1 files changed, 341 insertions, 238 deletions

diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 457238a2f3f..3fb04a86444 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -34,8 +34,7 @@
 
 #define IEEE80211_TX_OK		0
 #define IEEE80211_TX_AGAIN	1
-#define IEEE80211_TX_FRAG_AGAIN	2
-#define IEEE80211_TX_PENDING	3
+#define IEEE80211_TX_PENDING	2
 
 /* misc utils */
 
@@ -193,7 +192,19 @@ ieee80211_tx_h_check_assoc(struct ieee80211_tx_data *tx)
 		return TX_CONTINUE;
 
 	if (unlikely(tx->local->sw_scanning) &&
-	    !ieee80211_is_probe_req(hdr->frame_control))
+	    !ieee80211_is_probe_req(hdr->frame_control) &&
+	    !ieee80211_is_nullfunc(hdr->frame_control))
+		/*
+		 * When software scanning only nullfunc frames (to notify
+		 * the sleep state to the AP) and probe requests (for the
+		 * active scan) are allowed, all other frames should not be
+		 * sent and we should not get here, but if we do
+		 * nonetheless, drop them to avoid sending them
+		 * off-channel. See the link below and
+		 * ieee80211_start_scan() for more.
+		 *
+		 * http://article.gmane.org/gmane.linux.kernel.wireless.general/30089
+		 */
 		return TX_DROP;
 
 	if (tx->sdata->vif.type == NL80211_IFTYPE_MESH_POINT)
@@ -690,17 +701,62 @@ ieee80211_tx_h_sequence(struct ieee80211_tx_data *tx)
 	return TX_CONTINUE;
 }
 
+static int ieee80211_fragment(struct ieee80211_local *local,
+			      struct sk_buff *skb, int hdrlen,
+			      int frag_threshold)
+{
+	struct sk_buff *tail = skb, *tmp;
+	int per_fragm = frag_threshold - hdrlen - FCS_LEN;
+	int pos = hdrlen + per_fragm;
+	int rem = skb->len - hdrlen - per_fragm;
+
+	if (WARN_ON(rem < 0))
+		return -EINVAL;
+
+	while (rem) {
+		int fraglen = per_fragm;
+
+		if (fraglen > rem)
+			fraglen = rem;
+		rem -= fraglen;
+		tmp = dev_alloc_skb(local->tx_headroom +
+				    frag_threshold +
+				    IEEE80211_ENCRYPT_HEADROOM +
+				    IEEE80211_ENCRYPT_TAILROOM);
+		if (!tmp)
+			return -ENOMEM;
+		tail->next = tmp;
+		tail = tmp;
+		skb_reserve(tmp, local->tx_headroom +
+				 IEEE80211_ENCRYPT_HEADROOM);
+		/* copy control information */
+		memcpy(tmp->cb, skb->cb, sizeof(tmp->cb));
+		skb_copy_queue_mapping(tmp, skb);
+		tmp->priority = skb->priority;
+		tmp->do_not_encrypt = skb->do_not_encrypt;
+		tmp->dev = skb->dev;
+		tmp->iif = skb->iif;
+
+		/* copy header and data */
+		memcpy(skb_put(tmp, hdrlen), skb->data, hdrlen);
+		memcpy(skb_put(tmp, fraglen), skb->data + pos, fraglen);
+
+		pos += fraglen;
+	}
+
+	skb->len = hdrlen + per_fragm;
+	return 0;
+}
+
 static ieee80211_tx_result debug_noinline
 ieee80211_tx_h_fragment(struct ieee80211_tx_data *tx)
 {
-	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(tx->skb);
-	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)tx->skb->data;
-	size_t hdrlen, per_fragm, num_fragm, payload_len, left;
-	struct sk_buff **frags, *first, *frag;
-	int i;
-	u16 seq;
-	u8 *pos;
+	struct sk_buff *skb = tx->skb;
+	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
+	struct ieee80211_hdr *hdr = (void *)skb->data;
 	int frag_threshold = tx->local->fragmentation_threshold;
+	int hdrlen;
+	int fragnum;
 
 	if (!(tx->flags & IEEE80211_TX_FRAGMENTED))
 		return TX_CONTINUE;
@@ -713,58 +769,35 @@ ieee80211_tx_h_fragment(struct ieee80211_tx_data *tx)
 	if (WARN_ON(info->flags & IEEE80211_TX_CTL_AMPDU))
 		return TX_DROP;
 
