TCP/IP task crashes with MQX_UNHANDLED_INTERRUPT

[bowe]

I am seeing an issue that causes the TCP/IP task to crash with "MQX_UNHANDLED_INTERRUPT".  I have installed the unexpected ISR [_int_install_unexpected_isr()], it is says it is vector 0x03, which is a hard fault.  In looking at the call stack in KDS when the hard fault interrupt occurs, there are numerous (at least 50) _isr_execute() instances (namely on the line that calls the actual interrupt handler, which is after an interrupt enable), with nothing in between them and _int_unexpected_isr() at the top.  Is this the cause of the hard fault, is this a KDS bug, or am I seeing something else?

-Bowe

[bowe]

In further debugging, it appears that adding small delay in the TCPIP_task() main while loop (in tcpip.c) after getting the TCP/IP message avoids the MQX_UNHANDLED_INTERRUPT issue (of course this is not an acceptable fix).  This seems like it indicates a race condition.  Any ideas as to what this would be and how to fix it?

while(1)
{
     TCPIP_EVENT_PTR queue = TCPIP_Event_head;

     tcpip_msg = (TCPIP_MESSAGE_PTR)RTCS_msgq_receive(tcpip_qid, timeout, RTCS_data_ptr->TCPIP_msg_pool);

     if (tcpip_msg)
     {
          _time_delay(1);  // This line seems to eliminate the MQX_UNHANDLED_INTERRUPT crash

          if (NULL != tcpip_msg->COMMAND)
          {
               tcpip_msg->COMMAND(tcpip_msg->DATA);
          }
          RTCS_msg_free(tcpip_msg);
     }

     ...
}

-Bowe

[RadekS]

I suppose that this will be probably caused by bug in TCP close in KSDK 1.1. There is problem with canceling timeout events.

As workaround please edit tcp_close.c file and add lines 9-13 below to TCP_Close_TCB() function:

  /*
  ** Cancel timers...
  */
(void)TCP_Timer_stop(&tcb->sndxmittq.t, tcp_cfg->qhead);
(void)TCP_Timer_stop(&tcb->sndacktq.t, tcp_cfg->qhead);
(void)TCP_Timer_stop(&tcb->sndtq.t, tcp_cfg->qhead);
(void)TCP_Timer_stop(&tcb->rcvtq.t, tcp_cfg->qhead);

if(tcb->expire.PRIVATE)
  {
      TCPIP_Event_cancel(&tcb->expire);
      tcb->expier = NULL;
  }

I hope it helps you.

Have a great day,
RadekS

-----------------------------------------------------------------------------------------------------------------------
Note: If this post answers your question, please click the Correct Answer button. Thank you!
-----------------------------------------------------------------------------------------------------------------------

[bowe · Accepted solution]

Ok, I have inserted lines 9-13 below (I assume line 12 is actually what you meant) in the TCP_Close_TCB() in TCP_clos.c, and it is seeming more stable in the 10 minutes of testing (usually the MQX_UNHANDLED_INTERRUPT occurred pretty quickly).  I will continue to test, and reply back in about a week if everything still looks good.

  /*
   ** Cancel timers...
   */
   (void)TCP_Timer_stop(&tcb->sndxmittq.t, tcp_cfg->qhead);
   (void)TCP_Timer_stop(&tcb->sndacktq.t, tcp_cfg->qhead);
   (void)TCP_Timer_stop(&tcb->sndtq.t, tcp_cfg->qhead);
   (void)TCP_Timer_stop(&tcb->rcvtq.t, tcp_cfg->qhead);

   if(tcb->expire.PRIVATE)
   {
       TCPIP_Event_cancel(&tcb->expire);
       tcb->expire.PRIVATE = NULL;
   }

Thanks,

-Bowe

[bowe]

Just following up.  This fix seems to have worked.  We have not seen a crash in over a week now.  Thanks again.

-Bowe