[PATCH net-next 0/5] sched: convert queues to single-linked list

[PATCH net-next 0/5] sched: convert queues to single-linked list

[Florian Westphal]

During Netfilter Workshop 2016 Eric Dumazet pointed out that qdisc
schedulers use doubly-linked lists, even though single-linked list
would be enough.

The double-linked skb lists incur one extra write on enqueue/dequeue
operations (to change ->prev pointer of next list elem).

This series converts qdiscs to single-linked version, listhead
maintains pointers to first (for dequeue) and last skb (for enqueue).

Most qdiscs don't queue at all and instead use a leaf qdisc (typically
pfifo_fast) so only a few schedulers needed changes.

I briefly tested netem and htb and they seemed fine.

UDP_STREAM netperf with 64 byte packets via veth+pfifo_fast shows
a small (~2%) improvement.

Florian Westphal (5):
      pie: use qdisc_dequeue_head wrapper
      sched: don't use skb queue helpers
      sched: remove qdisc arg from __qdisc_dequeue_head
      sched: replace __skb_dequeue with __qdisc_dequeue_head
      sched: add and use qdisc_skb_head helpers

 include/net/sch_generic.h |   72 +++++++++++++++++++++++++++++++++++-----------
 net/sched/sch_codel.c     |    4 +-
 net/sched/sch_fifo.c      |    4 +-
 net/sched/sch_generic.c   |   30 +++++++++++--------
 net/sched/sch_htb.c       |   24 ++++++++++++---
 net/sched/sch_netem.c     |   20 +++++++++---
 net/sched/sch_pie.c       |    4 +-
 7 files changed, 115 insertions(+), 43 deletions(-)

[PATCH net-next 1/5] pie: use qdisc_dequeue_head wrapper

[Florian Westphal]

Doesn't change generated code.

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 net/sched/sch_pie.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/sched/sch_pie.c b/net/sched/sch_pie.c
index a570b0b..d976d74 100644
--- a/net/sched/sch_pie.c
+++ b/net/sched/sch_pie.c
@@ -511,7 +511,7 @@ static int pie_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 static struct sk_buff *pie_qdisc_dequeue(struct Qdisc *sch)
 {
 	struct sk_buff *skb;
-	skb = __qdisc_dequeue_head(sch, &sch->q);
+	skb = qdisc_dequeue_head(sch);
 
 	if (!skb)
 		return NULL;
-- 
2.7.3

[PATCH net-next 2/5] sched: don't use skb queue helpers

[Florian Westphal]

A followup change will replace the sk_buff_head in the qdisc
struct with a slightly different list.

Use of the sk_buff_head helpers will thus cause compiler
warnings.

Open-code these accesses in an extra change to ease review.

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 net/sched/sch_fifo.c    | 4 ++--
 net/sched/sch_generic.c | 2 +-
 net/sched/sch_netem.c   | 4 ++--
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/net/sched/sch_fifo.c b/net/sched/sch_fifo.c
index baeed6a..1e37247 100644
--- a/net/sched/sch_fifo.c
+++ b/net/sched/sch_fifo.c
@@ -31,7 +31,7 @@ static int bfifo_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 static int pfifo_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			 struct sk_buff **to_free)
 {
-	if (likely(skb_queue_len(&sch->q) < sch->limit))
+	if (likely(sch->q.qlen < sch->limit))
 		return qdisc_enqueue_tail(skb, sch);
 
 	return qdisc_drop(skb, sch, to_free);
@@ -42,7 +42,7 @@ static int pfifo_tail_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 {
 	unsigned int prev_backlog;
 
