mac80211: add improved HW queue control

mac80211 currently only supports one hardware queue
per AC. This is already problematic for off-channel
uses since if we go off channel while the BE queue
is full and then try to send an off-channel frame
the frame will never go out. This will become worse
when we support multi-channel since then a queue on
one channel might be full, but we have to stop the
software queue for all channels. That is obviously
not desirable.

To address this problem allow drivers to register
more hardware queues, and allow them to map them to
virtual interfaces. When they stop a hardware queue
the corresponding AC software queues on the correct
interfaces will be stopped as well. Additionally,
there's an off-channel queue to solve that problem
and a per-interface after-DTIM beacon queue. This
allows drivers to manage software queues closer to
how the hardware works.

Currently, there's a limit of 16 hardware queues.
This may or may not be sufficient, we can adjust it
as needed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

7 files changed, 162 insertions, 38 deletions

diff --git a/net/mac80211/agg-tx.c b/net/mac80211/agg-tx.c
index 9628a1892441..5b7053c58732 100644
--- a/net/mac80211/agg-tx.c
+++ b/net/mac80211/agg-tx.c
@@ -286,25 +286,25 @@ static inline int ieee80211_ac_from_tid(int tid)
  * a global "agg_queue_stop" refcount.
  */
 static void __acquires(agg_queue)
-ieee80211_stop_queue_agg(struct ieee80211_local *local, int tid)
+ieee80211_stop_queue_agg(struct ieee80211_sub_if_data *sdata, int tid)
 {
-	int queue = ieee80211_ac_from_tid(tid);
+	int queue = sdata->vif.hw_queue[ieee80211_ac_from_tid(tid)];
 
-	if (atomic_inc_return(&local->agg_queue_stop[queue]) == 1)
+	if (atomic_inc_return(&sdata->local->agg_queue_stop[queue]) == 1)
 		ieee80211_stop_queue_by_reason(
-			&local->hw, queue,
+			&sdata->local->hw, queue,
 			IEEE80211_QUEUE_STOP_REASON_AGGREGATION);
 	__acquire(agg_queue);
 }
 
 static void __releases(agg_queue)
-ieee80211_wake_queue_agg(struct ieee80211_local *local, int tid)
+ieee80211_wake_queue_agg(struct ieee80211_sub_if_data *sdata, int tid)
 {
-	int queue = ieee80211_ac_from_tid(tid);
+	int queue = sdata->vif.hw_queue[ieee80211_ac_from_tid(tid)];
 
-	if (atomic_dec_return(&local->agg_queue_stop[queue]) == 0)
+	if (atomic_dec_return(&sdata->local->agg_queue_stop[queue]) == 0)
 		ieee80211_wake_queue_by_reason(
-			&local->hw, queue,
+			&sdata->local->hw, queue,
 			IEEE80211_QUEUE_STOP_REASON_AGGREGATION);
 	__release(agg_queue);
 }
@@ -314,13 +314,14 @@ ieee80211_wake_queue_agg(struct ieee80211_local *local, int tid)
  * requires a call to ieee80211_agg_splice_finish later
  */
 static void __acquires(agg_queue)
-ieee80211_agg_splice_packets(struct ieee80211_local *local,
+ieee80211_agg_splice_packets(struct ieee80211_sub_if_data *sdata,
 			     struct tid_ampdu_tx *tid_tx, u16 tid)
 {
-	int queue = ieee80211_ac_from_tid(tid);
+	struct ieee80211_local *local = sdata->local;
+	int queue = sdata->vif.hw_queue[ieee80211_ac_from_tid(tid)];
 	unsigned long flags;
 
-	ieee80211_stop_queue_agg(local, tid);
+	ieee80211_stop_queue_agg(sdata, tid);
 
 	if (WARN(!tid_tx, "TID %d gone but expected when splicing aggregates"
 			  " from the pending queue\n", tid))
@@ -336,9 +337,9 @@ ieee80211_agg_splice_packets(struct ieee80211_local *local,
 }
 
 static void __releases(agg_queue)
-ieee80211_agg_splice_finish(struct ieee80211_local *local, u16 tid)
+ieee80211_agg_splice_finish(struct ieee80211_sub_if_data *sdata, u16 tid)
 {
-	ieee80211_wake_queue_agg(local, tid);
+	ieee80211_wake_queue_agg(sdata, tid);
 }
 
