Loading...
1/*
2 * net/sched/sch_htb.c Hierarchical token bucket, feed tree version
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Martin Devera, <devik@cdi.cz>
10 *
11 * Credits (in time order) for older HTB versions:
12 * Stef Coene <stef.coene@docum.org>
13 * HTB support at LARTC mailing list
14 * Ondrej Kraus, <krauso@barr.cz>
15 * found missing INIT_QDISC(htb)
16 * Vladimir Smelhaus, Aamer Akhter, Bert Hubert
17 * helped a lot to locate nasty class stall bug
18 * Andi Kleen, Jamal Hadi, Bert Hubert
19 * code review and helpful comments on shaping
20 * Tomasz Wrona, <tw@eter.tym.pl>
21 * created test case so that I was able to fix nasty bug
22 * Wilfried Weissmann
23 * spotted bug in dequeue code and helped with fix
24 * Jiri Fojtasek
25 * fixed requeue routine
26 * and many others. thanks.
27 */
28#include <linux/module.h>
29#include <linux/moduleparam.h>
30#include <linux/types.h>
31#include <linux/kernel.h>
32#include <linux/string.h>
33#include <linux/errno.h>
34#include <linux/skbuff.h>
35#include <linux/list.h>
36#include <linux/compiler.h>
37#include <linux/rbtree.h>
38#include <linux/workqueue.h>
39#include <linux/slab.h>
40#include <net/netlink.h>
41#include <net/sch_generic.h>
42#include <net/pkt_sched.h>
43
44/* HTB algorithm.
45 Author: devik@cdi.cz
46 ========================================================================
47 HTB is like TBF with multiple classes. It is also similar to CBQ because
48 it allows to assign priority to each class in hierarchy.
49 In fact it is another implementation of Floyd's formal sharing.
50
51 Levels:
52 Each class is assigned level. Leaf has ALWAYS level 0 and root
53 classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level
54 one less than their parent.
55*/
56
57static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */
58#define HTB_VER 0x30011 /* major must be matched with number suplied by TC as version */
59
60#if HTB_VER >> 16 != TC_HTB_PROTOVER
61#error "Mismatched sch_htb.c and pkt_sch.h"
62#endif
63
64/* Module parameter and sysfs export */
65module_param (htb_hysteresis, int, 0640);
66MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate");
67
68static int htb_rate_est = 0; /* htb classes have a default rate estimator */
69module_param(htb_rate_est, int, 0640);
70MODULE_PARM_DESC(htb_rate_est, "setup a default rate estimator (4sec 16sec) for htb classes");
71
72/* used internaly to keep status of single class */
73enum htb_cmode {
74 HTB_CANT_SEND, /* class can't send and can't borrow */
75 HTB_MAY_BORROW, /* class can't send but may borrow */
76 HTB_CAN_SEND /* class can send */
77};
78
79struct htb_prio {
80 union {
81 struct rb_root row;
82 struct rb_root feed;
83 };
84 struct rb_node *ptr;
85 /* When class changes from state 1->2 and disconnects from
86 * parent's feed then we lost ptr value and start from the
87 * first child again. Here we store classid of the
88 * last valid ptr (used when ptr is NULL).
89 */
90 u32 last_ptr_id;
91};
92
93/* interior & leaf nodes; props specific to leaves are marked L:
94 * To reduce false sharing, place mostly read fields at beginning,
95 * and mostly written ones at the end.
96 */
97struct htb_class {
98 struct Qdisc_class_common common;
99 struct psched_ratecfg rate;
100 struct psched_ratecfg ceil;
101 s64 buffer, cbuffer;/* token bucket depth/rate */
102 s64 mbuffer; /* max wait time */
103 u32 prio; /* these two are used only by leaves... */
104 int quantum; /* but stored for parent-to-leaf return */
105
106 struct tcf_proto __rcu *filter_list; /* class attached filters */
107 int filter_cnt;
108 int refcnt; /* usage count of this class */
109
110 int level; /* our level (see above) */
111 unsigned int children;
112 struct htb_class *parent; /* parent class */
113
114 struct gnet_stats_rate_est64 rate_est;
115
116 /*
117 * Written often fields
118 */
119 struct gnet_stats_basic_packed bstats;
120 struct gnet_stats_queue qstats;
121 struct tc_htb_xstats xstats; /* our special stats */
122
123 /* token bucket parameters */
124 s64 tokens, ctokens;/* current number of tokens */
125 s64 t_c; /* checkpoint time */
126
127 union {
128 struct htb_class_leaf {
129 struct list_head drop_list;
130 int deficit[TC_HTB_MAXDEPTH];
131 struct Qdisc *q;
132 } leaf;
133 struct htb_class_inner {
134 struct htb_prio clprio[TC_HTB_NUMPRIO];
135 } inner;
136 } un;
137 s64 pq_key;
138
139 int prio_activity; /* for which prios are we active */
140 enum htb_cmode cmode; /* current mode of the class */
141 struct rb_node pq_node; /* node for event queue */
142 struct rb_node node[TC_HTB_NUMPRIO]; /* node for self or feed tree */
143};
144
145struct htb_level {
146 struct rb_root wait_pq;
147 struct htb_prio hprio[TC_HTB_NUMPRIO];
148};
149
150struct htb_sched {
151 struct Qdisc_class_hash clhash;
152 int defcls; /* class where unclassified flows go to */
153 int rate2quantum; /* quant = rate / rate2quantum */
154
155 /* filters for qdisc itself */
156 struct tcf_proto __rcu *filter_list;
157
158#define HTB_WARN_TOOMANYEVENTS 0x1
159 unsigned int warned; /* only one warning */
160 int direct_qlen;
161 struct work_struct work;
162
163 /* non shaped skbs; let them go directly thru */
164 struct sk_buff_head direct_queue;
165 long direct_pkts;
166
167 struct qdisc_watchdog watchdog;
168
169 s64 now; /* cached dequeue time */
170 struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */
171
172 /* time of nearest event per level (row) */
173 s64 near_ev_cache[TC_HTB_MAXDEPTH];
174
175 int row_mask[TC_HTB_MAXDEPTH];
176
177 struct htb_level hlevel[TC_HTB_MAXDEPTH];
178};
179
180/* find class in global hash table using given handle */
181static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch)
182{
183 struct htb_sched *q = qdisc_priv(sch);
184 struct Qdisc_class_common *clc;
185
186 clc = qdisc_class_find(&q->clhash, handle);
187 if (clc == NULL)
188 return NULL;
189 return container_of(clc, struct htb_class, common);
190}
191
192/**
193 * htb_classify - classify a packet into class
194 *
195 * It returns NULL if the packet should be dropped or -1 if the packet
196 * should be passed directly thru. In all other cases leaf class is returned.
197 * We allow direct class selection by classid in priority. The we examine
198 * filters in qdisc and in inner nodes (if higher filter points to the inner
199 * node). If we end up with classid MAJOR:0 we enqueue the skb into special
200 * internal fifo (direct). These packets then go directly thru. If we still
201 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessful
202 * then finish and return direct queue.
203 */
204#define HTB_DIRECT ((struct htb_class *)-1L)
205
206static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
207 int *qerr)
208{
209 struct htb_sched *q = qdisc_priv(sch);
210 struct htb_class *cl;
211 struct tcf_result res;
212 struct tcf_proto *tcf;
213 int result;
214
215 /* allow to select class by setting skb->priority to valid classid;
216 * note that nfmark can be used too by attaching filter fw with no
217 * rules in it
218 */
219 if (skb->priority == sch->handle)
220 return HTB_DIRECT; /* X:0 (direct flow) selected */
221 cl = htb_find(skb->priority, sch);
222 if (cl) {
223 if (cl->level == 0)
224 return cl;
225 /* Start with inner filter chain if a non-leaf class is selected */
226 tcf = rcu_dereference_bh(cl->filter_list);
227 } else {
228 tcf = rcu_dereference_bh(q->filter_list);
229 }
230
231 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
232 while (tcf && (result = tc_classify(skb, tcf, &res, false)) >= 0) {
233#ifdef CONFIG_NET_CLS_ACT
234 switch (result) {
235 case TC_ACT_QUEUED:
236 case TC_ACT_STOLEN:
237 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
238 case TC_ACT_SHOT:
239 return NULL;
240 }
241#endif
242 cl = (void *)res.class;
243 if (!cl) {
244 if (res.classid == sch->handle)
245 return HTB_DIRECT; /* X:0 (direct flow) */
246 cl = htb_find(res.classid, sch);
247 if (!cl)
248 break; /* filter selected invalid classid */
249 }
250 if (!cl->level)
251 return cl; /* we hit leaf; return it */
252
253 /* we have got inner class; apply inner filter chain */
254 tcf = rcu_dereference_bh(cl->filter_list);
255 }
256 /* classification failed; try to use default class */
257 cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch);
258 if (!cl || cl->level)
259 return HTB_DIRECT; /* bad default .. this is safe bet */
260 return cl;
261}
262
263/**
264 * htb_add_to_id_tree - adds class to the round robin list
265 *
266 * Routine adds class to the list (actually tree) sorted by classid.
267 * Make sure that class is not already on such list for given prio.
268 */
269static void htb_add_to_id_tree(struct rb_root *root,
270 struct htb_class *cl, int prio)
271{
272 struct rb_node **p = &root->rb_node, *parent = NULL;
273
274 while (*p) {
275 struct htb_class *c;
276 parent = *p;
277 c = rb_entry(parent, struct htb_class, node[prio]);
278
279 if (cl->common.classid > c->common.classid)
280 p = &parent->rb_right;
281 else
282 p = &parent->rb_left;
283 }
284 rb_link_node(&cl->node[prio], parent, p);
285 rb_insert_color(&cl->node[prio], root);
286}
287
288/**
289 * htb_add_to_wait_tree - adds class to the event queue with delay
290 *
291 * The class is added to priority event queue to indicate that class will
292 * change its mode in cl->pq_key microseconds. Make sure that class is not
293 * already in the queue.
294 */
295static void htb_add_to_wait_tree(struct htb_sched *q,
296 struct htb_class *cl, s64 delay)
297{
298 struct rb_node **p = &q->hlevel[cl->level].wait_pq.rb_node, *parent = NULL;
299
300 cl->pq_key = q->now + delay;
301 if (cl->pq_key == q->now)
302 cl->pq_key++;
303
304 /* update the nearest event cache */
305 if (q->near_ev_cache[cl->level] > cl->pq_key)
306 q->near_ev_cache[cl->level] = cl->pq_key;
307
308 while (*p) {
309 struct htb_class *c;
310 parent = *p;
311 c = rb_entry(parent, struct htb_class, pq_node);
312 if (cl->pq_key >= c->pq_key)
313 p = &parent->rb_right;
314 else
315 p = &parent->rb_left;
316 }
317 rb_link_node(&cl->pq_node, parent, p);
318 rb_insert_color(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
319}
320
321/**
322 * htb_next_rb_node - finds next node in binary tree
323 *
324 * When we are past last key we return NULL.
325 * Average complexity is 2 steps per call.
326 */
327static inline void htb_next_rb_node(struct rb_node **n)
328{
329 *n = rb_next(*n);
330}
331
332/**
333 * htb_add_class_to_row - add class to its row
334 *
335 * The class is added to row at priorities marked in mask.
336 * It does nothing if mask == 0.
337 */
338static inline void htb_add_class_to_row(struct htb_sched *q,
339 struct htb_class *cl, int mask)
340{
341 q->row_mask[cl->level] |= mask;
342 while (mask) {
343 int prio = ffz(~mask);
344 mask &= ~(1 << prio);
345 htb_add_to_id_tree(&q->hlevel[cl->level].hprio[prio].row, cl, prio);
346 }
347}
348
349/* If this triggers, it is a bug in this code, but it need not be fatal */
350static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root)
351{
352 if (RB_EMPTY_NODE(rb)) {
353 WARN_ON(1);
354 } else {
355 rb_erase(rb, root);
356 RB_CLEAR_NODE(rb);
357 }
358}
359
360
361/**
362 * htb_remove_class_from_row - removes class from its row
363 *
364 * The class is removed from row at priorities marked in mask.
365 * It does nothing if mask == 0.
366 */
367static inline void htb_remove_class_from_row(struct htb_sched *q,
368 struct htb_class *cl, int mask)
369{
370 int m = 0;
371 struct htb_level *hlevel = &q->hlevel[cl->level];
372
373 while (mask) {
374 int prio = ffz(~mask);
375 struct htb_prio *hprio = &hlevel->hprio[prio];
376
377 mask &= ~(1 << prio);
378 if (hprio->ptr == cl->node + prio)
379 htb_next_rb_node(&hprio->ptr);
380
381 htb_safe_rb_erase(cl->node + prio, &hprio->row);
382 if (!hprio->row.rb_node)
383 m |= 1 << prio;
384 }
385 q->row_mask[cl->level] &= ~m;
386}
387
388/**
389 * htb_activate_prios - creates active classe's feed chain
390 *
391 * The class is connected to ancestors and/or appropriate rows
392 * for priorities it is participating on. cl->cmode must be new
393 * (activated) mode. It does nothing if cl->prio_activity == 0.
394 */
395static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl)
396{
397 struct htb_class *p = cl->parent;
398 long m, mask = cl->prio_activity;
399
400 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
401 m = mask;
402 while (m) {
403 int prio = ffz(~m);
404 m &= ~(1 << prio);
405
406 if (p->un.inner.clprio[prio].feed.rb_node)
407 /* parent already has its feed in use so that
408 * reset bit in mask as parent is already ok
409 */
410 mask &= ~(1 << prio);
411
412 htb_add_to_id_tree(&p->un.inner.clprio[prio].feed, cl, prio);
413 }
414 p->prio_activity |= mask;
415 cl = p;
416 p = cl->parent;
417
418 }
419 if (cl->cmode == HTB_CAN_SEND && mask)
420 htb_add_class_to_row(q, cl, mask);
421}
422
423/**
424 * htb_deactivate_prios - remove class from feed chain
425 *
426 * cl->cmode must represent old mode (before deactivation). It does
427 * nothing if cl->prio_activity == 0. Class is removed from all feed
428 * chains and rows.
