Hi Fabrice, On Wed, Dec 14, 2011 at 2:22 PM, Fabrice Deyber <[email protected]> wrote: >> Perhaps this is an argument for moving the MPM entirely to userspace >> in order to centralize the bugs? > That could be a good idea.
MPM, the open peering protocol is very lightweight, and having it in the kernel allows users to create open mesh networks without the need for a userspace daemon. Considering your inteded uses for o11s, would it be acceptable if you needed a userspace daemon even to establish unsecured links? Of course, before we do that, we would have to improve authsae a bit. As you've noticed, it still has some rough edges... > I ran more tests today here are some additional details. > In my case once the mesh plink expires (I changed the default timeout to 60 > secs), > the kernel destroys the entry, (along with the keys). > This event is ignored by authsae and the plink gets stuck in the LISTEN state > waiting > for mgmt frames. Here the kernel and authsae are out of sync. > The "NL80211_CMD_DEL_STATION" cmd should be processed by authsae and the > corresponding peer entry destroyed. > I'm not sure if it's related to the issue seen by Yeoh > (http://www.mail-archive.com/[email protected]/msg01247.html) > There too the plink is stuck in the LISTEN state, where no mgmt frame seem to > make it to authsae. > Could it be that the kernel and authsae went out of sync? Clearly. authsae currently ignores NL80211_CMD_DEL_STATION and it should not. I've just pushed a preliminary patch for that to this branch: https://github.com/cozybit/authsae/tree/ft-handle-sta-del I will not be able to test it until maybe tomorrow, but if in a hurry, you can try it. Cheers, Javier _______________________________________________ Devel mailing list [email protected] http://open80211s.com/mailman/listinfo/devel
