sch_api.c 46.3 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, bool invisible)
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
		if (invisible)
			q->flags |= TCQ_F_INVISIBLE;
287
	}
288
}
289
EXPORT_SYMBOL(qdisc_hash_add);
290

291
void qdisc_hash_del(struct Qdisc *q)
292
{
293 294
	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS)) {
		ASSERT_RTNL();
295
		hash_del_rcu(&q->hash);
296
	}
297
}
298
EXPORT_SYMBOL(qdisc_hash_del);
299

300
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
L
Linus Torvalds 已提交
301
{
302 303
	struct Qdisc *q;

304 305 306
	q = qdisc_match_from_root(dev->qdisc, handle);
	if (q)
		goto out;
307

308 309 310 311
	if (dev_ingress_queue(dev))
		q = qdisc_match_from_root(
			dev_ingress_queue(dev)->qdisc_sleeping,
			handle);
312
out:
313
	return q;
L
Linus Torvalds 已提交
314 315 316 317 318 319
}

static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid)
{
	unsigned long cl;
	struct Qdisc *leaf;
320
	const struct Qdisc_class_ops *cops = p->ops->cl_ops;
L
Linus Torvalds 已提交
321 322 323 324 325 326 327 328 329 330 331 332 333 334

	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 */

335
static struct Qdisc_ops *qdisc_lookup_ops(struct nlattr *kind)
L
Linus Torvalds 已提交
336 337 338 339 340 341
{
	struct Qdisc_ops *q = NULL;

	if (kind) {
		read_lock(&qdisc_mod_lock);
		for (q = qdisc_base; q; q = q->next) {
342
			if (nla_strcmp(kind, q->id) == 0) {
L
Linus Torvalds 已提交
343 344 345 346 347 348 349 350 351 352
				if (!try_module_get(q->owner))
					q = NULL;
				break;
			}
		}
		read_unlock(&qdisc_mod_lock);
	}
	return q;
}

353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391
/* 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
392 393
static struct qdisc_rate_table *qdisc_rtab_list;

394 395
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r,
					struct nlattr *tab)
L
Linus Torvalds 已提交
396 397 398
{
	struct qdisc_rate_table *rtab;

399 400 401 402
	if (tab == NULL || r->rate == 0 || r->cell_log == 0 ||
	    nla_len(tab) != TC_RTAB_SIZE)
		return NULL;

L
Linus Torvalds 已提交
403
	for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) {
404 405
		if (!memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) &&
		    !memcmp(&rtab->data, nla_data(tab), 1024)) {
L
Linus Torvalds 已提交
406 407 408 409 410 411 412 413 414
			rtab->refcnt++;
			return rtab;
		}
	}

	rtab = kmalloc(sizeof(*rtab), GFP_KERNEL);
	if (rtab) {
		rtab->rate = *r;
		rtab->refcnt = 1;
415
		memcpy(rtab->data, nla_data(tab), 1024);
416 417
		if (r->linklayer == TC_LINKLAYER_UNAWARE)
			r->linklayer = __detect_linklayer(r, rtab->data);
L
Linus Torvalds 已提交
418 419 420 421 422
		rtab->next = qdisc_rtab_list;
		qdisc_rtab_list = rtab;
	}
	return rtab;
}
423
EXPORT_SYMBOL(qdisc_get_rtab);
L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431

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
432 433 434
	for (rtabp = &qdisc_rtab_list;
	     (rtab = *rtabp) != NULL;
	     rtabp = &rtab->next) {
L
Linus Torvalds 已提交
435 436 437 438 439 440 441
		if (rtab == tab) {
			*rtabp = rtab->next;
			kfree(rtab);
			return;
		}
	}
}
442
EXPORT_SYMBOL(qdisc_put_rtab);
L
Linus Torvalds 已提交
443

444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
static LIST_HEAD(qdisc_stab_list);

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;

460
	err = nla_parse_nested(tb, TCA_STAB_MAX, opt, stab_policy, NULL);
461 462 463 464 465 466 467 468 469 470 471 472 473 474
	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);
	}

475
	if (tsize != s->tsize || (!tab && tsize > 0))
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
		return ERR_PTR(-EINVAL);

	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++;
		return stab;
	}

	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));

	list_add_tail(&stab->list, &qdisc_stab_list);

	return stab;
}

501 502 503 504 505
static void stab_kfree_rcu(struct rcu_head *head)
{
	kfree(container_of(head, struct qdisc_size_table, rcu));
}

506 507 508 509 510 511 512
void qdisc_put_stab(struct qdisc_size_table *tab)
{
	if (!tab)
		return;

	if (--tab->refcnt == 0) {
		list_del(&tab->list);
513
		call_rcu_bh(&tab->rcu, stab_kfree_rcu);
514 515 516 517 518 519 520 521 522
	}
}
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);
523 524
	if (nest == NULL)
		goto nla_put_failure;
525 526
	if (nla_put(skb, TCA_STAB_BASE, sizeof(stab->szopts), &stab->szopts))
		goto nla_put_failure;
527 528 529 530 531 532 533 534
	nla_nest_end(skb, nest);

	return skb->len;

nla_put_failure:
	return -1;
}

535 536
void __qdisc_calculate_pkt_len(struct sk_buff *skb,
			       const struct qdisc_size_table *stab)
537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
{
	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;
}
562
EXPORT_SYMBOL(__qdisc_calculate_pkt_len);
563

