Re: [reTurn-devel] Found two bug in reTurn server
These two issues sound very familiar. I've fixed a bunch of reTurn issues since 1.4.1 after updating to behave-turn-12 and running against the stun test vectors (draft-ietf-behave-stun-test-vectors-04). You can these fixes directly from the resiprocate SVN repository. They will make it into the 1.5 cut of resiprocate that is in the works.
Let me know if you still experience issues with latest code.
Thanks,
Scott
On Mon, Apr 27, 2009 at 3:16 AM, Frans Maas
<FMaas@xxxxxxxxxxxx> wrote:
Hi,
I discovered the following two bugs in 1.4 and 1.4.1 while using Adobe Flash 10 the TURN client.
The StunMagicCookie is tested and created in reversed order. Xor-ing of peer address with this cookie works fine, though.
The problem is in StunMessage.cxx lines 156 and 487.
May be relevant to know that I am working on a RHEL5 Linux server.
My work around was to create an additional const in StunMessage.hxx with cookie value in reversed order and use that on the two places indicated.
In TurnAllocation::sendDataToClient I always got “sendDataToClient RemotePeer info not found”. It appeared that no entries are made if no ChannelBind request is received.
I solved this by letting ChannelManager::findRemotePeerByPeerAddress return new RemotePeer(peerAddress, 0) on line 87.
Hope this helps.
As I cannot oversee possible side effects for other use cases I kept these changes local to my environment.
Regards,
Frans Maas
_______________________________________________
reTurn-devel mailing list
reTurn-devel@xxxxxxxxxxxxxxx
List Archive: http://resiprocate.org/archive/return-devel/