From: Paolo Abeni pabeni@redhat.com

After commit 30e51b923e43 ("mptcp: fix unreleased socket in accept queue") unaccepted msk sockets go throu complete shutdown, we don't need anymore to delay inserting the first subflow into the subflow lists.

The reference counting deserve some extra care, as __mptcp_close() is unaware of the request socket linkage to the first subflow.

Please note that this is more a refactoring than a fix but because this modification is needed to include other corrections, see the following commits. Then a Fixes tag has been added here to help the stable team.

Fixes: 30e51b923e43 ("mptcp: fix unreleased socket in accept queue") Cc: stable@vger.kernel.org # v6.0+ Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Tested-by: Christoph Paasch cpaasch@apple.com Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/protocol.c | 17 ----------------- net/mptcp/subflow.c | 27 +++++++++++++++++++++------ 2 files changed, 21 insertions(+), 23 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 3ad9c46202fc..447641d34c2c 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -825,7 +825,6 @@ static bool __mptcp_finish_join(struct mptcp_sock *msk, struct sock *ssk) if (sk->sk_socket && !ssk->sk_socket) mptcp_sock_graft(ssk, sk->sk_socket);

- mptcp_propagate_sndbuf((struct sock *)msk, ssk); mptcp_sockopt_sync_locked(msk, ssk); return true; } @@ -3708,22 +3707,6 @@ static int mptcp_stream_accept(struct socket *sock, struct socket *newsock,

lock_sock(newsk);

- /* PM/worker can now acquire the first subflow socket - * lock without racing with listener queue cleanup, - * we can notify it, if needed. - * - * Even if remote has reset the initial subflow by now - * the refcnt is still at least one. - */ - subflow = mptcp_subflow_ctx(msk->first); - list_add(&subflow->node, &msk->conn_list); - sock_hold(msk->first); - if (mptcp_is_fully_established(newsk)) - mptcp_pm_fully_established(msk, msk->first, GFP_KERNEL); - - mptcp_rcv_space_init(msk, msk->first); - mptcp_propagate_sndbuf(newsk, msk->first); - /* set ssk->sk_socket of accept()ed flows to mptcp socket. * This is needed so NOSPACE flag can be set from tcp stack. */ diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c index 5070dc33675d..a631a5e6fc7b 100644 --- a/net/mptcp/subflow.c +++ b/net/mptcp/subflow.c @@ -397,6 +397,12 @@ void mptcp_subflow_reset(struct sock *ssk) struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); struct sock *sk = subflow->conn;

+ /* mptcp_mp_fail_no_response() can reach here on an already closed + * socket + */ + if (ssk->sk_state == TCP_CLOSE) + return; + /* must hold: tcp_done() could drop last reference on parent */ sock_hold(sk);

@@ -750,6 +756,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk, struct mptcp_options_received mp_opt; bool fallback, fallback_is_fatal; struct sock *new_msk = NULL; + struct mptcp_sock *owner; struct sock *child;

pr_debug("listener=%p, req=%p, conn=%p", listener, req, listener->conn); @@ -824,6 +831,8 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk, ctx->setsockopt_seq = listener->setsockopt_seq;

if (ctx->mp_capable) { + owner = mptcp_sk(new_msk); + /* this can't race with mptcp_close(), as the msk is * not yet exposted to user-space */ @@ -832,14 +841,14 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk, /* record the newly created socket as the first msk * subflow, but don't link it yet into conn_list */ - WRITE_ONCE(mptcp_sk(new_msk)->first, child); + WRITE_ONCE(owner->first, child);

/* new mpc subflow takes ownership of the newly * created mptcp socket */ mptcp_sk(new_msk)->setsockopt_seq = ctx->setsockopt_seq; - mptcp_pm_new_connection(mptcp_sk(new_msk), child, 1); - mptcp_token_accept(subflow_req, mptcp_sk(new_msk)); + mptcp_pm_new_connection(owner, child, 1); + mptcp_token_accept(subflow_req, owner); ctx->conn = new_msk; new_msk = NULL;

@@ -847,15 +856,21 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk, * uses the correct data */ mptcp_copy_inaddrs(ctx->conn, child); + mptcp_propagate_sndbuf(ctx->conn, child); + + mptcp_rcv_space_init(owner, child); + list_add(&ctx->node, &owner->conn_list); + sock_hold(child);

