sch_htb.c 41.5 KB
Newer Older
S
Stephen Hemminger 已提交
1
/*
L
Linus Torvalds 已提交
2 3 4 5 6 7 8 9 10 11 12 13
 * net/sched/sch_htb.c	Hierarchical token bucket, feed tree version
 *
 *		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:	Martin Devera, <devik@cdi.cz>
 *
 * Credits (in time order) for older HTB versions:
 *              Stef Coene <stef.coene@docum.org>
 *			HTB support at LARTC mailing list
14
 *		Ondrej Kraus, <krauso@barr.cz>
L
Linus Torvalds 已提交
15 16 17 18 19 20 21 22 23 24 25 26 27 28
 *			found missing INIT_QDISC(htb)
 *		Vladimir Smelhaus, Aamer Akhter, Bert Hubert
 *			helped a lot to locate nasty class stall bug
 *		Andi Kleen, Jamal Hadi, Bert Hubert
 *			code review and helpful comments on shaping
 *		Tomasz Wrona, <tw@eter.tym.pl>
 *			created test case so that I was able to fix nasty bug
 *		Wilfried Weissmann
 *			spotted bug in dequeue code and helped with fix
 *		Jiri Fojtasek
 *			fixed requeue routine
 *		and many others. thanks.
 */
#include <linux/module.h>
29
#include <linux/moduleparam.h>
L
Linus Torvalds 已提交
30 31 32 33 34 35 36
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <linux/list.h>
#include <linux/compiler.h>
37
#include <linux/rbtree.h>
38
#include <linux/workqueue.h>
39
#include <linux/slab.h>
40
#include <net/netlink.h>
L
Linus Torvalds 已提交
41 42 43 44 45 46
#include <net/pkt_sched.h>

/* HTB algorithm.
    Author: devik@cdi.cz
    ========================================================================
    HTB is like TBF with multiple classes. It is also similar to CBQ because
47
    it allows to assign priority to each class in hierarchy.
L
Linus Torvalds 已提交
48 49 50
    In fact it is another implementation of Floyd's formal sharing.

    Levels:
51
    Each class is assigned level. Leaf has ALWAYS level 0 and root
L
Linus Torvalds 已提交
52 53 54 55
    classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level
    one less than their parent.
*/

56
static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */
S
Stephen Hemminger 已提交
57
#define HTB_VER 0x30011		/* major must be matched with number suplied by TC as version */
L
Linus Torvalds 已提交
58 59 60 61 62

#if HTB_VER >> 16 != TC_HTB_PROTOVER
#error "Mismatched sch_htb.c and pkt_sch.h"
#endif

63 64 65 66
/* Module parameter and sysfs export */
module_param    (htb_hysteresis, int, 0640);
MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate");

L
Linus Torvalds 已提交
67 68
/* used internaly to keep status of single class */
enum htb_cmode {
S
Stephen Hemminger 已提交
69 70 71
	HTB_CANT_SEND,		/* class can't send and can't borrow */
	HTB_MAY_BORROW,		/* class can't send but may borrow */
	HTB_CAN_SEND		/* class can send */
L
Linus Torvalds 已提交
72 73 74
};

/* interior & leaf nodes; props specific to leaves are marked L: */
S
Stephen Hemminger 已提交
75
struct htb_class {
76
	struct Qdisc_class_common common;
S
Stephen Hemminger 已提交
77
	/* general class parameters */
78
	struct gnet_stats_basic_packed bstats;
S
Stephen Hemminger 已提交
79 80 81 82
	struct gnet_stats_queue qstats;
	struct gnet_stats_rate_est rate_est;
	struct tc_htb_xstats xstats;	/* our special stats */
	int refcnt;		/* usage count of this class */
L
Linus Torvalds 已提交
83

S
Stephen Hemminger 已提交
84 85
	/* topology */
	int level;		/* our level (see above) */
86
	unsigned int children;
S
Stephen Hemminger 已提交
87 88
	struct htb_class *parent;	/* parent class */

89 90 91
	int prio;		/* these two are used only by leaves... */
	int quantum;		/* but stored for parent-to-leaf return */

S
Stephen Hemminger 已提交
92 93 94 95 96 97 98 99 100 101
	union {
		struct htb_class_leaf {
			struct Qdisc *q;
			int deficit[TC_HTB_MAXDEPTH];
			struct list_head drop_list;
		} leaf;
		struct htb_class_inner {
			struct rb_root feed[TC_HTB_NUMPRIO];	/* feed trees */
			struct rb_node *ptr[TC_HTB_NUMPRIO];	/* current class ptr */
			/* When class changes from state 1->2 and disconnects from
E
Eric Dumazet 已提交
102 103 104 105
			 * parent's feed then we lost ptr value and start from the
			 * first child again. Here we store classid of the
			 * last valid ptr (used when ptr is NULL).
			 */
S
Stephen Hemminger 已提交
106 107 108 109 110
			u32 last_ptr_id[TC_HTB_NUMPRIO];
		} inner;
	} un;
	struct rb_node node[TC_HTB_NUMPRIO];	/* node for self or feed tree */
	struct rb_node pq_node;	/* node for event queue */
111
	psched_time_t pq_key;
S
Stephen Hemminger 已提交
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126

	int prio_activity;	/* for which prios are we active */
	enum htb_cmode cmode;	/* current mode of the class */

	/* class attached filters */
	struct tcf_proto *filter_list;
	int filter_cnt;

	/* token bucket parameters */
	struct qdisc_rate_table *rate;	/* rate table of the class itself */
	struct qdisc_rate_table *ceil;	/* ceiling rate (limits borrows too) */
	long buffer, cbuffer;	/* token bucket depth/rate */
	psched_tdiff_t mbuffer;	/* max wait time */
	long tokens, ctokens;	/* current number of tokens */
	psched_time_t t_c;	/* checkpoint time */
L
Linus Torvalds 已提交
127 128
};

S
Stephen Hemminger 已提交
129
struct htb_sched {
130
	struct Qdisc_class_hash clhash;
131
	struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */
S
Stephen Hemminger 已提交
132 133 134 135 136 137

	/* self list - roots of self generating tree */
	struct rb_root row[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
	int row_mask[TC_HTB_MAXDEPTH];
	struct rb_node *ptr[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
	u32 last_ptr_id[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
L
Linus Torvalds 已提交
138

S
Stephen Hemminger 已提交
139 140
	/* self wait list - roots of wait PQs per row */
	struct rb_root wait_pq[TC_HTB_MAXDEPTH];
L
Linus Torvalds 已提交
141

S
Stephen Hemminger 已提交
142
	/* time of nearest event per level (row) */
143
	psched_time_t near_ev_cache[TC_HTB_MAXDEPTH];
L
Linus Torvalds 已提交
144

S
Stephen Hemminger 已提交
145
	int defcls;		/* class where unclassified flows go to */
L
Linus Torvalds 已提交
146

S
Stephen Hemminger 已提交
147 148
	/* filters for qdisc itself */
	struct tcf_proto *filter_list;
L
Linus Torvalds 已提交
149

S
Stephen Hemminger 已提交
150 151
	int rate2quantum;	/* quant = rate / rate2quantum */
	psched_time_t now;	/* cached dequeue time */
152
	struct qdisc_watchdog watchdog;
L
Linus Torvalds 已提交
153

S
Stephen Hemminger 已提交
154 155 156 157 158
	/* non shaped skbs; let them go directly thru */
	struct sk_buff_head direct_queue;
	int direct_qlen;	/* max qlen of above */

	long direct_pkts;
159 160 161

#define HTB_WARN_TOOMANYEVENTS	0x1
	unsigned int warned;	/* only one warning */
162
	struct work_struct work;
L
Linus Torvalds 已提交
163 164 165
};

/* find class in global hash table using given handle */
S
Stephen Hemminger 已提交
166
static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch)
L
Linus Torvalds 已提交
167 168
{
	struct htb_sched *q = qdisc_priv(sch);
169
	struct Qdisc_class_common *clc;
170

171 172
	clc = qdisc_class_find(&q->clhash, handle);
	if (clc == NULL)
L
Linus Torvalds 已提交
173
		return NULL;
174
	return container_of(clc, struct htb_class, common);
L
Linus Torvalds 已提交
175 176 177 178 179 180 181 182 183 184
}

/**
 * htb_classify - classify a packet into class
 *
 * It returns NULL if the packet should be dropped or -1 if the packet
 * should be passed directly thru. In all other cases leaf class is returned.
 * We allow direct class selection by classid in priority. The we examine
 * filters in qdisc and in inner nodes (if higher filter points to the inner
 * node). If we end up with classid MAJOR:0 we enqueue the skb into special
185
 * internal fifo (direct). These packets then go directly thru. If we still
L
Lucas De Marchi 已提交
186
 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessful
L
Linus Torvalds 已提交
187 188
 * then finish and return direct queue.
 */
E
Eric Dumazet 已提交
189
#define HTB_DIRECT ((struct htb_class *)-1L)
L
Linus Torvalds 已提交
190

S
Stephen Hemminger 已提交
191 192
static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
				      int *qerr)
L
Linus Torvalds 已提交
193 194 195 196 197 198 199 200
{
	struct htb_sched *q = qdisc_priv(sch);
	struct htb_class *cl;
	struct tcf_result res;
	struct tcf_proto *tcf;
	int result;

