sch_generic.c 21.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
/*
 * net/sched/sch_generic.c	Generic packet scheduler routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 *              Jamal Hadi Salim, <hadi@cyberus.ca> 990601
 *              - Ingress support
 */

#include <linux/bitops.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>
#include <linux/init.h>
#include <linux/rcupdate.h>
#include <linux/list.h>
27
#include <linux/slab.h>
L
Linus Torvalds 已提交
28
#include <net/pkt_sched.h>
E
Eric Dumazet 已提交
29
#include <net/dst.h>
L
Linus Torvalds 已提交
30 31 32

/* Main transmission queue. */

33
/* Modifications to data participating in scheduling must be protected with
34
 * qdisc_lock(qdisc) spinlock.
35 36
 *
 * The idea is the following:
37 38
 * - enqueue, dequeue are serialized via qdisc root lock
 * - ingress filtering is also serialized via qdisc root lock
39
 * - updates to tree and tree walking are only done under the rtnl mutex.
L
Linus Torvalds 已提交
40 41
 */

42
static inline int dev_requeue_skb(struct sk_buff *skb, struct Qdisc *q)
43
{
E
Eric Dumazet 已提交
44
	skb_dst_force(skb);
45
	q->gso_skb = skb;
46
	q->qstats.requeues++;
47
	q->q.qlen++;	/* it's still part of the queue */
48
	__netif_schedule(q);
49

50 51 52
	return 0;
}

53
static inline struct sk_buff *dequeue_skb(struct Qdisc *q)
54
{
55 56
	struct sk_buff *skb = q->gso_skb;

57 58 59 60 61 62
	if (unlikely(skb)) {
		struct net_device *dev = qdisc_dev(q);
		struct netdev_queue *txq;

		/* check the reason of requeuing without tx lock first */
		txq = netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
63
		if (!netif_tx_queue_frozen_or_stopped(txq)) {
64
			q->gso_skb = NULL;
65 66
			q->q.qlen--;
		} else
67 68
			skb = NULL;
	} else {
69
		skb = q->dequeue(q);
70
	}
71 72 73 74

	return skb;
}

75
static inline int handle_dev_cpu_collision(struct sk_buff *skb,
76
					   struct netdev_queue *dev_queue,
77
					   struct Qdisc *q)
78
{
79
	int ret;
80

81
	if (unlikely(dev_queue->xmit_lock_owner == smp_processor_id())) {
82 83 84 85 86 87
		/*
		 * Same CPU holding the lock. It may be a transient
		 * configuration error, when hard_start_xmit() recurses. We
		 * detect it by checking xmit owner and drop the packet when
		 * deadloop is detected. Return OK to try the next skb.
		 */
88
		kfree_skb(skb);
89
		if (net_ratelimit())
E
Eric Dumazet 已提交
90 91
			pr_warning("Dead loop on netdevice %s, fix it urgently!\n",
				   dev_queue->dev->name);
92 93 94 95 96 97
		ret = qdisc_qlen(q);
	} else {
		/*
		 * Another cpu is holding lock, requeue & delay xmits for
		 * some time.
		 */
E
Eric Dumazet 已提交
98
		__this_cpu_inc(softnet_data.cpu_collision);
99
		ret = dev_requeue_skb(skb, q);
100 101
	}

102
	return ret;
103 104
}

105
/*
106 107 108
 * Transmit one skb, and handle the return status as required. Holding the
 * __QDISC_STATE_RUNNING bit guarantees that only one CPU can execute this
 * function.
109 110 111 112 113
 *
 * Returns to the caller:
 *				0  - queue is empty or throttled.
 *				>0 - queue is not empty.
 */
114 115 116
int sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
		    struct net_device *dev, struct netdev_queue *txq,
		    spinlock_t *root_lock)
L
Linus Torvalds 已提交
117
{
118
	int ret = NETDEV_TX_BUSY;
119 120 121

	/* And release qdisc */
	spin_unlock(root_lock);
122

123
	HARD_TX_LOCK(dev, txq, smp_processor_id());
124
	if (!netif_tx_queue_frozen_or_stopped(txq))
125
		ret = dev_hard_start_xmit(skb, dev, txq);
126

127
	HARD_TX_UNLOCK(dev, txq);
128

129
	spin_lock(root_lock);
130

131 132
	if (dev_xmit_complete(ret)) {
		/* Driver sent out skb successfully or skb was consumed */
133
		ret = qdisc_qlen(q);
134
	} else if (ret == NETDEV_TX_LOCKED) {
135
		/* Driver try lock failed */
136
		ret = handle_dev_cpu_collision(skb, txq, q);
137
	} else {
138 139
		/* Driver returned NETDEV_TX_BUSY - requeue skb */
		if (unlikely (ret != NETDEV_TX_BUSY && net_ratelimit()))
E
Eric Dumazet 已提交
140 141
			pr_warning("BUG %s code %d qlen %d\n",
				   dev->name, ret, q->q.qlen);
142

143
		ret = dev_requeue_skb(skb, q);
144
	}
145

146
	if (ret && netif_tx_queue_frozen_or_stopped(txq))
147 148
		ret = 0;

149
	return ret;
L
Linus Torvalds 已提交
150 151
}

152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181
/*
 * NOTE: Called under qdisc_lock(q) with locally disabled BH.
 *
 * __QDISC_STATE_RUNNING guarantees only one CPU can process
 * this qdisc at a time. qdisc_lock(q) serializes queue accesses for
 * this queue.
 *
 *  netif_tx_lock serializes accesses to device driver.
 *
 *  qdisc_lock(q) and netif_tx_lock are mutually exclusive,
 *  if one is grabbed, another must be free.
 *
 * Note, that this procedure can be called by a watchdog timer
 *
 * Returns to the caller:
 *				0  - queue is empty or throttled.
 *				>0 - queue is not empty.
 *
 */
static inline int qdisc_restart(struct Qdisc *q)
{
	struct netdev_queue *txq;
	struct net_device *dev;
	spinlock_t *root_lock;
	struct sk_buff *skb;

	/* Dequeue packet */
	skb = dequeue_skb(q);
	if (unlikely(!skb))
		return 0;
E
Eric Dumazet 已提交
182
	WARN_ON_ONCE(skb_dst_is_noref(skb));
183 184 185 186 187 188 189
	root_lock = qdisc_lock(q);
	dev = qdisc_dev(q);
	txq = netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));

	return sch_direct_xmit(skb, q, dev, txq, root_lock);
}

190
void __qdisc_run(struct Qdisc *q)
H
Herbert Xu 已提交
191
{
192 193
	unsigned long start_time = jiffies;

194
	while (qdisc_restart(q)) {
195 196 197 198 199 200
		/*
		 * Postpone processing if
		 * 1. another process needs the CPU;
		 * 2. we've been doing it for too long.
		 */
		if (need_resched() || jiffies != start_time) {
201
			__netif_schedule(q);
202
			break;
203 204
		}
	}
H
Herbert Xu 已提交
205

206
	qdisc_run_end(q);
H
Herbert Xu 已提交
207 208
}

209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
unsigned long dev_trans_start(struct net_device *dev)
{
	unsigned long val, res = dev->trans_start;
	unsigned int i;

	for (i = 0; i < dev->num_tx_queues; i++) {
		val = netdev_get_tx_queue(dev, i)->trans_start;
		if (val && time_after(val, res))
			res = val;
	}
	dev->trans_start = res;
	return res;
}
EXPORT_SYMBOL(dev_trans_start);

L
Linus Torvalds 已提交
224 225 226 227
static void dev_watchdog(unsigned long arg)
{
	struct net_device *dev = (struct net_device *)arg;

H
Herbert Xu 已提交
228
	netif_tx_lock(dev);
229
	if (!qdisc_tx_is_noop(dev)) {
L
Linus Torvalds 已提交
230 231 232
		if (netif_device_present(dev) &&
		    netif_running(dev) &&
		    netif_carrier_ok(dev)) {
233
			int some_queue_timedout = 0;
234
			unsigned int i;
235
			unsigned long trans_start;
236 237 238 239 240

			for (i = 0; i < dev->num_tx_queues; i++) {
				struct netdev_queue *txq;

				txq = netdev_get_tx_queue(dev, i);
241 242 243 244 245 246 247 248
				/*
				 * old device drivers set dev->trans_start
				 */
				trans_start = txq->trans_start ? : dev->trans_start;
				if (netif_tx_queue_stopped(txq) &&
				    time_after(jiffies, (trans_start +
							 dev->watchdog_timeo))) {
					some_queue_timedout = 1;
249 250 251
					break;
				}
			}
252