564
void qdisc_warn_nonwc(const char *txt, struct Qdisc *qdisc)
565 566
{
	if (!(qdisc->flags & TCQ_F_WARN_NONWC)) {
E
Eric Dumazet 已提交
567 568
		pr_warn("%s: %s qdisc %X: is non-work-conserving?\n",
			txt, qdisc->ops->id, qdisc->handle >> 16);
569 570 571 572 573
		qdisc->flags |= TCQ_F_WARN_NONWC;
	}
}
EXPORT_SYMBOL(qdisc_warn_nonwc);

574 575 576
static enum hrtimer_restart qdisc_watchdog(struct hrtimer *timer)
{
	struct qdisc_watchdog *wd = container_of(timer, struct qdisc_watchdog,
577
						 timer);
578

579
	rcu_read_lock();
580
	__netif_schedule(qdisc_root(wd->qdisc));
581
	rcu_read_unlock();
582

583 584 585 586 587
	return HRTIMER_NORESTART;
}

void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc)
{
588
	hrtimer_init(&wd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_PINNED);
589
	wd->timer.function = qdisc_watchdog;
590 591 592 593
	wd->qdisc = qdisc;
}
EXPORT_SYMBOL(qdisc_watchdog_init);

594
void qdisc_watchdog_schedule_ns(struct qdisc_watchdog *wd, u64 expires)
595
{
596 597 598 599
	if (test_bit(__QDISC_STATE_DEACTIVATED,
		     &qdisc_root_sleeping(wd->qdisc)->state))
		return;

600 601 602 603
	if (wd->last_expires == expires)
		return;

	wd->last_expires = expires;
604
	hrtimer_start(&wd->timer,
605
		      ns_to_ktime(expires),
606
		      HRTIMER_MODE_ABS_PINNED);
607
}
608
EXPORT_SYMBOL(qdisc_watchdog_schedule_ns);
609 610 611

void qdisc_watchdog_cancel(struct qdisc_watchdog *wd)
{
612
	hrtimer_cancel(&wd->timer);
613 614
}
EXPORT_SYMBOL(qdisc_watchdog_cancel);
L
Linus Torvalds 已提交
615

616
static struct hlist_head *qdisc_class_hash_alloc(unsigned int n)
617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646
{
	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;
647
	struct hlist_node *next;
648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665
	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++) {
666
		hlist_for_each_entry_safe(cl, next, &ohash[i], hnode) {
667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719
			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);

720 721 722
/* Allocate an unique handle from space managed by kernel
 * Possible range is [8000-FFFF]:0000 (0x8000 values)
 */
L
Linus Torvalds 已提交
723 724
static u32 qdisc_alloc_handle(struct net_device *dev)
{
725
	int i = 0x8000;
L
Linus Torvalds 已提交
726 727 728 729 730 731
	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);
732 733 734 735
		if (!qdisc_lookup(dev, autohandle))
			return autohandle;
		cond_resched();
	} while	(--i > 0);
L
Linus Torvalds 已提交
736

737
	return 0;
L
Linus Torvalds 已提交
738 739
}

740 741
void qdisc_tree_reduce_backlog(struct Qdisc *sch, unsigned int n,
			       unsigned int len)
742
{
743
	const struct Qdisc_class_ops *cops;
744 745
	unsigned long cl;
	u32 parentid;
746
	int drops;
747

748
	if (n == 0 && len == 0)
749
		return;
750
	drops = max_t(int, n, 0);
751
	rcu_read_lock();
752
	while ((parentid = sch->parent)) {
753
		if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS))
754
			break;
755

756 757 758
		if (sch->flags & TCQ_F_NOPARENT)
			break;
		/* TODO: perform the search on a per txq basis */
759
		sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid));
760
		if (sch == NULL) {
761 762
			WARN_ON_ONCE(parentid != TC_H_ROOT);
			break;
763
		}
764 765 766 767 768 769 770
		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;
771
		sch->qstats.backlog -= len;
772
		__qdisc_qstats_drop(sch, drops);
773
	}
774
	rcu_read_unlock();
775
}
776
EXPORT_SYMBOL(qdisc_tree_reduce_backlog);
L
Linus Torvalds 已提交
777

778 779
static void notify_and_destroy(struct net *net, struct sk_buff *skb,
			       struct nlmsghdr *n, u32 clid,
780 781 782
			       struct Qdisc *old, struct Qdisc *new)
{
	if (new || old)
783
		qdisc_notify(net, skb, n, clid, old, new);
L
Linus Torvalds 已提交
784

785
	if (old)
786 787 788 789 790 791 792 793 794 795
		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
796 797 798
 */

static int qdisc_graft(struct net_device *dev, struct Qdisc *parent,
799 800
		       struct sk_buff *skb, struct nlmsghdr *n, u32 classid,
		       struct Qdisc *new, struct Qdisc *old)
L
Linus Torvalds 已提交
801
{
802
	struct Qdisc *q = old;
803
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
804 805
	int err = 0;

806
	if (parent == NULL) {
807 808 809 810
		unsigned int i, num_q, ingress;

		ingress = 0;
		num_q = dev->num_tx_queues;
811 812
		if ((q && q->flags & TCQ_F_INGRESS) ||
		    (new && new->flags & TCQ_F_INGRESS)) {
813 814
			num_q = 1;
			ingress = 1;
815 816
			if (!dev_ingress_queue(dev))
				return -ENOENT;
817 818 819 820 821
		}

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

822 823
		if (new && new->ops->attach)
			goto skip;
824

825
		for (i = 0; i < num_q; i++) {
826
			struct netdev_queue *dev_queue = dev_ingress_queue(dev);
827 828 829 830