-	first = tx->skb;
-
 	hdrlen = ieee80211_hdrlen(hdr->frame_control);
-	payload_len = first->len - hdrlen;
-	per_fragm = frag_threshold - hdrlen - FCS_LEN;
-	num_fragm = DIV_ROUND_UP(payload_len, per_fragm);
-
-	frags = kzalloc(num_fragm * sizeof(struct sk_buff *), GFP_ATOMIC);
-	if (!frags)
-		goto fail;
-
-	hdr->frame_control |= cpu_to_le16(IEEE80211_FCTL_MOREFRAGS);
-	seq = le16_to_cpu(hdr->seq_ctrl) & IEEE80211_SCTL_SEQ;
-	pos = first->data + hdrlen + per_fragm;
-	left = payload_len - per_fragm;
-	for (i = 0; i < num_fragm - 1; i++) {
-		struct ieee80211_hdr *fhdr;
-		size_t copylen;
-
-		if (left <= 0)
-			goto fail;
 
-		/* reserve enough extra head and tail room for possible
-		 * encryption */
-		frag = frags[i] =
-			dev_alloc_skb(tx->local->tx_headroom +
-				      frag_threshold +
-				      IEEE80211_ENCRYPT_HEADROOM +
-				      IEEE80211_ENCRYPT_TAILROOM);
-		if (!frag)
-			goto fail;
-
-		/* Make sure that all fragments use the same priority so
-		 * that they end up using the same TX queue */
-		frag->priority = first->priority;
+	/* internal error, why is TX_FRAGMENTED set? */
+	if (WARN_ON(skb->len <= frag_threshold))
+		return TX_DROP;
 
-		skb_reserve(frag, tx->local->tx_headroom +
-				  IEEE80211_ENCRYPT_HEADROOM);
+	/*
+	 * Now fragment the frame. This will allocate all the fragments and
+	 * chain them (using skb as the first fragment) to skb->next.
+	 * During transmission, we will remove the successfully transmitted
+	 * fragments from this list. When the low-level driver rejects one
+	 * of the fragments then we will simply pretend to accept the skb
+	 * but store it away as pending.
+	 */
+	if (ieee80211_fragment(tx->local, skb, hdrlen, frag_threshold))
+		return TX_DROP;
 
-		/* copy TX information */
-		info = IEEE80211_SKB_CB(frag);
-		memcpy(info, first->cb, sizeof(frag->cb));
+	/* update duration/seq/flags of fragments */
+	fragnum = 0;
+	do {
+		int next_len;
+		const __le16 morefrags = cpu_to_le16(IEEE80211_FCTL_MOREFRAGS);
 
-		/* copy/fill in 802.11 header */
-		fhdr = (struct ieee80211_hdr *) skb_put(frag, hdrlen);
-		memcpy(fhdr, first->data, hdrlen);
-		fhdr->seq_ctrl = cpu_to_le16(seq | ((i + 1) & IEEE80211_SCTL_FRAG));
+		hdr = (void *)skb->data;
+		info = IEEE80211_SKB_CB(skb);
 