429 */
430static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl)
431{
432 struct htb_class *p = cl->parent;
433 long m, mask = cl->prio_activity;
434
435 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
436 m = mask;
437 mask = 0;
438 while (m) {
439 int prio = ffz(~m);
440 m &= ~(1 << prio);
441
442 if (p->un.inner.clprio[prio].ptr == cl->node + prio) {
443 /* we are removing child which is pointed to from
444 * parent feed - forget the pointer but remember
445 * classid
446 */
447 p->un.inner.clprio[prio].last_ptr_id = cl->common.classid;
448 p->un.inner.clprio[prio].ptr = NULL;
449 }
450
451 htb_safe_rb_erase(cl->node + prio,
452 &p->un.inner.clprio[prio].feed);
453
454 if (!p->un.inner.clprio[prio].feed.rb_node)
455 mask |= 1 << prio;
456 }
457
458 p->prio_activity &= ~mask;
459 cl = p;
460 p = cl->parent;
461
462 }
463 if (cl->cmode == HTB_CAN_SEND && mask)
464 htb_remove_class_from_row(q, cl, mask);
465}
466
467static inline s64 htb_lowater(const struct htb_class *cl)
468{
469 if (htb_hysteresis)
470 return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0;
471 else
472 return 0;
473}
474static inline s64 htb_hiwater(const struct htb_class *cl)
475{
476 if (htb_hysteresis)
477 return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0;
478 else
479 return 0;
480}
481
482
483/**
484 * htb_class_mode - computes and returns current class mode
485 *
486 * It computes cl's mode at time cl->t_c+diff and returns it. If mode
487 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference
488 * from now to time when cl will change its state.
489 * Also it is worth to note that class mode doesn't change simply
490 * at cl->{c,}tokens == 0 but there can rather be hysteresis of
491 * 0 .. -cl->{c,}buffer range. It is meant to limit number of
492 * mode transitions per time unit. The speed gain is about 1/6.
493 */
494static inline enum htb_cmode
495htb_class_mode(struct htb_class *cl, s64 *diff)
496{
497 s64 toks;
498
499 if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) {
500 *diff = -toks;
501 return HTB_CANT_SEND;
502 }
503
504 if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl))
505 return HTB_CAN_SEND;
506
507 *diff = -toks;
508 return HTB_MAY_BORROW;
509}
510
511/**
512 * htb_change_class_mode - changes classe's mode
513 *
514 * This should be the only way how to change classe's mode under normal
515 * cirsumstances. Routine will update feed lists linkage, change mode
516 * and add class to the wait event queue if appropriate. New mode should
517 * be different from old one and cl->pq_key has to be valid if changing
518 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree).
519 */
520static void
521htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, s64 *diff)
522{
523 enum htb_cmode new_mode = htb_class_mode(cl, diff);
524
525 if (new_mode == cl->cmode)
526 return;
527
528 if (cl->prio_activity) { /* not necessary: speed optimization */
529 if (cl->cmode != HTB_CANT_SEND)
530 htb_deactivate_prios(q, cl);
531 cl->cmode = new_mode;
532 if (new_mode != HTB_CANT_SEND)
533 htb_activate_prios(q, cl);
534 } else
535 cl->cmode = new_mode;
536}
537
538/**
539 * htb_activate - inserts leaf cl into appropriate active feeds
540 *
541 * Routine learns (new) priority of leaf and activates feed chain
542 * for the prio. It can be called on already active leaf safely.
543 * It also adds leaf into droplist.
544 */
545static inline void htb_activate(struct htb_sched *q, struct htb_class *cl)
546{
547 WARN_ON(cl->level || !cl->un.leaf.q || !cl->un.leaf.q->q.qlen);
548
549 if (!cl->prio_activity) {
550 cl->prio_activity = 1 << cl->prio;
551 htb_activate_prios(q, cl);
552 list_add_tail(&cl->un.leaf.drop_list,
553 q->drops + cl->prio);
554 }
555}
556
557/**
558 * htb_deactivate - remove leaf cl from active feeds
559 *
560 * Make sure that leaf is active. In the other words it can't be called
561 * with non-active leaf. It also removes class from the drop list.
562 */
563static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
564{
565 WARN_ON(!cl->prio_activity);
566
567 htb_deactivate_prios(q, cl);
568 cl->prio_activity = 0;
569 list_del_init(&cl->un.leaf.drop_list);
570}
571
572static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch)
573{
574 int uninitialized_var(ret);
575 struct htb_sched *q = qdisc_priv(sch);
576 struct htb_class *cl = htb_classify(skb, sch, &ret);
577
578 if (cl == HTB_DIRECT) {
579 /* enqueue to helper queue */
580 if (q->direct_queue.qlen < q->direct_qlen) {
581 __skb_queue_tail(&q->direct_queue, skb);
582 q->direct_pkts++;
583 } else {
584 return qdisc_drop(skb, sch);
585 }
586#ifdef CONFIG_NET_CLS_ACT
587 } else if (!cl) {
588 if (ret & __NET_XMIT_BYPASS)
589 qdisc_qstats_drop(sch);
590 kfree_skb(skb);
591 return ret;
592#endif
593 } else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q)) != NET_XMIT_SUCCESS) {
594 if (net_xmit_drop_count(ret)) {
595 qdisc_qstats_drop(sch);
596 cl->qstats.drops++;
597 }
598 return ret;
599 } else {
600 htb_activate(q, cl);
601 }
602
603 qdisc_qstats_backlog_inc(sch, skb);
604 sch->q.qlen++;
605 return NET_XMIT_SUCCESS;
606}
607
608static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, s64 diff)
609{
610 s64 toks = diff + cl->tokens;
611
612 if (toks > cl->buffer)
613 toks = cl->buffer;
614 toks -= (s64) psched_l2t_ns(&cl->rate, bytes);
615 if (toks <= -cl->mbuffer)
616 toks = 1 - cl->mbuffer;
617
618 cl->tokens = toks;
619}
620
621static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, s64 diff)
622{
623 s64 toks = diff + cl->ctokens;
624
625 if (toks > cl->cbuffer)
626 toks = cl->cbuffer;
627 toks -= (s64) psched_l2t_ns(&cl->ceil, bytes);
628 if (toks <= -cl->mbuffer)
629 toks = 1 - cl->mbuffer;
630
631 cl->ctokens = toks;
632}
633
634/**
635 * htb_charge_class - charges amount "bytes" to leaf and ancestors
636 *
637 * Routine assumes that packet "bytes" long was dequeued from leaf cl
638 * borrowing from "level". It accounts bytes to ceil leaky bucket for
639 * leaf and all ancestors and to rate bucket for ancestors at levels
640 * "level" and higher. It also handles possible change of mode resulting
641 * from the update. Note that mode can also increase here (MAY_BORROW to
642 * CAN_SEND) because we can use more precise clock that event queue here.
643 * In such case we remove class from event queue first.
644 */
645static void htb_charge_class(struct htb_sched *q, struct htb_class *cl,
646 int level, struct sk_buff *skb)
647{
648 int bytes = qdisc_pkt_len(skb);
649 enum htb_cmode old_mode;
650 s64 diff;
651
652 while (cl) {
653 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
654 if (cl->level >= level) {
655 if (cl->level == level)
656 cl->xstats.lends++;
657 htb_accnt_tokens(cl, bytes, diff);
658 } else {
659 cl->xstats.borrows++;
660 cl->tokens += diff; /* we moved t_c; update tokens */
661 }
662 htb_accnt_ctokens(cl, bytes, diff);
663 cl->t_c = q->now;
664
665 old_mode = cl->cmode;
666 diff = 0;
667 htb_change_class_mode(q, cl, &diff);
668 if (old_mode != cl->cmode) {
669 if (old_mode != HTB_CAN_SEND)
670 htb_safe_rb_erase(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
671 if (cl->cmode != HTB_CAN_SEND)
672 htb_add_to_wait_tree(q, cl, diff);
673 }
674
675 /* update basic stats except for leaves which are already updated */
676 if (cl->level)
677 bstats_update(&cl->bstats, skb);
678
679 cl = cl->parent;
680 }
681}
682
683/**
684 * htb_do_events - make mode changes to classes at the level
685 *
686 * Scans event queue for pending events and applies them. Returns time of
687 * next pending event (0 for no event in pq, q->now for too many events).
688 * Note: Applied are events whose have cl->pq_key <= q->now.
689 */
690static s64 htb_do_events(struct htb_sched *q, const int level,
691 unsigned long start)
692{
693 /* don't run for longer than 2 jiffies; 2 is used instead of
694 * 1 to simplify things when jiffy is going to be incremented
695 * too soon
696 */
697 unsigned long stop_at = start + 2;
698 struct rb_root *wait_pq = &q->hlevel[level].wait_pq;
699
700 while (time_before(jiffies, stop_at)) {
701 struct htb_class *cl;
702 s64 diff;
703 struct rb_node *p = rb_first(wait_pq);
704
705 if (!p)
706 return 0;
707
708 cl = rb_entry(p, struct htb_class, pq_node);
709 if (cl->pq_key > q->now)
710 return cl->pq_key;
711
712 htb_safe_rb_erase(p, wait_pq);
713 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
714 htb_change_class_mode(q, cl, &diff);
715 if (cl->cmode != HTB_CAN_SEND)
716 htb_add_to_wait_tree(q, cl, diff);
717 }
718
719 /* too much load - let's continue after a break for scheduling */
720 if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) {
721 pr_warn("htb: too many events!\n");
722 q->warned |= HTB_WARN_TOOMANYEVENTS;
723 }
724
725 return q->now;
726}
727
728/* Returns class->node+prio from id-tree where classe's id is >= id. NULL
729 * is no such one exists.
730 */
731static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n,
732 u32 id)
733{
734 struct rb_node *r = NULL;
735 while (n) {
736 struct htb_class *cl =
737 rb_entry(n, struct htb_class, node[prio]);
738
739 if (id > cl->common.classid) {
740 n = n->rb_right;
741 } else if (id < cl->common.classid) {
742 r = n;
743 n = n->rb_left;
744 } else {
745 return n;
746 }
747 }
748 return r;
749}
750
751/**
752 * htb_lookup_leaf - returns next leaf class in DRR order
753 *
754 * Find leaf where current feed pointers points to.