/* with OoO packets we can reach here without ingress * mpc option */ - if (mp_opt.suboptions & OPTION_MPTCP_MPC_ACK) + if (mp_opt.suboptions & OPTION_MPTCP_MPC_ACK) { mptcp_subflow_fully_established(ctx, &mp_opt); + mptcp_pm_fully_established(owner, child, GFP_ATOMIC); + ctx->pm_notified = 1; + } } else if (ctx->mp_join) { - struct mptcp_sock *owner; - owner = subflow_req->msk; if (!owner) { subflow_add_reset_reason(skb, MPTCP_RST_EPROHIBIT);

From: Paolo Abeni pabeni@redhat.com

As reported by Christoph after having refactored the passive socket initialization, the mptcp listener shutdown path is prone to an UaF issue.

BUG: KASAN: use-after-free in _raw_spin_lock_bh+0x73/0xe0 Write of size 4 at addr ffff88810cb23098 by task syz-executor731/1266

CPU: 1 PID: 1266 Comm: syz-executor731 Not tainted 6.2.0-rc59af4eaa31c1f6c00c8f1e448ed99a45c66340dd5 #6 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x6e/0x91 print_report+0x16a/0x46f kasan_report+0xad/0x130 kasan_check_range+0x14a/0x1a0 _raw_spin_lock_bh+0x73/0xe0 subflow_error_report+0x6d/0x110 sk_error_report+0x3b/0x190 tcp_disconnect+0x138c/0x1aa0 inet_child_forget+0x6f/0x2e0 inet_csk_listen_stop+0x209/0x1060 __mptcp_close_ssk+0x52d/0x610 mptcp_destroy_common+0x165/0x640 mptcp_destroy+0x13/0x80 __mptcp_destroy_sock+0xe7/0x270 __mptcp_close+0x70e/0x9b0 mptcp_close+0x2b/0x150 inet_release+0xe9/0x1f0 __sock_release+0xd2/0x280 sock_close+0x15/0x20 __fput+0x252/0xa20 task_work_run+0x169/0x250 exit_to_user_mode_prepare+0x113/0x120 syscall_exit_to_user_mode+0x1d/0x40 do_syscall_64+0x48/0x90 entry_SYSCALL_64_after_hwframe+0x72/0xdc

The msk grace period can legitly expire in between the last reference count dropped in mptcp_subflow_queue_clean() and the later eventual access in inet_csk_listen_stop()

After the previous patch we don't need anymore special-casing msk listener socket cleanup: the mptcp worker will process each of the unaccepted msk sockets.

Just drop the now unnecessary code.

Please note this commit depends on the two parent ones:

mptcp: refactor passive socket initialization mptcp: use the workqueue to destroy unaccepted sockets

Fixes: 6aeed9045071 ("mptcp: fix race on unaccepted mptcp sockets") Cc: stable@vger.kernel.org # v6.0+ Reported-and-tested-by: Christoph Paasch cpaasch@apple.com Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/346 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 --- net/mptcp/protocol.c | 1 - net/mptcp/protocol.h | 1 - net/mptcp/subflow.c | 72 ---------------------------------------------------- 3 files changed, 74 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index b7014f939236..420d6616da7d 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -2355,7 +2355,6 @@ static void __mptcp_close_ssk(struct sock *sk, struct sock *ssk, /* 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); } diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h index 901c9da8fe66..bda5ad723d38 100644 --- a/net/mptcp/protocol.h +++ b/net/mptcp/protocol.h @@ -629,7 +629,6 @@ void mptcp_close_ssk(struct sock *sk, struct sock *ssk, struct mptcp_subflow_context *subflow); void __mptcp_subflow_send_ack(struct sock *ssk); void mptcp_subflow_reset(struct sock *ssk); -void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk); void mptcp_sock_graft(struct sock *sk, struct socket *parent); struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); bool __mptcp_close(struct sock *sk, long timeout); diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c index 9d5bf2a020ef..5a3b17811b6b 100644 --- a/net/mptcp/subflow.c +++ b/net/mptcp/subflow.c @@ -1826,78 +1826,6 @@ static void subflow_state_change(struct sock *sk) } }

