From 836c1addd3bf42d47ca5f7a5780a7ec52abf332e Mon Sep 17 00:00:00 2001 From: Felix Fietkau Date: Fri, 24 Jul 2026 11:54:28 +0000 Subject: [PATCH] wifi: mac80211: add AQL support for multicast packets Excessive multicast traffic with little competing unicast traffic can easily flood hardware queues, leading to throughput issues. Additionally, filling the hardware queues with too many packets breaks FQ for multicast data. Fix this by enabling AQL for multicast packets. Signed-off-by: Felix Fietkau Link: https://patch.msgid.link/20260724115429.3921457-3-nbd@nbd.name Signed-off-by: Johannes Berg --- include/net/cfg80211.h | 1 + include/net/mac80211.h | 3 ++- net/mac80211/debugfs.c | 13 ++++++++-- net/mac80211/ieee80211_i.h | 2 ++ net/mac80211/main.c | 1 + net/mac80211/sta_info.c | 17 ++++++++++++- net/mac80211/sta_info.h | 3 ++- net/mac80211/status.c | 5 ++-- net/mac80211/tx.c | 52 ++++++++++++++++++++------------------ 9 files changed, 66 insertions(+), 31 deletions(-) diff --git a/include/net/cfg80211.h b/include/net/cfg80211.h index a201979f066e..dcbb0e55c32e 100644 --- a/include/net/cfg80211.h +++ b/include/net/cfg80211.h @@ -3742,6 +3742,7 @@ enum wiphy_params_flags { /* The per TXQ device queue limit in airtime */ #define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_L 5000 #define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_H 12000 +#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_MC 50000 /* The per interface airtime threshold to switch to lower queue limit */ #define IEEE80211_AQL_THRESHOLD 24000 diff --git a/include/net/mac80211.h b/include/net/mac80211.h index 999e5189f113..a68bccebc525 100644 --- a/include/net/mac80211.h +++ b/include/net/mac80211.h @@ -1323,6 +1323,7 @@ ieee80211_rate_get_vht_nss(const struct ieee80211_tx_rate *rate) * @status_data: internal data for TX status handling, assigned privately, * see also &enum ieee80211_status_data for the internal documentation * @status_data_idr: indicates status data is IDR allocated ID for ack frame + * @tx_time_mc: TX time estimate is for a multicast frame, used internally * @tx_time_est: TX time estimate in units of 4us, used internally * @control: union part for control data * @control.rates: TX rates array to try @@ -1365,8 +1366,8 @@ struct ieee80211_tx_info { status_data_idr:1, status_data:13, hw_queue:4, + tx_time_mc:1, tx_time_est:10; - /* 1 free bit */ union { struct { diff --git a/net/mac80211/debugfs.c b/net/mac80211/debugfs.c index a4d5461f6480..8ebf5bcf3c0e 100644 --- a/net/mac80211/debugfs.c +++ b/net/mac80211/debugfs.c @@ -210,11 +210,13 @@ static ssize_t aql_pending_read(struct file *file, "VI %u us\n" "BE %u us\n" "BK %u us\n" + "MC %u us\n" "total %u us\n", atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_VO]), atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_VI]), atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_BE]), atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_BK]), + atomic_read(&local->aql_mc_pending_airtime), atomic_read(&local->aql_total_pending_airtime)); return simple_read_from_buffer(user_buf, count, ppos, buf, len); @@ -239,7 +241,8 @@ static ssize_t aql_txq_limit_read(struct file *file, "VO %u %u\n" "VI %u %u\n" "BE %u %u\n" - "BK %u %u\n", + "BK %u %u\n" + "MC %u\n", local->aql_txq_limit_low[IEEE80211_AC_VO], local->aql_txq_limit_high[IEEE80211_AC_VO], local->aql_txq_limit_low[IEEE80211_AC_VI], @@ -247,7 +250,8 @@ static ssize_t aql_txq_limit_read(struct file *file, local->aql_txq_limit_low[IEEE80211_AC_BE], local->aql_txq_limit_high[IEEE80211_AC_BE], local->aql_txq_limit_low[IEEE80211_AC_BK], - local->aql_txq_limit_high[IEEE80211_AC_BK]); + local->aql_txq_limit_high[IEEE80211_AC_BK], + local->aql_txq_limit_mc); return simple_read_from_buffer(user_buf, count, ppos, buf, len); } @@ -273,6 +277,11 @@ static ssize_t aql_txq_limit_write(struct file *file, else buf[count] = '\0'; + if (sscanf(buf, "mcast %u", &q_limit_low) == 1) { + local->aql_txq_limit_mc = q_limit_low; + return count; + } + if (sscanf(buf, "%u %u %u", &ac, &q_limit_low, &q_limit_high) != 3) return -EINVAL; diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h index 4034b71a31cf..a1ef88fe846d 100644 --- a/net/mac80211/ieee80211_i.h +++ b/net/mac80211/ieee80211_i.h @@ -1464,10 +1464,12 @@ struct ieee80211_local { spinlock_t handle_wake_tx_queue_lock; u16 airtime_flags; + u32 aql_txq_limit_mc; u32 aql_txq_limit_low[IEEE80211_NUM_ACS]; u32 aql_txq_limit_high[IEEE80211_NUM_ACS]; u32 aql_threshold; atomic_t aql_total_pending_airtime; + atomic_t aql_mc_pending_airtime; atomic_t aql_ac_pending_airtime[IEEE80211_NUM_ACS]; const struct ieee80211_ops *ops; diff --git a/net/mac80211/main.c b/net/mac80211/main.c index f996e15e3d9e..a59837b9f480 100644 --- a/net/mac80211/main.c +++ b/net/mac80211/main.c @@ -986,6 +986,7 @@ struct ieee80211_hw *ieee80211_alloc_hw_nm(size_t priv_data_len, spin_lock_init(&local->rx_path_lock); spin_lock_init(&local->queue_stop_reason_lock); + local->aql_txq_limit_mc = IEEE80211_DEFAULT_AQL_TXQ_LIMIT_MC; for (i = 0; i < IEEE80211_NUM_ACS; i++) { INIT_LIST_HEAD(&local->active_txqs[i]); spin_lock_init(&local->active_txq_lock[i]); diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c index c03aca57eb73..d12aed9c1756 100644 --- a/net/mac80211/sta_info.c +++ b/net/mac80211/sta_info.c @@ -2491,13 +2491,28 @@ EXPORT_SYMBOL(ieee80211_sta_recalc_aggregates); void ieee80211_sta_update_pending_airtime(struct ieee80211_local *local, struct sta_info *sta, u8 ac, - u16 tx_airtime, bool tx_completed) + u16 tx_airtime, bool tx_completed, + bool mcast) { int tx_pending; if (!wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) return; + if (mcast) { + if (!tx_completed) { + atomic_add(tx_airtime, &local->aql_mc_pending_airtime); + return; + } + + tx_pending = atomic_sub_return(tx_airtime, + &local->aql_mc_pending_airtime); + if (tx_pending < 0) + atomic_cmpxchg(&local->aql_mc_pending_airtime, + tx_pending, 0); + return; + } + if (!tx_completed) { if (sta) atomic_add(tx_airtime, diff --git a/net/mac80211/sta_info.h b/net/mac80211/sta_info.h index 5da3142d8516..ee0d32877c5b 100644 --- a/net/mac80211/sta_info.h +++ b/net/mac80211/sta_info.h @@ -147,7 +147,8 @@ struct airtime_info { void ieee80211_sta_update_pending_airtime(struct ieee80211_local *local, struct sta_info *sta, u8 ac, - u16 tx_airtime, bool tx_completed); + u16 tx_airtime, bool tx_completed, + bool mcast); struct sta_info; diff --git a/net/mac80211/status.c b/net/mac80211/status.c index d635490f59d3..3d811f652603 100644 --- a/net/mac80211/status.c +++ b/net/mac80211/status.c @@ -777,7 +777,7 @@ static void ieee80211_report_used_skb(struct ieee80211_local *local, ieee80211_sta_update_pending_airtime(local, sta, skb_get_queue_mapping(skb), tx_time_est, - true); + true, info->tx_time_mc); rcu_read_unlock(); } @@ -1189,10 +1189,11 @@ void ieee80211_tx_status_ext(struct ieee80211_hw *hw, /* Do this here to avoid the expensive lookup of the sta * in ieee80211_report_used_skb(). */ + bool mcast = IEEE80211_SKB_CB(skb)->tx_time_mc; ieee80211_sta_update_pending_airtime(local, sta, skb_get_queue_mapping(skb), tx_time_est, - true); + true, mcast); ieee80211_info_set_tx_time_est(IEEE80211_SKB_CB(skb), 0); } diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c index 76489129e9a1..49ef9494cc94 100644 --- a/net/mac80211/tx.c +++ b/net/mac80211/tx.c @@ -4064,20 +4064,20 @@ struct sk_buff *ieee80211_tx_dequeue(struct ieee80211_hw *hw, encap_out: info->control.vif = vif; - if (tx.sta && - wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) { - bool ampdu = txq->ac != IEEE80211_AC_VO; + if (wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) { + bool ampdu = txq->sta && txq->ac != IEEE80211_AC_VO; u32 airtime; airtime = ieee80211_calc_expected_tx_airtime(hw, vif, txq->sta, skb->len, ampdu); - if (airtime) { - airtime = ieee80211_info_set_tx_time_est(info, airtime); - ieee80211_sta_update_pending_airtime(local, tx.sta, - txq->ac, - airtime, - false); - } + if (!airtime) + return skb; + + airtime = ieee80211_info_set_tx_time_est(info, airtime); + info->tx_time_mc = !tx.sta; + ieee80211_sta_update_pending_airtime(local, tx.sta, txq->ac, + airtime, false, + info->tx_time_mc); } return skb; @@ -4129,6 +4129,7 @@ struct ieee80211_txq *ieee80211_next_txq(struct ieee80211_hw *hw, u8 ac) struct ieee80211_txq *ret = NULL; struct txq_info *txqi = NULL, *head = NULL; bool found_eligible_txq = false; + bool aql_check; spin_lock_bh(&local->active_txq_lock[ac]); @@ -4152,26 +4153,28 @@ struct ieee80211_txq *ieee80211_next_txq(struct ieee80211_hw *hw, u8 ac) if (!head) head = txqi; + aql_check = ieee80211_txq_airtime_check(hw, &txqi->txq); + if (aql_check) + found_eligible_txq = true; + if (txqi->txq.sta) { struct sta_info *sta = container_of(txqi->txq.sta, struct sta_info, sta); - bool aql_check = ieee80211_txq_airtime_check(hw, &txqi->txq); - s32 deficit = ieee80211_sta_deficit(sta, txqi->txq.ac); - if (aql_check) - found_eligible_txq = true; - - if (deficit < 0) + if (ieee80211_sta_deficit(sta, txqi->txq.ac) < 0) { sta->airtime[txqi->txq.ac].deficit += sta->airtime_weight; - if (deficit < 0 || !aql_check) { - list_move_tail(&txqi->schedule_order, - &local->active_txqs[txqi->txq.ac]); - goto begin; + aql_check = false; } } + if (!aql_check) { + list_move_tail(&txqi->schedule_order, + &local->active_txqs[txqi->txq.ac]); + goto begin; + } + if (txqi->schedule_round == local->schedule_round[ac]) goto out; @@ -4238,7 +4241,8 @@ bool ieee80211_txq_airtime_check(struct ieee80211_hw *hw, return true; if (!txq->sta) - return true; + return atomic_read(&local->aql_mc_pending_airtime) < + local->aql_txq_limit_mc; if (unlikely(txq->tid == IEEE80211_NUM_TIDS)) return true; @@ -4287,15 +4291,15 @@ bool ieee80211_txq_may_transmit(struct ieee80211_hw *hw, spin_lock_bh(&local->active_txq_lock[ac]); - if (!txqi->txq.sta) - goto out; - if (list_empty(&txqi->schedule_order)) goto out; if (!ieee80211_txq_schedule_airtime_check(local, ac)) goto out; + if (!txqi->txq.sta) + goto out; + list_for_each_entry_safe(iter, tmp, &local->active_txqs[ac], schedule_order) { if (iter == txqi)