[lwip-users] ARP (was Re: Possible bug in src/api/tcpip.c)

[steinarl]

> > Rishi Khan wrote:
> > So, if you call NETIF_FLAG_ETHARP, they you should leave the header  
> > alone, but if >> you don't call NETIF_FLAG_ETHARP, you should leave  
> > it there. Doesn't that seem weird? Why should the TCP/IP stack  
> > handle ARP? This is inherently an ethernet problem.

> Just for better understanding, the handling of ARP packets has  
> changed in 1.3.0:
> Previously they were indeed handled by the 'ethernetif' code. The  
> problem was >that this way, the ARP table wasn't protected against  
> concurrent access (from >receive thread an tcpip_thread - when  
> sending). This was solved by letting the >tcpip_thread handle ARP  
> packets, which is the reason why ethernet netifs have >to pass the  
> complete packet, not only the IP part.

> Simon

Leon Woestenberg had this suggestion in 2006: [RFC] Remove calls to  
etharp_ip_input().
(http://lists.nongnu.org/archive/html/lwip-devel/2006-04/msg00012.html)

Is there a problem with concurrent writes to the ARP table in the  
approach he outlined?

Steinar