diff --git a/drivers/net/ifb.c b/drivers/net/ifb.c
index af233b591534..bc3de272a829 100644
--- a/drivers/net/ifb.c
+++ b/drivers/net/ifb.c
@@ -229,12 +229,12 @@ module_param(numifbs, int, 0);
 MODULE_PARM_DESC(numifbs, "Number of ifb devices");
 
 /*
- * dev_ifb->queue_lock is usually taken after dev->ingress_lock,
+ * dev_ifb->tx_queue.lock is usually taken after dev->ingress_lock,
  * reversely to e.g. qdisc_lock_tree(). It should be safe until
- * ifb doesn't take dev->queue_lock with dev_ifb->ingress_lock.
+ * ifb doesn't take dev->tx_queue.lock with dev_ifb->ingress_lock.
  * But lockdep should know that ifb has different locks from dev.
  */
-static struct lock_class_key ifb_queue_lock_key;
+static struct lock_class_key ifb_tx_queue_lock_key;
 static struct lock_class_key ifb_ingress_lock_key;
 
 
@@ -258,7 +258,7 @@ static int __init ifb_init_one(int index)
 	if (err < 0)
 		goto err;
 
-	lockdep_set_class(&dev_ifb->queue_lock, &ifb_queue_lock_key);
+	lockdep_set_class(&dev_ifb->tx_queue.lock, &ifb_tx_queue_lock_key);
 	lockdep_set_class(&dev_ifb->ingress_lock, &ifb_ingress_lock_key);
 
 	return 0;
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 515fd25bf0fc..e835acacb479 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -449,6 +449,7 @@ static inline void napi_synchronize(const struct napi_struct *n)
 #endif
 
 struct netdev_queue {
+	spinlock_t		lock;
 	struct net_device	*dev;
 };
 
@@ -629,7 +630,7 @@ struct net_device
 	unsigned char		broadcast[MAX_ADDR_LEN];	/* hw bcast add	*/
 
 	struct netdev_queue	rx_queue;
-	struct netdev_queue	tx_queue;
+	struct netdev_queue	tx_queue ____cacheline_aligned_in_smp;
 
 	/* ingress path synchronizer */
 	spinlock_t		ingress_lock;
@@ -639,7 +640,6 @@ struct net_device
  * Cache line mostly used on queue transmit path (qdisc)
  */
 	/* device queue lock */
-	spinlock_t		queue_lock ____cacheline_aligned_in_smp;
 	struct Qdisc		*qdisc;
 	struct Qdisc		*qdisc_sleeping;
 	struct list_head	qdisc_list;
diff --git a/net/core/dev.c b/net/core/dev.c
index 9b281c906eb0..05011048b86c 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1667,6 +1667,7 @@ int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 int dev_queue_xmit(struct sk_buff *skb)
 {
 	struct net_device *dev = skb->dev;
+	struct netdev_queue *txq;
 	struct Qdisc *q;
 	int rc = -ENOMEM;
 
@@ -1699,14 +1700,15 @@ int dev_queue_xmit(struct sk_buff *skb)
 	}
 
 gso:
-	spin_lock_prefetch(&dev->queue_lock);
+	txq = &dev->tx_queue;
+	spin_lock_prefetch(&txq->lock);
 
 	/* Disable soft irqs for various locks below. Also
 	 * stops preemption for RCU.
 	 */
 	rcu_read_lock_bh();
 
-	/* Updates of qdisc are serialized by queue_lock.
+	/* Updates of qdisc are serialized by queue->lock.
 	 * The struct Qdisc which is pointed to by qdisc is now a
 	 * rcu structure - it may be accessed without acquiring
 	 * a lock (but the structure may be stale.) The freeing of the
@@ -1714,7 +1716,7 @@ int dev_queue_xmit(struct sk_buff *skb)
 	 * more references to it.
 	 *
 	 * If the qdisc has an enqueue function, we still need to
-	 * hold the queue_lock before calling it, since queue_lock
+	 * hold the queue->lock before calling it, since queue->lock
 	 * also serializes access to the device queue.
 	 */
 
