sch_api.c 46.0 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 27 28
/*
 * net/sched/sch_api.c	Packet scheduler API.
 *
 *		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>
 *
 * Fixes:
 *
 * Rani Assaf <rani@magic.metawire.com> :980802: JIFFIES and CPU clock sources are repaired.
 * Eduardo J. Blanco <ejbs@netlabs.com.uy> :990222: kmod support
 * Jamal Hadi Salim <hadi@nortelnetworks.com>: 990601: ingress support
 */

#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <linux/init.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/kmod.h>
#include <linux/list.h>
29
#include <linux/hrtimer.h>
30
#include <linux/lockdep.h>
31
#include <linux/slab.h>
32
#include <linux/hashtable.h>
L
Linus Torvalds 已提交
33

34
#include <net/net_namespace.h>
35
#include <net/sock.h>
36
#include <net/netlink.h>
L
Linus Torvalds 已提交
37 38
#include <net/pkt_sched.h>

39 40
static int qdisc_notify(struct net *net, struct sk_buff *oskb,
			struct nlmsghdr *n, u32 clid,
L
Linus Torvalds 已提交
41
			struct Qdisc *old, struct Qdisc *new);
42 43 44
static int tclass_notify(struct net *net, struct sk_buff *oskb,
			 struct nlmsghdr *n, struct Qdisc *q,
			 unsigned long cl, int event);
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101

/*

   Short review.
   -------------

   This file consists of two interrelated parts:

   1. queueing disciplines manager frontend.
   2. traffic classes manager frontend.

   Generally, queueing discipline ("qdisc") is a black box,
   which is able to enqueue packets and to dequeue them (when
   device is ready to send something) in order and at times
   determined by algorithm hidden in it.

   qdisc's are divided to two categories:
   - "queues", which have no internal structure visible from outside.
   - "schedulers", which split all the packets to "traffic classes",
     using "packet classifiers" (look at cls_api.c)

   In turn, classes may have child qdiscs (as rule, queues)
   attached to them etc. etc. etc.

   The goal of the routines in this file is to translate
   information supplied by user in the form of handles
   to more intelligible for kernel form, to make some sanity
   checks and part of work, which is common to all qdiscs
   and to provide rtnetlink notifications.

   All real intelligent work is done inside qdisc modules.



   Every discipline has two major routines: enqueue and dequeue.

   ---dequeue

   dequeue usually returns a skb to send. It is allowed to return NULL,
   but it does not mean that queue is empty, it just means that
   discipline does not want to send anything this time.
   Queue is really empty if q->q.qlen == 0.
   For complicated disciplines with multiple queues q->q is not
   real packet queue, but however q->q.qlen must be valid.

   ---enqueue

   enqueue returns 0, if packet was enqueued successfully.
   If packet (this one or another one) was dropped, it returns
   not zero error code.
   NET_XMIT_DROP 	- this packet dropped
     Expected action: do not backoff, but wait until queue will clear.
   NET_XMIT_CN	 	- probably this packet enqueued, but another one dropped.
     Expected action: backoff or ignore

   Auxiliary routines:

102 103 104 105
   ---peek

   like dequeue but without removing a packet from the queue

L
Linus Torvalds 已提交
106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
   ---reset

   returns qdisc to initial state: purge all buffers, clear all
   timers, counters (except for statistics) etc.

   ---init

   initializes newly created qdisc.

   ---destroy

   destroys resources allocated by init and during lifetime of qdisc.

   ---change

   changes qdisc parameters.
 */

/* Protects list of registered TC modules. It is pure SMP lock. */
static DEFINE_RWLOCK(qdisc_mod_lock);


/************************************************
 *	Queueing disciplines manipulation.	*
 ************************************************/


/* The list of all installed queueing disciplines. */

static struct Qdisc_ops *qdisc_base;

137
/* Register/unregister queueing discipline */
L
Linus Torvalds 已提交
138 139 140 141 142 143 144 145 146 147 148 149 150

int register_qdisc(struct Qdisc_ops *qops)
{
	struct Qdisc_ops *q, **qp;
	int rc = -EEXIST;

	write_lock(&qdisc_mod_lock);
	for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next)
		if (!strcmp(qops->id, q->id))
			goto out;

	if (qops->enqueue == NULL)
		qops->enqueue = noop_qdisc_ops.enqueue;
151
	if (qops->peek == NULL) {
152
		if (qops->dequeue == NULL)
153
			qops->peek = noop_qdisc_ops.peek;
154 155
		else
			goto out_einval;
156
	}
L
Linus Torvalds 已提交
157 158 159
	if (qops->dequeue == NULL)
		qops->dequeue = noop_qdisc_ops.dequeue;

160 161 162
	if (qops->cl_ops) {
		const struct Qdisc_class_ops *cops = qops->cl_ops;

163
		if (!(cops->get && cops->put && cops->walk && cops->leaf))
164 165 166 167 168 169
			goto out_einval;

		if (cops->tcf_chain && !(cops->bind_tcf && cops->unbind_tcf))
			goto out_einval;
	}

L
Linus Torvalds 已提交
170 171 172 173 174 175
	qops->next = NULL;
	*qp = qops;
	rc = 0;
out:
	write_unlock(&qdisc_mod_lock);
	return rc;
176 177 178 179

out_einval:
	rc = -EINVAL;
	goto out;
L
Linus Torvalds 已提交
180
}
181
EXPORT_SYMBOL(register_qdisc);
L
Linus Torvalds 已提交
182 183 184 185 186 187 188

int unregister_qdisc(struct Qdisc_ops *qops)
{
	struct Qdisc_ops *q, **qp;
	int err = -ENOENT;

	write_lock(&qdisc_mod_lock);
E
Eric Dumazet 已提交
189
	for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next)
L
Linus Torvalds 已提交
190 191 192 193 194 195 196 197 198 199
		if (q == qops)
			break;
	if (q) {
		*qp = q->next;
		q->next = NULL;
		err = 0;
	}
	write_unlock(&qdisc_mod_lock);
	return err;
}
200
EXPORT_SYMBOL(unregister_qdisc);
L
Linus Torvalds 已提交
201

202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
/* Get default qdisc if not otherwise specified */
void qdisc_get_default(char *name, size_t len)
{
	read_lock(&qdisc_mod_lock);
	strlcpy(name, default_qdisc_ops->id, len);
	read_unlock(&qdisc_mod_lock);
}

static struct Qdisc_ops *qdisc_lookup_default(const char *name)
{
	struct Qdisc_ops *q = NULL;

	for (q = qdisc_base; q; q = q->next) {
		if (!strcmp(name, q->id)) {
			if (!try_module_get(q->owner))
				q = NULL;
			break;
		}
	}

	return q;
}

/* Set new default qdisc to use */
int qdisc_set_default(const char *name)
{
	const struct Qdisc_ops *ops;

	if (!capable(CAP_NET_ADMIN))
		return -EPERM;

	write_lock(&qdisc_mod_lock);
	ops = qdisc_lookup_default(name);
	if (!ops) {
		/* Not found, drop lock and try to load module */
		write_unlock(&qdisc_mod_lock);
		request_module("sch_%s", name);
		write_lock(&qdisc_mod_lock);

		ops = qdisc_lookup_default(name);
	}

	if (ops) {
		/* Set new default */
		module_put(default_qdisc_ops->owner);
		default_qdisc_ops = ops;
	}
	write_unlock(&qdisc_mod_lock);

	return ops ? 0 : -ENOENT;
}

L
Linus Torvalds 已提交
254
/* We know handle. Find qdisc among all qdisc's attached to device
255 256
 * (root qdisc, all its children, children of children etc.)
 * Note: caller either uses rtnl or rcu_read_lock()
L
Linus Torvalds 已提交
257 258
 */

259
static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle)
260 261 262
{
	struct Qdisc *q;

263 264 265
	if (!qdisc_dev(root))
		return (root->handle == handle ? root : NULL);

266 267 268 269
	if (!(root->flags & TCQ_F_BUILTIN) &&
	    root->handle == handle)
		return root;

270
	hash_for_each_possible_rcu(qdisc_dev(root)->qdisc_hash, q, hash, handle) {
271 272 273 274 275 276
		if (q->handle == handle)
			return q;
	}
	return NULL;
}

277
void qdisc_hash_add(struct Qdisc *q)
278
{
279 280
	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) {
		struct Qdisc *root = qdisc_dev(q)->qdisc;
281

282
		WARN_ON_ONCE(root == &noop_qdisc);
283
		ASSERT_RTNL();
284
		hash_add_rcu(qdisc_dev(q)->qdisc_hash, &q->hash, q->handle);
285
	}
286
}
287
EXPORT_SYMBOL(qdisc_hash_add);
288