 void ieee80211_tx_ba_session_handle_start(struct sta_info *sta, int tid)
@@ -376,9 +377,9 @@ void ieee80211_tx_ba_session_handle_start(struct sta_info *sta, int tid)
 					" tid %d\n", tid);
 #endif
 		spin_lock_bh(&sta->lock);
-		ieee80211_agg_splice_packets(local, tid_tx, tid);
+		ieee80211_agg_splice_packets(sdata, tid_tx, tid);
 		ieee80211_assign_tid_tx(sta, tid, NULL);
-		ieee80211_agg_splice_finish(local, tid);
+		ieee80211_agg_splice_finish(sdata, tid);
 		spin_unlock_bh(&sta->lock);
 
 		kfree_rcu(tid_tx, rcu_head);
@@ -598,14 +599,14 @@ static void ieee80211_agg_tx_operational(struct ieee80211_local *local,
 	 */
 	spin_lock_bh(&sta->lock);
 
-	ieee80211_agg_splice_packets(local, tid_tx, tid);
+	ieee80211_agg_splice_packets(sta->sdata, tid_tx, tid);
 	/*
 	 * Now mark as operational. This will be visible
 	 * in the TX path, and lets it go lock-free in
 	 * the common case.
 	 */
 	set_bit(HT_AGG_STATE_OPERATIONAL, &tid_tx->state);
-	ieee80211_agg_splice_finish(local, tid);
+	ieee80211_agg_splice_finish(sta->sdata, tid);
 
 	spin_unlock_bh(&sta->lock);
 }
@@ -790,12 +791,12 @@ void ieee80211_stop_tx_ba_cb(struct ieee80211_vif *vif, u8 *ra, u8 tid)
 	 * more.
 	 */
 
-	ieee80211_agg_splice_packets(local, tid_tx, tid);
+	ieee80211_agg_splice_packets(sta->sdata, tid_tx, tid);
 
 	/* future packets must not find the tid_tx struct any more */
 	ieee80211_assign_tid_tx(sta, tid, NULL);
 
-	ieee80211_agg_splice_finish(local, tid);
+	ieee80211_agg_splice_finish(sta->sdata, tid);
 
 	kfree_rcu(tid_tx, rcu_head);
 
diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c
index 667d93943399..d6163b98f7b7 100644
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -2106,6 +2106,10 @@ static int ieee80211_mgmt_tx(struct wiphy *wiphy, struct net_device *dev,
 
 	IEEE80211_SKB_CB(skb)->flags = flags;
 
+	if (flags & IEEE80211_TX_CTL_TX_OFFCHAN)
+		IEEE80211_SKB_CB(skb)->hw_queue =
+			local->hw.offchannel_tx_hw_queue;
+
 	skb->dev = sdata->dev;
 
 	*cookie = (unsigned long) skb;
@@ -2147,6 +2151,8 @@ static int ieee80211_mgmt_tx(struct wiphy *wiphy, struct net_device *dev,
 		/* modify cookie to prevent API mismatches */
 		*cookie ^= 2;
 		IEEE80211_SKB_CB(skb)->flags |= IEEE80211_TX_CTL_TX_OFFCHAN;
+		IEEE80211_SKB_CB(skb)->hw_queue =
+			local->hw.offchannel_tx_hw_queue;
 		local->hw_roc_skb = skb;
 		local->hw_roc_skb_for_status = skb;
 		mutex_unlock(&local->mtx);
diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h
index 8ed074f7e6cd..4be11ea3dfc4 100644
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -1449,6 +1449,7 @@ void ieee80211_wake_queue_by_reason(struct ieee80211_hw *hw, int queue,
 				    enum queue_stop_reason reason);
 void ieee80211_stop_queue_by_reason(struct ieee80211_hw *hw, int queue,
 				    enum queue_stop_reason reason);
+void ieee80211_propagate_queue_wake(struct ieee80211_local *local, int queue);
 void ieee80211_add_pending_skb(struct ieee80211_local *local,
 			       struct sk_buff *skb);
 void ieee80211_add_pending_skbs_fn(struct ieee80211_local *local,
diff --git a/net/mac80211/iface.c b/net/mac80211/iface.c
index 2b88cb278fc4..ed297649c577 100644
--- a/net/mac80211/iface.c
+++ b/net/mac80211/iface.c
@@ -149,6 +149,34 @@ static int ieee80211_check_concurrent_iface(struct ieee80211_sub_if_data *sdata,
 	return 0;
 }
 