755 */
756static struct htb_class *htb_lookup_leaf(struct htb_prio *hprio, const int prio)
757{
758 int i;
759 struct {
760 struct rb_node *root;
761 struct rb_node **pptr;
762 u32 *pid;
763 } stk[TC_HTB_MAXDEPTH], *sp = stk;
764
765 BUG_ON(!hprio->row.rb_node);
766 sp->root = hprio->row.rb_node;
767 sp->pptr = &hprio->ptr;
768 sp->pid = &hprio->last_ptr_id;
769
770 for (i = 0; i < 65535; i++) {
771 if (!*sp->pptr && *sp->pid) {
772 /* ptr was invalidated but id is valid - try to recover
773 * the original or next ptr
774 */
775 *sp->pptr =
776 htb_id_find_next_upper(prio, sp->root, *sp->pid);
777 }
778 *sp->pid = 0; /* ptr is valid now so that remove this hint as it
779 * can become out of date quickly
780 */
781 if (!*sp->pptr) { /* we are at right end; rewind & go up */
782 *sp->pptr = sp->root;
783 while ((*sp->pptr)->rb_left)
784 *sp->pptr = (*sp->pptr)->rb_left;
785 if (sp > stk) {
786 sp--;
787 if (!*sp->pptr) {
788 WARN_ON(1);
789 return NULL;
790 }
791 htb_next_rb_node(sp->pptr);
792 }
793 } else {
794 struct htb_class *cl;
795 struct htb_prio *clp;
796
797 cl = rb_entry(*sp->pptr, struct htb_class, node[prio]);
798 if (!cl->level)
799 return cl;
800 clp = &cl->un.inner.clprio[prio];
801 (++sp)->root = clp->feed.rb_node;
802 sp->pptr = &clp->ptr;
803 sp->pid = &clp->last_ptr_id;
804 }
805 }
806 WARN_ON(1);
807 return NULL;
808}
809
810/* dequeues packet at given priority and level; call only if
811 * you are sure that there is active class at prio/level
812 */
813static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, const int prio,
814 const int level)
815{
816 struct sk_buff *skb = NULL;
817 struct htb_class *cl, *start;
818 struct htb_level *hlevel = &q->hlevel[level];
819 struct htb_prio *hprio = &hlevel->hprio[prio];
820
821 /* look initial class up in the row */
822 start = cl = htb_lookup_leaf(hprio, prio);
823
824 do {
825next:
826 if (unlikely(!cl))
827 return NULL;
828
829 /* class can be empty - it is unlikely but can be true if leaf
830 * qdisc drops packets in enqueue routine or if someone used
831 * graft operation on the leaf since last dequeue;
832 * simply deactivate and skip such class
833 */
834 if (unlikely(cl->un.leaf.q->q.qlen == 0)) {
835 struct htb_class *next;
836 htb_deactivate(q, cl);
837
838 /* row/level might become empty */
839 if ((q->row_mask[level] & (1 << prio)) == 0)
840 return NULL;
841
842 next = htb_lookup_leaf(hprio, prio);
843
844 if (cl == start) /* fix start if we just deleted it */
845 start = next;
846 cl = next;
847 goto next;
848 }
849
850 skb = cl->un.leaf.q->dequeue(cl->un.leaf.q);
851 if (likely(skb != NULL))
852 break;
853
854 qdisc_warn_nonwc("htb", cl->un.leaf.q);
855 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr:
856 &q->hlevel[0].hprio[prio].ptr);
857 cl = htb_lookup_leaf(hprio, prio);
858
859 } while (cl != start);
860
861 if (likely(skb != NULL)) {
862 bstats_update(&cl->bstats, skb);
863 cl->un.leaf.deficit[level] -= qdisc_pkt_len(skb);
864 if (cl->un.leaf.deficit[level] < 0) {
865 cl->un.leaf.deficit[level] += cl->quantum;
866 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr :
867 &q->hlevel[0].hprio[prio].ptr);
868 }
869 /* this used to be after charge_class but this constelation
870 * gives us slightly better performance
871 */
872 if (!cl->un.leaf.q->q.qlen)
873 htb_deactivate(q, cl);
874 htb_charge_class(q, cl, level, skb);
875 }
876 return skb;
877}
878
879static struct sk_buff *htb_dequeue(struct Qdisc *sch)
880{
881 struct sk_buff *skb;
882 struct htb_sched *q = qdisc_priv(sch);
883 int level;
884 s64 next_event;
885 unsigned long start_at;
886
887 /* try to dequeue direct packets as high prio (!) to minimize cpu work */
888 skb = __skb_dequeue(&q->direct_queue);
889 if (skb != NULL) {
890ok:
891 qdisc_bstats_update(sch, skb);
892 qdisc_unthrottled(sch);
893 qdisc_qstats_backlog_dec(sch, skb);
894 sch->q.qlen--;
895 return skb;
896 }
897
898 if (!sch->q.qlen)
899 goto fin;
900 q->now = ktime_get_ns();
901 start_at = jiffies;
902
903 next_event = q->now + 5LLU * NSEC_PER_SEC;
904
905 for (level = 0; level < TC_HTB_MAXDEPTH; level++) {
906 /* common case optimization - skip event handler quickly */
907 int m;
908 s64 event = q->near_ev_cache[level];
909
910 if (q->now >= event) {
911 event = htb_do_events(q, level, start_at);
912 if (!event)
913 event = q->now + NSEC_PER_SEC;
914 q->near_ev_cache[level] = event;
915 }
916
917 if (next_event > event)
918 next_event = event;
919
920 m = ~q->row_mask[level];
921 while (m != (int)(-1)) {
922 int prio = ffz(m);
923
924 m |= 1 << prio;
925 skb = htb_dequeue_tree(q, prio, level);
926 if (likely(skb != NULL))
927 goto ok;
928 }
929 }
930 qdisc_qstats_overlimit(sch);
931 if (likely(next_event > q->now)) {
932 if (!test_bit(__QDISC_STATE_DEACTIVATED,
933 &qdisc_root_sleeping(q->watchdog.qdisc)->state)) {
934 ktime_t time = ns_to_ktime(next_event);
935 qdisc_throttled(q->watchdog.qdisc);
936 hrtimer_start(&q->watchdog.timer, time,
937 HRTIMER_MODE_ABS_PINNED);
938 }
939 } else {
940 schedule_work(&q->work);
941 }
942fin:
943 return skb;
944}
945
946/* try to drop from each class (by prio) until one succeed */
947static unsigned int htb_drop(struct Qdisc *sch)
948{
949 struct htb_sched *q = qdisc_priv(sch);
950 int prio;
951
952 for (prio = TC_HTB_NUMPRIO - 1; prio >= 0; prio--) {
953 struct list_head *p;
954 list_for_each(p, q->drops + prio) {
955 struct htb_class *cl = list_entry(p, struct htb_class,
956 un.leaf.drop_list);
957 unsigned int len;
958 if (cl->un.leaf.q->ops->drop &&
959 (len = cl->un.leaf.q->ops->drop(cl->un.leaf.q))) {
960 sch->qstats.backlog -= len;
961 sch->q.qlen--;
962 if (!cl->un.leaf.q->q.qlen)
963 htb_deactivate(q, cl);
964 return len;
965 }
966 }
967 }
968 return 0;
969}
970
971/* reset all classes */
972/* always caled under BH & queue lock */
973static void htb_reset(struct Qdisc *sch)
974{
975 struct htb_sched *q = qdisc_priv(sch);
976 struct htb_class *cl;
977 unsigned int i;
978
979 for (i = 0; i < q->clhash.hashsize; i++) {
980 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
981 if (cl->level)
982 memset(&cl->un.inner, 0, sizeof(cl->un.inner));
983 else {
984 if (cl->un.leaf.q)
985 qdisc_reset(cl->un.leaf.q);
986 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
987 }
988 cl->prio_activity = 0;
989 cl->cmode = HTB_CAN_SEND;
990 }
991 }
992 qdisc_watchdog_cancel(&q->watchdog);
993 __skb_queue_purge(&q->direct_queue);
994 sch->q.qlen = 0;
995 sch->qstats.backlog = 0;
996 memset(q->hlevel, 0, sizeof(q->hlevel));
997 memset(q->row_mask, 0, sizeof(q->row_mask));
998 for (i = 0; i < TC_HTB_NUMPRIO; i++)
999 INIT_LIST_HEAD(q->drops + i);
1000}
1001
1002static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = {
1003 [TCA_HTB_PARMS] = { .len = sizeof(struct tc_htb_opt) },
1004 [TCA_HTB_INIT] = { .len = sizeof(struct tc_htb_glob) },
1005 [TCA_HTB_CTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
1006 [TCA_HTB_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
1007 [TCA_HTB_DIRECT_QLEN] = { .type = NLA_U32 },
1008 [TCA_HTB_RATE64] = { .type = NLA_U64 },
1009 [TCA_HTB_CEIL64] = { .type = NLA_U64 },
1010};
1011
1012static void htb_work_func(struct work_struct *work)
1013{
1014 struct htb_sched *q = container_of(work, struct htb_sched, work);
1015 struct Qdisc *sch = q->watchdog.qdisc;
1016
1017 __netif_schedule(qdisc_root(sch));
1018}
1019
1020static int htb_init(struct Qdisc *sch, struct nlattr *opt)
1021{
1022 struct htb_sched *q = qdisc_priv(sch);
1023 struct nlattr *tb[TCA_HTB_MAX + 1];
1024 struct tc_htb_glob *gopt;
1025 int err;
1026 int i;
1027
1028 if (!opt)
1029 return -EINVAL;
1030
1031 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy);
1032 if (err < 0)
1033 return err;
1034
1035 if (!tb[TCA_HTB_INIT])
1036 return -EINVAL;
1037
1038 gopt = nla_data(tb[TCA_HTB_INIT]);
1039 if (gopt->version != HTB_VER >> 16)
1040 return -EINVAL;
1041
1042 err = qdisc_class_hash_init(&q->clhash);
1043 if (err < 0)
1044 return err;
1045 for (i = 0; i < TC_HTB_NUMPRIO; i++)
1046 INIT_LIST_HEAD(q->drops + i);
1047
1048 qdisc_watchdog_init(&q->watchdog, sch);
1049 INIT_WORK(&q->work, htb_work_func);
1050 __skb_queue_head_init(&q->direct_queue);
1051
1052 if (tb[TCA_HTB_DIRECT_QLEN])
1053 q->direct_qlen = nla_get_u32(tb[TCA_HTB_DIRECT_QLEN]);
1054 else
1055 q->direct_qlen = qdisc_dev(sch)->tx_queue_len;
1056
1057 if ((q->rate2quantum = gopt->rate2quantum) < 1)
1058 q->rate2quantum = 1;
1059 q->defcls = gopt->defcls;
1060
1061 return 0;
1062}
1063
1064static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
1065{
1066 struct htb_sched *q = qdisc_priv(sch);
1067 struct nlattr *nest;
1068 struct tc_htb_glob gopt;
1069
1070 /* Its safe to not acquire qdisc lock. As we hold RTNL,
1071 * no change can happen on the qdisc parameters.
1072 */
1073
1074 gopt.direct_pkts = q->direct_pkts;
1075 gopt.version = HTB_VER;
1076 gopt.rate2quantum = q->rate2quantum;
1077 gopt.defcls = q->defcls;
1078 gopt.debug = 0;
1079
1080 nest = nla_nest_start(skb, TCA_OPTIONS);
1081 if (nest == NULL)
1082 goto nla_put_failure;
1083 if (nla_put(skb, TCA_HTB_INIT, sizeof(gopt), &gopt) ||
1084 nla_put_u32(skb, TCA_HTB_DIRECT_QLEN, q->direct_qlen))
1085 goto nla_put_failure;
1086
1087 return nla_nest_end(skb, nest);
1088
1089nla_put_failure:
1090 nla_nest_cancel(skb, nest);
1091 return -1;
1092}
1093
1094static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
1095 struct sk_buff *skb, struct tcmsg *tcm)
1096{
1097 struct htb_class *cl = (struct htb_class *)arg;
1098 struct nlattr *nest;
1099 struct tc_htb_opt opt;
1100
1101 /* Its safe to not acquire qdisc lock. As we hold RTNL,
1102 * no change can happen on the class parameters.
1103 */
1104 tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT;
1105 tcm->tcm_handle = cl->common.classid;
1106 if (!cl->level && cl->un.leaf.q)
1107 tcm->tcm_info = cl->un.leaf.q->handle;
1108
1109 nest = nla_nest_start(skb, TCA_OPTIONS);
1110 if (nest == NULL)
1111 goto nla_put_failure;
1112
1113 memset(&opt, 0, sizeof(opt));
1114
1115 psched_ratecfg_getrate(&opt.rate, &cl->rate);
1116 opt.buffer = PSCHED_NS2TICKS(cl->buffer);
1117 psched_ratecfg_getrate(&opt.ceil, &cl->ceil);
1118 opt.cbuffer = PSCHED_NS2TICKS(cl->cbuffer);
1119 opt.quantum = cl->quantum;
1120 opt.prio = cl->prio;
1121 opt.level = cl->level;
1122 if (nla_put(skb, TCA_HTB_PARMS, sizeof(opt), &opt))
1123 goto nla_put_failure;
1124 if ((cl->rate.rate_bytes_ps >= (1ULL << 32)) &&
1125 nla_put_u64(skb, TCA_HTB_RATE64, cl->rate.rate_bytes_ps))
1126 goto nla_put_failure;
1127 if ((cl->ceil.rate_bytes_ps >= (1ULL << 32)) &&
1128 nla_put_u64(skb, TCA_HTB_CEIL64, cl->ceil.rate_bytes_ps))
1129 goto nla_put_failure;
1130
1131 return nla_nest_end(skb, nest);
1132
1133nla_put_failure:
1134 nla_nest_cancel(skb, nest);
1135 return -1;
1136}
1137
1138static int
1139htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
1140{
1141 struct htb_class *cl = (struct htb_class *)arg;
1142 __u32 qlen = 0;
1143
1144 if (!cl->level && cl->un.leaf.q)
1145 qlen = cl->un.leaf.q->q.qlen;
1146 cl->xstats.tokens = PSCHED_NS2TICKS(cl->tokens);
1147 cl->xstats.ctokens = PSCHED_NS2TICKS(cl->ctokens);
1148
1149 if (gnet_stats_copy_basic(d, NULL, &cl->bstats) < 0 ||
1150 gnet_stats_copy_rate_est(d, NULL, &cl->rate_est) < 0 ||
1151 gnet_stats_copy_queue(d, NULL, &cl->qstats, qlen) < 0)
1152 return -1;
1153
1154 return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
1155}
1156
1157static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
1158 struct Qdisc **old)
1159{
1160 struct htb_class *cl = (struct htb_class *)arg;
1161
1162 if (cl->level)
1163 return -EINVAL;
1164 if (new == NULL &&
1165 (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1166 cl->common.classid)) == NULL)
1167 return -ENOBUFS;
1168
1169 *old = qdisc_replace(sch, new, &cl->un.leaf.q);
1170 return 0;
1171}
1172
1173static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg)
1174{
1175 struct htb_class *cl = (struct htb_class *)arg;
1176 return !cl->level ? cl->un.leaf.q : NULL;
1177}
1178
1179static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
1180{
1181 struct htb_class *cl = (struct htb_class *)arg;
1182
1183 if (cl->un.leaf.q->q.qlen == 0)
1184 htb_deactivate(qdisc_priv(sch), cl);
1185}
1186
1187static unsigned long htb_get(struct Qdisc *sch, u32 classid)
1188{
1189 struct htb_class *cl = htb_find(classid, sch);
1190 if (cl)
1191 cl->refcnt++;
1192 return (unsigned long)cl;
1193}
1194
1195static inline int htb_parent_last_child(struct htb_class *cl)
1196{
1197 if (!cl->parent)
1198 /* the root class */
1199 return 0;
1200 if (cl->parent->children > 1)
1201 /* not the last child */
1202 return 0;
1203 return 1;
1204}
1205
1206static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl,
1207 struct Qdisc *new_q)
1208{
1209 struct htb_class *parent = cl->parent;
1210
1211 WARN_ON(cl->level || !cl->un.leaf.q || cl->prio_activity);
1212
1213 if (parent->cmode != HTB_CAN_SEND)
1214 htb_safe_rb_erase(&parent->pq_node,
1215 &q->hlevel[parent->level].wait_pq);
1216
1217 parent->level = 0;
1218 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1219 INIT_LIST_HEAD(&parent->un.leaf.drop_list);
1220 parent->un.leaf.q = new_q ? new_q : &noop_qdisc;
1221 parent->tokens = parent->buffer;
1222 parent->ctokens = parent->cbuffer;
1223 parent->t_c = ktime_get_ns();
1224 parent->cmode = HTB_CAN_SEND;
1225}
1226
1227static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl)
1228{
1229 if (!cl->level) {
1230 WARN_ON(!cl->un.leaf.q);
1231 qdisc_destroy(cl->un.leaf.q);
1232 }
1233 gen_kill_estimator(&cl->bstats, &cl->rate_est);
1234 tcf_destroy_chain(&cl->filter_list);
1235 kfree(cl);
1236}
1237
1238static void htb_destroy(struct Qdisc *sch)
1239{
1240 struct htb_sched *q = qdisc_priv(sch);
1241 struct hlist_node *next;
1242 struct htb_class *cl;
1243 unsigned int i;
1244
1245 cancel_work_sync(&q->work);
1246 qdisc_watchdog_cancel(&q->watchdog);
1247 /* This line used to be after htb_destroy_class call below
1248 * and surprisingly it worked in 2.4. But it must precede it
1249 * because filter need its target class alive to be able to call
1250 * unbind_filter on it (without Oops).
