[Olsr-dev] OLSR Basic Multicast Forwarding (BMF) plugin v1.6.2 available

[Erik Tromp]

Hello all,

A new version of the Basic Multicast Forwarding (BMF) plugin is available.

Here is the change log of version 1.6.2 with respect to the previous version
1.6.1:

* Fixed a bug that prevented the route for multicast traffic to be updated
  when a network interface was added at runtime --> thanks to Daniele
Lacamera
  for finding and solving this bug.

I also updated BMF source code for olsrd-0.5.6 such that it compiles with
the current
olsrd-0.5.6-r2 release.

As always, the source package can be downloaded at:
http://sourceforge.net/projects/olsr-bmf


To Joerg (see e-mails below):

I checked your observation. It appears that even if two different types of
sockets (one cooked IP packet socket and one datagram over IP socket) are
waiting for the same packet, the packet is received only by the datagram
over IP socket (encapsulating) and not also by the cooked IP packet socket
(listening).

The encapsulating socket is always needed, for writing to. The listening
socket is also always needed, to 'overhear' the unicast packets that are not
destined to the local node.

Thanks anyway!

> -----Oorspronkelijk bericht-----
> Van: Erik Tromp [mailto:(spam-protected)]
> Verzonden: woensdag 27 augustus 2008 21:25
> Aan: '(spam-protected)'; '(spam-protected)'
> Onderwerp: RE: [Olsr-users] BMF-Plugin in UnicastPromiscuous mode
> 
> Hi Joerg,
> 
> I guess you're right. I will make this optimization in the next 
> release.
> 
> Thanks for the sharp observation!
> 
> Best regards,
> Erik
> 
> 
> > Hello all (esp. Erik),
> > I'm wondering about something I noticed in the bmf-1.6.1 plugin:
> > 
> > In UnicastPromiscuous it creates for every OLSR interface both an 
> > AF_INET and an AF_PACKET socket and listens for incoming packets on 
> > both sockets (both CreateListeningSocket and CreateEncapsulateSocket 
> > call AddDescriptorToInputSet).
> > Doesn't this mean that the plugin sees a BMF-Packet addressed to 
> > this node twice (on purpose, not just by accident)?
> > One copy should fail in the duplicate packet detection, but I 
> > thought this was not part of the normal operation and only meant to 
> > prevent loops.
> > 
> > --
> > Regards
> >        Joerg


Regards,
Erik