2 * net/sched/sch_htb.c Hierarchical token bucket, feed tree version
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.
9 * Authors: Martin Devera, <devik@cdi.cz>
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
23 * spotted bug in dequeue code and helped with fix
25 * fixed requeue routine
26 * and many others. thanks.
28 * $Id: sch_htb.c,v 1.25 2003/12/07 11:08:25 devik Exp devik $
30 #include <linux/module.h>
31 #include <linux/types.h>
32 #include <linux/kernel.h>
33 #include <linux/string.h>
34 #include <linux/errno.h>
35 #include <linux/skbuff.h>
36 #include <linux/list.h>
37 #include <linux/compiler.h>
38 #include <linux/rbtree.h>
39 #include <net/netlink.h>
40 #include <net/pkt_sched.h>
44 ========================================================================
45 HTB is like TBF with multiple classes. It is also similar to CBQ because
46 it allows to assign priority to each class in hierarchy.
47 In fact it is another implementation of Floyd's formal sharing.
50 Each class is assigned level. Leaf has ALWAYS level 0 and root
51 classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level
52 one less than their parent.
55 #define HTB_HSIZE 16 /* classid hash size */
56 #define HTB_HYSTERESIS 1 /* whether to use mode hysteresis for speedup */
57 #define HTB_VER 0x30011 /* major must be matched with number suplied by TC as version */
59 #if HTB_VER >> 16 != TC_HTB_PROTOVER
60 #error "Mismatched sch_htb.c and pkt_sch.h"
63 /* used internaly to keep status of single class */
65 HTB_CANT_SEND, /* class can't send and can't borrow */
66 HTB_MAY_BORROW, /* class can't send but may borrow */
67 HTB_CAN_SEND /* class can send */
70 /* interior & leaf nodes; props specific to leaves are marked L: */
72 /* general class parameters */
74 struct gnet_stats_basic bstats;
75 struct gnet_stats_queue qstats;
76 struct gnet_stats_rate_est rate_est;
77 struct tc_htb_xstats xstats; /* our special stats */
78 int refcnt; /* usage count of this class */
81 int level; /* our level (see above) */
82 struct htb_class *parent; /* parent class */
83 struct hlist_node hlist; /* classid hash list item */
84 struct list_head sibling; /* sibling list item */
85 struct list_head children; /* children list */
88 struct htb_class_leaf {
93 int deficit[TC_HTB_MAXDEPTH];
94 struct list_head drop_list;
96 struct htb_class_inner {
97 struct rb_root feed[TC_HTB_NUMPRIO]; /* feed trees */
98 struct rb_node *ptr[TC_HTB_NUMPRIO]; /* current class ptr */
99 /* When class changes from state 1->2 and disconnects from
100 parent's feed then we lost ptr value and start from the
101 first child again. Here we store classid of the
102 last valid ptr (used when ptr is NULL). */
103 u32 last_ptr_id[TC_HTB_NUMPRIO];
106 struct rb_node node[TC_HTB_NUMPRIO]; /* node for self or feed tree */
107 struct rb_node pq_node; /* node for event queue */
108 psched_time_t pq_key;
110 int prio_activity; /* for which prios are we active */
111 enum htb_cmode cmode; /* current mode of the class */
113 /* class attached filters */
114 struct tcf_proto *filter_list;
117 int warned; /* only one warning about non work conserving .. */
119 /* token bucket parameters */
120 struct qdisc_rate_table *rate; /* rate table of the class itself */
121 struct qdisc_rate_table *ceil; /* ceiling rate (limits borrows too) */
122 long buffer, cbuffer; /* token bucket depth/rate */
123 psched_tdiff_t mbuffer; /* max wait time */
124 long tokens, ctokens; /* current number of tokens */
125 psched_time_t t_c; /* checkpoint time */
127 int prio; /* For parent to leaf return possible here */
128 int quantum; /* we do backup. Finally full replacement */
129 /* of un.leaf originals should be done. */
132 /* TODO: maybe compute rate when size is too large .. or drop ? */
133 static inline long L2T(struct htb_class *cl, struct qdisc_rate_table *rate,
136 int slot = size >> rate->rate.cell_log;
141 return rate->data[slot];
145 struct list_head root; /* root classes list */
146 struct hlist_head hash[HTB_HSIZE]; /* hashed by classid */
147 struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */
149 /* self list - roots of self generating tree */
150 struct rb_root row[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
151 int row_mask[TC_HTB_MAXDEPTH];
152 struct rb_node *ptr[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
153 u32 last_ptr_id[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO];
155 /* self wait list - roots of wait PQs per row */
156 struct rb_root wait_pq[TC_HTB_MAXDEPTH];
158 /* time of nearest event per level (row) */
159 psched_time_t near_ev_cache[TC_HTB_MAXDEPTH];
161 /* whether we hit non-work conserving class during this dequeue; we use */
162 int nwc_hit; /* this to disable mindelay complaint in dequeue */
164 int defcls; /* class where unclassified flows go to */
166 /* filters for qdisc itself */
167 struct tcf_proto *filter_list;
170 int rate2quantum; /* quant = rate / rate2quantum */
171 psched_time_t now; /* cached dequeue time */
172 struct qdisc_watchdog watchdog;
174 /* non shaped skbs; let them go directly thru */
175 struct sk_buff_head direct_queue;
176 int direct_qlen; /* max qlen of above */
181 /* compute hash of size HTB_HSIZE for given handle */
182 static inline int htb_hash(u32 h)
185 #error "Declare new hash for your HTB_HSIZE"
187 h ^= h >> 8; /* stolen from cbq_hash */
192 /* find class in global hash table using given handle */
193 static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch)
195 struct htb_sched *q = qdisc_priv(sch);
196 struct hlist_node *p;
197 struct htb_class *cl;
199 if (TC_H_MAJ(handle) != sch->handle)
202 hlist_for_each_entry(cl, p, q->hash + htb_hash(handle), hlist) {
203 if (cl->classid == handle)
210 * htb_classify - classify a packet into class
212 * It returns NULL if the packet should be dropped or -1 if the packet
213 * should be passed directly thru. In all other cases leaf class is returned.
