sch_generic.c 22.4 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>
28
#include <net/sch_generic.h>
L
Linus Torvalds 已提交
29
#include <net/pkt_sched.h>
E
Eric Dumazet 已提交
30
#include <net/dst.h>
L
Linus Torvalds 已提交
31 32 33

/* Main transmission queue. */

34
/* Modifications to data participating in scheduling must be protected with
35
 * qdisc_lock(qdisc) spinlock.
36 37
 *
 * The idea is the following:
38 39
 * - enqueue, dequeue are serialized via qdisc root lock
 * - ingress filtering is also serialized via qdisc root lock
40
 * - updates to tree and tree walking are only done under the rtnl mutex.
L
Linus Torvalds 已提交
41 42
 */

43
static inline int dev_requeue_skb(struct sk_buff *skb, struct Qdisc *q)
44
{
E
Eric Dumazet 已提交
45
	skb_dst_force(skb);
46
	q->gso_skb = skb;
47
	q->qstats.requeues++;
48
	q->q.qlen++;	/* it's still part of the queue */
49
	__netif_schedule(q);
50

51 52 53
	return 0;
}

54
static inline struct sk_buff *dequeue_skb(struct Qdisc *q)
55
{
56
	struct sk_buff *skb = q->gso_skb;
57
	const struct netdev_queue *txq = q->dev_queue;
58

59 60
	if (unlikely(skb)) {
		/* check the reason of requeuing without tx lock first */
61
		txq = netdev_get_tx_queue(txq->dev, skb_get_queue_mapping(skb));
62
		if (!netif_xmit_frozen_or_stopped(txq)) {
63
			q->gso_skb = NULL;
64 65
			q->q.qlen--;
		} else
66 67
			skb = NULL;
	} else {
68 69
		if (!(q->flags & TCQ_F_ONETXQUEUE) || !netif_xmit_frozen_or_stopped(txq))
			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 90
		net_warn_ratelimited("Dead loop on netdevice %s, fix it urgently!\n",
				     dev_queue->dev->name);
91 92 93 94 95 96
		ret = qdisc_qlen(q);
	} else {
		/*
		 * Another cpu is holding lock, requeue & delay xmits for
		 * some time.
		 */
E
Eric Dumazet 已提交
97
		__this_cpu_inc(softnet_data.cpu_collision);
98
		ret = dev_requeue_skb(skb, q);
99 100
	}

101
	return ret;
102 103
}

104
/*
105 106 107
 * 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.
108 109 110 111 112
 *
 * Returns to the caller:
 *				0  - queue is empty or throttled.
 *				>0 - queue is not empty.
 */
113 114 115
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
116
{
117
	int ret = NETDEV_TX_BUSY;
118 119 120

	/* And release qdisc */
	spin_unlock(root_lock);
121

122
	HARD_TX_LOCK(dev, txq, smp_processor_id());
123
	if (!netif_xmit_frozen_or_stopped(txq))
124
		ret = dev_hard_start_xmit(skb, dev, txq);
125

126
	HARD_TX_UNLOCK(dev, txq);
127

128
	spin_lock(root_lock);
129

130 131
	if (dev_xmit_complete(ret)) {
		/* Driver sent out skb successfully or skb was consumed */
132
		ret = qdisc_qlen(q);
133
	} else if (ret == NETDEV_TX_LOCKED) {
134
		/* Driver try lock failed */
135
		ret = handle_dev_cpu_collision(skb, txq, q);
136
	} else {
137
		/* Driver returned NETDEV_TX_BUSY - requeue skb */
138 139 140
		if (unlikely(ret != NETDEV_TX_BUSY))
			net_warn_ratelimited("BUG %s code %d qlen %d\n",
					     dev->name, ret, q->q.qlen);
141

142
		ret = dev_requeue_skb(skb, q);
143
	}
144

145
	if (ret && netif_xmit_frozen_or_stopped(txq))
146 147
		ret = 0;

148
	return ret;
L
Linus Torvalds 已提交
149 150
}

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
/*
 * 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
181
	WARN_ON_ONCE(skb_dst_is_noref(skb));
182 183 184 185 186 187 188
	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);
}