-	if (likely(skb_queue_len(&sch->q) < sch->limit))
+	if (likely(sch->q.qlen < sch->limit))
 		return qdisc_enqueue_tail(skb, sch);
 
 	prev_backlog = sch->qstats.backlog;
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 0d21b56..5e63bf6 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -486,7 +486,7 @@ static inline struct sk_buff_head *band2list(struct pfifo_fast_priv *priv,
 static int pfifo_fast_enqueue(struct sk_buff *skb, struct Qdisc *qdisc,
 			      struct sk_buff **to_free)
 {
-	if (skb_queue_len(&qdisc->q) < qdisc_dev(qdisc)->tx_queue_len) {
+	if (qdisc->q.qlen < qdisc_dev(qdisc)->tx_queue_len) {
 		int band = prio2band[skb->priority & TC_PRIO_MAX];
 		struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
 		struct sk_buff_head *list = band2list(priv, band);
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index aaaf021..1832d77 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -502,7 +502,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			1<<(prandom_u32() % 8);
 	}
 
-	if (unlikely(skb_queue_len(&sch->q) >= sch->limit))
+	if (unlikely(sch->q.qlen >= sch->limit))
 		return qdisc_drop(skb, sch, to_free);
 
 	qdisc_qstats_backlog_inc(sch, skb);
@@ -522,7 +522,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		if (q->rate) {
 			struct sk_buff *last;
 
-			if (!skb_queue_empty(&sch->q))
+			if (sch->q.qlen)
 				last = skb_peek_tail(&sch->q);
 			else
 				last = netem_rb_to_skb(rb_last(&q->t_root));
-- 
2.7.3

[PATCH net-next 3/5] sched: remove qdisc arg from __qdisc_dequeue_head

[Florian Westphal]

Moves qdisc stat accouting to qdisc_dequeue_head.

The only direct caller of the __qdisc_dequeue_head version open-codes
this now.

This allows us to later use __qdisc_dequeue_head as a replacement
of __skb_dequeue() (which operates on sk_buff_head list).

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 include/net/sch_generic.h | 15 ++++++++-------
 net/sched/sch_generic.c   |  7 ++++++-
 2 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 52a2015..0741ed4 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -614,11 +614,17 @@ static inline int qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch)
 	return __qdisc_enqueue_tail(skb, sch, &sch->q);
 }
 
-static inline struct sk_buff *__qdisc_dequeue_head(struct Qdisc *sch,
-						   struct sk_buff_head *list)
+static inline struct sk_buff *__qdisc_dequeue_head(struct sk_buff_head *list)
 {
 	struct sk_buff *skb = __skb_dequeue(list);
 
+	return skb;
+}
+
+static inline struct sk_buff *qdisc_dequeue_head(struct Qdisc *sch)
+{
+	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
+
 	if (likely(skb != NULL)) {
 		qdisc_qstats_backlog_dec(sch, skb);
 		qdisc_bstats_update(sch, skb);
@@ -627,11 +633,6 @@ static inline struct sk_buff *__qdisc_dequeue_head(struct Qdisc *sch,
 	return skb;
 }
 
-static inline struct sk_buff *qdisc_dequeue_head(struct Qdisc *sch)
-{
-	return __qdisc_dequeue_head(sch, &sch->q);
-}
-
 /* Instead of calling kfree_skb() while root qdisc lock is held,
  * queue the skb for future freeing at end of __dev_xmit_skb()
  */
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 5e63bf6..73877d9 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -506,7 +506,12 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
 
 	if (likely(band >= 0)) {
 		struct sk_buff_head *list = band2list(priv, band);
-		struct sk_buff *skb = __qdisc_dequeue_head(qdisc, list);
+		struct sk_buff *skb = __qdisc_dequeue_head(list);
+
+		if (likely(skb != NULL)) {
+			qdisc_qstats_backlog_dec(qdisc, skb);
+			qdisc_bstats_update(qdisc, skb);
+		}
 
 		qdisc->q.qlen--;
 		if (skb_queue_empty(list))
-- 
2.7.3

[PATCH net-next 4/5] sched: replace __skb_dequeue with __qdisc_dequeue_head

[Florian Westphal]

After previous patch these functions are identical.
Replace __skb_dequeue in qdiscs with __qdisc_dequeue_head.

Next patch will then make __qdisc_dequeue_head handle
single-linked list instead of strcut sk_buff_head argument.