214 * We allow direct class selection by classid in priority. The we examine
215 * filters in qdisc and in inner nodes (if higher filter points to the inner
216 * node). If we end up with classid MAJOR:0 we enqueue the skb into special
217 * internal fifo (direct). These packets then go directly thru. If we still
218 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessfull
219 * then finish and return direct queue.
221 #define HTB_DIRECT (struct htb_class*)-1
222 static inline u32 htb_classid(struct htb_class *cl)
224 return (cl && cl != HTB_DIRECT) ? cl->classid : TC_H_UNSPEC;
227 static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch,
230 struct htb_sched *q = qdisc_priv(sch);
231 struct htb_class *cl;
232 struct tcf_result res;
233 struct tcf_proto *tcf;
236 /* allow to select class by setting skb->priority to valid classid;
237 note that nfmark can be used too by attaching filter fw with no
239 if (skb->priority == sch->handle)
240 return HTB_DIRECT; /* X:0 (direct flow) selected */
241 if ((cl = htb_find(skb->priority, sch)) != NULL && cl->level == 0)
244 *qerr = NET_XMIT_BYPASS;
245 tcf = q->filter_list;
246 while (tcf && (result = tc_classify(skb, tcf, &res)) >= 0) {
247 #ifdef CONFIG_NET_CLS_ACT
251 *qerr = NET_XMIT_SUCCESS;
255 #elif defined(CONFIG_NET_CLS_POLICE)
256 if (result == TC_POLICE_SHOT)
259 if ((cl = (void *)res.class) == NULL) {
260 if (res.classid == sch->handle)
261 return HTB_DIRECT; /* X:0 (direct flow) */
262 if ((cl = htb_find(res.classid, sch)) == NULL)
263 break; /* filter selected invalid classid */
266 return cl; /* we hit leaf; return it */
268 /* we have got inner class; apply inner filter chain */
269 tcf = cl->filter_list;
271 /* classification failed; try to use default class */
272 cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch);
273 if (!cl || cl->level)
274 return HTB_DIRECT; /* bad default .. this is safe bet */
279 * htb_add_to_id_tree - adds class to the round robin list
281 * Routine adds class to the list (actually tree) sorted by classid.
282 * Make sure that class is not already on such list for given prio.
284 static void htb_add_to_id_tree(struct rb_root *root,
285 struct htb_class *cl, int prio)
287 struct rb_node **p = &root->rb_node, *parent = NULL;
292 c = rb_entry(parent, struct htb_class, node[prio]);
294 if (cl->classid > c->classid)
295 p = &parent->rb_right;
297 p = &parent->rb_left;
299 rb_link_node(&cl->node[prio], parent, p);
300 rb_insert_color(&cl->node[prio], root);
304 * htb_add_to_wait_tree - adds class to the event queue with delay
306 * The class is added to priority event queue to indicate that class will
307 * change its mode in cl->pq_key microseconds. Make sure that class is not
308 * already in the queue.
