptcp_prot = { @@ -3696,18 +3693,6 @@ unlock: return err; } -static int mptcp_stream_connect(struct socket *sock, struct sockaddr *uaddr, - int addr_len, int flags) -{ - int ret; - - lock_sock(sock->sk); - mptcp_sk(sock->sk)->connect_flags = flags; - ret = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); - release_sock(sock->sk); - return ret; -} - static int mptcp_listen(struct socket *sock, int backlog) { struct mptcp_sock *msk = mptcp_sk(sock->sk); @@ -3859,7 +3844,7 @@ static const struct proto_ops mptcp_stream_ops = { .owner = THIS_MODULE, .release = inet_release, .bind = mptcp_bind, - .connect = mptcp_stream_connect, + .connect = inet_stream_connect, .socketpair = sock_no_socketpair, .accept = mptcp_stream_accept, .getname = inet_getname, @@ -3954,7 +3939,7 @@ static const struct proto_ops mptcp_v6_stream_ops = { .owner = THIS_MODULE, .release = inet6_release, .bind = mptcp_bind, - .connect = mptcp_stream_connect, + .connect = inet_stream_connect, .socketpair = sock_no_socketpair, .accept = mptcp_stream_accept, .getname = inet6_getname, -- cgit From 5b825727d0871b23e8867f6371183e61628b4a26 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:04 -0700 Subject: mptcp: add annotations around msk->subflow accesses The MPTCP can access the first subflow socket in a few spots outside the socket lock scope. That is actually safe, as MPTCP will delete the socket itself only after the msk sock close(). Still the such accesses causes a few KCSAN splats, as reported by Christoph. Silence the harmless warning adding a few annotation around the relevant accesses. Fixes: 71ba088ce0aa ("mptcp: cleanup accept and poll") Reported-by: Christoph Paasch Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/402 Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 9cafd3b89908..ce9de2c946b0 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -91,7 +91,7 @@ static int __mptcp_socket_create(struct mptcp_sock *msk) return err; msk->first = ssock->sk; - msk->subflow = ssock; + WRITE_ONCE(msk->subflow, ssock); subflow = mptcp_subflow_ctx(ssock->sk); list_add(&subflow->node, &msk->conn_list); sock_hold(ssock->sk); @@ -2282,7 +2282,7 @@ static void mptcp_dispose_initial_subflow(struct mptcp_sock *msk) { if (msk->subflow) { iput(SOCK_INODE(msk->subflow)); - msk->subflow = NULL; + WRITE_ONCE(msk->subflow, NULL); } } @@ -3136,7 +3136,7 @@ struct sock *mptcp_sk_clone(const struct sock *sk, msk = mptcp_sk(nsk); msk->local_key = subflow_req->local_key; msk->token = subflow_req->token; - msk->subflow = NULL; + WRITE_ONCE(msk->subflow, NULL); msk->in_accept_queue = 1; WRITE_ONCE(msk->fully_established, false); if (mp_opt->suboptions & OPTION_MPTCP_CSUMREQD) @@ -3184,7 +3184,7 @@ static struct sock *mptcp_accept(struct sock *sk, int flags, int *err, struct socket *listener; struct sock *newsk; - listener = msk->subflow; + listener = READ_ONCE(msk->subflow); if (WARN_ON_ONCE(!listener)) { *err = -EINVAL; return NULL; @@ -3736,10 +3736,10 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock, pr_debug("msk=%p", msk); - /* buggy applications can call accept on socket states other then LISTEN + /* Buggy applications can call accept on socket states other then LISTEN * but no need to allocate the first subflow just to error out. */ - ssock = msk->subflow; + ssock = READ_ONCE(msk->subflow); if (!ssock) return -EINVAL; @@ -3813,10 +3813,12 @@ static __poll_t mptcp_poll(struct file *file, struct socket *sock, state = inet_sk_state_load(sk); pr_debug("msk=%p state=%d flags=%lx", msk, state, msk->flags); if (state == TCP_LISTEN) { - if (WARN_ON_ONCE(!msk->subflow || !msk->subflow->sk)) + struct socket *ssock = READ_ONCE(msk->subflow); + + if (WARN_ON_ONCE(!ssock || !ssock->sk)) return 0; - return inet_csk_listen_poll(msk->subflow->sk); + return inet_csk_listen_poll(ssock->sk); } if (state != TCP_SYN_SENT && state != TCP_SYN_RECV) { -- cgit From 7e8b88ec35eef363040e08d99536d2bebef83774 