xref: /OK3568_Linux_fs/kernel/net/sched/sch_htb.c (revision 4882a59341e53eb6f0b4789bf948001014eff981)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * net/sched/sch_htb.c	Hierarchical token bucket, feed tree version
4  *
5  * Authors:	Martin Devera, <devik@cdi.cz>
6  *
7  * Credits (in time order) for older HTB versions:
8  *              Stef Coene <stef.coene@docum.org>
9  *			HTB support at LARTC mailing list
10  *		Ondrej Kraus, <krauso@barr.cz>
11  *			found missing INIT_QDISC(htb)
12  *		Vladimir Smelhaus, Aamer Akhter, Bert Hubert
13  *			helped a lot to locate nasty class stall bug
14  *		Andi Kleen, Jamal Hadi, Bert Hubert
15  *			code review and helpful comments on shaping
16  *		Tomasz Wrona, <tw@eter.tym.pl>
17  *			created test case so that I was able to fix nasty bug
18  *		Wilfried Weissmann
19  *			spotted bug in dequeue code and helped with fix
20  *		Jiri Fojtasek
21  *			fixed requeue routine
22  *		and many others. thanks.
23  */
24 #include <linux/module.h>
25 #include <linux/moduleparam.h>
26 #include <linux/types.h>
27 #include <linux/kernel.h>
28 #include <linux/string.h>
29 #include <linux/errno.h>
30 #include <linux/skbuff.h>
31 #include <linux/list.h>
32 #include <linux/compiler.h>
33 #include <linux/rbtree.h>
34 #include <linux/workqueue.h>
35 #include <linux/slab.h>
36 #include <net/netlink.h>
37 #include <net/sch_generic.h>
38 #include <net/pkt_sched.h>
39 #include <net/pkt_cls.h>
40 
41 /* HTB algorithm.
42     Author: devik@cdi.cz
43     ========================================================================
44     HTB is like TBF with multiple classes. It is also similar to CBQ because
45     it allows to assign priority to each class in hierarchy.
46     In fact it is another implementation of Floyd's formal sharing.
47 
48     Levels:
49     Each class is assigned level. Leaf has ALWAYS level 0 and root
50     classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level
51     one less than their parent.
52 */
53 
54 static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */
55 #define HTB_VER 0x30011		/* major must be matched with number suplied by TC as version */
56 
57 #if HTB_VER >> 16 != TC_HTB_PROTOVER
58 #error "Mismatched sch_htb.c and pkt_sch.h"
59 #endif
60 
61 /* Module parameter and sysfs export */
62 module_param    (htb_hysteresis, int, 0640);
63 MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate");
64 
65 static int htb_rate_est = 0; /* htb classes have a default rate estimator */
66 module_param(htb_rate_est, int, 0640);
67 MODULE_PARM_DESC(htb_rate_est, "setup a default rate estimator (4sec 16sec) for htb classes");
68 
69 /* used internaly to keep status of single class */
70 enum htb_cmode {
71 	HTB_CANT_SEND,		/* class can't send and can't borrow */
72 	HTB_MAY_BORROW,		/* class can't send but may borrow */
73 	HTB_CAN_SEND		/* class can send */
74 };
75 
76 struct htb_prio {
77 	union {
78 		struct rb_root	row;
79 		struct rb_root	feed;
80 	};
81 	struct rb_node	*ptr;
82 	/* When class changes from state 1->2 and disconnects from
83 	 * parent's feed then we lost ptr value and start from the
84 	 * first child again. Here we store classid of the
85 	 * last valid ptr (used when ptr is NULL).
86 	 */
87 	u32		last_ptr_id;
88 };
89 
90 /* interior & leaf nodes; props specific to leaves are marked L:
91  * To reduce false sharing, place mostly read fields at beginning,
92  * and mostly written ones at the end.
93  */
94 struct htb_class {
95 	struct Qdisc_class_common common;
96 	struct psched_ratecfg	rate;
97 	struct psched_ratecfg	ceil;
98 	s64			buffer, cbuffer;/* token bucket depth/rate */
99 	s64			mbuffer;	/* max wait time */
100 	u32			prio;		/* these two are used only by leaves... */
101 	int			quantum;	/* but stored for parent-to-leaf return */
102 
103 	struct tcf_proto __rcu	*filter_list;	/* class attached filters */
104 	struct tcf_block	*block;
105 	int			filter_cnt;
106 
107 	int			level;		/* our level (see above) */
108 	unsigned int		children;
109 	struct htb_class	*parent;	/* parent class */
110 
111 	struct net_rate_estimator __rcu *rate_est;
112 
113 	/*
114 	 * Written often fields
115 	 */
116 	struct gnet_stats_basic_packed bstats;
117 	struct tc_htb_xstats	xstats;	/* our special stats */
118 
119 	/* token bucket parameters */
120 	s64			tokens, ctokens;/* current number of tokens */
121 	s64			t_c;		/* checkpoint time */
122 
123 	union {
124 		struct htb_class_leaf {
125 			int		deficit[TC_HTB_MAXDEPTH];
126 			struct Qdisc	*q;
127 		} leaf;
128 		struct htb_class_inner {
129 			struct htb_prio clprio[TC_HTB_NUMPRIO];
130 		} inner;
131 	};
132 	s64			pq_key;
133 
134 	int			prio_activity;	/* for which prios are we active */
135 	enum htb_cmode		cmode;		/* current mode of the class */
136 	struct rb_node		pq_node;	/* node for event queue */
137 	struct rb_node		node[TC_HTB_NUMPRIO];	/* node for self or feed tree */
138 
139 	unsigned int drops ____cacheline_aligned_in_smp;
140 	unsigned int		overlimits;
141 };
142 
143 struct htb_level {
144 	struct rb_root	wait_pq;
145 	struct htb_prio hprio[TC_HTB_NUMPRIO];
146 };
147 
148 struct htb_sched {
149 	struct Qdisc_class_hash clhash;
150 	int			defcls;		/* class where unclassified flows go to */
151 	int			rate2quantum;	/* quant = rate / rate2quantum */
152 
153 	/* filters for qdisc itself */
154 	struct tcf_proto __rcu	*filter_list;
155 	struct tcf_block	*block;
156 
157 #define HTB_WARN_TOOMANYEVENTS	0x1
158 	unsigned int		warned;	/* only one warning */
159 	int			direct_qlen;
160 	struct work_struct	work;
161 
162 	/* non shaped skbs; let them go directly thru */
163 	struct qdisc_skb_head	direct_queue;
164 	u32			direct_pkts;
165 	u32			overlimits;
166 
167 	struct qdisc_watchdog	watchdog;
168 
169 	s64			now;	/* cached dequeue time */
170 
171 	/* time of nearest event per level (row) */
172 	s64			near_ev_cache[TC_HTB_MAXDEPTH];
173 
174 	int			row_mask[TC_HTB_MAXDEPTH];
175 
176 	struct htb_level	hlevel[TC_HTB_MAXDEPTH];
177 };
178 
179 /* find class in global hash table using given handle */
htb_find(u32 handle,struct Qdisc * sch)180 static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch)
181 {
182 	struct htb_sched *q = qdisc_priv(sch);
183 	struct Qdisc_class_common *clc;
184 
185 	clc = qdisc_class_find(&q->clhash, handle);
186 	if (clc == NULL)
187 		return NULL;
188 	return container_of(clc, struct htb_class, common);
189 }
190 
htb_search(struct Qdisc * sch,u32 handle)191 static unsigned long htb_search(struct Qdisc *sch, u32 handle)
192 {
193 	return (unsigned long)htb_find(handle, sch);
194 }
195 /**
196  * htb_classify - classify a packet into class
197  *
198  * It returns NULL if the packet should be dropped or -1 if the packet
199  * should be passed directly thru. In all other cases leaf class is returned.
