sch_api.c 44.3 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>
L
Linus Torvalds 已提交
32

33
#include <net/net_namespace.h>
34
#include <net/sock.h>
35
#include <net/netlink.h>
L
Linus Torvalds 已提交
36 37
#include <net/pkt_sched.h>

38 39
static int qdisc_notify(struct net *net, struct sk_buff *oskb,
			struct nlmsghdr *n, u32 clid,
L
Linus Torvalds 已提交
40
			struct Qdisc *old, struct Qdisc *new);
41 42 43
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
44 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 99 100 101 102

/*

   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
   NET_XMIT_POLICED	- dropped by police.
     Expected action: backoff or error to real-time apps.

   Auxiliary routines:

103 104 105 106
   ---peek

   like dequeue but without removing a packet from the queue

L
Linus Torvalds 已提交
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
   ---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;

138
/* Register/unregister queueing discipline */
L
Linus Torvalds 已提交
139 140 141 142 143 144 145 146 147 148 149 150 151

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;
152
	if (qops->peek == NULL) {
153
		if (qops->dequeue == NULL)
154
			qops->peek = noop_qdisc_ops.peek;
155 156
		else
			goto out_einval;
157
	}
L
Linus Torvalds 已提交
158 159 160
	if (qops->dequeue == NULL)
		qops->dequeue = noop_qdisc_ops.dequeue;

161 162 163
	if (qops->cl_ops) {
		const struct Qdisc_class_ops *cops = qops->cl_ops;

164
		if (!(cops->get && cops->put && cops->walk && cops->leaf))
165 166 167 168 169 170
			goto out_einval;

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

L
Linus Torvalds 已提交
171 172 173 174 175 176
	qops->next = NULL;
	*qp = qops;
	rc = 0;
out:
	write_unlock(&qdisc_mod_lock);
	return rc;
177 178 179 180

out_einval:
	rc = -EINVAL;
	goto out;
L
Linus Torvalds 已提交
181
}
182
EXPORT_SYMBOL(register_qdisc);
L
Linus Torvalds 已提交
183 184 185 186 187 188 189

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
190
	for (qp = &qdisc_base; (q = *qp) != NULL; qp = &q->next)
L
Linus Torvalds 已提交
191 192 193 194 195 196 197 198 199 200
		if (q == qops)
			break;
	if (q) {
		*qp = q->next;
		q->next = NULL;
		err = 0;
	}
	write_unlock(&qdisc_mod_lock);
	return err;
}
201
EXPORT_SYMBOL(unregister_qdisc);
L
Linus Torvalds 已提交
202

203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
/* 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
255 256 257 258
/* We know handle. Find qdisc among all qdisc's attached to device
   (root qdisc, all its children, children of children etc.)
 */

H
Hannes Eder 已提交
259
static struct Qdisc *qdisc_match_from_root(struct Qdisc *root, u32 handle)
260 261 262 263 264 265 266 267 268 269 270 271 272 273
{
	struct Qdisc *q;

	if (!(root->flags & TCQ_F_BUILTIN) &&
	    root->handle == handle)
		return root;

	list_for_each_entry(q, &root->list, list) {
		if (q->handle == handle)
			return q;
	}
	return NULL;
}

274
void qdisc_list_add(struct Qdisc *q)
275
{
276 277 278
	struct Qdisc *root = qdisc_dev(q)->qdisc;

	WARN_ON_ONCE(root == &noop_qdisc);
279
	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS))
280
		list_add_tail(&q->list, &root->list);
281
}
282
EXPORT_SYMBOL(qdisc_list_add);
283 284 285

void qdisc_list_del(struct Qdisc *q)
{
286
	if ((q->parent != TC_H_ROOT) && !(q->flags & TCQ_F_INGRESS))
287 288 289 290
		list_del(&q->list);
}
EXPORT_SYMBOL(qdisc_list_del);

291
struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle)
L
Linus Torvalds 已提交
292
{
293 294
	struct Qdisc *q;

295 296 297
	q = qdisc_match_from_root(dev->qdisc, handle);
	if (q)
		goto out;
298

299 300 301 302
	if (dev_ingress_queue(dev))
		q = qdisc_match_from_root(
			dev_ingress_queue(dev)->qdisc_sleeping,
			handle);
303
out:
304
	return q;
L
Linus Torvalds 已提交
305 306 307 308 309 310
}

static struct Qdisc *qdisc_leaf(struct Qdisc *p, u32 classid)
{
	unsigned long cl;
	struct Qdisc *leaf;
311
	const struct Qdisc_class_ops *cops = p->ops->cl_ops;
L
Linus Torvalds 已提交
312 313 314 315 316 317 318 319 320 321 322 323 324 325

	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 */

326
static struct Qdisc_ops *qdisc_lookup_ops(struct nlattr *kind)
L
Linus Torvalds 已提交
327 328 329 330 331 332
{
	struct Qdisc_ops *q = NULL;

	if (kind) {
		read_lock(&qdisc_mod_lock);
		for (q = qdisc_base; q; q = q->next) {
333
			if (nla_strcmp(kind, q->id) == 0) {
L
Linus Torvalds 已提交
334 335 336 337 338 339 340 341 342 343
				if (!try_module_get(q->owner))
					q = NULL;
				break;
			}
		}
		read_unlock(&qdisc_mod_lock);
	}
	return q;
}

344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382
/* 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
383 384
static struct qdisc_rate_table *qdisc_rtab_list;

385
struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r, struct nlattr *tab)
L
Linus Torvalds 已提交
386 387 388
{
	struct qdisc_rate_table *rtab;

389 390 391 392
	if (tab == NULL || r->rate == 0 || r->cell_log == 0 ||
	    nla_len(tab) != TC_RTAB_SIZE)
		return NULL;

L
Linus Torvalds 已提交
393
	for (rtab = qdisc_rtab_list; rtab; rtab = rtab->next) {
394 395
		if (!memcmp(&rtab->rate, r, sizeof(struct tc_ratespec)) &&
		    !memcmp(&rtab->data, nla_data(tab), 1024)) {
L
Linus Torvalds 已提交
396 397 398 399 400 401 402 403 404
			rtab->refcnt++;
			return rtab;
		}
	}

	rtab = kmalloc(sizeof(*rtab), GFP_KERNEL);
	if (rtab) {
		rtab->rate = *r;
		rtab->refcnt = 1;
405
		memcpy(rtab->data, nla_data(tab), 1024);
406 407
		if (r->linklayer == TC_LINKLAYER_UNAWARE)
			r->linklayer = __detect_linklayer(r, rtab->data);
L
Linus Torvalds 已提交
408 409 410 411 412
		rtab->next = qdisc_rtab_list;
		qdisc_rtab_list = rtab;
	}
	return rtab;
}
413
EXPORT_SYMBOL(qdisc_get_rtab);
L
Linus Torvalds 已提交
414 415 416 417 418 419 420 421