	/* allow to select class by setting skb->priority to valid classid;
E
Eric Dumazet 已提交
201 202 203
	 * note that nfmark can be used too by attaching filter fw with no
	 * rules in it
	 */
L
Linus Torvalds 已提交
204
	if (skb->priority == sch->handle)
S
Stephen Hemminger 已提交
205
		return HTB_DIRECT;	/* X:0 (direct flow) selected */
E
Eric Dumazet 已提交
206 207
	cl = htb_find(skb->priority, sch);
	if (cl && cl->level == 0)
L
Linus Torvalds 已提交
208 209
		return cl;

210
	*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
L
Linus Torvalds 已提交
211 212 213 214 215
	tcf = q->filter_list;
	while (tcf && (result = tc_classify(skb, tcf, &res)) >= 0) {
#ifdef CONFIG_NET_CLS_ACT
		switch (result) {
		case TC_ACT_QUEUED:
S
Stephen Hemminger 已提交
216
		case TC_ACT_STOLEN:
217
			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
L
Linus Torvalds 已提交
218 219 220 221
		case TC_ACT_SHOT:
			return NULL;
		}
#endif
E
Eric Dumazet 已提交
222 223
		cl = (void *)res.class;
		if (!cl) {
L
Linus Torvalds 已提交
224
			if (res.classid == sch->handle)
S
Stephen Hemminger 已提交
225
				return HTB_DIRECT;	/* X:0 (direct flow) */
E
Eric Dumazet 已提交
226 227
			cl = htb_find(res.classid, sch);
			if (!cl)
S
Stephen Hemminger 已提交
228
				break;	/* filter selected invalid classid */
L
Linus Torvalds 已提交
229 230
		}
		if (!cl->level)
S
Stephen Hemminger 已提交
231
			return cl;	/* we hit leaf; return it */
L
Linus Torvalds 已提交
232 233 234 235 236

		/* we have got inner class; apply inner filter chain */
		tcf = cl->filter_list;
	}
	/* classification failed; try to use default class */
S
Stephen Hemminger 已提交
237
	cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch);
L
Linus Torvalds 已提交
238
	if (!cl || cl->level)
S
Stephen Hemminger 已提交
239
		return HTB_DIRECT;	/* bad default .. this is safe bet */
L
Linus Torvalds 已提交
240 241 242 243 244 245 246 247 248
	return cl;
}

/**
 * htb_add_to_id_tree - adds class to the round robin list
 *
 * Routine adds class to the list (actually tree) sorted by classid.
 * Make sure that class is not already on such list for given prio.
 */
S
Stephen Hemminger 已提交
249 250
static void htb_add_to_id_tree(struct rb_root *root,
			       struct htb_class *cl, int prio)
L
Linus Torvalds 已提交
251 252
{
	struct rb_node **p = &root->rb_node, *parent = NULL;
253

L
Linus Torvalds 已提交
254
	while (*p) {
S
Stephen Hemminger 已提交
255 256
		struct htb_class *c;
		parent = *p;
L
Linus Torvalds 已提交
257
		c = rb_entry(parent, struct htb_class, node[prio]);
258

259
		if (cl->common.classid > c->common.classid)
L
Linus Torvalds 已提交
260
			p = &parent->rb_right;
S
Stephen Hemminger 已提交
261
		else
L
Linus Torvalds 已提交
262 263 264 265 266 267 268 269 270 271 272 273 274
			p = &parent->rb_left;
	}
	rb_link_node(&cl->node[prio], parent, p);
	rb_insert_color(&cl->node[prio], root);
}

/**
 * htb_add_to_wait_tree - adds class to the event queue with delay
 *
 * The class is added to priority event queue to indicate that class will
 * change its mode in cl->pq_key microseconds. Make sure that class is not
 * already in the queue.
 */
S
Stephen Hemminger 已提交
275 276
static void htb_add_to_wait_tree(struct htb_sched *q,
				 struct htb_class *cl, long delay)
L
Linus Torvalds 已提交
277 278
{
	struct rb_node **p = &q->wait_pq[cl->level].rb_node, *parent = NULL;
279

280 281
	cl->pq_key = q->now + delay;
	if (cl->pq_key == q->now)
L
Linus Torvalds 已提交
282 283 284
		cl->pq_key++;

	/* update the nearest event cache */
285
	if (q->near_ev_cache[cl->level] > cl->pq_key)
L
Linus Torvalds 已提交
286
		q->near_ev_cache[cl->level] = cl->pq_key;
S
Stephen Hemminger 已提交
287

L
Linus Torvalds 已提交
288
	while (*p) {
S
Stephen Hemminger 已提交
289 290
		struct htb_class *c;
		parent = *p;
L
Linus Torvalds 已提交
291
		c = rb_entry(parent, struct htb_class, pq_node);
292
		if (cl->pq_key >= c->pq_key)
L
Linus Torvalds 已提交
293
			p = &parent->rb_right;
S
Stephen Hemminger 已提交
294
		else
L
Linus Torvalds 已提交
295 296 297 298 299 300 301 302 303 304 305 306
			p = &parent->rb_left;
	}
	rb_link_node(&cl->pq_node, parent, p);
	rb_insert_color(&cl->pq_node, &q->wait_pq[cl->level]);
}

/**
 * htb_next_rb_node - finds next node in binary tree
 *
 * When we are past last key we return NULL.
 * Average complexity is 2 steps per call.
 */
S
Stephen Hemminger 已提交
307
static inline void htb_next_rb_node(struct rb_node **n)
L
Linus Torvalds 已提交
308 309 310 311 312 313 314 315 316 317
{
	*n = rb_next(*n);
}

/**
 * htb_add_class_to_row - add class to its row
 *
 * The class is added to row at priorities marked in mask.
 * It does nothing if mask == 0.
 */
S
Stephen Hemminger 已提交
318 319
static inline void htb_add_class_to_row(struct htb_sched *q,
					struct htb_class *cl, int mask)
L
Linus Torvalds 已提交
320 321 322 323 324
{
	q->row_mask[cl->level] |= mask;
	while (mask) {
		int prio = ffz(~mask);
		mask &= ~(1 << prio);
S
Stephen Hemminger 已提交
325
		htb_add_to_id_tree(q->row[cl->level] + prio, cl, prio);
L
Linus Torvalds 已提交
326 327 328
	}
}

S
Stephen Hemminger 已提交
329 330 331
/* If this triggers, it is a bug in this code, but it need not be fatal */
static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root)
{
332
	if (RB_EMPTY_NODE(rb)) {
S
Stephen Hemminger 已提交
333 334 335 336 337 338 339 340
		WARN_ON(1);
	} else {
		rb_erase(rb, root);
		RB_CLEAR_NODE(rb);
	}
}


L
Linus Torvalds 已提交
341 342 343 344 345 346
/**
 * htb_remove_class_from_row - removes class from its row
 *
 * The class is removed from row at priorities marked in mask.
 * It does nothing if mask == 0.
 */
S
Stephen Hemminger 已提交
347 348
static inline void htb_remove_class_from_row(struct htb_sched *q,
						 struct htb_class *cl, int mask)
L
Linus Torvalds 已提交
349 350
{
	int m = 0;
351

L
Linus Torvalds 已提交
352 353
	while (mask) {
		int prio = ffz(~mask);
S
Stephen Hemminger 已提交
354

L
Linus Torvalds 已提交
355
		mask &= ~(1 << prio);
S
Stephen Hemminger 已提交
356 357
		if (q->ptr[cl->level][prio] == cl->node + prio)
			htb_next_rb_node(q->ptr[cl->level] + prio);
S
Stephen Hemminger 已提交
358 359

		htb_safe_rb_erase(cl->node + prio, q->row[cl->level] + prio);
S
Stephen Hemminger 已提交
360
		if (!q->row[cl->level][prio].rb_node)
L
Linus Torvalds 已提交
361 362 363 364 365 366 367 368 369
			m |= 1 << prio;
	}
	q->row_mask[cl->level] &= ~m;
}

/**
 * htb_activate_prios - creates active classe's feed chain
 *
 * The class is connected to ancestors and/or appropriate rows
370
 * for priorities it is participating on. cl->cmode must be new
L
Linus Torvalds 已提交
371 372
 * (activated) mode. It does nothing if cl->prio_activity == 0.
 */
S
Stephen Hemminger 已提交
373
static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl)
L
Linus Torvalds 已提交
374 375
{
	struct htb_class *p = cl->parent;
S
Stephen Hemminger 已提交
376
	long m, mask = cl->prio_activity;
L
Linus Torvalds 已提交
377 378

	while (cl->cmode == HTB_MAY_BORROW && p && mask) {
S
Stephen Hemminger 已提交
379 380
		m = mask;
		while (m) {
L
Linus Torvalds 已提交
381 382
			int prio = ffz(~m);
			m &= ~(1 << prio);
S
Stephen Hemminger 已提交
383

L
Linus Torvalds 已提交
384 385
			if (p->un.inner.feed[prio].rb_node)
				/* parent already has its feed in use so that
E
Eric Dumazet 已提交
386 387
				 * reset bit in mask as parent is already ok
				 */
L
Linus Torvalds 已提交
388
				mask &= ~(1 << prio);
S
Stephen Hemminger 已提交
389 390

			htb_add_to_id_tree(p->un.inner.feed + prio, cl, prio);
L
Linus Torvalds 已提交
391 392
		}
		p->prio_activity |= mask;
S
Stephen Hemminger 已提交
393 394
		cl = p;
		p = cl->parent;
395