253
			if (some_queue_timedout) {
254
				char drivername[64];
255 256
				WARN_ONCE(1, KERN_INFO "NETDEV WATCHDOG: %s (%s): transmit queue %u timed out\n",
				       dev->name, netdev_drivername(dev, drivername, 64), i);
257
				dev->netdev_ops->ndo_tx_timeout(dev);
L
Linus Torvalds 已提交
258
			}
259 260 261
			if (!mod_timer(&dev->watchdog_timer,
				       round_jiffies(jiffies +
						     dev->watchdog_timeo)))
L
Linus Torvalds 已提交
262 263 264
				dev_hold(dev);
		}
	}
H
Herbert Xu 已提交
265
	netif_tx_unlock(dev);
L
Linus Torvalds 已提交
266 267 268 269 270 271

	dev_put(dev);
}

void __netdev_watchdog_up(struct net_device *dev)
{
272
	if (dev->netdev_ops->ndo_tx_timeout) {
L
Linus Torvalds 已提交
273 274
		if (dev->watchdog_timeo <= 0)
			dev->watchdog_timeo = 5*HZ;
275 276
		if (!mod_timer(&dev->watchdog_timer,
			       round_jiffies(jiffies + dev->watchdog_timeo)))
L
Linus Torvalds 已提交
277 278 279 280 281 282 283 284 285 286 287
			dev_hold(dev);
	}
}

static void dev_watchdog_up(struct net_device *dev)
{
	__netdev_watchdog_up(dev);
}

static void dev_watchdog_down(struct net_device *dev)
{
H
Herbert Xu 已提交
288
	netif_tx_lock_bh(dev);
L
Linus Torvalds 已提交
289
	if (del_timer(&dev->watchdog_timer))
290
		dev_put(dev);
H
Herbert Xu 已提交
291
	netif_tx_unlock_bh(dev);
L
Linus Torvalds 已提交
292 293
}

294 295 296 297 298 299
/**
 *	netif_carrier_on - set carrier
 *	@dev: network device
 *
 * Device has detected that carrier.
 */
300 301
void netif_carrier_on(struct net_device *dev)
{
J
Jeff Garzik 已提交
302
	if (test_and_clear_bit(__LINK_STATE_NOCARRIER, &dev->state)) {
303 304
		if (dev->reg_state == NETREG_UNINITIALIZED)
			return;
305
		linkwatch_fire_event(dev);
J
Jeff Garzik 已提交
306 307 308
		if (netif_running(dev))
			__netdev_watchdog_up(dev);
	}
309
}
310
EXPORT_SYMBOL(netif_carrier_on);
311

312 313 314 315 316 317
/**
 *	netif_carrier_off - clear carrier
 *	@dev: network device
 *
 * Device has detected loss of carrier.
 */
318 319
void netif_carrier_off(struct net_device *dev)
{
320 321 322
	if (!test_and_set_bit(__LINK_STATE_NOCARRIER, &dev->state)) {
		if (dev->reg_state == NETREG_UNINITIALIZED)
			return;
323
		linkwatch_fire_event(dev);
324
	}
325
}
326
EXPORT_SYMBOL(netif_carrier_off);
327

328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345
/**
 * 	netif_notify_peers - notify network peers about existence of @dev
 * 	@dev: network device
 *
 * Generate traffic such that interested network peers are aware of
 * @dev, such as by generating a gratuitous ARP. This may be used when
 * a device wants to inform the rest of the network about some sort of
 * reconfiguration such as a failover event or virtual machine
 * migration.
 */
void netif_notify_peers(struct net_device *dev)
{
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
}
EXPORT_SYMBOL(netif_notify_peers);

L
Linus Torvalds 已提交
346 347 348 349 350
/* "NOOP" scheduler: the best scheduler, recommended for all interfaces
   under all circumstances. It is difficult to invent anything faster or
   cheaper.
 */

351
static int noop_enqueue(struct sk_buff *skb, struct Qdisc * qdisc)
L
Linus Torvalds 已提交
352 353 354 355 356
{
	kfree_skb(skb);
	return NET_XMIT_CN;
}

357
static struct sk_buff *noop_dequeue(struct Qdisc * qdisc)
L
Linus Torvalds 已提交
358 359 360 361
{
	return NULL;
}

362
struct Qdisc_ops noop_qdisc_ops __read_mostly = {
L
Linus Torvalds 已提交
363 364 365 366
	.id		=	"noop",
	.priv_size	=	0,
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
367
	.peek		=	noop_dequeue,
L
Linus Torvalds 已提交
368 369 370
	.owner		=	THIS_MODULE,
};

