[hipl-users] Re: HIP mobility
- From: Miika Komu <miika@xxxxxx>
- To: hipl-users@xxxxxxxxxxxxx
- Date: Tue, 22 Jun 2004 16:35:57 +0300 (EEST)
On Mon, 21 Jun 2004, Simon Schuetz wrote:
> I experienced some problems with readdressing.
> Are there any limits on the frequency of HIP readdressing?
> I run the following experiments:
> - mobile node can use two interfaces, but at most one is
> configured at the same time
> - interface1 is up for 17 seconds, then for another 17 seconds
> no interface is configured
> - then interfaces are switched: interface2 is configured for
> 17 seconds and then no interface is configured for 17 seconds
> - repeating this until all data is transferred
>
> I observed, that HIP readdressing fails quite often in this scenario
> (checked entries in /proc/net/hip/sdb_state while switching interfaces).
> However, it seems to work quite fine when using longer times than 17
> seconds.
The endhost multihoming readdressing works better (surprise, surprise) if
you have always at least one interface up. You're scenario is not probably
working, because you're using link local addresses? In both the 2.4 and
2.6 series, the link local addresses are not sent anymore in the
readdressing packets because it is very difficult to detect when the
peer is in the same LAN. So, you need to do something like...
% initially eth1 is down
host1 # ifconfig eth0 down
host1 # ifconfig eth1 up
host1 # ifconfig eth1 add 3ffe::55/64
% Now, the host sends readdressing packet to the peer (host2). Look the
% /proc/net/hip/sdb_state on the peer and you should see the 3ffe::55
% address there.
Does this help you?
> I am currently using HIPL for Kernel 2.4.20 with patch version 60.
>
> Is the mobility extension of current HIPL version (kernel 2.6) more
> reliable,
> such that it is worth to switch to the new version?
We just tested readdressing on 2.6 with the vmware and it (still) seems to
work :) The 2.6 series is better than the 2.4 series because...
* We are working on the 2.6 series now, we won't do any new features to
the 2.4 series. We have a very limited time table and we may not have
time for doing bug fixes on the 2.4.
* Mika probably provides the mm-02-pre support sometime during the summer
to the 2.6 series. The mm-02 is incompatible with the mm-00 support.
On the other hand, we have not tested the 2.6 series so much as the 2.4.
We don't want to make changes to the old mm-00, which is still present in
the 2.6 series, until Mika replaces it with the mm-02 support.
P.S. The link local addresses are not signalled in the 2.6 series either.
--
Miika Komu miika@xxxxxx http://www.iki.fi/miika/
Other related posts:
