[lwip-users] Re: [lwip] changes in IP/UDP for DHCP

[Adam Dunkels]

Hi Leon!

On Thursday 10 January 2002 13.56, you wrote:
> I used:
>
> udp_bind(state->pcb, IP_ADDR_ANY, DHCP_CLIENT_PORT);
> udp_connect(state->pcb, IP_ADDR_ANY, DHCP_SERVER_PORT);
>
> the problem is, IP_ADDR_ANY is defined 0, interpreted NULL by
> udp_connect, and so, not set. It remains the old value.

You are right - udp_connect() (and udp_bind()) are flawed. This is among the 
oldest code is lwIP, and it hasn't been given enough attention until now. 
Both functions should be changed so that they use ip_addr_set() instead of 
the current implementation. This is the way the code should look like (from 
the current CVS):

err_t
udp_bind(struct udp_pcb *pcb, struct ip_addr *ipaddr, u16_t port)
{
  struct udp_pcb *ipcb;
  ip_addr_set(&pcb->local_ip, ipaddr);
  pcb->local_port = port;
[...]

err_t
udp_connect(struct udp_pcb *pcb, struct ip_addr *ipaddr, u16_t port)
{
  struct udp_pcb *ipcb;
  ip_addr_set(&pcb->dest_ip, ipaddr);
  pcb->dest_port = port;
[...]

> Also, as a small sidenote, I found it confusing having
> pcb->local_* and pcb->dest_*.
>
> More appropriate are pcb->local_* and pcb->remote_* which
> are location based identifiers. dest and src are always
> respective to a communication direction.

Yes, you are most definately right. I have been meaning to change this for a 
long time, but I hadn't come around to it until now. Thanks! The CVS code now 
is updated with s/dest_/remote_/.

/adam
-- 
Adam Dunkels <address@hidden>
http://www.sics.se/~adam
[This message was sent through the lwip discussion list.]