@@ -1724,19 +1726,19 @@ int dev_queue_xmit(struct sk_buff *skb)
 #endif
 	if (q->enqueue) {
 		/* Grab device queue */
-		spin_lock(&dev->queue_lock);
+		spin_lock(&txq->lock);
 		q = dev->qdisc;
 		if (q->enqueue) {
 			/* reset queue_mapping to zero */
 			skb_set_queue_mapping(skb, 0);
 			rc = q->enqueue(skb, q);
 			qdisc_run(dev);
-			spin_unlock(&dev->queue_lock);
+			spin_unlock(&txq->lock);
 
 			rc = rc == NET_XMIT_BYPASS ? NET_XMIT_SUCCESS : rc;
 			goto out;
 		}
-		spin_unlock(&dev->queue_lock);
+		spin_unlock(&txq->lock);
 	}
 
 	/* The device has no queue. Common case for software devices:
@@ -1919,14 +1921,17 @@ static void net_tx_action(struct softirq_action *h)
 
 		while (head) {
 			struct net_device *dev = head;
+			struct netdev_queue *txq;
 			head = head->next_sched;
 
+			txq = &dev->tx_queue;
+
 			smp_mb__before_clear_bit();
 			clear_bit(__LINK_STATE_SCHED, &dev->state);
 
-			if (spin_trylock(&dev->queue_lock)) {
+			if (spin_trylock(&txq->lock)) {
 				qdisc_run(dev);
-				spin_unlock(&dev->queue_lock);
+				spin_unlock(&txq->lock);
 			} else {
 				netif_schedule(dev);
 			}
@@ -3787,7 +3792,6 @@ int register_netdevice(struct net_device *dev)
 	BUG_ON(!dev_net(dev));
 	net = dev_net(dev);
 
-	spin_lock_init(&dev->queue_lock);
 	spin_lock_init(&dev->_xmit_lock);
 	netdev_set_lockdep_class(&dev->_xmit_lock, dev->type);
 	dev->xmit_lock_owner = -1;
@@ -4072,10 +4076,17 @@ static struct net_device_stats *internal_stats(struct net_device *dev)
 	return &dev->stats;
 }
 
+static void netdev_init_one_queue(struct net_device *dev,
+				  struct netdev_queue *queue)
+{
+	spin_lock_init(&queue->lock);
+	queue->dev = dev;
+}
+
 static void netdev_init_queues(struct net_device *dev)
 {
-	dev->rx_queue.dev = dev;
-	dev->tx_queue.dev = dev;
+	netdev_init_one_queue(dev, &dev->rx_queue);
+	netdev_init_one_queue(dev, &dev->tx_queue);
 }
 
 /**
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index cf477ad39dac..12aeaf78ae75 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -636,7 +636,7 @@ int ieee80211_start_tx_ba_session(struct ieee80211_hw *hw, u8 *ra, u16 tid)
 
 	/* ensure that TX flow won't interrupt us
 	 * until the end of the call to requeue function */
-	spin_lock_bh(&local->mdev->queue_lock);
+	spin_lock_bh(&local->mdev->tx_queue.lock);
 
 	/* create a new queue for this aggregation */
 	ret = ieee80211_ht_agg_queue_add(local, sta, tid);
@@ -675,7 +675,7 @@ int ieee80211_start_tx_ba_session(struct ieee80211_hw *hw, u8 *ra, u16 tid)
 
 	/* Will put all the packets in the new SW queue */
 	ieee80211_requeue(local, ieee802_1d_to_ac[tid]);
-	spin_unlock_bh(&local->mdev->queue_lock);
+	spin_unlock_bh(&local->mdev->tx_queue.lock);
 	spin_unlock_bh(&sta->lock);
 