1251 */
1252 tcf_destroy_chain(&q->filter_list);
1253
1254 for (i = 0; i < q->clhash.hashsize; i++) {
1255 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode)
1256 tcf_destroy_chain(&cl->filter_list);
1257 }
1258 for (i = 0; i < q->clhash.hashsize; i++) {
1259 hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i],
1260 common.hnode)
1261 htb_destroy_class(sch, cl);
1262 }
1263 qdisc_class_hash_destroy(&q->clhash);
1264 __skb_queue_purge(&q->direct_queue);
1265}
1266
1267static int htb_delete(struct Qdisc *sch, unsigned long arg)
1268{
1269 struct htb_sched *q = qdisc_priv(sch);
1270 struct htb_class *cl = (struct htb_class *)arg;
1271 struct Qdisc *new_q = NULL;
1272 int last_child = 0;
1273
1274 /* TODO: why don't allow to delete subtree ? references ? does
1275 * tc subsys guarantee us that in htb_destroy it holds no class
1276 * refs so that we can remove children safely there ?
1277 */
1278 if (cl->children || cl->filter_cnt)
1279 return -EBUSY;
1280
1281 if (!cl->level && htb_parent_last_child(cl)) {
1282 new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1283 cl->parent->common.classid);
1284 last_child = 1;
1285 }
1286
1287 sch_tree_lock(sch);
1288
1289 if (!cl->level) {
1290 unsigned int qlen = cl->un.leaf.q->q.qlen;
1291 unsigned int backlog = cl->un.leaf.q->qstats.backlog;
1292
1293 qdisc_reset(cl->un.leaf.q);
1294 qdisc_tree_reduce_backlog(cl->un.leaf.q, qlen, backlog);
1295 }
1296
1297 /* delete from hash and active; remainder in destroy_class */
1298 qdisc_class_hash_remove(&q->clhash, &cl->common);
1299 if (cl->parent)
1300 cl->parent->children--;
1301
1302 if (cl->prio_activity)
1303 htb_deactivate(q, cl);
1304
1305 if (cl->cmode != HTB_CAN_SEND)
1306 htb_safe_rb_erase(&cl->pq_node,
1307 &q->hlevel[cl->level].wait_pq);
1308
1309 if (last_child)
1310 htb_parent_to_leaf(q, cl, new_q);
1311
1312 BUG_ON(--cl->refcnt == 0);
1313 /*
1314 * This shouldn't happen: we "hold" one cops->get() when called
1315 * from tc_ctl_tclass; the destroy method is done from cops->put().
1316 */
1317
1318 sch_tree_unlock(sch);
1319 return 0;
1320}
1321
1322static void htb_put(struct Qdisc *sch, unsigned long arg)
1323{
1324 struct htb_class *cl = (struct htb_class *)arg;
1325
1326 if (--cl->refcnt == 0)
1327 htb_destroy_class(sch, cl);
1328}
1329
1330static int htb_change_class(struct Qdisc *sch, u32 classid,
1331 u32 parentid, struct nlattr **tca,
1332 unsigned long *arg)
1333{
1334 int err = -EINVAL;
1335 struct htb_sched *q = qdisc_priv(sch);
1336 struct htb_class *cl = (struct htb_class *)*arg, *parent;
1337 struct nlattr *opt = tca[TCA_OPTIONS];
1338 struct nlattr *tb[TCA_HTB_MAX + 1];
1339 struct tc_htb_opt *hopt;
1340 u64 rate64, ceil64;
1341
1342 /* extract all subattrs from opt attr */
1343 if (!opt)
1344 goto failure;
1345
1346 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy);
1347 if (err < 0)
1348 goto failure;
1349
1350 err = -EINVAL;
1351 if (tb[TCA_HTB_PARMS] == NULL)
1352 goto failure;
1353
1354 parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch);
1355
1356 hopt = nla_data(tb[TCA_HTB_PARMS]);
1357 if (!hopt->rate.rate || !hopt->ceil.rate)
1358 goto failure;
1359
1360 /* Keeping backward compatible with rate_table based iproute2 tc */
1361 if (hopt->rate.linklayer == TC_LINKLAYER_UNAWARE)
1362 qdisc_put_rtab(qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB]));
1363
1364 if (hopt->ceil.linklayer == TC_LINKLAYER_UNAWARE)
1365 qdisc_put_rtab(qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB]));
1366
1367 if (!cl) { /* new class */
1368 struct Qdisc *new_q;
1369 int prio;
1370 struct {
1371 struct nlattr nla;
1372 struct gnet_estimator opt;
1373 } est = {
1374 .nla = {
1375 .nla_len = nla_attr_size(sizeof(est.opt)),
1376 .nla_type = TCA_RATE,
1377 },
1378 .opt = {
1379 /* 4s interval, 16s averaging constant */
1380 .interval = 2,
1381 .ewma_log = 2,
1382 },
1383 };
1384
1385 /* check for valid classid */
1386 if (!classid || TC_H_MAJ(classid ^ sch->handle) ||
1387 htb_find(classid, sch))
1388 goto failure;
1389
1390 /* check maximal depth */
1391 if (parent && parent->parent && parent->parent->level < 2) {
1392 pr_err("htb: tree is too deep\n");
1393 goto failure;
1394 }
1395 err = -ENOBUFS;
1396 cl = kzalloc(sizeof(*cl), GFP_KERNEL);
1397 if (!cl)
1398 goto failure;
1399
1400 if (htb_rate_est || tca[TCA_RATE]) {
1401 err = gen_new_estimator(&cl->bstats, NULL,
1402 &cl->rate_est,
1403 qdisc_root_sleeping_lock(sch),
1404 tca[TCA_RATE] ? : &est.nla);
1405 if (err) {
1406 kfree(cl);
1407 goto failure;
1408 }
1409 }
1410
1411 cl->refcnt = 1;
1412 cl->children = 0;
1413 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
1414 RB_CLEAR_NODE(&cl->pq_node);
1415
1416 for (prio = 0; prio < TC_HTB_NUMPRIO; prio++)
1417 RB_CLEAR_NODE(&cl->node[prio]);
1418
1419 /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL)
1420 * so that can't be used inside of sch_tree_lock
1421 * -- thanks to Karlis Peisenieks
1422 */
1423 new_q = qdisc_create_dflt(sch->dev_queue,
1424 &pfifo_qdisc_ops, classid);
1425 sch_tree_lock(sch);
1426 if (parent && !parent->level) {
1427 unsigned int qlen = parent->un.leaf.q->q.qlen;
1428 unsigned int backlog = parent->un.leaf.q->qstats.backlog;
1429
1430 /* turn parent into inner node */
1431 qdisc_reset(parent->un.leaf.q);
1432 qdisc_tree_reduce_backlog(parent->un.leaf.q, qlen, backlog);
1433 qdisc_destroy(parent->un.leaf.q);
1434 if (parent->prio_activity)
1435 htb_deactivate(q, parent);
1436
1437 /* remove from evt list because of level change */
1438 if (parent->cmode != HTB_CAN_SEND) {
1439 htb_safe_rb_erase(&parent->pq_node, &q->hlevel[0].wait_pq);
1440 parent->cmode = HTB_CAN_SEND;
1441 }
1442 parent->level = (parent->parent ? parent->parent->level
1443 : TC_HTB_MAXDEPTH) - 1;
1444 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1445 }
1446 /* leaf (we) needs elementary qdisc */
1447 cl->un.leaf.q = new_q ? new_q : &noop_qdisc;
1448
1449 cl->common.classid = classid;
1450 cl->parent = parent;
1451
1452 /* set class to be in HTB_CAN_SEND state */
1453 cl->tokens = PSCHED_TICKS2NS(hopt->buffer);
1454 cl->ctokens = PSCHED_TICKS2NS(hopt->cbuffer);
1455 cl->mbuffer = 60ULL * NSEC_PER_SEC; /* 1min */
1456 cl->t_c = ktime_get_ns();
1457 cl->cmode = HTB_CAN_SEND;
1458
1459 /* attach to the hash list and parent's family */
1460 qdisc_class_hash_insert(&q->clhash, &cl->common);
1461 if (parent)
1462 parent->children++;
1463 } else {
1464 if (tca[TCA_RATE]) {
1465 spinlock_t *lock = qdisc_root_sleeping_lock(sch);
1466
1467 err = gen_replace_estimator(&cl->bstats, NULL,
1468 &cl->rate_est,
1469 lock,
1470 tca[TCA_RATE]);
1471 if (err)
1472 return err;
1473 }
1474 sch_tree_lock(sch);
1475 }
1476
1477 rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0;
1478
1479 ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0;
1480
1481 psched_ratecfg_precompute(&cl->rate, &hopt->rate, rate64);
1482 psched_ratecfg_precompute(&cl->ceil, &hopt->ceil, ceil64);
1483
1484 /* it used to be a nasty bug here, we have to check that node
1485 * is really leaf before changing cl->un.leaf !
1486 */
1487 if (!cl->level) {
1488 u64 quantum = cl->rate.rate_bytes_ps;
1489
1490 do_div(quantum, q->rate2quantum);
1491 cl->quantum = min_t(u64, quantum, INT_MAX);
1492
1493 if (!hopt->quantum && cl->quantum < 1000) {
1494 pr_warn("HTB: quantum of class %X is small. Consider r2q change.\n",
1495 cl->common.classid);
1496 cl->quantum = 1000;
1497 }
1498 if (!hopt->quantum && cl->quantum > 200000) {
1499 pr_warn("HTB: quantum of class %X is big. Consider r2q change.\n",
1500 cl->common.classid);
1501 cl->quantum = 200000;
1502 }
1503 if (hopt->quantum)
1504 cl->quantum = hopt->quantum;
1505 if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO)
1506 cl->prio = TC_HTB_NUMPRIO - 1;
1507 }
1508
1509 cl->buffer = PSCHED_TICKS2NS(hopt->buffer);
1510 cl->cbuffer = PSCHED_TICKS2NS(hopt->cbuffer);
1511
1512 sch_tree_unlock(sch);
1513
1514 qdisc_class_hash_grow(sch, &q->clhash);
1515
1516 *arg = (unsigned long)cl;
1517 return 0;
1518
1519failure:
1520 return err;
1521}
1522
1523static struct tcf_proto __rcu **htb_find_tcf(struct Qdisc *sch,
1524 unsigned long arg)
1525{
1526 struct htb_sched *q = qdisc_priv(sch);
1527 struct htb_class *cl = (struct htb_class *)arg;
1528 struct tcf_proto __rcu **fl = cl ? &cl->filter_list : &q->filter_list;
1529
1530 return fl;
1531}
1532
1533static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent,
1534 u32 classid)
1535{
1536 struct htb_class *cl = htb_find(classid, sch);
1537
1538 /*if (cl && !cl->level) return 0;
1539 * The line above used to be there to prevent attaching filters to
1540 * leaves. But at least tc_index filter uses this just to get class
1541 * for other reasons so that we have to allow for it.
1542 * ----
1543 * 19.6.2002 As Werner explained it is ok - bind filter is just
1544 * another way to "lock" the class - unlike "get" this lock can
1545 * be broken by class during destroy IIUC.
1546 */
1547 if (cl)
1548 cl->filter_cnt++;
1549 return (unsigned long)cl;
1550}
1551
1552static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg)
1553{
1554 struct htb_class *cl = (struct htb_class *)arg;
1555
1556 if (cl)
1557 cl->filter_cnt--;
1558}
1559
1560static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg)
1561{
1562 struct htb_sched *q = qdisc_priv(sch);
1563 struct htb_class *cl;
1564 unsigned int i;
1565
1566 if (arg->stop)
1567 return;
1568
1569 for (i = 0; i < q->clhash.hashsize; i++) {
1570 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
1571 if (arg->count < arg->skip) {
1572 arg->count++;
1573 continue;
1574 }
1575 if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
1576 arg->stop = 1;
1577 return;
1578 }
1579 arg->count++;
1580 }
1581 }
1582}
1583
1584static const struct Qdisc_class_ops htb_class_ops = {
1585 .graft = htb_graft,
1586 .leaf = htb_leaf,
1587 .qlen_notify = htb_qlen_notify,
1588 .get = htb_get,
1589 .put = htb_put,
1590 .change = htb_change_class,
1591 .delete = htb_delete,
1592 .walk = htb_walk,
1593 .tcf_chain = htb_find_tcf,
1594 .bind_tcf = htb_bind_filter,
1595 .unbind_tcf = htb_unbind_filter,
1596 .dump = htb_dump_class,
1597 .dump_stats = htb_dump_class_stats,
1598};
1599
1600static struct Qdisc_ops htb_qdisc_ops __read_mostly = {
1601 .cl_ops = &htb_class_ops,
1602 .id = "htb",
1603 .priv_size = sizeof(struct htb_sched),
1604 .enqueue = htb_enqueue,
1605 .dequeue = htb_dequeue,
1606 .peek = qdisc_peek_dequeued,
1607 .drop = htb_drop,
1608 .init = htb_init,
1609 .reset = htb_reset,
1610 .destroy = htb_destroy,
1611 .dump = htb_dump,
1612 .owner = THIS_MODULE,
1613};
1614
1615static int __init htb_module_init(void)
1616{
1617 return register_qdisc(&htb_qdisc_ops);
1618}
1619static void __exit htb_module_exit(void)
1620{
1621 unregister_qdisc(&htb_qdisc_ops);
1622}
1623
1624module_init(htb_module_init)
1625module_exit(htb_module_exit)
1626MODULE_LICENSE("GPL");
1/*
2 * net/sched/sch_htb.c Hierarchical token bucket, feed tree version
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Authors: Martin Devera, <devik@cdi.cz>
10 *
11 * Credits (in time order) for older HTB versions:
12 * Stef Coene <stef.coene@docum.org>
13 * HTB support at LARTC mailing list
14 * Ondrej Kraus, <krauso@barr.cz>
15 * found missing INIT_QDISC(htb)
16 * Vladimir Smelhaus, Aamer Akhter, Bert Hubert
17 * helped a lot to locate nasty class stall bug
18 * Andi Kleen, Jamal Hadi, Bert Hubert
19 * code review and helpful comments on shaping
20 * Tomasz Wrona, <tw@eter.tym.pl>
21 * created test case so that I was able to fix nasty bug
22 * Wilfried Weissmann
23 * spotted bug in dequeue code and helped with fix
24 * Jiri Fojtasek
25 * fixed requeue routine
26 * and many others. thanks.
