net_sched: reintroduce dev->qdisc for use by sch_api

Currently the multiqueue integration with the qdisc API suffers from a few problems: - with multiple queues, all root qdiscs use the same handle. This means they can't be exposed to userspace in a backwards compatible fashion. - all API operations always refer to queue number 0. Newly created qdiscs are automatically shared between all queues, its not possible to address individual queues or restore multiqueue behaviour once a shared qdisc has been attached. - Dumps only contain the root qdisc of queue 0, in case of non-shared qdiscs this means the statistics are incomplete. This patch reintroduces dev->qdisc, which points to the (single) root qdisc from userspace's point of view. Currently it either points to the first (non-shared) default qdisc, or a qdisc shared between all queues. The following patches will introduce a classful dummy qdisc, which will be used as root qdisc and contain the per-queue qdiscs as children. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: reintroduce dev->qdisc for use by sch_api
Currently the multiqueue integration with the qdisc API suffers from a few problems: - with multiple queues, all root qdiscs use the same handle. This means they can't be exposed to userspace in a backwards compatible fashion. - all API operations always refer to queue number 0. Newly created qdiscs are automatically shared between all queues, its not possible to address individual queues or restore multiqueue behaviour once a shared qdisc has been attached. - Dumps only contain the root qdisc of queue 0, in case of non-shared qdiscs this means the statistics are incomplete. This patch reintroduces dev->qdisc, which points to the (single) root qdisc from userspace's point of view. Currently it either points to the first (non-shared) default qdisc, or a qdisc shared between all queues. The following patches will introduce a classful dummy qdisc, which will be used as root qdisc and contain the per-queue qdiscs as children. Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy · David S. Miller
1 parent 5b9a9ccfad
Showing 5 changed files with 34 additions and 48 deletions Side-by-side Diff
include/linux/netdevice.h
net/core/rtnetlink.c
net/sched/cls_api.c
net/sched/sch_api.c
net/sched/sch_generic.c
@@ -832,6 +832,9 @@
 	/* Number of TX queues currently active in device  */
 	unsigned int		real_num_tx_queues;
  
+	/* root qdisc from userspace point of view */
+	struct Qdisc		*qdisc;
+
 	unsigned long		tx_queue_len;	/* Max frames per queue allowed */
 	spinlock_t		tx_global_lock;
 /*
@@ -606,7 +606,6 @@
 			    int type, u32 pid, u32 seq, u32 change,
 			    unsigned int flags)
 {
-	struct netdev_queue *txq;
 	struct ifinfomsg *ifm;
 	struct nlmsghdr *nlh;
 	const struct net_device_stats *stats;
@@ -637,9 +636,8 @@
 	if (dev->master)
 		NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
  
-	txq = netdev_get_tx_queue(dev, 0);
-	if (txq->qdisc_sleeping)
-		NLA_PUT_STRING(skb, IFLA_QDISC, txq->qdisc_sleeping->ops->id);
+	if (dev->qdisc)
+		NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
  
 	if (dev->ifalias)
 		NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
@@ -168,8 +168,7 @@
  
 	/* Find qdisc */
 	if (!parent) {
-		struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, 0);
-		q = dev_queue->qdisc_sleeping;
+		q = dev->qdisc;
 		parent = q->handle;
 	} else {
 		q = qdisc_lookup(dev, TC_H_MAJ(t->tcm_parent));
@@ -408,7 +407,6 @@
 static int tc_dump_tfilter(struct sk_buff *skb, struct netlink_callback *cb)
 {
 	struct net *net = sock_net(skb->sk);
-	struct netdev_queue *dev_queue;
 	int t;
 	int s_t;
 	struct net_device *dev;
  
@@ -427,9 +425,8 @@
 	if ((dev = dev_get_by_index(&init_net, tcm->tcm_ifindex)) == NULL)
 		return skb->len;
  
-	dev_queue = netdev_get_tx_queue(dev, 0);
 	if (!tcm->tcm_parent)
-		q = dev_queue->qdisc_sleeping;
+		q = dev->qdisc;
 	else
 		q = qdisc_lookup(dev, TC_H_MAJ(tcm->tcm_parent));
 	if (!q)
@@ -207,7 +207,7 @@
 static void qdisc_list_add(struct Qdisc *q)
 {
 	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS))
-		list_add_tail(&q->list, &qdisc_root_sleeping(q)->list);
+		list_add_tail(&q->list, &qdisc_dev(q)->qdisc->list);
 }
  
 void qdisc_list_del(struct Qdisc *q)
  
