[NET] Attach inner qdiscs to TBF

include/net/pkt_sched.h

@@ -61,7 +61,7 @@ struct Qdisc_ops
 	int 			(*enqueue)(struct sk_buff *, struct Qdisc *);
 	struct sk_buff *	(*dequeue)(struct Qdisc *);
 	int 			(*requeue)(struct sk_buff *, struct Qdisc *);
-	int			(*drop)(struct Qdisc *);
+	unsigned int		(*drop)(struct Qdisc *);
 
 	int			(*init)(struct Qdisc *, struct rtattr *arg);
 	void			(*reset)(struct Qdisc *);

net/netsyms.c

@@ -641,6 +641,7 @@ EXPORT_SYMBOL(qdisc_tree_lock);
 #ifdef CONFIG_NET_SCHED
 PSCHED_EXPORTLIST;
 EXPORT_SYMBOL(pfifo_qdisc_ops);
+EXPORT_SYMBOL(bfifo_qdisc_ops);
 EXPORT_SYMBOL(register_qdisc);
 EXPORT_SYMBOL(unregister_qdisc);
 EXPORT_SYMBOL(qdisc_get_rtab);

net/sched/sch_atm.c

@@ -545,15 +545,16 @@ static int atm_tc_requeue(struct sk_buff *skb,struct Qdisc *sch)
 }
 
 
-static int atm_tc_drop(struct Qdisc *sch)
+static unsigned int atm_tc_drop(struct Qdisc *sch)
 {
 	struct atm_qdisc_data *p = PRIV(sch);
 	struct atm_flow_data *flow;
+	unsigned int len;
 
 	DPRINTK("atm_tc_drop(sch %p,[qdisc %p])\n",sch,p);
 	for (flow = p->flows; flow; flow = flow->next)
-		if (flow->q->ops->drop && flow->q->ops->drop(flow->q))
-			return 1;
+		if (flow->q->ops->drop && (len = flow->q->ops->drop(flow->q)))
+			return len;
 	return 0;
 }
 

net/sched/sch_cbq.c

@@ -1231,11 +1231,12 @@ static void cbq_link_class(struct cbq_class *this)
 	}
 }
 
-static int cbq_drop(struct Qdisc* sch)
+static unsigned int cbq_drop(struct Qdisc* sch)
 {
 	struct cbq_sched_data *q = (struct cbq_sched_data *)sch->data;
 	struct cbq_class *cl, *cl_head;
 	int prio;
+	unsigned int len;
 
 	for (prio = TC_CBQ_MAXPRIO; prio >= 0; prio--) {
 		if ((cl_head = q->active[prio]) == NULL)
@@ -1243,9 +1244,9 @@ static int cbq_drop(struct Qdisc* sch)
 
 		cl = cl_head;
 		do {
-			if (cl->q->ops->drop && cl->q->ops->drop(cl->q)) {
+			if (cl->q->ops->drop && (len = cl->q->ops->drop(cl->q))) {
 				sch->q.qlen--;
-				return 1;
+				return len;
 			}
 		} while ((cl = cl->next_alive) != cl_head);
 	}

net/sched/sch_dsmark.c

@@ -302,17 +302,18 @@ static int dsmark_requeue(struct sk_buff *skb,struct Qdisc *sch)
 }
 
 
-static int dsmark_drop(struct Qdisc *sch)
+static unsigned int dsmark_drop(struct Qdisc *sch)
 {
 	struct dsmark_qdisc_data *p = PRIV(sch);
+	unsigned int len;
 	
 	DPRINTK("dsmark_reset(sch %p,[qdisc %p])\n",sch,p);
 	if (!p->q->ops->drop)
 		return 0;
-	if (!p->q->ops->drop(p->q))
+	if (!(len = p->q->ops->drop(p->q)))
 		return 0;
 	sch->q.qlen--;
-	return 1;
+	return len;
 }
 