27 */
28#include <linux/module.h>
29#include <linux/moduleparam.h>
30#include <linux/types.h>
31#include <linux/kernel.h>
32#include <linux/string.h>
33#include <linux/errno.h>
34#include <linux/skbuff.h>
35#include <linux/list.h>
36#include <linux/compiler.h>
37#include <linux/rbtree.h>
38#include <linux/workqueue.h>
39#include <linux/slab.h>
40#include <net/netlink.h>
41#include <net/sch_generic.h>
42#include <net/pkt_sched.h>
43#include <net/pkt_cls.h>
44
45/* HTB algorithm.
46 Author: devik@cdi.cz
47 ========================================================================
48 HTB is like TBF with multiple classes. It is also similar to CBQ because
49 it allows to assign priority to each class in hierarchy.
50 In fact it is another implementation of Floyd's formal sharing.
51
52 Levels:
53 Each class is assigned level. Leaf has ALWAYS level 0 and root
54 classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level
55 one less than their parent.
56*/
57
58static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */
59#define HTB_VER 0x30011 /* major must be matched with number suplied by TC as version */
60
61#if HTB_VER >> 16 != TC_HTB_PROTOVER
62#error "Mismatched sch_htb.c and pkt_sch.h"
63#endif
64
65/* Module parameter and sysfs export */
66module_param (htb_hysteresis, int, 0640);
67MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate");
68
69static int htb_rate_est = 0; /* htb classes have a default rate estimator */
70module_param(htb_rate_est, int, 0640);
71MODULE_PARM_DESC(htb_rate_est, "setup a default rate estimator (4sec 16sec) for htb classes");
72
73/* used internaly to keep status of single class */
74enum htb_cmode {
75 HTB_CANT_SEND, /* class can't send and can't borrow */
76 HTB_MAY_BORROW, /* class can't send but may borrow */
77 HTB_CAN_SEND /* class can send */
78};
79
80struct htb_prio {
81 union {
82 struct rb_root row;
83 struct rb_root feed;
84 };
85 struct rb_node *ptr;
86 /* When class changes from state 1->2 and disconnects from
87 * parent's feed then we lost ptr value and start from the
88 * first child again. Here we store classid of the
89 * last valid ptr (used when ptr is NULL).
90 */
91 u32 last_ptr_id;
92};
93
94/* interior & leaf nodes; props specific to leaves are marked L:
95 * To reduce false sharing, place mostly read fields at beginning,
96 * and mostly written ones at the end.
97 */
98struct htb_class {
99 struct Qdisc_class_common common;
100 struct psched_ratecfg rate;
101 struct psched_ratecfg ceil;
102 s64 buffer, cbuffer;/* token bucket depth/rate */
103 s64 mbuffer; /* max wait time */
104 u32 prio; /* these two are used only by leaves... */
105 int quantum; /* but stored for parent-to-leaf return */
106
107 struct tcf_proto __rcu *filter_list; /* class attached filters */
108 struct tcf_block *block;
109 int filter_cnt;
110
111 int level; /* our level (see above) */
112 unsigned int children;
113 struct htb_class *parent; /* parent class */
114
115 struct net_rate_estimator __rcu *rate_est;
116
117 /*
118 * Written often fields
119 */
120 struct gnet_stats_basic_packed bstats;
121 struct tc_htb_xstats xstats; /* our special stats */
122
123 /* token bucket parameters */
124 s64 tokens, ctokens;/* current number of tokens */
125 s64 t_c; /* checkpoint time */
126
127 union {
128 struct htb_class_leaf {
129 struct list_head drop_list;
130 int deficit[TC_HTB_MAXDEPTH];
131 struct Qdisc *q;
132 } leaf;
133 struct htb_class_inner {
134 struct htb_prio clprio[TC_HTB_NUMPRIO];
135 } inner;
136 } un;
137 s64 pq_key;
138
139 int prio_activity; /* for which prios are we active */
140 enum htb_cmode cmode; /* current mode of the class */
141 struct rb_node pq_node; /* node for event queue */
142 struct rb_node node[TC_HTB_NUMPRIO]; /* node for self or feed tree */
143
144 unsigned int drops ____cacheline_aligned_in_smp;
145 unsigned int overlimits;
146};
147
148struct htb_level {
149 struct rb_root wait_pq;
150 struct htb_prio hprio[TC_HTB_NUMPRIO];
151};
152
153struct htb_sched {
154 struct Qdisc_class_hash clhash;
155 int defcls; /* class where unclassified flows go to */
156 int rate2quantum; /* quant = rate / rate2quantum */
157
158 /* filters for qdisc itself */
159 struct tcf_proto __rcu *filter_list;
160 struct tcf_block *block;
161
162#define HTB_WARN_TOOMANYEVENTS 0x1
163 unsigned int warned; /* only one warning */
164 int direct_qlen;
165 struct work_struct work;
166
167 /* non shaped skbs; let them go directly thru */
168 struct qdisc_skb_head direct_queue;
169 long direct_pkts;
170
171 struct qdisc_watchdog watchdog;
172
173 s64 now; /* cached dequeue time */
174 struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */
175
176 /* time of nearest event per level (row) */
177 s64 near_ev_cache[TC_HTB_MAXDEPTH];
178
179 int row_mask[TC_HTB_MAXDEPTH];
180
181 struct htb_level hlevel[TC_HTB_MAXDEPTH];
182};
183
184/* find class in global hash table using given handle */
185static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch)
186{
187 struct htb_sched *q = qdisc_priv(sch);
188 struct Qdisc_class_common *clc;
189
190 clc = qdisc_class_find(&q->clhash, handle);
191 if (clc == NULL)
192 return NULL;
193 return container_of(clc, struct htb_class, common);
194}
195
196static unsigned long htb_search(struct Qdisc *sch, u32 handle)
197{
198 return (unsigned long)htb_find(handle, sch);
199}
200/**
201 * htb_classify - classify a packet into class
202 *
203 * It returns NULL if the packet should be dropped or -1 if the packet
204 * should be passed directly thru. In all other cases leaf class is returned.
205 * We allow direct class selection by classid in priority. The we examine
206 * filters in qdisc and in inner nodes (if higher filter points to the inner
207 * node). If we end up with classid MAJOR:0 we enqueue the skb into special
208 * internal fifo (direct). These packets then go directly thru. If we still
209 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessful
210 * then finish and return direct queue.
211 */
212#define HTB_DIRECT ((struct htb_class *)-1L)
213
214static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
215 int *qerr)
216{
217 struct htb_sched *q = qdisc_priv(sch);
218 struct htb_class *cl;
219 struct tcf_result res;
220 struct tcf_proto *tcf;
221 int result;
222
223 /* allow to select class by setting skb->priority to valid classid;
224 * note that nfmark can be used too by attaching filter fw with no
225 * rules in it
226 */
227 if (skb->priority == sch->handle)
228 return HTB_DIRECT; /* X:0 (direct flow) selected */
229 cl = htb_find(skb->priority, sch);
230 if (cl) {
231 if (cl->level == 0)
232 return cl;
233 /* Start with inner filter chain if a non-leaf class is selected */
234 tcf = rcu_dereference_bh(cl->filter_list);
235 } else {
236 tcf = rcu_dereference_bh(q->filter_list);
237 }
238
239 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
240 while (tcf && (result = tcf_classify(skb, tcf, &res, false)) >= 0) {
241#ifdef CONFIG_NET_CLS_ACT
242 switch (result) {
243 case TC_ACT_QUEUED:
244 case TC_ACT_STOLEN:
245 case TC_ACT_TRAP:
246 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
247 /* fall through */
248 case TC_ACT_SHOT:
249 return NULL;
250 }
251#endif
252 cl = (void *)res.class;
253 if (!cl) {
254 if (res.classid == sch->handle)
255 return HTB_DIRECT; /* X:0 (direct flow) */
256 cl = htb_find(res.classid, sch);
257 if (!cl)
258 break; /* filter selected invalid classid */
259 }
260 if (!cl->level)
261 return cl; /* we hit leaf; return it */
262
263 /* we have got inner class; apply inner filter chain */
264 tcf = rcu_dereference_bh(cl->filter_list);
265 }
266 /* classification failed; try to use default class */
267 cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch);
268 if (!cl || cl->level)
269 return HTB_DIRECT; /* bad default .. this is safe bet */
270 return cl;
271}
272
273/**
274 * htb_add_to_id_tree - adds class to the round robin list
275 *
276 * Routine adds class to the list (actually tree) sorted by classid.
277 * Make sure that class is not already on such list for given prio.
278 */
279static void htb_add_to_id_tree(struct rb_root *root,
280 struct htb_class *cl, int prio)
281{
282 struct rb_node **p = &root->rb_node, *parent = NULL;
283
284 while (*p) {
285 struct htb_class *c;
286 parent = *p;
287 c = rb_entry(parent, struct htb_class, node[prio]);
288
289 if (cl->common.classid > c->common.classid)
290 p = &parent->rb_right;
291 else
292 p = &parent->rb_left;
293 }
294 rb_link_node(&cl->node[prio], parent, p);
295 rb_insert_color(&cl->node[prio], root);
296}
297
298/**
299 * htb_add_to_wait_tree - adds class to the event queue with delay
300 *
301 * The class is added to priority event queue to indicate that class will
302 * change its mode in cl->pq_key microseconds. Make sure that class is not
303 * already in the queue.
304 */
305static void htb_add_to_wait_tree(struct htb_sched *q,
306 struct htb_class *cl, s64 delay)
307{
308 struct rb_node **p = &q->hlevel[cl->level].wait_pq.rb_node, *parent = NULL;
309
310 cl->pq_key = q->now + delay;
311 if (cl->pq_key == q->now)
312 cl->pq_key++;
313
314 /* update the nearest event cache */
315 if (q->near_ev_cache[cl->level] > cl->pq_key)
316 q->near_ev_cache[cl->level] = cl->pq_key;
317
318 while (*p) {
319 struct htb_class *c;
320 parent = *p;
321 c = rb_entry(parent, struct htb_class, pq_node);
322 if (cl->pq_key >= c->pq_key)
323 p = &parent->rb_right;
324 else
325 p = &parent->rb_left;
326 }
327 rb_link_node(&cl->pq_node, parent, p);
328 rb_insert_color(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
329}
330
331/**
332 * htb_next_rb_node - finds next node in binary tree
333 *
334 * When we are past last key we return NULL.
335 * Average complexity is 2 steps per call.
336 */
337static inline void htb_next_rb_node(struct rb_node **n)
338{
339 *n = rb_next(*n);
340}
341
342/**
343 * htb_add_class_to_row - add class to its row
344 *
345 * The class is added to row at priorities marked in mask.
346 * It does nothing if mask == 0.
347 */
348static inline void htb_add_class_to_row(struct htb_sched *q,
349 struct htb_class *cl, int mask)
350{
351 q->row_mask[cl->level] |= mask;
352 while (mask) {
353 int prio = ffz(~mask);
354 mask &= ~(1 << prio);
355 htb_add_to_id_tree(&q->hlevel[cl->level].hprio[prio].row, cl, prio);
356 }
357}
358
359/* If this triggers, it is a bug in this code, but it need not be fatal */
360static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root)
361{
362 if (RB_EMPTY_NODE(rb)) {
363 WARN_ON(1);
364 } else {
365 rb_erase(rb, root);
366 RB_CLEAR_NODE(rb);
367 }
368}
369
370
371/**
372 * htb_remove_class_from_row - removes class from its row
373 *
374 * The class is removed from row at priorities marked in mask.
375 * It does nothing if mask == 0.
376 */
377static inline void htb_remove_class_from_row(struct htb_sched *q,
378 struct htb_class *cl, int mask)
379{
380 int m = 0;
381 struct htb_level *hlevel = &q->hlevel[cl->level];
382
383 while (mask) {
384 int prio = ffz(~mask);
385 struct htb_prio *hprio = &hlevel->hprio[prio];
386
387 mask &= ~(1 << prio);
388 if (hprio->ptr == cl->node + prio)
389 htb_next_rb_node(&hprio->ptr);
390
391 htb_safe_rb_erase(cl->node + prio, &hprio->row);
392 if (!hprio->row.rb_node)
393 m |= 1 << prio;
394 }
395 q->row_mask[cl->level] &= ~m;
396}
397
398/**
399 * htb_activate_prios - creates active classe's feed chain
400 *
401 * The class is connected to ancestors and/or appropriate rows
402 * for priorities it is participating on. cl->cmode must be new
403 * (activated) mode. It does nothing if cl->prio_activity == 0.