  
@@ -219,18 +219,12 @@
  
 struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
 {
-	unsigned int i;
 	struct Qdisc *q;
  
-	for (i = 0; i < dev->num_tx_queues; i++) {
-		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
-		struct Qdisc *txq_root = txq->qdisc_sleeping;
+	q = qdisc_match_from_root(dev->qdisc, handle);
+	if (q)
+		goto out;
  
-		q = qdisc_match_from_root(txq_root, handle);
-		if (q)
-			goto out;
-	}
-
 	q = qdisc_match_from_root(dev->rx_queue.qdisc_sleeping, handle);
 out:
 	return q;
  
@@ -720,9 +714,14 @@
 			if (new && i > 0)
 				atomic_inc(&new->refcnt);
  
-			notify_and_destroy(skb, n, classid, old, new);
+			qdisc_destroy(old);
 		}
  
+		notify_and_destroy(skb, n, classid, dev->qdisc, new);
+		if (new)
+			atomic_inc(&new->refcnt);
+		dev->qdisc = new ? : &noop_qdisc;
+
 		if (dev->flags & IFF_UP)
 			dev_activate(dev);
 	} else {
@@ -974,9 +973,7 @@
 				q = dev->rx_queue.qdisc_sleeping;
 			}
 		} else {
-			struct netdev_queue *dev_queue;
-			dev_queue = netdev_get_tx_queue(dev, 0);
-			q = dev_queue->qdisc_sleeping;
+			q = dev->qdisc;
 		}
 		if (!q)
 			return -ENOENT;
@@ -1044,9 +1041,7 @@
 				q = dev->rx_queue.qdisc_sleeping;
 			}
 		} else {
-			struct netdev_queue *dev_queue;
-			dev_queue = netdev_get_tx_queue(dev, 0);
-			q = dev_queue->qdisc_sleeping;
+			q = dev->qdisc;
 		}
  
 		/* It may be default qdisc, ignore it */
@@ -1291,8 +1286,7 @@
 			s_q_idx = 0;
 		q_idx = 0;
  
-		dev_queue = netdev_get_tx_queue(dev, 0);
-		if (tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb, &q_idx, s_q_idx) < 0)
+		if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx) < 0)
 			goto done;
  
 		dev_queue = &dev->rx_queue;
@@ -1323,7 +1317,6 @@
 static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
 {
 	struct net *net = sock_net(skb->sk);
-	struct netdev_queue *dev_queue;
 	struct tcmsg *tcm = NLMSG_DATA(n);
 	struct nlattr *tca[TCA_MAX + 1];
 	struct net_device *dev;
@@ -1361,7 +1354,6 @@
  
 	/* Step 1. Determine qdisc handle X:0 */
  
-	dev_queue = netdev_get_tx_queue(dev, 0);
 	if (pid != TC_H_ROOT) {
 		u32 qid1 = TC_H_MAJ(pid);
  
@@ -1372,7 +1364,7 @@
 		} else if (qid1) {
 			qid = qid1;
 		} else if (qid == 0)
-			qid = dev_queue->qdisc_sleeping->handle;
+			qid = dev->qdisc->handle;
  
 		/* Now qid is genuine qdisc handle consistent
 		   both with parent and child.
@@ -1383,7 +1375,7 @@
 			pid = TC_H_MAKE(qid, pid);
 	} else {
 		if (qid == 0)
-			qid = dev_queue->qdisc_sleeping->handle;
+			qid = dev->qdisc->handle;
 	}
  
 	/* OK. Locate qdisc */
@@ -1588,8 +1580,7 @@
 	s_t = cb->args[0];
 	t = 0;
  
-	dev_queue = netdev_get_tx_queue(dev, 0);
-	if (tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb, &t, s_t) < 0)
+	if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0)
 		goto done;
  
 	dev_queue = &dev->rx_queue;
@@ -623,19 +623,6 @@
 }
 EXPORT_SYMBOL(qdisc_destroy);
  