 	/* send an addBA request */
@@ -701,7 +701,7 @@ int ieee80211_start_tx_ba_session(struct ieee80211_hw *hw, u8 *ra, u16 tid)
 err_unlock_queue:
 	kfree(sta->ampdu_mlme.tid_tx[tid]);
 	sta->ampdu_mlme.tid_tx[tid] = NULL;
-	spin_unlock_bh(&local->mdev->queue_lock);
+	spin_unlock_bh(&local->mdev->tx_queue.lock);
 	ret = -EBUSY;
 err_unlock_sta:
 	spin_unlock_bh(&sta->lock);
@@ -875,10 +875,10 @@ void ieee80211_stop_tx_ba_cb(struct ieee80211_hw *hw, u8 *ra, u8 tid)
 
 	/* avoid ordering issues: we are the only one that can modify
 	 * the content of the qdiscs */
-	spin_lock_bh(&local->mdev->queue_lock);
+	spin_lock_bh(&local->mdev->tx_queue.lock);
 	/* remove the queue for this aggregation */
 	ieee80211_ht_agg_queue_remove(local, sta, tid, 1);
-	spin_unlock_bh(&local->mdev->queue_lock);
+	spin_unlock_bh(&local->mdev->tx_queue.lock);
 
 	/* we just requeued the all the frames that were in the removed
 	 * queue, and since we might miss a softirq we do netif_schedule.
diff --git a/net/mac80211/wme.c b/net/mac80211/wme.c
index 2fbc171130bf..59ed9cae66b9 100644
--- a/net/mac80211/wme.c
+++ b/net/mac80211/wme.c
@@ -648,7 +648,7 @@ int ieee80211_ht_agg_queue_add(struct ieee80211_local *local,
 }
 
 /**
- * the caller needs to hold local->mdev->queue_lock
+ * the caller needs to hold local->mdev->tx_queue.lock
  */
 void ieee80211_ht_agg_queue_remove(struct ieee80211_local *local,
 				   struct sta_info *sta, u16 tid,
diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c
index 1f893082a4f6..2a1834f8c7d8 100644
--- a/net/sched/sch_api.c
+++ b/net/sched/sch_api.c
@@ -606,7 +606,7 @@ qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue,
 		sch->stats_lock = &dev->ingress_lock;
 		handle = TC_H_MAKE(TC_H_INGRESS, 0);
 	} else {
-		sch->stats_lock = &dev->queue_lock;
+		sch->stats_lock = &dev_queue->lock;
 		if (handle == 0) {
 			handle = qdisc_alloc_handle(dev);
 			err = -ENOMEM;
diff --git a/net/sched/sch_cbq.c b/net/sched/sch_cbq.c
index 9f2ace585fd6..99ce3da2b0a4 100644
--- a/net/sched/sch_cbq.c
+++ b/net/sched/sch_cbq.c
@@ -1746,10 +1746,10 @@ static void cbq_put(struct Qdisc *sch, unsigned long arg)
 #ifdef CONFIG_NET_CLS_ACT
 		struct cbq_sched_data *q = qdisc_priv(sch);
 
-		spin_lock_bh(&qdisc_dev(sch)->queue_lock);
+		spin_lock_bh(&sch->dev_queue->lock);
 		if (q->rx_class == cl)
 			q->rx_class = NULL;
-		spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+		spin_unlock_bh(&sch->dev_queue->lock);
 #endif
 
 		cbq_destroy_class(sch, cl);
@@ -1828,7 +1828,7 @@ cbq_change_class(struct Qdisc *sch, u32 classid, u32 parentid, struct nlattr **t
 
 		if (tca[TCA_RATE])
 			gen_replace_estimator(&cl->bstats, &cl->rate_est,
-					      &qdisc_dev(sch)->queue_lock,
+					      &sch->dev_queue->lock,
 					      tca[TCA_RATE]);
 		return 0;
 	}
@@ -1919,7 +1919,7 @@ cbq_change_class(struct Qdisc *sch, u32 classid, u32 parentid, struct nlattr **t
 
 	if (tca[TCA_RATE])
 		gen_new_estimator(&cl->bstats, &cl->rate_est,
-				  &qdisc_dev(sch)->queue_lock, tca[TCA_RATE]);
+				  &sch->dev_queue->lock, tca[TCA_RATE]);
 