-		if (i == num_fragm - 2) {
-			/* clear MOREFRAGS bit for the last fragment */
-			fhdr->frame_control &= cpu_to_le16(~IEEE80211_FCTL_MOREFRAGS);
-		} else {
+		if (skb->next) {
+			hdr->frame_control |= morefrags;
+			next_len = skb->next->len;
 			/*
 			 * No multi-rate retries for fragmented frames, that
 			 * would completely throw off the NAV at other STAs.
@@ -775,37 +808,16 @@ ieee80211_tx_h_fragment(struct ieee80211_tx_data *tx)
 			info->control.rates[4].idx = -1;
 			BUILD_BUG_ON(IEEE80211_TX_MAX_RATES != 5);
 			info->flags &= ~IEEE80211_TX_CTL_RATE_CTRL_PROBE;
+		} else {
+			hdr->frame_control &= ~morefrags;
+			next_len = 0;
 		}
-
-		/* copy data */
-		copylen = left > per_fragm ? per_fragm : left;
-		memcpy(skb_put(frag, copylen), pos, copylen);
-
-		skb_copy_queue_mapping(frag, first);
-
-		frag->do_not_encrypt = first->do_not_encrypt;
-		frag->dev = first->dev;
-		frag->iif = first->iif;
-
-		pos += copylen;
-		left -= copylen;
-	}
-	skb_trim(first, hdrlen + per_fragm);
-
-	tx->num_extra_frag = num_fragm - 1;
-	tx->extra_frag = frags;
+		hdr->duration_id = ieee80211_duration(tx, 0, next_len);
+		hdr->seq_ctrl |= cpu_to_le16(fragnum & IEEE80211_SCTL_FRAG);
+		fragnum++;
+	} while ((skb = skb->next));
 
 	return TX_CONTINUE;
-
- fail:
-	if (frags) {
-		for (i = 0; i < num_fragm - 1; i++)
-			if (frags[i])
-				dev_kfree_skb(frags[i]);
-		kfree(frags);
-	}
-	I802_DEBUG_INC(tx->local->tx_handlers_drop_fragment);
-	return TX_DROP;
 }
 
 static ieee80211_tx_result debug_noinline
@@ -833,27 +845,19 @@ ieee80211_tx_h_encrypt(struct ieee80211_tx_data *tx)
 static ieee80211_tx_result debug_noinline
 ieee80211_tx_h_calculate_duration(struct ieee80211_tx_data *tx)
 {
-	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)tx->skb->data;
-	int next_len, i;
-	int group_addr = is_multicast_ether_addr(hdr->addr1);
-
-	if (!(tx->flags & IEEE80211_TX_FRAGMENTED)) {
-		hdr->duration_id = ieee80211_duration(tx, group_addr, 0);
-		return TX_CONTINUE;
-	}
+	struct sk_buff *skb = tx->skb;
+	struct ieee80211_hdr *hdr;
+	int next_len;
+	bool group_addr;
 
-	hdr->duration_id = ieee80211_duration(tx, group_addr,
-					      tx->extra_frag[0]->len);
+	do {
+		hdr = (void *) skb->data;
+		next_len = skb->next ? skb->next->len : 0;
+		group_addr = is_multicast_ether_addr(hdr->addr1);
 
-	for (i = 0; i < tx->num_extra_frag; i++) {
-		if (i + 1 < tx->num_extra_frag)
-			next_len = tx->extra_frag[i + 1]->len;
-		else
-			next_len = 0;
-
-		hdr = (struct ieee80211_hdr *)tx->extra_frag[i]->data;
-		hdr->duration_id = ieee80211_duration(tx, 0, next_len);
-	}
+		hdr->duration_id =
+			ieee80211_duration(tx, group_addr, next_len);
+	} while ((skb = skb->next));
 
 	return TX_CONTINUE;
 }
@@ -861,19 +865,16 @@ ieee80211_tx_h_calculate_duration(struct ieee80211_tx_data *tx)
 static ieee80211_tx_result debug_noinline
 ieee80211_tx_h_stats(struct ieee80211_tx_data *tx)
 {
-	int i;
+	struct sk_buff *skb = tx->skb;
 
 	if (!tx->sta)
 		return TX_CONTINUE;
 
 	tx->sta->tx_packets++;
-	tx->sta->tx_fragments++;
-	tx->sta->tx_bytes += tx->skb->len;
-	if (tx->extra_frag) {
-		tx->sta->tx_fragments += tx->num_extra_frag;
-		for (i = 0; i < tx->num_extra_frag; i++)
-			tx->sta->tx_bytes += tx->extra_frag[i]->len;
-	}
+	do {
+		tx->sta->tx_fragments++;
+		tx->sta->tx_bytes += skb->len;
+	} while ((skb = skb->next));
 
 	return TX_CONTINUE;
 }
@@ -983,9 +984,9 @@ __ieee80211_tx_prepare(struct ieee80211_tx_data *tx,
 	struct ieee80211_hdr *hdr;
 	struct ieee80211_sub_if_data *sdata;
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
-
 	int hdrlen, tid;
 	u8 *qc, *state;
+	bool queued = false;
 
 	memset(tx, 0, sizeof(*tx));
 	tx->skb = skb;
@@ -1012,25 +1013,53 @@ __ieee80211_tx_prepare(struct ieee80211_tx_data *tx,
 		 */
 	}
 
