OK so I'm done with some patches that don't oops, at least immediately, and its nothing pretty. I have patchsets for 2.6.27-rc8 and for wireless-testing as of today. I only tested the patchset for 2.6.27-rc8 but remember -- this is just for distributions who want some 11n throughput on 2.6.27 or developers who want to test/hack on aggregation. This is in no way shape or form should be considered to be merged as its not up to upstream merge inclusion standards: * The patch does too much rework * I think ieee80211_reque() is requeuing onto the the same queue upon removal of a queue, could be wrong * I see no point in using ieee80211_requeue() being called upon rmmod path * I just uncommented ieee80211_ht_agg_queue_remove() out which fixes rmmod panic on scheduling while atomic -- this path only gets hit upon ieee80211_stop_tx_ba_cb() which so far I've seen happen upon rmmod, but I'm sure there are some other code paths it hits somewhere I guess. If indeed ignoring ieee80211_ht_agg_queue_remove() is OK under ieee80211_stop_tx_ba_cb() I'd feel more comfortable on recommending this strongly but at this point I am just not sure. Anyway at this point I'm going to stop looking at 2.6.27 as we just need to give aggregation a good cleaning and this needs to happen on wireless-testing. First priority is to get it working, and if I get to it first I plan on doing that by removing the whole tid to queue mapping, and simply mapping skbs to ACs. Then the driver can do what it needs for scheduling. Let me know what you think. Use at your own risk. http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/mac80211/aggrv9-2.6.27.tar http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/mac80211/aggrv9-wl.tar Luis -- 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