289
void qdisc_hash_del(struct Qdisc *q)
290
{
291 292
	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) {
		ASSERT_RTNL();
293
		hash_del_rcu(&q->hash);
294
	}
295
}
296
EXPORT_SYMBOL(qdisc_hash_del);
297

298
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
L
Linus Torvalds 已提交
299
{
300 301
	struct Qdisc *q;

302 303 304
	q = qdisc_match_from_root(dev->qdisc, handle);
	if (q)
		goto out;
305

306 307 308 309
	if (dev_ingress_queue(dev))
		q = qdisc_match_from_root(
			dev_ingress_queue(dev)->qdisc_sleeping,
			handle);
310
out:
311
	return q;
L
Linus Torvalds 已提交
312 313 314 315 316 317
}

static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid)
{
	unsigned long cl;
	struct Qdisc *leaf;
318
	const struct Qdisc_class_ops *cops = p->ops->cl_ops;
L
Linus Torvalds 已提交
319 320 321 322 323 324 325 326 327 328 329 330 331 332

	if (cops == NULL)
		return NULL;
	cl = cops->get(p, classid);

	if (cl == 0)
		return NULL;
	leaf = cops->leaf(p, cl);
	cops->put(p, cl);
	return leaf;
}

/* Find queueing discipline by name */

333
static struct Qdisc_ops *qdisc_lookup_ops(struct nlattr *kind)
L
Linus Torvalds 已提交
334 335 336 337 338 339
{
	struct Qdisc_ops *q = NULL;

	if (kind) {
		read_lock(&qdisc_mod_lock);
		for (q = qdisc_base; q; q = q->next) {
340
			if (nla_strcmp(kind, q->id) == 0) {
L
Linus Torvalds 已提交
341 342 343 344 345 346 347 348 349 350
				if (!try_module_get(q->owner))
					q = NULL;
				break;
			}
		}
		read_unlock(&qdisc_mod_lock);
	}
	return q;
}

351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
/* The linklayer setting were not transferred from iproute2, in older
 * versions, and the rate tables lookup systems have been dropped in
 * the kernel. To keep backward compatible with older iproute2 tc
 * utils, we detect the linklayer setting by detecting if the rate
 * table were modified.
 *
 * For linklayer ATM table entries, the rate table will be aligned to
 * 48 bytes, thus some table entries will contain the same value.  The
 * mpu (min packet unit) is also encoded into the old rate table, thus
 * starting from the mpu, we find low and high table entries for
 * mapping this cell.  If these entries contain the same value, when
 * the rate tables have been modified for linklayer ATM.
 *
 * This is done by rounding mpu to the nearest 48 bytes cell/entry,
 * and then roundup to the next cell, calc the table entry one below,
 * and compare.
 */
static __u8 __detect_linklayer(struct tc_ratespec *r, __u32 *rtab)
{
	int low       = roundup(r->mpu, 48);
	int high      = roundup(low+1, 48);
	int cell_low  = low >> r->cell_log;
	int cell_high = (high >> r->cell_log) - 1;

	/* rtab is too inaccurate at rates > 100Mbit/s */
	if ((r->rate > (100000000/8)) || (rtab[0] == 0)) {
		pr_debug("TC linklayer: Giving up ATM detection\n");
		return TC_LINKLAYER_ETHERNET;
	}

	if ((cell_high > cell_low) && (cell_high < 256)
	    && (rtab[cell_low] == rtab[cell_high])) {
		pr_debug("TC linklayer: Detected ATM, low(%d)=high(%d)=%u\n",
			 cell_low, cell_high, rtab[cell_high]);
		return TC_LINKLAYER_ATM;
	}
	return TC_LINKLAYER_ETHERNET;
}

L
Linus Torvalds 已提交
390 391
static struct qdisc_rate_table *qdisc_rtab_list;

392 393
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r,
					struct nlattr *tab)
L
Linus Torvalds 已提交
394 395 396
{
	struct qdisc_rate_table *rtab;

397 398 399 400
	if (tab == NULL || r->rate == 0 || r->cell_log == 0 ||
	    nla_len(tab) != TC_RTAB_SIZE)
		return NULL;

L
Linus Torvalds 已提交
401
	for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) {
402 403
		if (!memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) &&
		    !memcmp(&rtab->data, nla_data(tab), 1024)) {
L
Linus Torvalds 已提交
404 405 406 407 408 409 410 411 412
			rtab->refcnt++;
			return rtab;
		}
	}

	rtab = kmalloc(sizeof(*rtab), GFP_KERNEL);
	if (rtab) {
		rtab->rate = *r;
		rtab->refcnt = 1;
413
		memcpy(rtab->data, nla_data(tab), 1024);
414 415
		if (r->linklayer == TC_LINKLAYER_UNAWARE)
			r->linklayer = __detect_linklayer(r, rtab->data);
L
Linus Torvalds 已提交
416 417 418 419 420
		rtab->next = qdisc_rtab_list;
		qdisc_rtab_list = rtab;
	}
	return rtab;
}
421
EXPORT_SYMBOL(qdisc_get_rtab);
L
Linus Torvalds 已提交
422 423 424 425 426 427 428 429

void qdisc_put_rtab(struct qdisc_rate_table *tab)
{
	struct qdisc_rate_table *rtab, **rtabp;

	if (!tab || --tab->refcnt)
		return;

E
Eric Dumazet 已提交
430 431 432
	for (rtabp = &qdisc_rtab_list;
	     (rtab = *rtabp) != NULL;
	     rtabp = &rtab->next) {
L
Linus Torvalds 已提交
433 434 435 436 437 438 439
		if (rtab == tab) {
			*rtabp = rtab->next;
			kfree(rtab);
			return;
		}
	}
}
440
EXPORT_SYMBOL(qdisc_put_rtab);
L
Linus Torvalds 已提交
441

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
static LIST_HEAD(qdisc_stab_list);
static DEFINE_SPINLOCK(qdisc_stab_lock);

static const struct nla_policy stab_policy[TCA_STAB_MAX + 1] = {
	[TCA_STAB_BASE]	= { .len = sizeof(struct tc_sizespec) },
	[TCA_STAB_DATA] = { .type = NLA_BINARY },
};

static struct qdisc_size_table *qdisc_get_stab(struct nlattr *opt)
{
	struct nlattr *tb[TCA_STAB_MAX + 1];
	struct qdisc_size_table *stab;
	struct tc_sizespec *s;
	unsigned int tsize = 0;
	u16 *tab = NULL;
	int err;

	err = nla_parse_nested(tb, TCA_STAB_MAX, opt, stab_policy);
	if (err < 0)
		return ERR_PTR(err);
	if (!tb[TCA_STAB_BASE])
		return ERR_PTR(-EINVAL);

	s = nla_data(tb[TCA_STAB_BASE]);

	if (s->tsize > 0) {
		if (!tb[TCA_STAB_DATA])
			return ERR_PTR(-EINVAL);
		tab = nla_data(tb[TCA_STAB_DATA]);
		tsize = nla_len(tb[TCA_STAB_DATA]) / sizeof(u16);
	}

474
	if (tsize != s->tsize || (!tab && tsize > 0))
475 476
		return ERR_PTR(-EINVAL);

477
	spin_lock(&qdisc_stab_lock);
478 479 480 481 482 483 484

	list_for_each_entry(stab, &qdisc_stab_list, list) {
		if (memcmp(&stab->szopts, s, sizeof(*s)))
			continue;
		if (tsize > 0 && memcmp(stab->data, tab, tsize * sizeof(u16)))
			continue;
		stab->refcnt++;
485
		spin_unlock(&qdisc_stab_lock);
486 487 488
		return stab;
	}

489
	spin_unlock(&qdisc_stab_lock);
490 491 492 493 494 495 496 497 498 499

	stab = kmalloc(sizeof(*stab) + tsize * sizeof(u16), GFP_KERNEL);
	if (!stab)
		return ERR_PTR(-ENOMEM);

	stab->refcnt = 1;
	stab->szopts = *s;
	if (tsize > 0)
		memcpy(stab->data, tab, tsize * sizeof(u16));

500
	spin_lock(&qdisc_stab_lock);
501
	list_add_tail(&stab->list, &qdisc_stab_list);
502
	spin_unlock(&qdisc_stab_lock);
503 504 505 506

	return stab;
}

507 508 509 510 511
static void stab_kfree_rcu(struct rcu_head *head)
{
	kfree(container_of(head, struct qdisc_size_table, rcu));
}

512 513 514 515 516
void qdisc_put_stab(struct qdisc_size_table *tab)
{
	if (!tab)
		return;

517
	spin_lock(&qdisc_stab_lock);
518 519 520

	if (--tab->refcnt == 0) {
		list_del(&tab->list);
521
		call_rcu_bh(&tab->rcu, stab_kfree_rcu);
522 523
	}