+	/*
+	 * If this flag is set to true anywhere, and we get here,
+	 * we are doing the needed processing, so remove the flag
+	 * now.
+	 */
+	info->flags &= ~IEEE80211_TX_INTFL_NEED_TXPROCESSING;
+
 	hdr = (struct ieee80211_hdr *) skb->data;
 
 	tx->sta = sta_info_get(local, hdr->addr1);
 
-	if (tx->sta && ieee80211_is_data_qos(hdr->frame_control)) {
+	if (tx->sta && ieee80211_is_data_qos(hdr->frame_control) &&
+	    (local->hw.flags & IEEE80211_HW_AMPDU_AGGREGATION)) {
 		unsigned long flags;
+		struct tid_ampdu_tx *tid_tx;
+
 		qc = ieee80211_get_qos_ctl(hdr);
 		tid = *qc & IEEE80211_QOS_CTL_TID_MASK;
 
 		spin_lock_irqsave(&tx->sta->lock, flags);
+		/*
+		 * XXX: This spinlock could be fairly expensive, but see the
+		 *	comment in agg-tx.c:ieee80211_agg_tx_operational().
+		 *	One way to solve this would be to do something RCU-like
+		 *	for managing the tid_tx struct and using atomic bitops
+		 *	for the actual state -- by introducing an actual
+		 *	'operational' bit that would be possible. It would
+		 *	require changing ieee80211_agg_tx_operational() to
+		 *	set that bit, and changing the way tid_tx is managed
+		 *	everywhere, including races between that bit and
+		 *	tid_tx going away (tid_tx being added can be easily
+		 *	committed to memory before the 'operational' bit).
+		 */
+		tid_tx = tx->sta->ampdu_mlme.tid_tx[tid];
 		state = &tx->sta->ampdu_mlme.tid_state_tx[tid];
 		if (*state == HT_AGG_STATE_OPERATIONAL) {
 			info->flags |= IEEE80211_TX_CTL_AMPDU;
-			if (local->hw.ampdu_queues)
-				skb_set_queue_mapping(
-					skb, tx->local->hw.queues +
-					     tx->sta->tid_to_tx_q[tid]);
+		} else if (*state != HT_AGG_STATE_IDLE) {
+			/* in progress */
+			queued = true;
+			info->flags |= IEEE80211_TX_INTFL_NEED_TXPROCESSING;
+			__skb_queue_tail(&tid_tx->pending, skb);
 		}
 		spin_unlock_irqrestore(&tx->sta->lock, flags);
+
+		if (unlikely(queued))
+			return TX_QUEUED;
 	}
 
 	if (is_multicast_ether_addr(hdr->addr1)) {
@@ -1081,51 +1110,55 @@ static int ieee80211_tx_prepare(struct ieee80211_local *local,
 	}
 	if (unlikely(!dev))
 		return -ENODEV;
-	/* initialises tx with control */
+	/*
+	 * initialises tx with control
+	 *
+	 * return value is safe to ignore here because this function
+	 * can only be invoked for multicast frames
+	 *
+	 * XXX: clean up
+	 */
 	__ieee80211_tx_prepare(tx, skb, dev);
 	dev_put(dev);
 	return 0;
 }
 
-static int __ieee80211_tx(struct ieee80211_local *local, struct sk_buff *skb,
-			  struct ieee80211_tx_data *tx)
+static int __ieee80211_tx(struct ieee80211_local *local,
+			  struct sk_buff **skbp,
+			  struct sta_info *sta)
 {
+	struct sk_buff *skb = *skbp, *next;
 	struct ieee80211_tx_info *info;
-	int ret, i;
+	int ret, len;
+	bool fragm = false;
 
-	if (skb) {
+	local->mdev->trans_start = jiffies;
+
+	while (skb) {
 		if (ieee80211_queue_stopped(&local->hw,
 					    skb_get_queue_mapping(skb)))
 			return IEEE80211_TX_PENDING;
 