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
422 423 424
	for (rtabp = &qdisc_rtab_list;
	     (rtab = *rtabp) != NULL;
	     rtabp = &rtab->next) {
L
Linus Torvalds 已提交
425 426 427 428 429 430 431
		if (rtab == tab) {
			*rtabp = rtab->next;
			kfree(rtab);
			return;
		}
	}
}
432
EXPORT_SYMBOL(qdisc_put_rtab);
L
Linus Torvalds 已提交
433

434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
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);
	}

466
	if (tsize != s->tsize || (!tab && tsize > 0))
467 468
		return ERR_PTR(-EINVAL);

469
	spin_lock(&qdisc_stab_lock);
470 471 472 473 474 475 476

	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++;
477
		spin_unlock(&qdisc_stab_lock);
478 479 480
		return stab;
	}

481
	spin_unlock(&qdisc_stab_lock);
482 483 484 485 486 487 488 489 490 491

	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));

492
	spin_lock(&qdisc_stab_lock);
493
	list_add_tail(&stab->list, &qdisc_stab_list);
494
	spin_unlock(&qdisc_stab_lock);
495 496 497 498

	return stab;
}

E
Eric Dumazet 已提交
499 500 501 502 503
static void stab_kfree_rcu(struct rcu_head *head)
{
	kfree(container_of(head, struct qdisc_size_table, rcu));
}

504 505 506 507 508
void qdisc_put_stab(struct qdisc_size_table *tab)
{
	if (!tab)
		return;

509
	spin_lock(&qdisc_stab_lock);
510 511 512

	if (--tab->refcnt == 0) {
		list_del(&tab->list);
E
Eric Dumazet 已提交
513
		call_rcu_bh(&tab->rcu, stab_kfree_rcu);
514 515
	}

516
	spin_unlock(&qdisc_stab_lock);
517 518 519 520 521 522 523 524
}
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);
525 526
	if (nest == NULL)
		goto nla_put_failure;
527 528
	if (nla_put(skb, TCA_STAB_BASE, sizeof(stab->szopts), &stab->szopts))
		goto nla_put_failure;
529 530 531 532 533 534 535 536
	nla_nest_end(skb, nest);

	return skb->len;

nla_put_failure:
	return -1;
}

E
Eric Dumazet 已提交
537
void __qdisc_calculate_pkt_len(struct sk_buff *skb, const struct qdisc_size_table *stab)
538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
{
	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;
}
E
Eric Dumazet 已提交
563
EXPORT_SYMBOL(__qdisc_calculate_pkt_len);
564

565 566 567
void qdisc_warn_nonwc(char *txt, struct Qdisc *qdisc)
{
	if (!(qdisc->flags & TCQ_F_WARN_NONWC)) {
E
Eric Dumazet 已提交
568 569
		pr_warn("%s: %s qdisc %X: is non-work-conserving?\n",
			txt, qdisc->ops->id, qdisc->handle >> 16);
570 571 572 573 574
		qdisc->flags |= TCQ_F_WARN_NONWC;
	}
}
EXPORT_SYMBOL(qdisc_warn_nonwc);

575 576 577
static enum hrtimer_restart qdisc_watchdog(struct hrtimer *timer)
{
	struct qdisc_watchdog *wd = container_of(timer, struct qdisc_watchdog,
578
						 timer);
579

580
	qdisc_unthrottled(wd->qdisc);
581
	__netif_schedule(qdisc_root(wd->qdisc));
582

583 584 585 586 587
	return HRTIMER_NORESTART;
}

void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc)
{
588 589
	hrtimer_init(&wd->timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
	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
	qdisc_throttled(wd->qdisc);
E
Eric Dumazet 已提交
601 602

	hrtimer_start(&wd->timer,
603
		      ns_to_ktime(expires),
E
Eric Dumazet 已提交
604
		      HRTIMER_MODE_ABS);
605
}
606
EXPORT_SYMBOL(qdisc_watchdog_schedule_ns);
607 608 609

void qdisc_watchdog_cancel(struct qdisc_watchdog *wd)
{
610
	hrtimer_cancel(&wd->timer);
611
	qdisc_unthrottled(wd->qdisc);
612 613
}
EXPORT_SYMBOL(qdisc_watchdog_cancel);
L
Linus Torvalds 已提交
614

615
static struct hlist_head *qdisc_class_hash_alloc(unsigned int n)
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
{
	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;
646
	struct hlist_node *next;
647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
	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++) {
665
		hlist_for_each_entry_safe(cl, next, &ohash[i], hnode) {
666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
			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);

719 720 721
/* Allocate an unique handle from space managed by kernel
 * Possible range is [8000-FFFF]:0000 (0x8000 values)
 */
L
Linus Torvalds 已提交
722 723
static u32 qdisc_alloc_handle(struct net_device *dev)
{
724
	int i = 0x8000;
L
Linus Torvalds 已提交
725 726 727 728 729 730
	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);
731 732 733 734
		if (!qdisc_lookup(dev, autohandle))
			return autohandle;
		cond_resched();
	} while	(--i > 0);
L
Linus Torvalds 已提交
735

736
	return 0;
L
Linus Torvalds 已提交
737 738
}

739 740
void qdisc_tree_decrease_qlen(struct Qdisc *sch, unsigned int n)
{
741
	const struct Qdisc_class_ops *cops;
742 743
	unsigned long cl;
	u32 parentid;
744
	int drops;
745 746 747

	if (n == 0)
		return;
748
	drops = max_t(int, n, 0);
749
	while ((parentid = sch->parent)) {
750 751 752
		if (TC_H_MAJ(parentid) == TC_H_MAJ(TC_H_INGRESS))
			return;