200  * We allow direct class selection by classid in priority. The we examine
201  * filters in qdisc and in inner nodes (if higher filter points to the inner
202  * node). If we end up with classid MAJOR:0 we enqueue the skb into special
203  * internal fifo (direct). These packets then go directly thru. If we still
204  * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessful
205  * then finish and return direct queue.
206  */
207 #define HTB_DIRECT ((struct htb_class *)-1L)
208 
htb_classify(struct sk_buff * skb,struct Qdisc * sch,int * qerr)209 static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
210 				      int *qerr)
211 {
212 	struct htb_sched *q = qdisc_priv(sch);
213 	struct htb_class *cl;
214 	struct tcf_result res;
215 	struct tcf_proto *tcf;
216 	int result;
217 
218 	/* allow to select class by setting skb->priority to valid classid;
219 	 * note that nfmark can be used too by attaching filter fw with no
220 	 * rules in it
221 	 */
222 	if (skb->priority == sch->handle)
223 		return HTB_DIRECT;	/* X:0 (direct flow) selected */
224 	cl = htb_find(skb->priority, sch);
225 	if (cl) {
226 		if (cl->level == 0)
227 			return cl;
228 		/* Start with inner filter chain if a non-leaf class is selected */
229 		tcf = rcu_dereference_bh(cl->filter_list);
230 	} else {
231 		tcf = rcu_dereference_bh(q->filter_list);
232 	}
233 
234 	*qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
235 	while (tcf && (result = tcf_classify(skb, tcf, &res, false)) >= 0) {
236 #ifdef CONFIG_NET_CLS_ACT
237 		switch (result) {
238 		case TC_ACT_QUEUED:
239 		case TC_ACT_STOLEN:
240 		case TC_ACT_TRAP:
241 			*qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
242 			fallthrough;
243 		case TC_ACT_SHOT:
244 			return NULL;
245 		}
246 #endif
247 		cl = (void *)res.class;
248 		if (!cl) {
249 			if (res.classid == sch->handle)
250 				return HTB_DIRECT;	/* X:0 (direct flow) */
251 			cl = htb_find(res.classid, sch);
252 			if (!cl)
253 				break;	/* filter selected invalid classid */
254 		}
255 		if (!cl->level)
256 			return cl;	/* we hit leaf; return it */
257 
258 		/* we have got inner class; apply inner filter chain */
259 		tcf = rcu_dereference_bh(cl->filter_list);
260 	}
261 	/* classification failed; try to use default class */
262 	cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch);
263 	if (!cl || cl->level)
264 		return HTB_DIRECT;	/* bad default .. this is safe bet */
265 	return cl;
266 }
267 
268 /**
269  * htb_add_to_id_tree - adds class to the round robin list
270  *
271  * Routine adds class to the list (actually tree) sorted by classid.
272  * Make sure that class is not already on such list for given prio.
273  */
htb_add_to_id_tree(struct rb_root * root,struct htb_class * cl,int prio)274 static void htb_add_to_id_tree(struct rb_root *root,
275 			       struct htb_class *cl, int prio)
276 {
277 	struct rb_node **p = &root->rb_node, *parent = NULL;
278 
279 	while (*p) {
280 		struct htb_class *c;
281 		parent = *p;
282 		c = rb_entry(parent, struct htb_class, node[prio]);
283 
284 		if (cl->common.classid > c->common.classid)
285 			p = &parent->rb_right;
286 		else
287 			p = &parent->rb_left;
288 	}
289 	rb_link_node(&cl->node[prio], parent, p);
290 	rb_insert_color(&cl->node[prio], root);
291 }
292 
293 /**
294  * htb_add_to_wait_tree - adds class to the event queue with delay
295  *
296  * The class is added to priority event queue to indicate that class will
297  * change its mode in cl->pq_key microseconds. Make sure that class is not
298  * already in the queue.
299  */
htb_add_to_wait_tree(struct htb_sched * q,struct htb_class * cl,s64 delay)300 static void htb_add_to_wait_tree(struct htb_sched *q,
301 				 struct htb_class *cl, s64 delay)
302 {
303 	struct rb_node **p = &q->hlevel[cl->level].wait_pq.rb_node, *parent = NULL;
304 
305 	cl->pq_key = q->now + delay;
306 	if (cl->pq_key == q->now)
307 		cl->pq_key++;
308 
309 	/* update the nearest event cache */
310 	if (q->near_ev_cache[cl->level] > cl->pq_key)
311 		q->near_ev_cache[cl->level] = cl->pq_key;
312 
313 	while (*p) {
314 		struct htb_class *c;
315 		parent = *p;
316 		c = rb_entry(parent, struct htb_class, pq_node);
317 		if (cl->pq_key >= c->pq_key)
318 			p = &parent->rb_right;
319 		else
320 			p = &parent->rb_left;
321 	}
322 	rb_link_node(&cl->pq_node, parent, p);
323 	rb_insert_color(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
324 }
325 
326 /**
327  * htb_next_rb_node - finds next node in binary tree
328  *
329  * When we are past last key we return NULL.
330  * Average complexity is 2 steps per call.
331  */
htb_next_rb_node(struct rb_node ** n)332 static inline void htb_next_rb_node(struct rb_node **n)
333 {
334 	*n = rb_next(*n);
335 }
336 
337 /**
338  * htb_add_class_to_row - add class to its row
339  *
340  * The class is added to row at priorities marked in mask.
341  * It does nothing if mask == 0.
342  */
htb_add_class_to_row(struct htb_sched * q,struct htb_class * cl,int mask)343 static inline void htb_add_class_to_row(struct htb_sched *q,
344 					struct htb_class *cl, int mask)
345 {
346 	q->row_mask[cl->level] |= mask;
347 	while (mask) {
348 		int prio = ffz(~mask);
349 		mask &= ~(1 << prio);
350 		htb_add_to_id_tree(&q->hlevel[cl->level].hprio[prio].row, cl, prio);
351 	}
352 }
353 
354 /* If this triggers, it is a bug in this code, but it need not be fatal */
htb_safe_rb_erase(struct rb_node * rb,struct rb_root * root)355 static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root)
356 {
357 	if (RB_EMPTY_NODE(rb)) {
358 		WARN_ON(1);
359 	} else {
360 		rb_erase(rb, root);
361 		RB_CLEAR_NODE(rb);
362 	}
363 }
364 
365 
366 /**
367  * htb_remove_class_from_row - removes class from its row
368  *
369  * The class is removed from row at priorities marked in mask.
370  * It does nothing if mask == 0.
371  */
htb_remove_class_from_row(struct htb_sched * q,struct htb_class * cl,int mask)372 static inline void htb_remove_class_from_row(struct htb_sched *q,
373 						 struct htb_class *cl, int mask)
374 {
375 	int m = 0;
376 	struct htb_level *hlevel = &q->hlevel[cl->level];
377 
378 	while (mask) {
379 		int prio = ffz(~mask);
380 		struct htb_prio *hprio = &hlevel->hprio[prio];
381 
382 		mask &= ~(1 << prio);
383 		if (hprio->ptr == cl->node + prio)
384 			htb_next_rb_node(&hprio->ptr);
385 
386 		htb_safe_rb_erase(cl->node + prio, &hprio->row);
387 		if (!hprio->row.rb_node)
388 			m |= 1 << prio;
389 	}
390 	q->row_mask[cl->level] &= ~m;
391 }
392 
393 /**
394  * htb_activate_prios - creates active classe's feed chain
395  *
396  * The class is connected to ancestors and/or appropriate rows
397  * for priorities it is participating on. cl->cmode must be new
398  * (activated) mode. It does nothing if cl->prio_activity == 0.