L
Linus Torvalds 已提交
396 397
	}
	if (cl->cmode == HTB_CAN_SEND && mask)
S
Stephen Hemminger 已提交
398
		htb_add_class_to_row(q, cl, mask);
L
Linus Torvalds 已提交
399 400 401 402 403
}

/**
 * htb_deactivate_prios - remove class from feed chain
 *
404
 * cl->cmode must represent old mode (before deactivation). It does
L
Linus Torvalds 已提交
405 406 407 408 409 410
 * nothing if cl->prio_activity == 0. Class is removed from all feed
 * chains and rows.
 */
static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl)
{
	struct htb_class *p = cl->parent;
S
Stephen Hemminger 已提交
411
	long m, mask = cl->prio_activity;
L
Linus Torvalds 已提交
412 413

	while (cl->cmode == HTB_MAY_BORROW && p && mask) {
S
Stephen Hemminger 已提交
414 415
		m = mask;
		mask = 0;
L
Linus Torvalds 已提交
416 417 418
		while (m) {
			int prio = ffz(~m);
			m &= ~(1 << prio);
S
Stephen Hemminger 已提交
419 420

			if (p->un.inner.ptr[prio] == cl->node + prio) {
L
Linus Torvalds 已提交
421
				/* we are removing child which is pointed to from
E
Eric Dumazet 已提交
422 423 424
				 * parent feed - forget the pointer but remember
				 * classid
				 */
425
				p->un.inner.last_ptr_id[prio] = cl->common.classid;
L
Linus Torvalds 已提交
426 427
				p->un.inner.ptr[prio] = NULL;
			}
S
Stephen Hemminger 已提交
428

S
Stephen Hemminger 已提交
429
			htb_safe_rb_erase(cl->node + prio, p->un.inner.feed + prio);
S
Stephen Hemminger 已提交
430 431

			if (!p->un.inner.feed[prio].rb_node)
L
Linus Torvalds 已提交
432 433
				mask |= 1 << prio;
		}
434

L
Linus Torvalds 已提交
435
		p->prio_activity &= ~mask;
S
Stephen Hemminger 已提交
436 437
		cl = p;
		p = cl->parent;
438

L
Linus Torvalds 已提交
439
	}
S
Stephen Hemminger 已提交
440 441
	if (cl->cmode == HTB_CAN_SEND && mask)
		htb_remove_class_from_row(q, cl, mask);
L
Linus Torvalds 已提交
442 443
}

444 445
static inline long htb_lowater(const struct htb_class *cl)
{
446 447 448 449
	if (htb_hysteresis)
		return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0;
	else
		return 0;
450 451 452
}
static inline long htb_hiwater(const struct htb_class *cl)
{
453 454 455 456
	if (htb_hysteresis)
		return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0;
	else
		return 0;
457
}
458

459

L
Linus Torvalds 已提交
460 461 462 463 464
/**
 * htb_class_mode - computes and returns current class mode
 *
 * It computes cl's mode at time cl->t_c+diff and returns it. If mode
 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference
465
 * from now to time when cl will change its state.
L
Linus Torvalds 已提交
466
 * Also it is worth to note that class mode doesn't change simply
467
 * at cl->{c,}tokens == 0 but there can rather be hysteresis of
L
Linus Torvalds 已提交
468 469 470
 * 0 .. -cl->{c,}buffer range. It is meant to limit number of
 * mode transitions per time unit. The speed gain is about 1/6.
 */
S
Stephen Hemminger 已提交
471 472
static inline enum htb_cmode
htb_class_mode(struct htb_class *cl, long *diff)
L
Linus Torvalds 已提交
473
{
S
Stephen Hemminger 已提交
474
	long toks;
L
Linus Torvalds 已提交
475

S
Stephen Hemminger 已提交
476 477 478 479
	if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) {
		*diff = -toks;
		return HTB_CANT_SEND;
	}
480

S
Stephen Hemminger 已提交
481 482
	if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl))
		return HTB_CAN_SEND;
L
Linus Torvalds 已提交
483

S
Stephen Hemminger 已提交
484 485
	*diff = -toks;
	return HTB_MAY_BORROW;
L
Linus Torvalds 已提交
486 487 488 489 490 491 492 493 494 495 496
}

/**
 * htb_change_class_mode - changes classe's mode
 *
 * This should be the only way how to change classe's mode under normal
 * cirsumstances. Routine will update feed lists linkage, change mode
 * and add class to the wait event queue if appropriate. New mode should
 * be different from old one and cl->pq_key has to be valid if changing
 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree).
 */
S
Stephen Hemminger 已提交
497
static void
L
Linus Torvalds 已提交
498
htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, long *diff)
S
Stephen Hemminger 已提交
499 500
{
	enum htb_cmode new_mode = htb_class_mode(cl, diff);
L
Linus Torvalds 已提交
501 502

	if (new_mode == cl->cmode)
S
Stephen Hemminger 已提交
503 504 505 506 507
		return;

	if (cl->prio_activity) {	/* not necessary: speed optimization */
		if (cl->cmode != HTB_CANT_SEND)
			htb_deactivate_prios(q, cl);
L
Linus Torvalds 已提交
508
		cl->cmode = new_mode;
S
Stephen Hemminger 已提交
509 510 511
		if (new_mode != HTB_CANT_SEND)
			htb_activate_prios(q, cl);
	} else
L
Linus Torvalds 已提交
512 513 514 515
		cl->cmode = new_mode;
}

/**
516
 * htb_activate - inserts leaf cl into appropriate active feeds
L
Linus Torvalds 已提交
517 518 519 520 521
 *
 * Routine learns (new) priority of leaf and activates feed chain
 * for the prio. It can be called on already active leaf safely.
 * It also adds leaf into droplist.
 */
S
Stephen Hemminger 已提交
522
static inline void htb_activate(struct htb_sched *q, struct htb_class *cl)
L
Linus Torvalds 已提交
523
{
524
	WARN_ON(cl->level || !cl->un.leaf.q || !cl->un.leaf.q->q.qlen);
525

L
Linus Torvalds 已提交
526
	if (!cl->prio_activity) {
527
		cl->prio_activity = 1 << cl->prio;
S
Stephen Hemminger 已提交
528 529
		htb_activate_prios(q, cl);
		list_add_tail(&cl->un.leaf.drop_list,
530
			      q->drops + cl->prio);
L
Linus Torvalds 已提交
531 532 533 534
	}
}

/**
535
 * htb_deactivate - remove leaf cl from active feeds
L
Linus Torvalds 已提交
536 537 538 539
 *
 * Make sure that leaf is active. In the other words it can't be called
 * with non-active leaf. It also removes class from the drop list.
 */
S
Stephen Hemminger 已提交
540
static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
L
Linus Torvalds 已提交
541
{
542
	WARN_ON(!cl->prio_activity);
543

S
Stephen Hemminger 已提交
544
	htb_deactivate_prios(q, cl);
L
Linus Torvalds 已提交
545 546 547 548 549 550
	cl->prio_activity = 0;
	list_del_init(&cl->un.leaf.drop_list);
}

static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch)
{
551
	int uninitialized_var(ret);
S
Stephen Hemminger 已提交
552 553 554 555 556 557 558 559 560 561 562 563 564
	struct htb_sched *q = qdisc_priv(sch);
	struct htb_class *cl = htb_classify(skb, sch, &ret);

	if (cl == HTB_DIRECT) {
		/* enqueue to helper queue */
		if (q->direct_queue.qlen < q->direct_qlen) {
			__skb_queue_tail(&q->direct_queue, skb);
			q->direct_pkts++;
		} else {
			kfree_skb(skb);
			sch->qstats.drops++;
			return NET_XMIT_DROP;
		}
L
Linus Torvalds 已提交
565
#ifdef CONFIG_NET_CLS_ACT
S
Stephen Hemminger 已提交
566
	} else if (!cl) {
567
		if (ret & __NET_XMIT_BYPASS)
S
Stephen Hemminger 已提交
568 569 570
			sch->qstats.drops++;
		kfree_skb(skb);
		return ret;
L
Linus Torvalds 已提交
571
#endif
572 573 574 575 576
	} else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q)) != NET_XMIT_SUCCESS) {
		if (net_xmit_drop_count(ret)) {
			sch->qstats.drops++;
			cl->qstats.drops++;
		}
577
		return ret;
S
Stephen Hemminger 已提交
578
	} else {
579
		bstats_update(&cl->bstats, skb);
S
Stephen Hemminger 已提交
580 581 582 583 584
		htb_activate(q, cl);
	}

	sch->q.qlen++;
	return NET_XMIT_SUCCESS;
L
Linus Torvalds 已提交
585 586
}

587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, long diff)
{
	long toks = diff + cl->tokens;

	if (toks > cl->buffer)
		toks = cl->buffer;
	toks -= (long) qdisc_l2t(cl->rate, bytes);
	if (toks <= -cl->mbuffer)
		toks = 1 - cl->mbuffer;

	cl->tokens = toks;
}

static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, long diff)
{
	long toks = diff + cl->ctokens;

	if (toks > cl->cbuffer)
		toks = cl->cbuffer;
	toks -= (long) qdisc_l2t(cl->ceil, bytes);
	if (toks <= -cl->mbuffer)
		toks = 1 - cl->mbuffer;

