diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index da6839a7ff50e264d694a2c2b2309375ff6aadcb..9dcb5bfe094a3c6b833f253c1e04cba17c3d89ae 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -438,6 +438,29 @@ static inline struct sk_buff *qdisc_peek_head(struct Qdisc *sch)
 	return skb_peek(&sch->q);
 }
 
+/* generic pseudo peek method for non-work-conserving qdisc */
+static inline struct sk_buff *qdisc_peek_dequeued(struct Qdisc *sch)
+{
+	/* we can reuse ->gso_skb because peek isn't called for root qdiscs */
+	if (!sch->gso_skb)
+		sch->gso_skb = sch->dequeue(sch);
+
+	return sch->gso_skb;
+}
+
+/* use instead of qdisc->dequeue() for all qdiscs queried with ->peek() */
+static inline struct sk_buff *qdisc_dequeue_peeked(struct Qdisc *sch)
+{
+	struct sk_buff *skb = sch->gso_skb;
+
+	if (skb)
+		sch->gso_skb = NULL;
+	else
+		skb = sch->dequeue(sch);
+
+	return skb;
+}
+
 static inline int __qdisc_requeue(struct sk_buff *skb, struct Qdisc *sch,
 				  struct sk_buff_head *list)
 {
diff --git a/net/sched/sch_atm.c b/net/sched/sch_atm.c
index 2ee0c1a8efa961fe05e73a00ba9e1cdc2233f0b9..6eb9a650b63da60888e5108ca4c3fb9df01f776f 100644
--- a/net/sched/sch_atm.c
+++ b/net/sched/sch_atm.c
@@ -484,7 +484,7 @@ static void sch_atm_dequeue(unsigned long data)
 			if (!atm_may_send(flow->vcc, skb->truesize))
 				break;
 
-			skb = flow->q->dequeue(flow->q);
+			skb = qdisc_dequeue_peeked(flow->q);
 			if (unlikely(!skb))
 				break;
 
@@ -519,7 +519,7 @@ static struct sk_buff *atm_tc_dequeue(struct Qdisc *sch)
 
 	pr_debug("atm_tc_dequeue(sch %p,[qdisc %p])\n", sch, p);
 	tasklet_schedule(&p->task);
-	skb = p->link.q->dequeue(p->link.q);
+	skb = qdisc_dequeue_peeked(p->link.q);
 	if (skb)
 		sch->q.qlen--;
 	return skb;
diff --git a/net/sched/sch_cbq.c b/net/sched/sch_cbq.c
index 03e389e8d945d2b93b2d31b408a2ee78c6f0c8b9..63efa70abbea4cbc431956001ed6a82ba12c5b06 100644
--- a/net/sched/sch_cbq.c
+++ b/net/sched/sch_cbq.c
@@ -2066,6 +2066,7 @@ static struct Qdisc_ops cbq_qdisc_ops __read_mostly = {
 	.priv_size	=	sizeof(struct cbq_sched_data),
 	.enqueue	=	cbq_enqueue,
 	.dequeue	=	cbq_dequeue,
+	.peek		=	qdisc_peek_dequeued,
 	.requeue	=	cbq_requeue,
 	.drop		=	cbq_drop,
 	.init		=	cbq_init,
diff --git a/net/sched/sch_hfsc.c b/net/sched/sch_hfsc.c
index ddfc40887848b1dd5a22ef1b6d20be6a311e771c..d90b1652f2af8f2a557d410a1790166ff42f8277 100644
--- a/net/sched/sch_hfsc.c
+++ b/net/sched/sch_hfsc.c
@@ -1634,7 +1634,7 @@ hfsc_dequeue(struct Qdisc *sch)
 		}
 	}
 
-	skb = cl->qdisc->dequeue(cl->qdisc);
+	skb = qdisc_dequeue_peeked(cl->qdisc);
 	if (skb == NULL) {
 		if (net_ratelimit())
 			printk("HFSC: Non-work-conserving qdisc ?\n");
@@ -1727,6 +1727,7 @@ static struct Qdisc_ops hfsc_qdisc_ops __read_mostly = {
 	.dump		= hfsc_dump_qdisc,
 	.enqueue	= hfsc_enqueue,
 	.dequeue	= hfsc_dequeue,
+	.peek		= qdisc_peek_dequeued,
 	.requeue	= hfsc_requeue,
 	.drop		= hfsc_drop,
 	.cl_ops		= &hfsc_class_ops,
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index d14f02056ae6dcc31dbfad8c90ba57424e4a6d45..3fda8199713de5e48f068a4585ab0916a8ee9790 100644
--- a/net/sched/sch_htb.c
+++ b/net/sched/sch_htb.c
@@ -1565,6 +1565,7 @@ static struct Qdisc_ops htb_qdisc_ops __read_mostly = {
 	.priv_size	=	sizeof(struct htb_sched),
 	.enqueue	=	htb_enqueue,
 	.dequeue	=	htb_dequeue,
+	.peek		=	qdisc_peek_dequeued,
 	.requeue	=	htb_requeue,
 	.drop		=	htb_drop,
 	.init		=	htb_init,
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 74fbdb52baedae504bf09d23a266e1abde884260..3080bd6ee33224b7f5469c0da8eb704117cc41f4 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -290,8 +290,8 @@ static struct sk_buff *netem_dequeue(struct Qdisc *sch)
 
 		/* if more time remaining? */
 		if (cb->time_to_send <= now) {
-			skb = q->qdisc->dequeue(q->qdisc);
-			if (!skb)
+			skb = qdisc_dequeue_peeked(q->qdisc);
+			if (unlikely(!skb))
 				return NULL;
 
 			pr_debug("netem_dequeue: return skb=%p\n", skb);
@@ -714,6 +714,7 @@ static struct Qdisc_ops netem_qdisc_ops __read_mostly = {
 	.priv_size	=	sizeof(struct netem_sched_data),
 	.enqueue	=	netem_enqueue,
 	.dequeue	=	netem_dequeue,
+	.peek		=	qdisc_peek_dequeued,
 	.requeue	=	netem_requeue,
 	.drop		=	netem_drop,
 	.init		=	netem_init,
diff --git a/net/sched/sch_tbf.c b/net/sched/sch_tbf.c
index 61fdc77a48d293cc4be42c3eee9539b4ec21c463..435076cf620ee405fd55e12763659cd222e24110 100644
--- a/net/sched/sch_tbf.c
+++ b/net/sched/sch_tbf.c
@@ -192,7 +192,7 @@ static struct sk_buff *tbf_dequeue(struct Qdisc* sch)
 		toks -= L2T(q, len);
 
 		if ((toks|ptoks) >= 0) {
-			skb = q->qdisc->dequeue(q->qdisc);
+			skb = qdisc_dequeue_peeked(q->qdisc);
 			if (unlikely(!skb))
 				return NULL;
 
@@ -467,6 +467,7 @@ static struct Qdisc_ops tbf_qdisc_ops __read_mostly = {
 	.priv_size	=	sizeof(struct tbf_sched_data),
 	.enqueue	=	tbf_enqueue,
 	.dequeue	=	tbf_dequeue,
+	.peek		=	qdisc_peek_dequeued,
 	.requeue	=	tbf_requeue,
 	.drop		=	tbf_drop,
 	.init		=	tbf_init,