 

net/sched/sch_fifo.c

@@ -81,16 +81,17 @@ bfifo_dequeue(struct Qdisc* sch)
 	return skb;
 }
 
-static int
+static unsigned int 
 fifo_drop(struct Qdisc* sch)
 {
 	struct sk_buff *skb;
 
 	skb = __skb_dequeue_tail(&sch->q);
 	if (skb) {
-		sch->stats.backlog -= skb->len;
+		unsigned int len = skb->len;
+		sch->stats.backlog -= len;
 		kfree_skb(skb);
-		return 1;
+		return len;
 	}
 	return 0;
 }

net/sched/sch_gred.c

@@ -259,8 +259,7 @@ gred_dequeue(struct Qdisc* sch)
 	return NULL;
 }
 
-static int
-gred_drop(struct Qdisc* sch)
+static unsigned int gred_drop(struct Qdisc* sch)
 {
 	struct sk_buff *skb;
 
@@ -269,11 +268,12 @@ gred_drop(struct Qdisc* sch)
 
 	skb = __skb_dequeue_tail(&sch->q);
 	if (skb) {
-		sch->stats.backlog -= skb->len;
+		unsigned int len = skb->len;
+		sch->stats.backlog -= len;
 		sch->stats.drops++;
 		q= t->tab[(skb->tc_index&0xf)];
 		if (q) {
-			q->backlog -= skb->len;
+			q->backlog -= len;
 			q->other++;
 			if (!q->backlog && !t->eqp)
 				PSCHED_GET_TIME(q->qidlestart);
@@ -282,7 +282,7 @@ gred_drop(struct Qdisc* sch)
 		}
 
 		kfree_skb(skb);
-		return 1;
+		return len;
 	}
 
 	q=t->tab[t->def];

net/sched/sch_htb.c

@@ -1059,7 +1059,7 @@ static struct sk_buff *htb_dequeue(struct Qdisc *sch)
 }
 
 /* try to drop from each class (by prio) until one succeed */
-static int htb_drop(struct Qdisc* sch)
+static unsigned int htb_drop(struct Qdisc* sch)
 {
 	struct htb_sched *q = (struct htb_sched *)sch->data;
 	int prio;
@@ -1067,14 +1067,15 @@ static int htb_drop(struct Qdisc* sch)
 	for (prio = TC_HTB_NUMPRIO - 1; prio >= 0; prio--) {
 		struct list_head *p;
 		list_for_each (p,q->drops+prio) {
-			struct htb_class *cl = list_entry(p,struct htb_class,
+			struct htb_class *cl = list_entry(p, struct htb_class,
 							  un.leaf.drop_list);
+			unsigned int len;
 			if (cl->un.leaf.q->ops->drop && 
-				cl->un.leaf.q->ops->drop(cl->un.leaf.q)) {
+				(len = cl->un.leaf.q->ops->drop(cl->un.leaf.q))) {
 				sch->q.qlen--;
 				if (!cl->un.leaf.q->q.qlen)
 					htb_deactivate (q,cl);
-				return 1;
+				return len;
 			}
 		}
 	}

net/sched/sch_ingress.c

@@ -190,7 +190,7 @@ static int ingress_requeue(struct sk_buff *skb,struct Qdisc *sch)
 	return 0;
 }
 
-static int ingress_drop(struct Qdisc *sch)
+static unsigned int ingress_drop(struct Qdisc *sch)
 {
 #ifdef DEBUG_INGRESS
 	struct ingress_qdisc_data *p = PRIV(sch);

net/sched/sch_prio.c

@@ -124,18 +124,18 @@ prio_dequeue(struct Qdisc* sch)
 