Doesn't change generated code.

Signed-off-by: Florian Westphal <fw@strlen.de>
---
 net/sched/sch_codel.c | 4 ++--
 net/sched/sch_netem.c | 2 +-
 net/sched/sch_pie.c   | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/net/sched/sch_codel.c b/net/sched/sch_codel.c
index 4002df3..5bfa79e 100644
--- a/net/sched/sch_codel.c
+++ b/net/sched/sch_codel.c
@@ -69,7 +69,7 @@ struct codel_sched_data {
 static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
 {
 	struct Qdisc *sch = ctx;
-	struct sk_buff *skb = __skb_dequeue(&sch->q);
+	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
 
 	if (skb)
 		sch->qstats.backlog -= qdisc_pkt_len(skb);
@@ -172,7 +172,7 @@ static int codel_change(struct Qdisc *sch, struct nlattr *opt)
 
 	qlen = sch->q.qlen;
 	while (sch->q.qlen > sch->limit) {
-		struct sk_buff *skb = __skb_dequeue(&sch->q);
+		struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
 
 		dropped += qdisc_pkt_len(skb);
 		qdisc_qstats_backlog_dec(sch, skb);
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 1832d77..0a964b3 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -587,7 +587,7 @@ static struct sk_buff *netem_dequeue(struct Qdisc *sch)
 	struct rb_node *p;
 
 tfifo_dequeue:
-	skb = __skb_dequeue(&sch->q);
+	skb = __qdisc_dequeue_head(&sch->q);
 	if (skb) {
 		qdisc_qstats_backlog_dec(sch, skb);
 deliver:
diff --git a/net/sched/sch_pie.c b/net/sched/sch_pie.c
index d976d74..5c3a99d 100644
--- a/net/sched/sch_pie.c
+++ b/net/sched/sch_pie.c
@@ -231,7 +231,7 @@ static int pie_change(struct Qdisc *sch, struct nlattr *opt)
 	/* Drop excess packets if new limit is lower */
 	qlen = sch->q.qlen;
 	while (sch->q.qlen > sch->limit) {
-		struct sk_buff *skb = __skb_dequeue(&sch->q);
+		struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
 
 		dropped += qdisc_pkt_len(skb);
 		qdisc_qstats_backlog_dec(sch, skb);
-- 
2.7.3

[PATCH net-next 5/5] sched: add and use qdisc_skb_head helpers

[Florian Westphal]

This change replaces sk_buff_head struct in Qdiscs with new qdisc_skb_head.

Its similar to the skb_buff_head api, but does not use skb->prev pointers.

Qdiscs will commonly enqueue at the tail of a list and dequeue at head.
While skb_buff_head works fine for this, enqueue/dequeue needs to also
adjust the prev pointer of next element.

The ->prev pointer is not required for qdiscs so we can just leave
it undefined and avoid one cacheline write access for en/dequeue.

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 include/net/sch_generic.h | 63 ++++++++++++++++++++++++++++++++++++++---------
 net/sched/sch_generic.c   | 21 ++++++++--------
 net/sched/sch_htb.c       | 24 +++++++++++++++---
 net/sched/sch_netem.c     | 14 +++++++++--
 4 files changed, 94 insertions(+), 28 deletions(-)

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 0741ed4..e6aa0a2 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -36,6 +36,14 @@ struct qdisc_size_table {
 	u16			data[];
 };
 
+/* similar to sk_buff_head, but skb->prev pointer is undefined. */
+struct qdisc_skb_head {
+	struct sk_buff	*head;
+	struct sk_buff	*tail;
+	__u32		qlen;
+	spinlock_t	lock;
+};
+
 struct Qdisc {
 	int 			(*enqueue)(struct sk_buff *skb,
 					   struct Qdisc *sch,
@@ -76,7 +84,7 @@ struct Qdisc {
 	 * For performance sake on SMP, we put highly modified fields at the end
 	 */
 	struct sk_buff		*gso_skb ____cacheline_aligned_in_smp;
-	struct sk_buff_head	q;
+	struct qdisc_skb_head	q;
 	struct gnet_stats_basic_packed bstats;
 	seqcount_t		running;
 	struct gnet_stats_queue	qstats;
@@ -600,10 +608,27 @@ static inline void qdisc_qstats_overlimit(struct Qdisc *sch)
 	sch->qstats.overlimits++;
 }
 