 	*arg = (unsigned long)cl;
 	return 0;
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index b626a4f32b6b..ee8f9f78a095 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -29,31 +29,31 @@
 /* Main transmission queue. */
 
 /* Modifications to data participating in scheduling must be protected with
- * dev->queue_lock spinlock.
+ * queue->lock spinlock.
  *
  * The idea is the following:
  * - enqueue, dequeue are serialized via top level device
- *   spinlock dev->queue_lock.
+ *   spinlock queue->lock.
  * - ingress filtering is serialized via top level device
  *   spinlock dev->ingress_lock.
  * - updates to tree and tree walking are only done under the rtnl mutex.
  */
 
 void qdisc_lock_tree(struct net_device *dev)
-	__acquires(dev->queue_lock)
+	__acquires(dev->tx_queue.lock)
 	__acquires(dev->ingress_lock)
 {
-	spin_lock_bh(&dev->queue_lock);
+	spin_lock_bh(&dev->tx_queue.lock);
 	spin_lock(&dev->ingress_lock);
 }
 EXPORT_SYMBOL(qdisc_lock_tree);
 
 void qdisc_unlock_tree(struct net_device *dev)
 	__releases(dev->ingress_lock)
-	__releases(dev->queue_lock)
+	__releases(dev->tx_queue.lock)
 {
 	spin_unlock(&dev->ingress_lock);
-	spin_unlock_bh(&dev->queue_lock);
+	spin_unlock_bh(&dev->tx_queue.lock);
 }
 EXPORT_SYMBOL(qdisc_unlock_tree);
 
@@ -118,15 +118,15 @@ static inline int handle_dev_cpu_collision(struct sk_buff *skb,
 }
 
 /*
- * NOTE: Called under dev->queue_lock with locally disabled BH.
+ * NOTE: Called under queue->lock with locally disabled BH.
  *
  * __LINK_STATE_QDISC_RUNNING guarantees only one CPU can process this
- * device at a time. dev->queue_lock serializes queue accesses for
+ * device at a time. queue->lock serializes queue accesses for
  * this device AND dev->qdisc pointer itself.
  *
  *  netif_tx_lock serializes accesses to device driver.
  *
- *  dev->queue_lock and netif_tx_lock are mutually exclusive,
+ *  queue->lock and netif_tx_lock are mutually exclusive,
  *  if one is grabbed, another must be free.
  *
  * Note, that this procedure can be called by a watchdog timer
@@ -148,14 +148,14 @@ static inline int qdisc_restart(struct net_device *dev)
 
 
 	/* And release queue */
-	spin_unlock(&dev->queue_lock);
+	spin_unlock(&q->dev_queue->lock);
 
 	HARD_TX_LOCK(dev, smp_processor_id());
 	if (!netif_subqueue_stopped(dev, skb))
 		ret = dev_hard_start_xmit(skb, dev);
 	HARD_TX_UNLOCK(dev);
 
-	spin_lock(&dev->queue_lock);
+	spin_lock(&q->dev_queue->lock);
 	q = dev->qdisc;
 
 	switch (ret) {
@@ -482,7 +482,7 @@ struct Qdisc * qdisc_create_dflt(struct net_device *dev,
 	sch = qdisc_alloc(dev_queue, ops);
 	if (IS_ERR(sch))
 		goto errout;
-	sch->stats_lock = &dev->queue_lock;
+	sch->stats_lock = &dev_queue->lock;
 	sch->parent = parentid;
 
 	if (!ops->init || ops->init(sch, NULL) == 0)
@@ -494,7 +494,7 @@ struct Qdisc * qdisc_create_dflt(struct net_device *dev,
 }
 EXPORT_SYMBOL(qdisc_create_dflt);
 