753
		sch = qdisc_lookup(qdisc_dev(sch), TC_H_MAJ(parentid));
754 755 756 757
		if (sch == NULL) {
			WARN_ON(parentid != TC_H_ROOT);
			return;
		}
758 759 760 761 762 763 764
		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;
765
		sch->qstats.drops += drops;
766 767 768
	}
}
EXPORT_SYMBOL(qdisc_tree_decrease_qlen);
L
Linus Torvalds 已提交
769

770 771
static void notify_and_destroy(struct net *net, struct sk_buff *skb,
			       struct nlmsghdr *n, u32 clid,
772 773 774
			       struct Qdisc *old, struct Qdisc *new)
{
	if (new || old)
775
		qdisc_notify(net, skb, n, clid, old, new);
L
Linus Torvalds 已提交
776

777
	if (old)
778 779 780 781 782 783 784 785 786 787
		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
788 789 790
 */

static int qdisc_graft(struct net_device *dev, struct Qdisc *parent,
791 792
		       struct sk_buff *skb, struct nlmsghdr *n, u32 classid,
		       struct Qdisc *new, struct Qdisc *old)
L
Linus Torvalds 已提交
793
{
794
	struct Qdisc *q = old;
795
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
796 797
	int err = 0;

798
	if (parent == NULL) {
799 800 801 802
		unsigned int i, num_q, ingress;

		ingress = 0;
		num_q = dev->num_tx_queues;
803 804
		if ((q && q->flags & TCQ_F_INGRESS) ||
		    (new && new->flags & TCQ_F_INGRESS)) {
805 806
			num_q = 1;
			ingress = 1;
807 808
			if (!dev_ingress_queue(dev))
				return -ENOENT;
809 810 811 812 813
		}

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

814 815 816 817 818
		if (new && new->ops->attach) {
			new->ops->attach(new);
			num_q = 0;
		}

819
		for (i = 0; i < num_q; i++) {
820
			struct netdev_queue *dev_queue = dev_ingress_queue(dev);
821 822 823 824

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

825 826 827 828
			old = dev_graft_qdisc(dev_queue, new);
			if (new && i > 0)
				atomic_inc(&new->refcnt);

829 830
			if (!ingress)
				qdisc_destroy(old);
L
Linus Torvalds 已提交
831
		}
832

833
		if (!ingress) {
834 835
			notify_and_destroy(net, skb, n, classid,
					   dev->qdisc, new);
836 837 838 839
			if (new && !new->ops->attach)
				atomic_inc(&new->refcnt);
			dev->qdisc = new ? : &noop_qdisc;
		} else {
840
			notify_and_destroy(net, skb, n, classid, old, new);
841
		}
842

843 844
		if (dev->flags & IFF_UP)
			dev_activate(dev);
L
Linus Torvalds 已提交
845
	} else {
846
		const struct Qdisc_class_ops *cops = parent->ops->cl_ops;
L
Linus Torvalds 已提交
847

848 849
		err = -EOPNOTSUPP;
		if (cops && cops->graft) {
L
Linus Torvalds 已提交
850 851
			unsigned long cl = cops->get(parent, classid);
			if (cl) {
852
				err = cops->graft(parent, cl, new, &old);
L
Linus Torvalds 已提交
853
				cops->put(parent, cl);
854 855
			} else
				err = -ENOENT;
L
Linus Torvalds 已提交
856
		}
857
		if (!err)
858
			notify_and_destroy(net, skb, n, classid, old, new);
L
Linus Torvalds 已提交
859 860 861 862
	}
	return err;
}

863 864 865 866
/* 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
867 868 869 870 871 872 873
/*
   Allocate and initialize new qdisc.

   Parameters are passed via opt.
 */

static struct Qdisc *
874
qdisc_create(struct net_device *dev, struct netdev_queue *dev_queue,
875 876
	     struct Qdisc *p, u32 parent, u32 handle,
	     struct nlattr **tca, int *errp)
L
Linus Torvalds 已提交
877 878
{
	int err;
879
	struct nlattr *kind = tca[TCA_KIND];
L
Linus Torvalds 已提交
880 881
	struct Qdisc *sch;
	struct Qdisc_ops *ops;
882
	struct qdisc_size_table *stab;
L
Linus Torvalds 已提交
883 884

	ops = qdisc_lookup_ops(kind);
885
#ifdef CONFIG_MODULES
L
Linus Torvalds 已提交
886 887
	if (ops == NULL && kind != NULL) {
		char name[IFNAMSIZ];
888
		if (nla_strlcpy(name, kind, IFNAMSIZ) < IFNAMSIZ) {
L
Linus Torvalds 已提交
889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
			/* 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

913
	err = -ENOENT;
L
Linus Torvalds 已提交
914 915 916
	if (ops == NULL)
		goto err_out;

917
	sch = qdisc_alloc(dev_queue, ops);
918 919
	if (IS_ERR(sch)) {
		err = PTR_ERR(sch);
L
Linus Torvalds 已提交
920
		goto err_out2;
921
	}
L
Linus Torvalds 已提交
922

923 924
	sch->parent = parent;

925
	if (handle == TC_H_INGRESS) {
L
Linus Torvalds 已提交
926
		sch->flags |= TCQ_F_INGRESS;
927
		handle = TC_H_MAKE(TC_H_INGRESS, 0);
928
		lockdep_set_class(qdisc_lock(sch), &qdisc_rx_lock);
929 930 931 932 933 934 935
	} else {
		if (handle == 0) {
			handle = qdisc_alloc_handle(dev);
			err = -ENOMEM;
			if (handle == 0)
				goto err_out3;
		}
936
		lockdep_set_class(qdisc_lock(sch), &qdisc_tx_lock);
937 938
		if (!netif_is_multiqueue(dev))
			sch->flags |= TCQ_F_ONETXQUEUE;
L
Linus Torvalds 已提交
939 940
	}

941
	sch->handle = handle;
L
Linus Torvalds 已提交
942

943
	if (!ops->init || (err = ops->init(sch, tca[TCA_OPTIONS])) == 0) {
944 945 946 947
		if (tca[TCA_STAB]) {
			stab = qdisc_get_stab(tca[TCA_STAB]);
			if (IS_ERR(stab)) {
				err = PTR_ERR(stab);
948
				goto err_out4;
949
			}
E
Eric Dumazet 已提交
950
			rcu_assign_pointer(sch->stab, stab);
951
		}
952
		if (tca[TCA_RATE]) {
953 954
			spinlock_t *root_lock;