371 372
static struct netdev_queue noop_netdev_queue = {
	.qdisc		=	&noop_qdisc,
373
	.qdisc_sleeping	=	&noop_qdisc,
374 375
};

L
Linus Torvalds 已提交
376 377 378 379
struct Qdisc noop_qdisc = {
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
	.flags		=	TCQ_F_BUILTIN,
380
	.ops		=	&noop_qdisc_ops,
L
Linus Torvalds 已提交
381
	.list		=	LIST_HEAD_INIT(noop_qdisc.list),
382
	.q.lock		=	__SPIN_LOCK_UNLOCKED(noop_qdisc.q.lock),
383
	.dev_queue	=	&noop_netdev_queue,
384
	.busylock	=	__SPIN_LOCK_UNLOCKED(noop_qdisc.busylock),
L
Linus Torvalds 已提交
385
};
386
EXPORT_SYMBOL(noop_qdisc);
L
Linus Torvalds 已提交
387

388
static struct Qdisc_ops noqueue_qdisc_ops __read_mostly = {
L
Linus Torvalds 已提交
389 390 391 392
	.id		=	"noqueue",
	.priv_size	=	0,
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
393
	.peek		=	noop_dequeue,
L
Linus Torvalds 已提交
394 395 396
	.owner		=	THIS_MODULE,
};

397 398 399
static struct Qdisc noqueue_qdisc;
static struct netdev_queue noqueue_netdev_queue = {
	.qdisc		=	&noqueue_qdisc,
400
	.qdisc_sleeping	=	&noqueue_qdisc,
401 402
};

L
Linus Torvalds 已提交
403 404 405 406 407 408
static struct Qdisc noqueue_qdisc = {
	.enqueue	=	NULL,
	.dequeue	=	noop_dequeue,
	.flags		=	TCQ_F_BUILTIN,
	.ops		=	&noqueue_qdisc_ops,
	.list		=	LIST_HEAD_INIT(noqueue_qdisc.list),
409 410
	.q.lock		=	__SPIN_LOCK_UNLOCKED(noqueue_qdisc.q.lock),
	.dev_queue	=	&noqueue_netdev_queue,
411
	.busylock	=	__SPIN_LOCK_UNLOCKED(noqueue_qdisc.busylock),
L
Linus Torvalds 已提交
412 413 414
};


E
Eric Dumazet 已提交
415 416 417
static const u8 prio2band[TC_PRIO_MAX + 1] = {
	1, 2, 2, 2, 1, 2, 0, 0 , 1, 1, 1, 1, 1, 1, 1, 1
};
418 419 420 421 422 423 424

/* 3-band FIFO queue: old style, but should be a bit faster than
   generic prio+fifo combination.
 */

#define PFIFO_FAST_BANDS 3

425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444
/*
 * Private data for a pfifo_fast scheduler containing:
 * 	- queues for the three band
 * 	- bitmap indicating which of the bands contain skbs
 */
struct pfifo_fast_priv {
	u32 bitmap;
	struct sk_buff_head q[PFIFO_FAST_BANDS];
};

/*
 * Convert a bitmap to the first band number where an skb is queued, where:
 * 	bitmap=0 means there are no skbs on any band.
 * 	bitmap=1 means there is an skb on band 0.
 *	bitmap=7 means there are skbs on all 3 bands, etc.
 */
static const int bitmap2band[] = {-1, 0, 1, 0, 2, 0, 1, 0};

static inline struct sk_buff_head *band2list(struct pfifo_fast_priv *priv,
					     int band)
445
{
446
	return priv->q + band;
447 448
}

E
Eric Dumazet 已提交
449
static int pfifo_fast_enqueue(struct sk_buff *skb, struct Qdisc *qdisc)
450
{
451 452 453 454
	if (skb_queue_len(&qdisc->q) < qdisc_dev(qdisc)->tx_queue_len) {
		int band = prio2band[skb->priority & TC_PRIO_MAX];
		struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
		struct sk_buff_head *list = band2list(priv, band);
L
Linus Torvalds 已提交
455