 }
 
-static int
-prio_drop(struct Qdisc* sch)
+static unsigned int prio_drop(struct Qdisc* sch)
 {
 	struct prio_sched_data *q = (struct prio_sched_data *)sch->data;
 	int prio;
+	unsigned int len;
 	struct Qdisc *qdisc;
 
 	for (prio = q->bands-1; prio >= 0; prio--) {
 		qdisc = q->queues[prio];
-		if (qdisc->ops->drop(qdisc)) {
+		if ((len = qdisc->ops->drop(qdisc)) != 0) {
 			sch->q.qlen--;
-			return 1;
+			return len;
 		}
 	}
 	return 0;

net/sched/sch_red.c

@@ -342,19 +342,19 @@ red_dequeue(struct Qdisc* sch)
 	return NULL;
 }
 
-static int
-red_drop(struct Qdisc* sch)
+static unsigned int red_drop(struct Qdisc* sch)
 {
 	struct sk_buff *skb;
 	struct red_sched_data *q = (struct red_sched_data *)sch->data;
 
 	skb = __skb_dequeue_tail(&sch->q);
 	if (skb) {
-		sch->stats.backlog -= skb->len;
+		unsigned int len = skb->len;
+		sch->stats.backlog -= len;
 		sch->stats.drops++;
 		q->st.other++;
 		kfree_skb(skb);
-		return 1;
+		return len;
 	}
 	PSCHED_GET_TIME(q->qidlestart);
 	return 0;

net/sched/sch_sfq.c

@@ -209,11 +209,12 @@ static inline void sfq_inc(struct sfq_sched_data *q, sfq_index x)
 	sfq_link(q, x);
 }
 
-static int sfq_drop(struct Qdisc *sch)
+static unsigned int sfq_drop(struct Qdisc *sch)
 {
 	struct sfq_sched_data *q = (struct sfq_sched_data *)sch->data;
 	sfq_index d = q->max_depth;
 	struct sk_buff *skb;
+	unsigned int len;
 
 	/* Queue is full! Find the longest slot and
 	   drop a packet from it */
@@ -221,12 +222,13 @@ static int sfq_drop(struct Qdisc *sch)
 	if (d > 1) {
 		sfq_index x = q->dep[d+SFQ_DEPTH].next;
 		skb = q->qs[x].prev;
+		len = skb->len;
 		__skb_unlink(skb, &q->qs[x]);
 		kfree_skb(skb);
 		sfq_dec(q, x);
 		sch->q.qlen--;
 		sch->stats.drops++;
-		return 1;
+		return len;
 	}
 
 	if (d == 1) {
@@ -235,13 +237,14 @@ static int sfq_drop(struct Qdisc *sch)
 		q->next[q->tail] = q->next[d];
 		q->allot[q->next[d]] += q->quantum;
 		skb = q->qs[d].prev;
+		len = skb->len;
 		__skb_unlink(skb, &q->qs[d]);
 		kfree_skb(skb);
 		sfq_dec(q, d);
 		sch->q.qlen--;
 		q->ht[q->hash[d]] = SFQ_DEPTH;
 		sch->stats.drops++;
-		return 1;
+		return len;
 	}
 
 	return 0;

net/sched/sch_tbf.c

@@ -7,6 +7,8 @@
  *		2 of the License, or (at your option) any later version.
  *
  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
+ *		Dmitry Torokhov <dtor@mail.ru> - allow attaching inner qdiscs -
+ *						 original idea by Martin Devera
  *
  */
 
@@ -123,62 +125,63 @@ struct tbf_sched_data
 	long	ptokens;		/* Current number of P tokens */
 	psched_time_t	t_c;		/* Time check-point */
 	struct timer_list wd_timer;	/* Watchdog timer */
+	struct Qdisc	*qdisc;		/* Inner qdisc, default - bfifo queue */
 };
 
 #define L2T(q,L)   ((q)->R_tab->data[(L)>>(q)->R_tab->rate.cell_log])
 #define L2T_P(q,L) ((q)->P_tab->data[(L)>>(q)->P_tab->rate.cell_log])
 