524
	spin_unlock(&qdisc_stab_lock);
525 526 527 528 529 530 531 532
}
EXPORT_SYMBOL(qdisc_put_stab);

static int qdisc_dump_stab(struct sk_buff *skb, struct qdisc_size_table *stab)
{
	struct nlattr *nest;

	nest = nla_nest_start(skb, TCA_STAB);
533 534
	if (nest == NULL)
		goto nla_put_failure;
535 536
	if (nla_put(skb, TCA_STAB_BASE, sizeof(stab->szopts), &stab->szopts))
		goto nla_put_failure;
537 538 539 540 541 542 543 544
	nla_nest_end(skb, nest);

	return skb->len;

nla_put_failure:
	return -1;
}

545 546
void __qdisc_calculate_pkt_len(struct sk_buff *skb,
			       const struct qdisc_size_table *stab)
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571
{
	int pkt_len, slot;

	pkt_len = skb->len + stab->szopts.overhead;
	if (unlikely(!stab->szopts.tsize))
		goto out;

	slot = pkt_len + stab->szopts.cell_align;
	if (unlikely(slot < 0))
		slot = 0;

	slot >>= stab->szopts.cell_log;
	if (likely(slot < stab->szopts.tsize))
		pkt_len = stab->data[slot];
	else
		pkt_len = stab->data[stab->szopts.tsize - 1] *
				(slot / stab->szopts.tsize) +
				stab->data[slot % stab->szopts.tsize];

	pkt_len <<= stab->szopts.size_log;
out:
	if (unlikely(pkt_len < 1))
		pkt_len = 1;
	qdisc_skb_cb(skb)->pkt_len = pkt_len;
}
572
EXPORT_SYMBOL(__qdisc_calculate_pkt_len);
573

574
void qdisc_warn_nonwc(const char *txt, struct Qdisc *qdisc)
575 576
{
	if (!(qdisc->flags & TCQ_F_WARN_NONWC)) {
E
Eric Dumazet 已提交
577 578
		pr_warn("%s: %s qdisc %X: is non-work-conserving?\n",
			txt, qdisc->ops->id, qdisc->handle >> 16);
579 580 581 582 583
		qdisc->flags |= TCQ_F_WARN_NONWC;
	}
}
EXPORT_SYMBOL(qdisc_warn_nonwc);

584 585 586
static enum hrtimer_restart qdisc_watchdog(struct hrtimer *timer)
{
	struct qdisc_watchdog *wd = container_of(timer, struct qdisc_watchdog,
587
						 timer);
588

589
	rcu_read_lock();
590
	__netif_schedule(qdisc_root(wd->qdisc));
591
	rcu_read_unlock();
592

593 594 595 596 597
	return HRTIMER_NORESTART;
}

void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc)
{
598
	hrtimer_init(&wd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_PINNED);
599
	wd->timer.function = qdisc_watchdog;
600 601 602 603
	wd->qdisc = qdisc;
}
EXPORT_SYMBOL(qdisc_watchdog_init);

604
void qdisc_watchdog_schedule_ns(struct qdisc_watchdog *wd, u64 expires)
605
{
606 607 608 609
	if (test_bit(__QDISC_STATE_DEACTIVATED,
		     &qdisc_root_sleeping(wd->qdisc)->state))
		return;

610 611 612 613
	if (wd->last_expires == expires)
		return;

	wd->last_expires = expires;
614
	hrtimer_start(&wd->timer,
615
		      ns_to_ktime(expires),
616
		      HRTIMER_MODE_ABS_PINNED);
617
}
618
EXPORT_SYMBOL(qdisc_watchdog_schedule_ns);
619 620 621

void qdisc_watchdog_cancel(struct qdisc_watchdog *wd)
{
622
	hrtimer_cancel(&wd->timer);
623 624
}
EXPORT_SYMBOL(qdisc_watchdog_cancel);
L
Linus Torvalds 已提交
625

626
static struct hlist_head *qdisc_class_hash_alloc(unsigned int n)
627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656
{
	unsigned int size = n * sizeof(struct hlist_head), i;
	struct hlist_head *h;

	if (size <= PAGE_SIZE)
		h = kmalloc(size, GFP_KERNEL);
	else
		h = (struct hlist_head *)
			__get_free_pages(GFP_KERNEL, get_order(size));

	if (h != NULL) {
		for (i = 0; i < n; i++)
			INIT_HLIST_HEAD(&h[i]);
	}
	return h;
}

static void qdisc_class_hash_free(struct hlist_head *h, unsigned int n)
{
	unsigned int size = n * sizeof(struct hlist_head);

	if (size <= PAGE_SIZE)
		kfree(h);
	else
		free_pages((unsigned long)h, get_order(size));
}

void qdisc_class_hash_grow(struct Qdisc *sch, struct Qdisc_class_hash *clhash)
{
	struct Qdisc_class_common *cl;
657
	struct hlist_node *next;
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
	struct hlist_head *nhash, *ohash;
	unsigned int nsize, nmask, osize;
	unsigned int i, h;

	/* Rehash when load factor exceeds 0.75 */
	if (clhash->hashelems * 4 <= clhash->hashsize * 3)
		return;
	nsize = clhash->hashsize * 2;
	nmask = nsize - 1;
	nhash = qdisc_class_hash_alloc(nsize);
	if (nhash == NULL)
		return;

	ohash = clhash->hash;
	osize = clhash->hashsize;

	sch_tree_lock(sch);
	for (i = 0; i < osize; i++) {
676
		hlist_for_each_entry_safe(cl, next, &ohash[i], hnode) {
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729
			h = qdisc_class_hash(cl->classid, nmask);
			hlist_add_head(&cl->hnode, &nhash[h]);
		}
	}
	clhash->hash     = nhash;
	clhash->hashsize = nsize;
	clhash->hashmask = nmask;
	sch_tree_unlock(sch);

	qdisc_class_hash_free(ohash, osize);
}
EXPORT_SYMBOL(qdisc_class_hash_grow);

int qdisc_class_hash_init(struct Qdisc_class_hash *clhash)
{
	unsigned int size = 4;

	clhash->hash = qdisc_class_hash_alloc(size);
	if (clhash->hash == NULL)
		return -ENOMEM;
	clhash->hashsize  = size;
	clhash->hashmask  = size - 1;
	clhash->hashelems = 0;
	return 0;
}
EXPORT_SYMBOL(qdisc_class_hash_init);

void qdisc_class_hash_destroy(struct Qdisc_class_hash *clhash)
{
	qdisc_class_hash_free(clhash->hash, clhash->hashsize);
}
EXPORT_SYMBOL(qdisc_class_hash_destroy);

void qdisc_class_hash_insert(struct Qdisc_class_hash *clhash,
			     struct Qdisc_class_common *cl)
{
	unsigned int h;

	INIT_HLIST_NODE(&cl->hnode);
	h = qdisc_class_hash(cl->classid, clhash->hashmask);
	hlist_add_head(&cl->hnode, &clhash->hash[h]);
	clhash->hashelems++;
}
EXPORT_SYMBOL(qdisc_class_hash_insert);

void qdisc_class_hash_remove(struct Qdisc_class_hash *clhash,
			     struct Qdisc_class_common *cl)
{
	hlist_del(&cl->hnode);
	clhash->hashelems--;
}
EXPORT_SYMBOL(qdisc_class_hash_remove);

730 731 732
/* Allocate an unique handle from space managed by kernel
 * Possible range is [8000-FFFF]:0000 (0x8000 values)
 */
L
Linus Torvalds 已提交
733 734
static u32 qdisc_alloc_handle(struct net_device *dev)
{
735
	int i = 0x8000;
L
Linus Torvalds 已提交
736 737 738 739 740 741
	static u32 autohandle = TC_H_MAKE(0x80000000U, 0);

	do {
		autohandle += TC_H_MAKE(0x10000U, 0);
		if (autohandle == TC_H_MAKE(TC_H_ROOT, 0))
			autohandle = TC_H_MAKE(0x80000000U, 0);
742 743 744 745
		if (!qdisc_lookup(dev, autohandle))
			return autohandle;
		cond_resched();
	} while	(--i > 0);
L
Linus Torvalds 已提交
746

747
	return 0;
L
Linus Torvalds 已提交
748 749
}

750 751
void qdisc_tree_reduce_backlog(struct Qdisc *sch, unsigned int n,
			       unsigned int len)
752
{
753
	const struct Qdisc_class_ops *cops;
754 755
	unsigned long cl;
	u32 parentid;
756
	int drops;
757

758
	if (n == 0 && len == 0)
759
		return;
760
	drops = max_t(int, n, 0);
761
	rcu_read_lock();
762
	while ((parentid = sch->parent)) {
763
		if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS))
764
			break;
765

