Hi! I try to setup a ipip6 tunnel. I have two endpoints, one is native IPv6 the other is SIXXs tunnel. The sixxs default MTU is 1280 while native is ethernet's 1500. On top of that sits the ip in ipv6 tunneling and here comes the issue. In Linux the ipip6 interface is treated as ipv6 interface and thus is not allowed for MTU lower than 1280 despite the fact that only ipv4 communication is about to happen over it. See: http://webcache.googleusercontent.com/search?q=cache:Qz87HRX1kE4J:kerneltrap.org/mailarchive/linux-netdev/2010/3/19/6272410+&cd=1&hl=en&ct=clnk&gl=pl (google cached) So while for ipip6 the minimum is 1280 the trivial solution is to bump up underlying aiccu tunneling to 1320 (1280+the ipip6 tunneling overhead). Sounds trivial but unfortunately the sixxs gateway is doing something very strange. If I tracepath FROM my sixxs tunnel to my native ipv6 i got: tracepath6 emsi 1?: [LOCALHOST] 0.130ms pmtu 1320 1: gw-306.waw-01.pl.sixxs.net 11.720ms 1: gw-306.waw-01.pl.sixxs.net 10.456ms 2: 2001:6a0:1:1::2 10.112ms asymm 1 3: kp-mm-r3.rtr.net.icm.edu.pl 10.766ms asymm 2 4: 2001:7f8:42::xxxx:xxxx:1 9.772ms 5: no reply 6: no reply 7: emsi 13.115ms reached Resume: pmtu 1320 hops 7 back 59 Note that I bumped the mtu of my aiccu to 1320. Seems it works right? Unfortunately the tracepath from my native ipv6 to sixxs back shows that sixxs gw is enforcing 1280 mtu: 1?: [LOCALHOST] 0.016ms pmtu 1500 1: 2001:4190:xxxx:xxxx::1 1.208ms 1: 2001:4190:xxxx:xxxx::1 1.190ms 2: 2001:4190:xxxx:xxxx:582 1.678ms 3: 2001:4190:xxxx:4::1 1.734ms 4: 2001:4190::3 2.580ms asymm 3 5: 2001:6a0:1:1::2 2.944ms 6: gw-306.waw-01.pl.sixxs.net 2.981ms pmtu 1280 ^^^^^^^^^ 6: gw-306.waw-01.pl.sixxs.net 3.026ms 7: rumunek 16.769ms reached Resume: pmtu 1280 hops 7 back 58 The very same gw had no problems of passing larger packets form the tunnel. Any idea why it happens or how could I solve the issue?