399  */
htb_activate_prios(struct htb_sched * q,struct htb_class * cl)400 static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl)
401 {
402 	struct htb_class *p = cl->parent;
403 	long m, mask = cl->prio_activity;
404 
405 	while (cl->cmode == HTB_MAY_BORROW && p && mask) {
406 		m = mask;
407 		while (m) {
408 			int prio = ffz(~m);
409 			m &= ~(1 << prio);
410 
411 			if (p->inner.clprio[prio].feed.rb_node)
412 				/* parent already has its feed in use so that
413 				 * reset bit in mask as parent is already ok
414 				 */
415 				mask &= ~(1 << prio);
416 
417 			htb_add_to_id_tree(&p->inner.clprio[prio].feed, cl, prio);
418 		}
419 		p->prio_activity |= mask;
420 		cl = p;
421 		p = cl->parent;
422 
423 	}
424 	if (cl->cmode == HTB_CAN_SEND && mask)
425 		htb_add_class_to_row(q, cl, mask);
426 }
427 
428 /**
429  * htb_deactivate_prios - remove class from feed chain
430  *
431  * cl->cmode must represent old mode (before deactivation). It does
432  * nothing if cl->prio_activity == 0. Class is removed from all feed
433  * chains and rows.
434  */
htb_deactivate_prios(struct htb_sched * q,struct htb_class * cl)435 static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl)
436 {
437 	struct htb_class *p = cl->parent;
438 	long m, mask = cl->prio_activity;
439 
440 	while (cl->cmode == HTB_MAY_BORROW && p && mask) {
441 		m = mask;
442 		mask = 0;
443 		while (m) {
444 			int prio = ffz(~m);
445 			m &= ~(1 << prio);
446 
447 			if (p->inner.clprio[prio].ptr == cl->node + prio) {
448 				/* we are removing child which is pointed to from
449 				 * parent feed - forget the pointer but remember
450 				 * classid
451 				 */
452 				p->inner.clprio[prio].last_ptr_id = cl->common.classid;
453 				p->inner.clprio[prio].ptr = NULL;
454 			}
455 
456 			htb_safe_rb_erase(cl->node + prio,
457 					  &p->inner.clprio[prio].feed);
458 
459 			if (!p->inner.clprio[prio].feed.rb_node)
460 				mask |= 1 << prio;
461 		}
462 
463 		p->prio_activity &= ~mask;
464 		cl = p;
465 		p = cl->parent;
466 
467 	}
468 	if (cl->cmode == HTB_CAN_SEND && mask)
469 		htb_remove_class_from_row(q, cl, mask);
470 }
471 
htb_lowater(const struct htb_class * cl)472 static inline s64 htb_lowater(const struct htb_class *cl)
473 {
474 	if (htb_hysteresis)
475 		return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0;
476 	else
477 		return 0;
478 }
htb_hiwater(const struct htb_class * cl)479 static inline s64 htb_hiwater(const struct htb_class *cl)
480 {
481 	if (htb_hysteresis)
482 		return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0;
483 	else
484 		return 0;
485 }
486 
487 
488 /**
489  * htb_class_mode - computes and returns current class mode
490  *
491  * It computes cl's mode at time cl->t_c+diff and returns it. If mode
492  * is not HTB_CAN_SEND then cl->pq_key is updated to time difference
493  * from now to time when cl will change its state.
494  * Also it is worth to note that class mode doesn't change simply
495  * at cl->{c,}tokens == 0 but there can rather be hysteresis of
496  * 0 .. -cl->{c,}buffer range. It is meant to limit number of
497  * mode transitions per time unit. The speed gain is about 1/6.
498  */
499 static inline enum htb_cmode
htb_class_mode(struct htb_class * cl,s64 * diff)500 htb_class_mode(struct htb_class *cl, s64 *diff)
501 {
502 	s64 toks;
503 
504 	if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) {
505 		*diff = -toks;
506 		return HTB_CANT_SEND;
507 	}
508 
509 	if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl))
510 		return HTB_CAN_SEND;
511 
512 	*diff = -toks;
513 	return HTB_MAY_BORROW;
514 }
515 
516 /**
517  * htb_change_class_mode - changes classe's mode
518  *
519  * This should be the only way how to change classe's mode under normal
520  * cirsumstances. Routine will update feed lists linkage, change mode
521  * and add class to the wait event queue if appropriate. New mode should
522  * be different from old one and cl->pq_key has to be valid if changing
523  * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree).
524  */
525 static void
htb_change_class_mode(struct htb_sched * q,struct htb_class * cl,s64 * diff)526 htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, s64 *diff)
527 {
528 	enum htb_cmode new_mode = htb_class_mode(cl, diff);
529 
530 	if (new_mode == cl->cmode)
531 		return;
532 
533 	if (new_mode == HTB_CANT_SEND) {
534 		cl->overlimits++;
535 		q->overlimits++;
536 	}
537 
538 	if (cl->prio_activity) {	/* not necessary: speed optimization */
539 		if (cl->cmode != HTB_CANT_SEND)
540 			htb_deactivate_prios(q, cl);
541 		cl->cmode = new_mode;
542 		if (new_mode != HTB_CANT_SEND)
543 			htb_activate_prios(q, cl);
544 	} else
545 		cl->cmode = new_mode;
546 }
547 
548 /**
549  * htb_activate - inserts leaf cl into appropriate active feeds
550  *
551  * Routine learns (new) priority of leaf and activates feed chain
552  * for the prio. It can be called on already active leaf safely.
553  * It also adds leaf into droplist.
554  */
htb_activate(struct htb_sched * q,struct htb_class * cl)555 static inline void htb_activate(struct htb_sched *q, struct htb_class *cl)
556 {
557 	WARN_ON(cl->level || !cl->leaf.q || !cl->leaf.q->q.qlen);
558 
559 	if (!cl->prio_activity) {
560 		cl->prio_activity = 1 << cl->prio;
561 		htb_activate_prios(q, cl);
562 	}
563 }
564 
565 /**
566  * htb_deactivate - remove leaf cl from active feeds
567  *
568  * Make sure that leaf is active. In the other words it can't be called
569  * with non-active leaf. It also removes class from the drop list.
570  */
htb_deactivate(struct htb_sched * q,struct htb_class * cl)571 static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
572 {
573 	WARN_ON(!cl->prio_activity);
574 
575 	htb_deactivate_prios(q, cl);
576 	cl->prio_activity = 0;
577 }
578 
htb_enqueue(struct sk_buff * skb,struct Qdisc * sch,struct sk_buff ** to_free)579 static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
580 		       struct sk_buff **to_free)
581 {
582 	int ret;
583 	unsigned int len = qdisc_pkt_len(skb);
584 	struct htb_sched *q = qdisc_priv(sch);
585 	struct htb_class *cl = htb_classify(skb, sch, &ret);
586 
587 	if (cl == HTB_DIRECT) {
588 		/* enqueue to helper queue */
589 		if (q->direct_queue.qlen < q->direct_qlen) {
590 			__qdisc_enqueue_tail(skb, &q->direct_queue);
591 			q->direct_pkts++;
592 		} else {
593 			return qdisc_drop(skb, sch, to_free);
594 		}
595 #ifdef CONFIG_NET_CLS_ACT
596 	} else if (!cl) {
597 		if (ret & __NET_XMIT_BYPASS)
598 			qdisc_qstats_drop(sch);
599 		__qdisc_drop(skb, to_free);
600 		return ret;
601 #endif
602 	} else if ((ret = qdisc_enqueue(skb, cl->leaf.q,
603 					to_free)) != NET_XMIT_SUCCESS) {
604 		if (net_xmit_drop_count(ret)) {
605 			qdisc_qstats_drop(sch);
606 			cl->drops++;
607 		}
608 		return ret;
609 	} else {
610 		htb_activate(q, cl);
611 	}
612 
613 	sch->qstats.backlog += len;
614 	sch->q.qlen++;
615 	return NET_XMIT_SUCCESS;
616 }
617 
htb_accnt_tokens(struct htb_class * cl,int bytes,s64 diff)618 static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, s64 diff)
619 {
620 	s64 toks = diff + cl->tokens;
621 
622 	if (toks > cl->buffer)
623 		toks = cl->buffer;
624 	toks -= (s64) psched_l2t_ns(&cl->rate, bytes);
625 	if (toks <= -cl->mbuffer)
626 		toks = 1 - cl->mbuffer;
627 
628 	cl->tokens = toks;
629 }
630 
htb_accnt_ctokens(struct htb_class * cl,int bytes,s64 diff)631 static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, s64 diff)
632 {
633 	s64 toks = diff + cl->ctokens;
634 
635 	if (toks > cl->cbuffer)
636 		toks = cl->cbuffer;
637 	toks -= (s64) psched_l2t_ns(&cl->ceil, bytes);
638 	if (toks <= -cl->mbuffer)
639 		toks = 1 - cl->mbuffer;
640 
641 	cl->ctokens = toks;
642 }
643 
644 /**
645  * htb_charge_class - charges amount "bytes" to leaf and ancestors
646  *
647  * Routine assumes that packet "bytes" long was dequeued from leaf cl
648  * borrowing from "level". It accounts bytes to ceil leaky bucket for
649  * leaf and all ancestors and to rate bucket for ancestors at levels
650  * "level" and higher. It also handles possible change of mode resulting
651  * from the update. Note that mode can also increase here (MAY_BORROW to
652  * CAN_SEND) because we can use more precise clock that event queue here.