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

831 832 833 834
			old = dev_graft_qdisc(dev_queue, new);
			if (new && i > 0)
				atomic_inc(&new->refcnt);

835 836
			if (!ingress)
				qdisc_destroy(old);
L
Linus Torvalds 已提交
837
		}
838

839
skip:
840
		if (!ingress) {
841 842
			notify_and_destroy(net, skb, n, classid,
					   dev->qdisc, new);
843 844 845
			if (new && !new->ops->attach)
				atomic_inc(&new->refcnt);
			dev->qdisc = new ? : &noop_qdisc;
846 847 848

			if (new && new->ops->attach)
				new->ops->attach(new);
849
		} else {
850
			notify_and_destroy(net, skb, n, classid, old, new);
851
		}
852

853 854
		if (dev->flags & IFF_UP)
			dev_activate(dev);
L
Linus Torvalds 已提交
855
	} else {
856
		const struct Qdisc_class_ops *cops = parent->ops->cl_ops;
L
Linus Torvalds 已提交
857

858 859
		err = -EOPNOTSUPP;
		if (cops && cops->graft) {
L
Linus Torvalds 已提交
860 861
			unsigned long cl = cops->get(parent, classid);
			if (cl) {
862
				err = cops->graft(parent, cl, new, &old);
L
Linus Torvalds 已提交
863
				cops->put(parent, cl);
864 865
			} else
				err = -ENOENT;
L
Linus Torvalds 已提交
866
		}
867
		if (!err)
868
			notify_and_destroy(net, skb, n, classid, old, new);
L
Linus Torvalds 已提交
869 870 871 872
	}
	return err;
}

873 874 875 876
/* 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
877 878 879 880 881 882
/*
   Allocate and initialize new qdisc.

   Parameters are passed via opt.
 */

883 884 885 886
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
887 888
{
	int err;
889
	struct nlattr *kind = tca[TCA_KIND];
L
Linus Torvalds 已提交
890 891
	struct Qdisc *sch;
	struct Qdisc_ops *ops;
892
	struct qdisc_size_table *stab;
L
Linus Torvalds 已提交
893 894

	ops = qdisc_lookup_ops(kind);
895
#ifdef CONFIG_MODULES
L
Linus Torvalds 已提交
896 897
	if (ops == NULL && kind != NULL) {
		char name[IFNAMSIZ];
898
		if (nla_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) {
L
Linus Torvalds 已提交
899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922
			/* 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

923
	err = -ENOENT;
L
Linus Torvalds 已提交
924 925 926
	if (ops == NULL)
		goto err_out;

927
	sch = qdisc_alloc(dev_queue, ops);
928 929
	if (IS_ERR(sch)) {
		err = PTR_ERR(sch);
L
Linus Torvalds 已提交
930
		goto err_out2;
931
	}
L
Linus Torvalds 已提交
932

933 934
	sch->parent = parent;

935
	if (handle == TC_H_INGRESS) {
L
Linus Torvalds 已提交
936
		sch->flags |= TCQ_F_INGRESS;
937
		handle = TC_H_MAKE(TC_H_INGRESS, 0);
938
		lockdep_set_class(qdisc_lock(sch), &qdisc_rx_lock);
939 940 941 942 943 944 945
	} else {
		if (handle == 0) {
			handle = qdisc_alloc_handle(dev);
			err = -ENOMEM;
			if (handle == 0)
				goto err_out3;
		}
946
		lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock);
947
		if (!netif_is_multiqueue(dev))
948
			sch->flags |= TCQ_F_ONETXQUEUE;
L
Linus Torvalds 已提交
949 950
	}

951
	sch->handle = handle;
L
Linus Torvalds 已提交
952

953 954 955 956 957 958 959 960 961 962 963
	/* 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");
	}

964
	if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS])) == 0) {
965 966
		if (qdisc_is_percpu_stats(sch)) {
			sch->cpu_bstats =
967
				netdev_alloc_pcpu_stats(struct gnet_stats_basic_cpu);
968 969
			if (!sch->cpu_bstats)
				goto err_out4;
970 971 972 973

			sch->cpu_qstats = alloc_percpu(struct gnet_stats_queue);
			if (!sch->cpu_qstats)
				goto err_out4;
974 975
		}

976 977 978 979
		if (tca[TCA_STAB]) {
			stab = qdisc_get_stab(tca[TCA_STAB]);
			if (IS_ERR(stab)) {
				err = PTR_ERR(stab);
980
				goto err_out4;
981
			}
982
			rcu_assign_pointer(sch->stab, stab);
983
		}
984
		if (tca[TCA_RATE]) {
985
			seqcount_t *running;
986

987 988 989 990
			err = -EOPNOTSUPP;
			if (sch->flags & TCQ_F_MQROOT)
				goto err_out4;

991
			if ((sch->parent != TC_H_ROOT) &&
992 993
			    !(sch->flags & TCQ_F_INGRESS) &&
			    (!p || !(p->flags & TCQ_F_MQROOT)))
994
				running = qdisc_root_sleeping_running(sch);
995
			else
996
				running = &sch->running;
997

998 999 1000
			err = gen_new_estimator(&sch->bstats,
						sch->cpu_bstats,
						&sch->rate_est,
1001 1002
						NULL,
						running,
1003
						tca[TCA_RATE]);
1004 1005
			if (err)
				goto err_out4;
1006
		}
1007