-static int
-tbf_enqueue(struct sk_buff *skb, struct Qdisc* sch)
+static int tbf_enqueue(struct sk_buff *skb, struct Qdisc* sch)
 {
 	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+	int ret;
 
-	if (skb->len > q->max_size)
-		goto drop;
-	__skb_queue_tail(&sch->q, skb);
-	if ((sch->stats.backlog += skb->len) <= q->limit) {
-		sch->stats.bytes += skb->len;
-		sch->stats.packets++;
-		return 0;
-	}
-
-	/* Drop action: undo the things that we just did,
-	 * i.e. make tail drop
-	 */
-
-	__skb_unlink(skb, &sch->q);
-	sch->stats.backlog -= skb->len;
-
-drop:
+	if (skb->len > q->max_size || sch->stats.backlog + skb->len > q->limit) {
 		sch->stats.drops++;
 #ifdef CONFIG_NET_CLS_POLICE
-	if (sch->reshape_fail==NULL || sch->reshape_fail(skb, sch))
+		if (sch->reshape_fail == NULL || sch->reshape_fail(skb, sch))
 #endif
 			kfree_skb(skb);
+	
 		return NET_XMIT_DROP;
+	}
+	
+	if ((ret = q->qdisc->enqueue(skb, q->qdisc)) != 0) {
+		sch->stats.drops++;
+		return ret;
+	}	
+	
+	sch->q.qlen++;
+	sch->stats.backlog += skb->len;
+	sch->stats.bytes += skb->len;
+	sch->stats.packets++;
+	return 0;
 }
 
-static int
-tbf_requeue(struct sk_buff *skb, struct Qdisc* sch)
+static int tbf_requeue(struct sk_buff *skb, struct Qdisc* sch)
 {
-	__skb_queue_head(&sch->q, skb);
+	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+	int ret;
+	
+	if ((ret = q->qdisc->ops->requeue(skb, q->qdisc)) == 0) {
+		sch->q.qlen++; 
 		sch->stats.backlog += skb->len;
-	return 0;
+	}
+	
+	return ret;
 }
 
-static int
-tbf_drop(struct Qdisc* sch)
+static unsigned int tbf_drop(struct Qdisc* sch)
 {
-	struct sk_buff *skb;
+	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+	unsigned int len;
 	
-	skb = __skb_dequeue_tail(&sch->q);
-	if (skb) {
-		sch->stats.backlog -= skb->len;
+	if ((len = q->qdisc->ops->drop(q->qdisc)) != 0) {
+		sch->q.qlen--;
+		sch->stats.backlog -= len;
 		sch->stats.drops++;
-		kfree_skb(skb);
-		return 1;
 	}
-	return 0;
+	return len;
 }
 
 static void tbf_watchdog(unsigned long arg)
@@ -189,18 +192,18 @@ static void tbf_watchdog(unsigned long arg)
 	netif_schedule(sch->dev);
 }
 
-static struct sk_buff *
-tbf_dequeue(struct Qdisc* sch)
+static struct sk_buff *tbf_dequeue(struct Qdisc* sch)
 {
 	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
 	struct sk_buff *skb;
 	
-	skb = __skb_dequeue(&sch->q);
+	skb = q->qdisc->dequeue(q->qdisc);
 
 	if (skb) {
 		psched_time_t now;
 		long toks;
 		long ptoks = 0;
+		unsigned int len = skb->len;
 		
 		PSCHED_GET_TIME(now);
 
