Odd issue with 6in4 (sit) tunnels

Ruben Laban r.laban at ism.nl
Mon Feb 28 16:03:57 UTC 2011


Hi,

I'm seeing something very odd with 6in4/sit tunnels: when I specify the local IPv4 address, the link local address of the tunnel gets a /128 mask assigned instead 
of the usual /64:

$ sudo ip tunnel add sit-ok  mode sit remote 10.10.10.10
$ sudo ip tunnel add sit-not-ok  mode sit local 192.168.178.30 remote 10.10.10.10
$ sudo ip link set sit-ok up
$ sudo ip link set sit-not-ok up
$ ip -6 address show sit-ok
21: sit-ok: <POINTOPOINT,NOARP,UP,LOWER_UP> mtu 1480 
    inet6 fe80::c0a8:b21e/64 scope link 
       valid_lft forever preferred_lft forever
$ ip -6 address show sit-not-ok
22: sit-not-ok: <POINTOPOINT,NOARP,UP,LOWER_UP> mtu 1480 
    inet6 fe80::c0a8:b21e/128 scope link 
       valid_lft forever preferred_lft forever
$ 

I can't think of any reason why this would be intended behaviour. Both cases do get a fe80::/64 route:

$ ip -6 route show | grep sit
fe80::/64 via :: dev sit-ok  proto kernel  metric 256  mtu 1480 advmss 1420 hoplimit 0
fe80::/64 via :: dev sit-not-ok  proto kernel  metric 256  mtu 1480 advmss 1420 hoplimit 0

Am I missing something here, or is it just some oversight somewhere in the IPv6 handling?

Regards,
Ruben Laban



More information about the kernel-team mailing list