189
void __qdisc_run(struct Qdisc *q)
H
Herbert Xu 已提交
190
{
J
jamal 已提交
191
	int quota = weight_p;
192

193
	while (qdisc_restart(q)) {
194
		/*
J
jamal 已提交
195 196 197
		 * Ordered by possible occurrence: Postpone processing if
		 * 1. we've exceeded packet quota
		 * 2. another process needs the CPU;
198
		 */
199
		if (--quota <= 0 || need_resched()) {
200
			__netif_schedule(q);
201
			break;
202 203
		}
	}
H
Herbert Xu 已提交
204

205
	qdisc_run_end(q);
H
Herbert Xu 已提交
206 207
}

208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
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
223 224 225 226
static void dev_watchdog(unsigned long arg)
{
	struct net_device *dev = (struct net_device *)arg;

H
Herbert Xu 已提交
227
	netif_tx_lock(dev);
228
	if (!qdisc_tx_is_noop(dev)) {
L
Linus Torvalds 已提交
229 230 231
		if (netif_device_present(dev) &&
		    netif_running(dev) &&
		    netif_carrier_ok(dev)) {
232
			int some_queue_timedout = 0;
233
			unsigned int i;
234
			unsigned long trans_start;
235 236 237 238 239

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

				txq = netdev_get_tx_queue(dev, i);
240 241 242 243
				/*
				 * old device drivers set dev->trans_start
				 */
				trans_start = txq->trans_start ? : dev->trans_start;
244
				if (netif_xmit_stopped(txq) &&
245 246 247
				    time_after(jiffies, (trans_start +
							 dev->watchdog_timeo))) {
					some_queue_timedout = 1;
248
					txq->trans_timeout++;
249 250 251
					break;
				}
			}
252

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

	dev_put(dev);
}

void __netdev_watchdog_up(struct net_device *dev)
{
271
	if (dev->netdev_ops->ndo_tx_timeout) {
L
Linus Torvalds 已提交
272 273
		if (dev->watchdog_timeo <= 0)
			dev->watchdog_timeo = 5*HZ;
274 275
		if (!mod_timer(&dev->watchdog_timer,
			       round_jiffies(jiffies + dev->watchdog_timeo)))
L
Linus Torvalds 已提交
276 277 278 279 280 281 282 283 284 285 286
			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
287
	netif_tx_lock_bh(dev);
L
Linus Torvalds 已提交
288
	if (del_timer(&dev->watchdog_timer))
289
		dev_put(dev);
H
Herbert Xu 已提交
290
	netif_tx_unlock_bh(dev);
L
Linus Torvalds 已提交
291 292
}

293 294 295 296 297 298
/**
 *	netif_carrier_on - set carrier
 *	@dev: network device
 *
 * Device has detected that carrier.
 */
299 300
void netif_carrier_on(struct net_device *dev)
{
J
Jeff Garzik 已提交
301
	if (test_and_clear_bit(__LINK_STATE_NOCARRIER, &dev->state)) {
302 303
		if (dev->reg_state == NETREG_UNINITIALIZED)
			return;
304
		linkwatch_fire_event(dev);
J
Jeff Garzik 已提交
305 306 307
		if (netif_running(dev))
			__netdev_watchdog_up(dev);
	}
308
}
309
EXPORT_SYMBOL(netif_carrier_on);
310

311 312 313 314 315 316
/**
 *	netif_carrier_off - clear carrier
 *	@dev: network device
 *
 * Device has detected loss of carrier.
 */
317 318
void netif_carrier_off(struct net_device *dev)
{
319 320 321
	if (!test_and_set_bit(__LINK_STATE_NOCARRIER, &dev->state)) {
		if (dev->reg_state == NETREG_UNINITIALIZED)
			return;
322
		linkwatch_fire_event(dev);
323
	}
324
}
325
EXPORT_SYMBOL(netif_carrier_off);
326

L
Linus Torvalds 已提交
327 328 329 330 331
/* "NOOP" scheduler: the best scheduler, recommended for all interfaces
   under all circumstances. It is difficult to invent anything faster or
   cheaper.
 */