955 956 957 958
			err = -EOPNOTSUPP;
			if (sch->flags & TCQ_F_MQROOT)
				goto err_out4;

959
			if ((sch->parent != TC_H_ROOT) &&
960 961
			    !(sch->flags & TCQ_F_INGRESS) &&
			    (!p || !(p->flags & TCQ_F_MQROOT)))
962 963 964 965
				root_lock = qdisc_root_sleeping_lock(sch);
			else
				root_lock = qdisc_lock(sch);

966
			err = gen_new_estimator(&sch->bstats, &sch->rate_est,
967
						root_lock, tca[TCA_RATE]);
968 969
			if (err)
				goto err_out4;
970
		}
971 972

		qdisc_list_add(sch);
L
Linus Torvalds 已提交
973 974 975 976 977

		return sch;
	}
err_out3:
	dev_put(dev);
978
	kfree((char *) sch - sch->padded);
L
Linus Torvalds 已提交
979 980 981 982 983
err_out2:
	module_put(ops->owner);
err_out:
	*errp = err;
	return NULL;
984 985 986 987 988 989

err_out4:
	/*
	 * Any broken qdiscs that would require a ops->reset() here?
	 * The qdisc was never in action so it shouldn't be necessary.
	 */
E
Eric Dumazet 已提交
990
	qdisc_put_stab(rtnl_dereference(sch->stab));
991 992 993
	if (ops->destroy)
		ops->destroy(sch);
	goto err_out3;
L
Linus Torvalds 已提交
994 995
}

996
static int qdisc_change(struct Qdisc *sch, struct nlattr **tca)
L
Linus Torvalds 已提交
997
{
E
Eric Dumazet 已提交
998
	struct qdisc_size_table *ostab, *stab = NULL;
999
	int err = 0;
L
Linus Torvalds 已提交
1000

1001
	if (tca[TCA_OPTIONS]) {
L
Linus Torvalds 已提交
1002 1003
		if (sch->ops->change == NULL)
			return -EINVAL;
1004
		err = sch->ops->change(sch, tca[TCA_OPTIONS]);
L
Linus Torvalds 已提交
1005 1006 1007
		if (err)
			return err;
	}
1008 1009 1010 1011 1012 1013 1014

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

E
Eric Dumazet 已提交
1015 1016 1017
	ostab = rtnl_dereference(sch->stab);
	rcu_assign_pointer(sch->stab, stab);
	qdisc_put_stab(ostab);
1018

1019
	if (tca[TCA_RATE]) {
1020 1021
		/* NB: ignores errors from replace_estimator
		   because change can't be undone. */
1022 1023
		if (sch->flags & TCQ_F_MQROOT)
			goto out;
L
Linus Torvalds 已提交
1024
		gen_replace_estimator(&sch->bstats, &sch->rate_est,
1025 1026
					    qdisc_root_sleeping_lock(sch),
					    tca[TCA_RATE]);
1027 1028
	}
out:
L
Linus Torvalds 已提交
1029 1030 1031
	return 0;
}

E
Eric Dumazet 已提交
1032 1033
struct check_loop_arg {
	struct qdisc_walker	w;
L
Linus Torvalds 已提交
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
	struct Qdisc		*p;
	int			depth;
};

static int check_loop_fn(struct Qdisc *q, unsigned long cl, struct qdisc_walker *w);

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;
1059
	const struct Qdisc_class_ops *cops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
	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.
 */

1075
static int tc_get_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1076
{
1077
	struct net *net = sock_net(skb->sk);
1078
	struct tcmsg *tcm = nlmsg_data(n);
1079
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1080
	struct net_device *dev;
1081
	u32 clid;
L
Linus Torvalds 已提交
1082 1083 1084 1085
	struct Qdisc *q = NULL;
	struct Qdisc *p = NULL;
	int err;

1086 1087 1088
	if ((n->nlmsg_type != RTM_GETQDISC) && !capable(CAP_NET_ADMIN))
		return -EPERM;

1089 1090 1091 1092
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1093 1094 1095 1096 1097
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

	clid = tcm->tcm_parent;
L
Linus Torvalds 已提交
1098 1099 1100
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (TC_H_MAJ(clid) != TC_H_MAJ(TC_H_INGRESS)) {
E
Eric Dumazet 已提交
1101 1102
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1103 1104
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1105 1106
			} else if (dev_ingress_queue(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
1107
			}
L
Linus Torvalds 已提交
1108
		} else {
1109
			q = dev->qdisc;
L
Linus Torvalds 已提交
1110 1111 1112 1113 1114 1115 1116
		}
		if (!q)
			return -ENOENT;

		if (tcm->tcm_handle && q->handle != tcm->tcm_handle)
			return -EINVAL;
	} else {
E
Eric Dumazet 已提交
1117 1118
		q = qdisc_lookup(dev, tcm->tcm_handle);
		if (!q)
L
Linus Torvalds 已提交
1119 1120 1121
			return -ENOENT;
	}

1122
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1123 1124 1125 1126 1127 1128 1129
		return -EINVAL;

	if (n->nlmsg_type == RTM_DELQDISC) {
		if (!clid)
			return -EINVAL;
		if (q->handle == 0)
			return -ENOENT;
E
Eric Dumazet 已提交
1130 1131
		err = qdisc_graft(dev, p, skb, n, clid, NULL, q);
		if (err != 0)
L
Linus Torvalds 已提交
1132 1133
			return err;
	} else {
1134
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139
	}
	return 0;
}

/*
E
Eric Dumazet 已提交
1140
 * Create/change qdisc.
L
Linus Torvalds 已提交
1141 1142
 */

1143
static int tc_modify_qdisc(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1144
{
1145
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1146
	struct tcmsg *tcm;
1147
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1148 1149 1150 1151 1152
	struct net_device *dev;
	u32 clid;
	struct Qdisc *q, *p;
	int err;

1153 1154 1155
	if (!capable(CAP_NET_ADMIN))
		return -EPERM;

L
Linus Torvalds 已提交
1156 1157
replay:
	/* Reinit, just in case something touches this. */
1158 1159 1160 1161
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1162
	tcm = nlmsg_data(n);
L
Linus Torvalds 已提交
1163 1164 1165
	clid = tcm->tcm_parent;
	q = p = NULL;

E
Eric Dumazet 已提交
1166 1167
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1168 1169
		return -ENODEV;