+static inline void qdisc_skb_head_init(struct qdisc_skb_head *qh)
+{
+	qh->head = NULL;
+	qh->tail = NULL;
+	qh->qlen = 0;
+}
+
 static inline int __qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,
-				       struct sk_buff_head *list)
+				       struct qdisc_skb_head *qh)
 {
-	__skb_queue_tail(list, skb);
+	struct sk_buff *last = qh->tail;
+
+	if (last) {
+		skb->next = NULL;
+		last->next = skb;
+		qh->tail = skb;
+	} else {
+		qh->tail = skb;
+		qh->head = skb;
+	}
+	qh->qlen++;
 	qdisc_qstats_backlog_inc(sch, skb);
 
 	return NET_XMIT_SUCCESS;
@@ -614,9 +639,17 @@ static inline int qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch)
 	return __qdisc_enqueue_tail(skb, sch, &sch->q);
 }
 
-static inline struct sk_buff *__qdisc_dequeue_head(struct sk_buff_head *list)
+static inline struct sk_buff *__qdisc_dequeue_head(struct qdisc_skb_head *qh)
 {
-	struct sk_buff *skb = __skb_dequeue(list);
+	struct sk_buff *skb = qh->head;
+
+	if (likely(skb != NULL)) {
+		qh->head = skb->next;
+		qh->qlen--;
+		if (qh->head == NULL)
+			qh->tail = NULL;
+		skb->next = NULL;
+	}
 
 	return skb;
 }
@@ -643,10 +676,10 @@ static inline void __qdisc_drop(struct sk_buff *skb, struct sk_buff **to_free)
 }
 
 static inline unsigned int __qdisc_queue_drop_head(struct Qdisc *sch,
-						   struct sk_buff_head *list,
+						   struct qdisc_skb_head *qh,
 						   struct sk_buff **to_free)
 {
-	struct sk_buff *skb = __skb_dequeue(list);
+	struct sk_buff *skb = __qdisc_dequeue_head(qh);
 
 	if (likely(skb != NULL)) {
 		unsigned int len = qdisc_pkt_len(skb);
@@ -667,7 +700,9 @@ static inline unsigned int qdisc_queue_drop_head(struct Qdisc *sch,
 
 static inline struct sk_buff *qdisc_peek_head(struct Qdisc *sch)
 {
-	return skb_peek(&sch->q);
+	const struct qdisc_skb_head *qh = &sch->q;
+
+	return qh->head;
 }
 
 /* generic pseudo peek method for non-work-conserving qdisc */
@@ -702,15 +737,19 @@ static inline struct sk_buff *qdisc_dequeue_peeked(struct Qdisc *sch)
 	return skb;
 }
 
-static inline void __qdisc_reset_queue(struct sk_buff_head *list)
+static inline void __qdisc_reset_queue(struct qdisc_skb_head *qh)
 {
 	/*
 	 * We do not know the backlog in bytes of this list, it
 	 * is up to the caller to correct it
 	 */
-	if (!skb_queue_empty(list)) {
-		rtnl_kfree_skbs(list->next, list->prev);
-		__skb_queue_head_init(list);
+	ASSERT_RTNL();
+	if (qh->qlen) {
+		rtnl_kfree_skbs(qh->head, qh->tail);
+
+		qh->head = NULL;
+		qh->tail = NULL;
+		qh->qlen = 0;
 	}
 }
 