332
static int noop_enqueue(struct sk_buff *skb, struct Qdisc * qdisc)
L
Linus Torvalds 已提交
333 334 335 336 337
{
	kfree_skb(skb);
	return NET_XMIT_CN;
}

338
static struct sk_buff *noop_dequeue(struct Qdisc * qdisc)
L
Linus Torvalds 已提交
339 340 341 342
{
	return NULL;
}

343
struct Qdisc_ops noop_qdisc_ops __read_mostly = {
L
Linus Torvalds 已提交
344 345 346 347
	.id		=	"noop",
	.priv_size	=	0,
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
348
	.peek		=	noop_dequeue,
L
Linus Torvalds 已提交
349 350 351
	.owner		=	THIS_MODULE,
};

352 353
static struct netdev_queue noop_netdev_queue = {
	.qdisc		=	&noop_qdisc,
354
	.qdisc_sleeping	=	&noop_qdisc,
355 356
};

L
Linus Torvalds 已提交
357 358 359 360
struct Qdisc noop_qdisc = {
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
	.flags		=	TCQ_F_BUILTIN,
361
	.ops		=	&noop_qdisc_ops,
L
Linus Torvalds 已提交
362
	.list		=	LIST_HEAD_INIT(noop_qdisc.list),
363
	.q.lock		=	__SPIN_LOCK_UNLOCKED(noop_qdisc.q.lock),
364
	.dev_queue	=	&noop_netdev_queue,
365
	.busylock	=	__SPIN_LOCK_UNLOCKED(noop_qdisc.busylock),
L
Linus Torvalds 已提交
366
};
367
EXPORT_SYMBOL(noop_qdisc);
L
Linus Torvalds 已提交
368

369
static struct Qdisc_ops noqueue_qdisc_ops __read_mostly = {
L
Linus Torvalds 已提交
370 371 372 373
	.id		=	"noqueue",
	.priv_size	=	0,
	.enqueue	=	noop_enqueue,
	.dequeue	=	noop_dequeue,
374
	.peek		=	noop_dequeue,
L
Linus Torvalds 已提交
375 376 377
	.owner		=	THIS_MODULE,
};

378 379 380
static struct Qdisc noqueue_qdisc;
static struct netdev_queue noqueue_netdev_queue = {
	.qdisc		=	&noqueue_qdisc,
381
	.qdisc_sleeping	=	&noqueue_qdisc,
382 383
};

L
Linus Torvalds 已提交
384 385 386 387 388 389
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),
390 391
	.q.lock		=	__SPIN_LOCK_UNLOCKED(noqueue_qdisc.q.lock),
	.dev_queue	=	&noqueue_netdev_queue,
392
	.busylock	=	__SPIN_LOCK_UNLOCKED(noqueue_qdisc.busylock),
L
Linus Torvalds 已提交
393 394 395
};


E
Eric Dumazet 已提交
396 397 398
static const u8 prio2band[TC_PRIO_MAX + 1] = {
	1, 2, 2, 2, 1, 2, 0, 0 , 1, 1, 1, 1, 1, 1, 1, 1
};
399 400 401 402 403 404 405

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

#define PFIFO_FAST_BANDS 3

406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
/*
 * 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)
426
{
427
	return priv->q + band;
428 429
}

E
Eric Dumazet 已提交
430
static int pfifo_fast_enqueue(struct sk_buff *skb, struct Qdisc *qdisc)
431
{
432 433 434 435
	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
436

437
		priv->bitmap |= (1 << band);
438
		qdisc->q.qlen++;
439
		return __qdisc_enqueue_tail(skb, qdisc, list);
440
	}
441 442

	return qdisc_drop(skb, qdisc);
L
Linus Torvalds 已提交
443 444
}

E
Eric Dumazet 已提交
445
static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
446
{
447 448
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
	int band = bitmap2band[priv->bitmap];
L
Linus Torvalds 已提交
449

450 451 452 453 454 455 456 457 458
	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;
459
	}
460