404 */
405static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl)
406{
407 struct htb_class *p = cl->parent;
408 long m, mask = cl->prio_activity;
409
410 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
411 m = mask;
412 while (m) {
413 int prio = ffz(~m);
414 m &= ~(1 << prio);
415
416 if (p->un.inner.clprio[prio].feed.rb_node)
417 /* parent already has its feed in use so that
418 * reset bit in mask as parent is already ok
419 */
420 mask &= ~(1 << prio);
421
422 htb_add_to_id_tree(&p->un.inner.clprio[prio].feed, cl, prio);
423 }
424 p->prio_activity |= mask;
425 cl = p;
426 p = cl->parent;
427
428 }
429 if (cl->cmode == HTB_CAN_SEND && mask)
430 htb_add_class_to_row(q, cl, mask);
431}
432
433/**
434 * htb_deactivate_prios - remove class from feed chain
435 *
436 * cl->cmode must represent old mode (before deactivation). It does
437 * nothing if cl->prio_activity == 0. Class is removed from all feed
438 * chains and rows.
439 */
440static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl)
441{
442 struct htb_class *p = cl->parent;
443 long m, mask = cl->prio_activity;
444
445 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
446 m = mask;
447 mask = 0;
448 while (m) {
449 int prio = ffz(~m);
450 m &= ~(1 << prio);
451
452 if (p->un.inner.clprio[prio].ptr == cl->node + prio) {
453 /* we are removing child which is pointed to from
454 * parent feed - forget the pointer but remember
455 * classid
456 */
457 p->un.inner.clprio[prio].last_ptr_id = cl->common.classid;
458 p->un.inner.clprio[prio].ptr = NULL;
459 }
460
461 htb_safe_rb_erase(cl->node + prio,
462 &p->un.inner.clprio[prio].feed);
463
464 if (!p->un.inner.clprio[prio].feed.rb_node)
465 mask |= 1 << prio;
466 }
467
468 p->prio_activity &= ~mask;
469 cl = p;
470 p = cl->parent;
471
472 }
473 if (cl->cmode == HTB_CAN_SEND && mask)
474 htb_remove_class_from_row(q, cl, mask);
475}
476
477static inline s64 htb_lowater(const struct htb_class *cl)
478{
479 if (htb_hysteresis)
480 return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0;
481 else
482 return 0;
483}
484static inline s64 htb_hiwater(const struct htb_class *cl)
485{
486 if (htb_hysteresis)
487 return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0;
488 else
489 return 0;
490}
491
492
493/**
494 * htb_class_mode - computes and returns current class mode
495 *
496 * It computes cl's mode at time cl->t_c+diff and returns it. If mode
497 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference
498 * from now to time when cl will change its state.
499 * Also it is worth to note that class mode doesn't change simply
500 * at cl->{c,}tokens == 0 but there can rather be hysteresis of
501 * 0 .. -cl->{c,}buffer range. It is meant to limit number of
502 * mode transitions per time unit. The speed gain is about 1/6.
503 */
504static inline enum htb_cmode
505htb_class_mode(struct htb_class *cl, s64 *diff)
506{
507 s64 toks;
508
509 if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) {
510 *diff = -toks;
511 return HTB_CANT_SEND;
512 }
513
514 if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl))
515 return HTB_CAN_SEND;
516
517 *diff = -toks;
518 return HTB_MAY_BORROW;
519}
520
521/**
522 * htb_change_class_mode - changes classe's mode
523 *
524 * This should be the only way how to change classe's mode under normal
525 * cirsumstances. Routine will update feed lists linkage, change mode
526 * and add class to the wait event queue if appropriate. New mode should
527 * be different from old one and cl->pq_key has to be valid if changing
528 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree).
529 */
530static void
531htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, s64 *diff)
532{
533 enum htb_cmode new_mode = htb_class_mode(cl, diff);
534
535 if (new_mode == cl->cmode)
536 return;
537
538 if (new_mode == HTB_CANT_SEND)
539 cl->overlimits++;
540
541 if (cl->prio_activity) { /* not necessary: speed optimization */
542 if (cl->cmode != HTB_CANT_SEND)
543 htb_deactivate_prios(q, cl);
544 cl->cmode = new_mode;
545 if (new_mode != HTB_CANT_SEND)
546 htb_activate_prios(q, cl);
547 } else
548 cl->cmode = new_mode;
549}
550
551/**
552 * htb_activate - inserts leaf cl into appropriate active feeds
553 *
554 * Routine learns (new) priority of leaf and activates feed chain
555 * for the prio. It can be called on already active leaf safely.
556 * It also adds leaf into droplist.
557 */
558static inline void htb_activate(struct htb_sched *q, struct htb_class *cl)
559{
560 WARN_ON(cl->level || !cl->un.leaf.q || !cl->un.leaf.q->q.qlen);
561
562 if (!cl->prio_activity) {
563 cl->prio_activity = 1 << cl->prio;
564 htb_activate_prios(q, cl);
565 list_add_tail(&cl->un.leaf.drop_list,
566 q->drops + cl->prio);
567 }
568}
569
570/**
571 * htb_deactivate - remove leaf cl from active feeds
572 *
573 * Make sure that leaf is active. In the other words it can't be called
574 * with non-active leaf. It also removes class from the drop list.
575 */
576static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
577{
578 WARN_ON(!cl->prio_activity);
579
580 htb_deactivate_prios(q, cl);
581 cl->prio_activity = 0;
582 list_del_init(&cl->un.leaf.drop_list);
583}
584
585static void htb_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch,
586 struct qdisc_skb_head *qh)
587{
588 struct sk_buff *last = qh->tail;
589
590 if (last) {
591 skb->next = NULL;
592 last->next = skb;
593 qh->tail = skb;
594 } else {
595 qh->tail = skb;
596 qh->head = skb;
597 }
598 qh->qlen++;
599}
600
601static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
602 struct sk_buff **to_free)
603{
604 int uninitialized_var(ret);
605 struct htb_sched *q = qdisc_priv(sch);
606 struct htb_class *cl = htb_classify(skb, sch, &ret);
607
608 if (cl == HTB_DIRECT) {
609 /* enqueue to helper queue */
610 if (q->direct_queue.qlen < q->direct_qlen) {
611 htb_enqueue_tail(skb, sch, &q->direct_queue);
612 q->direct_pkts++;
613 } else {
614 return qdisc_drop(skb, sch, to_free);
615 }
616#ifdef CONFIG_NET_CLS_ACT
617 } else if (!cl) {
618 if (ret & __NET_XMIT_BYPASS)
619 qdisc_qstats_drop(sch);
620 __qdisc_drop(skb, to_free);
621 return ret;
622#endif
623 } else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q,
624 to_free)) != NET_XMIT_SUCCESS) {
625 if (net_xmit_drop_count(ret)) {
626 qdisc_qstats_drop(sch);
627 cl->drops++;
628 }
629 return ret;
630 } else {
631 htb_activate(q, cl);
632 }
633
634 qdisc_qstats_backlog_inc(sch, skb);
635 sch->q.qlen++;
636 return NET_XMIT_SUCCESS;
637}
638
639static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, s64 diff)
640{
641 s64 toks = diff + cl->tokens;
642
643 if (toks > cl->buffer)
644 toks = cl->buffer;
645 toks -= (s64) psched_l2t_ns(&cl->rate, bytes);
646 if (toks <= -cl->mbuffer)
647 toks = 1 - cl->mbuffer;
648
649 cl->tokens = toks;
650}
651
652static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, s64 diff)
653{
654 s64 toks = diff + cl->ctokens;
655
656 if (toks > cl->cbuffer)
657 toks = cl->cbuffer;
658 toks -= (s64) psched_l2t_ns(&cl->ceil, bytes);
659 if (toks <= -cl->mbuffer)
660 toks = 1 - cl->mbuffer;
661
662 cl->ctokens = toks;
663}
664
665/**
666 * htb_charge_class - charges amount "bytes" to leaf and ancestors
667 *
668 * Routine assumes that packet "bytes" long was dequeued from leaf cl
669 * borrowing from "level". It accounts bytes to ceil leaky bucket for
670 * leaf and all ancestors and to rate bucket for ancestors at levels
671 * "level" and higher. It also handles possible change of mode resulting
672 * from the update. Note that mode can also increase here (MAY_BORROW to
673 * CAN_SEND) because we can use more precise clock that event queue here.
674 * In such case we remove class from event queue first.
675 */
676static void htb_charge_class(struct htb_sched *q, struct htb_class *cl,
677 int level, struct sk_buff *skb)
678{
679 int bytes = qdisc_pkt_len(skb);
680 enum htb_cmode old_mode;
681 s64 diff;
682
683 while (cl) {
684 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
685 if (cl->level >= level) {
686 if (cl->level == level)
687 cl->xstats.lends++;
688 htb_accnt_tokens(cl, bytes, diff);
689 } else {
690 cl->xstats.borrows++;
691 cl->tokens += diff; /* we moved t_c; update tokens */
692 }
693 htb_accnt_ctokens(cl, bytes, diff);
694 cl->t_c = q->now;
695
696 old_mode = cl->cmode;
697 diff = 0;
698 htb_change_class_mode(q, cl, &diff);
699 if (old_mode != cl->cmode) {
700 if (old_mode != HTB_CAN_SEND)
701 htb_safe_rb_erase(&cl->pq_node, &q->hlevel[cl->level].wait_pq);
702 if (cl->cmode != HTB_CAN_SEND)
703 htb_add_to_wait_tree(q, cl, diff);
704 }
705
706 /* update basic stats except for leaves which are already updated */
707 if (cl->level)
708 bstats_update(&cl->bstats, skb);
709
710 cl = cl->parent;
711 }
712}
713
714/**
715 * htb_do_events - make mode changes to classes at the level
716 *
717 * Scans event queue for pending events and applies them. Returns time of
718 * next pending event (0 for no event in pq, q->now for too many events).
719 * Note: Applied are events whose have cl->pq_key <= q->now.
720 */
721static s64 htb_do_events(struct htb_sched *q, const int level,
722 unsigned long start)
723{
724 /* don't run for longer than 2 jiffies; 2 is used instead of
725 * 1 to simplify things when jiffy is going to be incremented
726 * too soon
727 */
728 unsigned long stop_at = start + 2;
729 struct rb_root *wait_pq = &q->hlevel[level].wait_pq;
730
731 while (time_before(jiffies, stop_at)) {
732 struct htb_class *cl;
733 s64 diff;
734 struct rb_node *p = rb_first(wait_pq);
735
736 if (!p)
737 return 0;
738
739 cl = rb_entry(p, struct htb_class, pq_node);
740 if (cl->pq_key > q->now)
741 return cl->pq_key;
742
743 htb_safe_rb_erase(p, wait_pq);
744 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer);
745 htb_change_class_mode(q, cl, &diff);
746 if (cl->cmode != HTB_CAN_SEND)
747 htb_add_to_wait_tree(q, cl, diff);
748 }
749
750 /* too much load - let's continue after a break for scheduling */
751 if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) {
752 pr_warn("htb: too many events!\n");
753 q->warned |= HTB_WARN_TOOMANYEVENTS;
754 }
755
756 return q->now;
757}
758
759/* Returns class->node+prio from id-tree where classe's id is >= id. NULL
760 * is no such one exists.
761 */
762static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n,
763 u32 id)
764{
765 struct rb_node *r = NULL;
766 while (n) {
767 struct htb_class *cl =
768 rb_entry(n, struct htb_class, node[prio]);
769
770 if (id > cl->common.classid) {
771 n = n->rb_right;
772 } else if (id < cl->common.classid) {
773 r = n;
774 n = n->rb_left;
775 } else {
776 return n;
777 }
778 }
779 return r;
780}
781
782/**
783 * htb_lookup_leaf - returns next leaf class in DRR order
784 *
785 * Find leaf where current feed pointers points to.
