Opened 12 years ago

Last modified 12 years ago

#168 new technical

Multi-Homing Support

Reported by: christian.vogt@… Owned by: Sri Gundavelli <sgundave@…
Priority: normal Milestone: PMIPv6 base specification
Component: draft-ietf-netlmm-proxymip6 Severity:
Keywords: multi-homing Cc:


P-MIPv6 currently supports mobility on the host level. This leads to problems when hosts are multi-homed. This problem was identified by Hesham Soliman at IETF 69 in Chicago, and Vidya has written down the following problem statement:

  • MN has two interfaces attached to MAG1 and MAG2 under the same LMA
  • Let's say MAG1 first created the binding at the LMA and acquired a prefix/address for the MN, using MN-ID1
  • MN obtains the prefix/address on Interface1; if the MN is using DHCP, it will only get an address (since there is no requirement on the MN to do DHCP-PD)
  • MAG2 now sends a PBU with MN-ID1 also (very possible when the ID is something like an NAI)
  • LMA thinks the MN moved, updates the binding and returns the same prefix for the MN to MAG2
  • MN receives the same prefix/address via DHCP or RA over Interface2; MN detects a duplicate address and per RFC2462, does not configure that address on Interface2; Interface2 is now not enabled for IP forwarding;
  • If the MN is using stateless autoconfiguration, it may end up with a different address for Interface2; in this case, Interface2 is enabled; the assigned prefix, however, is no longer just over a point-to-point link, although, the MAG thinks it is
  • The LMA has a binding for MAG2 (corresponding to Interface2)
  • If the MN never enabled Interface2 due to duplicate address issues, the MN will try to use Interface1, but, unsuccessfully, since the LMA will drop those packets; LMA will send packets to MAG2 (Interface2), which will not be received by the MN
  • If the MN has enabled Interface2 with a different address, depending on the implementation, the MN may use Interface1 or Interface2 (some OSs will just pick one); any communication attempted via Interface 1 will fail
  • In the meantime, the binding may keep toggling between MAG1 and MAG2 at the LMA, due to corresponding PBUs received; the MN may have intermittent connectivity as a result of this

Change History (1)

comment:1 Changed 12 years ago by christian.vogt@…

Since this issue could so far not be resolved depite a thorough mailing list discussion, the working group chairs have decided to proceed as follows:

1) Vidya has asked for outside-NETLMM expert opinions about the topic.

2) The working group will wait until Friday, Sep 5, to get answers back from these outside experts. The results will be then posted on the mailing list for discussion.

3) If the discussion starting on Friday does not result to consensus a consensus call will be held. Jonne will do the consensus call, and Vidya will contribute to the technical discussion.

Note: See TracTickets for help on using tickets.