653  * In such case we remove class from event queue first.
654  */
htb_charge_class(struct htb_sched * q,struct htb_class * cl,int level,struct sk_buff * skb)655 static void htb_charge_class(struct htb_sched *q, struct htb_class *cl,
656 			     int level, struct sk_buff *skb)
657 {
658 	int bytes = qdisc_pkt_len(skb);
659 	enum htb_cmode old_mode;
660 	s64 diff;
661 
662 	while (cl) {
663 		diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
664 		if (cl->level >= level) {
665 			if (cl->level == level)
666 				cl->xstats.lends++;
667 			htb_accnt_tokens(cl, bytes, diff);
668 		} else {
669 			cl->xstats.borrows++;
670 			cl->tokens += diff;	/* we moved t_c; update tokens */
671 		}
672 		htb_accnt_ctokens(cl, bytes, diff);
673 		cl->t_c = q->now;
674 
675 		old_mode = cl->cmode;
676 		diff = 0;
677 		htb_change_class_mode(q, cl, &diff);
678 		if (old_mode != cl->cmode) {
679 			if (old_mode != HTB_CAN_SEND)
680 				htb_safe_rb_erase(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
681 			if (cl->cmode != HTB_CAN_SEND)
682 				htb_add_to_wait_tree(q, cl, diff);
683 		}
684 
685 		/* update basic stats except for leaves which are already updated */
686 		if (cl->level)
687 			bstats_update(&cl->bstats, skb);
688 
689 		cl = cl->parent;
690 	}
691 }
692 
693 /**
694  * htb_do_events - make mode changes to classes at the level
695  *
696  * Scans event queue for pending events and applies them. Returns time of
697  * next pending event (0 for no event in pq, q->now for too many events).
698  * Note: Applied are events whose have cl->pq_key <= q->now.
699  */
htb_do_events(struct htb_sched * q,const int level,unsigned long start)700 static s64 htb_do_events(struct htb_sched *q, const int level,
701 			 unsigned long start)
702 {
703 	/* don't run for longer than 2 jiffies; 2 is used instead of
704 	 * 1 to simplify things when jiffy is going to be incremented
705 	 * too soon
706 	 */
707 	unsigned long stop_at = start + 2;
708 	struct rb_root *wait_pq = &q->hlevel[level].wait_pq;
709 
710 	while (time_before(jiffies, stop_at)) {
711 		struct htb_class *cl;
712 		s64 diff;
713 		struct rb_node *p = rb_first(wait_pq);
714 
715 		if (!p)
716 			return 0;
717 
718 		cl = rb_entry(p, struct htb_class, pq_node);
719 		if (cl->pq_key > q->now)
720 			return cl->pq_key;
721 
722 		htb_safe_rb_erase(p, wait_pq);
723 		diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
724 		htb_change_class_mode(q, cl, &diff);
725 		if (cl->cmode != HTB_CAN_SEND)
726 			htb_add_to_wait_tree(q, cl, diff);
727 	}
728 
729 	/* too much load - let's continue after a break for scheduling */
730 	if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) {
731 		pr_warn("htb: too many events!\n");
732 		q->warned |= HTB_WARN_TOOMANYEVENTS;
733 	}
734 
735 	return q->now;
736 }
737 
738 /* Returns class->node+prio from id-tree where classe's id is >= id. NULL
739  * is no such one exists.
740  */
htb_id_find_next_upper(int prio,struct rb_node * n,u32 id)741 static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n,
742 					      u32 id)
743 {
744 	struct rb_node *r = NULL;
745 	while (n) {
746 		struct htb_class *cl =
747 		    rb_entry(n, struct htb_class, node[prio]);
748 
749 		if (id > cl->common.classid) {
750 			n = n->rb_right;
751 		} else if (id < cl->common.classid) {
752 			r = n;
753 			n = n->rb_left;
754 		} else {
755 			return n;
756 		}
757 	}
758 	return r;
759 }
760 
761 /**
762  * htb_lookup_leaf - returns next leaf class in DRR order
763  *
764  * Find leaf where current feed pointers points to.
765  */
htb_lookup_leaf(struct htb_prio * hprio,const int prio)766 static struct htb_class *htb_lookup_leaf(struct htb_prio *hprio, const int prio)
767 {
768 	int i;
769 	struct {
770 		struct rb_node *root;
771 		struct rb_node **pptr;
772 		u32 *pid;
773 	} stk[TC_HTB_MAXDEPTH], *sp = stk;
774 
775 	BUG_ON(!hprio->row.rb_node);
776 	sp->root = hprio->row.rb_node;
777 	sp->pptr = &hprio->ptr;
778 	sp->pid = &hprio->last_ptr_id;
779 
780 	for (i = 0; i < 65535; i++) {
781 		if (!*sp->pptr && *sp->pid) {
782 			/* ptr was invalidated but id is valid - try to recover
783 			 * the original or next ptr
784 			 */
785 			*sp->pptr =
786 			    htb_id_find_next_upper(prio, sp->root, *sp->pid);
787 		}
788 		*sp->pid = 0;	/* ptr is valid now so that remove this hint as it
789 				 * can become out of date quickly
790 				 */
791 		if (!*sp->pptr) {	/* we are at right end; rewind & go up */
792 			*sp->pptr = sp->root;
793 			while ((*sp->pptr)->rb_left)
794 				*sp->pptr = (*sp->pptr)->rb_left;
795 			if (sp > stk) {
796 				sp--;
797 				if (!*sp->pptr) {
798 					WARN_ON(1);
799 					return NULL;
800 				}
801 				htb_next_rb_node(sp->pptr);
802 			}
803 		} else {
804 			struct htb_class *cl;
805 			struct htb_prio *clp;
806 
807 			cl = rb_entry(*sp->pptr, struct htb_class, node[prio]);
808 			if (!cl->level)
809 				return cl;
810 			clp = &cl->inner.clprio[prio];
811 			(++sp)->root = clp->feed.rb_node;
812 			sp->pptr = &clp->ptr;
813 			sp->pid = &clp->last_ptr_id;
814 		}
815 	}
816 	WARN_ON(1);
817 	return NULL;
818 }
819 
820 /* dequeues packet at given priority and level; call only if
821  * you are sure that there is active class at prio/level
822  */
htb_dequeue_tree(struct htb_sched * q,const int prio,const int level)823 static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, const int prio,
824 					const int level)
825 {
826 	struct sk_buff *skb = NULL;
827 	struct htb_class *cl, *start;
828 	struct htb_level *hlevel = &q->hlevel[level];
829 	struct htb_prio *hprio = &hlevel->hprio[prio];
830 
831 	/* look initial class up in the row */
832 	start = cl = htb_lookup_leaf(hprio, prio);
833 
834 	do {
835 next:
836 		if (unlikely(!cl))
837 			return NULL;
838 
839 		/* class can be empty - it is unlikely but can be true if leaf
840 		 * qdisc drops packets in enqueue routine or if someone used
841 		 * graft operation on the leaf since last dequeue;
842 		 * simply deactivate and skip such class
843 		 */
844 		if (unlikely(cl->leaf.q->q.qlen == 0)) {
845 			struct htb_class *next;