1170

L
Linus Torvalds 已提交
1171 1172 1173
	if (clid) {
		if (clid != TC_H_ROOT) {
			if (clid != TC_H_INGRESS) {
E
Eric Dumazet 已提交
1174 1175
				p = qdisc_lookup(dev, TC_H_MAJ(clid));
				if (!p)
L
Linus Torvalds 已提交
1176 1177
					return -ENOENT;
				q = qdisc_leaf(p, clid);
E
Eric Dumazet 已提交
1178 1179
			} else if (dev_ingress_queue_create(dev)) {
				q = dev_ingress_queue(dev)->qdisc_sleeping;
L
Linus Torvalds 已提交
1180 1181
			}
		} else {
1182
			q = dev->qdisc;
L
Linus Torvalds 已提交
1183 1184 1185 1186 1187 1188 1189 1190
		}

		/* 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
1191
				if (q && !(n->nlmsg_flags & NLM_F_REPLACE))
L
Linus Torvalds 已提交
1192 1193 1194
					return -EEXIST;
				if (TC_H_MIN(tcm->tcm_handle))
					return -EINVAL;
E
Eric Dumazet 已提交
1195 1196
				q = qdisc_lookup(dev, tcm->tcm_handle);
				if (!q)
L
Linus Torvalds 已提交
1197
					goto create_n_graft;
E
Eric Dumazet 已提交
1198
				if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1199
					return -EEXIST;
1200
				if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1201 1202 1203 1204 1205 1206 1207
					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
1208
				if (!q)
L
Linus Torvalds 已提交
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
					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
1230 1231 1232
				if ((n->nlmsg_flags & NLM_F_CREATE) &&
				    (n->nlmsg_flags & NLM_F_REPLACE) &&
				    ((n->nlmsg_flags & NLM_F_EXCL) ||
1233 1234
				     (tca[TCA_KIND] &&
				      nla_strcmp(tca[TCA_KIND], q->ops->id))))
L
Linus Torvalds 已提交
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
					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
1247
	if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1248
		return -EEXIST;
1249
	if (tca[TCA_KIND] && nla_strcmp(tca[TCA_KIND], q->ops->id))
L
Linus Torvalds 已提交
1250 1251 1252
		return -EINVAL;
	err = qdisc_change(q, tca);
	if (err == 0)
1253
		qdisc_notify(net, skb, n, clid, NULL, q);
L
Linus Torvalds 已提交
1254 1255 1256
	return err;

create_n_graft:
E
Eric Dumazet 已提交
1257
	if (!(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1258
		return -ENOENT;
1259 1260 1261 1262 1263 1264 1265 1266
	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 {
1267
		struct netdev_queue *dev_queue;
1268 1269

		if (p && p->ops->cl_ops && p->ops->cl_ops->select_queue)
1270 1271 1272 1273 1274
			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);
1275

1276
		q = qdisc_create(dev, dev_queue, p,
1277
				 tcm->tcm_parent, tcm->tcm_handle,
1278
				 tca, &err);
1279
	}
L
Linus Torvalds 已提交
1280 1281 1282 1283 1284 1285 1286
	if (q == NULL) {
		if (err == -EAGAIN)
			goto replay;
		return err;
	}

graft:
1287 1288 1289 1290 1291
	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
1292
	}
1293

L
Linus Torvalds 已提交
1294 1295 1296 1297
	return 0;
}

static int tc_fill_qdisc(struct sk_buff *skb, struct Qdisc *q, u32 clid,
1298
			 u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1299 1300 1301
{
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1302
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1303
	struct gnet_dump d;
E
Eric Dumazet 已提交
1304
	struct qdisc_size_table *stab;
L
Linus Torvalds 已提交
1305

1306
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1307 1308 1309
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1310
	tcm->tcm_family = AF_UNSPEC;
1311 1312
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1313
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1314 1315 1316
	tcm->tcm_parent = clid;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = atomic_read(&q->refcnt);
1317 1318
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1319
	if (q->ops->dump && q->ops->dump(q, skb) < 0)
1320
		goto nla_put_failure;
L
Linus Torvalds 已提交
1321 1322
	q->qstats.qlen = q->q.qlen;

E
Eric Dumazet 已提交
1323 1324
	stab = rtnl_dereference(q->stab);
	if (stab && qdisc_dump_stab(skb, stab) < 0)
1325 1326
		goto nla_put_failure;

1327 1328
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
					 qdisc_root_sleeping_lock(q), &d) < 0)
1329
		goto nla_put_failure;
L
Linus Torvalds 已提交
1330 1331

	if (q->ops->dump_stats && q->ops->dump_stats(q, &d) < 0)
1332
		goto nla_put_failure;
L
Linus Torvalds 已提交
1333 1334

	if (gnet_stats_copy_basic(&d, &q->bstats) < 0 ||
1335
	    gnet_stats_copy_rate_est(&d, &q->bstats, &q->rate_est) < 0 ||
L
Linus Torvalds 已提交
1336
	    gnet_stats_copy_queue(&d, &q->qstats) < 0)
1337
		goto nla_put_failure;
1338

L
Linus Torvalds 已提交
1339
	if (gnet_stats_finish_copy(&d) < 0)
1340
		goto nla_put_failure;
1341

1342
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1343 1344
	return skb->len;

1345
out_nlmsg_trim:
1346
nla_put_failure:
1347
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1348 1349 1350
	return -1;
}

E
Eric Dumazet 已提交
1351 1352 1353 1354 1355
static bool tc_qdisc_dump_ignore(struct Qdisc *q)
{
	return (q->flags & TCQ_F_BUILTIN) ? true : false;
}

1356 1357 1358
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
1359 1360
{
	struct sk_buff *skb;
1361
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1362 1363 1364 1365 1366

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

E
Eric Dumazet 已提交
1367
	if (old && !tc_qdisc_dump_ignore(old)) {
1368
		if (tc_fill_qdisc(skb, old, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1369
				  0, RTM_DELQDISC) < 0)
L
Linus Torvalds 已提交
1370 1371
			goto err_out;
	}
E
Eric Dumazet 已提交
1372
	if (new && !tc_qdisc_dump_ignore(new)) {
1373
		if (tc_fill_qdisc(skb, new, clid, portid, n->nlmsg_seq,
E
Eric Dumazet 已提交
1374
				  old ? NLM_F_REPLACE : 0, RTM_NEWQDISC) < 0)
L
Linus Torvalds 已提交
1375 1376 1377 1378
			goto err_out;
	}

	if (skb->len)
1379
		return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1380
				      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1381 1382 1383 1384 1385 1386

