长时间运行无法联网之sal_closesocket函数的bug

发布于 2019-03-23 17:27:31 浏览：2078 订阅该版

即前段时间使用at socket编程实现了产品功能，现在产品在现场运行经过24小时或更长时间后，设备无法再连接服务器，打印信息显示`socket(AF_AT, SOCK_STREAM, 0)`执行失败。
经过查看相关代码发现，sal_closesocket在执行，当4G通信模块 AT指令 `“AT+CIPCLOSE”`失败时，会导致sal socket不能释放。
当发生多次这种异常后，导致程序中sock资源用尽，无法socket_alloc分配到新的socket，就出现了设备运行了很长时间后，无法分配出新`socket`，`socket(AF_AT, SOCK_STREAM, 0)`执行失败，无法连接网络的情况。解决办法如下：

```c
      sal_closesocket函数的修改方法（标蓝色的是修改部分）：
 int sal_closesocket(int socket)
{
    struct sal_socket *sock;
    int     error = 0;
    sock = sal_get_socket(socket);
    if (!sock)
    {
        return -1;
    }
    if (sock->ops->closesocket == RT_NULL)
    {
        return -RT_ENOSYS;
    }
    if (sock->ops->closesocket((int) sock->user_data) == 0)
    {
#ifdef SAL_USING_TLS
        if (SAL_SOCKOPS_PROTO_TLS_VALID(sock, closesocket))
        {
            if (proto_tls->ops->closesocket(sock->user_data_tls) < 0)
            {
                return -1;
            }
        }
#endif
        error = 0;
    }
    else
    {
        error = -1;
    }
    /*zhaoshimin 2019.03.23 sometimes at socket close faile should free the sock */
    rt_free(sock);        
    socket_table.sockets[socket] = RT_NULL;
    return error;
}
```

另外socket_alloc函数中当socket分配出最大数量st->max_socket后，程序设计的用意为再增加4个socket资源，实际程序书写有误，并没有实现这个想法，应做如下修改
```c
static int socket_alloc(struct sal_socket_table *st, int f_socket)
{
    int idx;
    /* find an empty socket entry */
    for (idx = f_socket; idx < (int) st->max_socket; idx++)
    {
        if (st->sockets[idx] == RT_NULL)
            break;
        if (st->sockets[idx]->ops == RT_NULL)
            break;
    }
    /* allocate a larger sockte container */
    if (idx == (int) st->max_socket &&  st->max_socket < SAL_SOCKETS_NUM)
    {
        int cnt, index;
        struct sal_socket **sockets;
        /* increase the number of socket with 4 step length */
        cnt = st->max_socket + SOCKET_TABLE_STEP_LEN;
        /*zhaoshimin 2019.03.23 add 4 sockets */
        cnt = cnt > SAL_SOCKETS_NUM ? cnt:SAL_SOCKETS_NUM;
        sockets = rt_realloc(st->sockets, cnt * sizeof(struct sal_socket *));
        if (sockets == RT_NULL)
            goto __result; /* return st->max_socket */
        /* clean the new allocated fds */
        for (index = st->max_socket; index < cnt; index++)
        {
            sockets[index] = RT_NULL;
        }
        st->sockets = sockets;
        st->max_socket = cnt;
    }
    /* allocate  'struct sal_socket' */
    if (idx < (int) st->max_socket && st->sockets[idx] == RT_NULL)
    {
        st->sockets[idx] = rt_calloc(1, sizeof(struct sal_socket));
        if (st->sockets[idx] == RT_NULL)
        {
            idx = st->max_socket;
        }
    }
__result:
    return idx;
}
```

请rt thread的专家帮助确认是否存在这些问题。