846 			htb_deactivate(q, cl);
847 
848 			/* row/level might become empty */
849 			if ((q->row_mask[level] & (1 << prio)) == 0)
850 				return NULL;
851 
852 			next = htb_lookup_leaf(hprio, prio);
853 
854 			if (cl == start)	/* fix start if we just deleted it */
855 				start = next;
856 			cl = next;
857 			goto next;
858 		}
859 
860 		skb = cl->leaf.q->dequeue(cl->leaf.q);
861 		if (likely(skb != NULL))
862 			break;
863 
864 		qdisc_warn_nonwc("htb", cl->leaf.q);
865 		htb_next_rb_node(level ? &cl->parent->inner.clprio[prio].ptr:
866 					 &q->hlevel[0].hprio[prio].ptr);
867 		cl = htb_lookup_leaf(hprio, prio);
868 
869 	} while (cl != start);
870 
871 	if (likely(skb != NULL)) {
872 		bstats_update(&cl->bstats, skb);
873 		cl->leaf.deficit[level] -= qdisc_pkt_len(skb);
874 		if (cl->leaf.deficit[level] < 0) {
875 			cl->leaf.deficit[level] += cl->quantum;
876 			htb_next_rb_node(level ? &cl->parent->inner.clprio[prio].ptr :
877 						 &q->hlevel[0].hprio[prio].ptr);
878 		}
879 		/* this used to be after charge_class but this constelation
880 		 * gives us slightly better performance
881 		 */
882 		if (!cl->leaf.q->q.qlen)
883 			htb_deactivate(q, cl);
884 		htb_charge_class(q, cl, level, skb);
885 	}
886 	return skb;
887 }
888 
htb_dequeue(struct Qdisc * sch)889 static struct sk_buff *htb_dequeue(struct Qdisc *sch)
890 {
891 	struct sk_buff *skb;
892 	struct htb_sched *q = qdisc_priv(sch);
893 	int level;
894 	s64 next_event;
895 	unsigned long start_at;
896 
897 	/* try to dequeue direct packets as high prio (!) to minimize cpu work */
898 	skb = __qdisc_dequeue_head(&q->direct_queue);
899 	if (skb != NULL) {
900 ok:
901 		qdisc_bstats_update(sch, skb);
902 		qdisc_qstats_backlog_dec(sch, skb);
903 		sch->q.qlen--;
904 		return skb;
905 	}
906 
907 	if (!sch->q.qlen)
908 		goto fin;
909 	q->now = ktime_get_ns();
910 	start_at = jiffies;
911 
912 	next_event = q->now + 5LLU * NSEC_PER_SEC;
913 
914 	for (level = 0; level < TC_HTB_MAXDEPTH; level++) {
915 		/* common case optimization - skip event handler quickly */
916 		int m;
917 		s64 event = q->near_ev_cache[level];
918 
919 		if (q->now >= event) {
920 			event = htb_do_events(q, level, start_at);
921 			if (!event)
922 				event = q->now + NSEC_PER_SEC;
923 			q->near_ev_cache[level] = event;
924 		}
925 
926 		if (next_event > event)
927 			next_event = event;
928 
929 		m = ~q->row_mask[level];
930 		while (m != (int)(-1)) {
931 			int prio = ffz(m);
932 
933 			m |= 1 << prio;
934 			skb = htb_dequeue_tree(q, prio, level);
935 			if (likely(skb != NULL))
936 				goto ok;
937 		}
938 	}
939 	if (likely(next_event > q->now))
940 		qdisc_watchdog_schedule_ns(&q->watchdog, next_event);
941 	else
942 		schedule_work(&q->work);
943 fin:
944 	return skb;
945 }
946 
947 /* reset all classes */
948 /* always caled under BH & queue lock */
htb_reset(struct Qdisc * sch)949 static void htb_reset(struct Qdisc *sch)
950 {
951 	struct htb_sched *q = qdisc_priv(sch);
952 	struct htb_class *cl;
953 	unsigned int i;
954 
955 	for (i = 0; i < q->clhash.hashsize; i++) {
956 		hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
957 			if (cl->level)
958 				memset(&cl->inner, 0, sizeof(cl->inner));
959 			else {
960 				if (cl->leaf.q)
961 					qdisc_reset(cl->leaf.q);
962 			}
963 			cl->prio_activity = 0;
964 			cl->cmode = HTB_CAN_SEND;
965 		}
966 	}
967 	qdisc_watchdog_cancel(&q->watchdog);
968 	__qdisc_reset_queue(&q->direct_queue);
969 	memset(q->hlevel, 0, sizeof(q->hlevel));
970 	memset(q->row_mask, 0, sizeof(q->row_mask));
971 }
972 
973 static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = {
974 	[TCA_HTB_PARMS]	= { .len = sizeof(struct tc_htb_opt) },
975 	[TCA_HTB_INIT]	= { .len = sizeof(struct tc_htb_glob) },
976 	[TCA_HTB_CTAB]	= { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
977 	[TCA_HTB_RTAB]	= { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
978 	[TCA_HTB_DIRECT_QLEN] = { .type = NLA_U32 },
979 	[TCA_HTB_RATE64] = { .type = NLA_U64 },
980 	[TCA_HTB_CEIL64] = { .type = NLA_U64 },
981 };
982 
htb_work_func(struct work_struct * work)983 static void htb_work_func(struct work_struct *work)
984 {
985 	struct htb_sched *q = container_of(work, struct htb_sched, work);
986 	struct Qdisc *sch = q->watchdog.qdisc;
987 
988 	rcu_read_lock();
989 	__netif_schedule(qdisc_root(sch));
990 	rcu_read_unlock();
991 }
992 
htb_init(struct Qdisc * sch,struct nlattr * opt,struct netlink_ext_ack * extack)993 static int htb_init(struct Qdisc *sch, struct nlattr *opt,
994 		    struct netlink_ext_ack *extack)
995 {
996 	struct htb_sched *q = qdisc_priv(sch);
997 	struct nlattr *tb[TCA_HTB_MAX + 1];
998 	struct tc_htb_glob *gopt;
999 	int err;
1000 
1001 	qdisc_watchdog_init(&q->watchdog, sch);
1002 	INIT_WORK(&q->work, htb_work_func);
1003 
1004 	if (!opt)
1005 		return -EINVAL;
1006 
1007 	err = tcf_block_get(&q->block, &q->filter_list, sch, extack);
1008 	if (err)
1009 		return err;
1010 
1011 	err = nla_parse_nested_deprecated(tb, TCA_HTB_MAX, opt, htb_policy,
1012 					  NULL);
1013 	if (err < 0)
1014 		return err;
1015 
1016 	if (!tb[TCA_HTB_INIT])
1017 		return -EINVAL;
1018 
1019 	gopt = nla_data(tb[TCA_HTB_INIT]);
1020 	if (gopt->version != HTB_VER >> 16)
1021 		return -EINVAL;
1022 
1023 	err = qdisc_class_hash_init(&q->clhash);
1024 	if (err < 0)
1025 		return err;
1026 
1027 	qdisc_skb_head_init(&q->direct_queue);
1028 
1029 	if (tb[TCA_HTB_DIRECT_QLEN])
1030 		q->direct_qlen = nla_get_u32(tb[TCA_HTB_DIRECT_QLEN]);
1031 	else
1032 		q->direct_qlen = qdisc_dev(sch)->tx_queue_len;
1033 
1034 	if ((q->rate2quantum = gopt->rate2quantum) < 1)
1035 		q->rate2quantum = 1;
1036 	q->defcls = gopt->defcls;
1037 
1038 	return 0;
1039 }
1040 
htb_dump(struct Qdisc * sch,struct sk_buff * skb)1041 static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
1042 {
1043 	struct htb_sched *q = qdisc_priv(sch);
1044 	struct nlattr *nest;
1045 	struct tc_htb_glob gopt;
1046 
1047 	sch->qstats.overlimits = q->overlimits;
1048 	/* Its safe to not acquire qdisc lock. As we hold RTNL,
1049 	 * no change can happen on the qdisc parameters.