766 767 768
		if (sch->flags & TCQ_F_NOPARENT)
			break;
		/* TODO: perform the search on a per txq basis */
769
		sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid));
770
		if (sch == NULL) {
771 772
			WARN_ON_ONCE(parentid != TC_H_ROOT);
			break;
773
		}
774 775 776 777 778 779 780
		cops = sch->ops->cl_ops;
		if (cops->qlen_notify) {
			cl = cops->get(sch, parentid);
			cops->qlen_notify(sch, cl);
			cops->put(sch, cl);
		}
		sch->q.qlen -= n;
781
		sch->qstats.backlog -= len;
782
		__qdisc_qstats_drop(sch, drops);
783
	}
784
	rcu_read_unlock();
785
}
786
EXPORT_SYMBOL(qdisc_tree_reduce_backlog);
L
Linus Torvalds 已提交
787

788 789
static void notify_and_destroy(struct net *net, struct sk_buff *skb,
			       struct nlmsghdr *n, u32 clid,
790 791 792
			       struct Qdisc *old, struct Qdisc *new)
{
	if (new || old)
793
		qdisc_notify(net, skb, n, clid, old, new);
L
Linus Torvalds 已提交
794

795
	if (old)
796 797 798 799 800 801 802 803 804 805
		qdisc_destroy(old);
}

/* Graft qdisc "new" to class "classid" of qdisc "parent" or
 * to device "dev".
 *
 * When appropriate send a netlink notification using 'skb'
 * and "n".
 *
 * On success, destroy old qdisc.
L
Linus Torvalds 已提交
806 807 808
 */

static int qdisc_graft(struct net_device *dev, struct Qdisc *parent,
809 810
		       struct sk_buff *skb, struct nlmsghdr *n, u32 classid,
		       struct Qdisc *new, struct Qdisc *old)
L
Linus Torvalds 已提交
811
{
812
	struct Qdisc *q = old;
813
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
814 815
	int err = 0;

816
	if (parent == NULL) {
817 818 819 820
		unsigned int i, num_q, ingress;

		ingress = 0;
		num_q = dev->num_tx_queues;
821 822
		if ((q && q->flags & TCQ_F_INGRESS) ||
		    (new && new->flags & TCQ_F_INGRESS)) {
823 824
			num_q = 1;
			ingress = 1;
825 826
			if (!dev_ingress_queue(dev))
				return -ENOENT;
827 828 829 830 831
		}

		if (dev->flags & IFF_UP)
			dev_deactivate(dev);

832 833
		if (new && new->ops->attach)
			goto skip;
834

835
		for (i = 0; i < num_q; i++) {
836
			struct netdev_queue *dev_queue = dev_ingress_queue(dev);
837 838 839 840

			if (!ingress)
				dev_queue = netdev_get_tx_queue(dev, i);

841 842 843 844
			old = dev_graft_qdisc(dev_queue, new);
			if (new && i > 0)
				atomic_inc(&new->refcnt);

845 846
			if (!ingress)
				qdisc_destroy(old);
L
Linus Torvalds 已提交
847
		}
848

849
skip:
850
		if (!ingress) {
851 852
			notify_and_destroy(net, skb, n, classid,
					   dev->qdisc, new);
853 854 855
			if (new && !new->ops->attach)
				atomic_inc(&new->refcnt);
			dev->qdisc = new ? : &noop_qdisc;
856 857 858

			if (new && new->ops->attach)
				new->ops->attach(new);
859
		} else {
860
			notify_and_destroy(net, skb, n, classid, old, new);
861
		}
862

863 864
		if (dev->flags & IFF_UP)
			dev_activate(dev);
L
Linus Torvalds 已提交
865
	} else {
866
		const struct Qdisc_class_ops *cops = parent->ops->cl_ops;
L
Linus Torvalds 已提交
867

868 869
		err = -EOPNOTSUPP;
		if (cops && cops->graft) {
L
Linus Torvalds 已提交
870 871
			unsigned long cl = cops->get(parent, classid);
			if (cl) {
872
				err = cops->graft(parent, cl, new, &old);
L
Linus Torvalds 已提交
873
				cops->put(parent, cl);
874 875
			} else
				err = -ENOENT;
L
Linus Torvalds 已提交
876
		}
877
		if (!err)
878
			notify_and_destroy(net, skb, n, classid, old, new);
L
Linus Torvalds 已提交
879 880 881 882
	}
	return err;
}

883 884 885 886
/* lockdep annotation is needed for ingress; egress gets it only for name */
static struct lock_class_key qdisc_tx_lock;
static struct lock_class_key qdisc_rx_lock;

L
Linus Torvalds 已提交
887 888 889 890 891 892
/*
   Allocate and initialize new qdisc.

   Parameters are passed via opt.
 */

893 894 895 896
static struct Qdisc *qdisc_create(struct net_device *dev,
				  struct netdev_queue *dev_queue,
				  struct Qdisc *p, u32 parent, u32 handle,
				  struct nlattr **tca, int *errp)
L
Linus Torvalds 已提交
897 898
{
	int err;
899
	struct nlattr *kind = tca[TCA_KIND];
L
Linus Torvalds 已提交
900 901
	struct Qdisc *sch;
	struct Qdisc_ops *ops;
902
	struct qdisc_size_table *stab;
L
Linus Torvalds 已提交
903 904

	ops = qdisc_lookup_ops(kind);
905
#ifdef CONFIG_MODULES
L
Linus Torvalds 已提交
906 907
	if (ops == NULL && kind != NULL) {
		char name[IFNAMSIZ];
908
		if (nla_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) {
L
Linus Torvalds 已提交
909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932
			/* We dropped the RTNL semaphore in order to
			 * perform the module load.  So, even if we
			 * succeeded in loading the module we have to
			 * tell the caller to replay the request.  We
			 * indicate this using -EAGAIN.
			 * We replay the request because the device may
			 * go away in the mean time.
			 */
			rtnl_unlock();
			request_module("sch_%s", name);
			rtnl_lock();
			ops = qdisc_lookup_ops(kind);
			if (ops != NULL) {
				/* We will try again qdisc_lookup_ops,
				 * so don't keep a reference.
				 */
				module_put(ops->owner);
				err = -EAGAIN;
				goto err_out;
			}
		}
	}
#endif

933
	err = -ENOENT;
L
Linus Torvalds 已提交
934 935 936
	if (ops == NULL)
		goto err_out;

937
	sch = qdisc_alloc(dev_queue, ops);
938 939
	if (IS_ERR(sch)) {
		err = PTR_ERR(sch);
L
Linus Torvalds 已提交
940
		goto err_out2;
941
	}
L
Linus Torvalds 已提交
942

943 944
	sch->parent = parent;

945
	if (handle == TC_H_INGRESS) {
L
Linus Torvalds 已提交
946
		sch->flags |= TCQ_F_INGRESS;
947
		handle = TC_H_MAKE(TC_H_INGRESS, 0);
948
		lockdep_set_class(qdisc_lock(sch), &qdisc_rx_lock);
949 950 951 952 953 954 955
	} else {
		if (handle == 0) {
			handle = qdisc_alloc_handle(dev);
			err = -ENOMEM;
			if (handle == 0)
				goto err_out3;
		}
956
		lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock);
957
		if (!netif_is_multiqueue(dev))
958
			sch->flags |= TCQ_F_ONETXQUEUE;
L
Linus Torvalds 已提交
959 960
	}

961
	sch->handle = handle;
L
Linus Torvalds 已提交
962

963 964 965 966 967 968 969 970 971 972 973
	/* This exist to keep backward compatible with a userspace
	 * loophole, what allowed userspace to get IFF_NO_QUEUE
	 * facility on older kernels by setting tx_queue_len=0 (prior
	 * to qdisc init), and then forgot to reinit tx_queue_len
	 * before again attaching a qdisc.
	 */
	if ((dev->priv_flags & IFF_NO_QUEUE) && (dev->tx_queue_len == 0)) {
		dev->tx_queue_len = DEFAULT_TX_QUEUE_LEN;
		netdev_info(dev, "Caught tx_queue_len zero misconfig\n");
	}

974
	if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS])) == 0) {
975 976
		if (qdisc_is_percpu_stats(sch)) {
			sch->cpu_bstats =
977
				netdev_alloc_pcpu_stats(struct gnet_stats_basic_cpu);
978 979
			if (!sch->cpu_bstats)
				goto err_out4;
980 981 982 983

			sch->cpu_qstats = alloc_percpu(struct gnet_stats_queue);
			if (!sch->cpu_qstats)
				goto err_out4;
984 985
		}