@@ -210,18 +213,19 @@ tbf_dequeue(struct Qdisc* sch)
 			ptoks = toks + q->ptokens;
 			if (ptoks > (long)q->mtu)
 				ptoks = q->mtu;
-			ptoks -= L2T_P(q, skb->len);
+			ptoks -= L2T_P(q, len);
 		}
 		toks += q->tokens;
 		if (toks > (long)q->buffer)
 			toks = q->buffer;
-		toks -= L2T(q, skb->len);
+		toks -= L2T(q, len);
 
 		if ((toks|ptoks) >= 0) {
 			q->t_c = now;
 			q->tokens = toks;
 			q->ptokens = ptoks;
-			sch->stats.backlog -= skb->len;
+			sch->stats.backlog -= len;
+			sch->q.qlen--;
 			sch->flags &= ~TCQ_F_THROTTLED;
 			return skb;
 		}
@@ -245,7 +249,13 @@ tbf_dequeue(struct Qdisc* sch)
 		   This is the main idea of all FQ algorithms
 		   (cf. CSZ, HPFQ, HFSC)
 		 */
-		__skb_queue_head(&sch->q, skb);
+		
+		if (q->qdisc->ops->requeue(skb, q->qdisc) != NET_XMIT_SUCCESS) {
+			/* When requeue fails skb is dropped */ 
+			sch->q.qlen--;
+			sch->stats.backlog -= len;
+			sch->stats.drops++;
+		}	
 		
 		sch->flags |= TCQ_F_THROTTLED;
 		sch->stats.overlimits++;
@@ -253,12 +263,11 @@ tbf_dequeue(struct Qdisc* sch)
 	return NULL;
 }
 
-
-static void
-tbf_reset(struct Qdisc* sch)
+static void tbf_reset(struct Qdisc* sch)
 {
 	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
 
+	qdisc_reset(q->qdisc);
 	skb_queue_purge(&sch->q);
 	sch->stats.backlog = 0;
 	PSCHED_GET_TIME(q->t_c);
@@ -268,6 +277,31 @@ tbf_reset(struct Qdisc* sch)
 	del_timer(&q->wd_timer);
 }
 
+static struct Qdisc *tbf_create_dflt_qdisc(struct net_device *dev, u32 limit)
+{
+	struct Qdisc *q = qdisc_create_dflt(dev, &bfifo_qdisc_ops);
+        struct rtattr *rta;
+	int ret;
+	
+	if (q) {
+		rta = kmalloc(RTA_LENGTH(sizeof(struct tc_fifo_qopt)), GFP_KERNEL);
+		if (rta) {
+			rta->rta_type = RTM_NEWQDISC;
+			rta->rta_len = RTA_LENGTH(sizeof(struct tc_fifo_qopt)); 
+			((struct tc_fifo_qopt *)RTA_DATA(rta))->limit = limit;
+			
+			ret = q->ops->change(q, rta);
+			kfree(rta);
+			
+			if (ret == 0)
+				return q;
+		}
+		qdisc_destroy(q);
+	}
+
+	return NULL;	
+}
+
 static int tbf_change(struct Qdisc* sch, struct rtattr *opt)
 {
 	int err = -EINVAL;
@@ -276,6 +310,7 @@ static int tbf_change(struct Qdisc* sch, struct rtattr *opt)
 	struct tc_tbf_qopt *qopt;
 	struct qdisc_rate_table *rtab = NULL;
 	struct qdisc_rate_table *ptab = NULL;
+	struct Qdisc *child = NULL;
 	int max_size,n;
 
 	if (rtattr_parse(tb, TCA_TBF_PTAB, RTA_DATA(opt), RTA_PAYLOAD(opt)) ||
@@ -309,7 +344,13 @@ static int tbf_change(struct Qdisc* sch, struct rtattr *opt)
 	if (max_size < 0)
 		goto done;
 	
+	if (q->qdisc == &noop_qdisc) {
+		if ((child = tbf_create_dflt_qdisc(sch->dev, qopt->limit)) == NULL)
+			goto done;
+	}
+
 	sch_tree_lock(sch);
+	if (child) q->qdisc = child;
 	q->limit = qopt->limit;
 	q->mtu = qopt->mtu;
 	q->max_size = max_size;
@@ -340,6 +381,8 @@ static int tbf_init(struct Qdisc* sch, struct rtattr *opt)
 	q->wd_timer.function = tbf_watchdog;
 	q->wd_timer.data = (unsigned long)sch;
 