L
Linus Torvalds 已提交
461 462 463
	return NULL;
}

E
Eric Dumazet 已提交
464
static struct sk_buff *pfifo_fast_peek(struct Qdisc *qdisc)
465
{
466 467 468 469 470
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
	int band = bitmap2band[priv->bitmap];

	if (band >= 0) {
		struct sk_buff_head *list = band2list(priv, band);
471

472
		return skb_peek(list);
473 474 475 476 477
	}

	return NULL;
}

E
Eric Dumazet 已提交
478
static void pfifo_fast_reset(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
479
{
480
	int prio;
481
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
482 483

	for (prio = 0; prio < PFIFO_FAST_BANDS; prio++)
484
		__qdisc_reset_queue(qdisc, band2list(priv, prio));
485

486
	priv->bitmap = 0;
487
	qdisc->qstats.backlog = 0;
488
	qdisc->q.qlen = 0;
L
Linus Torvalds 已提交
489 490
}

491 492 493 494
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
495
	memcpy(&opt.priomap, prio2band, TC_PRIO_MAX + 1);
496 497
	if (nla_put(skb, TCA_OPTIONS, sizeof(opt), &opt))
		goto nla_put_failure;
498 499 500 501 502 503 504 505 506
	return skb->len;

nla_put_failure:
	return -1;
}

static int pfifo_fast_init(struct Qdisc *qdisc, struct nlattr *opt)
{
	int prio;
507
	struct pfifo_fast_priv *priv = qdisc_priv(qdisc);
508 509

	for (prio = 0; prio < PFIFO_FAST_BANDS; prio++)
510
		skb_queue_head_init(band2list(priv, prio));
511

512 513
	/* Can by-pass the queue discipline */
	qdisc->flags |= TCQ_F_CAN_BYPASS;
514 515 516
	return 0;
}

517
struct Qdisc_ops pfifo_fast_ops __read_mostly = {
518
	.id		=	"pfifo_fast",
519
	.priv_size	=	sizeof(struct pfifo_fast_priv),
520 521
	.enqueue	=	pfifo_fast_enqueue,
	.dequeue	=	pfifo_fast_dequeue,
522
	.peek		=	pfifo_fast_peek,
523 524 525
	.init		=	pfifo_fast_init,
	.reset		=	pfifo_fast_reset,
	.dump		=	pfifo_fast_dump,
L
Linus Torvalds 已提交
526 527
	.owner		=	THIS_MODULE,
};
528
EXPORT_SYMBOL(pfifo_fast_ops);
L
Linus Torvalds 已提交
529

530 531
static struct lock_class_key qdisc_tx_busylock;

532
struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
533
			  struct Qdisc_ops *ops)
L
Linus Torvalds 已提交
534 535 536
{
	void *p;
	struct Qdisc *sch;
E
Eric Dumazet 已提交
537
	unsigned int size = QDISC_ALIGN(sizeof(*sch)) + ops->priv_size;
538
	int err = -ENOBUFS;
539
	struct net_device *dev = dev_queue->dev;
L
Linus Torvalds 已提交
540

541 542 543
	p = kzalloc_node(size, GFP_KERNEL,
			 netdev_queue_numa_node_read(dev_queue));

L
Linus Torvalds 已提交
544
	if (!p)
545 546
		goto errout;
	sch = (struct Qdisc *) QDISC_ALIGN((unsigned long) p);
E
Eric Dumazet 已提交
547 548 549 550 551 552 553 554 555 556
	/* if we got non aligned memory, ask more and do alignment ourself */
	if (sch != p) {
		kfree(p);
		p = kzalloc_node(size + QDISC_ALIGNTO - 1, GFP_KERNEL,
				 netdev_queue_numa_node_read(dev_queue));
		if (!p)
			goto errout;
		sch = (struct Qdisc *) QDISC_ALIGN((unsigned long) p);
		sch->padded = (char *) sch - (char *) p;
	}
L
Linus Torvalds 已提交
557 558
	INIT_LIST_HEAD(&sch->list);
	skb_queue_head_init(&sch->q);
559

560
	spin_lock_init(&sch->busylock);
561 562 563
	lockdep_set_class(&sch->busylock,
			  dev->qdisc_tx_busylock ?: &qdisc_tx_busylock);