1050 	 */
1051 
1052 	gopt.direct_pkts = q->direct_pkts;
1053 	gopt.version = HTB_VER;
1054 	gopt.rate2quantum = q->rate2quantum;
1055 	gopt.defcls = q->defcls;
1056 	gopt.debug = 0;
1057 
1058 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
1059 	if (nest == NULL)
1060 		goto nla_put_failure;
1061 	if (nla_put(skb, TCA_HTB_INIT, sizeof(gopt), &gopt) ||
1062 	    nla_put_u32(skb, TCA_HTB_DIRECT_QLEN, q->direct_qlen))
1063 		goto nla_put_failure;
1064 
1065 	return nla_nest_end(skb, nest);
1066 
1067 nla_put_failure:
1068 	nla_nest_cancel(skb, nest);
1069 	return -1;
1070 }
1071 
htb_dump_class(struct Qdisc * sch,unsigned long arg,struct sk_buff * skb,struct tcmsg * tcm)1072 static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
1073 			  struct sk_buff *skb, struct tcmsg *tcm)
1074 {
1075 	struct htb_class *cl = (struct htb_class *)arg;
1076 	struct nlattr *nest;
1077 	struct tc_htb_opt opt;
1078 
1079 	/* Its safe to not acquire qdisc lock. As we hold RTNL,
1080 	 * no change can happen on the class parameters.
1081 	 */
1082 	tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT;
1083 	tcm->tcm_handle = cl->common.classid;
1084 	if (!cl->level && cl->leaf.q)
1085 		tcm->tcm_info = cl->leaf.q->handle;
1086 
1087 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
1088 	if (nest == NULL)
1089 		goto nla_put_failure;
1090 
1091 	memset(&opt, 0, sizeof(opt));
1092 
1093 	psched_ratecfg_getrate(&opt.rate, &cl->rate);
1094 	opt.buffer = PSCHED_NS2TICKS(cl->buffer);
1095 	psched_ratecfg_getrate(&opt.ceil, &cl->ceil);
1096 	opt.cbuffer = PSCHED_NS2TICKS(cl->cbuffer);
1097 	opt.quantum = cl->quantum;
1098 	opt.prio = cl->prio;
1099 	opt.level = cl->level;
1100 	if (nla_put(skb, TCA_HTB_PARMS, sizeof(opt), &opt))
1101 		goto nla_put_failure;
1102 	if ((cl->rate.rate_bytes_ps >= (1ULL << 32)) &&
1103 	    nla_put_u64_64bit(skb, TCA_HTB_RATE64, cl->rate.rate_bytes_ps,
1104 			      TCA_HTB_PAD))
1105 		goto nla_put_failure;
1106 	if ((cl->ceil.rate_bytes_ps >= (1ULL << 32)) &&
1107 	    nla_put_u64_64bit(skb, TCA_HTB_CEIL64, cl->ceil.rate_bytes_ps,
1108 			      TCA_HTB_PAD))
1109 		goto nla_put_failure;
1110 
1111 	return nla_nest_end(skb, nest);
1112 
1113 nla_put_failure:
1114 	nla_nest_cancel(skb, nest);
1115 	return -1;
1116 }
1117 
1118 static int
htb_dump_class_stats(struct Qdisc * sch,unsigned long arg,struct gnet_dump * d)1119 htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
1120 {
1121 	struct htb_class *cl = (struct htb_class *)arg;
1122 	struct gnet_stats_queue qs = {
1123 		.drops = cl->drops,
1124 		.overlimits = cl->overlimits,
1125 	};
1126 	__u32 qlen = 0;
1127 
1128 	if (!cl->level && cl->leaf.q)
1129 		qdisc_qstats_qlen_backlog(cl->leaf.q, &qlen, &qs.backlog);
1130 
1131 	cl->xstats.tokens = clamp_t(s64, PSCHED_NS2TICKS(cl->tokens),
1132 				    INT_MIN, INT_MAX);
1133 	cl->xstats.ctokens = clamp_t(s64, PSCHED_NS2TICKS(cl->ctokens),
1134 				     INT_MIN, INT_MAX);
1135 
1136 	if (gnet_stats_copy_basic(qdisc_root_sleeping_running(sch),
1137 				  d, NULL, &cl->bstats) < 0 ||
1138 	    gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 ||
1139 	    gnet_stats_copy_queue(d, NULL, &qs, qlen) < 0)
1140 		return -1;
1141 
1142 	return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
1143 }
1144 
htb_graft(struct Qdisc * sch,unsigned long arg,struct Qdisc * new,struct Qdisc ** old,struct netlink_ext_ack * extack)1145 static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
1146 		     struct Qdisc **old, struct netlink_ext_ack *extack)
1147 {
1148 	struct htb_class *cl = (struct htb_class *)arg;
1149 
1150 	if (cl->level)
1151 		return -EINVAL;
1152 	if (new == NULL &&
1153 	    (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1154 				     cl->common.classid, extack)) == NULL)
1155 		return -ENOBUFS;
1156 
1157 	*old = qdisc_replace(sch, new, &cl->leaf.q);
1158 	return 0;
1159 }
1160 
htb_leaf(struct Qdisc * sch,unsigned long arg)1161 static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg)
1162 {
1163 	struct htb_class *cl = (struct htb_class *)arg;
1164 	return !cl->level ? cl->leaf.q : NULL;
1165 }
1166 
htb_qlen_notify(struct Qdisc * sch,unsigned long arg)1167 static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
1168 {
1169 	struct htb_class *cl = (struct htb_class *)arg;
1170 
1171 	htb_deactivate(qdisc_priv(sch), cl);
1172 }
1173 
htb_parent_last_child(struct htb_class * cl)1174 static inline int htb_parent_last_child(struct htb_class *cl)
1175 {
1176 	if (!cl->parent)
1177 		/* the root class */
1178 		return 0;
1179 	if (cl->parent->children > 1)
1180 		/* not the last child */
1181 		return 0;
1182 	return 1;
1183 }
1184 
htb_parent_to_leaf(struct htb_sched * q,struct htb_class * cl,struct Qdisc * new_q)1185 static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl,
1186 			       struct Qdisc *new_q)
1187 {
1188 	struct htb_class *parent = cl->parent;
1189 
1190 	WARN_ON(cl->level || !cl->leaf.q || cl->prio_activity);
1191 
1192 	if (parent->cmode != HTB_CAN_SEND)
1193 		htb_safe_rb_erase(&parent->pq_node,
1194 				  &q->hlevel[parent->level].wait_pq);
1195 
1196 	parent->level = 0;
1197 	memset(&parent->inner, 0, sizeof(parent->inner));
1198 	parent->leaf.q = new_q ? new_q : &noop_qdisc;
1199 	parent->tokens = parent->buffer;
1200 	parent->ctokens = parent->cbuffer;
1201 	parent->t_c = ktime_get_ns();
1202 	parent->cmode = HTB_CAN_SEND;
1203 }
1204 
htb_destroy_class(struct Qdisc * sch,struct htb_class * cl)1205 static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl)
1206 {
1207 	if (!cl->level) {
1208 		WARN_ON(!cl->leaf.q);
1209 		qdisc_put(cl->leaf.q);
1210 	}
1211 	gen_kill_estimator(&cl->rate_est);
1212 	tcf_block_put(cl->block);
1213 	kfree(cl);
1214 }
1215 
htb_destroy(struct Qdisc * sch)1216 static void htb_destroy(struct Qdisc *sch)
1217 {
1218 	struct htb_sched *q = qdisc_priv(sch);
1219 	struct hlist_node *next;
1220 	struct htb_class *cl;
1221 	unsigned int i;
1222 
1223 	cancel_work_sync(&q->work);
1224 	qdisc_watchdog_cancel(&q->watchdog);
1225 	/* This line used to be after htb_destroy_class call below
1226 	 * and surprisingly it worked in 2.4. But it must precede it
1227 	 * because filter need its target class alive to be able to call
1228 	 * unbind_filter on it (without Oops).