456
		priv->bitmap |= (1 << band);
457
		qdisc->q.qlen++;
458
		return __qdisc_enqueue_tail(skb, qdisc, list);
459
	}
460 461

	return qdisc_drop(skb, qdisc);
L
Linus Torvalds 已提交
462 463
}

E
Eric Dumazet 已提交
464
static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
465
{
466 467
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
	int band = bitmap2band[priv->bitmap];
L
Linus Torvalds 已提交
468

469 470 471 472 473 474 475 476 477
	if (likely(band >= 0)) {
		struct sk_buff_head *list = band2list(priv, band);
		struct sk_buff *skb = __qdisc_dequeue_head(qdisc, list);

		qdisc->q.qlen--;
		if (skb_queue_empty(list))
			priv->bitmap &= ~(1 << band);

		return skb;
478
	}
479

L
Linus Torvalds 已提交
480 481 482
	return NULL;
}

E
Eric Dumazet 已提交
483
static struct sk_buff *pfifo_fast_peek(struct Qdisc *qdisc)
484
{
485 486 487 488 489
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
	int band = bitmap2band[priv->bitmap];

	if (band >= 0) {
		struct sk_buff_head *list = band2list(priv, band);
490

491
		return skb_peek(list);
492 493 494 495 496
	}

	return NULL;
}

E
Eric Dumazet 已提交
497
static void pfifo_fast_reset(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
498
{
499
	int prio;
500
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
501 502

	for (prio = 0; prio < PFIFO_FAST_BANDS; prio++)
503
		__qdisc_reset_queue(qdisc, band2list(priv, prio));
504

505
	priv->bitmap = 0;
506
	qdisc->qstats.backlog = 0;
507
	qdisc->q.qlen = 0;
L
Linus Torvalds 已提交
508 509
}

510 511 512 513
static int pfifo_fast_dump(struct Qdisc *qdisc, struct sk_buff *skb)
{
	struct tc_prio_qopt opt = { .bands = PFIFO_FAST_BANDS };

E
Eric Dumazet 已提交
514
	memcpy(&opt.priomap, prio2band, TC_PRIO_MAX + 1);
515 516 517 518 519 520 521 522 523 524
	NLA_PUT(skb, TCA_OPTIONS, sizeof(opt), &opt);
	return skb->len;

nla_put_failure:
	return -1;
}

static int pfifo_fast_init(struct Qdisc *qdisc, struct nlattr *opt)
{
	int prio;
525
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
526 527

	for (prio = 0; prio < PFIFO_FAST_BANDS; prio++)
528
		skb_queue_head_init(band2list(priv, prio));
529 530 531 532

	return 0;
}

533
struct Qdisc_ops pfifo_fast_ops __read_mostly = {
534
	.id		=	"pfifo_fast",
535
	.priv_size	=	sizeof(struct pfifo_fast_priv),
536 537
	.enqueue	=	pfifo_fast_enqueue,
	.dequeue	=	pfifo_fast_dequeue,
538
	.peek		=	pfifo_fast_peek,
539 540 541
	.init		=	pfifo_fast_init,
	.reset		=	pfifo_fast_reset,
	.dump		=	pfifo_fast_dump,
L
Linus Torvalds 已提交
542 543
	.owner		=	THIS_MODULE,
};
544
EXPORT_SYMBOL(pfifo_fast_ops);
L
Linus Torvalds 已提交
545

546
struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
547
			  struct Qdisc_ops *ops)
L
Linus Torvalds 已提交
548 549 550
{
	void *p;
	struct Qdisc *sch;
551 552
	unsigned int size;
	int err = -ENOBUFS;
L
Linus Torvalds 已提交
553

E
Eric Dumazet 已提交
554
	/* ensure that the Qdisc and the private data are 64-byte aligned */
555 556
	size = QDISC_ALIGN(sizeof(*sch));
	size += ops->priv_size + (QDISC_ALIGNTO - 1);
L
Linus Torvalds 已提交
557

558 559 560
	p = kzalloc_node(size, GFP_KERNEL,
			 netdev_queue_numa_node_read(dev_queue));

L
Linus Torvalds 已提交
561
	if (!p)
562 563 564
		goto errout;
	sch = (struct Qdisc *) QDISC_ALIGN((unsigned long) p);
	sch->padded = (char *) sch - (char *) p;
L
Linus Torvalds 已提交
565 566 567