	cl->ctokens = toks;
}

L
Linus Torvalds 已提交
613 614 615 616 617 618 619 620 621 622 623
/**
 * htb_charge_class - charges amount "bytes" to leaf and ancestors
 *
 * Routine assumes that packet "bytes" long was dequeued from leaf cl
 * borrowing from "level". It accounts bytes to ceil leaky bucket for
 * leaf and all ancestors and to rate bucket for ancestors at levels
 * "level" and higher. It also handles possible change of mode resulting
 * from the update. Note that mode can also increase here (MAY_BORROW to
 * CAN_SEND) because we can use more precise clock that event queue here.
 * In such case we remove class from event queue first.
 */
S
Stephen Hemminger 已提交
624
static void htb_charge_class(struct htb_sched *q, struct htb_class *cl,
625
			     int level, struct sk_buff *skb)
S
Stephen Hemminger 已提交
626
{
627
	int bytes = qdisc_pkt_len(skb);
L
Linus Torvalds 已提交
628
	enum htb_cmode old_mode;
629
	long diff;
L
Linus Torvalds 已提交
630 631

	while (cl) {
632
		diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer);
L
Linus Torvalds 已提交
633
		if (cl->level >= level) {
S
Stephen Hemminger 已提交
634 635
			if (cl->level == level)
				cl->xstats.lends++;
636
			htb_accnt_tokens(cl, bytes, diff);
L
Linus Torvalds 已提交
637 638
		} else {
			cl->xstats.borrows++;
S
Stephen Hemminger 已提交
639
			cl->tokens += diff;	/* we moved t_c; update tokens */
L
Linus Torvalds 已提交
640
		}
641
		htb_accnt_ctokens(cl, bytes, diff);
L
Linus Torvalds 已提交
642 643
		cl->t_c = q->now;

S
Stephen Hemminger 已提交
644 645 646
		old_mode = cl->cmode;
		diff = 0;
		htb_change_class_mode(q, cl, &diff);
L
Linus Torvalds 已提交
647 648
		if (old_mode != cl->cmode) {
			if (old_mode != HTB_CAN_SEND)
S
Stephen Hemminger 已提交
649
				htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level);
L
Linus Torvalds 已提交
650
			if (cl->cmode != HTB_CAN_SEND)
S
Stephen Hemminger 已提交
651
				htb_add_to_wait_tree(q, cl, diff);
L
Linus Torvalds 已提交
652 653
		}

654 655 656 657
		/* update basic stats except for leaves which are already updated */
		if (cl->level)
			bstats_update(&cl->bstats, skb);

L
Linus Torvalds 已提交
658 659 660 661 662 663 664
		cl = cl->parent;
	}
}

/**
 * htb_do_events - make mode changes to classes at the level
 *
665
 * Scans event queue for pending events and applies them. Returns time of
666
 * next pending event (0 for no event in pq, q->now for too many events).
667
 * Note: Applied are events whose have cl->pq_key <= q->now.
L
Linus Torvalds 已提交
668
 */
669 670
static psched_time_t htb_do_events(struct htb_sched *q, int level,
				   unsigned long start)
L
Linus Torvalds 已提交
671
{
672
	/* don't run for longer than 2 jiffies; 2 is used instead of
E
Eric Dumazet 已提交
673 674 675
	 * 1 to simplify things when jiffy is going to be incremented
	 * too soon
	 */
676
	unsigned long stop_at = start + 2;
677
	while (time_before(jiffies, stop_at)) {
L
Linus Torvalds 已提交
678 679
		struct htb_class *cl;
		long diff;
680 681
		struct rb_node *p = rb_first(&q->wait_pq[level]);

S
Stephen Hemminger 已提交
682 683
		if (!p)
			return 0;
L
Linus Torvalds 已提交
684 685

		cl = rb_entry(p, struct htb_class, pq_node);
686 687 688
		if (cl->pq_key > q->now)
			return cl->pq_key;

S
Stephen Hemminger 已提交
689
		htb_safe_rb_erase(p, q->wait_pq + level);
690
		diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer);
S
Stephen Hemminger 已提交
691
		htb_change_class_mode(q, cl, &diff);
L
Linus Torvalds 已提交
692
		if (cl->cmode != HTB_CAN_SEND)
S
Stephen Hemminger 已提交
693
			htb_add_to_wait_tree(q, cl, diff);
L
Linus Torvalds 已提交
694
	}
695 696

	/* too much load - let's continue after a break for scheduling */
697
	if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) {
E
Eric Dumazet 已提交
698
		pr_warning("htb: too many events!\n");
699 700
		q->warned |= HTB_WARN_TOOMANYEVENTS;
	}
701 702

	return q->now;
L
Linus Torvalds 已提交
703 704 705
}

/* Returns class->node+prio from id-tree where classe's id is >= id. NULL
E
Eric Dumazet 已提交
706 707
 * is no such one exists.
 */
S
Stephen Hemminger 已提交
708 709
static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n,
					      u32 id)
L
Linus Torvalds 已提交
710 711 712
{
	struct rb_node *r = NULL;
	while (n) {
S
Stephen Hemminger 已提交
713 714 715
		struct htb_class *cl =
		    rb_entry(n, struct htb_class, node[prio]);

716
		if (id > cl->common.classid) {
L
Linus Torvalds 已提交
717
			n = n->rb_right;
718
		} else if (id < cl->common.classid) {
L
Linus Torvalds 已提交
719 720
			r = n;
			n = n->rb_left;
721 722
		} else {
			return n;
L
Linus Torvalds 已提交
723 724 725 726 727 728 729 730 731 732
		}
	}
	return r;
}

/**
 * htb_lookup_leaf - returns next leaf class in DRR order
 *
 * Find leaf where current feed pointers points to.
 */
S
Stephen Hemminger 已提交
733 734
static struct htb_class *htb_lookup_leaf(struct rb_root *tree, int prio,
					 struct rb_node **pptr, u32 * pid)
L
Linus Torvalds 已提交
735 736 737 738 739 740
{
	int i;
	struct {
		struct rb_node *root;
		struct rb_node **pptr;
		u32 *pid;
S
Stephen Hemminger 已提交
741 742
	} stk[TC_HTB_MAXDEPTH], *sp = stk;

743
	BUG_ON(!tree->rb_node);
L
Linus Torvalds 已提交
744 745 746 747 748
	sp->root = tree->rb_node;
	sp->pptr = pptr;
	sp->pid = pid;

	for (i = 0; i < 65535; i++) {
S
Stephen Hemminger 已提交
749
		if (!*sp->pptr && *sp->pid) {
750
			/* ptr was invalidated but id is valid - try to recover
E
Eric Dumazet 已提交
751 752
			 * the original or next ptr
			 */
S
Stephen Hemminger 已提交
753 754
			*sp->pptr =
			    htb_id_find_next_upper(prio, sp->root, *sp->pid);
L
Linus Torvalds 已提交
755
		}
S
Stephen Hemminger 已提交
756
		*sp->pid = 0;	/* ptr is valid now so that remove this hint as it
E
Eric Dumazet 已提交
757 758
				 * can become out of date quickly
				 */
S
Stephen Hemminger 已提交
759
		if (!*sp->pptr) {	/* we are at right end; rewind & go up */
L
Linus Torvalds 已提交
760
			*sp->pptr = sp->root;
S
Stephen Hemminger 已提交
761
			while ((*sp->pptr)->rb_left)
L
Linus Torvalds 已提交
762 763 764
				*sp->pptr = (*sp->pptr)->rb_left;
			if (sp > stk) {
				sp--;
765 766
				if (!*sp->pptr) {
					WARN_ON(1);
S
Stephen Hemminger 已提交
767
					return NULL;
768
				}
S
Stephen Hemminger 已提交
769
				htb_next_rb_node(sp->pptr);
L
Linus Torvalds 已提交
770 771 772
			}
		} else {
			struct htb_class *cl;
S
Stephen Hemminger 已提交
773 774
			cl = rb_entry(*sp->pptr, struct htb_class, node[prio]);
			if (!cl->level)
L
Linus Torvalds 已提交
775 776
				return cl;
			(++sp)->root = cl->un.inner.feed[prio].rb_node;
S
Stephen Hemminger 已提交
777 778
			sp->pptr = cl->un.inner.ptr + prio;
			sp->pid = cl->un.inner.last_ptr_id + prio;
L
Linus Torvalds 已提交
779 780
		}
	}
781
	WARN_ON(1);
L
Linus Torvalds 已提交
782 783 784 785
	return NULL;
}

/* dequeues packet at given priority and level; call only if
E
Eric Dumazet 已提交
786 787
 * you are sure that there is active class at prio/level
 */
S
Stephen Hemminger 已提交
788 789
static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, int prio,
					int level)
L
Linus Torvalds 已提交
790 791
{
	struct sk_buff *skb = NULL;
S
Stephen Hemminger 已提交
792
	struct htb_class *cl, *start;
L
Linus Torvalds 已提交
793
	/* look initial class up in the row */
S
Stephen Hemminger 已提交
794 795 796 797
	start = cl = htb_lookup_leaf(q->row[level] + prio, prio,
				     q->ptr[level] + prio,
				     q->last_ptr_id[level] + prio);

L
Linus Torvalds 已提交
798 799
	do {
next:
800
		if (unlikely(!cl))
S
Stephen Hemminger 已提交
801
			return NULL;
L
Linus Torvalds 已提交
802 803