786 */
787static struct htb_class *htb_lookup_leaf(struct htb_prio *hprio, const int prio)
788{
789 int i;
790 struct {
791 struct rb_node *root;
792 struct rb_node **pptr;
793 u32 *pid;
794 } stk[TC_HTB_MAXDEPTH], *sp = stk;
795
796 BUG_ON(!hprio->row.rb_node);
797 sp->root = hprio->row.rb_node;
798 sp->pptr = &hprio->ptr;
799 sp->pid = &hprio->last_ptr_id;
800
801 for (i = 0; i < 65535; i++) {
802 if (!*sp->pptr && *sp->pid) {
803 /* ptr was invalidated but id is valid - try to recover
804 * the original or next ptr
805 */
806 *sp->pptr =
807 htb_id_find_next_upper(prio, sp->root, *sp->pid);
808 }
809 *sp->pid = 0; /* ptr is valid now so that remove this hint as it
810 * can become out of date quickly
811 */
812 if (!*sp->pptr) { /* we are at right end; rewind & go up */
813 *sp->pptr = sp->root;
814 while ((*sp->pptr)->rb_left)
815 *sp->pptr = (*sp->pptr)->rb_left;
816 if (sp > stk) {
817 sp--;
818 if (!*sp->pptr) {
819 WARN_ON(1);
820 return NULL;
821 }
822 htb_next_rb_node(sp->pptr);
823 }
824 } else {
825 struct htb_class *cl;
826 struct htb_prio *clp;
827
828 cl = rb_entry(*sp->pptr, struct htb_class, node[prio]);
829 if (!cl->level)
830 return cl;
831 clp = &cl->un.inner.clprio[prio];
832 (++sp)->root = clp->feed.rb_node;
833 sp->pptr = &clp->ptr;
834 sp->pid = &clp->last_ptr_id;
835 }
836 }
837 WARN_ON(1);
838 return NULL;
839}
840
841/* dequeues packet at given priority and level; call only if
842 * you are sure that there is active class at prio/level
843 */
844static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, const int prio,
845 const int level)
846{
847 struct sk_buff *skb = NULL;
848 struct htb_class *cl, *start;
849 struct htb_level *hlevel = &q->hlevel[level];
850 struct htb_prio *hprio = &hlevel->hprio[prio];
851
852 /* look initial class up in the row */
853 start = cl = htb_lookup_leaf(hprio, prio);
854
855 do {
856next:
857 if (unlikely(!cl))
858 return NULL;
859
860 /* class can be empty - it is unlikely but can be true if leaf
861 * qdisc drops packets in enqueue routine or if someone used
862 * graft operation on the leaf since last dequeue;
863 * simply deactivate and skip such class
864 */
865 if (unlikely(cl->un.leaf.q->q.qlen == 0)) {
866 struct htb_class *next;
867 htb_deactivate(q, cl);
868
869 /* row/level might become empty */
870 if ((q->row_mask[level] & (1 << prio)) == 0)
871 return NULL;
872
873 next = htb_lookup_leaf(hprio, prio);
874
875 if (cl == start) /* fix start if we just deleted it */
876 start = next;
877 cl = next;
878 goto next;
879 }
880
881 skb = cl->un.leaf.q->dequeue(cl->un.leaf.q);
882 if (likely(skb != NULL))
883 break;
884
885 qdisc_warn_nonwc("htb", cl->un.leaf.q);
886 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr:
887 &q->hlevel[0].hprio[prio].ptr);
888 cl = htb_lookup_leaf(hprio, prio);
889
890 } while (cl != start);
891
892 if (likely(skb != NULL)) {
893 bstats_update(&cl->bstats, skb);
894 cl->un.leaf.deficit[level] -= qdisc_pkt_len(skb);
895 if (cl->un.leaf.deficit[level] < 0) {
896 cl->un.leaf.deficit[level] += cl->quantum;
897 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr :
898 &q->hlevel[0].hprio[prio].ptr);
899 }
900 /* this used to be after charge_class but this constelation
901 * gives us slightly better performance
902 */
903 if (!cl->un.leaf.q->q.qlen)
904 htb_deactivate(q, cl);
905 htb_charge_class(q, cl, level, skb);
906 }
907 return skb;
908}
909
910static struct sk_buff *htb_dequeue(struct Qdisc *sch)
911{
912 struct sk_buff *skb;
913 struct htb_sched *q = qdisc_priv(sch);
914 int level;
915 s64 next_event;
916 unsigned long start_at;
917
918 /* try to dequeue direct packets as high prio (!) to minimize cpu work */
919 skb = __qdisc_dequeue_head(&q->direct_queue);
920 if (skb != NULL) {
921ok:
922 qdisc_bstats_update(sch, skb);
923 qdisc_qstats_backlog_dec(sch, skb);
924 sch->q.qlen--;
925 return skb;
926 }
927
928 if (!sch->q.qlen)
929 goto fin;
930 q->now = ktime_get_ns();
931 start_at = jiffies;
932
933 next_event = q->now + 5LLU * NSEC_PER_SEC;
934
935 for (level = 0; level < TC_HTB_MAXDEPTH; level++) {
936 /* common case optimization - skip event handler quickly */
937 int m;
938 s64 event = q->near_ev_cache[level];
939
940 if (q->now >= event) {
941 event = htb_do_events(q, level, start_at);
942 if (!event)
943 event = q->now + NSEC_PER_SEC;
944 q->near_ev_cache[level] = event;
945 }
946
947 if (next_event > event)
948 next_event = event;
949
950 m = ~q->row_mask[level];
951 while (m != (int)(-1)) {
952 int prio = ffz(m);
953
954 m |= 1 << prio;
955 skb = htb_dequeue_tree(q, prio, level);
956 if (likely(skb != NULL))
957 goto ok;
958 }
959 }
960 qdisc_qstats_overlimit(sch);
961 if (likely(next_event > q->now))
962 qdisc_watchdog_schedule_ns(&q->watchdog, next_event);
963 else
964 schedule_work(&q->work);
965fin:
966 return skb;
967}
968
969/* reset all classes */
970/* always caled under BH & queue lock */
971static void htb_reset(struct Qdisc *sch)
972{
973 struct htb_sched *q = qdisc_priv(sch);
974 struct htb_class *cl;
975 unsigned int i;
976
977 for (i = 0; i < q->clhash.hashsize; i++) {
978 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
979 if (cl->level)
980 memset(&cl->un.inner, 0, sizeof(cl->un.inner));
981 else {
982 if (cl->un.leaf.q)
983 qdisc_reset(cl->un.leaf.q);
984 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
985 }
986 cl->prio_activity = 0;
987 cl->cmode = HTB_CAN_SEND;
988 }
989 }
990 qdisc_watchdog_cancel(&q->watchdog);
991 __qdisc_reset_queue(&q->direct_queue);
992 sch->q.qlen = 0;
993 sch->qstats.backlog = 0;
994 memset(q->hlevel, 0, sizeof(q->hlevel));
995 memset(q->row_mask, 0, sizeof(q->row_mask));
996 for (i = 0; i < TC_HTB_NUMPRIO; i++)
997 INIT_LIST_HEAD(q->drops + i);
998}
999
1000static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = {
1001 [TCA_HTB_PARMS] = { .len = sizeof(struct tc_htb_opt) },
1002 [TCA_HTB_INIT] = { .len = sizeof(struct tc_htb_glob) },
1003 [TCA_HTB_CTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
1004 [TCA_HTB_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
1005 [TCA_HTB_DIRECT_QLEN] = { .type = NLA_U32 },
1006 [TCA_HTB_RATE64] = { .type = NLA_U64 },
1007 [TCA_HTB_CEIL64] = { .type = NLA_U64 },
1008};
1009
1010static void htb_work_func(struct work_struct *work)
1011{
1012 struct htb_sched *q = container_of(work, struct htb_sched, work);
1013 struct Qdisc *sch = q->watchdog.qdisc;
1014
1015 rcu_read_lock();
1016 __netif_schedule(qdisc_root(sch));
1017 rcu_read_unlock();
1018}
1019
1020static int htb_init(struct Qdisc *sch, struct nlattr *opt,
1021 struct netlink_ext_ack *extack)
1022{
1023 struct htb_sched *q = qdisc_priv(sch);
1024 struct nlattr *tb[TCA_HTB_MAX + 1];
1025 struct tc_htb_glob *gopt;
1026 int err;
1027 int i;
1028
1029 qdisc_watchdog_init(&q->watchdog, sch);
1030 INIT_WORK(&q->work, htb_work_func);
1031
1032 if (!opt)
1033 return -EINVAL;
1034
1035 err = tcf_block_get(&q->block, &q->filter_list, sch, extack);
1036 if (err)
1037 return err;
1038
1039 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy, NULL);
1040 if (err < 0)
1041 return err;
1042
1043 if (!tb[TCA_HTB_INIT])
1044 return -EINVAL;
1045
1046 gopt = nla_data(tb[TCA_HTB_INIT]);
1047 if (gopt->version != HTB_VER >> 16)
1048 return -EINVAL;
1049
1050 err = qdisc_class_hash_init(&q->clhash);
1051 if (err < 0)
1052 return err;
1053 for (i = 0; i < TC_HTB_NUMPRIO; i++)
1054 INIT_LIST_HEAD(q->drops + i);
1055
1056 qdisc_skb_head_init(&q->direct_queue);
1057
1058 if (tb[TCA_HTB_DIRECT_QLEN])
1059 q->direct_qlen = nla_get_u32(tb[TCA_HTB_DIRECT_QLEN]);
1060 else
1061 q->direct_qlen = qdisc_dev(sch)->tx_queue_len;
1062
1063 if ((q->rate2quantum = gopt->rate2quantum) < 1)
1064 q->rate2quantum = 1;
1065 q->defcls = gopt->defcls;
1066
1067 return 0;
1068}
1069
1070static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
1071{
1072 struct htb_sched *q = qdisc_priv(sch);
1073 struct nlattr *nest;
1074 struct tc_htb_glob gopt;
1075
1076 /* Its safe to not acquire qdisc lock. As we hold RTNL,
1077 * no change can happen on the qdisc parameters.
1078 */
1079
1080 gopt.direct_pkts = q->direct_pkts;
1081 gopt.version = HTB_VER;
1082 gopt.rate2quantum = q->rate2quantum;
1083 gopt.defcls = q->defcls;
1084 gopt.debug = 0;
1085
1086 nest = nla_nest_start(skb, TCA_OPTIONS);
1087 if (nest == NULL)
1088 goto nla_put_failure;
1089 if (nla_put(skb, TCA_HTB_INIT, sizeof(gopt), &gopt) ||
1090 nla_put_u32(skb, TCA_HTB_DIRECT_QLEN, q->direct_qlen))
1091 goto nla_put_failure;
1092
1093 return nla_nest_end(skb, nest);
1094
1095nla_put_failure:
1096 nla_nest_cancel(skb, nest);
1097 return -1;
1098}
1099
1100static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
1101 struct sk_buff *skb, struct tcmsg *tcm)
1102{
1103 struct htb_class *cl = (struct htb_class *)arg;
1104 struct nlattr *nest;
1105 struct tc_htb_opt opt;
1106
1107 /* Its safe to not acquire qdisc lock. As we hold RTNL,
1108 * no change can happen on the class parameters.
1109 */
1110 tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT;
1111 tcm->tcm_handle = cl->common.classid;
1112 if (!cl->level && cl->un.leaf.q)
1113 tcm->tcm_info = cl->un.leaf.q->handle;
1114
1115 nest = nla_nest_start(skb, TCA_OPTIONS);
1116 if (nest == NULL)
1117 goto nla_put_failure;
1118
1119 memset(&opt, 0, sizeof(opt));
1120
1121 psched_ratecfg_getrate(&opt.rate, &cl->rate);
1122 opt.buffer = PSCHED_NS2TICKS(cl->buffer);
1123 psched_ratecfg_getrate(&opt.ceil, &cl->ceil);
1124 opt.cbuffer = PSCHED_NS2TICKS(cl->cbuffer);
1125 opt.quantum = cl->quantum;
1126 opt.prio = cl->prio;
1127 opt.level = cl->level;
1128 if (nla_put(skb, TCA_HTB_PARMS, sizeof(opt), &opt))
1129 goto nla_put_failure;
1130 if ((cl->rate.rate_bytes_ps >= (1ULL << 32)) &&
1131 nla_put_u64_64bit(skb, TCA_HTB_RATE64, cl->rate.rate_bytes_ps,
1132 TCA_HTB_PAD))
1133 goto nla_put_failure;
1134 if ((cl->ceil.rate_bytes_ps >= (1ULL << 32)) &&
1135 nla_put_u64_64bit(skb, TCA_HTB_CEIL64, cl->ceil.rate_bytes_ps,
1136 TCA_HTB_PAD))
1137 goto nla_put_failure;
1138
1139 return nla_nest_end(skb, nest);
1140
1141nla_put_failure:
1142 nla_nest_cancel(skb, nest);
1143 return -1;
1144}
1145
1146static int
1147htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
1148{
1149 struct htb_class *cl = (struct htb_class *)arg;
1150 struct gnet_stats_queue qs = {
1151 .drops = cl->drops,
1152 .overlimits = cl->overlimits,
1153 };
1154 __u32 qlen = 0;
1155
1156 if (!cl->level && cl->un.leaf.q) {
1157 qlen = cl->un.leaf.q->q.qlen;
1158 qs.backlog = cl->un.leaf.q->qstats.backlog;
1159 }
1160 cl->xstats.tokens = clamp_t(s64, PSCHED_NS2TICKS(cl->tokens),
1161 INT_MIN, INT_MAX);
1162 cl->xstats.ctokens = clamp_t(s64, PSCHED_NS2TICKS(cl->ctokens),
1163 INT_MIN, INT_MAX);
1164
1165 if (gnet_stats_copy_basic(qdisc_root_sleeping_running(sch),
1166 d, NULL, &cl->bstats) < 0 ||
1167 gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 ||
1168 gnet_stats_copy_queue(d, NULL, &qs, qlen) < 0)
1169 return -1;
1170
1171 return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
1172}
1173
1174static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
1175 struct Qdisc **old, struct netlink_ext_ack *extack)
1176{
1177 struct htb_class *cl = (struct htb_class *)arg;
1178
1179 if (cl->level)
1180 return -EINVAL;
1181 if (new == NULL &&
1182 (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1183 cl->common.classid, extack)) == NULL)
1184 return -ENOBUFS;
1185
1186 *old = qdisc_replace(sch, new, &cl->un.leaf.q);
1187 return 0;
1188}
1189
1190static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg)
1191{
1192 struct htb_class *cl = (struct htb_class *)arg;
1193 return !cl->level ? cl->un.leaf.q : NULL;
1194}
1195
1196static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
1197{
1198 struct htb_class *cl = (struct htb_class *)arg;
1199
1200 htb_deactivate(qdisc_priv(sch), cl);
1201}
1202
1203static inline int htb_parent_last_child(struct htb_class *cl)
1204{
1205 if (!cl->parent)
1206 /* the root class */
1207 return 0;
1208 if (cl->parent->children > 1)
1209 /* not the last child */
1210 return 0;
1211 return 1;
1212}
1213
1214static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl,
1215 struct Qdisc *new_q)
1216{
1217 struct htb_class *parent = cl->parent;
1218
1219 WARN_ON(cl->level || !cl->un.leaf.q || cl->prio_activity);
1220
1221 if (parent->cmode != HTB_CAN_SEND)
1222 htb_safe_rb_erase(&parent->pq_node,
1223 &q->hlevel[parent->level].wait_pq);
1224
1225 parent->level = 0;
1226 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1227 INIT_LIST_HEAD(&parent->un.leaf.drop_list);
1228 parent->un.leaf.q = new_q ? new_q : &noop_qdisc;
1229 parent->tokens = parent->buffer;
1230 parent->ctokens = parent->cbuffer;
1231 parent->t_c = ktime_get_ns();
1232 parent->cmode = HTB_CAN_SEND;
1233}
1234
1235static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl)
1236{
1237 if (!cl->level) {
1238 WARN_ON(!cl->un.leaf.q);
1239 qdisc_destroy(cl->un.leaf.q);
1240 }
1241 gen_kill_estimator(&cl->rate_est);
1242 tcf_block_put(cl->block);
1243 kfree(cl);
1244}
1245
1246static void htb_destroy(struct Qdisc *sch)
1247{
1248 struct htb_sched *q = qdisc_priv(sch);
1249 struct hlist_node *next;
1250 struct htb_class *cl;
1251 unsigned int i;
1252
1253 cancel_work_sync(&q->work);
1254 qdisc_watchdog_cancel(&q->watchdog);
1255 /* This line used to be after htb_destroy_class call below
1256 * and surprisingly it worked in 2.4. But it must precede it
1257 * because filter need its target class alive to be able to call
1258 * unbind_filter on it (without Oops).
