net_sched: Add qdisc __NET_XMIT_STOLEN flag

Patrick McHardy <[e-mail]> noticed:
"The other problem that affects all qdiscs supporting actions is
TC_ACT_QUEUED/TC_ACT_STOLEN getting mapped to NET_XMIT_SUCCESS
even though the packet is not queued, corrupting upper qdiscs'
qlen counters."

and later explained:
"The reason why it translates it at all seems to be to not increase
the drops counter. Within a single qdisc this could be avoided by
other means easily, upper qdiscs would still increase the counter
when we return anything besides NET_XMIT_SUCCESS though.

This means we need a new NET_XMIT return value to indicate this to
the upper qdiscs. So I'd suggest to introduce NET_XMIT_STOLEN,
return that to upper qdiscs and translate it to NET_XMIT_SUCCESS
in dev_queue_xmit, similar to NET_XMIT_BYPASS."

David Miller <[e-mail]> noticed:
"Maybe these NET_XMIT_* values being passed around should be a set of
bits. They could be composed of base meanings, combined with specific
attributes.

So you could say "NET_XMIT_DROP | __NET_XMIT_NO_DROP_COUNT"

The attributes get masked out by the top-level ->enqueue() caller,
such that the base meanings are the only thing that make their
way up into the stack. If it's only about communication within the
qdisc tree, let's simply code it that way."

This patch is trying to realize these ideas.

Signed-off-by: Jarek Poplawski <[e-mail]>
Signed-off-by: David S. Miller <[e-mail]>

include/linux/netdevice.h

···
 #define NET_XMIT_BYPASS		4	/* packet does not leave via dequeue;
 					   (TC use only - dev_queue_xmit
 					   returns this as NET_XMIT_SUCCESS) */
+#define NET_XMIT_MASK		0xFFFF	/* qdisc flags in net/sch_generic.h */
 
 /* Backlog congestion levels */
 #define NET_RX_SUCCESS		0   /* keep 'em coming, baby */

include/net/sch_generic.h

···
 	return qdisc_skb_cb(skb)->pkt_len;
 }
 
+#ifdef CONFIG_NET_CLS_ACT
+/* additional qdisc xmit flags */
+enum net_xmit_qdisc_t {
+	__NET_XMIT_STOLEN = 0x00010000,
+};
+
+#define net_xmit_drop_count(e)	((e) & __NET_XMIT_STOLEN ? 0 : 1)
+
+#else
+#define net_xmit_drop_count(e)	(1)
+#endif
+
 static inline int qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 {
 #ifdef CONFIG_NET_SCHED
···
 static inline int qdisc_enqueue_root(struct sk_buff *skb, struct Qdisc *sch)
 {
 	qdisc_skb_cb(skb)->pkt_len = skb->len;
-	return qdisc_enqueue(skb, sch);
+	return qdisc_enqueue(skb, sch) & NET_XMIT_MASK;
 }
 
 static inline int __qdisc_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,

net/sched/sch_atm.c

···
 		case TC_ACT_QUEUED:
 		case TC_ACT_STOLEN:
 			kfree_skb(skb);
-			return NET_XMIT_SUCCESS;
+			return NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			kfree_skb(skb);
 			goto drop;
···
 	ret = qdisc_enqueue(skb, flow->q);
 	if (ret != 0) {
 drop: __maybe_unused
-		sch->qstats.drops++;
-		if (flow)
-			flow->qstats.drops++;
+		if (net_xmit_drop_count(ret)) {
+			sch->qstats.drops++;
+			if (flow)
+				flow->qstats.drops++;
+		}
 		return ret;
 	}
 	sch->bstats.bytes += qdisc_pkt_len(skb);
···
 	if (!ret) {
 		sch->q.qlen++;
 		sch->qstats.requeues++;
-	} else {
+	} else if (net_xmit_drop_count(ret)) {
 		sch->qstats.drops++;
 		p->link.qstats.drops++;
 	}

net/sched/sch_cbq.c

···
 		switch (result) {
 		case TC_ACT_QUEUED:
 		case TC_ACT_STOLEN:
-			*qerr = NET_XMIT_SUCCESS;
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			return NULL;
 		case TC_ACT_RECLASSIFY:
···
 		return ret;
 	}
 
-	sch->qstats.drops++;
-	cbq_mark_toplevel(q, cl);
-	cl->qstats.drops++;
+	if (net_xmit_drop_count(ret)) {
+		sch->qstats.drops++;
+		cbq_mark_toplevel(q, cl);
+		cl->qstats.drops++;
+	}
 	return ret;
 }
 
···
 			cbq_activate_class(cl);
 		return 0;
 	}
-	sch->qstats.drops++;
-	cl->qstats.drops++;
+	if (net_xmit_drop_count(ret)) {
+		sch->qstats.drops++;
+		cl->qstats.drops++;
+	}
 	return ret;
 }
 
···
 	q->rx_class = NULL;
 
 	if (cl && (cl = cbq_reclassify(skb, cl)) != NULL) {
+		int ret;
 
 		cbq_mark_toplevel(q, cl);
 
 		q->rx_class = cl;
 		cl->q->__parent = sch;
 