		/* class can be empty - it is unlikely but can be true if leaf
E
Eric Dumazet 已提交
804 805 806 807
		 * qdisc drops packets in enqueue routine or if someone used
		 * graft operation on the leaf since last dequeue;
		 * simply deactivate and skip such class
		 */
L
Linus Torvalds 已提交
808 809
		if (unlikely(cl->un.leaf.q->q.qlen == 0)) {
			struct htb_class *next;
S
Stephen Hemminger 已提交
810
			htb_deactivate(q, cl);
L
Linus Torvalds 已提交
811 812 813

			/* row/level might become empty */
			if ((q->row_mask[level] & (1 << prio)) == 0)
S
Stephen Hemminger 已提交
814
				return NULL;
L
Linus Torvalds 已提交
815

S
Stephen Hemminger 已提交
816 817 818 819 820
			next = htb_lookup_leaf(q->row[level] + prio,
					       prio, q->ptr[level] + prio,
					       q->last_ptr_id[level] + prio);

			if (cl == start)	/* fix start if we just deleted it */
L
Linus Torvalds 已提交
821 822 823 824
				start = next;
			cl = next;
			goto next;
		}
S
Stephen Hemminger 已提交
825 826 827

		skb = cl->un.leaf.q->dequeue(cl->un.leaf.q);
		if (likely(skb != NULL))
L
Linus Torvalds 已提交
828
			break;
829

830
		qdisc_warn_nonwc("htb", cl->un.leaf.q);
S
Stephen Hemminger 已提交
831 832 833 834 835
		htb_next_rb_node((level ? cl->parent->un.inner.ptr : q->
				  ptr[0]) + prio);
		cl = htb_lookup_leaf(q->row[level] + prio, prio,
				     q->ptr[level] + prio,
				     q->last_ptr_id[level] + prio);
L
Linus Torvalds 已提交
836 837 838 839

	} while (cl != start);

	if (likely(skb != NULL)) {
840 841
		cl->un.leaf.deficit[level] -= qdisc_pkt_len(skb);
		if (cl->un.leaf.deficit[level] < 0) {
842
			cl->un.leaf.deficit[level] += cl->quantum;
S
Stephen Hemminger 已提交
843 844
			htb_next_rb_node((level ? cl->parent->un.inner.ptr : q->
					  ptr[0]) + prio);
L
Linus Torvalds 已提交
845 846
		}
		/* this used to be after charge_class but this constelation
E
Eric Dumazet 已提交
847 848
		 * gives us slightly better performance
		 */
L
Linus Torvalds 已提交
849
		if (!cl->un.leaf.q->q.qlen)
S
Stephen Hemminger 已提交
850
			htb_deactivate(q, cl);
851
		htb_charge_class(q, cl, level, skb);
L
Linus Torvalds 已提交
852 853 854 855 856 857
	}
	return skb;
}

static struct sk_buff *htb_dequeue(struct Qdisc *sch)
{
858
	struct sk_buff *skb;
L
Linus Torvalds 已提交
859 860
	struct htb_sched *q = qdisc_priv(sch);
	int level;
861
	psched_time_t next_event;
862
	unsigned long start_at;
L
Linus Torvalds 已提交
863 864

	/* try to dequeue direct packets as high prio (!) to minimize cpu work */
S
Stephen Hemminger 已提交
865 866
	skb = __skb_dequeue(&q->direct_queue);
	if (skb != NULL) {
867 868
ok:
		qdisc_bstats_update(sch, skb);
869
		qdisc_unthrottled(sch);
L
Linus Torvalds 已提交
870 871 872 873
		sch->q.qlen--;
		return skb;
	}

S
Stephen Hemminger 已提交
874 875
	if (!sch->q.qlen)
		goto fin;
876
	q->now = psched_get_time();
877
	start_at = jiffies;
L
Linus Torvalds 已提交
878

879
	next_event = q->now + 5 * PSCHED_TICKS_PER_SEC;
880

L
Linus Torvalds 已提交
881 882 883
	for (level = 0; level < TC_HTB_MAXDEPTH; level++) {
		/* common case optimization - skip event handler quickly */
		int m;
884 885 886
		psched_time_t event;

		if (q->now >= q->near_ev_cache[level]) {
887
			event = htb_do_events(q, level, start_at);
888 889 890
			if (!event)
				event = q->now + PSCHED_TICKS_PER_SEC;
			q->near_ev_cache[level] = event;
L
Linus Torvalds 已提交
891
		} else
892 893
			event = q->near_ev_cache[level];

894
		if (next_event > event)
895
			next_event = event;
S
Stephen Hemminger 已提交
896

L
Linus Torvalds 已提交
897 898
		m = ~q->row_mask[level];
		while (m != (int)(-1)) {
S
Stephen Hemminger 已提交
899
			int prio = ffz(m);
E
Eric Dumazet 已提交
900

L
Linus Torvalds 已提交
901
			m |= 1 << prio;
S
Stephen Hemminger 已提交
902
			skb = htb_dequeue_tree(q, prio, level);
903 904
			if (likely(skb != NULL))
				goto ok;
L
Linus Torvalds 已提交
905 906
		}
	}
907
	sch->qstats.overlimits++;
908 909 910 911
	if (likely(next_event > q->now))
		qdisc_watchdog_schedule(&q->watchdog, next_event);
	else
		schedule_work(&q->work);
L
Linus Torvalds 已提交
912 913 914 915 916
fin:
	return skb;
}

/* try to drop from each class (by prio) until one succeed */
S
Stephen Hemminger 已提交
917
static unsigned int htb_drop(struct Qdisc *sch)
L
Linus Torvalds 已提交
918 919 920 921 922 923
{
	struct htb_sched *q = qdisc_priv(sch);
	int prio;

	for (prio = TC_HTB_NUMPRIO - 1; prio >= 0; prio--) {
		struct list_head *p;
S
Stephen Hemminger 已提交
924
		list_for_each(p, q->drops + prio) {
L
Linus Torvalds 已提交
925 926 927
			struct htb_class *cl = list_entry(p, struct htb_class,
							  un.leaf.drop_list);
			unsigned int len;
S
Stephen Hemminger 已提交
928 929
			if (cl->un.leaf.q->ops->drop &&
			    (len = cl->un.leaf.q->ops->drop(cl->un.leaf.q))) {
L
Linus Torvalds 已提交
930 931
				sch->q.qlen--;
				if (!cl->un.leaf.q->q.qlen)
S
Stephen Hemminger 已提交
932
					htb_deactivate(q, cl);
L
Linus Torvalds 已提交
933 934 935 936 937 938 939 940 941
				return len;
			}
		}
	}
	return 0;
}

/* reset all classes */
/* always caled under BH & queue lock */
S
Stephen Hemminger 已提交
942
static void htb_reset(struct Qdisc *sch)
L
Linus Torvalds 已提交
943 944
{
	struct htb_sched *q = qdisc_priv(sch);
945 946 947
	struct htb_class *cl;
	struct hlist_node *n;
	unsigned int i;
948

949 950
	for (i = 0; i < q->clhash.hashsize; i++) {
		hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) {
L
Linus Torvalds 已提交
951
			if (cl->level)
S
Stephen Hemminger 已提交
952
				memset(&cl->un.inner, 0, sizeof(cl->un.inner));
L
Linus Torvalds 已提交
953
			else {
S
Stephen Hemminger 已提交
954
				if (cl->un.leaf.q)
L
Linus Torvalds 已提交
955 956 957 958 959 960 961 962
					qdisc_reset(cl->un.leaf.q);
				INIT_LIST_HEAD(&cl->un.leaf.drop_list);
			}
			cl->prio_activity = 0;
			cl->cmode = HTB_CAN_SEND;

		}
	}
963
	qdisc_watchdog_cancel(&q->watchdog);
L
Linus Torvalds 已提交
964 965
	__skb_queue_purge(&q->direct_queue);
	sch->q.qlen = 0;
S
Stephen Hemminger 已提交
966 967 968 969
	memset(q->row, 0, sizeof(q->row));
	memset(q->row_mask, 0, sizeof(q->row_mask));
	memset(q->wait_pq, 0, sizeof(q->wait_pq));
	memset(q->ptr, 0, sizeof(q->ptr));
L
Linus Torvalds 已提交
970
	for (i = 0; i < TC_HTB_NUMPRIO; i++)
S
Stephen Hemminger 已提交
971
		INIT_LIST_HEAD(q->drops + i);
L
Linus Torvalds 已提交
972 973
}

974 975 976 977 978 979 980
static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = {
	[TCA_HTB_PARMS]	= { .len = sizeof(struct tc_htb_opt) },
	[TCA_HTB_INIT]	= { .len = sizeof(struct tc_htb_glob) },
	[TCA_HTB_CTAB]	= { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
	[TCA_HTB_RTAB]	= { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
};

981 982 983 984 985 986 987 988
static void htb_work_func(struct work_struct *work)
{
	struct htb_sched *q = container_of(work, struct htb_sched, work);
	struct Qdisc *sch = q->watchdog.qdisc;

	__netif_schedule(qdisc_root(sch));
}

989
static int htb_init(struct Qdisc *sch, struct nlattr *opt)
L
Linus Torvalds 已提交
990 991
{
	struct htb_sched *q = qdisc_priv(sch);
992
	struct nlattr *tb[TCA_HTB_INIT + 1];
L
Linus Torvalds 已提交
993
	struct tc_htb_glob *gopt;
994
	int err;
L
Linus Torvalds 已提交
995
	int i;
996 997 998 999

	if (!opt)
		return -EINVAL;

