Heho, and a quick follow up on the root cause: The HE tunnel host continuously sent packet-to-big messages indicating an MTU of 1480. The packets that triggered the message had an MTU of 1476, so my openbsd just resent those packets as before (honoring RFC8201 by the letter, i.e., reducing the MTU to the value communicated in the ICMP packet; in this case not changing it, triggering the next ICMP message...):
When a node receives a Packet Too Big message, it must reduce its estimate of the PMTU for the relevant path, based on the value of the MTU field in the message. The root cause seems to have been a pppoe link on path which dropped the MTU to 1472 effectively, while the tunnel was still configured at 1480. I am not yet sure whether this is a sit/gif implementation issue at HE or a general thing. I will try to reproduce the setup next week, also to test if this is maybe an OpenBSD only thing; Thinking about it, being able to make a system send ~1-3GB as long as you can make it open a tcp connection to you sounds like a rather not so fun thing. With best regards, Tobias On Sun, 2023-03-12 at 12:32 +0100, Tobias Fiebig via mailop wrote: > Heho, > ok, just had the confirmation in my inbox that this is a .gov > measurement/research project. Will discuss the matter of 'probe > attribution' with them. > > With best regards, > Tobias > _______________________________________________ > mailop mailing list > [email protected] > https://list.mailop.org/listinfo/mailop -- Dr.-Ing. Tobias Fiebig T +31 616 80 98 99 M [email protected] _______________________________________________ mailop mailing list [email protected] https://list.mailop.org/listinfo/mailop