-static bool dev_all_qdisc_sleeping_noop(struct net_device *dev)
-{
-	unsigned int i;
-
-	for (i = 0; i < dev->num_tx_queues; i++) {
-		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
-
-		if (txq->qdisc_sleeping != &noop_qdisc)
-			return false;
-	}
-	return true;
-}
-
 static void attach_one_default_qdisc(struct net_device *dev,
 				     struct netdev_queue *dev_queue,
 				     void *_unused)
@@ -677,6 +664,7 @@
  
 void dev_activate(struct net_device *dev)
 {
+	struct netdev_queue *txq;
 	int need_watchdog;
  
 	/* No queueing discipline is attached to device;
  
@@ -685,9 +673,14 @@
 	   virtual interfaces
 	 */
  
-	if (dev_all_qdisc_sleeping_noop(dev))
+	if (dev->qdisc == &noop_qdisc) {
 		netdev_for_each_tx_queue(dev, attach_one_default_qdisc, NULL);
  
+		txq = netdev_get_tx_queue(dev, 0);
+		dev->qdisc = txq->qdisc_sleeping;
+		atomic_inc(&dev->qdisc->refcnt);
+	}
+
 	if (!netif_carrier_ok(dev))
 		/* Delay activation until next carrier-on event */
 		return;
@@ -777,6 +770,7 @@
  
 void dev_init_scheduler(struct net_device *dev)
 {
+	dev->qdisc = &noop_qdisc;
 	netdev_for_each_tx_queue(dev, dev_init_scheduler_queue, &noop_qdisc);
 	dev_init_scheduler_queue(dev, &dev->rx_queue, &noop_qdisc);
  
