W5500 IP 续约会重新 DHCP_init() 的问题

发布于 2022-01-13 19:35:07 浏览：1804 订阅该版

发现设备偶尔会有丢帧的问题，用 Wireshark 抓包发现了以下的问题

![image.png](https://oss-club.rt-thread.org/uploads/20220113/281ed1622a0ac990029eeb4aa118fcd4.png.webp)

正好我的路由器的租约时间也是1小时，感觉每次续约的过程中会重新 DHCP
看了下源码，感觉这里是有BUG

排查过程：
工作流初始化：将 wiz_dhcp_work 加入到工作流中，time时间一到，就会自动调用
![image.png](https://oss-club.rt-thread.org/uploads/20220113/a4009350a26eb367c87f2b19081d6f1a.png)

同时此函数包含 DHCP_run()
![image.png](https://oss-club.rt-thread.org/uploads/20220113/27126e04cbf8d9e0f6b022d60e7e93f8.png)

DHCP_run()中的 dhcp_state 分支源头：`INIT -> DISCOVER -> REQUEST -> LEASED -> REREQUEST -> CHANGED`

但是 DHCP_run() 中 一旦进入到 LEASED 分支，就会返回 DHCP_IP_LEASED, 从而在 wiz_dhcp_work 的 LEASED分支中关闭 DHCP定时器，退出当前任务	等待1/2租约时间，又重启该任务，然后又重新 DHCP_init() , 最终导致掐断了正常的 ip 续约逻辑

![image.png](https://oss-club.rt-thread.org/uploads/20220113/386bf167bbceaf195834a4f544f558a2.png)

不知道为什么进入到 LEASED 中需要关闭 DHCP的定时器 和 DHCP_stop()?
但是可以看出，应该是希望等到 1/2租约时间到后，重新进入 DHCP_run() 中 跳到 REREQUEST 过程中 续约ip ，当然这里还需要对 DHCP_init() 进行条件约束 避免重新DHCP。

**以下是修改后的代码**
```c
static void wiz_dhcp_work(struct rt_work *dhcp_work, void *dhcp_work_data)
{
#define WIZ_DHCP_WORK_RETRY         3 /* DHCP will have 3 times handshake */
#define WIZ_DHCP_WORK_RETRY_TIME    (2 * RT_TICK_PER_SECOND)
    static int wiz_dhcp_retry_times = WIZ_DHCP_WORK_RETRY * 20;

RT_ASSERT(dhcp_work_data != RT_NULL);

struct netdev *netdev = (struct netdev *)dhcp_work_data;

uint8_t dhcp_times = 0;
    static uint8_t data_buffer[1024];
    static uint32_t dhcp_status = 0;

if(dhcp_status == 0)
	{
		rt_timer_start(dhcp_timer);
		DHCP_init(WIZ_DHCP_SOCKET, data_buffer);
	}

while (1)
    {
        /* DHCP start, return DHCP_IP_LEASED is success. */
        dhcp_status = DHCP_run();

switch (dhcp_status)
        {
        case DHCP_IP_ASSIGN:
        case DHCP_IP_CHANGED:
        {
            /* to update netdev information */
            wiz_netdev_info_update(netdev, RT_FALSE);
            break;
        }
        case DHCP_IP_LEASED:
        {
            int hour, min;
//            DHCP_stop();
//            rt_timer_stop(dhcp_timer);
            /* to update netdev information */
            wiz_netdev_info_update(netdev, RT_FALSE);

/* reset the previous work configure */
            rt_work_cancel(dhcp_work);

/* according to the DHCP leaset time, config next DHCP produce */
            rt_work_submit(dhcp_work, (getDHCPLeasetime() / 2) * RT_TICK_PER_SECOND);
            hour = getDHCPLeasetime() / 3600;
            min = (getDHCPLeasetime() % 3600) / 60;
            LOG_D("DHCP countdown to lease renewal [%dH: %dMin], retry time[%04d]", hour, min, dhcp_times);
            //wiz_dhcp_retry_times = dhcp_times;
			wiz_dhcp_retry_times = 3;
            return;
        }
        case DHCP_STOPPED:
			DHCP_stop();
            rt_timer_stop(dhcp_timer);
			return;
        case DHCP_FAILED:
        {
            dhcp_times = wiz_dhcp_retry_times;
            LOG_E("dhcp handshake failed!");
            break;
        }
        default:
        {
            dhcp_times++;

/* DHCP_RUNNING status, include don't receive data */
            rt_thread_mdelay(10);
            break;
        }
        }

if (dhcp_times >= wiz_dhcp_retry_times)
        {
            LOG_D("DHCP work in %d seconds, [%03d|%03d]", WIZ_DHCP_WORK_RETRY_TIME / RT_TICK_PER_SECOND, dhcp_times, wiz_dhcp_retry_times);

/* if dhcp service is too busy to manger IP, increase retry times */
            wiz_dhcp_retry_times = wiz_dhcp_retry_times + WIZ_DHCP_WORK_RETRY;
			
		dhcp_status = 0;
            DHCP_stop();
            rt_timer_stop(dhcp_timer);

rt_work_cancel(dhcp_work);

/* according to WIZ_DHCP_WORK_RETRY_TIME, reconfigure in 2 seconds */
            rt_work_submit(dhcp_work, WIZ_DHCP_WORK_RETRY_TIME);
            break;
        }
    }
}
```

**测试结果：**
可以看出确实不会 再重新DHCP 来续约ip，但是没有抓到请求的过程，但是打断点又确实有进入`send_DHCP_REQUEST();`
不知道是不是请求的过程没经过电脑的网卡所以抓不到？

![image.png](https://oss-club.rt-thread.org/uploads/20220113/deaa0b2bf831100c601c1e74b4f52fab.png)