-		ret = local->ops->tx(local_to_hw(local), skb);
-		if (ret)
-			return IEEE80211_TX_AGAIN;
-		local->mdev->trans_start = jiffies;
-		ieee80211_led_tx(local, 1);
-	}
-	if (tx->extra_frag) {
-		for (i = 0; i < tx->num_extra_frag; i++) {
-			if (!tx->extra_frag[i])
-				continue;
-			info = IEEE80211_SKB_CB(tx->extra_frag[i]);
+		info = IEEE80211_SKB_CB(skb);
+
+		if (fragm)
 			info->flags &= ~(IEEE80211_TX_CTL_CLEAR_PS_FILT |
 					 IEEE80211_TX_CTL_FIRST_FRAGMENT);
-			if (ieee80211_queue_stopped(&local->hw,
-					skb_get_queue_mapping(tx->extra_frag[i])))
-				return IEEE80211_TX_FRAG_AGAIN;
-
-			ret = local->ops->tx(local_to_hw(local),
-					    tx->extra_frag[i]);
-			if (ret)
-				return IEEE80211_TX_FRAG_AGAIN;
-			local->mdev->trans_start = jiffies;
-			ieee80211_led_tx(local, 1);
-			tx->extra_frag[i] = NULL;
+
+		next = skb->next;
+		len = skb->len;
+		ret = local->ops->tx(local_to_hw(local), skb);
+		if (WARN_ON(ret != NETDEV_TX_OK && skb->len != len)) {
+			dev_kfree_skb(skb);
+			ret = NETDEV_TX_OK;
 		}
-		kfree(tx->extra_frag);
-		tx->extra_frag = NULL;
+		if (ret != NETDEV_TX_OK)
+			return IEEE80211_TX_AGAIN;
+		*skbp = skb = next;
+		ieee80211_led_tx(local, 1);
+		fragm = true;
 	}
+
 	return IEEE80211_TX_OK;
 }
 
@@ -1137,7 +1170,6 @@ static int invoke_tx_handlers(struct ieee80211_tx_data *tx)
 {
 	struct sk_buff *skb = tx->skb;
 	ieee80211_tx_result res = TX_DROP;
-	int i;
 
 #define CALL_TXH(txh)		\
 	res = txh(tx);		\
@@ -1161,11 +1193,13 @@ static int invoke_tx_handlers(struct ieee80211_tx_data *tx)
  txh_done:
 	if (unlikely(res == TX_DROP)) {
 		I802_DEBUG_INC(tx->local->tx_handlers_drop);
-		dev_kfree_skb(skb);
-		for (i = 0; i < tx->num_extra_frag; i++)
-			if (tx->extra_frag[i])
-				dev_kfree_skb(tx->extra_frag[i]);
-		kfree(tx->extra_frag);
+		while (skb) {
+			struct sk_buff *next;
+
+			next = skb->next;
+			dev_kfree_skb(skb);
+			skb = next;
+		}
 		return -1;
 	} else if (unlikely(res == TX_QUEUED)) {
 		I802_DEBUG_INC(tx->local->tx_handlers_queued);
@@ -1175,23 +1209,26 @@ static int invoke_tx_handlers(struct ieee80211_tx_data *tx)
 	return 0;
 }
 
-static int ieee80211_tx(struct net_device *dev, struct sk_buff *skb)
+static void ieee80211_tx(struct net_device *dev, struct sk_buff *skb,
+			 bool txpending)
 {
 	struct ieee80211_local *local = wdev_priv(dev->ieee80211_ptr);
 	struct sta_info *sta;
 	struct ieee80211_tx_data tx;
 	ieee80211_tx_result res_prepare;
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
-	int ret, i;
+	struct sk_buff *next;
+	unsigned long flags;
+	int ret, retries;
 	u16 queue;
 
 	queue = skb_get_queue_mapping(skb);
 
