[support] Closing MPS/MPA issue

[Arnaud Ebalard]

Hi Mattias,

I have just updated (0.4-15) the umip package in the debian repo on
natisbad.org. It includes the following additional patches:

    prevent_OptimisticHandoff_on_home_return.patch
    configurable_InitialSolicitTimer.patch
    reuse_previous_id_in_mps_retransmission.patch
    check_netlink_pid.patch
    check_nd_opt_size.patch

I think the second and third patches address your needs.

Note that the InitialSolicitTimer configuration option added by second
patch is not public (no advertisement in the man page) but I will
maintain the patch.

For the third patch, I asked for some thoughts on IETF MEXT WG mailing
list [1] but got not reply. I implemented the following: the ID in an
MPS reemission is not randomized. So, even if the RTT is just a bit
higher than the reemission timer, the second MPS sent by the MN will
have the same ID as the previous one. When the MPA will be received, it
will have current ID and will be processed as expected.

AFAICT, this is compliant with what RFC 3775 specifies. If you can test
it and see how things work, I'd be interested by some feedback. Note
that I wanted to remove the code which replies with a new MPS when
receiving an MPA with a mismatching ID but did not do it yet.

As a side note, current package does not yet address the packing issues
reported on the list but this is a work in progress. 

Cheers,

a+

[1]: http://permalink.gmane.org/gmane.ietf.mip6/9433