1229 	 */
1230 	tcf_block_put(q->block);
1231 
1232 	for (i = 0; i < q->clhash.hashsize; i++) {
1233 		hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
1234 			tcf_block_put(cl->block);
1235 			cl->block = NULL;
1236 		}
1237 	}
1238 	for (i = 0; i < q->clhash.hashsize; i++) {
1239 		hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i],
1240 					  common.hnode)
1241 			htb_destroy_class(sch, cl);
1242 	}
1243 	qdisc_class_hash_destroy(&q->clhash);
1244 	__qdisc_reset_queue(&q->direct_queue);
1245 }
1246 
htb_delete(struct Qdisc * sch,unsigned long arg)1247 static int htb_delete(struct Qdisc *sch, unsigned long arg)
1248 {
1249 	struct htb_sched *q = qdisc_priv(sch);
1250 	struct htb_class *cl = (struct htb_class *)arg;
1251 	struct Qdisc *new_q = NULL;
1252 	int last_child = 0;
1253 
1254 	/* TODO: why don't allow to delete subtree ? references ? does
1255 	 * tc subsys guarantee us that in htb_destroy it holds no class
1256 	 * refs so that we can remove children safely there ?
1257 	 */
1258 	if (cl->children || cl->filter_cnt)
1259 		return -EBUSY;
1260 
1261 	if (!cl->level && htb_parent_last_child(cl)) {
1262 		new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1263 					  cl->parent->common.classid,
1264 					  NULL);
1265 		last_child = 1;
1266 	}
1267 
1268 	sch_tree_lock(sch);
1269 
1270 	if (!cl->level)
1271 		qdisc_purge_queue(cl->leaf.q);
1272 
1273 	/* delete from hash and active; remainder in destroy_class */
1274 	qdisc_class_hash_remove(&q->clhash, &cl->common);
1275 	if (cl->parent)
1276 		cl->parent->children--;
1277 
1278 	if (cl->prio_activity)
1279 		htb_deactivate(q, cl);
1280 
1281 	if (cl->cmode != HTB_CAN_SEND)
1282 		htb_safe_rb_erase(&cl->pq_node,
1283 				  &q->hlevel[cl->level].wait_pq);
1284 
1285 	if (last_child)
1286 		htb_parent_to_leaf(q, cl, new_q);
1287 
1288 	sch_tree_unlock(sch);
1289 
1290 	htb_destroy_class(sch, cl);
1291 	return 0;
1292 }
1293 
htb_change_class(struct Qdisc * sch,u32 classid,u32 parentid,struct nlattr ** tca,unsigned long * arg,struct netlink_ext_ack * extack)1294 static int htb_change_class(struct Qdisc *sch, u32 classid,
1295 			    u32 parentid, struct nlattr **tca,
1296 			    unsigned long *arg, struct netlink_ext_ack *extack)
1297 {
1298 	int err = -EINVAL;
1299 	struct htb_sched *q = qdisc_priv(sch);
1300 	struct htb_class *cl = (struct htb_class *)*arg, *parent;
1301 	struct nlattr *opt = tca[TCA_OPTIONS];
1302 	struct nlattr *tb[TCA_HTB_MAX + 1];
1303 	struct Qdisc *parent_qdisc = NULL;
1304 	struct tc_htb_opt *hopt;
1305 	u64 rate64, ceil64;
1306 	int warn = 0;
1307 
1308 	/* extract all subattrs from opt attr */
1309 	if (!opt)
1310 		goto failure;
1311 
1312 	err = nla_parse_nested_deprecated(tb, TCA_HTB_MAX, opt, htb_policy,
1313 					  NULL);
1314 	if (err < 0)
1315 		goto failure;
1316 
1317 	err = -EINVAL;
1318 	if (tb[TCA_HTB_PARMS] == NULL)
1319 		goto failure;
1320 
1321 	parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch);
1322 
1323 	hopt = nla_data(tb[TCA_HTB_PARMS]);
1324 	if (!hopt->rate.rate || !hopt->ceil.rate)
1325 		goto failure;
1326 
1327 	/* Keeping backward compatible with rate_table based iproute2 tc */
1328 	if (hopt->rate.linklayer == TC_LINKLAYER_UNAWARE)
1329 		qdisc_put_rtab(qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB],
1330 					      NULL));
1331 
1332 	if (hopt->ceil.linklayer == TC_LINKLAYER_UNAWARE)
1333 		qdisc_put_rtab(qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB],
1334 					      NULL));
1335 
1336 	if (!cl) {		/* new class */
1337 		struct Qdisc *new_q;
1338 		int prio;
1339 		struct {
1340 			struct nlattr		nla;
1341 			struct gnet_estimator	opt;
1342 		} est = {
1343 			.nla = {
1344 				.nla_len	= nla_attr_size(sizeof(est.opt)),
1345 				.nla_type	= TCA_RATE,
1346 			},
1347 			.opt = {
1348 				/* 4s interval, 16s averaging constant */
1349 				.interval	= 2,
1350 				.ewma_log	= 2,
1351 			},
1352 		};
1353 
1354 		/* check for valid classid */
1355 		if (!classid || TC_H_MAJ(classid ^ sch->handle) ||
1356 		    htb_find(classid, sch))
1357 			goto failure;
1358 
1359 		/* check maximal depth */
1360 		if (parent && parent->parent && parent->parent->level < 2) {
1361 			pr_err("htb: tree is too deep\n");
1362 			goto failure;
1363 		}
1364 		err = -ENOBUFS;
1365 		cl = kzalloc(sizeof(*cl), GFP_KERNEL);
1366 		if (!cl)
1367 			goto failure;
1368 
1369 		err = tcf_block_get(&cl->block, &cl->filter_list, sch, extack);
1370 		if (err) {
1371 			kfree(cl);
1372 			goto failure;
1373 		}
1374 		if (htb_rate_est || tca[TCA_RATE]) {
1375 			err = gen_new_estimator(&cl->bstats, NULL,
1376 						&cl->rate_est,
1377 						NULL,
1378 						qdisc_root_sleeping_running(sch),
1379 						tca[TCA_RATE] ? : &est.nla);
1380 			if (err) {
1381 				tcf_block_put(cl->block);
1382 				kfree(cl);
1383 				goto failure;
1384 			}
1385 		}
1386 
1387 		cl->children = 0;
1388 		RB_CLEAR_NODE(&cl->pq_node);
1389 
1390 		for (prio = 0; prio < TC_HTB_NUMPRIO; prio++)
1391 			RB_CLEAR_NODE(&cl->node[prio]);
1392 
1393 		/* create leaf qdisc early because it uses kmalloc(GFP_KERNEL)
1394 		 * so that can't be used inside of sch_tree_lock
1395 		 * -- thanks to Karlis Peisenieks
1396 		 */
1397 		new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1398 					  classid, NULL);
1399 		sch_tree_lock(sch);
1400 		if (parent && !parent->level) {
1401 			/* turn parent into inner node */
1402 			qdisc_purge_queue(parent->leaf.q);
1403 			parent_qdisc = parent->leaf.q;
1404 			if (parent->prio_activity)
1405 				htb_deactivate(q, parent);
1406 
1407 			/* remove from evt list because of level change */
1408 			if (parent->cmode != HTB_CAN_SEND) {
1409 				htb_safe_rb_erase(&parent->pq_node, &q->hlevel[0].wait_pq);
1410 				parent->cmode = HTB_CAN_SEND;
1411 			}
1412 			parent->level = (parent->parent ? parent->parent->level
1413 					 : TC_HTB_MAXDEPTH) - 1;
1414 			memset(&parent->inner, 0, sizeof(parent->inner));
1415 		}
1416 		/* leaf (we) needs elementary qdisc */
1417 		cl->leaf.q = new_q ? new_q : &noop_qdisc;
1418 
1419 		cl->common.classid = classid;
1420 		cl->parent = parent;
1421 
1422 		/* set class to be in HTB_CAN_SEND state */
1423 		cl->tokens = PSCHED_TICKS2NS(hopt->buffer);
1424 		cl->ctokens = PSCHED_TICKS2NS(hopt->cbuffer);
1425 		cl->mbuffer = 60ULL * NSEC_PER_SEC;	/* 1min */
1426 		cl->t_c = ktime_get_ns();
1427 		cl->cmode = HTB_CAN_SEND;
1428 
1429 		/* attach to the hash list and parent's family */
1430 		qdisc_class_hash_insert(&q->clhash, &cl->common);
1431 		if (parent)
1432 			parent->children++;
1433 		if (cl->leaf.q != &noop_qdisc)
1434 			qdisc_hash_add(cl->leaf.q, true);
1435 	} else {
1436 		if (tca[TCA_RATE]) {
1437 			err = gen_replace_estimator(&cl->bstats, NULL,
1438 						    &cl->rate_est,
1439 						    NULL,
1440 						    qdisc_root_sleeping_running(sch),
1441 						    tca[TCA_RATE]);
1442 			if (err)
1443 				return err;
1444 		}
1445 		sch_tree_lock(sch);
1446 	}
1447 
1448 	rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0;
1449 
1450 	ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0;
1451 
1452 	psched_ratecfg_precompute(&cl->rate, &hopt->rate, rate64);
1453 	psched_ratecfg_precompute(&cl->ceil, &hopt->ceil, ceil64);
1454 
1455 	/* it used to be a nasty bug here, we have to check that node
1456 	 * is really leaf before changing cl->leaf !