-/* Under dev->queue_lock and BH! */
+/* Under queue->lock and BH! */
 
 void qdisc_reset(struct Qdisc *qdisc)
 {
@@ -514,7 +514,7 @@ static void __qdisc_destroy(struct rcu_head *head)
 	kfree((char *) qdisc - qdisc->padded);
 }
 
-/* Under dev->queue_lock and BH! */
+/* Under queue->lock and BH! */
 
 void qdisc_destroy(struct Qdisc *qdisc)
 {
@@ -566,13 +566,13 @@ void dev_activate(struct net_device *dev)
 		/* Delay activation until next carrier-on event */
 		return;
 
-	spin_lock_bh(&dev->queue_lock);
+	spin_lock_bh(&dev->tx_queue.lock);
 	rcu_assign_pointer(dev->qdisc, dev->qdisc_sleeping);
 	if (dev->qdisc != &noqueue_qdisc) {
 		dev->trans_start = jiffies;
 		dev_watchdog_up(dev);
 	}
-	spin_unlock_bh(&dev->queue_lock);
+	spin_unlock_bh(&dev->tx_queue.lock);
 }
 
 void dev_deactivate(struct net_device *dev)
@@ -581,7 +581,7 @@ void dev_deactivate(struct net_device *dev)
 	struct sk_buff *skb;
 	int running;
 
-	spin_lock_bh(&dev->queue_lock);
+	spin_lock_bh(&dev->tx_queue.lock);
 	qdisc = dev->qdisc;
 	dev->qdisc = &noop_qdisc;
 
@@ -589,7 +589,7 @@ void dev_deactivate(struct net_device *dev)
 
 	skb = dev->gso_skb;
 	dev->gso_skb = NULL;
-	spin_unlock_bh(&dev->queue_lock);
+	spin_unlock_bh(&dev->tx_queue.lock);
 
 	kfree_skb(skb);
 
@@ -607,9 +607,9 @@ void dev_deactivate(struct net_device *dev)
 		 * Double-check inside queue lock to ensure that all effects
 		 * of the queue run are visible when we return.
 		 */
-		spin_lock_bh(&dev->queue_lock);
+		spin_lock_bh(&dev->tx_queue.lock);
 		running = test_bit(__LINK_STATE_QDISC_RUNNING, &dev->state);
-		spin_unlock_bh(&dev->queue_lock);
+		spin_unlock_bh(&dev->tx_queue.lock);
 
 		/*
 		 * The running flag should never be set at this point because
diff --git a/net/sched/sch_hfsc.c b/net/sched/sch_hfsc.c
index 333525422f45..997d520ca580 100644
--- a/net/sched/sch_hfsc.c
+++ b/net/sched/sch_hfsc.c
@@ -1045,7 +1045,7 @@ hfsc_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
 
 		if (tca[TCA_RATE])
 			gen_replace_estimator(&cl->bstats, &cl->rate_est,
-					      &qdisc_dev(sch)->queue_lock,
+					      &sch->dev_queue->lock,
 					      tca[TCA_RATE]);
 		return 0;
 	}
@@ -1104,7 +1104,7 @@ hfsc_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
 
 	if (tca[TCA_RATE])
 		gen_new_estimator(&cl->bstats, &cl->rate_est,
-				  &qdisc_dev(sch)->queue_lock, tca[TCA_RATE]);
+				  &sch->dev_queue->lock, tca[TCA_RATE]);
 	*arg = (unsigned long)cl;
 	return 0;
 }
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index 31f7d1536e6d..c8ca54cc26b0 100644
--- a/net/sched/sch_htb.c
+++ b/net/sched/sch_htb.c
@@ -1043,7 +1043,7 @@ static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
 	struct nlattr *nest;
 	struct tc_htb_glob gopt;
 