1000
	err = nla_parse_nested(tb, TCA_HTB_INIT, opt, htb_policy);
1001 1002 1003
	if (err < 0)
		return err;

1004
	if (tb[TCA_HTB_INIT] == NULL) {
E
Eric Dumazet 已提交
1005
		pr_err("HTB: hey probably you have bad tc tool ?\n");
L
Linus Torvalds 已提交
1006 1007
		return -EINVAL;
	}
1008
	gopt = nla_data(tb[TCA_HTB_INIT]);
L
Linus Torvalds 已提交
1009
	if (gopt->version != HTB_VER >> 16) {
E
Eric Dumazet 已提交
1010
		pr_err("HTB: need tc/htb version %d (minor is %d), you have %d\n",
S
Stephen Hemminger 已提交
1011
		       HTB_VER >> 16, HTB_VER & 0xffff, gopt->version);
L
Linus Torvalds 已提交
1012 1013 1014
		return -EINVAL;
	}

1015 1016 1017
	err = qdisc_class_hash_init(&q->clhash);
	if (err < 0)
		return err;
L
Linus Torvalds 已提交
1018
	for (i = 0; i < TC_HTB_NUMPRIO; i++)
S
Stephen Hemminger 已提交
1019
		INIT_LIST_HEAD(q->drops + i);
L
Linus Torvalds 已提交
1020

1021
	qdisc_watchdog_init(&q->watchdog, sch);
1022
	INIT_WORK(&q->work, htb_work_func);
L
Linus Torvalds 已提交
1023 1024
	skb_queue_head_init(&q->direct_queue);

1025
	q->direct_qlen = qdisc_dev(sch)->tx_queue_len;
S
Stephen Hemminger 已提交
1026
	if (q->direct_qlen < 2)	/* some devices have zero tx_queue_len */
L
Linus Torvalds 已提交
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
		q->direct_qlen = 2;

	if ((q->rate2quantum = gopt->rate2quantum) < 1)
		q->rate2quantum = 1;
	q->defcls = gopt->defcls;

	return 0;
}

static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
{
1038
	spinlock_t *root_lock = qdisc_root_sleeping_lock(sch);
L
Linus Torvalds 已提交
1039
	struct htb_sched *q = qdisc_priv(sch);
1040
	struct nlattr *nest;
L
Linus Torvalds 已提交
1041
	struct tc_htb_glob gopt;
1042

1043
	spin_lock_bh(root_lock);
L
Linus Torvalds 已提交
1044

1045
	gopt.direct_pkts = q->direct_pkts;
L
Linus Torvalds 已提交
1046 1047 1048
	gopt.version = HTB_VER;
	gopt.rate2quantum = q->rate2quantum;
	gopt.defcls = q->defcls;
1049
	gopt.debug = 0;
1050 1051 1052 1053

	nest = nla_nest_start(skb, TCA_OPTIONS);
	if (nest == NULL)
		goto nla_put_failure;
1054
	NLA_PUT(skb, TCA_HTB_INIT, sizeof(gopt), &gopt);
1055 1056
	nla_nest_end(skb, nest);

1057
	spin_unlock_bh(root_lock);
L
Linus Torvalds 已提交
1058
	return skb->len;
1059

1060
nla_put_failure:
1061
	spin_unlock_bh(root_lock);
1062
	nla_nest_cancel(skb, nest);
L
Linus Torvalds 已提交
1063 1064 1065 1066
	return -1;
}

static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
S
Stephen Hemminger 已提交
1067
			  struct sk_buff *skb, struct tcmsg *tcm)
L
Linus Torvalds 已提交
1068
{
S
Stephen Hemminger 已提交
1069
	struct htb_class *cl = (struct htb_class *)arg;
1070
	spinlock_t *root_lock = qdisc_root_sleeping_lock(sch);
1071
	struct nlattr *nest;
L
Linus Torvalds 已提交
1072 1073
	struct tc_htb_opt opt;

1074
	spin_lock_bh(root_lock);
1075 1076
	tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT;
	tcm->tcm_handle = cl->common.classid;
L
Linus Torvalds 已提交
1077 1078 1079
	if (!cl->level && cl->un.leaf.q)
		tcm->tcm_info = cl->un.leaf.q->handle;

1080 1081 1082
	nest = nla_nest_start(skb, TCA_OPTIONS);
	if (nest == NULL)
		goto nla_put_failure;
L
Linus Torvalds 已提交
1083

S
Stephen Hemminger 已提交
1084
	memset(&opt, 0, sizeof(opt));
L
Linus Torvalds 已提交
1085

S
Stephen Hemminger 已提交
1086 1087 1088 1089
	opt.rate = cl->rate->rate;
	opt.buffer = cl->buffer;
	opt.ceil = cl->ceil->rate;
	opt.cbuffer = cl->cbuffer;
1090 1091
	opt.quantum = cl->quantum;
	opt.prio = cl->prio;
S
Stephen Hemminger 已提交
1092
	opt.level = cl->level;
1093
	NLA_PUT(skb, TCA_HTB_PARMS, sizeof(opt), &opt);
1094 1095

	nla_nest_end(skb, nest);
1096
	spin_unlock_bh(root_lock);
L
Linus Torvalds 已提交
1097
	return skb->len;
1098

1099
nla_put_failure:
1100
	spin_unlock_bh(root_lock);
1101
	nla_nest_cancel(skb, nest);
L
Linus Torvalds 已提交
1102 1103 1104 1105
	return -1;
}

static int
S
Stephen Hemminger 已提交
1106
htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
L
Linus Torvalds 已提交
1107
{
S
Stephen Hemminger 已提交
1108
	struct htb_class *cl = (struct htb_class *)arg;
L
Linus Torvalds 已提交
1109 1110 1111 1112 1113 1114 1115

	if (!cl->level && cl->un.leaf.q)
		cl->qstats.qlen = cl->un.leaf.q->q.qlen;
	cl->xstats.tokens = cl->tokens;
	cl->xstats.ctokens = cl->ctokens;

	if (gnet_stats_copy_basic(d, &cl->bstats) < 0 ||
1116
	    gnet_stats_copy_rate_est(d, NULL, &cl->rate_est) < 0 ||
L
Linus Torvalds 已提交
1117 1118 1119 1120 1121 1122 1123
	    gnet_stats_copy_queue(d, &cl->qstats) < 0)
		return -1;

	return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
}

static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
S
Stephen Hemminger 已提交
1124
		     struct Qdisc **old)
L
Linus Torvalds 已提交
1125
{
S
Stephen Hemminger 已提交
1126
	struct htb_class *cl = (struct htb_class *)arg;
L
Linus Torvalds 已提交
1127

1128 1129 1130
	if (cl->level)
		return -EINVAL;
	if (new == NULL &&
1131
	    (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1132 1133 1134 1135 1136 1137 1138 1139 1140
				     cl->common.classid)) == NULL)
		return -ENOBUFS;

	sch_tree_lock(sch);
	*old = cl->un.leaf.q;
	cl->un.leaf.q = new;
	if (*old != NULL) {
		qdisc_tree_decrease_qlen(*old, (*old)->q.qlen);
		qdisc_reset(*old);
L
Linus Torvalds 已提交
1141
	}
1142 1143
	sch_tree_unlock(sch);
	return 0;
L
Linus Torvalds 已提交
1144 1145
}

S
Stephen Hemminger 已提交
1146
static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg)
L
Linus Torvalds 已提交
1147
{
S
Stephen Hemminger 已提交
1148
	struct htb_class *cl = (struct htb_class *)arg;
1149
	return !cl->level ? cl->un.leaf.q : NULL;
L
Linus Torvalds 已提交
1150 1151
}

1152 1153 1154 1155 1156 1157 1158 1159
static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
{
	struct htb_class *cl = (struct htb_class *)arg;

	if (cl->un.leaf.q->q.qlen == 0)
		htb_deactivate(qdisc_priv(sch), cl);
}

L
Linus Torvalds 已提交
1160 1161
static unsigned long htb_get(struct Qdisc *sch, u32 classid)
{
S
Stephen Hemminger 已提交
1162 1163
	struct htb_class *cl = htb_find(classid, sch);
	if (cl)
L
Linus Torvalds 已提交
1164 1165 1166 1167
		cl->refcnt++;
	return (unsigned long)cl;
}

1168 1169 1170 1171 1172
static inline int htb_parent_last_child(struct htb_class *cl)
{
	if (!cl->parent)
		/* the root class */
		return 0;
1173
	if (cl->parent->children > 1)
1174 1175 1176 1177 1178
		/* not the last child */
		return 0;
	return 1;
}

1179 1180
static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl,
			       struct Qdisc *new_q)
1181 1182 1183
{
	struct htb_class *parent = cl->parent;

1184
	WARN_ON(cl->level || !cl->un.leaf.q || cl->prio_activity);
1185

1186 1187 1188
	if (parent->cmode != HTB_CAN_SEND)
		htb_safe_rb_erase(&parent->pq_node, q->wait_pq + parent->level);

1189 1190 1191 1192 1193 1194
	parent->level = 0;
	memset(&parent->un.inner, 0, sizeof(parent->un.inner));
	INIT_LIST_HEAD(&parent->un.leaf.drop_list);
	parent->un.leaf.q = new_q ? new_q : &noop_qdisc;
	parent->tokens = parent->buffer;
	parent->ctokens = parent->cbuffer;
1195
	parent->t_c = psched_get_time();
1196 1197 1198
	parent->cmode = HTB_CAN_SEND;
}