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

1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
static int tc_dump_qdisc_root(struct Qdisc *root, struct sk_buff *skb,
			      struct netlink_callback *cb,
			      int *q_idx_p, int s_q_idx)
{
	int ret = 0, q_idx = *q_idx_p;
	struct Qdisc *q;

	if (!root)
		return 0;

	q = root;
	if (q_idx < s_q_idx) {
		q_idx++;
	} else {
		if (!tc_qdisc_dump_ignore(q) &&
1402
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1403 1404 1405 1406 1407 1408 1409 1410 1411
				  cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0)
			goto done;
		q_idx++;
	}
	list_for_each_entry(q, &root->list, list) {
		if (q_idx < s_q_idx) {
			q_idx++;
			continue;
		}
E
Eric Dumazet 已提交
1412
		if (!tc_qdisc_dump_ignore(q) &&
1413
		    tc_fill_qdisc(skb, q, q->parent, NETLINK_CB(cb->skb).portid,
1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426
				  cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWQDISC) <= 0)
			goto done;
		q_idx++;
	}

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

L
Linus Torvalds 已提交
1427 1428
static int tc_dump_qdisc(struct sk_buff *skb, struct netlink_callback *cb)
{
1429
	struct net *net = sock_net(skb->sk);
L
Linus Torvalds 已提交
1430 1431 1432 1433 1434 1435
	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];
1436 1437

	rcu_read_lock();
1438
	idx = 0;
1439
	for_each_netdev_rcu(net, dev) {
1440 1441
		struct netdev_queue *dev_queue;

L
Linus Torvalds 已提交
1442
		if (idx < s_idx)
1443
			goto cont;
L
Linus Torvalds 已提交
1444 1445 1446
		if (idx > s_idx)
			s_q_idx = 0;
		q_idx = 0;
1447

1448
		if (tc_dump_qdisc_root(dev->qdisc, skb, cb, &q_idx, s_q_idx) < 0)
1449 1450
			goto done;

1451 1452 1453 1454
		dev_queue = dev_ingress_queue(dev);
		if (dev_queue &&
		    tc_dump_qdisc_root(dev_queue->qdisc_sleeping, skb, cb,
				       &q_idx, s_q_idx) < 0)
1455 1456
			goto done;

1457 1458
cont:
		idx++;
L
Linus Torvalds 已提交
1459 1460 1461
	}

done:
1462
	rcu_read_unlock();
L
Linus Torvalds 已提交
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477

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

	return skb->len;
}



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



1478
static int tc_ctl_tclass(struct sk_buff *skb, struct nlmsghdr *n)
L
Linus Torvalds 已提交
1479
{
1480
	struct net *net = sock_net(skb->sk);
1481
	struct tcmsg *tcm = nlmsg_data(n);
1482
	struct nlattr *tca[TCA_MAX + 1];
L
Linus Torvalds 已提交
1483 1484
	struct net_device *dev;
	struct Qdisc *q = NULL;
1485
	const struct Qdisc_class_ops *cops;
L
Linus Torvalds 已提交
1486 1487
	unsigned long cl = 0;
	unsigned long new_cl;
1488 1489 1490
	u32 portid;
	u32 clid;
	u32 qid;
L
Linus Torvalds 已提交
1491 1492
	int err;

1493 1494 1495
	if ((n->nlmsg_type != RTM_GETTCLASS) && !capable(CAP_NET_ADMIN))
		return -EPERM;

1496 1497 1498 1499
	err = nlmsg_parse(n, sizeof(*tcm), tca, TCA_MAX, NULL);
	if (err < 0)
		return err;

1500 1501 1502 1503
	dev = __dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
		return -ENODEV;

L
Linus Torvalds 已提交
1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
	/*
	   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 */

1519 1520 1521 1522
	portid = tcm->tcm_parent;
	clid = tcm->tcm_handle;
	qid = TC_H_MAJ(clid);

1523 1524
	if (portid != TC_H_ROOT) {
		u32 qid1 = TC_H_MAJ(portid);
L
Linus Torvalds 已提交
1525 1526 1527 1528 1529 1530 1531 1532

		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)
1533
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1534 1535

		/* Now qid is genuine qdisc handle consistent
E
Eric Dumazet 已提交
1536 1537
		 * both with parent and child.
		 *
1538
		 * TC_H_MAJ(portid) still may be unspecified, complete it now.
L
Linus Torvalds 已提交
1539
		 */
1540 1541
		if (portid)
			portid = TC_H_MAKE(qid, portid);
L
Linus Torvalds 已提交
1542 1543
	} else {
		if (qid == 0)
1544
			qid = dev->qdisc->handle;
L
Linus Torvalds 已提交
1545 1546 1547
	}

	/* OK. Locate qdisc */
E
Eric Dumazet 已提交
1548 1549
	q = qdisc_lookup(dev, qid);
	if (!q)
L
Linus Torvalds 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558
		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) {
1559
		if (portid == TC_H_ROOT)
L
Linus Torvalds 已提交
1560 1561 1562 1563 1564 1565 1566 1567 1568
			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
1569 1570
		if (n->nlmsg_type != RTM_NEWTCLASS ||
		    !(n->nlmsg_flags & NLM_F_CREATE))
L
Linus Torvalds 已提交
1571 1572 1573
			goto out;
	} else {
		switch (n->nlmsg_type) {
1574
		case RTM_NEWTCLASS:
L
Linus Torvalds 已提交
1575
			err = -EEXIST;
E
Eric Dumazet 已提交
1576
			if (n->nlmsg_flags & NLM_F_EXCL)
L
Linus Torvalds 已提交
1577 1578 1579
				goto out;
			break;
		case RTM_DELTCLASS:
1580 1581 1582
			err = -EOPNOTSUPP;
			if (cops->delete)
				err = cops->delete(q, cl);
L
Linus Torvalds 已提交
1583
			if (err == 0)
1584
				tclass_notify(net, skb, n, q, cl, RTM_DELTCLASS);
L
Linus Torvalds 已提交
1585 1586
			goto out;
		case RTM_GETTCLASS:
1587
			err = tclass_notify(net, skb, n, q, cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1588 1589 1590 1591 1592 1593 1594 1595
			goto out;
		default:
			err = -EINVAL;
			goto out;
		}
	}

	new_cl = cl;
1596 1597
	err = -EOPNOTSUPP;
	if (cops->change)
1598
		err = cops->change(q, clid, portid, tca, &new_cl);
L
Linus Torvalds 已提交
1599
	if (err == 0)
1600
		tclass_notify(net, skb, n, q, new_cl, RTM_NEWTCLASS);
L
Linus Torvalds 已提交
1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611