+static int ieee80211_check_queues(struct ieee80211_sub_if_data *sdata)
+{
+	int n_queues = sdata->local->hw.queues;
+	int i;
+
+	for (i = 0; i < IEEE80211_NUM_ACS; i++) {
+		if (WARN_ON_ONCE(sdata->vif.hw_queue[i] ==
+				 IEEE80211_INVAL_HW_QUEUE))
+			return -EINVAL;
+		if (WARN_ON_ONCE(sdata->vif.hw_queue[i] >=
+				 n_queues))
+			return -EINVAL;
+	}
+
+	if (sdata->vif.type != NL80211_IFTYPE_AP) {
+		sdata->vif.cab_queue = IEEE80211_INVAL_HW_QUEUE;
+		return 0;
+	}
+
+	if (WARN_ON_ONCE(sdata->vif.cab_queue == IEEE80211_INVAL_HW_QUEUE))
+		return -EINVAL;
+
+	if (WARN_ON_ONCE(sdata->vif.cab_queue >= n_queues))
+		return -EINVAL;
+
+	return 0;
+}
+
 void ieee80211_adjust_monitor_flags(struct ieee80211_sub_if_data *sdata,
 				    const int offset)
 {
@@ -169,6 +197,20 @@ void ieee80211_adjust_monitor_flags(struct ieee80211_sub_if_data *sdata,
 #undef ADJUST
 }
 
+static void ieee80211_set_default_queues(struct ieee80211_sub_if_data *sdata)
+{
+	struct ieee80211_local *local = sdata->local;
+	int i;
+
+	for (i = 0; i < IEEE80211_NUM_ACS; i++) {
+		if (local->hw.flags & IEEE80211_HW_QUEUE_CONTROL)
+			sdata->vif.hw_queue[i] = IEEE80211_INVAL_HW_QUEUE;
+		else
+			sdata->vif.hw_queue[i] = i;
+	}
+	sdata->vif.cab_queue = IEEE80211_INVAL_HW_QUEUE;
+}
+
 static int ieee80211_add_virtual_monitor(struct ieee80211_local *local)
 {
 	struct ieee80211_sub_if_data *sdata;
@@ -190,6 +232,8 @@ static int ieee80211_add_virtual_monitor(struct ieee80211_local *local)
 	snprintf(sdata->name, IFNAMSIZ, "%s-monitor",
 		 wiphy_name(local->hw.wiphy));
 
+	ieee80211_set_default_queues(sdata);
+
 	ret = drv_add_interface(local, sdata);
 	if (WARN_ON(ret)) {
 		/* ok .. stupid driver, it asked for this! */
@@ -197,6 +241,12 @@ static int ieee80211_add_virtual_monitor(struct ieee80211_local *local)
 		return ret;
 	}
 
+	ret = ieee80211_check_queues(sdata);
+	if (ret) {
+		kfree(sdata);
+		return ret;
+	}
+
 	rcu_assign_pointer(local->monitor_sdata, sdata);
 
 	return 0;
@@ -344,6 +394,9 @@ static int ieee80211_do_open(struct net_device *dev, bool coming_up)
 			res = drv_add_interface(local, sdata);
 			if (res)
 				goto err_stop;
+			res = ieee80211_check_queues(sdata);
+			if (res)
+				goto err_del_interface;
 		}
 
 		if (sdata->vif.type == NL80211_IFTYPE_AP) {
@@ -1040,6 +1093,13 @@ static int ieee80211_runtime_change_iftype(struct ieee80211_sub_if_data *sdata,
 	if (ret)
 		type = sdata->vif.type;
 
+	/*
+	 * Ignore return value here, there's not much we can do since
+	 * the driver changed the interface type internally already.
+	 * The warnings will hopefully make driver authors fix it :-)
+	 */
+	ieee80211_check_queues(sdata);
+
 	ieee80211_setup_sdata(sdata, type);
 
 	err = ieee80211_do_open(sdata->dev, false);
@@ -1266,6 +1326,8 @@ int ieee80211_if_add(struct ieee80211_local *local, const char *name,
 			       sizeof(sdata->rc_rateidx_mcs_mask[i]));
 	}
 
+	ieee80211_set_default_queues(sdata);
+
 	/* setup type-dependent data */
 	ieee80211_setup_sdata(sdata, type);
 
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index d019f0d3a0fe..ac79d5e8e0d0 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -591,6 +591,7 @@ struct ieee80211_hw *ieee80211_alloc_hw(size_t priv_data_len,
 	local->hw.max_report_rates = 0;
 	local->hw.max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF;
 	local->hw.max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF;
+	local->hw.offchannel_tx_hw_queue = IEEE80211_INVAL_HW_QUEUE;
 	local->hw.conf.long_frame_max_tx_count = wiphy->retry_long;
 	local->hw.conf.short_frame_max_tx_count = wiphy->retry_short;
 	local->user_power_level = -1;
@@ -687,6 +688,11 @@ int ieee80211_register_hw(struct ieee80211_hw *hw)
 		WLAN_CIPHER_SUITE_AES_CMAC
 	};
 