986 987 988 989
		if (tca[TCA_STAB]) {
			stab = qdisc_get_stab(tca[TCA_STAB]);
			if (IS_ERR(stab)) {
				err = PTR_ERR(stab);
990
				goto err_out4;
991
			}
992
			rcu_assign_pointer(sch->stab, stab);
993
		}
994
		if (tca[TCA_RATE]) {
995
			seqcount_t *running;
996

997 998 999 1000
			err = -EOPNOTSUPP;
			if (sch->flags & TCQ_F_MQROOT)
				goto err_out4;

1001
			if ((sch->parent != TC_H_ROOT) &&
1002 1003
			    !(sch->flags & TCQ_F_INGRESS) &&
			    (!p || !(p->flags & TCQ_F_MQROOT)))
1004
				running = qdisc_root_sleeping_running(sch);
1005
			else
1006
				running = &sch->running;
1007

1008 1009 1010
			err = gen_new_estimator(&sch->bstats,
						sch->cpu_bstats,
						&sch->rate_est,
1011 1012
						NULL,
						running,
1013
						tca[TCA_RATE]);
1014 1015
			if (err)
				goto err_out4;
1016
		}
1017

1018
		qdisc_hash_add(sch);
L
Linus Torvalds 已提交
1019 1020 1021

		return sch;
	}
1022 1023
	/* ops->init() failed, we call ->destroy() like qdisc_create_dflt() */
	ops->destroy(sch);
L
Linus Torvalds 已提交
1024 1025
err_out3:
	dev_put(dev);
1026
	kfree((char *) sch - sch->padded);
L
Linus Torvalds 已提交
1027 1028 1029 1030 1031
err_out2:
	module_put(ops->owner);
err_out:
	*errp = err;
	return NULL;
1032 1033

err_out4:
1034
	free_percpu(sch->cpu_bstats);
1035
	free_percpu(sch->cpu_qstats);
1036 1037 1038 1039
	/*
	 * Any broken qdiscs that would require a ops->reset() here?
	 * The qdisc was never in action so it shouldn't be necessary.
	 */
1040
	qdisc_put_stab(rtnl_dereference(sch->stab));
1041 1042 1043
	if (ops->destroy)
		ops->destroy(sch);
	goto err_out3;
L
Linus Torvalds 已提交
1044 1045
}

1046
static int qdisc_change(struct Qdisc *sch, struct nlattr **tca)
L
Linus Torvalds 已提交
1047
{
1048
	struct qdisc_size_table *ostab, *stab = NULL;
1049
	int err = 0;
L
Linus Torvalds 已提交
1050

1051
	if (tca[TCA_OPTIONS]) {
L
Linus Torvalds 已提交
1052 1053
		if (sch->ops->change == NULL)
			return -EINVAL;
1054
		err = sch->ops->change(sch, tca[TCA_OPTIONS]);
L
Linus Torvalds 已提交
1055 1056 1057
		if (err)
			return err;
	}
1058 1059 1060 1061 1062 1063 1064

	if (tca[TCA_STAB]) {
		stab = qdisc_get_stab(tca[TCA_STAB]);
		if (IS_ERR(stab))
			return PTR_ERR(stab);
	}

1065 1066 1067
	ostab = rtnl_dereference(sch->stab);
	rcu_assign_pointer(sch->stab, stab);
	qdisc_put_stab(ostab);
1068

1069
	if (tca[TCA_RATE]) {
1070 1071
		/* NB: ignores errors from replace_estimator
		   because change can't be undone. */
1072 1073
		if (sch->flags & TCQ_F_MQROOT)
			goto out;
1074 1075 1076
		gen_replace_estimator(&sch->bstats,
				      sch->cpu_bstats,
				      &sch->rate_est,
1077 1078
				      NULL,
				      qdisc_root_sleeping_running(sch),
1079
				      tca[TCA_RATE]);
1080 1081
	}
out:
L
Linus Torvalds 已提交
1082 1083 1084
	return 0;
}

E
Eric Dumazet 已提交
1085 1086
struct check_loop_arg {
	struct qdisc_walker	w;
L
Linus Torvalds 已提交
1087 1088 1089 1090
	struct Qdisc		*p;
	int			depth;
};

1091 1092
static int check_loop_fn(struct Qdisc *q, unsigned long cl,
			 struct qdisc_walker *w);
L
Linus Torvalds 已提交
1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112

static int check_loop(struct Qdisc *q, struct Qdisc *p, int depth)
{
	struct check_loop_arg	arg;

	if (q->ops->cl_ops == NULL)
		return 0;

	arg.w.stop = arg.w.skip = arg.w.count = 0;
	arg.w.fn = check_loop_fn;
	arg.depth = depth;
	arg.p = p;
	q->ops->cl_ops->walk(q, &arg.w);
	return arg.w.stop ? -ELOOP : 0;
}

static int
check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w)
{
	struct Qdisc *leaf;
1113
	const struct Qdisc_class_ops *cops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
	struct check_loop_arg *arg = (struct check_loop_arg *)w;

	leaf = cops->leaf(q, cl);
	if (leaf) {
		if (leaf == arg->p || arg->depth > 7)
			return -ELOOP;
		return check_loop(leaf, arg->p, arg->depth + 1);
	}
	return 0;
}

/*
 * Delete/get qdisc.
 */

1129
static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1130
{
1131
	struct net *net = sock_net(skb->sk);
1132
	struct tcmsg *tcm = nlmsg_data(n);
1133
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1134
	struct net_device *dev;
1135
	u32 clid;
L
Linus Torvalds 已提交
1136 1137 1138 1139
	struct Qdisc *q = NULL;
	struct Qdisc *p = NULL;
	int err;

1140
	if ((n->nlmsg_type != RTM_GETQDISC) &&
1141
	    !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1142 1143
		return -EPERM;

1144 1145 1146 1147
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1148 1149 1150 1151 1152
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

	clid = tcm->tcm_parent;
L
Linus Torvalds 已提交
1153 1154 1155
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) {
E
Eric Dumazet 已提交
1156 1157
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1158 1159
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1160 1161
			} else if (dev_ingress_queue(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
1162
			}
L
Linus Torvalds 已提交
1163
		} else {
1164
			q = dev->qdisc;
L
Linus Torvalds 已提交
1165 1166 1167 1168 1169 1170 1171
		}
		if (!q)
			return -ENOENT;

		if (tcm->tcm_handle && q->handle != tcm->tcm_handle)
			return -EINVAL;
	} else {
E
Eric Dumazet 已提交
1172 1173
		q = qdisc_lookup(dev, tcm->tcm_handle);
		if (!q)
L
Linus Torvalds 已提交
1174 1175 1176
			return -ENOENT;
	}

1177
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1178 1179 1180 1181 1182 1183 1184
		return -EINVAL;

	if (n->nlmsg_type == RTM_DELQDISC) {
		if (!clid)
			return -EINVAL;
		if (q->handle == 0)
			return -ENOENT;
E
Eric Dumazet 已提交
1185 1186
		err = qdisc_graft(dev, p, skb, n, clid, NULL, q);
		if (err != 0)
L
Linus Torvalds 已提交
1187 1188
			return err;
	} else {
1189
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194
	}
	return 0;
}

/*
E
Eric Dumazet 已提交
1195
 * Create/change qdisc.
L
Linus Torvalds 已提交
1196 1197
 */

1198
static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1199
{
1200
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1201
	struct tcmsg *tcm;
1202
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1203 1204 1205 1206 1207
	struct net_device *dev;
	u32 clid;
	struct Qdisc *q, *p;
	int err;

1208
	if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1209 1210
		return -EPERM;

L
Linus Torvalds 已提交
1211 1212
replay:
	/* Reinit, just in case something touches this. */
1213 1214 1215 1216
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1217
	tcm = nlmsg_data(n);
L
Linus Torvalds 已提交
1218 1219 1220
	clid = tcm->tcm_parent;
	q = p = NULL;

E
Eric Dumazet 已提交
1221 1222
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1223 1224
		return -ENODEV;

1225

L
Linus Torvalds 已提交
1226 1227 1228
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (clid != TC_H_INGRESS) {
E
Eric Dumazet 已提交
1229 1230
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1231 1232
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1233 1234
			} else if (dev_ingress_queue_create(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
L
Linus Torvalds 已提交
1235 1236
			}
		} else {
1237
			q = dev->qdisc;
L
Linus Torvalds 已提交
1238 1239 1240 1241 1242 1243 1244 1245
		}