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:05 -0700 Subject: mptcp: consolidate passive msk socket initialization When the msk socket is cloned at MPC handshake time, a few fields are initialized in a racy way outside mptcp_sk_clone() and the msk socket lock. The above is due historical reasons: before commit a88d0092b24b ("mptcp: simplify subflow_syn_recv_sock()") as the first subflow socket carrying all the needed date was not available yet at msk creation time We can now refactor the code moving the missing initialization bit under the socket lock, removing the init race and avoiding some code duplication. This will also simplify the next patch, as all msk->first write access are now under the msk socket lock. Fixes: 0397c6d85f9c ("mptcp: keep unaccepted MPC subflow into join list") Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 35 ++++++++++++++++++++++++++++------- 1 file changed, 28 insertions(+), 7 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index ce9de2c946b0..2ecd0117ab1b 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -3038,7 +3038,7 @@ static void mptcp_close(struct sock *sk, long timeout) sock_put(sk); } -void mptcp_copy_inaddrs(struct sock *msk, const struct sock *ssk) +static void mptcp_copy_inaddrs(struct sock *msk, const struct sock *ssk) { #if IS_ENABLED(CONFIG_MPTCP_IPV6) const struct ipv6_pinfo *ssk6 = inet6_sk(ssk); @@ -3115,9 +3115,10 @@ static struct ipv6_pinfo *mptcp_inet6_sk(const struct sock *sk) } #endif -struct sock *mptcp_sk_clone(const struct sock *sk, - const struct mptcp_options_received *mp_opt, - struct request_sock *req) +struct sock *mptcp_sk_clone_init(const struct sock *sk, + const struct mptcp_options_received *mp_opt, + struct sock *ssk, + struct request_sock *req) { struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); struct sock *nsk = sk_clone_lock(sk, GFP_ATOMIC); @@ -3149,10 +3150,30 @@ struct sock *mptcp_sk_clone(const struct sock *sk, msk->setsockopt_seq = mptcp_sk(sk)->setsockopt_seq; sock_reset_flag(nsk, SOCK_RCU_FREE); - /* will be fully established after successful MPC subflow creation */ - inet_sk_state_store(nsk, TCP_SYN_RECV); - security_inet_csk_clone(nsk, req); + + /* this can't race with mptcp_close(), as the msk is + * not yet exposted to user-space + */ + inet_sk_state_store(nsk, TCP_ESTABLISHED); + + /* The msk maintain a ref to each subflow in the connections list */ + WRITE_ONCE(msk->first, ssk); + list_add(&mptcp_subflow_ctx(ssk)->node, &msk->conn_list); + sock_hold(ssk); + + /* new mpc subflow takes ownership of the newly + * created mptcp socket + */ + mptcp_token_accept(subflow_req, msk); + + /* set msk addresses early to ensure mptcp_pm_get_local_id() + * uses the correct data + */ + mptcp_copy_inaddrs(nsk, ssk); + mptcp_propagate_sndbuf(nsk, ssk); + + mptcp_rcv_space_init(msk, ssk); bh_unlock_sock(nsk); /* note: the newly allocated socket refcount is 2 now */ -- cgit From 1b1b43ee7a208096ecd79e626f2fc90d4a321111 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:06 -0700 Subject: mptcp: fix data race around msk->first access The first subflow socket is accessed outside the msk socket lock by mptcp_subflow_fail(), we need to annotate each write access with WRITE_ONCE, but a few spots still lacks it. Fixes: 76a13b315709 ("mptcp: invoke MP_FAIL response when needed") Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 2ecd0117ab1b..a7dd7d8c9af2 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -90,7 +90,7 @@ static int __mptcp_socket_create(struct mptcp_sock *msk) if (err) return err; - msk->first = ssock->sk; + WRITE_ONCE(msk->first, ssock->sk); WRITE_ONCE(msk->subflow, ssock); subflow = mptcp_subflow_ctx(ssock->sk); list_add(&subflow->node, &msk->conn_list); @@ -2419,7 +2419,7 @@ out_release: sock_put(ssk); if (ssk == msk->first) - msk->first = NULL; + WRITE_ONCE(msk->first, NULL); out: if (ssk == msk->last_snd) @@ -2720,7 +2720,7 @@ static int __mptcp_init_sock(struct sock *sk) WRITE_ONCE(msk->rmem_released, 0); msk->timer_ival = TCP_RTO_MIN; - msk->first = NULL; + WRITE_ONCE(msk->first, NULL); inet_csk(sk)->icsk_sync_mss = mptcp_sync_mss; WRITE_ONCE(msk->csum_enabled, mptcp_is_checksum_enabled(sock_net(sk))); WRITE_ONCE(msk->allow_infinite_fallback, true); -- cgit From 6b9831bfd9322b297eb6d44257808cc055fdc586 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:07 -0700 Subject: mptcp: add annotations around sk->sk_shutdown accesses Christoph reported the mptcp variant of a recently addressed plain TCP issue. Similar to commit e14cadfd80d7 ("tcp: add annotations around sk->sk_shutdown accesses") add READ/WRITE ONCE annotations to silence KCSAN reports around lockless sk_shutdown access. Fixes: 71ba088ce0aa ("mptcp: cleanup accept and poll") Reported-by: Christoph Paasch Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/401 Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 29 ++++++++++++++++------------- 1 file changed, 16 insertions(+), 13 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index a7dd7d8c9af2..af54a878ac27 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -603,7 +603,7 @@ static bool mptcp_check_data_fin(struct sock *sk) WRITE_ONCE(msk->ack_seq, msk->ack_seq + 1); WRITE_ONCE(msk->rcv_data_fin, 0); - sk->sk_shutdown |= RCV_SHUTDOWN; + WRITE_ONCE(sk->sk_shutdown, sk->sk_shutdown | RCV_SHUTDOWN); smp_mb__before_atomic(); /* SHUTDOWN must be visible first */ switch (sk->sk_state) { @@ -910,7 +910,7 @@ static void mptcp_check_for_eof(struct mptcp_sock *msk) /* hopefully temporary hack: propagate shutdown status * to msk, when all subflows agree on it */ - sk->sk_shutdown |= RCV_SHUTDOWN; + WRITE_ONCE(sk->sk_shutdown, sk->sk_shutdown | RCV_SHUTDOWN); smp_mb__before_atomic(); /* SHUTDOWN must be visible first */ sk->sk_data_ready(sk); @@ -2526,7 +2526,7 @@ static void mptcp_check_fastclose(struct mptcp_sock *msk) } inet_sk_state_store(sk, TCP_CLOSE); - sk->sk_shutdown = SHUTDOWN_MASK; + WRITE_ONCE(sk->sk_shutdown, SHUTDOWN_MASK); smp_mb__before_atomic(); /* SHUTDOWN must be visible first */ set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &msk->flags); @@ -2958,7 +2958,7 @@ bool __mptcp_close(struct sock *sk, long timeout) bool do_cancel_work = false; int subflows_alive = 0; - sk->sk_shutdown = SHUTDOWN_MASK; + WRITE_ONCE(sk->sk_shutdown, SHUTDOWN_MASK); if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) { mptcp_listen_inuse_dec(sk); @@ -3101,7 +3101,7 @@ static int mptcp_disconnect(struct sock *sk, int flags) mptcp_pm_data_reset(msk); mptcp_ca_reset(sk); - sk->sk_shutdown = 0; + WRITE_ONCE(sk->sk_shutdown, 0); sk_error_report(sk); return 0; } @@ -3806,9 +3806,6 @@ static __poll_t mptcp_check_writeable(struct mptcp_sock *msk) { struct sock *sk = (struct sock *)msk; - if (unlikely(sk->sk_shutdown & SEND_SHUTDOWN)) - return EPOLLOUT | EPOLLWRNORM; - if (sk_stream_is_writeable(sk)) return EPOLLOUT | EPOLLWRNORM; @@ -3826,6 +3823,7 @@ static __poll_t mptcp_poll(struct file *file, struct socket *sock, struct sock *sk = sock->sk; struct mptcp_sock *msk; __poll_t mask = 0; + u8 shutdown; int state; msk = mptcp_sk(sk); @@ -3842,17 +3840,22 @@ static __poll_t mptcp_poll(struct file *file, struct socket *sock, return inet_csk_listen_poll(ssock->sk); } + shutdown = READ_ONCE(sk->sk_shutdown); + if (shutdown == SHUTDOWN_MASK || state == TCP_CLOSE) + mask |= EPOLLHUP; + if (shutdown & RCV_SHUTDOWN) + mask |= EPOLLIN | EPOLLRDNORM | EPOLLRDHUP; + if (state != TCP_SYN_SENT && state != TCP_SYN_RECV) { mask |= mptcp_check_readable(msk); - mask |= mptcp_check_writeable(msk); + if (shutdown & SEND_SHUTDOWN) + mask |= EPOLLOUT | EPOLLWRNORM; + else + mask |= mptcp_check_writeable(msk); } else if (state == TCP_SYN_SENT && inet_sk(sk)->defer_connect) { /* cf tcp_poll() note about TFO */ mask |= EPOLLOUT | EPOLLWRNORM; } - if (sk->sk_shutdown == SHUTDOWN_MASK || state == TCP_CLOSE) - mask |= EPOLLHUP; - if (sk->sk_shutdown & RCV_SHUTDOWN) - mask |= EPOLLIN | EPOLLRDNORM | EPOLLRDHUP; /* This barrier is coupled with smp_wmb() in __mptcp_error_report() */ smp_rmb(); -- cgit From 55b47ca7d80814ceb63d64e032e96cd6777811e5 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Wed, 31 May 2023 12:37:08 -0700 Subject: mptcp: fix active subflow finalization Active subflow are inserted into the connection list at creation time. When the MPJ handshake completes successfully, a new subflow creation netlink event is generated correctly, but the current code wrongly avoid initializing a couple of subflow data. The above will cause misbehavior on a few exceptional events: unneeded mptcp-level retransmission on msk-level sequence wrap-around and infinite mapping fallback even when a MPJ socket is present. Address the issue factoring out the needed initialization in a new helper and invoking the latter from __mptcp_finish_join() time for passive subflow and from mptcp_finish_join() for active ones. Fixes: 0530020a7c8f ("mptcp: track and update contiguous data status") Cc: stable@vger.kernel.org Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Mat Martineau Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index af54a878ac27..67311e7d5b21 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -825,6 +825,13 @@ void mptcp_data_ready(struct sock *sk, struct sock *ssk) mptcp_data_unlock(sk); } +static void mptcp_subflow_joined(struct mptcp_sock *msk, struct sock *ssk) +{ + mptcp_subflow_ctx(ssk)->map_seq = READ_ONCE(msk->ack_seq); + WRITE_ONCE(msk->allow_infinite_fallback, false); + mptcp_event(MPTCP_EVENT_SUB_ESTABLISHED, msk, ssk, GFP_ATOMIC); +} + static bool __mptcp_finish_join(struct mptcp_sock *msk, struct sock *ssk) { struct sock *sk = (struct sock *)msk; @@ -839,6 +846,7 @@ static bool __mptcp_finish_join(struct mptcp_sock *msk, struct sock *ssk) mptcp_sock_graft(ssk, sk->sk_socket); mptcp_sockopt_sync_locked(msk, ssk); + mptcp_subflow_joined(msk, ssk); return true; } @@ -3485,14 +3493,16 @@ bool mptcp_finish_join(struct sock *ssk) return false; } - if (!list_empty(&subflow->node)) - goto out; + /* active subflow, already present inside the conn_list */ + if (!list_empty(&subflow->node)) { + mptcp_subflow_joined(msk, ssk); + return true; + } if (!mptcp_pm_allow_new_subflow(msk)) goto err_prohibited; - /* active connections are already on conn_list. - * If we can't acquire msk socket lock here, let the release callback + /* If we can't acquire msk socket lock here, let the release callback * handle it */ mptcp_data_lock(parent); @@ -3515,11 +3525,6 @@ err_prohibited: return false; } - subflow->map_seq = READ_ONCE(msk->ack_seq); - WRITE_ONCE(msk->allow_infinite_fallback, false); - -out: - mptcp_event(MPTCP_EVENT_SUB_ESTABLISHED, msk, ssk, GFP_ATOMIC); return true; } -- cgit From c2b2ae3925b65070adb27d5a31a31c376f26dec7 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:18 +0200 Subject: mptcp: handle correctly disconnect() failures Currently the mptcp code has assumes that disconnect() can fail only at mptcp_sendmsg_fastopen() time - to avoid a deadlock scenario - and don't even bother returning an error code. Soon mptcp_disconnect() will handle more error conditions: let's track them explicitly. As a bonus, explicitly annotate TCP-level disconnect as not failing: the mptcp code never blocks for event on the subflows. Fixes: 7d803344fdc3 ("mptcp: fix deadlock in fastopen error path") Cc: stable@vger.kernel.org Signed-off-by: Paolo Abeni Tested-by: Christoph Paasch Reviewed-by: Matthieu Baerts Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 67311e7d5b21..86f8a7621aff 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -1727,7 +1727,13 @@ static int mptcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg, if (ret && ret != -EINPROGRESS && ret != -ERESTARTSYS && ret != -EINTR) *copied_syn = 0; } else if (ret && ret != -EINPROGRESS) { - mptcp_disconnect(sk, 0); + /* The disconnect() op called by tcp_sendmsg_fastopen()/ + * __inet_stream_connect() can fail, due to looking check, + * see mptcp_disconnect(). + * Attempt it again outside the problematic scope. + */ + if (!mptcp_disconnect(sk, 0)) + sk->sk_socket->state = SS_UNCONNECTED; } inet_sk(sk)->defer_connect = 0; @@ -2389,7 +2395,10 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk, need_push = (flags & MPTCP_CF_PUSH) && __mptcp_retransmit_pending_data(sk); if (!dispose_it) { - tcp_disconnect(ssk, 0); + /* The MPTCP code never wait on the subflow sockets, TCP-level + * disconnect should never fail + */ + WARN_ON_ONCE(tcp_disconnect(ssk, 0)); msk->subflow->state = SS_UNCONNECTED; mptcp_subflow_ctx_reset(subflow); release_sock(ssk); @@ -2812,7 +2821,7 @@ void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how) break; fallthrough; case TCP_SYN_SENT: - tcp_disconnect(ssk, O_NONBLOCK); + WARN_ON_ONCE(tcp_disconnect(ssk, O_NONBLOCK)); break; default: if (__mptcp_check_fallback(mptcp_sk(sk))) { @@ -3075,11 +3084,10 @@ static int mptcp_disconnect(struct sock *sk, int flags) /* 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). Do nothing and leave the cleanup to the - * caller. + * msk->firstsocket lock). */ if (msk->fastopening) - return 0; + return -EBUSY; mptcp_listen_inuse_dec(sk); inet_sk_state_store(sk, TCP_CLOSE); -- cgit From 0ad529d9fd2bfa3fc619552a8d2fb2f2ef0bce2e Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:19 +0200 Subject: mptcp: fix possible divide by zero in recvmsg() Christoph reported a divide by zero bug in mptcp_recvmsg(): divide error: 0000 [#1] PREEMPT SMP CPU: 1 PID: 19978 Comm: syz-executor.6 Not tainted 6.4.0-rc2-gffcc7899081b #20 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.el7 04/01/2014 RIP: 0010:__tcp_select_window+0x30e/0x420 net/ipv4/tcp_output.c:3018 Code: 11 ff 0f b7 cd c1 e9 0c b8 ff ff ff ff d3 e0 89 c1 f7 d1 01 cb 21 c3 eb 17 e8 2e 83 11 ff 31 db eb 0e e8 25 83 11 ff 89 d8 99 7c 24 04 29 d3 65 48 8b 04 25 28 00 00 00 48 3b 44 24 10 75 60 RSP: 0018:ffffc90000a07a18 EFLAGS: 00010246 RAX: 000000000000ffd7 RBX: 000000000000ffd7 RCX: 0000000000040000 RDX: 0000000000000000 RSI: 000000000003ffff RDI: 0000000000040000 RBP: 000000000000ffd7 R08: ffffffff820cf297 R09: 0000000000000001 R10: 0000000000000000 R11: ffffffff8103d1a0 R12: 0000000000003f00 R13: 0000000000300000 R14: ffff888101cf3540 R15: 0000000000180000 FS: 00007f9af4c09640(0000) GS:ffff88813bd00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000001b33824000 CR3: 000000012f241001 CR4: 0000000000170ee0 Call Trace: __tcp_cleanup_rbuf+0x138/0x1d0 net/ipv4/tcp.c:1611 mptcp_recvmsg+0xcb8/0xdd0 net/mptcp/protocol.c:2034 inet_recvmsg+0x127/0x1f0 net/ipv4/af_inet.c:861 ____sys_recvmsg+0x269/0x2b0 net/socket.c:1019 ___sys_recvmsg+0xe6/0x260 net/socket.c:2764 do_recvmmsg+0x1a5/0x470 net/socket.c:2858 __do_sys_recvmmsg net/socket.c:2937 [inline] __se_sys_recvmmsg net/socket.c:2953 [inline] __x64_sys_recvmmsg+0xa6/0x130 net/socket.c:2953 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x47/0xa0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x72/0xdc RIP: 0033:0x7f9af58fc6a9 Code: 5c c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 4f 37 0d 00 f7 d8 64 89 01 48 RSP: 002b:00007f9af4c08cd8 EFLAGS: 00000246 ORIG_RAX: 000000000000012b RAX: ffffffffffffffda RBX: 00000000006bc050 RCX: 00007f9af58fc6a9 RDX: 0000000000000001 RSI: 0000000020000140 RDI: 0000000000000004 RBP: 0000000000000000 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000f00 R11: 0000000000000246 R12: 00000000006bc05c R13: fffffffffffffea8 R14: 00000000006bc050 R15: 000000000001fe40 mptcp_recvmsg is allowed to release the msk socket lock when blocking, and before re-acquiring it another thread could have switched the sock to TCP_LISTEN status - with a prior connect(AF_UNSPEC) - also clearing icsk_ack.rcv_mss. Address