The start_seq_num is taken from the station's tid_seq[tid]. (agg-tx.c / line 304). This would be fine, but there is a small pitfall that we know too well: tid_seq is meant to produce the sequence, which can be insert into an IEEE 802.11 frame sequence_control field without the 4 bit left shift (which is usually necessary to accommodate for frame fragmenting). So tid_seq is _increased_ in 0x10 steps, but some of the code in agg-tx.c doesn't know about that, e.g: ieee80211_send_addba_request: (line 91) mgmt->u.action.u.addba_req.start_seq_num = cpu_to_le16(start_seq_num << 4); Now, ADDBA recipient certainly gets a wrong BA starting point and may decide to drop some of the incoming aggregated frames, since it might think that they were already too old. --- Hmm, I'm not sure what to do with the debug_sta.c counters? Obviously, they are now >> 4 as well, but on the other hand just like the tid_rx side always was... Note: By the looks of it: iwlagn/ath9k are probably unaffected, but carl9170 was: so no need for stable. --- diff --git a/net/mac80211/agg-tx.c b/net/mac80211/agg-tx.c index ceda366..5aa8f4a 100644 --- a/net/mac80211/agg-tx.c +++ b/net/mac80211/agg-tx.c @@ -301,7 +301,7 @@ int ieee80211_start_tx_ba_session(struct ieee80211_sta *pubsta, u16 tid) * call back right away, it must see that the flow has begun */ *state |= HT_ADDBA_REQUESTED_MSK; - start_seq_num = sta->tid_seq[tid]; + start_seq_num = sta->tid_seq[tid] >> 4; ret = drv_ampdu_action(local, sdata, IEEE80211_AMPDU_TX_START, pubsta, tid, &start_seq_num); -- 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