Re: [lwip-users] lwip 1.2.0 + freeRTOS

[Sirjee Rooplall]

Hi Simon,

I am new to LWIP, so I copied that code from an example I have seen.

I removed the sleep and I have removed the access to the internal pcb, and 
only rely on the API now.

The problem I am explaining, seems to also exist in LWIP 1.3.0, as I have 
seen in some of the recent lwip archives.

Another thing:
Because I am connecting through a PPP->GPRS interface the possibility exists 
where someone removes the antenna and the connection longer exists, however 
the user can trigger a server connection externally through a push button, 
which will attempt to perform a neconn_connect which blocks and returns with 
ERR_ABRT(-3), how do I handle this senario without crashing the stack. At 
the moment, I keep trying every 30 seconds, and eventually I cause the stack 
to crash with too many netconn_connect without a physical connection 
available.

What should my application code do in this case.

struct netconn* ConnectProcess( U16 connPortNo)
{
U8 ip_addrTemp[4];
 struct ip_addr ipaddrToConnectTo;
err_t error;

ipStr2ipAddr((U8 *)&ip_addrTemp[0], (U8 *)&DestIP[0]);
IP4_ADDR(&ipaddrToConnectTo, ip_addrTemp[0], ip_addrTemp[1], 
ip_addrTemp[2], ip_addrTemp[3]);

  if (pxSeverConnectionHandle==NULL)
  {
    //we make a connection
#ifdef XCOM_DEBUG
     SendDebugString("ConnectProcess->Calling netconn_new");
#endif
   pxSeverConnectionHandle = netconn_new(NETCONN_TCP);
  }

#ifdef XCOM_DEBUG
  SendDebugString("ConnectProcess->Calling netconn_connect");
#endif
   error=netconn_connect(pxSeverConnectionHandle, &ipaddrToConnectTo, 
connPortNo);
  switch (error)
  {
   case ERR_VAL:
#ifdef XCOM_DEBUG
          SendDebugString("ConnectProcess->ERR_VAL=%d", error);
#endif
     return(NULL);
    break;
   case ERR_MEM:
#ifdef XCOM_DEBUG
          SendDebugString("ConnectProcess->ERR_MEM=%d", error);
#endif
     return(NULL);
    break;
   case ERR_OK:
#ifdef XCOM_DEBUG
          SendDebugString("ConnectProcess->ERR_OK=%d, CONNECTED", error);
#endif
    break;
    case ERR_ABRT:
#ifdef XCOM_DEBUG
          SendDebugString("ConnectProcess->ERR_ABRT=%d, ABORTED", error);
#endif
          return(NULL);
     break;
   default :
#ifdef XCOM_DEBUG
          SendDebugString("ConnectProcess->Unhandled Error=%d", error);
#endif
    return(NULL);
    break;
    }
   return pxSeverConnectionHandle;
}


void CloseConnection(void)
{
 err_count = 4;
 while( netconn_close( pxSeverConnectionHandle ) != ERR_OK )
 {
    if (--err_count == 0) break;
    vTaskDelay( 10 );
 }
 err_count = 4;

 err_count = 4;
 while( netconn_delete( pxSeverConnectionHandle ) != ERR_OK )
 {
    if (--err_count == 0) break;
    vTaskDelay( 10 );
 }

 vPortEnterCritical();
 pxSeverConnectionHandle = NULL;
 vPortExitCritical();
}

Kind Regards, Sirjee Rooplall Figment Design laboratories (Pty) Ltd mailto: 
address@hidden Mobile: +27 (0)83 230 8466
----- Original Message ----- 
From: "Simon Goldschmidt" <address@hidden>
To: "Mailing list for lwIP users" <address@hidden>
Sent: Monday, January 25, 2010 2:09 PM
Subject: Re: [lwip-users] lwip 1.2.0 + freeRTOS


> You're using an old version, added an extra sleep and violate api 
> layering. Any reasons for that? ;-)
>
> Sirjee Rooplall wrote:
> > Hi, I am using lwip 1.2.0 under freeRTOS, I am also using PPP + external
> > modem to establish a PPP connection to a server over GPRS.
>
> Using such an old version is not recommended and makes it hard to track 
> down bugs as they might already have been fixed by now. 1.2.0 is over 
> three years old by now.
>
> > All works well, provided I can establish a connection to the server and
> > the connection is closed.
> >
> > If for some reason the connection is not closed, i.e the TCP/IP stack 
> > does
> > not get to the FIN_2 closing state, all hell breaks loose and my system
> > crashes when I attempt to re-connect to the server.
> >
> > I s there an elegant way to remove the connection witout causing future
> > failure.
> >
> > To connect I use:
> >
> > conn = netconn_new(NETCONN_TCP);
> > netconn_connect(conn, IP, PORT);
> > vTaskDelay(100); // wait 100ms
>
> What's that for? (I mean, I could have guessed the comment if it wasn't 
> there, but why sleep?)
>
> > Then I check the state of the pcb:
> > if (conn->pcb.tcp->state == ESTABLISHED)
>
> This is a total no-go! Don't ever (I really mean never!) access the pcb 
> from your netconn application code. The netconn API is exists for you not 
> having to worry about the raw API. Just because you have the struct 
> definition doesn't mean you know what you are doing (see below).
>
> > {
> >   conn->state = NETCONN_CONNECT;
>
> And where did you get that from? Please, don't "mess around" with internal 
> netconn members. In this case, you tell the netconn it is in state 
> 'connecting' after netconn_connect has returned (which means successfully 
> or not) which no doubt leads to problems!
>
> Please have a look at other (example-) netconn apps (you can find some by 
> downloading the contrib module from cvs) and make sure you use the api 
> correctly!
>
> Simon
> --
> Jetzt kostenlos herunterladen: Internet Explorer 8 und Mozilla Firefox 
> 3.5 -
> sicherer, schneller und einfacher! http://portal.gmx.net/de/go/atbrowser
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users