> Do you have jumbograms enabled on your physical network, or are the two VMs > running on the same machine connected by a virtual network? > > That aside, tinc should handle that case, so that is definitely a bug. I'll > try > to find out what is happening.
Yes, for testing purposes, I do this test on 2 VM, on a same host, connected to a Linux bridge. But everywhere, MTU is 1500. And that installation worked fine with Tinc 1.0. Setting 192.168.0.0/24 interface to MTU 1400 in VM worked a few minutes and then failed (back to 1518 bytes packets). For the few seconds I could test, performances are better than Tinc 1.0 but not awesome in my tests (179 Mbit/s vs 135 Mbit/s). Here is my sptps_speed for both VM (aes & pclmulqdq instructions are available in both VM, host is an i5-2400) : root@host2:~# sptps_speed Generating keys for 10 seconds: 580.77 op/s ECDSA sign for 10 seconds: 538.36 op/s ECDSA verify for 10 seconds: 457.77 op/s ECDH for 10 seconds: 272.66 op/s SPTPS/TCP authenticate for 10 seconds: 130.10 op/s SPTPS/TCP transmit for 10 seconds: 5.01 Gbit/s SPTPS/UDP authenticate for 10 seconds: 130.27 op/s SPTPS/UDP transmit for 10 seconds: 5.24 Gbit/s _______________________________________________ tinc mailing list [email protected] http://www.tinc-vpn.org/cgi-bin/mailman/listinfo/tinc