1008
		qdisc_hash_add(sch, false);
L
Linus Torvalds 已提交
1009 1010 1011

		return sch;
	}
1012 1013
	/* ops->init() failed, we call ->destroy() like qdisc_create_dflt() */
	ops->destroy(sch);
L
Linus Torvalds 已提交
1014 1015
err_out3:
	dev_put(dev);
1016
	kfree((char *) sch - sch->padded);
L
Linus Torvalds 已提交
1017 1018 1019 1020 1021
err_out2:
	module_put(ops->owner);
err_out:
	*errp = err;
	return NULL;
1022 1023

err_out4:
1024
	free_percpu(sch->cpu_bstats);
1025
	free_percpu(sch->cpu_qstats);
1026 1027 1028 1029
	/*
	 * Any broken qdiscs that would require a ops->reset() here?
	 * The qdisc was never in action so it shouldn't be necessary.
	 */
1030
	qdisc_put_stab(rtnl_dereference(sch->stab));
1031 1032 1033
	if (ops->destroy)
		ops->destroy(sch);
	goto err_out3;
L
Linus Torvalds 已提交
1034 1035
}

1036
static int qdisc_change(struct Qdisc *sch, struct nlattr **tca)
L
Linus Torvalds 已提交
1037
{
1038
	struct qdisc_size_table *ostab, *stab = NULL;
1039
	int err = 0;
L
Linus Torvalds 已提交
1040

1041
	if (tca[TCA_OPTIONS]) {
L
Linus Torvalds 已提交
1042 1043
		if (sch->ops->change == NULL)
			return -EINVAL;
1044
		err = sch->ops->change(sch, tca[TCA_OPTIONS]);
L
Linus Torvalds 已提交
1045 1046 1047
		if (err)
			return err;
	}
1048 1049 1050 1051 1052 1053 1054

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

1055 1056 1057
	ostab = rtnl_dereference(sch->stab);
	rcu_assign_pointer(sch->stab, stab);
	qdisc_put_stab(ostab);
1058

1059
	if (tca[TCA_RATE]) {
1060 1061
		/* NB: ignores errors from replace_estimator
		   because change can't be undone. */
1062 1063
		if (sch->flags & TCQ_F_MQROOT)
			goto out;
1064 1065 1066
		gen_replace_estimator(&sch->bstats,
				      sch->cpu_bstats,
				      &sch->rate_est,
1067 1068
				      NULL,
				      qdisc_root_sleeping_running(sch),
1069
				      tca[TCA_RATE]);
1070 1071
	}
out:
L
Linus Torvalds 已提交
1072 1073 1074
	return 0;
}

E
Eric Dumazet 已提交
1075 1076
struct check_loop_arg {
	struct qdisc_walker	w;
L
Linus Torvalds 已提交
1077 1078 1079 1080
	struct Qdisc		*p;
	int			depth;
};

1081 1082
static int check_loop_fn(struct Qdisc *q, unsigned long cl,
			 struct qdisc_walker *w);
L
Linus Torvalds 已提交
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102

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;
1103
	const struct Qdisc_class_ops *cops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
	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.
 */

1119
static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1120
{
1121
	struct net *net = sock_net(skb->sk);
1122
	struct tcmsg *tcm = nlmsg_data(n);
1123
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1124
	struct net_device *dev;
1125
	u32 clid;
L
Linus Torvalds 已提交
1126 1127 1128 1129
	struct Qdisc *q = NULL;
	struct Qdisc *p = NULL;
	int err;

1130
	if ((n->nlmsg_type != RTM_GETQDISC) &&
1131
	    !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1132 1133
		return -EPERM;

1134
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1135 1136 1137
	if (err < 0)
		return err;

1138 1139 1140 1141 1142
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

	clid = tcm->tcm_parent;
L
Linus Torvalds 已提交
1143 1144 1145
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) {
E
Eric Dumazet 已提交
1146 1147
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1148 1149
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1150 1151
			} else if (dev_ingress_queue(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
1152
			}
L
Linus Torvalds 已提交
1153
		} else {
1154
			q = dev->qdisc;
L
Linus Torvalds 已提交
1155 1156 1157 1158 1159 1160 1161
		}
		if (!q)
			return -ENOENT;

		if (tcm->tcm_handle && q->handle != tcm->tcm_handle)
			return -EINVAL;
	} else {
E
Eric Dumazet 已提交
1162 1163
		q = qdisc_lookup(dev, tcm->tcm_handle);
		if (!q)
L
Linus Torvalds 已提交
1164 1165 1166
			return -ENOENT;
	}

1167
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1168 1169 1170 1171 1172 1173 1174
		return -EINVAL;

	if (n->nlmsg_type == RTM_DELQDISC) {
		if (!clid)
			return -EINVAL;
		if (q->handle == 0)
			return -ENOENT;
E
Eric Dumazet 已提交
1175 1176
		err = qdisc_graft(dev, p, skb, n, clid, NULL, q);
		if (err != 0)
L
Linus Torvalds 已提交
1177 1178
			return err;
	} else {
1179
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1180 1181 1182 1183 1184
	}
	return 0;
}

/*
E
Eric Dumazet 已提交
1185
 * Create/change qdisc.
L
Linus Torvalds 已提交
1186 1187
 */

1188
static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1189
{
1190
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1191
	struct tcmsg *tcm;
1192
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197
	struct net_device *dev;
	u32 clid;
	struct Qdisc *q, *p;
	int err;

1198
	if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1199 1200
		return -EPERM;

L
Linus Torvalds 已提交
1201 1202
replay:
	/* Reinit, just in case something touches this. */
1203
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1204 1205 1206
	if (err < 0)
		return err;

