On Wed, Aug 06, 2014 at 01:06:54PM +0000, Stuart Henderson wrote: > On 2014-08-01, Mike Larkin <mlar...@azathoth.net> wrote: > > I know other people have said that hostap mode is not > > stable but on this machine it's been rock solid. > > Historically there have been problems where a setup could be completely > stable, then move it to a different environment (different clients around? > different other APs? I'm not sure) and it would hang frequently. I'm not > sure of the current status, I stopped using hostap on OpenBSD some years > ago - athn(4) would be my first choice if I were to try it again.
This might have been the "full node cache is never cleared" bug. It couldn't be noticed much in an environment with few wifi devices around. Passing non-associated devices (laptops, other APs, phones...) filled up the node cache (i.e. the list of visible stations) and were never purged. Once the maximum cache size was reached (100 nodes) things stopped working. If I remember correctly, whenever new a node popped up the AP would try to purge the node cache but didn't actually remove any node. And on top of that it sent a de-auth frame to each node without discrimination whenever it tried to purge the cache. In plain English the AP told every device trying to join the network or just passing by to piss off until ifconfig down/up actually cleared the cache. These issues were fixed for 5.1. But many other problems (such as low throughput and range with some ral(4)-based APs) remain. athn works well given the card is supported. I use one as my primary AP, too.