		/* It may be default qdisc, ignore it */
		if (q && q->handle == 0)
			q = NULL;

		if (!q || !tcm->tcm_handle || q->handle != tcm->tcm_handle) {
			if (tcm->tcm_handle) {
E
Eric Dumazet 已提交
1246
				if (q && !(n->nlmsg_flags & NLM_F_REPLACE))
L
Linus Torvalds 已提交
1247 1248 1249
					return -EEXIST;
				if (TC_H_MIN(tcm->tcm_handle))
					return -EINVAL;
E
Eric Dumazet 已提交
1250 1251
				q = qdisc_lookup(dev, tcm->tcm_handle);
				if (!q)
L
Linus Torvalds 已提交
1252
					goto create_n_graft;
E
Eric Dumazet 已提交
1253
				if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1254
					return -EEXIST;
1255
				if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1256 1257 1258 1259 1260 1261 1262
					return -EINVAL;
				if (q == p ||
				    (p && check_loop(q, p, 0)))
					return -ELOOP;
				atomic_inc(&q->refcnt);
				goto graft;
			} else {
E
Eric Dumazet 已提交
1263
				if (!q)
L
Linus Torvalds 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
					goto create_n_graft;

				/* This magic test requires explanation.
				 *
				 *   We know, that some child q is already
				 *   attached to this parent and have choice:
				 *   either to change it or to create/graft new one.
				 *
				 *   1. We are allowed to create/graft only
				 *   if CREATE and REPLACE flags are set.
				 *
				 *   2. If EXCL is set, requestor wanted to say,
				 *   that qdisc tcm_handle is not expected
				 *   to exist, so that we choose create/graft too.
				 *
				 *   3. The last case is when no flags are set.
				 *   Alas, it is sort of hole in API, we
				 *   cannot decide what to do unambiguously.
				 *   For now we select create/graft, if
				 *   user gave KIND, which does not match existing.
				 */
E
Eric Dumazet 已提交
1285 1286 1287
				if ((n->nlmsg_flags & NLM_F_CREATE) &&
				    (n->nlmsg_flags & NLM_F_REPLACE) &&
				    ((n->nlmsg_flags & NLM_F_EXCL) ||
1288 1289
				     (tca[TCA_KIND] &&
				      nla_strcmp(tca[TCA_KIND], q->ops->id))))
L
Linus Torvalds 已提交
1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
					goto create_n_graft;
			}
		}
	} else {
		if (!tcm->tcm_handle)
			return -EINVAL;
		q = qdisc_lookup(dev, tcm->tcm_handle);
	}

	/* Change qdisc parameters */
	if (q == NULL)
		return -ENOENT;
E
Eric Dumazet 已提交
1302
	if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1303
		return -EEXIST;
1304
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1305 1306 1307
		return -EINVAL;
	err = qdisc_change(q, tca);
	if (err == 0)
1308
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1309 1310 1311
	return err;

create_n_graft:
E
Eric Dumazet 已提交
1312
	if (!(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1313
		return -ENOENT;
1314 1315 1316 1317 1318 1319 1320 1321
	if (clid == TC_H_INGRESS) {
		if (dev_ingress_queue(dev))
			q = qdisc_create(dev, dev_ingress_queue(dev), p,
					 tcm->tcm_parent, tcm->tcm_parent,
					 tca, &err);
		else
			err = -ENOENT;
	} else {
1322
		struct netdev_queue *dev_queue;
1323 1324

		if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue)
1325 1326 1327 1328 1329
			dev_queue = p->ops->cl_ops->select_queue(p, tcm);
		else if (p)
			dev_queue = p->dev_queue;
		else
			dev_queue = netdev_get_tx_queue(dev, 0);
1330

1331
		q = qdisc_create(dev, dev_queue, p,
1332
				 tcm->tcm_parent, tcm->tcm_handle,
1333
				 tca, &err);
1334
	}
L
Linus Torvalds 已提交
1335 1336 1337 1338 1339 1340 1341
	if (q == NULL) {
		if (err == -EAGAIN)
			goto replay;
		return err;
	}

graft:
1342 1343 1344 1345 1346
	err = qdisc_graft(dev, p, skb, n, clid, q, NULL);
	if (err) {
		if (q)
			qdisc_destroy(q);
		return err;
L
Linus Torvalds 已提交
1347
	}
1348

L
Linus Torvalds 已提交
1349 1350 1351 1352
	return 0;
}

static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid,
1353
			 u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1354
{
1355
	struct gnet_stats_basic_cpu __percpu *cpu_bstats = NULL;
1356
	struct gnet_stats_queue __percpu *cpu_qstats = NULL;
L
Linus Torvalds 已提交
1357 1358
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1359
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1360
	struct gnet_dump d;
1361
	struct qdisc_size_table *stab;
1362
	__u32 qlen;
L
Linus Torvalds 已提交
1363

1364
	cond_resched();
1365
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1366 1367 1368
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1369
	tcm->tcm_family = AF_UNSPEC;
1370 1371
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1372
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1373 1374 1375
	tcm->tcm_parent = clid;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = atomic_read(&q->refcnt);
1376 1377
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1378
	if (q->ops->dump && q->ops->dump(q, skb) < 0)
1379
		goto nla_put_failure;
1380
	qlen = q->q.qlen;
L
Linus Torvalds 已提交
1381

1382 1383
	stab = rtnl_dereference(q->stab);
	if (stab && qdisc_dump_stab(skb, stab) < 0)
1384 1385
		goto nla_put_failure;

1386
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
1387
					 NULL, &d, TCA_PAD) < 0)
1388
		goto nla_put_failure;
L
Linus Torvalds 已提交
1389 1390

	if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0)
1391
		goto nla_put_failure;
L
Linus Torvalds 已提交
1392

1393
	if (qdisc_is_percpu_stats(q)) {
1394
		cpu_bstats = q->cpu_bstats;
1395 1396
		cpu_qstats = q->cpu_qstats;
	}
1397

1398 1399
	if (gnet_stats_copy_basic(qdisc_root_sleeping_running(q),
				  &d, cpu_bstats, &q->bstats) < 0 ||
1400
	    gnet_stats_copy_rate_est(&d, &q->rate_est) < 0 ||
1401
	    gnet_stats_copy_queue(&d, cpu_qstats, &q->qstats, qlen) < 0)
1402
		goto nla_put_failure;
1403

L
Linus Torvalds 已提交
1404
	if (gnet_stats_finish_copy(&d) < 0)
1405
		goto nla_put_failure;
1406

1407
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1408 1409
	return skb->len;

1410
out_nlmsg_trim:
1411
nla_put_failure:
1412
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1413 1414 1415
	return -1;
}

1416 1417 1418 1419 1420
static bool tc_qdisc_dump_ignore(struct Qdisc *q)
{
	return (q->flags & TCQ_F_BUILTIN) ? true : false;
}

1421 1422 1423
static int qdisc_notify(struct net *net, struct sk_buff *oskb,
			struct nlmsghdr *n, u32 clid,
			struct Qdisc *old, struct Qdisc *new)
L
Linus Torvalds 已提交
1424 1425
{
	struct sk_buff *skb;
1426
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1427 1428 1429 1430 1431

	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
	if (!skb)
		return -ENOBUFS;

1432
	if (old && !tc_qdisc_dump_ignore(old)) {
1433
		if (tc_fill_qdisc(skb, old, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1434
				  0, RTM_DELQDISC) < 0)
L
Linus Torvalds 已提交
1435 1436
			goto err_out;
	}
1437
	if (new && !tc_qdisc_dump_ignore(new)) {
1438
		if (tc_fill_qdisc(skb, new, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1439
				  old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0)
L
Linus Torvalds 已提交
1440 1441 1442 1443
			goto err_out;
	}

	if (skb->len)
1444
		return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1445
				      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1446 1447 1448 1449 1450 1451

err_out:
	kfree_skb(skb);
	return -EINVAL;
}

1452 1453
static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb,
			      struct netlink_callback *cb,
1454
			      int *q_idx_p, int s_q_idx, bool recur)
1455 1456 1457
{
	int ret = 0, q_idx = *q_idx_p;
	struct Qdisc *q;
1458
	int b;
1459 1460 1461 1462 1463 1464 1465 1466 1467

	if (!root)
		return 0;

	q = root;
	if (q_idx < s_q_idx) {
		q_idx++;
	} else {
		if (!tc_qdisc_dump_ignore(q) &&
1468
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1469 1470
				  cb->nlh->nlmsg_seq, NLM_F_MULTI,
				  RTM_NEWQDISC) <= 0)
1471 1472 1473
			goto done;
		q_idx++;
	}
1474

1475 1476 1477 1478 1479 1480 1481
	/* If dumping singletons, there is no qdisc_dev(root) and the singleton
	 * itself has already been dumped.
	 *
	 * If we've already dumped the top-level (ingress) qdisc above and the global
	 * qdisc hashtable, we don't want to hit it again
	 */
	if (!qdisc_dev(root) || !recur)
1482 1483
		goto out;

