[lwip-users] [lwip]Stack crashing with huge ICMP packets

[Dennis Borgmann]

Hi lwIP-users!

I am sorry, if this mail seems to be "old stuff" for most of you, but I
could not find any detailed information about what I am facing here.

I have lwIP 1.3.2 running combined with FreeRTOS on an AVR32UC3A0512. It
works fine, but as soon as I start a ping with packetsize 2000(or
something similar), the stack seems to crash - meaning, I can't reach
the device via IP anymore. As long as I stay below 1536, everything just
runs fine.

What could be the option, that I'd have to increase in order to get a
"stable" lwIP-stack with packets bigger than 1536(let's say even 65000...)?

Thank you a lot in advance for any single hint(or even a good link to
some info about my question)!

Best regards,
Dennis

_______________________________________________
lwip-users mailing list
lwip-...@nongnu.org
https://lists.nongnu.org/mailman/listinfo/lwip-users

[Kieran Mansley]

On 24 Jan 2012, at 11:10, Dennis Borgmann wrote:

> Hi lwIP-users!
>
> I am sorry, if this mail seems to be "old stuff" for most of you, but I could not find any detailed information about what I am facing here.
>
> I have lwIP 1.3.2 running combined with FreeRTOS on an AVR32UC3A0512. It works fine, but as soon as I start a ping with packetsize 2000(or something similar), the stack seems to crash - meaning, I can't reach the device via IP anymore. As long as I stay below 1536, everything just runs fine.

I think it could just be that the ping code in that release assumes that the whole datagram will fit in a single packet. When you exceed the MTU the datagram will be fragmented and I doubt the code has been written to cope with that.

> What could be the option, that I'd have to increase in order to get a "stable" lwIP-stack with packets bigger than 1536(let's say even 65000...)?

I would try and see if (i) the more recent versions of lwIP do what you want, or (ii) if the problem is just with ICMP; I would hope that large UDP datagrams are handled correctly for example.

Kieran

[Simon Goldschmidt]

Kieran Mansley <kie...@recoil.org> wrote:

>> As long as I stay below 1536, everything just runs fine.
>
> I think it could just be that the ping code in that release assumes that the whole datagram will fit in a single packet. When you exceed the MTU the datagram will be fragmented and I doubt the code has been written to cope with that.

Hmm, I also thought about that, but 1536 is already bigger than the MTU, isn't it?

> I would try and see if (i) the more recent versions of lwIP do what you want

Definitively. And if not, it will be much easier to try and reproduce the problems you are seeing.

Simon

[Dennis Borgmann]

Hi Kieran,
hi Simon,

thank you for your thoughts on my problem. So far I have solved the
problem by simply setting IP_REASSEMBLY and IP_FRAG both to "0". That
way no fragmented packes are processed at all and since I am only using
a tftp, this shouldn't happen anyway and it works for me.

Thanks anyway. Maybe one day, if I get the time to do so, I might try
lwIP 1.4.

Best regards,
Dennis

[Arvind Dandotiya]

Hi lwIP-users! and Dennis Borgmann

I have lwIP 1.3.0 running combined with standalone on Spartan 3E. It
works fine, but it getting stuck(crush) after random time.

    sometimes work full day or sometimes getting stuck after one hour,
  one more problem faced, that is parallel ping (fping) ,when i start fping immediately getting stuck(crash).
   if any body face like that problem please reply

 

Best regards,
Arvind