the issue preventing the disconnect if some other process is concurrently performing a blocking syscall on the same socket, alike commit 4faeee0cf8a5 ("tcp: deny tcp_disconnect() when threads are waiting"). Fixes: a6b118febbab ("mptcp: add receive buffer auto-tuning") Cc: stable@vger.kernel.org Reported-by: Christoph Paasch Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/404 Signed-off-by: Paolo Abeni Tested-by: Christoph Paasch Reviewed-by: Matthieu Baerts Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 86f8a7621aff..ee357700b27b 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -3082,6 +3082,12 @@ 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). @@ -3148,6 +3154,7 @@ 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); -- cgit From 56a666c48b038e91b76471289e2cf60c79d326b9 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:20 +0200 Subject: mptcp: fix possible list corruption on passive MPJ At passive MPJ time, if the msk socket lock is held by the user, the new subflow is appended to the msk->join_list under the msk data lock. In mptcp_release_cb()/__mptcp_flush_join_list(), the subflows in that list are moved from the join_list into the conn_list under the msk socket lock. Append and removal could race, possibly corrupting such list. Address the issue splicing the join list into a temporary one while still under the msk data lock. Found by code inspection, the race itself should be almost impossible to trigger in practice. Fixes: 3e5014909b56 ("mptcp: cleanup MPJ subflow list handling") Cc: stable@vger.kernel.org Signed-off-by: Paolo Abeni Reviewed-by: Matthieu Baerts Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index ee357700b27b..9a40dae31cec 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -850,12 +850,12 @@ static bool __mptcp_finish_join(struct mptcp_sock *msk, struct sock *ssk) return true; } -static void __mptcp_flush_join_list(struct sock *sk) +static void __mptcp_flush_join_list(struct sock *sk, struct list_head *join_list) { struct mptcp_subflow_context *tmp, *subflow; struct mptcp_sock *msk = mptcp_sk(sk); - list_for_each_entry_safe(subflow, tmp, &msk->join_list, node) { + list_for_each_entry_safe(subflow, tmp, join_list, node) { struct sock *ssk = mptcp_subflow_tcp_sock(subflow); bool slow = lock_sock_fast(ssk); @@ -3342,9 +3342,14 @@ static void mptcp_release_cb(struct sock *sk) for (;;) { unsigned long flags = (msk->cb_flags & MPTCP_FLAGS_PROCESS_CTX_NEED) | msk->push_pending; + struct list_head join_list; + if (!flags) break; + INIT_LIST_HEAD(&join_list); + list_splice_init(&msk->join_list, &join_list); + /* the following actions acquire the subflow socket lock * * 1) can't be invoked in atomic scope @@ -3355,8 +3360,9 @@ static void mptcp_release_cb(struct sock *sk) msk->push_pending = 0; msk->cb_flags &= ~flags; spin_unlock_bh(&sk->sk_lock.slock); + if (flags & BIT(MPTCP_FLUSH_JOIN_LIST)) - __mptcp_flush_join_list(sk); + __mptcp_flush_join_list(sk, &join_list); if (flags & BIT(MPTCP_PUSH_PENDING)) __mptcp_push_pending(sk, 0); if (flags & BIT(MPTCP_RETRANSMIT)) -- cgit From 81c1d029016001f994ce1c46849c5e9900d8eab8 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:21 +0200 Subject: mptcp: consolidate fallback and non fallback state machine An orphaned msk releases the used resources via the worker, when the latter first see the msk in CLOSED status. If the msk status transitions to TCP_CLOSE in the release callback invoked by the worker's final release_sock(), such instance of the workqueue will not take any action. Additionally the MPTCP code prevents scheduling the worker once the socket reaches the CLOSE status: such msk resources will be leaked. The only code path that can trigger the above scenario is the __mptcp_check_send_data_fin() in fallback mode. Address the issue removing the special handling of fallback socket in __mptcp_check_send_data_fin(), consolidating the state machine for