Re: [lwip-users] Randomly delayed frame (STM32070 package including LwIP v1.4.1)

[address@hidden]

Am 02.01.2019 um 20:14 schrieb stevestrong:
> I finally managed to get DHCP working, I get a valid IP address assigned by
> the router.

Well, the holidays and new year is a bit of a bad time to ask ;-)

> I post here the working version from netconf.c, not sure if it is really ok:
> ********************************************
> void LwIP_Periodic_Handle(__IO uint32_t localtime)
> {
>      /* Cyclic lwIP timers check */
>      sys_check_timeouts(); // v2
>
> #ifdef USE_DHCP
>    /* Fine DHCP periodic process every 500ms */
>    if ( (localtime - DHCPfineTimer) >= DHCP_FINE_TIMER_MSECS)
>    {
>      DHCPfineTimer = localtime;
>       if ((DHCP_state != DHCP_ADDRESS_ASSIGNED) &&
>          (DHCP_state != DHCP_TIMEOUT) &&
>            (DHCP_state != DHCP_LINK_DOWN))
>      {
>        /* toggle LED1 to indicate DHCP on-going process */
>           ToggleLED(); // STM_EVAL_LEDToggle(LED1);
>
>        /* process DHCP state machine */
>        LwIP_DHCP_Process_Handle();
>      }
>    }
> #else
>         (void)(localtime);
> #endif
> }
> ********************************************
> void LwIP_DHCP_Process_Handle()
> {
>    struct ip4_addr ipaddr;
>    struct ip4_addr netmask;
>    struct ip4_addr gw;
>
>    switch (DHCP_state)
>    {
>    case DHCP_START:
>      {
>        DHCP_state = DHCP_WAIT_ADDRESS;
>           PRINTF(". DHCP_WAIT_ADDRESS .\n");
>        dhcp_start(&gnetif);
>        /* IP address should be set to 0 every time we want to assign a new
> DHCP address */
>        IPaddress = 0;
>      }
>      break;
>
>    case DHCP_WAIT_ADDRESS:
>      {
>        /* Read the new IP address */
>        IPaddress = gnetif.ip_addr.addr;
>
>        if (IPaddress!=0)
>        {
>          DHCP_state = DHCP_ADDRESS_ASSIGNED;    
>
>             PRINTF(". DHCP_ADDRESS_ASSIGNED .\n");
>          /* Stop DHCP */
>          dhcp_stop(&gnetif);
>
>                 LEDOn(1); // STM_EVAL_LEDOn(LED1);
>        }
>        else
>        {
>                 struct dhcp * dhcp;
>                 dhcp = netif_get_client_data(&gnetif, 
> LWIP_NETIF_CLIENT_DATA_INDEX_DHCP);
>          /* DHCP timeout */
>          if (dhcp->tries > MAX_DHCP_TRIES)
>          {
>            DHCP_state = DHCP_TIMEOUT;
>
>               PRINTF(". DHCP_TIMEOUT .\n");
>            /* Stop DHCP */
>            dhcp_stop(&gnetif);
>
>            /* Static address used */
>            IP4_ADDR(&ipaddr, IP_ADDR0 ,IP_ADDR1 , IP_ADDR2 , IP_ADDR3 );
>            IP4_ADDR(&netmask, NETMASK_ADDR0, NETMASK_ADDR1, NETMASK_ADDR2,
> NETMASK_ADDR3);
>            IP4_ADDR(&gw, GW_ADDR0, GW_ADDR1, GW_ADDR2, GW_ADDR3);
>            netif_set_addr(&gnetif, &ipaddr , &netmask, &gw);
>
>                   LEDOff(1); // STM_EVAL_LEDOn(LED1);
>          }
>        }
>      }
>      break;
>    default: break;
>    }
> }
> ********************************************
>
> Basicaly, I added sys_check_timeouts() in the periodic handler and removed
> the DHCP fine and course timeout routines from DHCP handler because they are
> anyway executed by the newly added timeout check function.

I have seen threading issues in the way ST does this DHCP thing. 
However, if you're using one big main loop only without interrupts, you 
should be fine...

> However, I am not sure whether the DHCP process is standard conform.
> It is weird that the gratuitous ARP request comes after the IP has been
> assigned by the router. Is that OK?

Yes, the gratuitous ARP is expected.

> I made a wireshark and debug log capture.
> dhcp3.jpg <http://lwip.100.n7.nabble.com/file/t2167/dhcp3.jpg>

Except for not seeing the DHCP server'ss responses, that looks ok.

> dhcp3.log <http://lwip.100.n7.nabble.com/file/t2167/dhcp3.log>

I cannot commen on the board specific logs but lwip output seems ok.

> It would be nice if someone could have a look on it.

Regards,
Simon