-	WARN_ON(test_bit(queue, local->queues_pending));
+	WARN_ON(!txpending && !skb_queue_empty(&local->pending[queue]));
 
 	if (unlikely(skb->len < 10)) {
 		dev_kfree_skb(skb);
-		return 0;
+		return;
 	}
 
 	rcu_read_lock();
@@ -1199,10 +1236,13 @@ static int ieee80211_tx(struct net_device *dev, struct sk_buff *skb)
 	/* initialises tx */
 	res_prepare = __ieee80211_tx_prepare(&tx, skb, dev);
 
-	if (res_prepare == TX_DROP) {
+	if (unlikely(res_prepare == TX_DROP)) {
 		dev_kfree_skb(skb);
 		rcu_read_unlock();
-		return 0;
+		return;
+	} else if (unlikely(res_prepare == TX_QUEUED)) {
+		rcu_read_unlock();
+		return;
 	}
 
 	sta = tx.sta;
@@ -1212,59 +1252,71 @@ static int ieee80211_tx(struct net_device *dev, struct sk_buff *skb)
 	if (invoke_tx_handlers(&tx))
 		goto out;
 
-retry:
-	ret = __ieee80211_tx(local, skb, &tx);
-	if (ret) {
-		struct ieee80211_tx_stored_packet *store;
-
+	retries = 0;
+ retry:
+	ret = __ieee80211_tx(local, &tx.skb, tx.sta);
+	switch (ret) {
+	case IEEE80211_TX_OK:
+		break;
+	case IEEE80211_TX_AGAIN:
 		/*
 		 * Since there are no fragmented frames on A-MPDU
 		 * queues, there's no reason for a driver to reject
 		 * a frame there, warn and drop it.
 		 */
-		if (ret != IEEE80211_TX_PENDING)
-			if (WARN_ON(info->flags & IEEE80211_TX_CTL_AMPDU))
-				goto drop;
+		if (WARN_ON(info->flags & IEEE80211_TX_CTL_AMPDU))
+			goto drop;
+		/* fall through */
+	case IEEE80211_TX_PENDING:
+		skb = tx.skb;
+
+		spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
+
+		if (__netif_subqueue_stopped(local->mdev, queue)) {
+			do {
+				next = skb->next;
+				skb->next = NULL;
+				if (unlikely(txpending))
+					skb_queue_head(&local->pending[queue],
+						       skb);
+				else
+					skb_queue_tail(&local->pending[queue],
+						       skb);
+			} while ((skb = next));
 
-		store = &local->pending_packet[queue];
+			/*
+			 * Make sure nobody will enable the queue on us
+			 * (without going through the tasklet) nor disable the
+			 * netdev queue underneath the pending handling code.
+			 */
+			__set_bit(IEEE80211_QUEUE_STOP_REASON_PENDING,
+				  &local->queue_stop_reasons[queue]);
 
-		if (ret == IEEE80211_TX_FRAG_AGAIN)
-			skb = NULL;
+			spin_unlock_irqrestore(&local->queue_stop_reason_lock,
+					       flags);
+		} else {
+			spin_unlock_irqrestore(&local->queue_stop_reason_lock,
+					       flags);
 
-		set_bit(queue, local->queues_pending);
-		smp_mb();
-		/*
-		 * When the driver gets out of buffers during sending of
-		 * fragments and calls ieee80211_stop_queue, the netif
-		 * subqueue is stopped. There is, however, a small window
-		 * in which the PENDING bit is not yet set. If a buffer
-		 * gets available in that window (i.e. driver calls
-		 * ieee80211_wake_queue), we would end up with ieee80211_tx
-		 * called with the PENDING bit still set. Prevent this by
-		 * continuing transmitting here when that situation is
-		 * possible to have happened.
-		 */
-		if (!__netif_subqueue_stopped(local->mdev, queue)) {
-			clear_bit(queue, local->queues_pending);
+			retries++;
+			if (WARN(retries > 10, "tx refused but queue active"))
+				goto drop;
 			goto retry;
 		}
-		store->skb = skb;
-		store->extra_frag = tx.extra_frag;
-		store->num_extra_frag = tx.num_extra_frag;
 	}
  out:
 	rcu_read_unlock();
-	return 0;
+	return;
 
  drop:
-	if (skb)
-		dev_kfree_skb(skb);
-	for (i = 0; i < tx.num_extra_frag; i++)
-		if (tx.extra_frag[i])
-			dev_kfree_skb(tx.extra_frag[i]);
-	kfree(tx.extra_frag);
 	rcu_read_unlock();
-	return 0;
+
+	skb = tx.skb;
+	while (skb) {
+		next = skb->next;
+		dev_kfree_skb(skb);
+		skb = next;
+	}
 }
 
