net_sched: make room for (struct qdisc_skb_cb)->pkt_segs

Add a new u16 field, next to pkt_len : pkt_segs

This will cache shinfo->gso_segs to speed up qdisc deqeue().

Move slave_dev_queue_mapping at the end of qdisc_skb_cb,
and move three bits from tc_skb_cb :
- post_ct
- post_ct_snat
- post_ct_dnat

Signed-off-by: Eric Dumazet <[e-mail]>
Link: https://patch.msgid.link/[e-mail]
Signed-off-by: Paolo Abeni <[e-mail]>

include/net/sch_generic.h

···
 };
 
 struct qdisc_skb_cb {
-	struct {
-		unsigned int		pkt_len;
-		u16			slave_dev_queue_mapping;
-		u16			tc_classid;
-	};
+	unsigned int		pkt_len;
+	u16			pkt_segs;
+	u16			tc_classid;
 #define QDISC_CB_PRIV_LEN 20
 	unsigned char		data[QDISC_CB_PRIV_LEN];
+
+	u16			slave_dev_queue_mapping;
+	u8			post_ct:1;
+	u8			post_ct_snat:1;
+	u8			post_ct_dnat:1;
 };
 
 typedef void tcf_chain_head_change_t(struct tcf_proto *tp_head, void *priv);
···
 	struct qdisc_skb_cb qdisc_cb;
 	u32 drop_reason;
 
-	u16 zone; /* Only valid if post_ct = true */
+	u16 zone; /* Only valid if qdisc_skb_cb(skb)->post_ct = true */
 	u16 mru;
-	u8 post_ct:1;
-	u8 post_ct_snat:1;
-	u8 post_ct_dnat:1;
 };
 
 static inline struct tc_skb_cb *tc_skb_cb(const struct sk_buff *skb)

net/core/dev.c

···
 		return ret;
 
 	tc_skb_cb(skb)->mru = 0;
-	tc_skb_cb(skb)->post_ct = false;
+	qdisc_skb_cb(skb)->post_ct = false;
 	tcf_set_drop_reason(skb, *drop_reason);
 
 	mini_qdisc_bstats_cpu_update(miniq, skb);

net/sched/act_ct.c

···
 		return err & NF_VERDICT_MASK;
 
 	if (action & BIT(NF_NAT_MANIP_SRC))
-		tc_skb_cb(skb)->post_ct_snat = 1;
+		qdisc_skb_cb(skb)->post_ct_snat = 1;
 	if (action & BIT(NF_NAT_MANIP_DST))
-		tc_skb_cb(skb)->post_ct_dnat = 1;
+		qdisc_skb_cb(skb)->post_ct_dnat = 1;
 
 	return err;
 #else
···
 	tcf_action_update_bstats(&c->common, skb);
 
 	if (clear) {
-		tc_skb_cb(skb)->post_ct = false;
+		qdisc_skb_cb(skb)->post_ct = false;
 		ct = nf_ct_get(skb, &ctinfo);
 		if (ct) {
 			nf_ct_put(ct);
···
 out_push:
 	skb_push_rcsum(skb, nh_ofs);
 
-	tc_skb_cb(skb)->post_ct = true;
+	qdisc_skb_cb(skb)->post_ct = true;
 	tc_skb_cb(skb)->zone = p->zone;
 out_clear:
 	if (defrag)

net/sched/cls_api.c

···
 			}
 			ext->chain = last_executed_chain;
 			ext->mru = cb->mru;
-			ext->post_ct = cb->post_ct;
-			ext->post_ct_snat = cb->post_ct_snat;
-			ext->post_ct_dnat = cb->post_ct_dnat;
+			ext->post_ct = qdisc_skb_cb(skb)->post_ct;
+			ext->post_ct_snat = qdisc_skb_cb(skb)->post_ct_snat;
+			ext->post_ct_dnat = qdisc_skb_cb(skb)->post_ct_dnat;
 			ext->zone = cb->zone;
 		}
 	}

net/sched/cls_flower.c

···
 				  struct tcf_result *res)
 {
 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
-	bool post_ct = tc_skb_cb(skb)->post_ct;
+	bool post_ct = qdisc_skb_cb(skb)->post_ct;
 	u16 zone = tc_skb_cb(skb)->zone;
 	struct fl_flow_key skb_key;
 	struct fl_flow_mask *mask;