ZeroMQ socket断开与ZMQ_EVENT_DISCONNECTED事件技术问询

根据ZeroMQ官方API文档说明，ZMQ_EVENT_DISCONNECTED事件的定义为：套接字发生意外断开，事件值为底层网络套接字的文件描述符（FD），需注意该套接字随后将被关闭。

核心疑问 #

• 上述定义中的*“意外”*具体判定标准是什么？
• 如何在REQ-REP类型的套接字对上触发“干净”的断开操作？
• 经实测验证：调用zmq_close(client)会触发ZMQ_EVENT_DISCONNECTED事件，但该操作本应属于干净关闭操作，与预期不符；若在调用zmq_close前先执行zmq_disconnect(client, addr)，监控到的事件序列没有变化，即使显式断开操作执行成功，依然会上报“意外断开”事件，原因是什么？
• 是否存在可行方法，可区分套接字的正常干净关闭操作，与客户端被执行kill -9强制终止的异常场景？（理论上后者仍会发送FIN报文，因此从原理上具备检测区分的可能性）

问题复现示例代码 #

以下为用于复现上述问题的示例代码，可通过命令gcc -lzmq zmq_test.c && ./a.out编译运行：

#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <assert.h>
#include <zmq.h>

// Read one event off the monitor socket; return value and address
// by reference, if not null, and event number by value. Returns -1
// in case of error.
static int get_monitor_event (void *monitor, int *value, char **address)
{
    // First frame in message contains event number and value
    zmq_msg_t msg;
    zmq_msg_init (&msg);
    if (zmq_msg_recv (&msg, monitor, 0) == -1)
        return -1; // Interruped, presumably
    assert (zmq_msg_more (&msg));

    uint8_t *data = (uint8_t *) zmq_msg_data (&msg);
    uint16_t event = *(uint16_t *) (data);
    if (value)
        *value = *(uint32_t *) (data + 2);

    // Second frame in message contains event address
    zmq_msg_init (&msg);
    if (zmq_msg_recv (&msg, monitor, 0) == -1)
        return -1; // Interruped, presumably
    assert (!zmq_msg_more (&msg));

    if (address) {
        uint8_t *data = (uint8_t *) zmq_msg_data (&msg);
        size_t size = zmq_msg_size (&msg);
        *address = (char *) malloc (size + 1);
        memcpy (*address, data, size);
        *address [size] = 0;
    }
    printf("event: %d\n", event);
    return event;
}

static void close_zero_linger(void *socket)
{
    int linger = 0;
    zmq_setsockopt(socket, ZMQ_LINGER, &linger, sizeof(linger));
    zmq_close(socket);
}

static void bounce(void *client, void *server)
{
    char send_buf[] = "Hello";
    char recv_buf[sizeof(send_buf) / sizeof(send_buf[0])];

    zmq_send(client, send_buf, sizeof(send_buf), 0);
    zmq_recv(server, recv_buf, sizeof(recv_buf), 0);
    zmq_send(server, send_buf, sizeof(send_buf), 0);
    zmq_recv(client, recv_buf, sizeof(recv_buf), 0);
}

int main (void)
{
    void *ctx = zmq_ctx_new ();
    void *client = zmq_socket (ctx, ZMQ_REQ);
    void *server = zmq_socket (ctx, ZMQ_REP);
    int rc = zmq_socket_monitor (client, "tcp://127.0.0.1:9999", 0);

    // Monitor all events on client and server sockets
    rc = zmq_socket_monitor (server, "inproc://monitor-server", ZMQ_EVENT_ALL)
    void *server_mon = zmq_socket (ctx, ZMQ_PAIR);
    rc = zmq_connect (server_mon, "inproc://monitor-server");

    // Now do a basic ping test
    rc = zmq_bind (server, "tcp://127.0.0.1:9998");
    rc = zmq_connect (client, "tcp://127.0.0.1:9998");
    bounce (client, server);

//  Close client
    rc = zmq_disconnect(client, "tcp://127.0.0.1:9998");
    printf("disconnect: %d\n", rc);
//    close_zero_linger (client);
    rc = zmq_close(client);
    printf("close: %d\n", rc);
//   sleep(1); // no difference

    // This is the flow of server events
    event = get_monitor_event (server_mon, NULL, NULL);
    assert (event == ZMQ_EVENT_LISTENING);
    event = get_monitor_event (server_mon, NULL, NULL);
    assert (event == ZMQ_EVENT_ACCEPTED);
    event = get_monitor_event (server_mon, NULL, NULL);
    assert (event == ZMQ_EVENT_HANDSHAKE_SUCCEEDED);
//   sleep(1); // no difference
    event = get_monitor_event (server_mon, NULL, NULL);
//    assert (event == ZMQ_EVENT_CLOSED);
//    event = get_monitor_event (server_mon, NULL, NULL);
//    assert (event == ZMQ_EVENT_MONITOR_STOPPED);
    assert (event == ZMQ_EVENT_DISCONNECTED); // why??

    // Close down the sockets
    close_zero_linger (server);
    close_zero_linger (server_mon);
    zmq_ctx_term (ctx);

    return 0;
}

内容的提问来源于stack exchange，提问作者Michael Schubert