S
Stephen Hemminger 已提交
1199
static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl)
L
Linus Torvalds 已提交
1200 1201
{
	if (!cl->level) {
1202
		WARN_ON(!cl->un.leaf.q);
L
Linus Torvalds 已提交
1203 1204
		qdisc_destroy(cl->un.leaf.q);
	}
1205
	gen_kill_estimator(&cl->bstats, &cl->rate_est);
L
Linus Torvalds 已提交
1206 1207
	qdisc_put_rtab(cl->rate);
	qdisc_put_rtab(cl->ceil);
S
Stephen Hemminger 已提交
1208

1209
	tcf_destroy_chain(&cl->filter_list);
L
Linus Torvalds 已提交
1210 1211 1212
	kfree(cl);
}

S
Stephen Hemminger 已提交
1213
static void htb_destroy(struct Qdisc *sch)
L
Linus Torvalds 已提交
1214 1215
{
	struct htb_sched *q = qdisc_priv(sch);
1216 1217 1218
	struct hlist_node *n, *next;
	struct htb_class *cl;
	unsigned int i;
L
Linus Torvalds 已提交
1219

1220
	cancel_work_sync(&q->work);
1221
	qdisc_watchdog_cancel(&q->watchdog);
L
Linus Torvalds 已提交
1222
	/* This line used to be after htb_destroy_class call below
E
Eric Dumazet 已提交
1223 1224 1225 1226
	 * and surprisingly it worked in 2.4. But it must precede it
	 * because filter need its target class alive to be able to call
	 * unbind_filter on it (without Oops).
	 */
1227
	tcf_destroy_chain(&q->filter_list);
S
Stephen Hemminger 已提交
1228

1229 1230
	for (i = 0; i < q->clhash.hashsize; i++) {
		hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode)
1231 1232
			tcf_destroy_chain(&cl->filter_list);
	}
1233 1234 1235
	for (i = 0; i < q->clhash.hashsize; i++) {
		hlist_for_each_entry_safe(cl, n, next, &q->clhash.hash[i],
					  common.hnode)
1236 1237
			htb_destroy_class(sch, cl);
	}
1238
	qdisc_class_hash_destroy(&q->clhash);
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244
	__skb_queue_purge(&q->direct_queue);
}

static int htb_delete(struct Qdisc *sch, unsigned long arg)
{
	struct htb_sched *q = qdisc_priv(sch);
S
Stephen Hemminger 已提交
1245
	struct htb_class *cl = (struct htb_class *)arg;
1246
	unsigned int qlen;
1247 1248
	struct Qdisc *new_q = NULL;
	int last_child = 0;
L
Linus Torvalds 已提交
1249 1250 1251 1252

	// TODO: why don't allow to delete subtree ? references ? does
	// tc subsys quarantee us that in htb_destroy it holds no class
	// refs so that we can remove children safely there ?
1253
	if (cl->children || cl->filter_cnt)
L
Linus Torvalds 已提交
1254
		return -EBUSY;
S
Stephen Hemminger 已提交
1255

1256
	if (!cl->level && htb_parent_last_child(cl)) {
1257
		new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1258
					  cl->parent->common.classid);
1259 1260 1261
		last_child = 1;
	}

L
Linus Torvalds 已提交
1262
	sch_tree_lock(sch);
S
Stephen Hemminger 已提交
1263

1264
	if (!cl->level) {
1265
		qlen = cl->un.leaf.q->q.qlen;
1266
		qdisc_reset(cl->un.leaf.q);
1267
		qdisc_tree_decrease_qlen(cl->un.leaf.q, qlen);
1268 1269
	}

1270 1271
	/* delete from hash and active; remainder in destroy_class */
	qdisc_class_hash_remove(&q->clhash, &cl->common);
1272 1273
	if (cl->parent)
		cl->parent->children--;
1274

L
Linus Torvalds 已提交
1275
	if (cl->prio_activity)
S
Stephen Hemminger 已提交
1276
		htb_deactivate(q, cl);
L
Linus Torvalds 已提交
1277

1278 1279 1280
	if (cl->cmode != HTB_CAN_SEND)
		htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level);

1281
	if (last_child)
1282
		htb_parent_to_leaf(q, cl, new_q);
1283

1284 1285 1286 1287 1288
	BUG_ON(--cl->refcnt == 0);
	/*
	 * This shouldn't happen: we "hold" one cops->get() when called
	 * from tc_ctl_tclass; the destroy method is done from cops->put().
	 */
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294 1295

	sch_tree_unlock(sch);
	return 0;
}

static void htb_put(struct Qdisc *sch, unsigned long arg)
{
S
Stephen Hemminger 已提交
1296
	struct htb_class *cl = (struct htb_class *)arg;
L
Linus Torvalds 已提交
1297 1298

	if (--cl->refcnt == 0)
S
Stephen Hemminger 已提交
1299
		htb_destroy_class(sch, cl);
L
Linus Torvalds 已提交
1300 1301
}

S
Stephen Hemminger 已提交
1302
static int htb_change_class(struct Qdisc *sch, u32 classid,
1303
			    u32 parentid, struct nlattr **tca,
S
Stephen Hemminger 已提交
1304
			    unsigned long *arg)
L
Linus Torvalds 已提交
1305 1306 1307
{
	int err = -EINVAL;
	struct htb_sched *q = qdisc_priv(sch);
S
Stephen Hemminger 已提交
1308
	struct htb_class *cl = (struct htb_class *)*arg, *parent;
1309
	struct nlattr *opt = tca[TCA_OPTIONS];
L
Linus Torvalds 已提交
1310
	struct qdisc_rate_table *rtab = NULL, *ctab = NULL;
1311
	struct nlattr *tb[__TCA_HTB_MAX];
L
Linus Torvalds 已提交
1312 1313 1314
	struct tc_htb_opt *hopt;

	/* extract all subattrs from opt attr */
1315 1316 1317
	if (!opt)
		goto failure;

1318
	err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy);
1319 1320 1321 1322
	if (err < 0)
		goto failure;

	err = -EINVAL;
1323
	if (tb[TCA_HTB_PARMS] == NULL)
L
Linus Torvalds 已提交
1324 1325
		goto failure;

S
Stephen Hemminger 已提交
1326 1327
	parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch);

1328
	hopt = nla_data(tb[TCA_HTB_PARMS]);
1329

1330 1331
	rtab = qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB]);
	ctab = qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB]);
S
Stephen Hemminger 已提交
1332 1333
	if (!rtab || !ctab)
		goto failure;
L
Linus Torvalds 已提交
1334

S
Stephen Hemminger 已提交
1335
	if (!cl) {		/* new class */
L
Linus Torvalds 已提交
1336
		struct Qdisc *new_q;
S
Stephen Hemminger 已提交
1337
		int prio;
1338
		struct {
1339
			struct nlattr		nla;
1340 1341
			struct gnet_estimator	opt;
		} est = {
1342 1343 1344
			.nla = {
				.nla_len	= nla_attr_size(sizeof(est.opt)),
				.nla_type	= TCA_RATE,
1345 1346 1347 1348 1349 1350 1351
			},
			.opt = {
				/* 4s interval, 16s averaging constant */
				.interval	= 2,
				.ewma_log	= 2,
			},
		};
S
Stephen Hemminger 已提交
1352

L
Linus Torvalds 已提交
1353
		/* check for valid classid */
1354 1355
		if (!classid || TC_H_MAJ(classid ^ sch->handle) ||
		    htb_find(classid, sch))
L
Linus Torvalds 已提交
1356 1357 1358 1359
			goto failure;

		/* check maximal depth */
		if (parent && parent->parent && parent->parent->level < 2) {
E
Eric Dumazet 已提交
1360
			pr_err("htb: tree is too deep\n");
L
Linus Torvalds 已提交
1361 1362 1363
			goto failure;
		}
		err = -ENOBUFS;
E
Eric Dumazet 已提交
1364 1365
		cl = kzalloc(sizeof(*cl), GFP_KERNEL);
		if (!cl)
L
Linus Torvalds 已提交
1366
			goto failure;
S
Stephen Hemminger 已提交
1367

1368 1369 1370 1371 1372 1373 1374 1375
		err = gen_new_estimator(&cl->bstats, &cl->rate_est,
					qdisc_root_sleeping_lock(sch),
					tca[TCA_RATE] ? : &est.nla);
		if (err) {
			kfree(cl);
			goto failure;
		}

L
Linus Torvalds 已提交
1376
		cl->refcnt = 1;
1377
		cl->children = 0;
L
Linus Torvalds 已提交
1378
		INIT_LIST_HEAD(&cl->un.leaf.drop_list);
S
Stephen Hemminger 已提交
1379 1380 1381 1382
		RB_CLEAR_NODE(&cl->pq_node);

		for (prio = 0; prio < TC_HTB_NUMPRIO; prio++)
			RB_CLEAR_NODE(&cl->node[prio]);
L
Linus Torvalds 已提交
1383 1384

		/* create leaf qdisc early because it uses kmalloc(GFP_KERNEL)
E
Eric Dumazet 已提交
1385 1386 1387
		 * so that can't be used inside of sch_tree_lock
		 * -- thanks to Karlis Peisenieks
		 */
1388
		new_q = qdisc_create_dflt(sch->dev_queue,
1389
					  &pfifo_qdisc_ops, classid);
L
Linus Torvalds 已提交
1390 1391
		sch_tree_lock(sch);
		if (parent && !parent->level) {
1392 1393
			unsigned int qlen = parent->un.leaf.q->q.qlen;