1207
	tcm = nlmsg_data(n);
L
Linus Torvalds 已提交
1208 1209 1210
	clid = tcm->tcm_parent;
	q = p = NULL;

E
Eric Dumazet 已提交
1211 1212
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1213 1214
		return -ENODEV;

1215

L
Linus Torvalds 已提交
1216 1217 1218
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (clid != TC_H_INGRESS) {
E
Eric Dumazet 已提交
1219 1220
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1221 1222
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1223 1224
			} else if (dev_ingress_queue_create(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
L
Linus Torvalds 已提交
1225 1226
			}
		} else {
1227
			q = dev->qdisc;
L
Linus Torvalds 已提交
1228 1229 1230 1231 1232 1233 1234 1235
		}

		/* 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
1236
				if (q && !(n->nlmsg_flags & NLM_F_REPLACE))
L
Linus Torvalds 已提交
1237 1238 1239
					return -EEXIST;
				if (TC_H_MIN(tcm->tcm_handle))
					return -EINVAL;
E
Eric Dumazet 已提交
1240 1241
				q = qdisc_lookup(dev, tcm->tcm_handle);
				if (!q)
L
Linus Torvalds 已提交
1242
					goto create_n_graft;
E
Eric Dumazet 已提交
1243
				if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1244
					return -EEXIST;
1245
				if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1246 1247 1248 1249 1250 1251 1252
					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
1253
				if (!q)
L
Linus Torvalds 已提交
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
					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
1275 1276 1277
				if ((n->nlmsg_flags & NLM_F_CREATE) &&
				    (n->nlmsg_flags & NLM_F_REPLACE) &&
				    ((n->nlmsg_flags & NLM_F_EXCL) ||
1278 1279
				     (tca[TCA_KIND] &&
				      nla_strcmp(tca[TCA_KIND], q->ops->id))))
L
Linus Torvalds 已提交
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
					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
1292
	if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1293
		return -EEXIST;
1294
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1295 1296 1297
		return -EINVAL;
	err = qdisc_change(q, tca);
	if (err == 0)
1298
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1299 1300 1301
	return err;

create_n_graft:
E
Eric Dumazet 已提交
1302
	if (!(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1303
		return -ENOENT;
1304 1305 1306 1307 1308 1309 1310 1311
	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 {
1312
		struct netdev_queue *dev_queue;
1313 1314

		if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue)
1315 1316 1317 1318 1319
			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);
1320

1321
		q = qdisc_create(dev, dev_queue, p,
1322
				 tcm->tcm_parent, tcm->tcm_handle,
1323
				 tca, &err);
1324
	}
L
Linus Torvalds 已提交
1325 1326 1327 1328 1329 1330 1331
	if (q == NULL) {
		if (err == -EAGAIN)
			goto replay;
		return err;
	}

graft:
1332 1333 1334 1335 1336
	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
1337
	}
1338

L
Linus Torvalds 已提交
1339 1340 1341 1342
	return 0;
}

static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid,
1343
			 u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1344
{
1345
	struct gnet_stats_basic_cpu __percpu *cpu_bstats = NULL;
1346
	struct gnet_stats_queue __percpu *cpu_qstats = NULL;
L
Linus Torvalds 已提交
1347 1348
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1349
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1350
	struct gnet_dump d;
1351
	struct qdisc_size_table *stab;
1352
	__u32 qlen;
L
Linus Torvalds 已提交
1353

1354
	cond_resched();
1355
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1356 1357 1358
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1359
	tcm->tcm_family = AF_UNSPEC;
1360 1361
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1362
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1363 1364 1365
	tcm->tcm_parent = clid;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = atomic_read(&q->refcnt);
1366 1367
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1368
	if (q->ops->dump && q->ops->dump(q, skb) < 0)
1369
		goto nla_put_failure;
1370
	qlen = q->q.qlen;
L
Linus Torvalds 已提交
1371

1372 1373
	stab = rtnl_dereference(q->stab);
	if (stab && qdisc_dump_stab(skb, stab) < 0)
1374 1375
		goto nla_put_failure;

1376
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
1377
					 NULL, &d, TCA_PAD) < 0)
1378
		goto nla_put_failure;
L
Linus Torvalds 已提交
1379 1380

	if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0)
1381
		goto nla_put_failure;
L
Linus Torvalds 已提交
1382

1383
	if (qdisc_is_percpu_stats(q)) {
1384
		cpu_bstats = q->cpu_bstats;
1385 1386
		cpu_qstats = q->cpu_qstats;
	}
1387

1388 1389
	if (gnet_stats_copy_basic(qdisc_root_sleeping_running(q),
				  &d, cpu_bstats, &q->bstats) < 0 ||
1390
	    gnet_stats_copy_rate_est(&d, &q->rate_est) < 0 ||
1391
	    gnet_stats_copy_queue(&d, cpu_qstats, &q->qstats, qlen) < 0)
1392
		goto nla_put_failure;
1393

L
Linus Torvalds 已提交
1394
	if (gnet_stats_finish_copy(&d) < 0)
1395
		goto nla_put_failure;
1396

1397
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1398 1399
	return skb->len;

1400
out_nlmsg_trim:
1401
nla_put_failure:
1402
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1403 1404 1405
	return -1;
}

1406
static bool tc_qdisc_dump_ignore(struct Qdisc *q, bool dump_invisible)
1407
{
1408 1409 1410 1411 1412 1413
	if (q->flags & TCQ_F_BUILTIN)
		return true;
	if ((q->flags & TCQ_F_INVISIBLE) && !dump_invisible)
		return true;

	return false;
1414 1415
}

1416 1417 1418
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
1419 1420
{
	struct sk_buff *skb;
1421
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1422 1423 1424 1425 1426

