When a socket is shutting down, we notify the peer node about the connection termination by reusing an incoming message if possible. If the last received message was a connection acknowledgment message, we reverse this message and set the error code to TIPC_ERR_NO_PORT and send it to peer. In tipc_sk_proto_rcv(), we never check for message errors while processing the connection acknowledgment or probe messages. Thus this message performs the usual flow control accounting and leaves the session hanging. In this commit, we terminate the connection when we receive such error messages.
Unified · Split
| @@ -866,6 +866,14 @@ static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb, · 865 unmodified lines | ||
| 866 | 866 | if (!tsk_peer_msg(tsk, hdr)) |
| 867 | 867 | goto exit; |
| 868 | 868 | |
| 869 | + if (unlikely(msg_errcode(hdr))) { | |
| 870 | + tipc_set_sk_state(sk, TIPC_DISCONNECTING); | |
| 871 | + tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk), | |
| 872 | + tsk_peer_port(tsk)); | |
| 873 | + sk->sk_state_change(sk); | |
| 874 | + goto exit; | |
| 875 | + } | |
| 876 | + | |
| 869 | 877 | tsk->probe_unacked = false; |
| 870 | 878 | |
| 871 | 879 | if (mtyp == CONN_PROBE) { |