SixXS::Sunset 2017-06-06

wrong source address (RA address rather than static)
[de] Shadow Hawkins on Tuesday, 21 April 2009 23:51:03
are there any known issues with bridged interfaces and IPv6 networking on Linux..? I recently reconfigured one of my boxes (Linux x86_64, Kernel 2.6.28.9, Gentoo) to use a bridged interface (br0) rather than the plain NIC (eth0). All I've done was setting up a bridge interface with the very same configuration that used to be eth0's configuration and adding eth0 to that bridge. results in [snipped the link local]:
magrathea ~ # ip -6 a s dev br0 4: br0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 inet6 2001:6f8:10ff:f101:21f:e2ff:fe6c:d7fe/64 scope global dynamic valid_lft 2489096sec preferred_lft 501896sec inet6 2001:6f8:10ff:f101::40/64 scope global valid_lft forever preferred_lft forever
So the first address comes advertised from the local radvd on my gateway, and the second is a static address from my subnet. The routing table reads as follows [snipped link local agagin]:
magrathea ~ # ip -6 r s dev br0 2001:6f8:10ff:f101::/64 proto kernel metric 256 expires 2147157sec mtu 1500 advmss 1440 hoplimit 4294967295 2000::/3 via 2001:6f8:10ff:f101::1 metric 1000 mtu 1500 advmss 1440 hoplimit 4294967295
Any originating packet now picks the advertised address rather than the static one.. e.g.:
magrathea ~ # traceroute6 -n www.sixxs.net traceroute to 2001:638:a00:6667::53 (2001:638:a00:6667::53) from 2001:6f8:10ff:f101:21f:e2ff:fe6c:d7fe, 30 hops max, 24 byte packets 1 2001:6f8:10ff:f101::1 0.543 ms 0.511 ms 0.568 ms [..] 11 2001:838:1:1:210:dcff:fe20:7c7c 55.581 ms 54.466 ms 54.258 ms
It used to be (and afaik that's intended by the spec/RFC) that it picked the static address by preference for outgoing packets. Im not aware of that I've changed anything else in e.g. sysctl values or Kernel configuration but creating that bridged interface.. Any other places to look for? Thanks Christian
wrong source address (RA address rather than static)
[de] Shadow Hawkins on Wednesday, 22 April 2009 00:00:16
(edit my own post) actually that traceroute6 above was an incorrect cut&paste from two different traces. correct paste:
magrathea ~ # traceroute6 -n www.sixxs.net traceroute to www.m.sixxs.net (2001:838:1:1:210:dcff:fe20:7c7c) from 2001:6f8:10ff:f101:21f:e2ff:fe6c:d7fe, 30 hops max, 24 byte packets 1 2001:6f8:10ff:f101::1 0.653 ms 0.495 ms 0.485 ms [..] 11 2001:838:1:1:210:dcff:fe20:7c7c 56.961 ms 53.334 ms 55.434 ms
wrong source address (RA address rather than static)
[de] Shadow Hawkins on Wednesday, 22 April 2009 07:57:35
Nothing you can do, at least not until See http://madduck.net/docs/ipv6/#choosing_source_addresses .
wrong source address (RA address rather than static)
[de] Shadow Hawkins on Wednesday, 22 April 2009 08:00:19
Sorry, I meant to say "at least not until addrlabel support is in the kernel". Maybe this is the case with 2.6.26 already, but I've had no time to play around with that.
wrong source address (RA address rather than static)
[de] Shadow Hawkins on Friday, 08 May 2009 00:24:54
Tag [/link] is not closed
wrong source address (RA address rather than static)
[de] Shadow Hawkins on Friday, 08 May 2009 00:29:41
maybe this post about recent Linux kernels following RFC 3484 fits in this issue. Alas in my case, that would mean to set the IP given by radvd a value of "preferred_lft 0" - which imho is impossible to do automagically by an init-script.. P.S. As this particular machine has it's static IPv6 assigned, it does not actually need an additional advertized one. So I've set:
net.ipv6.conf.default.accept_ra=0 net.ipv6.conf.all.accept_ra=0 net.ipv6.conf.default.autoconf=0 net.ipv6.conf.all.autoconf=0
in /etc/sysctl.conf. Christian (P.P.S About "what can be done better": The Forum needs a "preview post" button and/or an "(re)edit my own post" button...)

Please note Posting is only allowed when you are logged in.

Static Sunset Edition of SixXS
©2001-2017 SixXS - IPv6 Deployment & Tunnel Broker