L
Linus Torvalds 已提交
564 565 566
	sch->ops = ops;
	sch->enqueue = ops->enqueue;
	sch->dequeue = ops->dequeue;
567
	sch->dev_queue = dev_queue;
568
	dev_hold(dev);
L
Linus Torvalds 已提交
569
	atomic_set(&sch->refcnt, 1);
570 571 572

	return sch;
errout:
573
	return ERR_PTR(err);
574 575
}

576 577
struct Qdisc *qdisc_create_dflt(struct netdev_queue *dev_queue,
				struct Qdisc_ops *ops, unsigned int parentid)
578 579
{
	struct Qdisc *sch;
580

581
	sch = qdisc_alloc(dev_queue, ops);
582 583
	if (IS_ERR(sch))
		goto errout;
584
	sch->parent = parentid;
585

L
Linus Torvalds 已提交
586 587 588
	if (!ops->init || ops->init(sch, NULL) == 0)
		return sch;

589
	qdisc_destroy(sch);
590
errout:
L
Linus Torvalds 已提交
591 592
	return NULL;
}
593
EXPORT_SYMBOL(qdisc_create_dflt);
L
Linus Torvalds 已提交
594

595
/* Under qdisc_lock(qdisc) and BH! */
L
Linus Torvalds 已提交
596 597 598

void qdisc_reset(struct Qdisc *qdisc)
{
599
	const struct Qdisc_ops *ops = qdisc->ops;
L
Linus Torvalds 已提交
600 601 602

	if (ops->reset)
		ops->reset(qdisc);
603

604 605 606 607 608
	if (qdisc->gso_skb) {
		kfree_skb(qdisc->gso_skb);
		qdisc->gso_skb = NULL;
		qdisc->q.qlen = 0;
	}
L
Linus Torvalds 已提交
609
}
610
EXPORT_SYMBOL(qdisc_reset);
L
Linus Torvalds 已提交
611

E
Eric Dumazet 已提交
612 613 614 615 616 617 618
static void qdisc_rcu_free(struct rcu_head *head)
{
	struct Qdisc *qdisc = container_of(head, struct Qdisc, rcu_head);

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

619
void qdisc_destroy(struct Qdisc *qdisc)
L
Linus Torvalds 已提交
620
{
621 622
	const struct Qdisc_ops  *ops = qdisc->ops;

623 624 625 626
	if (qdisc->flags & TCQ_F_BUILTIN ||
	    !atomic_dec_and_test(&qdisc->refcnt))
		return;

627
#ifdef CONFIG_NET_SCHED
628 629
	qdisc_list_del(qdisc);

E
Eric Dumazet 已提交
630
	qdisc_put_stab(rtnl_dereference(qdisc->stab));
631
#endif
632 633 634 635 636 637 638 639 640
	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));

641
	kfree_skb(qdisc->gso_skb);
E
Eric Dumazet 已提交
642 643 644 645 646
	/*
	 * 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
647
}
648
EXPORT_SYMBOL(qdisc_destroy);
L
Linus Torvalds 已提交
649

650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
/* 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;
}
674
EXPORT_SYMBOL(dev_graft_qdisc);
675

676 677 678 679
static void attach_one_default_qdisc(struct net_device *dev,
				     struct netdev_queue *dev_queue,
				     void *_unused)
{
E
Eric Dumazet 已提交
680
	struct Qdisc *qdisc = &noqueue_qdisc;
681 682

	if (dev->tx_queue_len) {
683
		qdisc = qdisc_create_dflt(dev_queue,
684
					  &pfifo_fast_ops, TC_H_ROOT);
685
		if (!qdisc) {
E
Eric Dumazet 已提交
686
			netdev_info(dev, "activation failed\n");
687 688
			return;
		}
689 690
		if (!netif_is_multiqueue(dev))
			qdisc->flags |= TCQ_F_ONETXQUEUE;
691 692 693 694
	}
	dev_queue->qdisc_sleeping = qdisc;
}

695 696 697 698 699 700 701 702 703 704 705 706
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 {
707
		qdisc = qdisc_create_dflt(txq, &mq_qdisc_ops, TC_H_ROOT);
708 709 710 711 712 713 714
		if (qdisc) {
			qdisc->ops->attach(qdisc);
			dev->qdisc = qdisc;
		}
	}
}