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

	return err;
}


static int tc_fill_tclass(struct sk_buff *skb, struct Qdisc *q,
			  unsigned long cl,
1612
			  u32 portid, u32 seq, u16 flags, int event)
L
Linus Torvalds 已提交
1613 1614 1615
{
	struct tcmsg *tcm;
	struct nlmsghdr  *nlh;
1616
	unsigned char *b = skb_tail_pointer(skb);
L
Linus Torvalds 已提交
1617
	struct gnet_dump d;
1618
	const struct Qdisc_class_ops *cl_ops = q->ops->cl_ops;
L
Linus Torvalds 已提交
1619

1620
	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
1621 1622 1623
	if (!nlh)
		goto out_nlmsg_trim;
	tcm = nlmsg_data(nlh);
L
Linus Torvalds 已提交
1624
	tcm->tcm_family = AF_UNSPEC;
1625 1626
	tcm->tcm__pad1 = 0;
	tcm->tcm__pad2 = 0;
1627
	tcm->tcm_ifindex = qdisc_dev(q)->ifindex;
L
Linus Torvalds 已提交
1628 1629 1630
	tcm->tcm_parent = q->handle;
	tcm->tcm_handle = q->handle;
	tcm->tcm_info = 0;
1631 1632
	if (nla_put_string(skb, TCA_KIND, q->ops->id))
		goto nla_put_failure;
L
Linus Torvalds 已提交
1633
	if (cl_ops->dump && cl_ops->dump(q, cl, skb, tcm) < 0)
1634
		goto nla_put_failure;
L
Linus Torvalds 已提交
1635

1636 1637
	if (gnet_stats_start_copy_compat(skb, TCA_STATS2, TCA_STATS, TCA_XSTATS,
					 qdisc_root_sleeping_lock(q), &d) < 0)
1638
		goto nla_put_failure;
L
Linus Torvalds 已提交
1639 1640

	if (cl_ops->dump_stats && cl_ops->dump_stats(q, cl, &d) < 0)
1641
		goto nla_put_failure;
L
Linus Torvalds 已提交
1642 1643

	if (gnet_stats_finish_copy(&d) < 0)
1644
		goto nla_put_failure;
L
Linus Torvalds 已提交
1645

1646
	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
L
Linus Torvalds 已提交
1647 1648
	return skb->len;

1649
out_nlmsg_trim:
1650
nla_put_failure:
1651
	nlmsg_trim(skb, b);
L
Linus Torvalds 已提交
1652 1653 1654
	return -1;
}

1655 1656 1657
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
1658 1659
{
	struct sk_buff *skb;
1660
	u32 portid = oskb ? NETLINK_CB(oskb).portid : 0;
L
Linus Torvalds 已提交
1661 1662 1663 1664 1665

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

1666
	if (tc_fill_tclass(skb, q, cl, portid, n->nlmsg_seq, 0, event) < 0) {
L
Linus Torvalds 已提交
1667 1668 1669 1670
		kfree_skb(skb);
		return -EINVAL;
	}

1671
	return rtnetlink_send(skb, net, portid, RTNLGRP_TC,
E
Eric Dumazet 已提交
1672
			      n->nlmsg_flags & NLM_F_ECHO);
L
Linus Torvalds 已提交
1673 1674
}

E
Eric Dumazet 已提交
1675 1676 1677 1678
struct qdisc_dump_args {
	struct qdisc_walker	w;
	struct sk_buff		*skb;
	struct netlink_callback	*cb;
L
Linus Torvalds 已提交
1679 1680 1681 1682 1683 1684
};

static int qdisc_class_dump(struct Qdisc *q, unsigned long cl, struct qdisc_walker *arg)
{
	struct qdisc_dump_args *a = (struct qdisc_dump_args *)arg;

1685
	return tc_fill_tclass(a->skb, q, cl, NETLINK_CB(a->cb->skb).portid,
L
Linus Torvalds 已提交
1686 1687 1688
			      a->cb->nlh->nlmsg_seq, NLM_F_MULTI, RTM_NEWTCLASS);
}

1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
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;

	if (!root)
		return 0;

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

	list_for_each_entry(q, &root->list, list) {
		if (tc_dump_tclass_qdisc(q, skb, tcm, cb, t_p, s_t) < 0)
			return -1;
	}

	return 0;
}

L
Linus Torvalds 已提交
1738 1739
static int tc_dump_tclass(struct sk_buff *skb, struct netlink_callback *cb)
{
1740
	struct tcmsg *tcm = nlmsg_data(cb->nlh);
1741
	struct net *net = sock_net(skb->sk);
1742
	struct netdev_queue *dev_queue;
L
Linus Torvalds 已提交
1743
	struct net_device *dev;
1744
	int t, s_t;
L
Linus Torvalds 已提交
1745

1746
	if (nlmsg_len(cb->nlh) < sizeof(*tcm))
L
Linus Torvalds 已提交
1747
		return 0;
E
Eric Dumazet 已提交
1748 1749
	dev = dev_get_by_index(net, tcm->tcm_ifindex);
	if (!dev)
L
Linus Torvalds 已提交
1750 1751 1752 1753 1754
		return 0;

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

1755
	if (tc_dump_tclass_root(dev->qdisc, skb, tcm, cb, &t, s_t) < 0)
1756 1757
		goto done;

1758 1759 1760 1761
	dev_queue = dev_ingress_queue(dev);
	if (dev_queue &&
	    tc_dump_tclass_root(dev_queue->qdisc_sleeping, skb, tcm, cb,
				&t, s_t) < 0)
1762
		goto done;
L
Linus Torvalds 已提交
1763

1764
done:
L
Linus Torvalds 已提交
1765 1766 1767 1768 1769 1770 1771
	cb->args[0] = t;