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

1427
	if (old && !tc_qdisc_dump_ignore(old, false)) {
1428
		if (tc_fill_qdisc(skb, old, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1429
				  0, RTM_DELQDISC) < 0)
L
Linus Torvalds 已提交
1430 1431
			goto err_out;
	}
1432
	if (new && !tc_qdisc_dump_ignore(new, false)) {
1433
		if (tc_fill_qdisc(skb, new, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1434
				  old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0)
L
Linus Torvalds 已提交
1435 1436 1437 1438
			goto err_out;
	}

	if (skb->len)
1439
		return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1440
				      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1441 1442 1443 1444 1445 1446

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

1447 1448
static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb,
			      struct netlink_callback *cb,
1449 1450
			      int *q_idx_p, int s_q_idx, bool recur,
			      bool dump_invisible)
1451 1452 1453
{
	int ret = 0, q_idx = *q_idx_p;
	struct Qdisc *q;
1454
	int b;
1455 1456 1457 1458 1459 1460 1461 1462

	if (!root)
		return 0;

	q = root;
	if (q_idx < s_q_idx) {
		q_idx++;
	} else {
1463
		if (!tc_qdisc_dump_ignore(q, dump_invisible) &&
1464
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1465 1466
				  cb->nlh->nlmsg_seq, NLM_F_MULTI,
				  RTM_NEWQDISC) <= 0)
1467 1468 1469
			goto done;
		q_idx++;
	}
1470

1471 1472 1473 1474 1475 1476 1477
	/* 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)
1478 1479
		goto out;

1480
	hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
1481 1482 1483 1484
		if (q_idx < s_q_idx) {
			q_idx++;
			continue;
		}
1485
		if (!tc_qdisc_dump_ignore(q, dump_invisible) &&
1486
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1487 1488
				  cb->nlh->nlmsg_seq, NLM_F_MULTI,
				  RTM_NEWQDISC) <= 0)
1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
			goto done;
		q_idx++;
	}

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

L
Linus Torvalds 已提交
1501 1502
static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb)
{
1503
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1504 1505 1506
	int idx, q_idx;
	int s_idx, s_q_idx;
	struct net_device *dev;
1507 1508 1509 1510
	const struct nlmsghdr *nlh = cb->nlh;
	struct tcmsg *tcm = nlmsg_data(nlh);
	struct nlattr *tca[TCA_MAX + 1];
	int err;
L
Linus Torvalds 已提交
1511 1512 1513

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

1515
	idx = 0;
1516
	ASSERT_RTNL();
1517

1518
	err = nlmsg_parse(nlh, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1519 1520 1521
	if (err < 0)
		return err;

1522
	for_each_netdev(net, dev) {
1523 1524
		struct netdev_queue *dev_queue;

L
Linus Torvalds 已提交
1525
		if (idx < s_idx)
1526
			goto cont;
L
Linus Torvalds 已提交
1527 1528 1529
		if (idx > s_idx)
			s_q_idx = 0;
		q_idx = 0;
1530

1531
		if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx,
1532
				       true, tca[TCA_DUMP_INVISIBLE]) < 0)
1533 1534
			goto done;

1535 1536 1537
		dev_queue = dev_ingress_queue(dev);
		if (dev_queue &&
		    tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb,
1538 1539
				       &q_idx, s_q_idx, false,
				       tca[TCA_DUMP_INVISIBLE]) < 0)
1540 1541
			goto done;

1542 1543
cont:
		idx++;
L
Linus Torvalds 已提交
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
	}

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

	return skb->len;
}



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



1561
static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1562
{
1563
	struct net *net = sock_net(skb->sk);
1564
	struct tcmsg *tcm = nlmsg_data(n);
1565
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1566 1567
	struct net_device *dev;
	struct Qdisc *q = NULL;
1568
	const struct Qdisc_class_ops *cops;
L
Linus Torvalds 已提交
1569 1570
	unsigned long cl = 0;
	unsigned long new_cl;
1571 1572 1573
	u32 portid;
	u32 clid;
	u32 qid;
L
Linus Torvalds 已提交
1574 1575
	int err;

1576
	if ((n->nlmsg_type != RTM_GETTCLASS) &&
1577
	    !netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN))
1578 1579
		return -EPERM;

1580
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL, NULL);
1581 1582 1583
	if (err < 0)
		return err;

1584 1585 1586 1587
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

L
Linus Torvalds 已提交
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
	/*
	   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 */

1603 1604 1605 1606
	portid = tcm->tcm_parent;
	clid = tcm->tcm_handle;
	qid = TC_H_MAJ(clid);

1607 1608
	if (portid != TC_H_ROOT) {
		u32 qid1 = TC_H_MAJ(portid);
L
Linus Torvalds 已提交
1609 1610 1611 1612 1613 1614 1615 1616

		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)
1617
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1618 1619

		/* Now qid is genuine qdisc handle consistent
E
Eric Dumazet 已提交
1620 1621
		 * both with parent and child.
		 *
1622
		 * TC_H_MAJ(portid) still may be unspecified, complete it now.
L
Linus Torvalds 已提交
1623
		 */
1624 1625
		if (portid)
			portid = TC_H_MAKE(qid, portid);
L
Linus Torvalds 已提交
1626 1627
	} else {
		if (qid == 0)
1628
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1629 1630 1631
	}