715 716 717 718
static void transition_one_qdisc(struct net_device *dev,
				 struct netdev_queue *dev_queue,
				 void *_need_watchdog)
{
719
	struct Qdisc *new_qdisc = dev_queue->qdisc_sleeping;
720 721
	int *need_watchdog_p = _need_watchdog;

722 723 724
	if (!(new_qdisc->flags & TCQ_F_BUILTIN))
		clear_bit(__QDISC_STATE_DEACTIVATED, &new_qdisc->state);

725
	rcu_assign_pointer(dev_queue->qdisc, new_qdisc);
726 727
	if (need_watchdog_p && new_qdisc != &noqueue_qdisc) {
		dev_queue->trans_start = 0;
728
		*need_watchdog_p = 1;
729
	}
730 731
}

L
Linus Torvalds 已提交
732 733
void dev_activate(struct net_device *dev)
{
734
	int need_watchdog;
735

L
Linus Torvalds 已提交
736
	/* No queueing discipline is attached to device;
737 738 739
	   create default one i.e. pfifo_fast for devices,
	   which need queueing and noqueue_qdisc for
	   virtual interfaces
L
Linus Torvalds 已提交
740 741
	 */

742 743
	if (dev->qdisc == &noop_qdisc)
		attach_default_qdiscs(dev);
744

745 746 747 748
	if (!netif_carrier_ok(dev))
		/* Delay activation until next carrier-on event */
		return;

749 750
	need_watchdog = 0;
	netdev_for_each_tx_queue(dev, transition_one_qdisc, &need_watchdog);
751 752
	if (dev_ingress_queue(dev))
		transition_one_qdisc(dev, dev_ingress_queue(dev), NULL);
753 754

	if (need_watchdog) {
L
Linus Torvalds 已提交
755 756 757
		dev->trans_start = jiffies;
		dev_watchdog_up(dev);
	}
758
}
759
EXPORT_SYMBOL(dev_activate);
760

761 762 763
static void dev_deactivate_queue(struct net_device *dev,
				 struct netdev_queue *dev_queue,
				 void *_qdisc_default)
764
{
765
	struct Qdisc *qdisc_default = _qdisc_default;
766 767 768
	struct Qdisc *qdisc;

	qdisc = dev_queue->qdisc;
769
	if (qdisc) {
770 771
		spin_lock_bh(qdisc_lock(qdisc));

772 773 774
		if (!(qdisc->flags & TCQ_F_BUILTIN))
			set_bit(__QDISC_STATE_DEACTIVATED, &qdisc->state);

775
		rcu_assign_pointer(dev_queue->qdisc, qdisc_default);
776
		qdisc_reset(qdisc);
777

778
		spin_unlock_bh(qdisc_lock(qdisc));
779
	}
L
Linus Torvalds 已提交
780 781
}

782
static bool some_qdisc_is_busy(struct net_device *dev)
783 784 785 786 787
{
	unsigned int i;

	for (i = 0; i < dev->num_tx_queues; i++) {
		struct netdev_queue *dev_queue;
788
		spinlock_t *root_lock;
789
		struct Qdisc *q;
790 791 792
		int val;

		dev_queue = netdev_get_tx_queue(dev, i);
793
		q = dev_queue->qdisc_sleeping;
794
		root_lock = qdisc_lock(q);
795

796
		spin_lock_bh(root_lock);
797

798
		val = (qdisc_is_running(q) ||
799
		       test_bit(__QDISC_STATE_SCHED, &q->state));
800

801
		spin_unlock_bh(root_lock);
802 803 804 805 806 807 808

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

809 810 811 812 813 814 815
/**
 * 	dev_deactivate_many - deactivate transmissions on several devices
 * 	@head: list of devices to deactivate
 *
 *	This function returns only when all outstanding transmissions
 *	have completed, unless all devices are in dismantle phase.
 */
816
void dev_deactivate_many(struct list_head *head)
L
Linus Torvalds 已提交
817
{
818
	struct net_device *dev;
819
	bool sync_needed = false;
820

821 822 823 824 825 826 827 828
	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);
829
		sync_needed |= !dev->dismantle;
830
	}
L
Linus Torvalds 已提交
831

