Re: [lwip-users] Non-blocking netconn callback and TCP accept backlog

[Sylvain Rochet]

Hi Nick,


On Sat, Jul 14, 2012 at 05:24:47PM +0100, Nick Brereton wrote:
> 
> When I get get a NETCONN_EVT_RECVPLUS event with 0 length for the 
> listening connection I add this to a queue of accept event, which I 
> then drain whenever I have memory available to process the connection. 
> As I have understood it data should not start flowing until I have 
> accepted the connection (presumably completion of the handshake should 
> be delayed until the netconn_accept call, thus preventing the client 
> from sending any data?) and the connection should be sitting in the 
> tcp listen queue.

I am not a lwIP API expert too but I played recently with the 
non-blocking netconn API. It seems that connections are automatically 
accept()ed when using the non blocking API.

This is what I did with this strange behavior:


static void telnet_tcp_callback(struct netconn *nc, enum netconn_evt 
  event, u16_t len) {
  if(nc == telnet_tcp_listen) {
    if(event == NETCONN_EVT_RCVPLUS) {

      /* We must fetch the new netconn pointer in the lwIP callback,
       * else we are going to receive events about an unknown session
       * ... because lwIP don't wait the session to be accept()'ed
       * before sending events about it.
       */
       err = netconn_accept(telnet_tcp_listen, &telnet->tcp);
       if(err != ERR_OK) {
          return;
       }
  ....


Of course, this can be a lwIP bug.

Sylvain

signature.asc
Description: Digital signature