	INIT_LIST_HEAD(&sch->list);
	skb_queue_head_init(&sch->q);
568
	spin_lock_init(&sch->busylock);
L
Linus Torvalds 已提交
569 570 571
	sch->ops = ops;
	sch->enqueue = ops->enqueue;
	sch->dequeue = ops->dequeue;
572
	sch->dev_queue = dev_queue;
573
	dev_hold(qdisc_dev(sch));
L
Linus Torvalds 已提交
574
	atomic_set(&sch->refcnt, 1);
575 576 577

	return sch;
errout:
578
	return ERR_PTR(err);
579 580
}

581 582
struct Qdisc *qdisc_create_dflt(struct netdev_queue *dev_queue,
				struct Qdisc_ops *ops, unsigned int parentid)
583 584
{
	struct Qdisc *sch;
585

586
	sch = qdisc_alloc(dev_queue, ops);
587 588
	if (IS_ERR(sch))
		goto errout;
589
	sch->parent = parentid;
590

L
Linus Torvalds 已提交
591 592 593
	if (!ops->init || ops->init(sch, NULL) == 0)
		return sch;

594
	qdisc_destroy(sch);
595
errout:
L
Linus Torvalds 已提交
596 597
	return NULL;
}
598
EXPORT_SYMBOL(qdisc_create_dflt);
L
Linus Torvalds 已提交
599

600
/* Under qdisc_lock(qdisc) and BH! */
L
Linus Torvalds 已提交
601 602 603

void qdisc_reset(struct Qdisc *qdisc)
{
604
	const struct Qdisc_ops *ops = qdisc->ops;
L
Linus Torvalds 已提交
605 606 607

	if (ops->reset)
		ops->reset(qdisc);
608

609 610 611 612 613
	if (qdisc->gso_skb) {
		kfree_skb(qdisc->gso_skb);
		qdisc->gso_skb = NULL;
		qdisc->q.qlen = 0;
	}
L
Linus Torvalds 已提交
614
}
615
EXPORT_SYMBOL(qdisc_reset);
L
Linus Torvalds 已提交
616

E
Eric Dumazet 已提交
617 618 619 620 621 622 623
static void qdisc_rcu_free(struct rcu_head *head)
{
	struct Qdisc *qdisc = container_of(head, struct Qdisc, rcu_head);

	kfree((char *) qdisc - qdisc->padded);
}

624
void qdisc_destroy(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
625
{
626 627
	const struct Qdisc_ops  *ops = qdisc->ops;

628 629 630 631
	if (qdisc->flags & TCQ_F_BUILTIN ||
	    !atomic_dec_and_test(&qdisc->refcnt))
		return;

632
#ifdef CONFIG_NET_SCHED
633 634
	qdisc_list_del(qdisc);

E
Eric Dumazet 已提交
635
	qdisc_put_stab(rtnl_dereference(qdisc->stab));
636
#endif
637 638 639 640 641 642 643 644 645
	gen_kill_estimator(&qdisc->bstats, &qdisc->rate_est);
	if (ops->reset)
		ops->reset(qdisc);
	if (ops->destroy)
		ops->destroy(qdisc);

	module_put(ops->owner);
	dev_put(qdisc_dev(qdisc));

646
	kfree_skb(qdisc->gso_skb);
E
Eric Dumazet 已提交
647 648 649 650 651
	/*
	 * gen_estimator est_timer() might access qdisc->q.lock,
	 * wait a RCU grace period before freeing qdisc.
	 */
	call_rcu(&qdisc->rcu_head, qdisc_rcu_free);
L
Linus Torvalds 已提交
652
}
653
EXPORT_SYMBOL(qdisc_destroy);
L
Linus Torvalds 已提交
654

655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
/* Attach toplevel qdisc to device queue. */
struct Qdisc *dev_graft_qdisc(struct netdev_queue *dev_queue,
			      struct Qdisc *qdisc)
{
	struct Qdisc *oqdisc = dev_queue->qdisc_sleeping;
	spinlock_t *root_lock;

	root_lock = qdisc_lock(oqdisc);
	spin_lock_bh(root_lock);

	/* Prune old scheduler */
	if (oqdisc && atomic_read(&oqdisc->refcnt) <= 1)
		qdisc_reset(oqdisc);