L
Linus Torvalds 已提交
1394
			/* turn parent into inner node */
1395 1396
			qdisc_reset(parent->un.leaf.q);
			qdisc_tree_decrease_qlen(parent->un.leaf.q, qlen);
S
Stephen Hemminger 已提交
1397 1398 1399
			qdisc_destroy(parent->un.leaf.q);
			if (parent->prio_activity)
				htb_deactivate(q, parent);
L
Linus Torvalds 已提交
1400 1401 1402

			/* remove from evt list because of level change */
			if (parent->cmode != HTB_CAN_SEND) {
S
Stephen Hemminger 已提交
1403
				htb_safe_rb_erase(&parent->pq_node, q->wait_pq);
L
Linus Torvalds 已提交
1404 1405 1406
				parent->cmode = HTB_CAN_SEND;
			}
			parent->level = (parent->parent ? parent->parent->level
S
Stephen Hemminger 已提交
1407 1408
					 : TC_HTB_MAXDEPTH) - 1;
			memset(&parent->un.inner, 0, sizeof(parent->un.inner));
L
Linus Torvalds 已提交
1409 1410 1411 1412
		}
		/* leaf (we) needs elementary qdisc */
		cl->un.leaf.q = new_q ? new_q : &noop_qdisc;

1413
		cl->common.classid = classid;
S
Stephen Hemminger 已提交
1414
		cl->parent = parent;
L
Linus Torvalds 已提交
1415 1416 1417 1418

		/* set class to be in HTB_CAN_SEND state */
		cl->tokens = hopt->buffer;
		cl->ctokens = hopt->cbuffer;
1419
		cl->mbuffer = 60 * PSCHED_TICKS_PER_SEC;	/* 1min */
1420
		cl->t_c = psched_get_time();
L
Linus Torvalds 已提交
1421 1422 1423
		cl->cmode = HTB_CAN_SEND;

		/* attach to the hash list and parent's family */
1424
		qdisc_class_hash_insert(&q->clhash, &cl->common);
1425 1426
		if (parent)
			parent->children++;
1427
	} else {
1428 1429 1430 1431 1432 1433 1434
		if (tca[TCA_RATE]) {
			err = gen_replace_estimator(&cl->bstats, &cl->rate_est,
						    qdisc_root_sleeping_lock(sch),
						    tca[TCA_RATE]);
			if (err)
				return err;
		}
S
Stephen Hemminger 已提交
1435
		sch_tree_lock(sch);
1436
	}
L
Linus Torvalds 已提交
1437 1438

	/* it used to be a nasty bug here, we have to check that node
E
Eric Dumazet 已提交
1439 1440
	 * is really leaf before changing cl->un.leaf !
	 */
L
Linus Torvalds 已提交
1441
	if (!cl->level) {
1442 1443
		cl->quantum = rtab->rate.rate / q->rate2quantum;
		if (!hopt->quantum && cl->quantum < 1000) {
E
Eric Dumazet 已提交
1444
			pr_warning(
S
Stephen Hemminger 已提交
1445
			       "HTB: quantum of class %X is small. Consider r2q change.\n",
1446
			       cl->common.classid);
1447
			cl->quantum = 1000;
L
Linus Torvalds 已提交
1448
		}
1449
		if (!hopt->quantum && cl->quantum > 200000) {
E
Eric Dumazet 已提交
1450
			pr_warning(
S
Stephen Hemminger 已提交
1451
			       "HTB: quantum of class %X is big. Consider r2q change.\n",
1452
			       cl->common.classid);
1453
			cl->quantum = 200000;
L
Linus Torvalds 已提交
1454 1455
		}
		if (hopt->quantum)
1456 1457 1458
			cl->quantum = hopt->quantum;
		if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO)
			cl->prio = TC_HTB_NUMPRIO - 1;
L
Linus Torvalds 已提交
1459 1460 1461 1462
	}

	cl->buffer = hopt->buffer;
	cl->cbuffer = hopt->cbuffer;
S
Stephen Hemminger 已提交
1463 1464 1465 1466 1467 1468
	if (cl->rate)
		qdisc_put_rtab(cl->rate);
	cl->rate = rtab;
	if (cl->ceil)
		qdisc_put_rtab(cl->ceil);
	cl->ceil = ctab;
L
Linus Torvalds 已提交
1469 1470
	sch_tree_unlock(sch);

1471 1472
	qdisc_class_hash_grow(sch, &q->clhash);

L
Linus Torvalds 已提交
1473 1474 1475 1476
	*arg = (unsigned long)cl;
	return 0;

failure:
S
Stephen Hemminger 已提交
1477 1478 1479 1480
	if (rtab)
		qdisc_put_rtab(rtab);
	if (ctab)
		qdisc_put_rtab(ctab);
L
Linus Torvalds 已提交
1481 1482 1483 1484 1485 1486 1487 1488
	return err;
}

static struct tcf_proto **htb_find_tcf(struct Qdisc *sch, unsigned long arg)
{
	struct htb_sched *q = qdisc_priv(sch);
	struct htb_class *cl = (struct htb_class *)arg;
	struct tcf_proto **fl = cl ? &cl->filter_list : &q->filter_list;
1489

L
Linus Torvalds 已提交
1490 1491 1492 1493
	return fl;
}

static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent,
S
Stephen Hemminger 已提交
1494
				     u32 classid)
L
Linus Torvalds 已提交
1495
{
S
Stephen Hemminger 已提交
1496
	struct htb_class *cl = htb_find(classid, sch);
1497

L
Linus Torvalds 已提交
1498
	/*if (cl && !cl->level) return 0;
E
Eric Dumazet 已提交
1499 1500 1501 1502 1503 1504 1505
	 * The line above used to be there to prevent attaching filters to
	 * leaves. But at least tc_index filter uses this just to get class
	 * for other reasons so that we have to allow for it.
	 * ----
	 * 19.6.2002 As Werner explained it is ok - bind filter is just
	 * another way to "lock" the class - unlike "get" this lock can
	 * be broken by class during destroy IIUC.
L
Linus Torvalds 已提交
1506
	 */
S
Stephen Hemminger 已提交
1507 1508
	if (cl)
		cl->filter_cnt++;
L
Linus Torvalds 已提交
1509 1510 1511 1512 1513 1514
	return (unsigned long)cl;
}

static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg)
{
	struct htb_class *cl = (struct htb_class *)arg;
1515

S
Stephen Hemminger 已提交
1516 1517
	if (cl)
		cl->filter_cnt--;
L
Linus Torvalds 已提交
1518 1519 1520 1521 1522
}

static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg)
{
	struct htb_sched *q = qdisc_priv(sch);
1523 1524 1525
	struct htb_class *cl;
	struct hlist_node *n;
	unsigned int i;
L
Linus Torvalds 已提交
1526 1527 1528 1529

	if (arg->stop)
		return;

1530 1531
	for (i = 0; i < q->clhash.hashsize; i++) {
		hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) {
L
Linus Torvalds 已提交
1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
			if (arg->count < arg->skip) {
				arg->count++;
				continue;
			}
			if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
				arg->stop = 1;
				return;
			}
			arg->count++;
		}
	}
}

1545
static const struct Qdisc_class_ops htb_class_ops = {
L
Linus Torvalds 已提交
1546 1547
	.graft		=	htb_graft,
	.leaf		=	htb_leaf,
1548
	.qlen_notify	=	htb_qlen_notify,
L
Linus Torvalds 已提交
1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
	.get		=	htb_get,
	.put		=	htb_put,
	.change		=	htb_change_class,
	.delete		=	htb_delete,
	.walk		=	htb_walk,
	.tcf_chain	=	htb_find_tcf,
	.bind_tcf	=	htb_bind_filter,
	.unbind_tcf	=	htb_unbind_filter,
	.dump		=	htb_dump_class,
	.dump_stats	=	htb_dump_class_stats,
};

1561
static struct Qdisc_ops htb_qdisc_ops __read_mostly = {
L
Linus Torvalds 已提交
1562 1563 1564 1565 1566
	.cl_ops		=	&htb_class_ops,
	.id		=	"htb",
	.priv_size	=	sizeof(struct htb_sched),
	.enqueue	=	htb_enqueue,
	.dequeue	=	htb_dequeue,
1567
	.peek		=	qdisc_peek_dequeued,
L
Linus Torvalds 已提交
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
	.drop		=	htb_drop,
	.init		=	htb_init,
	.reset		=	htb_reset,
	.destroy	=	htb_destroy,
	.dump		=	htb_dump,
	.owner		=	THIS_MODULE,
};

static int __init htb_module_init(void)
{
S
Stephen Hemminger 已提交
1578
	return register_qdisc(&htb_qdisc_ops);
L
Linus Torvalds 已提交
1579
}
S
Stephen Hemminger 已提交
1580
static void __exit htb_module_exit(void)
L
Linus Torvalds 已提交
1581
{
S
Stephen Hemminger 已提交
1582
	unregister_qdisc(&htb_qdisc_ops);
L
Linus Torvalds 已提交
1583
}
S
Stephen Hemminger 已提交
1584

L
Linus Torvalds 已提交
1585 1586 1587
module_init(htb_module_init)
module_exit(htb_module_exit)
MODULE_LICENSE("GPL");