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 73877d9..6cfb6e9 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -466,7 +466,7 @@ static const u8 prio2band[TC_PRIO_MAX + 1] = {
  */
 struct pfifo_fast_priv {
 	u32 bitmap;
-	struct sk_buff_head q[PFIFO_FAST_BANDS];
+	struct qdisc_skb_head q[PFIFO_FAST_BANDS];
 };
 
 /*
@@ -477,7 +477,7 @@ struct pfifo_fast_priv {
  */
 static const int bitmap2band[] = {-1, 0, 1, 0, 2, 0, 1, 0};
 
-static inline struct sk_buff_head *band2list(struct pfifo_fast_priv *priv,
+static inline struct qdisc_skb_head *band2list(struct pfifo_fast_priv *priv,
 					     int band)
 {
 	return priv->q + band;
@@ -489,7 +489,7 @@ static int pfifo_fast_enqueue(struct sk_buff *skb, struct Qdisc *qdisc,
 	if (qdisc->q.qlen < qdisc_dev(qdisc)->tx_queue_len) {
 		int band = prio2band[skb->priority & TC_PRIO_MAX];
 		struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
-		struct sk_buff_head *list = band2list(priv, band);
+		struct qdisc_skb_head *list = band2list(priv, band);
 
 		priv->bitmap |= (1 << band);
 		qdisc->q.qlen++;
@@ -505,8 +505,8 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
 	int band = bitmap2band[priv->bitmap];
 
 	if (likely(band >= 0)) {
-		struct sk_buff_head *list = band2list(priv, band);
-		struct sk_buff *skb = __qdisc_dequeue_head(list);
+		struct qdisc_skb_head *qh = band2list(priv, band);
+		struct sk_buff *skb = __qdisc_dequeue_head(qh);
 
 		if (likely(skb != NULL)) {
 			qdisc_qstats_backlog_dec(qdisc, skb);
@@ -514,7 +514,7 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
 		}
 
 		qdisc->q.qlen--;
-		if (skb_queue_empty(list))
+		if (qh->qlen == 0)
 			priv->bitmap &= ~(1 << band);
 
 		return skb;
@@ -529,9 +529,9 @@ static struct sk_buff *pfifo_fast_peek(struct Qdisc *qdisc)
 	int band = bitmap2band[priv->bitmap];
 
 	if (band >= 0) {
-		struct sk_buff_head *list = band2list(priv, band);
+		struct qdisc_skb_head *qh = band2list(priv, band);
 
-		return skb_peek(list);
+		return qh->head;
 	}
 
 	return NULL;
@@ -569,7 +569,7 @@ static int pfifo_fast_init(struct Qdisc *qdisc, struct nlattr *opt)
 	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
 
 	for (prio = 0; prio < PFIFO_FAST_BANDS; prio++)
-		__skb_queue_head_init(band2list(priv, prio));
+		qdisc_skb_head_init(band2list(priv, prio));
 
 	/* Can by-pass the queue discipline */
 	qdisc->flags |= TCQ_F_CAN_BYPASS;
@@ -617,7 +617,8 @@ struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
 		sch = (struct Qdisc *) QDISC_ALIGN((unsigned long) p);
 		sch->padded = (char *) sch - (char *) p;
 	}
-	skb_queue_head_init(&sch->q);
+	qdisc_skb_head_init(&sch->q);
+	spin_lock_init(&sch->q.lock);
 
 	spin_lock_init(&sch->busylock);
 	lockdep_set_class(&sch->busylock,
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index 53dbfa1..c798d0d 100644
--- a/net/sched/sch_htb.c
+++ b/net/sched/sch_htb.c
@@ -162,7 +162,7 @@ struct htb_sched {
 	struct work_struct	work;
 
 	/* non shaped skbs; let them go directly thru */
-	struct sk_buff_head	direct_queue;
+	struct qdisc_skb_head	direct_queue;
 	long			direct_pkts;
 
 	struct qdisc_watchdog	watchdog;
@@ -570,6 +570,22 @@ static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
 	list_del_init(&cl->un.leaf.drop_list);
 }
 