	/* OK. Locate qdisc */
E
Eric Dumazet 已提交
1632 1633
	q = qdisc_lookup(dev, qid);
	if (!q)
L
Linus Torvalds 已提交
1634 1635 1636 1637 1638 1639 1640 1641 1642
		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) {
1643
		if (portid == TC_H_ROOT)
L
Linus Torvalds 已提交
1644 1645 1646 1647 1648 1649 1650 1651 1652
			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
1653 1654
		if (n->nlmsg_type != RTM_NEWTCLASS ||
		    !(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1655 1656 1657
			goto out;
	} else {
		switch (n->nlmsg_type) {
1658
		case RTM_NEWTCLASS:
L
Linus Torvalds 已提交
1659
			err = -EEXIST;
E
Eric Dumazet 已提交
1660
			if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1661 1662 1663
				goto out;
			break;
		case RTM_DELTCLASS:
1664 1665 1666
			err = -EOPNOTSUPP;
			if (cops->delete)
				err = cops->delete(q, cl);
L
Linus Torvalds 已提交
1667
			if (err == 0)
1668 1669
				tclass_notify(net, skb, n, q, cl,
					      RTM_DELTCLASS);
L
Linus Torvalds 已提交
1670 1671
			goto out;
		case RTM_GETTCLASS:
1672
			err = tclass_notify(net, skb, n, q, cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1673 1674 1675 1676 1677 1678 1679 1680
			goto out;
		default:
			err = -EINVAL;
			goto out;
		}
	}

	new_cl = cl;
1681 1682
	err = -EOPNOTSUPP;
	if (cops->change)
1683
		err = cops->change(q, clid, portid, tca, &new_cl);
L
Linus Torvalds 已提交
1684
	if (err == 0)
1685
		tclass_notify(net, skb, n, q, new_cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696

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

	return err;
}


static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q,
			  unsigned long cl,
1697
			  u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1698 1699 1700
{
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1701
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1702
	struct gnet_dump d;
1703
	const struct Qdisc_class_ops *cl_ops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1704

1705
	cond_resched();
1706
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1707 1708 1709
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1710
	tcm->tcm_family = AF_UNSPEC;
1711 1712
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1713
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1714 1715 1716
	tcm->tcm_parent = q->handle;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = 0;
1717 1718
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1719
	if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0)
1720
		goto nla_put_failure;
L
Linus Torvalds 已提交
1721

1722
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
1723
					 NULL, &d, TCA_PAD) < 0)
1724
		goto nla_put_failure;
L
Linus Torvalds 已提交
1725 1726

	if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0)
1727
		goto nla_put_failure;
L
Linus Torvalds 已提交
1728 1729

	if (gnet_stats_finish_copy(&d) < 0)
1730
		goto nla_put_failure;
L
Linus Torvalds 已提交
1731

1732
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1733 1734
	return skb->len;

1735
out_nlmsg_trim:
1736
nla_put_failure:
1737
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1738 1739 1740
	return -1;
}

1741 1742 1743
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
1744 1745
{
	struct sk_buff *skb;
1746
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1747 1748 1749 1750 1751

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

1752
	if (tc_fill_tclass(skb, q, cl, portid, n->nlmsg_seq, 0, event) < 0) {
L
Linus Torvalds 已提交
1753 1754 1755 1756
		kfree_skb(skb);
		return -EINVAL;
	}

1757
	return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1758
			      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1759 1760
}

E
Eric Dumazet 已提交
1761 1762 1763 1764
struct qdisc_dump_args {
	struct qdisc_walker	w;
	struct sk_buff		*skb;
	struct netlink_callback	*cb;
L
Linus Torvalds 已提交
1765 1766
};

1767 1768
static int qdisc_class_dump(struct Qdisc *q, unsigned long cl,
			    struct qdisc_walker *arg)
L
Linus Torvalds 已提交
1769 1770 1771
{
	struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg;

1772
	return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).portid,
1773 1774
			      a->cb->nlh->nlmsg_seq, NLM_F_MULTI,
			      RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1775 1776
}

1777 1778 1779 1780 1781 1782
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;

1783
	if (tc_qdisc_dump_ignore(q, false) ||
1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810
	    *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;
1811
	int b;
1812 1813 1814 1815 1816 1817 1818

	if (!root)
		return 0;

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

1819 1820 1821
	if (!qdisc_dev(root))
		return 0;

1822
	hash_for_each(qdisc_dev(root)->qdisc_hash, b, q, hash) {
1823 1824 1825 1826 1827 1828 1829
		if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0)
			return -1;
	}

	return 0;
}

L
Linus Torvalds 已提交
1830 1831
static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb)
{
1832
	struct tcmsg *tcm = nlmsg_data(cb->nlh);
1833
	struct net *net = sock_net(skb->sk);
1834
	struct netdev_queue *dev_queue;
L
Linus Torvalds 已提交
1835
	struct net_device *dev;
1836
	int t, s_t;
L
Linus Torvalds 已提交
1837

1838
	if (nlmsg_len(cb->nlh) < sizeof(*tcm))
L
Linus Torvalds 已提交
1839
		return 0;
E
Eric Dumazet 已提交
1840 1841
	dev = dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1842 1843 1844 1845 1846
		return 0;

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

1847
	if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0)
1848 1849
		goto done;

1850 1851 1852 1853
	dev_queue = dev_ingress_queue(dev);
	if (dev_queue &&
	    tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb,
				&t, s_t) < 0)
1854
		goto done;
L
Linus Torvalds 已提交
1855