1457 	 */
1458 	if (!cl->level) {
1459 		u64 quantum = cl->rate.rate_bytes_ps;
1460 
1461 		do_div(quantum, q->rate2quantum);
1462 		cl->quantum = min_t(u64, quantum, INT_MAX);
1463 
1464 		if (!hopt->quantum && cl->quantum < 1000) {
1465 			warn = -1;
1466 			cl->quantum = 1000;
1467 		}
1468 		if (!hopt->quantum && cl->quantum > 200000) {
1469 			warn = 1;
1470 			cl->quantum = 200000;
1471 		}
1472 		if (hopt->quantum)
1473 			cl->quantum = hopt->quantum;
1474 		if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO)
1475 			cl->prio = TC_HTB_NUMPRIO - 1;
1476 	}
1477 
1478 	cl->buffer = PSCHED_TICKS2NS(hopt->buffer);
1479 	cl->cbuffer = PSCHED_TICKS2NS(hopt->cbuffer);
1480 
1481 	sch_tree_unlock(sch);
1482 	qdisc_put(parent_qdisc);
1483 
1484 	if (warn)
1485 		pr_warn("HTB: quantum of class %X is %s. Consider r2q change.\n",
1486 			    cl->common.classid, (warn == -1 ? "small" : "big"));
1487 
1488 	qdisc_class_hash_grow(sch, &q->clhash);
1489 
1490 	*arg = (unsigned long)cl;
1491 	return 0;
1492 
1493 failure:
1494 	return err;
1495 }
1496 
htb_tcf_block(struct Qdisc * sch,unsigned long arg,struct netlink_ext_ack * extack)1497 static struct tcf_block *htb_tcf_block(struct Qdisc *sch, unsigned long arg,
1498 				       struct netlink_ext_ack *extack)
1499 {
1500 	struct htb_sched *q = qdisc_priv(sch);
1501 	struct htb_class *cl = (struct htb_class *)arg;
1502 
1503 	return cl ? cl->block : q->block;
1504 }
1505 
htb_bind_filter(struct Qdisc * sch,unsigned long parent,u32 classid)1506 static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent,
1507 				     u32 classid)
1508 {
1509 	struct htb_class *cl = htb_find(classid, sch);
1510 
1511 	/*if (cl && !cl->level) return 0;
1512 	 * The line above used to be there to prevent attaching filters to
1513 	 * leaves. But at least tc_index filter uses this just to get class
1514 	 * for other reasons so that we have to allow for it.
1515 	 * ----
1516 	 * 19.6.2002 As Werner explained it is ok - bind filter is just
1517 	 * another way to "lock" the class - unlike "get" this lock can
1518 	 * be broken by class during destroy IIUC.
1519 	 */
1520 	if (cl)
1521 		cl->filter_cnt++;
1522 	return (unsigned long)cl;
1523 }
1524 
htb_unbind_filter(struct Qdisc * sch,unsigned long arg)1525 static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg)
1526 {
1527 	struct htb_class *cl = (struct htb_class *)arg;
1528 
1529 	if (cl)
1530 		cl->filter_cnt--;
1531 }
1532 
htb_walk(struct Qdisc * sch,struct qdisc_walker * arg)1533 static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg)
1534 {
1535 	struct htb_sched *q = qdisc_priv(sch);
1536 	struct htb_class *cl;
1537 	unsigned int i;
1538 
1539 	if (arg->stop)
1540 		return;
1541 
1542 	for (i = 0; i < q->clhash.hashsize; i++) {
1543 		hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
1544 			if (arg->count < arg->skip) {
1545 				arg->count++;
1546 				continue;
1547 			}
1548 			if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
1549 				arg->stop = 1;
1550 				return;
1551 			}
1552 			arg->count++;
1553 		}
1554 	}
1555 }
1556 
1557 static const struct Qdisc_class_ops htb_class_ops = {
1558 	.graft		=	htb_graft,
1559 	.leaf		=	htb_leaf,
1560 	.qlen_notify	=	htb_qlen_notify,
1561 	.find		=	htb_search,
1562 	.change		=	htb_change_class,
1563 	.delete		=	htb_delete,
1564 	.walk		=	htb_walk,
1565 	.tcf_block	=	htb_tcf_block,
1566 	.bind_tcf	=	htb_bind_filter,
1567 	.unbind_tcf	=	htb_unbind_filter,
1568 	.dump		=	htb_dump_class,
1569 	.dump_stats	=	htb_dump_class_stats,
1570 };
1571 
1572 static struct Qdisc_ops htb_qdisc_ops __read_mostly = {
1573 	.cl_ops		=	&htb_class_ops,
1574 	.id		=	"htb",
1575 	.priv_size	=	sizeof(struct htb_sched),
1576 	.enqueue	=	htb_enqueue,
1577 	.dequeue	=	htb_dequeue,
1578 	.peek		=	qdisc_peek_dequeued,
1579 	.init		=	htb_init,
1580 	.reset		=	htb_reset,
1581 	.destroy	=	htb_destroy,
1582 	.dump		=	htb_dump,
1583 	.owner		=	THIS_MODULE,
1584 };
1585 
htb_module_init(void)1586 static int __init htb_module_init(void)
1587 {
1588 	return register_qdisc(&htb_qdisc_ops);
1589 }
htb_module_exit(void)1590 static void __exit htb_module_exit(void)
1591 {
1592 	unregister_qdisc(&htb_qdisc_ops);
1593 }
1594 
1595 module_init(htb_module_init)
1596 module_exit(htb_module_exit)
1597 MODULE_LICENSE("GPL");
1598