@@ -802,6 +796,9 @@
 {
 	netdev_for_each_tx_queue(dev, shutdown_scheduler_queue, &noop_qdisc);
 	shutdown_scheduler_queue(dev, &dev->rx_queue, &noop_qdisc);
+	qdisc_destroy(dev->qdisc);
+	dev->qdisc = &noop_qdisc;
+
 	WARN_ON(timer_pending(&dev->watchdog_timer));
 }
...	...	@@ -832,6 +832,9 @@
832	832	/* Number of TX queues currently active in device */
833	833	unsigned int real_num_tx_queues;
834	834
	835	+ /* root qdisc from userspace point of view */
	836	+ struct Qdisc *qdisc;
	837	+
835	838	unsigned long tx_queue_len; /* Max frames per queue allowed */
836	839	spinlock_t tx_global_lock;
837	840	/*
...	...	@@ -606,7 +606,6 @@
606	606	int type, u32 pid, u32 seq, u32 change,
607	607	unsigned int flags)
608	608	{
609		- struct netdev_queue *txq;
610	609	struct ifinfomsg *ifm;
611	610	struct nlmsghdr *nlh;
612	611	const struct net_device_stats *stats;
...	...	@@ -637,9 +636,8 @@
637	636	if (dev->master)
638	637	NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
639	638
640		- txq = netdev_get_tx_queue(dev, 0);
641		- if (txq->qdisc_sleeping)
642		- NLA_PUT_STRING(skb, IFLA_QDISC, txq->qdisc_sleeping->ops->id);
	639	+ if (dev->qdisc)
	640	+ NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
643	641
644	642	if (dev->ifalias)
645	643	NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
...	...	@@ -168,8 +168,7 @@
168	168
169	169	/* Find qdisc */
170	170	if (!parent) {
171		- struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, 0);
172		- q = dev_queue->qdisc_sleeping;
	171	+ q = dev->qdisc;
173	172	parent = q->handle;
174	173	} else {
175	174	q = qdisc_lookup(dev, TC_H_MAJ(t->tcm_parent));
...	...	@@ -408,7 +407,6 @@
408	407	static int tc_dump_tfilter(struct sk_buff skb, struct netlink_callback cb)
409	408	{
410	409	struct net *net = sock_net(skb->sk);
411		- struct netdev_queue *dev_queue;
412	410	int t;
413	411	int s_t;
414	412	struct net_device *dev;
415	413
...	...	@@ -427,9 +425,8 @@
427	425	if ((dev = dev_get_by_index(&init_net, tcm->tcm_ifindex)) == NULL)
428	426	return skb->len;
429	427
430		- dev_queue = netdev_get_tx_queue(dev, 0);
431	428	if (!tcm->tcm_parent)
432		- q = dev_queue->qdisc_sleeping;
	429	+ q = dev->qdisc;
433	430	else
434	431	q = qdisc_lookup(dev, TC_H_MAJ(tcm->tcm_parent));
435	432	if (!q)
...	...	@@ -207,7 +207,7 @@
207	207	static void qdisc_list_add(struct Qdisc *q)
208	208	{
209	209	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS))
210		- list_add_tail(&q->list, &qdisc_root_sleeping(q)->list);
	210	+ list_add_tail(&q->list, &qdisc_dev(q)->qdisc->list);
211	211	}
212	212
213	213	void qdisc_list_del(struct Qdisc *q)
214	214
215	215
...	...	@@ -219,18 +219,12 @@
219	219
220	220	struct Qdisc qdisc_lookup(struct net_device dev, u32 handle)
221	221	{
222		- unsigned int i;
223	222	struct Qdisc *q;
224	223
225		- for (i = 0; i < dev->num_tx_queues; i++) {
226		- struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
227		- struct Qdisc *txq_root = txq->qdisc_sleeping;
	224	+ q = qdisc_match_from_root(dev->qdisc, handle);
	225	+ if (q)
	226	+ goto out;
228	227
229		- q = qdisc_match_from_root(txq_root, handle);
230		- if (q)
231		- goto out;
232		- }
233		-
234	228	q = qdisc_match_from_root(dev->rx_queue.qdisc_sleeping, handle);
235	229	out:
236	230	return q;
237	231
...	...	@@ -720,9 +714,14 @@
720	714	if (new && i > 0)
721	715	atomic_inc(&new->refcnt);
722	716
723		- notify_and_destroy(skb, n, classid, old, new);
	717	+ qdisc_destroy(old);
724	718	}
725	719
	720	+ notify_and_destroy(skb, n, classid, dev->qdisc, new);
	721	+ if (new)
	722	+ atomic_inc(&new->refcnt);
	723	+ dev->qdisc = new ? : &noop_qdisc;
	724	+
726	725	if (dev->flags & IFF_UP)
727	726	dev_activate(dev);
728	727	} else {
...	...	@@ -974,9 +973,7 @@
974	973	q = dev->rx_queue.qdisc_sleeping;
975	974	}
976	975	} else {
977		- struct netdev_queue *dev_queue;
978		- dev_queue = netdev_get_tx_queue(dev, 0);
979		- q = dev_queue->qdisc_sleeping;
	976	+ q = dev->qdisc;
980	977	}
981	978	if (!q)
982	979	return -ENOENT;
...	...	@@ -1044,9 +1041,7 @@
