AW: [lwip-users] Sending UDP data beyond MTU size

[Becker, Andreas]

Hi Timmy, 

did you tried to reduce teh TCP_MMS too?
For me things are working with a TCP_MMS of 1392 on a Medium, that support only 
480Byte payload.
-> So TCP and UDP Packets are fragmented most times.
I use this with FTP in both directions. 
The second netif is a Ethernet Link, with an MTU of 1500 byte.
I use lwip 1.3.2. For me fragmentation works, I like to check my interfaces 
with 3000 byte Pings.

Regards,
Andreas.

Mit freundlichen Grüßen

i.A. Andreas Becker
Entwicklung Embedded
________________________________
SE-Elektronic GmbH
Eythstrasse 16
73037 Göppingen
Deutschland

-----Ursprüngliche Nachricht-----
Von: address@hidden [mailto:address@hidden Im Auftrag von Timmy Brolin
Gesendet: Sonntag, 6. Februar 2011 23:04
An: address@hidden
Betreff: Re: [lwip-users] Sending UDP data beyond MTU size

I do work with another popular BSD based commercial embedded TCP/IP stack.
In a certain specialty industrial protocol there is a requirement to
reduce the Ethernet MTU for TCP/IP traffic. But when I implemented this,
the TCP/IP stack did not work very well. Turned out that the
fragmentation code in the TCP/IP stack was broken. No one had ever
noticed, since the TCP MSS and maximum outgoing UDP frame size was set
to match the standard Ethernet MTU by default, so the fragmentation code
was never used in practice.

Regards,
Timmy Brolin

On 2011-02-04 16:45, Simon Goldschmidt wrote:
>  "Bill Auerbach" <address@hidden> wrote:
>
>> The other concern is being positive the receiver can handle this - if you're
>> talking to another embedded device it's possible it's not supported.  Isn't
>> this kind of like Jumbo Frames - you can't guarantee that the other end is
>> capable of supporting it?
> Hehe, that's what happens if developers ignore the RFCs. As far as I 
> remember, the requirements for internet hosts makes IP reassembly mandatory 
> while fragmentation is optional...
> Of course, it can be disabled in lwIP, too :-)
>
> Ore than that, IP fragmentation and reassembly works with lwIP. I haven't 
> tested it for a while now, but I see no reason it shouldn't work as we 
> haven't been changing anything around that code path lately.
>
> Simon
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users
>

_______________________________________________
lwip-users mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/lwip-users