-	spin_lock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_lock_bh(&sch->dev_queue->lock);
 
 	gopt.direct_pkts = q->direct_pkts;
 	gopt.version = HTB_VER;
@@ -1057,11 +1057,11 @@ static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
 	NLA_PUT(skb, TCA_HTB_INIT, sizeof(gopt), &gopt);
 	nla_nest_end(skb, nest);
 
-	spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_unlock_bh(&sch->dev_queue->lock);
 	return skb->len;
 
 nla_put_failure:
-	spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_unlock_bh(&sch->dev_queue->lock);
 	nla_nest_cancel(skb, nest);
 	return -1;
 }
@@ -1073,7 +1073,7 @@ static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
 	struct nlattr *nest;
 	struct tc_htb_opt opt;
 
-	spin_lock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_lock_bh(&sch->dev_queue->lock);
 	tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT;
 	tcm->tcm_handle = cl->common.classid;
 	if (!cl->level && cl->un.leaf.q)
@@ -1095,11 +1095,11 @@ static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
 	NLA_PUT(skb, TCA_HTB_PARMS, sizeof(opt), &opt);
 
 	nla_nest_end(skb, nest);
-	spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_unlock_bh(&sch->dev_queue->lock);
 	return skb->len;
 
 nla_put_failure:
-	spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_unlock_bh(&sch->dev_queue->lock);
 	nla_nest_cancel(skb, nest);
 	return -1;
 }
@@ -1365,7 +1365,7 @@ static int htb_change_class(struct Qdisc *sch, u32 classid,
 			goto failure;
 
 		gen_new_estimator(&cl->bstats, &cl->rate_est,
-				  &qdisc_dev(sch)->queue_lock,
+				  &sch->dev_queue->lock,
 				  tca[TCA_RATE] ? : &est.nla);
 		cl->refcnt = 1;
 		cl->children = 0;
@@ -1420,7 +1420,7 @@ static int htb_change_class(struct Qdisc *sch, u32 classid,
 	} else {
 		if (tca[TCA_RATE])
 			gen_replace_estimator(&cl->bstats, &cl->rate_est,
-					      &qdisc_dev(sch)->queue_lock,
+					      &sch->dev_queue->lock,
 					      tca[TCA_RATE]);
 		sch_tree_lock(sch);
 	}
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 790582960444..71b73c528f9b 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -333,9 +333,9 @@ static int get_dist_table(struct Qdisc *sch, const struct nlattr *attr)
 	for (i = 0; i < n; i++)
 		d->table[i] = data[i];
 
-	spin_lock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_lock_bh(&sch->dev_queue->lock);
 	d = xchg(&q->delay_dist, d);
-	spin_unlock_bh(&qdisc_dev(sch)->queue_lock);
+	spin_unlock_bh(&sch->dev_queue->lock);
 
 	kfree(d);
 	return 0;
diff --git a/net/sched/sch_teql.c b/net/sched/sch_teql.c
index b3fc82623fc6..4f3054e8e1ab 100644
--- a/net/sched/sch_teql.c
+++ b/net/sched/sch_teql.c
@@ -154,9 +154,9 @@ teql_destroy(struct Qdisc* sch)
 					master->slaves = NEXT_SLAVE(q);
 					if (q == master->slaves) {
 						master->slaves = NULL;
-						spin_lock_bh(&master->dev->queue_lock);
+						spin_lock_bh(&master->dev->tx_queue.lock);
 						qdisc_reset(master->dev->qdisc);
-						spin_unlock_bh(&master->dev->queue_lock);
+						spin_unlock_bh(&master->dev->tx_queue.lock);
 					}
 				}
 				skb_queue_purge(&dat->q);