On Wed, Mar 10, 2010 at 03:59:57PM -0800, Johannes Berg wrote: > I have objections. Or maybe not, but I don't think this makes a whole > lot of sense since the BE queue should be programmed in the "qos turned > off" way normally. That may be broken right now, but to me your patches > seem like a workaround rather than a solution which would require an > analysis of what's really going on. I looked at this on the air. Everything is fine until device send QoS NULL frame (even with default parameters: normal ACK, TXOP request). Then AP send two strange Probe Response frames and Disassociate frame with "Micheal MIC failure" error. So I think problem is with that, when WMM is configured device send QoS frames, which AP do not expect. So when AP does not support WMM, we should have QoS disabled. Currently in mac80211 there is no such option, so I add it. In kernels before 2.6.31 everything works fine, because drivers use own tx queue defaults with QoS disabled. I guess we can do something better with that, for example change ieee80211_set_wmm_default() to do: if (data->vif.type == NL80211_IFTYPE_AP || data->vif.type == NL80211_IFTYPE_ADHOC) set_default_tx_queues(); else disable_qos(); But I wrote fix to do not affect other drivers, which I'm unable to test and which may relay on current mac80211 behaviour. Fix is intended to be backported. During next development cycle I plan revise my patch and maybe send some more code changes related with that area. Stanislaw -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