fallback and non fallback socket. Since non-fallback sockets do not send and do not receive data_fin, the mptcp code can update the msk internal status to match the next step in the SM every time data fin (ack) should be generated or received. As a consequence we can remove a bunch of checks for fallback from the fastpath. Fixes: 6e628cd3a8f7 ("mptcp: use mptcp release_cb for delayed tasks") Cc: stable@vger.kernel.org Signed-off-by: Paolo Abeni Reviewed-by: Mat Martineau Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 41 +++++++++++++++-------------------------- 1 file changed, 15 insertions(+), 26 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 9a40dae31cec..27d206f7af62 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -44,7 +44,7 @@ enum { static struct percpu_counter mptcp_sockets_allocated ____cacheline_aligned_in_smp; static void __mptcp_destroy_sock(struct sock *sk); -static void __mptcp_check_send_data_fin(struct sock *sk); +static void mptcp_check_send_data_fin(struct sock *sk); DEFINE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); static struct net_device mptcp_napi_dev; @@ -424,8 +424,7 @@ static bool mptcp_pending_data_fin_ack(struct sock *sk) { struct mptcp_sock *msk = mptcp_sk(sk); - return !__mptcp_check_fallback(msk) && - ((1 << sk->sk_state) & + return ((1 << sk->sk_state) & (TCPF_FIN_WAIT1 | TCPF_CLOSING | TCPF_LAST_ACK)) && msk->write_seq == READ_ONCE(msk->snd_una); } @@ -583,9 +582,6 @@ static bool mptcp_check_data_fin(struct sock *sk) u64 rcv_data_fin_seq; bool ret = false; - if (__mptcp_check_fallback(msk)) - return ret; - /* Need to ack a DATA_FIN received from a peer while this side * of the connection is in ESTABLISHED, FIN_WAIT1, or FIN_WAIT2. * msk->rcv_data_fin was set when parsing the incoming options @@ -623,7 +619,8 @@ static bool mptcp_check_data_fin(struct sock *sk) } ret = true; - mptcp_send_ack(msk); + if (!__mptcp_check_fallback(msk)) + mptcp_send_ack(msk); mptcp_close_wake_up(sk); } return ret; @@ -1609,7 +1606,7 @@ out: if (!mptcp_timer_pending(sk)) mptcp_reset_timer(sk); if (do_check_data_fin) - __mptcp_check_send_data_fin(sk); + mptcp_check_send_data_fin(sk); } static void __mptcp_subflow_push_pending(struct sock *sk, struct sock *ssk, bool first) @@ -2680,8 +2677,6 @@ static void mptcp_worker(struct work_struct *work) if (unlikely((1 << state) & (TCPF_CLOSE | TCPF_LISTEN))) goto unlock; - mptcp_check_data_fin_ack(sk); - mptcp_check_fastclose(msk); mptcp_pm_nl_work(msk); @@ -2689,7 +2684,8 @@ static void mptcp_worker(struct work_struct *work) if (test_and_clear_bit(MPTCP_WORK_EOF, &msk->flags)) mptcp_check_for_eof(msk); - __mptcp_check_send_data_fin(sk); + mptcp_check_send_data_fin(sk); + mptcp_check_data_fin_ack(sk); mptcp_check_data_fin(sk); if (test_and_clear_bit(MPTCP_WORK_CLOSE_SUBFLOW, &msk->flags)) @@ -2828,6 +2824,12 @@ void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how) pr_debug("Fallback"); ssk->sk_shutdown |= how; tcp_shutdown(ssk, how); + + /* simulate the data_fin ack reception to let the state + * machine move forward + */ + WRITE_ONCE(mptcp_sk(sk)->snd_una, mptcp_sk(sk)->snd_nxt); + mptcp_schedule_work(sk); } else { pr_debug("Sending DATA_FIN on subflow %p", ssk); tcp_send_ack(ssk); @@ -2867,7 +2869,7 @@ static int mptcp_close_state(struct sock *sk) return next & TCP_ACTION_FIN; } -static void __mptcp_check_send_data_fin(struct sock *sk) +static void mptcp_check_send_data_fin(struct sock *sk) { struct mptcp_subflow_context *subflow; struct mptcp_sock *msk = mptcp_sk(sk); @@ -2885,19 +2887,6 @@ static void __mptcp_check_send_data_fin(struct sock *sk) WRITE_ONCE(msk->snd_nxt, msk->write_seq); - /* fallback socket will not get data_fin/ack, can move to the next - * state now - */ - if (__mptcp_check_fallback(msk)) { - WRITE_ONCE(msk->snd_una, msk->write_seq); - if ((1 << sk->sk_state) & (TCPF_CLOSING | TCPF_LAST_ACK)) { - inet_sk_state_store(sk, TCP_CLOSE); - mptcp_close_wake_up(sk); - } else