1259 */
1260 tcf_block_put(q->block);
1261
1262 for (i = 0; i < q->clhash.hashsize; i++) {
1263 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
1264 tcf_block_put(cl->block);
1265 cl->block = NULL;
1266 }
1267 }
1268 for (i = 0; i < q->clhash.hashsize; i++) {
1269 hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i],
1270 common.hnode)
1271 htb_destroy_class(sch, cl);
1272 }
1273 qdisc_class_hash_destroy(&q->clhash);
1274 __qdisc_reset_queue(&q->direct_queue);
1275}
1276
1277static int htb_delete(struct Qdisc *sch, unsigned long arg)
1278{
1279 struct htb_sched *q = qdisc_priv(sch);
1280 struct htb_class *cl = (struct htb_class *)arg;
1281 struct Qdisc *new_q = NULL;
1282 int last_child = 0;
1283
1284 /* TODO: why don't allow to delete subtree ? references ? does
1285 * tc subsys guarantee us that in htb_destroy it holds no class
1286 * refs so that we can remove children safely there ?
1287 */
1288 if (cl->children || cl->filter_cnt)
1289 return -EBUSY;
1290
1291 if (!cl->level && htb_parent_last_child(cl)) {
1292 new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1293 cl->parent->common.classid,
1294 NULL);
1295 last_child = 1;
1296 }
1297
1298 sch_tree_lock(sch);
1299
1300 if (!cl->level) {
1301 unsigned int qlen = cl->un.leaf.q->q.qlen;
1302 unsigned int backlog = cl->un.leaf.q->qstats.backlog;
1303
1304 qdisc_reset(cl->un.leaf.q);
1305 qdisc_tree_reduce_backlog(cl->un.leaf.q, qlen, backlog);
1306 }
1307
1308 /* delete from hash and active; remainder in destroy_class */
1309 qdisc_class_hash_remove(&q->clhash, &cl->common);
1310 if (cl->parent)
1311 cl->parent->children--;
1312
1313 if (cl->prio_activity)
1314 htb_deactivate(q, cl);
1315
1316 if (cl->cmode != HTB_CAN_SEND)
1317 htb_safe_rb_erase(&cl->pq_node,
1318 &q->hlevel[cl->level].wait_pq);
1319
1320 if (last_child)
1321 htb_parent_to_leaf(q, cl, new_q);
1322
1323 sch_tree_unlock(sch);
1324
1325 htb_destroy_class(sch, cl);
1326 return 0;
1327}
1328
1329static int htb_change_class(struct Qdisc *sch, u32 classid,
1330 u32 parentid, struct nlattr **tca,
1331 unsigned long *arg, struct netlink_ext_ack *extack)
1332{
1333 int err = -EINVAL;
1334 struct htb_sched *q = qdisc_priv(sch);
1335 struct htb_class *cl = (struct htb_class *)*arg, *parent;
1336 struct nlattr *opt = tca[TCA_OPTIONS];
1337 struct nlattr *tb[TCA_HTB_MAX + 1];
1338 struct tc_htb_opt *hopt;
1339 u64 rate64, ceil64;
1340 int warn = 0;
1341
1342 /* extract all subattrs from opt attr */
1343 if (!opt)
1344 goto failure;
1345
1346 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy, NULL);
1347 if (err < 0)
1348 goto failure;
1349
1350 err = -EINVAL;
1351 if (tb[TCA_HTB_PARMS] == NULL)
1352 goto failure;
1353
1354 parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch);
1355
1356 hopt = nla_data(tb[TCA_HTB_PARMS]);
1357 if (!hopt->rate.rate || !hopt->ceil.rate)
1358 goto failure;
1359
1360 /* Keeping backward compatible with rate_table based iproute2 tc */
1361 if (hopt->rate.linklayer == TC_LINKLAYER_UNAWARE)
1362 qdisc_put_rtab(qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB],
1363 NULL));
1364
1365 if (hopt->ceil.linklayer == TC_LINKLAYER_UNAWARE)
1366 qdisc_put_rtab(qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB],
1367 NULL));
1368
1369 if (!cl) { /* new class */
1370 struct Qdisc *new_q;
1371 int prio;
1372 struct {
1373 struct nlattr nla;
1374 struct gnet_estimator opt;
1375 } est = {
1376 .nla = {
1377 .nla_len = nla_attr_size(sizeof(est.opt)),
1378 .nla_type = TCA_RATE,
1379 },
1380 .opt = {
1381 /* 4s interval, 16s averaging constant */
1382 .interval = 2,
1383 .ewma_log = 2,
1384 },
1385 };
1386
1387 /* check for valid classid */
1388 if (!classid || TC_H_MAJ(classid ^ sch->handle) ||
1389 htb_find(classid, sch))
1390 goto failure;
1391
1392 /* check maximal depth */
1393 if (parent && parent->parent && parent->parent->level < 2) {
1394 pr_err("htb: tree is too deep\n");
1395 goto failure;
1396 }
1397 err = -ENOBUFS;
1398 cl = kzalloc(sizeof(*cl), GFP_KERNEL);
1399 if (!cl)
1400 goto failure;
1401
1402 err = tcf_block_get(&cl->block, &cl->filter_list, sch, extack);
1403 if (err) {
1404 kfree(cl);
1405 goto failure;
1406 }
1407 if (htb_rate_est || tca[TCA_RATE]) {
1408 err = gen_new_estimator(&cl->bstats, NULL,
1409 &cl->rate_est,
1410 NULL,
1411 qdisc_root_sleeping_running(sch),
1412 tca[TCA_RATE] ? : &est.nla);
1413 if (err) {
1414 tcf_block_put(cl->block);
1415 kfree(cl);
1416 goto failure;
1417 }
1418 }
1419
1420 cl->children = 0;
1421 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
1422 RB_CLEAR_NODE(&cl->pq_node);
1423
1424 for (prio = 0; prio < TC_HTB_NUMPRIO; prio++)
1425 RB_CLEAR_NODE(&cl->node[prio]);
1426
1427 /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL)
1428 * so that can't be used inside of sch_tree_lock
1429 * -- thanks to Karlis Peisenieks
1430 */
1431 new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1432 classid, NULL);
1433 sch_tree_lock(sch);
1434 if (parent && !parent->level) {
1435 unsigned int qlen = parent->un.leaf.q->q.qlen;
1436 unsigned int backlog = parent->un.leaf.q->qstats.backlog;
1437
1438 /* turn parent into inner node */
1439 qdisc_reset(parent->un.leaf.q);
1440 qdisc_tree_reduce_backlog(parent->un.leaf.q, qlen, backlog);
1441 qdisc_destroy(parent->un.leaf.q);
1442 if (parent->prio_activity)
1443 htb_deactivate(q, parent);
1444
1445 /* remove from evt list because of level change */
1446 if (parent->cmode != HTB_CAN_SEND) {
1447 htb_safe_rb_erase(&parent->pq_node, &q->hlevel[0].wait_pq);
1448 parent->cmode = HTB_CAN_SEND;
1449 }
1450 parent->level = (parent->parent ? parent->parent->level
1451 : TC_HTB_MAXDEPTH) - 1;
1452 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1453 }
1454 /* leaf (we) needs elementary qdisc */
1455 cl->un.leaf.q = new_q ? new_q : &noop_qdisc;
1456
1457 cl->common.classid = classid;
1458 cl->parent = parent;
1459
1460 /* set class to be in HTB_CAN_SEND state */
1461 cl->tokens = PSCHED_TICKS2NS(hopt->buffer);
1462 cl->ctokens = PSCHED_TICKS2NS(hopt->cbuffer);
1463 cl->mbuffer = 60ULL * NSEC_PER_SEC; /* 1min */
1464 cl->t_c = ktime_get_ns();
1465 cl->cmode = HTB_CAN_SEND;
1466
1467 /* attach to the hash list and parent's family */
1468 qdisc_class_hash_insert(&q->clhash, &cl->common);
1469 if (parent)
1470 parent->children++;
1471 if (cl->un.leaf.q != &noop_qdisc)
1472 qdisc_hash_add(cl->un.leaf.q, true);
1473 } else {
1474 if (tca[TCA_RATE]) {
1475 err = gen_replace_estimator(&cl->bstats, NULL,
1476 &cl->rate_est,
1477 NULL,
1478 qdisc_root_sleeping_running(sch),
1479 tca[TCA_RATE]);
1480 if (err)
1481 return err;
1482 }
1483 sch_tree_lock(sch);
1484 }
1485
1486 rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0;
1487
1488 ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0;
1489
1490 psched_ratecfg_precompute(&cl->rate, &hopt->rate, rate64);
1491 psched_ratecfg_precompute(&cl->ceil, &hopt->ceil, ceil64);
1492
1493 /* it used to be a nasty bug here, we have to check that node
1494 * is really leaf before changing cl->un.leaf !
1495 */
1496 if (!cl->level) {
1497 u64 quantum = cl->rate.rate_bytes_ps;
1498
1499 do_div(quantum, q->rate2quantum);
1500 cl->quantum = min_t(u64, quantum, INT_MAX);
1501
1502 if (!hopt->quantum && cl->quantum < 1000) {
1503 warn = -1;
1504 cl->quantum = 1000;
1505 }
1506 if (!hopt->quantum && cl->quantum > 200000) {
1507 warn = 1;
1508 cl->quantum = 200000;
1509 }
1510 if (hopt->quantum)
1511 cl->quantum = hopt->quantum;
1512 if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO)
1513 cl->prio = TC_HTB_NUMPRIO - 1;
1514 }
1515
1516 cl->buffer = PSCHED_TICKS2NS(hopt->buffer);
1517 cl->cbuffer = PSCHED_TICKS2NS(hopt->cbuffer);
1518
1519 sch_tree_unlock(sch);
1520
1521 if (warn)
1522 pr_warn("HTB: quantum of class %X is %s. Consider r2q change.\n",
1523 cl->common.classid, (warn == -1 ? "small" : "big"));
1524
1525 qdisc_class_hash_grow(sch, &q->clhash);
1526
1527 *arg = (unsigned long)cl;
1528 return 0;
1529
1530failure:
1531 return err;
1532}
1533
1534static struct tcf_block *htb_tcf_block(struct Qdisc *sch, unsigned long arg,
1535 struct netlink_ext_ack *extack)
1536{
1537 struct htb_sched *q = qdisc_priv(sch);
1538 struct htb_class *cl = (struct htb_class *)arg;
1539
1540 return cl ? cl->block : q->block;
1541}
1542
1543static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent,
1544 u32 classid)
1545{
1546 struct htb_class *cl = htb_find(classid, sch);
1547
1548 /*if (cl && !cl->level) return 0;
1549 * The line above used to be there to prevent attaching filters to
1550 * leaves. But at least tc_index filter uses this just to get class
1551 * for other reasons so that we have to allow for it.
1552 * ----
1553 * 19.6.2002 As Werner explained it is ok - bind filter is just
1554 * another way to "lock" the class - unlike "get" this lock can
1555 * be broken by class during destroy IIUC.
1556 */
1557 if (cl)
1558 cl->filter_cnt++;
1559 return (unsigned long)cl;
1560}
1561
1562static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg)
1563{
1564 struct htb_class *cl = (struct htb_class *)arg;
1565
1566 if (cl)
1567 cl->filter_cnt--;
1568}
1569
1570static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg)
1571{
1572 struct htb_sched *q = qdisc_priv(sch);
1573 struct htb_class *cl;
1574 unsigned int i;
1575
1576 if (arg->stop)
1577 return;
1578
1579 for (i = 0; i < q->clhash.hashsize; i++) {
1580 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) {
1581 if (arg->count < arg->skip) {
1582 arg->count++;
1583 continue;
1584 }
1585 if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
1586 arg->stop = 1;
1587 return;
1588 }
1589 arg->count++;
1590 }
1591 }
1592}
1593
1594static const struct Qdisc_class_ops htb_class_ops = {
1595 .graft = htb_graft,
1596 .leaf = htb_leaf,
1597 .qlen_notify = htb_qlen_notify,
1598 .find = htb_search,
1599 .change = htb_change_class,
1600 .delete = htb_delete,
1601 .walk = htb_walk,
1602 .tcf_block = htb_tcf_block,
1603 .bind_tcf = htb_bind_filter,
1604 .unbind_tcf = htb_unbind_filter,
1605 .dump = htb_dump_class,
1606 .dump_stats = htb_dump_class_stats,
1607};
1608
1609static struct Qdisc_ops htb_qdisc_ops __read_mostly = {
1610 .cl_ops = &htb_class_ops,
1611 .id = "htb",
1612 .priv_size = sizeof(struct htb_sched),
1613 .enqueue = htb_enqueue,
1614 .dequeue = htb_dequeue,
1615 .peek = qdisc_peek_dequeued,
1616 .init = htb_init,
1617 .reset = htb_reset,
1618 .destroy = htb_destroy,
1619 .dump = htb_dump,
1620 .owner = THIS_MODULE,
1621};
1622
1623static int __init htb_module_init(void)
1624{
1625 return register_qdisc(&htb_qdisc_ops);
1626}
1627static void __exit htb_module_exit(void)
1628{
1629 unregister_qdisc(&htb_qdisc_ops);
1630}
1631
1632module_init(htb_module_init)
1633module_exit(htb_module_exit)
1634MODULE_LICENSE("GPL");