1484
	hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
1485 1486 1487 1488
		if (q_idx < s_q_idx) {
			q_idx++;
			continue;
		}
E
Eric Dumazet 已提交
1489
		if (!tc_qdisc_dump_ignore(q) &&
1490
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1491 1492
				  cb->nlh->nlmsg_seq, NLM_F_MULTI,
				  RTM_NEWQDISC) <= 0)
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
			goto done;
		q_idx++;
	}

out:
	*q_idx_p = q_idx;
	return ret;
done:
	ret = -1;
	goto out;
}

L
Linus Torvalds 已提交
1505 1506
static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb)
{
1507
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1508 1509 1510 1511 1512 1513
	int idx, q_idx;
	int s_idx, s_q_idx;
	struct net_device *dev;

	s_idx = cb->args[0];
	s_q_idx = q_idx = cb->args[1];
1514

1515
	idx = 0;
1516 1517
	ASSERT_RTNL();
	for_each_netdev(net, dev) {
1518 1519
		struct netdev_queue *dev_queue;

L
Linus Torvalds 已提交
1520
		if (idx < s_idx)
1521
			goto cont;
L
Linus Torvalds 已提交
1522 1523 1524
		if (idx > s_idx)
			s_q_idx = 0;
		q_idx = 0;
1525

1526 1527
		if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx,
				       true) < 0)
1528 1529
			goto done;

1530 1531 1532
		dev_queue = dev_ingress_queue(dev);
		if (dev_queue &&
		    tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb,
1533
				       &q_idx, s_q_idx, false) < 0)
1534 1535
			goto done;

1536 1537
cont:
		idx++;
L
Linus Torvalds 已提交
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554
	}

done:
	cb->args[0] = idx;
	cb->args[1] = q_idx;

	return skb->len;
}



/************************************************
 *	Traffic classes manipulation.		*
 ************************************************/



1555
static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1556
{
1557
	struct net *net = sock_net(skb->sk);
1558
	struct tcmsg *tcm = nlmsg_data(n);
1559
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1560 1561
	struct net_device *dev;
	struct Qdisc *q = NULL;
1562
	const struct Qdisc_class_ops *cops;
L
Linus Torvalds 已提交
1563 1564
	unsigned long cl = 0;
	unsigned long new_cl;
1565 1566 1567
	u32 portid;
	u32 clid;
	u32 qid;
L
Linus Torvalds 已提交
1568 1569
	int err;

1570
	if ((n->nlmsg_type != RTM_GETTCLASS) &&
1571
	    !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1572 1573
		return -EPERM;

1574 1575 1576 1577
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1578 1579 1580 1581
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

L
Linus Torvalds 已提交
1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
	/*
	   parent == TC_H_UNSPEC - unspecified parent.
	   parent == TC_H_ROOT   - class is root, which has no parent.
	   parent == X:0	 - parent is root class.
	   parent == X:Y	 - parent is a node in hierarchy.
	   parent == 0:Y	 - parent is X:Y, where X:0 is qdisc.

	   handle == 0:0	 - generate handle from kernel pool.
	   handle == 0:Y	 - class is X:Y, where X:0 is qdisc.
	   handle == X:Y	 - clear.
	   handle == X:0	 - root class.
	 */

	/* Step 1. Determine qdisc handle X:0 */

1597 1598 1599 1600
	portid = tcm->tcm_parent;
	clid = tcm->tcm_handle;
	qid = TC_H_MAJ(clid);

1601 1602
	if (portid != TC_H_ROOT) {
		u32 qid1 = TC_H_MAJ(portid);
L
Linus Torvalds 已提交
1603 1604 1605 1606 1607 1608 1609 1610

		if (qid && qid1) {
			/* If both majors are known, they must be identical. */
			if (qid != qid1)
				return -EINVAL;
		} else if (qid1) {
			qid = qid1;
		} else if (qid == 0)
1611
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1612 1613

		/* Now qid is genuine qdisc handle consistent
E
Eric Dumazet 已提交
1614 1615
		 * both with parent and child.
		 *
1616
		 * TC_H_MAJ(portid) still may be unspecified, complete it now.
L
Linus Torvalds 已提交
1617
		 */
1618 1619
		if (portid)
			portid = TC_H_MAKE(qid, portid);
L
Linus Torvalds 已提交
1620 1621
	} else {
		if (qid == 0)
1622
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1623 1624 1625
	}

	/* OK. Locate qdisc */
E
Eric Dumazet 已提交
1626 1627
	q = qdisc_lookup(dev, qid);
	if (!q)
L
Linus Torvalds 已提交
1628 1629 1630 1631 1632 1633 1634 1635 1636
		return -ENOENT;

	/* An check that it supports classes */
	cops = q->ops->cl_ops;
	if (cops == NULL)
		return -EINVAL;

	/* Now try to get class */
	if (clid == 0) {
1637
		if (portid == TC_H_ROOT)
L
Linus Torvalds 已提交
1638 1639 1640 1641 1642 1643 1644 1645 1646
			clid = qid;
	} else
		clid = TC_H_MAKE(qid, clid);

	if (clid)
		cl = cops->get(q, clid);

	if (cl == 0) {
		err = -ENOENT;
E
Eric Dumazet 已提交
1647 1648
		if (n->nlmsg_type != RTM_NEWTCLASS ||
		    !(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1649 1650 1651
			goto out;
	} else {
		switch (n->nlmsg_type) {
1652
		case RTM_NEWTCLASS:
L
Linus Torvalds 已提交
1653
			err = -EEXIST;
E
Eric Dumazet 已提交
1654
			if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1655 1656 1657
				goto out;
			break;
		case RTM_DELTCLASS:
1658 1659 1660
			err = -EOPNOTSUPP;
			if (cops->delete)
				err = cops->delete(q, cl);
L
Linus Torvalds 已提交
1661
			if (err == 0)
1662 1663
				tclass_notify(net, skb, n, q, cl,
					      RTM_DELTCLASS);
L
Linus Torvalds 已提交
1664 1665
			goto out;
		case RTM_GETTCLASS:
1666
			err = tclass_notify(net, skb, n, q, cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1667 1668 1669 1670 1671 1672 1673 1674
			goto out;
		default:
			err = -EINVAL;
			goto out;
		}
	}

	new_cl = cl;
1675 1676
	err = -EOPNOTSUPP;
	if (cops->change)
1677
		err = cops->change(q, clid, portid, tca, &new_cl);
L
Linus Torvalds 已提交
1678
	if (err == 0)
1679
		tclass_notify(net, skb, n, q, new_cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690

out:
	if (cl)
		cops->put(q, cl);

	return err;
}


static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q,
			  unsigned long cl,
1691
			  u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1692 1693 1694
{
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1695
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1696
	struct gnet_dump d;
1697
	const struct Qdisc_class_ops *cl_ops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1698

1699
	cond_resched();
1700
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1701 1702 1703
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1704
	tcm->tcm_family = AF_UNSPEC;
1705 1706
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1707
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1708 1709 1710
	tcm->tcm_parent = q->handle;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = 0;
1711 1712
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1713
	if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0)
1714
		goto nla_put_failure;
L
Linus Torvalds 已提交
1715

1716
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
1717
					 NULL, &d, TCA_PAD) < 0)
1718
		goto nla_put_failure;
L
Linus Torvalds 已提交
1719 1720

	if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0)
1721
		goto nla_put_failure;
L
Linus Torvalds 已提交
1722 1723

	if (gnet_stats_finish_copy(&d) < 0)
1724
		goto nla_put_failure;
L
Linus Torvalds 已提交
1725

1726
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1727 1728
	return skb->len;

1729
out_nlmsg_trim:
1730
nla_put_failure:
1731
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1732 1733 1734
	return -1;
}

1735 1736 1737
static int tclass_notify(struct net *net, struct sk_buff *oskb,
			 struct nlmsghdr *n, struct Qdisc *q,
			 unsigned long cl, int event)
L
Linus Torvalds 已提交
1738 1739
{
	struct sk_buff *skb;
1740
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1741 1742 1743 1744 1745

	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
	if (!skb)
		return -ENOBUFS;

1746
	if (tc_fill_tclass(skb, q, cl, portid, n->nlmsg_seq, 0, event) < 0) {
L
Linus Torvalds 已提交
1747 1748 1749 1750
		kfree_skb(skb);
		return -EINVAL;
	}

1751
	return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1752
			      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1753 1754
}

E
Eric Dumazet 已提交
1755 1756 1757 1758
struct qdisc_dump_args {
	struct qdisc_walker	w;
	struct sk_buff		*skb;
	struct netlink_callback	*cb;
L
Linus Torvalds 已提交
1759 1760
};

1761 1762
static int qdisc_class_dump(struct Qdisc *q, unsigned long cl,
			    struct qdisc_walker *arg)
L
Linus Torvalds 已提交
1763 1764 1765
{
	struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg;