1044	1041	q = dev->rx_queue.qdisc_sleeping;
1045	1042	}
1046	1043	} else {
1047		- struct netdev_queue *dev_queue;
1048		- dev_queue = netdev_get_tx_queue(dev, 0);
1049		- q = dev_queue->qdisc_sleeping;
	1044	+ q = dev->qdisc;
1050	1045	}
1051	1046
1052	1047	/* It may be default qdisc, ignore it */
...	...	@@ -1291,8 +1286,7 @@
1291	1286	s_q_idx = 0;
1292	1287	q_idx = 0;
1293	1288
1294		- dev_queue = netdev_get_tx_queue(dev, 0);
1295		- if (tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb, &q_idx, s_q_idx) < 0)
	1289	+ if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx) < 0)
1296	1290	goto done;
1297	1291
1298	1292	dev_queue = &dev->rx_queue;
...	...	@@ -1323,7 +1317,6 @@
1323	1317	static int tc_ctl_tclass(struct sk_buff skb, struct nlmsghdr n, void *arg)
1324	1318	{
1325	1319	struct net *net = sock_net(skb->sk);
1326		- struct netdev_queue *dev_queue;
1327	1320	struct tcmsg *tcm = NLMSG_DATA(n);
1328	1321	struct nlattr *tca[TCA_MAX + 1];
1329	1322	struct net_device *dev;
...	...	@@ -1361,7 +1354,6 @@
1361	1354
1362	1355	/* Step 1. Determine qdisc handle X:0 */
1363	1356
1364		- dev_queue = netdev_get_tx_queue(dev, 0);
1365	1357	if (pid != TC_H_ROOT) {
1366	1358	u32 qid1 = TC_H_MAJ(pid);
1367	1359
...	...	@@ -1372,7 +1364,7 @@
1372	1364	} else if (qid1) {
1373	1365	qid = qid1;
1374	1366	} else if (qid == 0)
1375		- qid = dev_queue->qdisc_sleeping->handle;
	1367	+ qid = dev->qdisc->handle;
1376	1368
1377	1369	/* Now qid is genuine qdisc handle consistent
1378	1370	both with parent and child.
...	...	@@ -1383,7 +1375,7 @@
1383	1375	pid = TC_H_MAKE(qid, pid);
1384	1376	} else {
1385	1377	if (qid == 0)
1386		- qid = dev_queue->qdisc_sleeping->handle;
	1378	+ qid = dev->qdisc->handle;
1387	1379	}
1388	1380
1389	1381	/* OK. Locate qdisc */
...	...	@@ -1588,8 +1580,7 @@
1588	1580	s_t = cb->args[0];
1589	1581	t = 0;
1590	1582
1591		- dev_queue = netdev_get_tx_queue(dev, 0);
1592		- if (tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb, &t, s_t) < 0)
	1583	+ if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0)
1593	1584	goto done;
1594	1585
1595	1586	dev_queue = &dev->rx_queue;
...	...	@@ -623,19 +623,6 @@
623	623	}
624	624	EXPORT_SYMBOL(qdisc_destroy);
625	625
626		-static bool dev_all_qdisc_sleeping_noop(struct net_device *dev)
627		-{
628		- unsigned int i;
629		-
630		- for (i = 0; i < dev->num_tx_queues; i++) {
631		- struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
632		-
633		- if (txq->qdisc_sleeping != &noop_qdisc)
634		- return false;
635		- }
636		- return true;
637		-}
638		-
639	626	static void attach_one_default_qdisc(struct net_device *dev,
640	627	struct netdev_queue *dev_queue,
641	628	void *_unused)
...	...	@@ -677,6 +664,7 @@
677	664
678	665	void dev_activate(struct net_device *dev)
679	666	{
	667	+ struct netdev_queue *txq;
680	668	int need_watchdog;
681	669
682	670	/* No queueing discipline is attached to device;
683	671
...	...	@@ -685,9 +673,14 @@
685	673	virtual interfaces
686	674	*/
687	675
688		- if (dev_all_qdisc_sleeping_noop(dev))
	676	+ if (dev->qdisc == &noop_qdisc) {
689	677	netdev_for_each_tx_queue(dev, attach_one_default_qdisc, NULL);
690	678
	679	+ txq = netdev_get_tx_queue(dev, 0);
	680	+ dev->qdisc = txq->qdisc_sleeping;
	681	+ atomic_inc(&dev->qdisc->refcnt);
	682	+ }
	683	+
691	684	if (!netif_carrier_ok(dev))
692	685	/* Delay activation until next carrier-on event */
693	686	return;
...	...	@@ -777,6 +770,7 @@
777	770
778	771	void dev_init_scheduler(struct net_device *dev)
779	772	{
	773	+ dev->qdisc = &noop_qdisc;
780	774	netdev_for_each_tx_queue(dev, dev_init_scheduler_queue, &noop_qdisc);
781	775	dev_init_scheduler_queue(dev, &dev->rx_queue, &noop_qdisc);
782	776
...	...	@@ -802,6 +796,9 @@
802	796	{
803	797	netdev_for_each_tx_queue(dev, shutdown_scheduler_queue, &noop_qdisc);
804	798	shutdown_scheduler_queue(dev, &dev->rx_queue, &noop_qdisc);
	799	+ qdisc_destroy(dev->qdisc);
	800	+ dev->qdisc = &noop_qdisc;
	801	+
805	802	WARN_ON(timer_pending(&dev->watchdog_timer));
806	803	}