+static void htb_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,
+			     struct qdisc_skb_head *qh)
+{
+	struct sk_buff *last = qh->tail;
+
+	if (last) {
+		skb->next = NULL;
+		last->next = skb;
+		qh->tail = skb;
+	} else {
+		qh->tail = skb;
+		qh->head = skb;
+	}
+	qh->qlen++;
+}
+
 static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		       struct sk_buff **to_free)
 {
@@ -580,7 +596,7 @@ static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	if (cl == HTB_DIRECT) {
 		/* enqueue to helper queue */
 		if (q->direct_queue.qlen < q->direct_qlen) {
-			__skb_queue_tail(&q->direct_queue, skb);
+			htb_enqueue_tail(skb, sch, &q->direct_queue);
 			q->direct_pkts++;
 		} else {
 			return qdisc_drop(skb, sch, to_free);
@@ -888,7 +904,7 @@ static struct sk_buff *htb_dequeue(struct Qdisc *sch)
 	unsigned long start_at;
 
 	/* try to dequeue direct packets as high prio (!) to minimize cpu work */
-	skb = __skb_dequeue(&q->direct_queue);
+	skb = __qdisc_dequeue_head(&q->direct_queue);
 	if (skb != NULL) {
 ok:
 		qdisc_bstats_update(sch, skb);
@@ -1019,7 +1035,7 @@ static int htb_init(struct Qdisc *sch, struct nlattr *opt)
 
 	qdisc_watchdog_init(&q->watchdog, sch);
 	INIT_WORK(&q->work, htb_work_func);
-	__skb_queue_head_init(&q->direct_queue);
+	qdisc_skb_head_init(&q->direct_queue);
 
 	if (tb[TCA_HTB_DIRECT_QLEN])
 		q->direct_qlen = nla_get_u32(tb[TCA_HTB_DIRECT_QLEN]);
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 0a964b3..9f7b380 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -413,6 +413,16 @@ static struct sk_buff *netem_segment(struct sk_buff *skb, struct Qdisc *sch,
 	return segs;
 }
 
+static void netem_enqueue_skb_head(struct qdisc_skb_head *qh, struct sk_buff *skb)
+{
+	skb->next = qh->head;
+
+	if (!qh->head)
+		qh->tail = skb;
+	qh->head = skb;
+	qh->qlen++;
+}
+
 /*
  * Insert one skb into qdisc.
  * Note: parent depends on return value to account for queue length.
@@ -523,7 +533,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			struct sk_buff *last;
 
 			if (sch->q.qlen)
-				last = skb_peek_tail(&sch->q);
+				last = sch->q.tail;
 			else
 				last = netem_rb_to_skb(rb_last(&q->t_root));
 			if (last) {
@@ -552,7 +562,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		cb->time_to_send = psched_get_time();
 		q->counter = 0;
 
-		__skb_queue_head(&sch->q, skb);
+		netem_enqueue_skb_head(&sch->q, skb);
 		sch->qstats.requeues++;
 	}
 
-- 
2.7.3

Re: [PATCH net-next 0/5] sched: convert queues to single-linked list

[David Miller]

From: Florian Westphal <fw@strlen.de>
Date: Sun, 18 Sep 2016 00:57:29 +0200

> During Netfilter Workshop 2016 Eric Dumazet pointed out that qdisc
> schedulers use doubly-linked lists, even though single-linked list
> would be enough.
> 
> The double-linked skb lists incur one extra write on enqueue/dequeue
> operations (to change ->prev pointer of next list elem).
> 
> This series converts qdiscs to single-linked version, listhead
> maintains pointers to first (for dequeue) and last skb (for enqueue).
> 
> Most qdiscs don't queue at all and instead use a leaf qdisc (typically
> pfifo_fast) so only a few schedulers needed changes.
> 
> I briefly tested netem and htb and they seemed fine.
> 
> UDP_STREAM netperf with 64 byte packets via veth+pfifo_fast shows
> a small (~2%) improvement.

Series applied, thanks Florian.