 /* device xmit handlers */
@@ -1323,7 +1375,6 @@ int ieee80211_master_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		FOUND_SDATA,
 		UNKNOWN_ADDRESS,
 	} monitor_iface = NOT_MONITOR;
-	int ret;
 
 	if (skb->iif)
 		odev = dev_get_by_index(&init_net, skb->iif);
@@ -1337,7 +1388,7 @@ int ieee80211_master_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		       "originating device\n", dev->name);
 #endif
 		dev_kfree_skb(skb);
-		return 0;
+		return NETDEV_TX_OK;
 	}
 
 	if ((local->hw.flags & IEEE80211_HW_PS_NULLFUNC_STACK) &&
@@ -1366,7 +1417,7 @@ int ieee80211_master_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		else
 			if (mesh_nexthop_lookup(skb, osdata)) {
 				dev_put(odev);
-				return 0;
+				return NETDEV_TX_OK;
 			}
 		if (memcmp(odev->dev_addr, hdr->addr4, ETH_ALEN) != 0)
 			IEEE80211_IFSTA_MESH_CTR_INC(&osdata->u.mesh,
@@ -1428,7 +1479,7 @@ int ieee80211_master_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	if (ieee80211_skb_resize(osdata->local, skb, headroom, may_encrypt)) {
 		dev_kfree_skb(skb);
 		dev_put(odev);
-		return 0;
+		return NETDEV_TX_OK;
 	}
 
 	if (osdata->vif.type == NL80211_IFTYPE_AP_VLAN)
@@ -1437,10 +1488,11 @@ int ieee80211_master_start_xmit(struct sk_buff *skb, struct net_device *dev)
 				      u.ap);
 	if (likely(monitor_iface != UNKNOWN_ADDRESS))
 		info->control.vif = &osdata->vif;
-	ret = ieee80211_tx(odev, skb);
+
+	ieee80211_tx(odev, skb, false);
 	dev_put(odev);
 
-	return ret;
+	return NETDEV_TX_OK;
 }
 
 int ieee80211_monitor_start_xmit(struct sk_buff *skb,
@@ -1666,8 +1718,7 @@ int ieee80211_subif_start_xmit(struct sk_buff *skb,
 	}
 
 	/* receiver and we are QoS enabled, use a QoS type frame */
-	if (sta_flags & WLAN_STA_WME &&
-	    ieee80211_num_regular_queues(&local->hw) >= 4) {
+	if ((sta_flags & WLAN_STA_WME) && local->hw.queues >= 4) {
 		fc |= cpu_to_le16(IEEE80211_STYPE_QOS_DATA);
 		hdrlen += 2;
 	}
@@ -1799,19 +1850,58 @@ int ieee80211_subif_start_xmit(struct sk_buff *skb,
  */
 void ieee80211_clear_tx_pending(struct ieee80211_local *local)
 {
-	int i, j;
-	struct ieee80211_tx_stored_packet *store;
+	int i;
 
