Re: [lwip-users] Usage of TCP_EVENT_RECV macro

[Leon Woestenberg]

Hello Roland, Lee,

Roland Winistörfer wrote:

> Hi
>
> After a few little problems I got the lwIP (version 1.0) with uC/OS 
> working on my hardware. But there is a little curiosity, if a try to 
> show the little HTML page over the lwIP.
>
> I often break into the "LWIP_ASSERT("p != NULL", p != NULL);" at the 
> beginning of the pbuf_free(). This is called from the tcp_input() with 
> the following code:
>
>    /* If a FIN segment was received, we call the callback
>        function with a NULL buffer to indicate EOF. */
>        if (recv_flags & TF_GOT_FIN) {
>            TCP_EVENT_RECV(pcb, NULL, ERR_OK, err);
>        }
>
> The call of the pbuf_free() is hidden in the TCP_EVENT_RECV macro. But 
> if I have look into this macro, there is an else case with a call of 
> pbuf_free(NULL) (in fact the "p" argument of the macro TCP_EVENT_RECV 
> is always NULL), which always results in the assert condition above.
>
> What's going wrong here ? Is this macro misplaced here (means the else 
> case) ?
For any packet p, the macro first looks if the recv() callback is 
non-NULL and passes the packet, otherwise drops (simply frees) the given 
packet p.

In this case, TCP_EVENT_RECV() is called with a NULL packet p to 
indicate EOF to the callback function.

**The real problem seems to be that the callback function is not 
installed, i.e. pcb->recv == NULL**

However, the macro should not try to free a NULL packet and shouldn't. I 
will fix this in CVS.

Regards, Leon.