if (sk->sk_state == TCP_FIN_WAIT1) { - inet_sk_state_store(sk, TCP_FIN_WAIT2); - } - } - mptcp_for_each_subflow(msk, subflow) { struct sock *tcp_sk = mptcp_subflow_tcp_sock(subflow); @@ -2917,7 +2906,7 @@ static void __mptcp_wr_shutdown(struct sock *sk) WRITE_ONCE(msk->write_seq, msk->write_seq + 1); WRITE_ONCE(msk->snd_data_fin_enable, 1); - __mptcp_check_send_data_fin(sk); + mptcp_check_send_data_fin(sk); } static void __mptcp_destroy_sock(struct sock *sk) -- cgit From b7535cfed223a9f02f9530853616f197b386d775 Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:22 +0200 Subject: mptcp: drop legacy code around RX EOF Thanks to the previous patch -- "mptcp: consolidate fallback and non fallback state machine" -- we can finally drop the "temporary hack" used to detect rx eof. Signed-off-by: Paolo Abeni Reviewed-by: Mat Martineau Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 49 ------------------------------------------------- 1 file changed, 49 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 27d206f7af62..a66ec341485e 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -894,49 +894,6 @@ bool mptcp_schedule_work(struct sock *sk) return false; } -void mptcp_subflow_eof(struct sock *sk) -{ - if (!test_and_set_bit(MPTCP_WORK_EOF, &mptcp_sk(sk)->flags)) - mptcp_schedule_work(sk); -} - -static void mptcp_check_for_eof(struct mptcp_sock *msk) -{ - struct mptcp_subflow_context *subflow; - struct sock *sk = (struct sock *)msk; - int receivers = 0; - - mptcp_for_each_subflow(msk, subflow) - receivers += !subflow->rx_eof; - if (receivers) - return; - - if (!(sk->sk_shutdown & RCV_SHUTDOWN)) { - /* hopefully temporary hack: propagate shutdown status - * to msk, when all subflows agree on it - */ - WRITE_ONCE(sk->sk_shutdown, sk->sk_shutdown | RCV_SHUTDOWN); - - smp_mb__before_atomic(); /* SHUTDOWN must be visible first */ - sk->sk_data_ready(sk); - } - - switch (sk->sk_state) { - case TCP_ESTABLISHED: - inet_sk_state_store(sk, TCP_CLOSE_WAIT); - break; - case TCP_FIN_WAIT1: - inet_sk_state_store(sk, TCP_CLOSING); - break; - case TCP_FIN_WAIT2: - inet_sk_state_store(sk, TCP_CLOSE); - break; - default: - return; - } - mptcp_close_wake_up(sk); -} - static struct sock *mptcp_subflow_recv_lookup(const struct mptcp_sock *msk) { struct mptcp_subflow_context *subflow; @@ -2161,9 +2118,6 @@ static int mptcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, break; } - if (test_and_clear_bit(MPTCP_WORK_EOF, &msk->flags)) - mptcp_check_for_eof(msk); - if (sk->sk_shutdown & RCV_SHUTDOWN) { /* race breaker: the shutdown could be after the * previous receive queue check @@ -2681,9 +2635,6 @@ static void mptcp_worker(struct work_struct *work) mptcp_pm_nl_work(msk); - if (test_and_clear_bit(MPTCP_WORK_EOF, &msk->flags)) - mptcp_check_for_eof(msk); - mptcp_check_send_data_fin(sk); mptcp_check_data_fin_ack(sk); mptcp_check_data_fin(sk); -- cgit From 57fc0f1ceaa4016354cf6f88533e20b56190e41a Mon Sep 17 00:00:00 2001 From: Paolo Abeni Date: Tue, 20 Jun 2023 18:24:23 +0200 Subject: mptcp: ensure listener is unhashed before updating the sk status 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 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 Reviewed-by: Matthieu Baerts Signed-off-by: Matthieu Baerts Signed-off-by: Jakub Kicinski --- net/mptcp/protocol.c | 31 +++++++++++++++++++------------ 1 file changed, 19 insertions(+), 12 deletions(-) (limited to 'net/mptcp/protocol.c') diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index a66ec341485e..a6c7f2d24909 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -2368,13 +2368,6 @@ 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 */ @@ -2902,10 +2895,24 @@ 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) @@ -2918,7 +2925,7 @@ 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; } @@ -3035,7 +3042,7 @@ 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); -- cgit