Hi Javier, Thanks for replying, answers in the text below. > Envoyé : jeudi 2 mai 2013 21:37 > On Thu, May 2, 2013 at 1:22 AM, Jean-Pierre Tosoni <[email protected]> > wrote: > > (...) > > I can see the same behaviour here. Steps to reproduce are, > > I am not sure the two problems are related given that David does not > seem to be forcing topologies via iw block commands. Indeed he is using iw...mpath del / new But, doesn't it make the path change after the next path refresh timeout ?
> > 1/ set up two peers (A and B) normally > > 2/ use iw on peer A to block peer B (iw wlan0 station set ... plink_action > > block > > => no frame is transmitted > > Keep in mind that blocking is directional: if you issue iw > plink_action block on node A, only incoming traffic from B -> A will > be blocked. > In most cases you will also want to block A on node B. Could you try that? I meant that after issuing the iw command on station A, and before issuing it on station B, I was expecting a PLINK CLOSE going from A to B, but I do not see any > > > 3/ use iw on peer A to restore the link. A PEER OPEN is sent and nothing > > else happens. > > How do you restore the link? With "iw wlan0 station set ... plink_action open" on station A (where I executed the block) > > > (...) > > >> 2 Another strange behavior that is repeated many times in my tests. A > > network with 4 nodes in line A-B-C-D (A only has a B peer..., B has A and > > C as peers...), > > > > I am using the same kind of setup. To achieve this I am using iw ... > > plink_action. How do you enforce the in-line topology ? Is there some > other > > way ? > > We routinely test linear topologies like the one you describe by > blocking traffic both ways. That should work. Ah, thanks. That is what I am doing, I just wondered about any other way. I saw that David uses mpaths, but there are timeouts involved, contrary to the peer links. Cheers, Jean-Pierre _______________________________________________ Devel mailing list [email protected] http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel
