XX-net/XX-Net · warning

add_sock_event %s conn:%d e:%r

Error message

add_sock_event %s conn:%d e:%r

What it means

Registering a connection's socket with the selector (select2.register_event) raised an exception; the connection is then closed as a safety measure. Typically a closed/invalid fd ('EBADF' / 'bad file descriptor') or closed selector.

Source

Thrown at code/default/x_tunnel/local/base_container.py:431

        self._debug_log("ConnectionPipe stop")

    def _debug_log(self, fmt, *args, **kwargs):
        if not self.session or not self.session.config.show_debug:
            return
        self.xlog.debug(fmt, *args, **kwargs)

    def add_sock_event(self, sock, conn, event):
        # this function can repeat without through an error.

        if not sock:
            return

        with self._lock:
            self._debug_log("add_sock_event conn:%d event:%s", conn.conn_id, event)
            try:
                self.select2.register_event(sock, event, conn)
            except Exception as e:
                self.xlog.warn("add_sock_event %s conn:%d e:%r", sock, conn.conn_id, e)
                self.close_sock(sock, str(e) + "_when_add_sock_event")
                return

            # if sys.platform == "win32" and (sock not in self.sock_conn_map or event == selectors.EVENT_WRITE):
            #     self.notice_select()

            self.sock_conn_map[sock] = conn

            if not self.th:
                self.th = threading.Thread(target=self.pipe_worker, name="x_tunnel_pipe_worker")
                self.th.start()
                self._debug_log("ConnectionPipe start")

    def remove_sock_event(self, sock, event):
        # this function can repeat without through an error.

        with self._lock:
            if sock not in self.sock_conn_map:

View on GitHub (pinned to cfa5bc17b6)

Solutions

1. Check if the pipe/worker is shutting down before issuing commands
2. Ensure sockets are unregistered before closing
3. Upgrade/patch race conditions around close_sock and register_event

if not pipe.running:
    return  # skip issuing commands while shutting down

try:
    pipe.add_sock_event(sock, event, conn)
except Exception:
    # already handled internally; conn is closed

When it happens

Trigger: add_sock_event called from start, put_cmd_data, process_cmd or send_to_sock after the socket was already closed, or during ConnectionPipe shutdown.

Common situations: Race between socket close and event registration; shutting down the tunnel while data still flows.

Related errors

• ConnectionPipe remove sock e:%r
• unregister %s e:%r
• XTunnelNotRunning
• proxy handler read error %r
• login session server is down, try get new server.

AI-assisted analysis of XX-net/XX-Net@cfa5bc17b6 (2026-08-27). Data as JSON: /api/errors/d02fdf50aac027ec. Report an issue: GitHub.