1856
done:
L
Linus Torvalds 已提交
1857 1858 1859 1860 1861 1862 1863
	cb->args[0] = t;

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

/* Main classifier routine: scans classifier chain attached
E
Eric Dumazet 已提交
1864 1865
 * to this qdisc, (optionally) tests for protocol and asks
 * specific classifiers.
L
Linus Torvalds 已提交
1866
 */
1867 1868
int tc_classify(struct sk_buff *skb, const struct tcf_proto *tp,
		struct tcf_result *res, bool compat_mode)
1869
{
1870
	__be16 protocol = tc_skb_protocol(skb);
1871
#ifdef CONFIG_NET_CLS_ACT
1872
	const int max_reclassify_loop = 4;
1873 1874
	const struct tcf_proto *old_tp = tp;
	int limit = 0;
1875

1876 1877
reclassify:
#endif
J
John Fastabend 已提交
1878
	for (; tp; tp = rcu_dereference_bh(tp->next)) {
1879 1880
		int err;

E
Eric Dumazet 已提交
1881 1882 1883 1884
		if (tp->protocol != protocol &&
		    tp->protocol != htons(ETH_P_ALL))
			continue;

1885 1886
		err = tp->classify(skb, tp, res);
#ifdef CONFIG_NET_CLS_ACT
1887
		if (unlikely(err == TC_ACT_RECLASSIFY && !compat_mode))
1888 1889
			goto reset;
#endif
1890
		if (err >= 0)
1891 1892 1893
			return err;
	}

1894
	return TC_ACT_UNSPEC; /* signal: continue lookup */
L
Linus Torvalds 已提交
1895
#ifdef CONFIG_NET_CLS_ACT
1896
reset:
1897
	if (unlikely(limit++ >= max_reclassify_loop)) {
1898 1899 1900
		net_notice_ratelimited("%s: reclassify loop, rule prio %u, protocol %02x\n",
				       tp->q->ops->id, tp->prio & 0xffff,
				       ntohs(tp->protocol));
1901
		return TC_ACT_SHOT;
L
Linus Torvalds 已提交
1902
	}
1903 1904

	tp = old_tp;
1905
	protocol = tc_skb_protocol(skb);
1906
	goto reclassify;
1907
#endif
L
Linus Torvalds 已提交
1908
}
1909
EXPORT_SYMBOL(tc_classify);
L
Linus Torvalds 已提交
1910 1911 1912 1913 1914

#ifdef CONFIG_PROC_FS
static int psched_show(struct seq_file *seq, void *v)
{
	seq_printf(seq, "%08x %08x %08x %08x\n",
1915
		   (u32)NSEC_PER_USEC, (u32)PSCHED_TICKS2NS(1),
1916
		   1000000,
1917
		   (u32)NSEC_PER_SEC / hrtimer_resolution);
L
Linus Torvalds 已提交
1918 1919 1920 1921 1922 1923

	return 0;
}

static int psched_open(struct inode *inode, struct file *file)
{
1924
	return single_open(file, psched_show, NULL);
L
Linus Torvalds 已提交
1925 1926
}

1927
static const struct file_operations psched_fops = {
L
Linus Torvalds 已提交
1928 1929 1930 1931 1932
	.owner = THIS_MODULE,
	.open = psched_open,
	.read  = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
1933
};
1934 1935 1936 1937 1938

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

1939
	e = proc_create("psched", 0, net->proc_net, &psched_fops);
1940 1941 1942 1943 1944 1945 1946 1947
	if (e == NULL)
		return -ENOMEM;

	return 0;
}

static void __net_exit psched_net_exit(struct net *net)
{
1948
	remove_proc_entry("psched", net->proc_net);
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
}
#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
1959 1960
#endif

1961 1962 1963 1964 1965
static struct pernet_operations psched_net_ops = {
	.init = psched_net_init,
	.exit = psched_net_exit,
};

L
Linus Torvalds 已提交
1966 1967
static int __init pktsched_init(void)
{
1968 1969 1970 1971
	int err;

	err = register_pernet_subsys(&psched_net_ops);
	if (err) {
E
Eric Dumazet 已提交
1972
		pr_err("pktsched_init: "
1973 1974 1975 1976
		       "cannot initialize per netns operations\n");
		return err;
	}

1977
	register_qdisc(&pfifo_fast_ops);
L
Linus Torvalds 已提交
1978 1979
	register_qdisc(&pfifo_qdisc_ops);
	register_qdisc(&bfifo_qdisc_ops);
1980
	register_qdisc(&pfifo_head_drop_qdisc_ops);
1981
	register_qdisc(&mq_qdisc_ops);
1982
	register_qdisc(&noqueue_qdisc_ops);
L
Linus Torvalds 已提交
1983

1984 1985
	rtnl_register(PF_UNSPEC, RTM_NEWQDISC, tc_modify_qdisc, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELQDISC, tc_get_qdisc, NULL, NULL);
1986 1987
	rtnl_register(PF_UNSPEC, RTM_GETQDISC, tc_get_qdisc, tc_dump_qdisc,
		      NULL);
1988 1989
	rtnl_register(PF_UNSPEC, RTM_NEWTCLASS, tc_ctl_tclass, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELTCLASS, tc_ctl_tclass, NULL, NULL);
1990 1991
	rtnl_register(PF_UNSPEC, RTM_GETTCLASS, tc_ctl_tclass, tc_dump_tclass,
		      NULL);
1992

L
Linus Torvalds 已提交
1993 1994 1995 1996
	return 0;
}

subsys_initcall(pktsched_init);
新手
引导
客服 返回
顶部