v4.10版本, 在telnet中使用netstat指令卡死

发布于 2022-03-27 19:21:11 浏览：1535 订阅该版

开启telnet后, 在msh中使用netstat指令, 程序进入死循环直至崩溃.
问题定位:
1. 使用netstat指令会调用ethernetif.c中list_tcps()函数, 调用该函数时使用了rt_enter_critical(); 随后程序调用了rt_kprintf();![微信截图_20220327190503.png](https://oss-club.rt-thread.org/uploads/20220327/e68b22e0f8babc90df1abfb87ddf6cd4.png)
2. 因为切换了telnet作为console, rt_kprintf()会引起telnet_write()调用, 该函数代码中使用了rt_mutex_take(telnet->tx_ringbuffer_lock, RT_WAITING_FOREVER);![微信截图_20220327191536.png](https://oss-club.rt-thread.org/uploads/20220327/65589236335ceaa929cf6ab16dfb8f9d.png)
3. rt_mutex_take()函数会检查scheduler是否被锁, 实际上因为之前调用rt_enter_critical()已经将scheduler上锁, 会打印调试信息调用rt_kprintf.![微信截图_20220327191630.png](https://oss-club.rt-thread.org/uploads/20220327/db7e8ce1d1d7b7994c305018582b3e37.png)![微信截图_20220327191709.png](https://oss-club.rt-thread.org/uploads/20220327/0f9afab7b3b41db2516a1ae96746fe0c.png)
4. 至此死循环形成rt_kprintf -> telnet_write -> rt_mutex_take -> rt_kprintf->...死循环形成.

解决方法:
1. 在list_tcps()中进入临界段后, 先将信息打印到buff, 退出临界段后, 将buff中的信息进行进行rt_kprintf. 目前尝试用这种方法解决.
2. 解决telnet_write中rt_mutex_take使用问题.

其他:
在其他使用设备方式的代码中是否也存在类似问题, 需要考虑.

3 个回答

aozima 2022-03-27

调网络不抓包，调I2C等时序不上逻辑分析仪，就像电工不用万用表！多用整理的好的文字，比截图更省流量，还能在整理过程中思考。

出出啊 2022-03-28

恃人不如自恃，人之为己者不如己之自为也

梦回唐朝 2022-03-28

这家伙很懒，什么也没写！

采用了第一种方法修改, 目前测试可以用.
同时需要修改:
1. void list_udps(void);
2. 如果使用list_fd命令也会出现类似问题, 需要修改dfs.c中的这个函数:int list_fd(void).

附: 修改的代码.
```
void list_tcps(void)
{
    rt_uint32_t num = 0;
    struct tcp_pcb *pcb;
    char local_ip_str[16];
    char remote_ip_str[16];

extern struct tcp_pcb *tcp_active_pcbs;
    extern union tcp_listen_pcbs_t tcp_listen_pcbs;
    extern struct tcp_pcb *tcp_tw_pcbs;
    
    char *buff = rt_malloc(1024);
    char *buff_start = buff;

rt_enter_critical();
    buff += rt_sprintf(buff, "Active PCB states:\n");
    for(pcb = tcp_active_pcbs; pcb != NULL; pcb = pcb->next)
    {
        strcpy(local_ip_str, ipaddr_ntoa(&(pcb->local_ip)));
        strcpy(remote_ip_str, ipaddr_ntoa(&(pcb->remote_ip)));

buff += rt_sprintf(buff,
                   "#%d %s:%d <==> %s:%d snd_nxt 0x%08X rcv_nxt 0x%08X ",
                   num++,
                   local_ip_str,
                   pcb->local_port,
                   remote_ip_str,
                   pcb->remote_port,
                   pcb->snd_nxt,
                   pcb->rcv_nxt);
        buff += rt_sprintf(buff, "state: %s\n", tcp_debug_state_str(pcb->state));
    }
    rt_exit_critical();
    rt_kprintf(buff_start);
    
    rt_enter_critical();
    buff = buff_start;
    buff += rt_sprintf(buff, "Listen PCB states:\n");
    num = 0;
    for(pcb = (struct tcp_pcb *)tcp_listen_pcbs.pcbs; pcb != NULL; pcb = pcb->next)
    {
        buff += rt_sprintf(buff,"#%d local port %d ", num++, pcb->local_port);
        buff += rt_sprintf(buff,"state: %s\n", tcp_debug_state_str(pcb->state));
    }
    rt_exit_critical();
    rt_kprintf(buff_start);
    
    rt_enter_critical();
    buff = buff_start;
    buff += rt_sprintf(buff,"TIME-WAIT PCB states:\n");
    num = 0;
    for(pcb = tcp_tw_pcbs; pcb != NULL; pcb = pcb->next)
    {
        strcpy(local_ip_str, ipaddr_ntoa(&(pcb->local_ip)));
        strcpy(remote_ip_str, ipaddr_ntoa(&(pcb->remote_ip)));