+	q->qdisc = &noop_qdisc;
+	
 	return tbf_change(sch, opt);
 }
 
@@ -353,6 +396,9 @@ static void tbf_destroy(struct Qdisc *sch)
 		qdisc_put_rtab(q->P_tab);
 	if (q->R_tab)
 		qdisc_put_rtab(q->R_tab);
+	
+	qdisc_destroy(q->qdisc);
+	q->qdisc = &noop_qdisc;
 }
 
 static int tbf_dump(struct Qdisc *sch, struct sk_buff *skb)
@@ -383,9 +429,92 @@ static int tbf_dump(struct Qdisc *sch, struct sk_buff *skb)
 	return -1;
 }
 
+static int tbf_dump_class(struct Qdisc *sch, unsigned long cl,
+	       		  struct sk_buff *skb, struct tcmsg *tcm)
+{
+	struct tbf_sched_data *q = (struct tbf_sched_data*)sch->data;
+
+	if (cl != 1) 	/* only one class */ 
+		return -ENOENT;
+    
+	tcm->tcm_parent = TC_H_ROOT;
+	tcm->tcm_handle = 1;
+	tcm->tcm_info = q->qdisc->handle;
+
+	return 0;
+}
+
+static int tbf_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
+		     struct Qdisc **old)
+{
+	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+
+	if (new == NULL)
+		new = &noop_qdisc;
+
+	sch_tree_lock(sch);	
+	*old = xchg(&q->qdisc, new);
+	qdisc_reset(*old);
+	sch_tree_unlock(sch);
+	
+	return 0;
+}
+
+static struct Qdisc *tbf_leaf(struct Qdisc *sch, unsigned long arg)
+{
+	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+	return q->qdisc;
+}
+
+static unsigned long tbf_get(struct Qdisc *sch, u32 classid)
+{
+	return 1;
+}
+
+static void tbf_put(struct Qdisc *sch, unsigned long arg)
+{
+}
+
+static int tbf_change_class(struct Qdisc *sch, u32 classid, u32 parentid, 
+			struct rtattr **tca, unsigned long *arg)
+{
+	return -ENOSYS;
+}
+
+static int tbf_delete(struct Qdisc *sch, unsigned long arg)
+{
+	return -ENOSYS;
+}
+
+static void tbf_walk(struct Qdisc *sch, struct qdisc_walker *walker)
+{
+	struct tbf_sched_data *q = (struct tbf_sched_data *)sch->data;
+
+	if (!walker->stop) {
+		if (walker->count >= walker->skip) 
+			if (walker->fn(sch, (unsigned long)q, walker) < 0) { 
+				walker->stop = 1;
+				return;
+			}
+		walker->count++;
+	}
+}
+
+static struct Qdisc_class_ops tbf_class_ops =
+{
+	.graft		= 	tbf_graft,
+	.leaf		=	tbf_leaf,
+	.get		=	tbf_get,
+	.put		=	tbf_put,
+	.change		=	tbf_change_class,
+	.delete		=	tbf_delete,
+	.walk		=	tbf_walk,
+	.dump		=	tbf_dump_class,
+};
+
 struct Qdisc_ops tbf_qdisc_ops = {
 	.next		=	NULL,
-	.cl_ops		=	NULL,
+	.cl_ops		=	&tbf_class_ops,
 	.id		=	"tbf",
 	.priv_size	=	sizeof(struct tbf_sched_data),
 	.enqueue	=	tbf_enqueue,
@@ -397,6 +526,7 @@ struct Qdisc_ops tbf_qdisc_ops = {
 	.destroy	=	tbf_destroy,
 	.change		=	tbf_change,
 	.dump		=	tbf_dump,
+	.owner		=	THIS_MODULE,
 };