-		if (qdisc_enqueue(skb, cl->q) == 0) {
+		ret = qdisc_enqueue(skb, cl->q);
+		if (ret == NET_XMIT_SUCCESS) {
 			sch->q.qlen++;
 			sch->bstats.packets++;
 			sch->bstats.bytes += qdisc_pkt_len(skb);
···
 				cbq_activate_class(cl);
 			return 0;
 		}
-		sch->qstats.drops++;
+		if (net_xmit_drop_count(ret))
+			sch->qstats.drops++;
 		return 0;
 	}
 

net/sched/sch_dsmark.c

···
 		case TC_ACT_QUEUED:
 		case TC_ACT_STOLEN:
 			kfree_skb(skb);
-			return NET_XMIT_SUCCESS;
+			return NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 
 		case TC_ACT_SHOT:
 			goto drop;
···
 
 	err = qdisc_enqueue(skb, p->q);
 	if (err != NET_XMIT_SUCCESS) {
-		sch->qstats.drops++;
+		if (net_xmit_drop_count(err))
+			sch->qstats.drops++;
 		return err;
 	}
 
···
 
 	err = p->q->ops->requeue(skb, p->q);
 	if (err != NET_XMIT_SUCCESS) {
-		sch->qstats.drops++;
+		if (net_xmit_drop_count(err))
+			sch->qstats.drops++;
 		return err;
 	}
 

net/sched/sch_hfsc.c

···
 		switch (result) {
 		case TC_ACT_QUEUED:
 		case TC_ACT_STOLEN:
-			*qerr = NET_XMIT_SUCCESS;
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			return NULL;
 		}
···
 
 	err = qdisc_enqueue(skb, cl->qdisc);
 	if (unlikely(err != NET_XMIT_SUCCESS)) {
-		cl->qstats.drops++;
-		sch->qstats.drops++;
+		if (net_xmit_drop_count(err)) {
+			cl->qstats.drops++;
+			sch->qstats.drops++;
+		}
 		return err;
 	}
 

net/sched/sch_htb.c

···
 		switch (result) {
 		case TC_ACT_QUEUED:
 		case TC_ACT_STOLEN:
-			*qerr = NET_XMIT_SUCCESS;
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			return NULL;
 		}
···
 		kfree_skb(skb);
 		return ret;
 #endif
-	} else if (qdisc_enqueue(skb, cl->un.leaf.q) != NET_XMIT_SUCCESS) {
-		sch->qstats.drops++;
-		cl->qstats.drops++;
+	} else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q)) != NET_XMIT_SUCCESS) {
+		if (net_xmit_drop_count(ret)) {
+			sch->qstats.drops++;
+			cl->qstats.drops++;
+		}
 		return NET_XMIT_DROP;
 	} else {
 		cl->bstats.packets +=
···
 		kfree_skb(skb);
 		return ret;
 #endif
-	} else if (cl->un.leaf.q->ops->requeue(skb, cl->un.leaf.q) !=
+	} else if ((ret = cl->un.leaf.q->ops->requeue(skb, cl->un.leaf.q)) !=
 		   NET_XMIT_SUCCESS) {
-		sch->qstats.drops++;
-		cl->qstats.drops++;
+		if (net_xmit_drop_count(ret)) {
+			sch->qstats.drops++;
+			cl->qstats.drops++;
+		}
 		return NET_XMIT_DROP;
 	} else
 		htb_activate(q, cl);

net/sched/sch_netem.c

···
 		sch->q.qlen++;
 		sch->bstats.bytes += qdisc_pkt_len(skb);
 		sch->bstats.packets++;
-	} else
+	} else if (net_xmit_drop_count(ret)) {
 		sch->qstats.drops++;
+	}
 
 	pr_debug("netem: enqueue ret %d\n", ret);
 	return ret;

net/sched/sch_prio.c

···
 		switch (err) {
 		case TC_ACT_STOLEN:
 		case TC_ACT_QUEUED:
-			*qerr = NET_XMIT_SUCCESS;
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			return NULL;
 		}
···
 		sch->q.qlen++;
 		return NET_XMIT_SUCCESS;
 	}
-	sch->qstats.drops++;
+	if (net_xmit_drop_count(ret))
+		sch->qstats.drops++;
 	return ret;
 }
 
···
 		sch->qstats.requeues++;
 		return 0;
 	}
-	sch->qstats.drops++;
+	if (net_xmit_drop_count(ret))
+		sch->qstats.drops++;
 	return NET_XMIT_DROP;
 }
 

net/sched/sch_red.c

···
 		sch->bstats.bytes += qdisc_pkt_len(skb);
 		sch->bstats.packets++;
 		sch->q.qlen++;
-	} else {
+	} else if (net_xmit_drop_count(ret)) {
 		q->stats.pdrop++;
 		sch->qstats.drops++;
 	}

net/sched/sch_sfq.c

···
 		switch (result) {
 		case TC_ACT_STOLEN:
 		case TC_ACT_QUEUED:
-			*qerr = NET_XMIT_SUCCESS;
+			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 		case TC_ACT_SHOT:
 			return 0;
 		}

net/sched/sch_tbf.c

···
 
 	ret = qdisc_enqueue(skb, q->qdisc);
 	if (ret != 0) {
-		sch->qstats.drops++;
+		if (net_xmit_drop_count(ret))
+			sch->qstats.drops++;
 		return ret;
 	}