+	if (hw->flags & IEEE80211_HW_QUEUE_CONTROL &&
+	    (local->hw.offchannel_tx_hw_queue == IEEE80211_INVAL_HW_QUEUE ||
+	     local->hw.offchannel_tx_hw_queue >= local->hw.queues))
+		return -EINVAL;
+
 	if ((hw->wiphy->wowlan.flags || hw->wiphy->wowlan.n_patterns)
 #ifdef CONFIG_PM
 	    && (!local->ops->suspend || !local->ops->resume)
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index a8d0188ab408..4f6aac16ac3a 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -400,6 +400,8 @@ ieee80211_tx_h_multicast_ps_buf(struct ieee80211_tx_data *tx)
 		return TX_CONTINUE;
 
 	info->flags |= IEEE80211_TX_CTL_SEND_AFTER_DTIM;
+	if (tx->local->hw.flags & IEEE80211_HW_QUEUE_CONTROL)
+		info->hw_queue = tx->sdata->vif.cab_queue;
 
 	/* device releases frame after DTIM beacon */
 	if (!(tx->local->hw.flags & IEEE80211_HW_HOST_BROADCAST_PS_BUFFERING))
@@ -1214,11 +1216,19 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local,
 			       bool txpending)
 {
 	struct sk_buff *skb, *tmp;
-	struct ieee80211_tx_info *info;
 	unsigned long flags;
 
 	skb_queue_walk_safe(skbs, skb, tmp) {
-		int q = skb_get_queue_mapping(skb);
+		struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
+		int q = info->hw_queue;
+
+#ifdef CONFIG_MAC80211_VERBOSE_DEBUG
+		if (WARN_ON_ONCE(q >= local->hw.queues)) {
+			__skb_unlink(skb, skbs);
+			dev_kfree_skb(skb);
+			continue;
+		}
+#endif
 
 		spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
 		if (local->queue_stop_reasons[q] ||
@@ -1240,7 +1250,6 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local,
 		}
 		spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
 
-		info = IEEE80211_SKB_CB(skb);
 		info->control.vif = vif;
 		info->control.sta = sta;
 
@@ -1284,9 +1293,14 @@ static bool __ieee80211_tx(struct ieee80211_local *local,
 	switch (sdata->vif.type) {
 	case NL80211_IFTYPE_MONITOR:
 		sdata = rcu_dereference(local->monitor_sdata);
-		if (sdata)
+		if (sdata) {
 			vif = &sdata->vif;
-		else
+			info->hw_queue =
+				vif->hw_queue[skb_get_queue_mapping(skb)];
+		} else if (local->hw.flags & IEEE80211_HW_QUEUE_CONTROL) {
+			dev_kfree_skb(skb);
+			return true;
+		} else
 			vif = NULL;
 		break;
 	case NL80211_IFTYPE_AP_VLAN:
@@ -1402,6 +1416,12 @@ static bool ieee80211_tx(struct ieee80211_sub_if_data *sdata,
 	tx.channel = local->hw.conf.channel;
 	info->band = tx.channel->band;
 
+	/* set up hw_queue value early */
+	if (!(info->flags & IEEE80211_TX_CTL_TX_OFFCHAN) ||
+	    !(local->hw.flags & IEEE80211_HW_QUEUE_CONTROL))
+		info->hw_queue =
+			sdata->vif.hw_queue[skb_get_queue_mapping(skb)];
+
 	if (!invoke_tx_handlers(&tx))
 		result = __ieee80211_tx(local, &tx.skbs, led_len,
 					tx.sta, txpending);
@@ -2172,7 +2192,6 @@ static bool ieee80211_tx_pending_skb(struct ieee80211_local *local,
 void ieee80211_tx_pending(unsigned long data)
 {
 	struct ieee80211_local *local = (struct ieee80211_local *)data;
-	struct ieee80211_sub_if_data *sdata;
 	unsigned long flags;
 	int i;
 	bool txok;
@@ -2209,8 +2228,7 @@ void ieee80211_tx_pending(unsigned long data)
 		}
 
 		if (skb_queue_empty(&local->pending[i]))
-			list_for_each_entry_rcu(sdata, &local->interfaces, list)
-				netif_wake_subqueue(sdata->dev, i);
+			ieee80211_propagate_queue_wake(local, i);
 	}
 	spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
 