	/* ... and graft new one */
	if (qdisc == NULL)
		qdisc = &noop_qdisc;
	dev_queue->qdisc_sleeping = qdisc;
	rcu_assign_pointer(dev_queue->qdisc, &noop_qdisc);

	spin_unlock_bh(root_lock);

	return oqdisc;
}
679
EXPORT_SYMBOL(dev_graft_qdisc);
680

681 682 683 684
static void attach_one_default_qdisc(struct net_device *dev,
				     struct netdev_queue *dev_queue,
				     void *_unused)
{
E
Eric Dumazet 已提交
685
	struct Qdisc *qdisc = &noqueue_qdisc;
686 687

	if (dev->tx_queue_len) {
688
		qdisc = qdisc_create_dflt(dev_queue,
689
					  &pfifo_fast_ops, TC_H_ROOT);
690
		if (!qdisc) {
E
Eric Dumazet 已提交
691
			netdev_info(dev, "activation failed\n");
692 693
			return;
		}
694 695 696

		/* Can by-pass the queue discipline for default qdisc */
		qdisc->flags |= TCQ_F_CAN_BYPASS;
697 698 699 700
	}
	dev_queue->qdisc_sleeping = qdisc;
}

701 702 703 704 705 706 707 708 709 710 711 712
static void attach_default_qdiscs(struct net_device *dev)
{
	struct netdev_queue *txq;
	struct Qdisc *qdisc;

	txq = netdev_get_tx_queue(dev, 0);

	if (!netif_is_multiqueue(dev) || dev->tx_queue_len == 0) {
		netdev_for_each_tx_queue(dev, attach_one_default_qdisc, NULL);
		dev->qdisc = txq->qdisc_sleeping;
		atomic_inc(&dev->qdisc->refcnt);
	} else {
713
		qdisc = qdisc_create_dflt(txq, &mq_qdisc_ops, TC_H_ROOT);
714 715 716 717 718 719 720
		if (qdisc) {
			qdisc->ops->attach(qdisc);
			dev->qdisc = qdisc;
		}
	}
}

721 722 723 724
static void transition_one_qdisc(struct net_device *dev,
				 struct netdev_queue *dev_queue,
				 void *_need_watchdog)
{
725
	struct Qdisc *new_qdisc = dev_queue->qdisc_sleeping;
726 727
	int *need_watchdog_p = _need_watchdog;

728 729 730
	if (!(new_qdisc->flags & TCQ_F_BUILTIN))
		clear_bit(__QDISC_STATE_DEACTIVATED, &new_qdisc->state);

731
	rcu_assign_pointer(dev_queue->qdisc, new_qdisc);
732 733
	if (need_watchdog_p && new_qdisc != &noqueue_qdisc) {
		dev_queue->trans_start = 0;
734
		*need_watchdog_p = 1;
735
	}
736 737
}

L
Linus Torvalds 已提交
738 739
void dev_activate(struct net_device *dev)
{
740
	int need_watchdog;
741

L
Linus Torvalds 已提交
742
	/* No queueing discipline is attached to device;
743 744 745
	   create default one i.e. pfifo_fast for devices,
	   which need queueing and noqueue_qdisc for
	   virtual interfaces
L
Linus Torvalds 已提交
746 747
	 */

748 749
	if (dev->qdisc == &noop_qdisc)
		attach_default_qdiscs(dev);
750

751 752 753 754
	if (!netif_carrier_ok(dev))
		/* Delay activation until next carrier-on event */
		return;

755 756
	need_watchdog = 0;
	netdev_for_each_tx_queue(dev, transition_one_qdisc, &need_watchdog);
757 758
	if (dev_ingress_queue(dev))
		transition_one_qdisc(dev, dev_ingress_queue(dev), NULL);
759 760

	if (need_watchdog) {
L
Linus Torvalds 已提交
761 762 763
		dev->trans_start = jiffies;
		dev_watchdog_up(dev);
	}
764
}
765
EXPORT_SYMBOL(dev_activate);
766

767 768 769
static void dev_deactivate_queue(struct net_device *dev,
				 struct netdev_queue *dev_queue,
				 void *_qdisc_default)
770
{
771
	struct Qdisc *qdisc_default = _qdisc_default;
772 773 774
	struct Qdisc *qdisc;

	qdisc = dev_queue->qdisc;
775
	if (qdisc) {
776 777
		spin_lock_bh(qdisc_lock(qdisc));

778 779 780
		if (!(qdisc->flags & TCQ_F_BUILTIN))
			set_bit(__QDISC_STATE_DEACTIVATED, &qdisc->state);

781
		rcu_assign_pointer(dev_queue->qdisc, qdisc_default);
782
		qdisc_reset(qdisc);
783

