Hello,

I can not reproduce the problem on 2.6.24, but i can on 2.6.22.19 with the same batman-adv source (rev1084). So it looks like the two kernel have a different behaviour for this problem.

@Paolo: Which kernel version did you use?

As i can reproduce it now, i'll look in this problem. :)

Thanks for the report, Simon

On Thu, Jun 05, 2008 at 10:38:13AM +0800, Marek Lindner wrote:

On Thursday, 5. June 2008 00:11:31 Sven Eckelmann wrote:

...

...

Someone can help me?

Can you please test it further? I used uml with linux 2.6.22.5 and uml_switch configured as hub to produce the same problem, but dotslash told me that he wasn't able to reproduce it with his uml image (2.6.24 based). Maybe I can help him to get the same effect it so he can try to find the problem.

Same with me - I can't reproduce the problem but I use a setup similar setup to Simons setup. Could you try to use 2.6.24 to see whether it depends on the kernel version ?

Greetings, Marek _______________________________________________ B.A.T.M.A.N mailing list B.A.T.M.A.N@open-mesh.net https://list.open-mesh.net/mm/listinfo/b.a.t.m.a.n

On Thursday, 5. June 2008 04:38:13 Marek Lindner wrote: Same with me - I can't reproduce the problem but I use a setup similar setup to Simons setup.

I gave Simon my (nearly) complete uml test environment and he was able to kill batman+kernel with a small ping packet.

Could you try to use 2.6.24 to see whether it depends on the kernel version ?

Yes, I tried it and it happened again (and this time no warning in dmesg at all). The problem seems to be gone when I configure my uml_switch as switch instead of hub (removing "-hub" from commandline) - but sometimes it crashes again when all nodes try to send pings to each other (2 -> 1, 3 -> 1 and 1 -> 2). So I am really sure that you know what it looks like :(

Greetings Sven

Hey Paolo,

thanks again for your testing. I've tried a few old revisions and could reproduce the "ping of death" problem in Svens 2.6.22 enviroment. Even Revision 1024, which worked flawlessly before in my UML, on Accton routers and other devices in various setups, showed this problem. :(

So there seems to be some hard to catch issue and i haven't found the bug yet (two others instead ;), but i'm working on it.

For the time being you may try your luck with revision 1024, maybe it works for you. It also has some bugs, but worked out in some scenarios i set up.

Thanks again Simon

On Thu, Jun 05, 2008 at 04:03:41PM +0200, cipollone wrote:

Hi,

I've tested B.A.T.M.A.N. Advanced Kernel Land version 6 with a different Linux OpenWRT Kernel Version. I've used the # 2.6.24.2 and , also in this case, I can't ping the other pc with the same kernel version.

Now, with kernel version 2.6.25.4, 2.6.22.2 and 2.6.24.2, Batman doesn't send packets. I think there is a problem passing in layer 3 because if I ping itself it answer at the request ( it is because it doesn't go in the layer 3, but stay at layer 2 with ARP).

I'm reading the log message in order to understand what's the problem and where is.

Now I have also putted vde switch in hub mode.

Regards

Paolo

---

B.A.T.M.A.N mailing list B.A.T.M.A.N@open-mesh.net https://list.open-mesh.net/mm/listinfo/b.a.t.m.a.n