[En-Nut-Discussion] changes in tcpip stack from 3.9.5 to 3.9.6

[Vesa Jääskeläinen]

I think this change is included in 3.9.7, so you could download that and 
test it.

William Baker wrote:
> 
> I see you're really on top of it.  Thank you for the help.  Hope to give
> 3.9.6 a try again with the next rev, or if you come out with a patch. 
> For now, 3.9.5 is working well.  I would rather be working and
> reporting bugs on the latest release of winavr and Nut/OS, but patience
> is a virtue.
> 
> bbaker
> 
>> I'd not insist on blaming GCC. No idea what drove me in
>> this direction. This change causes the different behavior:
>>
>> * net/udpin.c: ICMP unreachable will be sent on incoming
>> udp packets with no local peer port.
>>
>> Looks like Ethernut wants to reject UDP packets even
>> if the datagram is broadcasted.
>>
>> All this exposes a general problem with the stack.
>> Like ICMP echo replies, this UDP reject is sent by
>> the Ethernet receiver thread. As long as this thread
>> gets blocked in ARP, no new incoming Ethernet packet is
>> processed.
>>
>> This is typically no problem, because Ethernut will
>> receive an ARP request before another hosts contacts
>> it. As Ethernut uses incoming ARP request to update
>> the ARP table, there is no need for another ARP request
>> on the response. Thus, limiting the ICMP unreachable
>> response to non-broadcasts will fix this for now.
>>
>> I'm thinking about setting MAX_ARPREQUESTS to 1
>> by default, which is the correct way, as Michael Jones
>> already pointed out earlier.
>>
>> Harald
>>
>> _______________________________________________
>> En-Nut-Discussion mailing list
>> En-Nut-Discussion at egnite.de
>> http://www.egnite.de/mailman/listinfo.cgi/en-nut-discussion
> 
> 
> _______________________________________________
> En-Nut-Discussion mailing list
> En-Nut-Discussion at egnite.de
> http://www.egnite.de/mailman/listinfo.cgi/en-nut-discussion
>