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

/* Main classifier routine: scans classifier chain attached
E
Eric Dumazet 已提交
1772 1773
 * to this qdisc, (optionally) tests for protocol and asks
 * specific classifiers.
L
Linus Torvalds 已提交
1774
 */
1775
int tc_classify_compat(struct sk_buff *skb, const struct tcf_proto *tp,
1776 1777 1778
		       struct tcf_result *res)
{
	__be16 protocol = skb->protocol;
E
Eric Dumazet 已提交
1779
	int err;
1780 1781

	for (; tp; tp = tp->next) {
E
Eric Dumazet 已提交
1782 1783 1784 1785 1786 1787
		if (tp->protocol != protocol &&
		    tp->protocol != htons(ETH_P_ALL))
			continue;
		err = tp->classify(skb, tp, res);

		if (err >= 0) {
1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798
#ifdef CONFIG_NET_CLS_ACT
			if (err != TC_ACT_RECLASSIFY && skb->tc_verd)
				skb->tc_verd = SET_TC_VERD(skb->tc_verd, 0);
#endif
			return err;
		}
	}
	return -1;
}
EXPORT_SYMBOL(tc_classify_compat);

1799
int tc_classify(struct sk_buff *skb, const struct tcf_proto *tp,
1800
		struct tcf_result *res)
L
Linus Torvalds 已提交
1801 1802 1803
{
	int err = 0;
#ifdef CONFIG_NET_CLS_ACT
1804
	const struct tcf_proto *otp = tp;
L
Linus Torvalds 已提交
1805
reclassify:
1806
#endif
L
Linus Torvalds 已提交
1807

1808
	err = tc_classify_compat(skb, tp, res);
L
Linus Torvalds 已提交
1809
#ifdef CONFIG_NET_CLS_ACT
1810 1811 1812 1813 1814
	if (err == TC_ACT_RECLASSIFY) {
		u32 verd = G_TC_VERD(skb->tc_verd);
		tp = otp;

		if (verd++ >= MAX_REC_LOOP) {
1815 1816 1817 1818
			net_notice_ratelimited("%s: packet reclassify loop rule prio %u protocol %02x\n",
					       tp->q->ops->id,
					       tp->prio & 0xffff,
					       ntohs(tp->protocol));
1819
			return TC_ACT_SHOT;
L
Linus Torvalds 已提交
1820
		}
1821 1822
		skb->tc_verd = SET_TC_VERD(skb->tc_verd, verd);
		goto reclassify;
L
Linus Torvalds 已提交
1823
	}
1824 1825
#endif
	return err;
L
Linus Torvalds 已提交
1826
}
1827
EXPORT_SYMBOL(tc_classify);
L
Linus Torvalds 已提交
1828

1829 1830 1831 1832 1833 1834 1835
void tcf_destroy(struct tcf_proto *tp)
{
	tp->ops->destroy(tp);
	module_put(tp->ops->owner);
	kfree(tp);
}

1836
void tcf_destroy_chain(struct tcf_proto **fl)
1837 1838 1839
{
	struct tcf_proto *tp;

1840 1841
	while ((tp = *fl) != NULL) {
		*fl = tp->next;
1842 1843 1844 1845 1846
		tcf_destroy(tp);
	}
}
EXPORT_SYMBOL(tcf_destroy_chain);

L
Linus Torvalds 已提交
1847 1848 1849
#ifdef CONFIG_PROC_FS
static int psched_show(struct seq_file *seq, void *v)
{
1850 1851 1852
	struct timespec ts;

	hrtimer_get_res(CLOCK_MONOTONIC, &ts);
L
Linus Torvalds 已提交
1853
	seq_printf(seq, "%08x %08x %08x %08x\n",
1854
		   (u32)NSEC_PER_USEC, (u32)PSCHED_TICKS2NS(1),
P
Patrick McHardy 已提交
1855
		   1000000,
1856
		   (u32)NSEC_PER_SEC/(u32)ktime_to_ns(timespec_to_ktime(ts)));
L
Linus Torvalds 已提交
1857 1858 1859 1860 1861 1862

	return 0;
}

static int psched_open(struct inode *inode, struct file *file)
{
1863
	return single_open(file, psched_show, NULL);
L
Linus Torvalds 已提交
1864 1865
}

1866
static const struct file_operations psched_fops = {
L
Linus Torvalds 已提交
1867 1868 1869 1870 1871
	.owner = THIS_MODULE,
	.open = psched_open,
	.read  = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
1872
};
1873 1874 1875 1876 1877

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

1878
	e = proc_create("psched", 0, net->proc_net, &psched_fops);
1879 1880 1881 1882 1883 1884 1885 1886
	if (e == NULL)
		return -ENOMEM;

	return 0;
}

static void __net_exit psched_net_exit(struct net *net)
{
1887
	remove_proc_entry("psched", net->proc_net);
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
}
#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
1898 1899
#endif

1900 1901 1902 1903 1904
static struct pernet_operations psched_net_ops = {
	.init = psched_net_init,
	.exit = psched_net_exit,
};

L
Linus Torvalds 已提交
1905 1906
static int __init pktsched_init(void)
{
1907 1908 1909 1910
	int err;

	err = register_pernet_subsys(&psched_net_ops);
	if (err) {
E
Eric Dumazet 已提交
1911
		pr_err("pktsched_init: "
1912 1913 1914 1915
		       "cannot initialize per netns operations\n");
		return err;
	}

1916
	register_qdisc(&pfifo_fast_ops);
L
Linus Torvalds 已提交
1917 1918
	register_qdisc(&pfifo_qdisc_ops);
	register_qdisc(&bfifo_qdisc_ops);
1919
	register_qdisc(&pfifo_head_drop_qdisc_ops);
1920
	register_qdisc(&mq_qdisc_ops);
L
Linus Torvalds 已提交
1921

1922 1923 1924 1925 1926 1927
	rtnl_register(PF_UNSPEC, RTM_NEWQDISC, tc_modify_qdisc, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELQDISC, tc_get_qdisc, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_GETQDISC, tc_get_qdisc, tc_dump_qdisc, NULL);
	rtnl_register(PF_UNSPEC, RTM_NEWTCLASS, tc_ctl_tclass, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELTCLASS, tc_ctl_tclass, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_GETTCLASS, tc_ctl_tclass, tc_dump_tclass, NULL);
1928

L
Linus Torvalds 已提交
1929 1930 1931 1932
	return 0;
}

subsys_initcall(pktsched_init);