Avoid up to two cache line misses in qdisc dequeue() to fetch
skb_shinfo(skb)->gso_segs/gso_size while qdisc spinlock is held.

This gives a 5 % improvement in a TX intensive workload.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 include/net/sch_generic.h | 13 ++++++++++---
 net/sched/sch_cake.c      |  1 +
 net/sched/sch_dualpi2.c   |  1 +
 net/sched/sch_netem.c     |  1 +
 net/sched/sch_qfq.c       |  2 +-
 net/sched/sch_taprio.c    |  1 +
 net/sched/sch_tbf.c       |  1 +
 7 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 9cd8b5d4b23698fd8959ef40c303468e31c1d4af..ae037e56088208ad17f664c43689aee895569cab 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -829,6 +829,15 @@ static inline unsigned int qdisc_pkt_len(const struct sk_buff *skb)
 	return qdisc_skb_cb(skb)->pkt_len;
 }
 
+static inline unsigned int qdisc_pkt_segs(const struct sk_buff *skb)
+{
+	u32 pkt_segs = qdisc_skb_cb(skb)->pkt_segs;
+
+	DEBUG_NET_WARN_ON_ONCE(pkt_segs !=
+			       skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1);
+	return pkt_segs;
+}
+
 /* additional qdisc xmit flags (NET_XMIT_MASK in linux/netdevice.h) */
 enum net_xmit_qdisc_t {
 	__NET_XMIT_STOLEN = 0x00010000,
@@ -870,9 +879,7 @@ static inline void _bstats_update(struct gnet_stats_basic_sync *bstats,
 static inline void bstats_update(struct gnet_stats_basic_sync *bstats,
 				 const struct sk_buff *skb)
 {
-	_bstats_update(bstats,
-		       qdisc_pkt_len(skb),
-		       skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1);
+	_bstats_update(bstats, qdisc_pkt_len(skb), qdisc_pkt_segs(skb));
 }
 
 static inline void qdisc_bstats_cpu_update(struct Qdisc *sch,
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 32bacfc314c260dccf94178d309ccb2be22d69e4..993ce808230fb7d4769c926f6c8368d927f5a45f 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -1800,6 +1800,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		skb_list_walk_safe(segs, segs, nskb) {
 			skb_mark_not_on_list(segs);
 			qdisc_skb_cb(segs)->pkt_len = segs->len;
+			qdisc_skb_cb(segs)->pkt_segs = 1;
 			cobalt_set_enqueue_time(segs, now);
 			get_cobalt_cb(segs)->adjusted_len = cake_overhead(q,
 									  segs);
diff --git a/net/sched/sch_dualpi2.c b/net/sched/sch_dualpi2.c
index 4b975feb52b1f3d3b37b31713d1477de5f5806d9..6d7e6389758dc8e645b1116efe4e11fb7290ac86 100644
--- a/net/sched/sch_dualpi2.c
+++ b/net/sched/sch_dualpi2.c
@@ -475,6 +475,7 @@ static int dualpi2_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			 * (3) Enqueue fragment & set ts in dualpi2_enqueue_skb
 			 */
 			qdisc_skb_cb(nskb)->pkt_len = nskb->len;
+			qdisc_skb_cb(nskb)->pkt_segs = 1;
 			dualpi2_skb_cb(nskb)->classified =
 				dualpi2_skb_cb(skb)->classified;
 			dualpi2_skb_cb(nskb)->ect = dualpi2_skb_cb(skb)->ect;
diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index eafc316ae319e3f8c23b0cb0c58fdf54be102213..32a5f33040461f3be952055c097b5f2fe760a858 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -429,6 +429,7 @@ static struct sk_buff *netem_segment(struct sk_buff *skb, struct Qdisc *sch,
 	struct sk_buff *segs;
 	netdev_features_t features = netif_skb_features(skb);
 
+	qdisc_skb_cb(skb)->pkt_segs = 1;
 	segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK);
 
 	if (IS_ERR_OR_NULL(segs)) {
diff --git a/net/sched/sch_qfq.c b/net/sched/sch_qfq.c
index 2255355e51d350eded4549c1584b60d4d9b00fff..d920f57dc6d7659c510a98956c6dd2ed9e5ee5b8 100644
--- a/net/sched/sch_qfq.c
+++ b/net/sched/sch_qfq.c
@@ -1250,7 +1250,7 @@ static int qfq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		}
 	}
 
-	gso_segs = skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1;
+	gso_segs = qdisc_pkt_segs(skb);
 	err = qdisc_enqueue(skb, cl->qdisc, to_free);
 	if (unlikely(err != NET_XMIT_SUCCESS)) {
 		pr_debug("qfq_enqueue: enqueue failed %d\n", err);
diff --git a/net/sched/sch_taprio.c b/net/sched/sch_taprio.c
index 39b735386996eb59712a1fc28f7bb903ec1b2220..300d577b328699eb42d2b829ecfc76464fd7b186 100644
--- a/net/sched/sch_taprio.c
+++ b/net/sched/sch_taprio.c
@@ -595,6 +595,7 @@ static int taprio_enqueue_segmented(struct sk_buff *skb, struct Qdisc *sch,
 	skb_list_walk_safe(segs, segs, nskb) {
 		skb_mark_not_on_list(segs);
 		qdisc_skb_cb(segs)->pkt_len = segs->len;
+		qdisc_skb_cb(segs)->pkt_segs = 1;
 		slen += segs->len;
 
 		/* FIXME: we should be segmenting to a smaller size
diff --git a/net/sched/sch_tbf.c b/net/sched/sch_tbf.c
index 4c977f049670a600eafd219c898e5f29597be2c1..f2340164f579a25431979e12ec3d23ab828edd16 100644
--- a/net/sched/sch_tbf.c
+++ b/net/sched/sch_tbf.c
@@ -221,6 +221,7 @@ static int tbf_segment(struct sk_buff *skb, struct Qdisc *sch,
 		skb_mark_not_on_list(segs);
 		seg_len = segs->len;
 		qdisc_skb_cb(segs)->pkt_len = seg_len;
+		qdisc_skb_cb(segs)->pkt_segs = 1;
 		ret = qdisc_enqueue(segs, q->qdisc, to_free);
 		if (ret != NET_XMIT_SUCCESS) {
 			if (net_xmit_drop_count(ret))
-- 
2.51.2.1041.gc1ab5b90ca-goog