1766
	return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).portid,
1767 1768
			      a->cb->nlh->nlmsg_seq, NLM_F_MULTI,
			      RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1769 1770
}

1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
static int tc_dump_tclass_qdisc(struct Qdisc *q, struct sk_buff *skb,
				struct tcmsg *tcm, struct netlink_callback *cb,
				int *t_p, int s_t)
{
	struct qdisc_dump_args arg;

	if (tc_qdisc_dump_ignore(q) ||
	    *t_p < s_t || !q->ops->cl_ops ||
	    (tcm->tcm_parent &&
	     TC_H_MAJ(tcm->tcm_parent) != q->handle)) {
		(*t_p)++;
		return 0;
	}
	if (*t_p > s_t)
		memset(&cb->args[1], 0, sizeof(cb->args)-sizeof(cb->args[0]));
	arg.w.fn = qdisc_class_dump;
	arg.skb = skb;
	arg.cb = cb;
	arg.w.stop  = 0;
	arg.w.skip = cb->args[1];
	arg.w.count = 0;
	q->ops->cl_ops->walk(q, &arg.w);
	cb->args[1] = arg.w.count;
	if (arg.w.stop)
		return -1;
	(*t_p)++;
	return 0;
}

static int tc_dump_tclass_root(struct Qdisc *root, struct sk_buff *skb,
			       struct tcmsg *tcm, struct netlink_callback *cb,
			       int *t_p, int s_t)
{
	struct Qdisc *q;
1805
	int b;
1806 1807 1808 1809 1810 1811 1812

	if (!root)
		return 0;

	if (tc_dump_tclass_qdisc(root, skb, tcm, cb, t_p, s_t) < 0)
		return -1;

1813 1814 1815
	if (!qdisc_dev(root))
		return 0;

1816
	hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
1817 1818 1819 1820 1821 1822 1823
		if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0)
			return -1;
	}

	return 0;
}

L
Linus Torvalds 已提交
1824 1825
static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb)
{
1826
	struct tcmsg *tcm = nlmsg_data(cb->nlh);
1827
	struct net *net = sock_net(skb->sk);
1828
	struct netdev_queue *dev_queue;
L
Linus Torvalds 已提交
1829
	struct net_device *dev;
1830
	int t, s_t;
L
Linus Torvalds 已提交
1831

1832
	if (nlmsg_len(cb->nlh) < sizeof(*tcm))
L
Linus Torvalds 已提交
1833
		return 0;
E
Eric Dumazet 已提交
1834 1835
	dev = dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840
		return 0;

	s_t = cb->args[0];
	t = 0;

1841
	if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0)
1842 1843
		goto done;

1844 1845 1846 1847
	dev_queue = dev_ingress_queue(dev);
	if (dev_queue &&
	    tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb,
				&t, s_t) < 0)
1848
		goto done;
L
Linus Torvalds 已提交
1849

1850
done:
L
Linus Torvalds 已提交
1851 1852 1853 1854 1855 1856 1857
	cb->args[0] = t;

	dev_put(dev);
	return skb->len;
}

/* Main classifier routine: scans classifier chain attached
E
Eric Dumazet 已提交
1858 1859
 * to this qdisc, (optionally) tests for protocol and asks
 * specific classifiers.
L
Linus Torvalds 已提交
1860
 */
1861 1862
int tc_classify(struct sk_buff *skb, const struct tcf_proto *tp,
		struct tcf_result *res, bool compat_mode)
1863
{
1864
	__be16 protocol = tc_skb_protocol(skb);
1865
#ifdef CONFIG_NET_CLS_ACT
1866
	const int max_reclassify_loop = 4;
1867 1868
	const struct tcf_proto *old_tp = tp;
	int limit = 0;
1869

1870 1871
reclassify:
#endif
J
John Fastabend 已提交
1872
	for (; tp; tp = rcu_dereference_bh(tp->next)) {
1873 1874
		int err;

E
Eric Dumazet 已提交
1875 1876 1877 1878
		if (tp->protocol != protocol &&
		    tp->protocol != htons(ETH_P_ALL))
			continue;

1879 1880
		err = tp->classify(skb, tp, res);
#ifdef CONFIG_NET_CLS_ACT
1881
		if (unlikely(err == TC_ACT_RECLASSIFY && !compat_mode))
1882 1883
			goto reset;
#endif
1884
		if (err >= 0)
1885 1886 1887
			return err;
	}

1888
	return TC_ACT_UNSPEC; /* signal: continue lookup */
L
Linus Torvalds 已提交
1889
#ifdef CONFIG_NET_CLS_ACT
1890
reset:
1891
	if (unlikely(limit++ >= max_reclassify_loop)) {
1892 1893 1894
		net_notice_ratelimited("%s: reclassify loop, rule prio %u, protocol %02x\n",
				       tp->q->ops->id, tp->prio & 0xffff,
				       ntohs(tp->protocol));
1895
		return TC_ACT_SHOT;
L
Linus Torvalds 已提交
1896
	}
1897 1898

	tp = old_tp;
1899
	protocol = tc_skb_protocol(skb);
1900
	goto reclassify;
1901
#endif
L
Linus Torvalds 已提交
1902
}
1903
EXPORT_SYMBOL(tc_classify);
L
Linus Torvalds 已提交
1904 1905 1906 1907 1908

#ifdef CONFIG_PROC_FS
static int psched_show(struct seq_file *seq, void *v)
{
	seq_printf(seq, "%08x %08x %08x %08x\n",
1909
		   (u32)NSEC_PER_USEC, (u32)PSCHED_TICKS2NS(1),
1910
		   1000000,
1911
		   (u32)NSEC_PER_SEC / hrtimer_resolution);
L
Linus Torvalds 已提交
1912 1913 1914 1915 1916 1917

	return 0;
}

static int psched_open(struct inode *inode, struct file *file)
{
1918
	return single_open(file, psched_show, NULL);
L
Linus Torvalds 已提交
1919 1920
}

1921
static const struct file_operations psched_fops = {
L
Linus Torvalds 已提交
1922 1923 1924 1925 1926
	.owner = THIS_MODULE,
	.open = psched_open,
	.read  = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
1927
};
1928 1929 1930 1931 1932

static int __net_init psched_net_init(struct net *net)
{
	struct proc_dir_entry *e;

1933
	e = proc_create("psched", 0, net->proc_net, &psched_fops);
1934 1935 1936 1937 1938 1939 1940 1941
	if (e == NULL)
		return -ENOMEM;

	return 0;
}

static void __net_exit psched_net_exit(struct net *net)
{
1942
	remove_proc_entry("psched", net->proc_net);
1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
}
#else
static int __net_init psched_net_init(struct net *net)
{
	return 0;
}

static void __net_exit psched_net_exit(struct net *net)
{
}
L
Linus Torvalds 已提交
1953 1954
#endif

1955 1956 1957 1958 1959
static struct pernet_operations psched_net_ops = {
	.init = psched_net_init,
	.exit = psched_net_exit,
};

L
Linus Torvalds 已提交
1960 1961
static int __init pktsched_init(void)
{
1962 1963 1964 1965
	int err;

	err = register_pernet_subsys(&psched_net_ops);
	if (err) {
E
Eric Dumazet 已提交
1966
		pr_err("pktsched_init: "
1967 1968 1969 1970
		       "cannot initialize per netns operations\n");
		return err;
	}

1971
	register_qdisc(&pfifo_fast_ops);
L
Linus Torvalds 已提交
1972 1973
	register_qdisc(&pfifo_qdisc_ops);
	register_qdisc(&bfifo_qdisc_ops);
1974
	register_qdisc(&pfifo_head_drop_qdisc_ops);
1975
	register_qdisc(&mq_qdisc_ops);
1976
	register_qdisc(&noqueue_qdisc_ops);
L
Linus Torvalds 已提交
1977

1978 1979
	rtnl_register(PF_UNSPEC, RTM_NEWQDISC, tc_modify_qdisc, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELQDISC, tc_get_qdisc, NULL, NULL);
1980 1981
	rtnl_register(PF_UNSPEC, RTM_GETQDISC, tc_get_qdisc, tc_dump_qdisc,
		      NULL);
1982 1983
	rtnl_register(PF_UNSPEC, RTM_NEWTCLASS, tc_ctl_tclass, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELTCLASS, tc_ctl_tclass, NULL, NULL);
1984 1985
	rtnl_register(PF_UNSPEC, RTM_GETTCLASS, tc_ctl_tclass, tc_dump_tclass,
		      NULL);
1986

L
Linus Torvalds 已提交
1987 1988 1989 1990
	return 0;
}

subsys_initcall(pktsched_init);
反馈
建议
客服 返回
顶部