832 833 834 835 836 837
	/* Wait for outstanding qdisc-less dev_queue_xmit calls.
	 * This is avoided if all devices are in dismantle phase :
	 * Caller will call synchronize_net() for us
	 */
	if (sync_needed)
		synchronize_net();
L
Linus Torvalds 已提交
838

839
	/* Wait for outstanding qdisc_run calls. */
840 841 842 843 844 845 846 847 848 849 850
	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);
851
	list_del(&single);
L
Linus Torvalds 已提交
852
}
853
EXPORT_SYMBOL(dev_deactivate);
L
Linus Torvalds 已提交
854

855 856
static void dev_init_scheduler_queue(struct net_device *dev,
				     struct netdev_queue *dev_queue,
857
				     void *_qdisc)
858
{
859 860
	struct Qdisc *qdisc = _qdisc;

861 862 863 864
	dev_queue->qdisc = qdisc;
	dev_queue->qdisc_sleeping = qdisc;
}

L
Linus Torvalds 已提交
865 866
void dev_init_scheduler(struct net_device *dev)
{
867
	dev->qdisc = &noop_qdisc;
868
	netdev_for_each_tx_queue(dev, dev_init_scheduler_queue, &noop_qdisc);
869 870
	if (dev_ingress_queue(dev))
		dev_init_scheduler_queue(dev, dev_ingress_queue(dev), &noop_qdisc);
L
Linus Torvalds 已提交
871

872
	setup_timer(&dev->watchdog_timer, dev_watchdog, (unsigned long)dev);
L
Linus Torvalds 已提交
873 874
}

875 876 877
static void shutdown_scheduler_queue(struct net_device *dev,
				     struct netdev_queue *dev_queue,
				     void *_qdisc_default)
L
Linus Torvalds 已提交
878
{
879
	struct Qdisc *qdisc = dev_queue->qdisc_sleeping;
880
	struct Qdisc *qdisc_default = _qdisc_default;
881 882

	if (qdisc) {
883
		rcu_assign_pointer(dev_queue->qdisc, qdisc_default);
884
		dev_queue->qdisc_sleeping = qdisc_default;
L
Linus Torvalds 已提交
885 886

		qdisc_destroy(qdisc);
887
	}
888 889 890 891
}

void dev_shutdown(struct net_device *dev)
{
892
	netdev_for_each_tx_queue(dev, shutdown_scheduler_queue, &noop_qdisc);
893 894
	if (dev_ingress_queue(dev))
		shutdown_scheduler_queue(dev, dev_ingress_queue(dev), &noop_qdisc);
895 896 897
	qdisc_destroy(dev->qdisc);
	dev->qdisc = &noop_qdisc;

898
	WARN_ON(timer_pending(&dev->watchdog_timer));
L
Linus Torvalds 已提交
899
}
900 901 902 903 904 905 906

void psched_ratecfg_precompute(struct psched_ratecfg *r, u32 rate)
{
	u64 factor;
	u64 mult;
	int shift;

907
	r->rate_bps = (u64)rate << 3;
908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935
	r->shift = 0;
	r->mult = 1;
	/*
	 * Calibrate mult, shift so that token counting is accurate
	 * for smallest packet size (64 bytes).  Token (time in ns) is
	 * computed as (bytes * 8) * NSEC_PER_SEC / rate_bps.  It will
	 * work as long as the smallest packet transfer time can be
	 * accurately represented in nanosec.
	 */
	if (r->rate_bps > 0) {
		/*
		 * Higher shift gives better accuracy.  Find the largest
		 * shift such that mult fits in 32 bits.
		 */
		for (shift = 0; shift < 16; shift++) {
			r->shift = shift;
			factor = 8LLU * NSEC_PER_SEC * (1 << r->shift);
			mult = div64_u64(factor, r->rate_bps);
			if (mult > UINT_MAX)
				break;
		}

		r->shift = shift - 1;
		factor = 8LLU * NSEC_PER_SEC * (1 << r->shift);
		r->mult = div64_u64(factor, r->rate_bps);
	}
}
EXPORT_SYMBOL(psched_ratecfg_precompute);