-	for (i = 0; i < ieee80211_num_regular_queues(&local->hw); i++) {
-		if (!test_bit(i, local->queues_pending))
-			continue;
-		store = &local->pending_packet[i];
-		kfree_skb(store->skb);
-		for (j = 0; j < store->num_extra_frag; j++)
-			kfree_skb(store->extra_frag[j]);
-		kfree(store->extra_frag);
-		clear_bit(i, local->queues_pending);
+	for (i = 0; i < local->hw.queues; i++)
+		skb_queue_purge(&local->pending[i]);
+}
+
+static bool ieee80211_tx_pending_skb(struct ieee80211_local *local,
+				     struct sk_buff *skb)
+{
+	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
+	struct ieee80211_sub_if_data *sdata;
+	struct sta_info *sta;
+	struct ieee80211_hdr *hdr;
+	struct net_device *dev;
+	int ret;
+	bool result = true;
+
+	/* does interface still exist? */
+	dev = dev_get_by_index(&init_net, skb->iif);
+	if (!dev) {
+		dev_kfree_skb(skb);
+		return true;
 	}
+
+	/* validate info->control.vif against skb->iif */
+	sdata = IEEE80211_DEV_TO_SUB_IF(dev);
+	if (sdata->vif.type == NL80211_IFTYPE_AP_VLAN)
+		sdata = container_of(sdata->bss,
+				     struct ieee80211_sub_if_data,
+				     u.ap);
+
+	if (unlikely(info->control.vif && info->control.vif != &sdata->vif)) {
+		dev_kfree_skb(skb);
+		result = true;
+		goto out;
+	}
+
+	if (info->flags & IEEE80211_TX_INTFL_NEED_TXPROCESSING) {
+		ieee80211_tx(dev, skb, true);
+	} else {
+		hdr = (struct ieee80211_hdr *)skb->data;
+		sta = sta_info_get(local, hdr->addr1);
+
+		ret = __ieee80211_tx(local, &skb, sta);
+		if (ret != IEEE80211_TX_OK)
+			result = false;
+	}
+
+ out:
+	dev_put(dev);
+
+	return result;
 }
 
 /*
@@ -1822,40 +1912,53 @@ void ieee80211_tx_pending(unsigned long data)
 {
 	struct ieee80211_local *local = (struct ieee80211_local *)data;
 	struct net_device *dev = local->mdev;
-	struct ieee80211_tx_stored_packet *store;
-	struct ieee80211_tx_data tx;
-	int i, ret;
+	unsigned long flags;
+	int i;
+	bool next;
 
+	rcu_read_lock();
 	netif_tx_lock_bh(dev);
-	for (i = 0; i < ieee80211_num_regular_queues(&local->hw); i++) {
-		/* Check that this queue is ok */
-		if (__netif_subqueue_stopped(local->mdev, i) &&
-		    !test_bit(i, local->queues_pending_run))
-			continue;
 
-		if (!test_bit(i, local->queues_pending)) {
-			clear_bit(i, local->queues_pending_run);
-			ieee80211_wake_queue(&local->hw, i);
+	for (i = 0; i < local->hw.queues; i++) {
+		/*
+		 * If queue is stopped by something other than due to pending
+		 * frames, or we have no pending frames, proceed to next queue.
+		 */
+		spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
+		next = false;
+		if (local->queue_stop_reasons[i] !=
+			BIT(IEEE80211_QUEUE_STOP_REASON_PENDING) ||
+		    skb_queue_empty(&local->pending[i]))
+			next = true;
+		spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
+
+		if (next)
 			continue;
-		}
 
-		clear_bit(i, local->queues_pending_run);
+		/*
+		 * start the queue now to allow processing our packets,
+		 * we're under the tx lock here anyway so nothing will
+		 * happen as a result of this
+		 */
 		netif_start_subqueue(local->mdev, i);
 
-		store = &local->pending_packet[i];
-		tx.extra_frag = store->extra_frag;
-		tx.num_extra_frag = store->num_extra_frag;
-		tx.flags = 0;
-		ret = __ieee80211_tx(local, store->skb, &tx);
-		if (ret) {
-			if (ret == IEEE80211_TX_FRAG_AGAIN)
-				store->skb = NULL;
-		} else {
-			clear_bit(i, local->queues_pending);
-			ieee80211_wake_queue(&local->hw, i);
+		while (!skb_queue_empty(&local->pending[i])) {
+			struct sk_buff *skb = skb_dequeue(&local->pending[i]);
+
+			if (!ieee80211_tx_pending_skb(local, skb)) {
+				skb_queue_head(&local->pending[i], skb);
+				break;
+			}
 		}
+
+		/* Start regular packet processing again. */
+		if (skb_queue_empty(&local->pending[i]))
+			ieee80211_wake_queue_by_reason(&local->hw, i,
+					IEEE80211_QUEUE_STOP_REASON_PENDING);
 	}
+
 	netif_tx_unlock_bh(dev);
+	rcu_read_unlock();
 }
 
 /* functions for drivers to get certain frames */