784
		spin_unlock_bh(qdisc_lock(qdisc));
785
	}
L
Linus Torvalds 已提交
786 787
}

788
static bool some_qdisc_is_busy(struct net_device *dev)
789 790 791 792 793
{
	unsigned int i;

	for (i = 0; i < dev->num_tx_queues; i++) {
		struct netdev_queue *dev_queue;
794
		spinlock_t *root_lock;
795
		struct Qdisc *q;
796 797 798
		int val;

		dev_queue = netdev_get_tx_queue(dev, i);
799
		q = dev_queue->qdisc_sleeping;
800
		root_lock = qdisc_lock(q);
801

802
		spin_lock_bh(root_lock);
803

804
		val = (qdisc_is_running(q) ||
805
		       test_bit(__QDISC_STATE_SCHED, &q->state));
806

807
		spin_unlock_bh(root_lock);
808 809 810 811 812 813 814

		if (val)
			return true;
	}
	return false;
}

815
void dev_deactivate_many(struct list_head *head)
L
Linus Torvalds 已提交
816
{
817
	struct net_device *dev;
818

819 820 821 822 823 824 825 826 827
	list_for_each_entry(dev, head, unreg_list) {
		netdev_for_each_tx_queue(dev, dev_deactivate_queue,
					 &noop_qdisc);
		if (dev_ingress_queue(dev))
			dev_deactivate_queue(dev, dev_ingress_queue(dev),
					     &noop_qdisc);

		dev_watchdog_down(dev);
	}
L
Linus Torvalds 已提交
828

829
	/* Wait for outstanding qdisc-less dev_queue_xmit calls. */
830
	synchronize_rcu();
L
Linus Torvalds 已提交
831

832
	/* Wait for outstanding qdisc_run calls. */
833 834 835 836 837 838 839 840 841 842 843
	list_for_each_entry(dev, head, unreg_list)
		while (some_qdisc_is_busy(dev))
			yield();
}

void dev_deactivate(struct net_device *dev)
{
	LIST_HEAD(single);

	list_add(&dev->unreg_list, &single);
	dev_deactivate_many(&single);
L
Linus Torvalds 已提交
844
}
845
EXPORT_SYMBOL(dev_deactivate);
L
Linus Torvalds 已提交
846

847 848
static void dev_init_scheduler_queue(struct net_device *dev,
				     struct netdev_queue *dev_queue,
849
				     void *_qdisc)
850
{
851 852
	struct Qdisc *qdisc = _qdisc;

853 854 855 856
	dev_queue->qdisc = qdisc;
	dev_queue->qdisc_sleeping = qdisc;
}

L
Linus Torvalds 已提交
857 858
void dev_init_scheduler(struct net_device *dev)
{
859
	dev->qdisc = &noop_qdisc;
860
	netdev_for_each_tx_queue(dev, dev_init_scheduler_queue, &noop_qdisc);
861 862
	if (dev_ingress_queue(dev))
		dev_init_scheduler_queue(dev, dev_ingress_queue(dev), &noop_qdisc);
L
Linus Torvalds 已提交
863

864
	setup_timer(&dev->watchdog_timer, dev_watchdog, (unsigned long)dev);
L
Linus Torvalds 已提交
865 866
}

867 868 869
static void shutdown_scheduler_queue(struct net_device *dev,
				     struct netdev_queue *dev_queue,
				     void *_qdisc_default)
L
Linus Torvalds 已提交
870
{
871
	struct Qdisc *qdisc = dev_queue->qdisc_sleeping;
872
	struct Qdisc *qdisc_default = _qdisc_default;
873 874

	if (qdisc) {
875
		rcu_assign_pointer(dev_queue->qdisc, qdisc_default);
876
		dev_queue->qdisc_sleeping = qdisc_default;
L
Linus Torvalds 已提交
877 878

		qdisc_destroy(qdisc);
879
	}
880 881 882 883
}

void dev_shutdown(struct net_device *dev)
{
884
	netdev_for_each_tx_queue(dev, shutdown_scheduler_queue, &noop_qdisc);
885 886
	if (dev_ingress_queue(dev))
		shutdown_scheduler_queue(dev, dev_ingress_queue(dev), &noop_qdisc);
887 888 889
	qdisc_destroy(dev->qdisc);
	dev->qdisc = &noop_qdisc;

890
	WARN_ON(timer_pending(&dev->watchdog_timer));
L
Linus Torvalds 已提交
891
}