[wpa_supplicant] Ad-Hoc/IBSS with 3+ Nodes

[Dan Williams]

On Wed, 2010-08-04 at 08:55 -0400, SJB89 at Drexel.edu wrote:
> Dan - Thanks for the info.  Surprisingly, the output from iwconfig is
> identical on all nodes.  I've done a little more testing since my
> original post and found the following (assuming three nodes only, and
> using group=CCMP in wpa_supplicant.conf):

Does group=TKIP work?  Just trying to narrow things down.

> If nodes A and B come "up" first, they can communicate as expected.
>  When C comes up, the A and B retain their communication.  If, for
> example, A tries to ping C, a series of ARP requests - varying in
> number between 4 and 30+ - are sent from A looking for C's MAC
> address, along with the ICMP_ECHO_REQUEST packets expected from
> running 'ping.'  Eventually, 1-10 ARP replies will be sent by C and
> communication between C and A will resemble the communication between
> A and B previously (low latency, no packet loss). However, the process
> of sending these ARP requests causes the first n packets, where n is
> roughly equal to the number of ARP requests sent by A, to not reach
> their destination.  If A needs to now communicate with B again, the
> same ARP request scenario occurs, even though there is already an
> entry for B in A's ARP table.
> 
> 
> Manually setting static ARP tables does not fix the problem.
>  Additionally, A and B can not simultaneously communicate with C -
> only the first node to successfully pass traffic to and from C can.
>  It appears that there is a pairwise authentication occurring that
> adjusts when a new pair begins communicating.
> 
> 
> 
> Sean