-void mptcp_subflow_queue_clean(struct sock *listener_sk, struct sock *listener_ssk) -{ - struct request_sock_queue *queue = &inet_csk(listener_ssk)->icsk_accept_queue; - struct mptcp_sock *msk, *next, *head = NULL; - struct request_sock *req; - - /* build a list of all unaccepted mptcp sockets */ - spin_lock_bh(&queue->rskq_lock); - for (req = queue->rskq_accept_head; req; req = req->dl_next) { - struct mptcp_subflow_context *subflow; - struct sock *ssk = req->sk; - struct mptcp_sock *msk; - - if (!sk_is_mptcp(ssk)) - continue; - - subflow = mptcp_subflow_ctx(ssk); - if (!subflow || !subflow->conn) - continue; - - /* skip if already in list */ - msk = mptcp_sk(subflow->conn); - if (msk->dl_next || msk == head) - continue; - - msk->dl_next = head; - head = msk; - } - spin_unlock_bh(&queue->rskq_lock); - if (!head) - return; - - /* can't acquire the msk socket lock under the subflow one, - * or will cause ABBA deadlock - */ - release_sock(listener_ssk); - - for (msk = head; msk; msk = next) { - struct sock *sk = (struct sock *)msk; - bool do_cancel_work; - - lock_sock_nested(sk, SINGLE_DEPTH_NESTING); - next = msk->dl_next; - msk->first = NULL; - msk->dl_next = NULL; - - do_cancel_work = __mptcp_close(sk, 0); - release_sock(sk); - if (do_cancel_work) { - /* lockdep will report a false positive ABBA deadlock - * between cancel_work_sync and the listener socket. - * The involved locks belong to different sockets WRT - * the existing AB chain. - * Using a per socket key is problematic as key - * deregistration requires process context and must be - * performed at socket disposal time, in atomic - * context. - * Just tell lockdep to consider the listener socket - * released here. - */ - mutex_release(&listener_sk->sk_lock.dep_map, _RET_IP_); - mptcp_cancel_work(sk); - mutex_acquire(&listener_sk->sk_lock.dep_map, - SINGLE_DEPTH_NESTING, 0, _RET_IP_); - } - sock_put(sk); - } - - /* we are still under the listener msk socket lock */ - lock_sock_nested(listener_ssk, SINGLE_DEPTH_NESTING); -} - static int subflow_ulp_init(struct sock *sk) { struct inet_connection_sock *icsk = inet_csk(sk);

From: Geliang Tang geliang.tang@suse.com

Add __ro_after_init labels for the variables tcp_prot_override and tcpv6_prot_override, just like other variables adjacent to them, to indicate that they are initialised from the init hooks and no writes occur afterwards.

Fixes: b19bc2945b40 ("mptcp: implement delegated actions") Cc: stable@vger.kernel.org Fixes: 51fa7f8ebf0e ("mptcp: mark ops structures as ro_after_init") Signed-off-by: Geliang Tang geliang.tang@suse.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/subflow.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/mptcp/subflow.c b/net/mptcp/subflow.c index 5a3b17811b6b..f6b4511b09b0 100644 --- a/net/mptcp/subflow.c +++ b/net/mptcp/subflow.c @@ -628,7 +628,7 @@ static struct request_sock_ops mptcp_subflow_v6_request_sock_ops __ro_after_init static struct tcp_request_sock_ops subflow_request_sock_ipv6_ops __ro_after_init; static struct inet_connection_sock_af_ops subflow_v6_specific __ro_after_init; static struct inet_connection_sock_af_ops subflow_v6m_specific __ro_after_init; -static struct proto tcpv6_prot_override; +static struct proto tcpv6_prot_override __ro_after_init;

static int subflow_v6_conn_request(struct sock *sk, struct sk_buff *skb) { @@ -926,7 +926,7 @@ static struct sock *subflow_syn_recv_sock(const struct sock *sk, }

static struct inet_connection_sock_af_ops subflow_specific __ro_after_init; -static struct proto tcp_prot_override; +static struct proto tcp_prot_override __ro_after_init;

enum mapping_status { MAPPING_OK,