@@ -2717,11 +2735,13 @@ EXPORT_SYMBOL(ieee80211_get_buffered_bc);
 void ieee80211_tx_skb_tid(struct ieee80211_sub_if_data *sdata,
 			  struct sk_buff *skb, int tid)
 {
+	int ac = ieee802_1d_to_ac[tid];
+
 	skb_set_mac_header(skb, 0);
 	skb_set_network_header(skb, 0);
 	skb_set_transport_header(skb, 0);
 
-	skb_set_queue_mapping(skb, ieee802_1d_to_ac[tid]);
+	skb_set_queue_mapping(skb, ac);
 	skb->priority = tid;
 
 	/*
diff --git a/net/mac80211/util.c b/net/mac80211/util.c
index 9e8f4b892555..e67fe5c1def9 100644
--- a/net/mac80211/util.c
+++ b/net/mac80211/util.c
@@ -265,11 +265,36 @@ __le16 ieee80211_ctstoself_duration(struct ieee80211_hw *hw,
 }
 EXPORT_SYMBOL(ieee80211_ctstoself_duration);
 
+void ieee80211_propagate_queue_wake(struct ieee80211_local *local, int queue)
+{
+	struct ieee80211_sub_if_data *sdata;
+
+	list_for_each_entry_rcu(sdata, &local->interfaces, list) {
+		int ac;
+
+		if (test_bit(SDATA_STATE_OFFCHANNEL, &sdata->state))
+			continue;
+
+		if (sdata->vif.cab_queue != IEEE80211_INVAL_HW_QUEUE &&
+		    local->queue_stop_reasons[sdata->vif.cab_queue] != 0)
+			continue;
+
+		for (ac = 0; ac < IEEE80211_NUM_ACS; ac++) {
+			int ac_queue = sdata->vif.hw_queue[ac];
+
+			if (ac_queue == queue ||
+			    (sdata->vif.cab_queue == queue &&
+			     local->queue_stop_reasons[ac_queue] == 0 &&
+			     skb_queue_empty(&local->pending[ac_queue])))
+				netif_wake_subqueue(sdata->dev, ac);
+		}
+	}
+}
+
 static void __ieee80211_wake_queue(struct ieee80211_hw *hw, int queue,
 				   enum queue_stop_reason reason)
 {
 	struct ieee80211_local *local = hw_to_local(hw);
-	struct ieee80211_sub_if_data *sdata;
 
 	trace_wake_queue(local, queue, reason);
 
@@ -287,11 +312,7 @@ static void __ieee80211_wake_queue(struct ieee80211_hw *hw, int queue,
 
 	if (skb_queue_empty(&local->pending[queue])) {
 		rcu_read_lock();
-		list_for_each_entry_rcu(sdata, &local->interfaces, list) {
-			if (test_bit(SDATA_STATE_OFFCHANNEL, &sdata->state))
-				continue;
-			netif_wake_subqueue(sdata->dev, queue);
-		}
+		ieee80211_propagate_queue_wake(local, queue);
 		rcu_read_unlock();
 	} else
 		tasklet_schedule(&local->tx_pending_tasklet);
@@ -332,8 +353,15 @@ static void __ieee80211_stop_queue(struct ieee80211_hw *hw, int queue,
 	__set_bit(reason, &local->queue_stop_reasons[queue]);
 
 	rcu_read_lock();
-	list_for_each_entry_rcu(sdata, &local->interfaces, list)
-		netif_stop_subqueue(sdata->dev, queue);
+	list_for_each_entry_rcu(sdata, &local->interfaces, list) {
+		int ac;
+
+		for (ac = 0; ac < IEEE80211_NUM_ACS; ac++) {
+			if (sdata->vif.hw_queue[ac] == queue ||
+			    sdata->vif.cab_queue == queue)
+				netif_stop_subqueue(sdata->dev, ac);
+		}
+	}
 	rcu_read_unlock();
 }
 
@@ -360,8 +388,8 @@ void ieee80211_add_pending_skb(struct ieee80211_local *local,
 {
 	struct ieee80211_hw *hw = &local->hw;
 	unsigned long flags;
-	int queue = skb_get_queue_mapping(skb);
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
+	int queue = info->hw_queue;
 
 	if (WARN_ON(!info->control.vif)) {
 		kfree_skb(skb);
@@ -393,7 +421,7 @@ void ieee80211_add_pending_skbs_fn(struct ieee80211_local *local,
 			continue;
 		}
 
-		queue = skb_get_queue_mapping(skb);
+		queue = info->hw_queue;
 
 		__ieee80211_stop_queue(hw, queue,
 				IEEE80211_QUEUE_STOP_REASON_SKB_ADD);