310 static void htb_add_to_wait_tree(struct htb_sched *q,
311 struct htb_class *cl, long delay)
313 struct rb_node **p = &q->wait_pq[cl->level].rb_node, *parent = NULL;
315 cl->pq_key = q->now + delay;
316 if (cl->pq_key == q->now)
319 /* update the nearest event cache */
320 if (q->near_ev_cache[cl->level] > cl->pq_key)
321 q->near_ev_cache[cl->level] = cl->pq_key;
326 c = rb_entry(parent, struct htb_class, pq_node);
327 if (cl->pq_key >= c->pq_key)
328 p = &parent->rb_right;
330 p = &parent->rb_left;
332 rb_link_node(&cl->pq_node, parent, p);
333 rb_insert_color(&cl->pq_node, &q->wait_pq[cl->level]);
337 * htb_next_rb_node - finds next node in binary tree
339 * When we are past last key we return NULL.
340 * Average complexity is 2 steps per call.
342 static inline void htb_next_rb_node(struct rb_node **n)
348 * htb_add_class_to_row - add class to its row
350 * The class is added to row at priorities marked in mask.
351 * It does nothing if mask == 0.
353 static inline void htb_add_class_to_row(struct htb_sched *q,
354 struct htb_class *cl, int mask)
356 q->row_mask[cl->level] |= mask;
358 int prio = ffz(~mask);
359 mask &= ~(1 << prio);
360 htb_add_to_id_tree(q->row[cl->level] + prio, cl, prio);
364 /* If this triggers, it is a bug in this code, but it need not be fatal */
365 static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root)
367 if (RB_EMPTY_NODE(rb)) {
377 * htb_remove_class_from_row - removes class from its row
379 * The class is removed from row at priorities marked in mask.
380 * It does nothing if mask == 0.
382 static inline void htb_remove_class_from_row(struct htb_sched *q,
383 struct htb_class *cl, int mask)
388 int prio = ffz(~mask);
390 mask &= ~(1 << prio);
391 if (q->ptr[cl->level][prio] == cl->node + prio)
392 htb_next_rb_node(q->ptr[cl->level] + prio);
394 htb_safe_rb_erase(cl->node + prio, q->row[cl->level] + prio);
395 if (!q->row[cl->level][prio].rb_node)
398 q->row_mask[cl->level] &= ~m;
402 * htb_activate_prios - creates active classe's feed chain
404 * The class is connected to ancestors and/or appropriate rows
405 * for priorities it is participating on. cl->cmode must be new
406 * (activated) mode. It does nothing if cl->prio_activity == 0.
408 static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl)
410 struct htb_class *p = cl->parent;
411 long m, mask = cl->prio_activity;
413 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
419 if (p->un.inner.feed[prio].rb_node)
420 /* parent already has its feed in use so that
421 reset bit in mask as parent is already ok */
422 mask &= ~(1 << prio);
424 htb_add_to_id_tree(p->un.inner.feed + prio, cl, prio);
426 p->prio_activity |= mask;
431 if (cl->cmode == HTB_CAN_SEND && mask)
432 htb_add_class_to_row(q, cl, mask);
436 * htb_deactivate_prios - remove class from feed chain
438 * cl->cmode must represent old mode (before deactivation). It does
439 * nothing if cl->prio_activity == 0. Class is removed from all feed
442 static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl)
444 struct htb_class *p = cl->parent;
445 long m, mask = cl->prio_activity;
447 while (cl->cmode == HTB_MAY_BORROW && p && mask) {
454 if (p->un.inner.ptr[prio] == cl->node + prio) {
455 /* we are removing child which is pointed to from
456 parent feed - forget the pointer but remember
458 p->un.inner.last_ptr_id[prio] = cl->classid;
459 p->un.inner.ptr[prio] = NULL;
462 htb_safe_rb_erase(cl->node + prio, p->un.inner.feed + prio);
464 if (!p->un.inner.feed[prio].rb_node)
468 p->prio_activity &= ~mask;
473 if (cl->cmode == HTB_CAN_SEND && mask)
474 htb_remove_class_from_row(q, cl, mask);
478 static inline long htb_lowater(const struct htb_class *cl)
480 return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0;
482 static inline long htb_hiwater(const struct htb_class *cl)
484 return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0;
487 #define htb_lowater(cl) (0)
488 #define htb_hiwater(cl) (0)
492 * htb_class_mode - computes and returns current class mode
494 * It computes cl's mode at time cl->t_c+diff and returns it. If mode
495 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference
496 * from now to time when cl will change its state.
497 * Also it is worth to note that class mode doesn't change simply
498 * at cl->{c,}tokens == 0 but there can rather be hysteresis of
499 * 0 .. -cl->{c,}buffer range. It is meant to limit number of
500 * mode transitions per time unit. The speed gain is about 1/6.
502 static inline enum htb_cmode
503 htb_class_mode(struct htb_class *cl, long *diff)
507 if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) {
509 return HTB_CANT_SEND;
512 if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl))
516 return HTB_MAY_BORROW;
520 * htb_change_class_mode - changes classe's mode
522 * This should be the only way how to change classe's mode under normal
523 * cirsumstances. Routine will update feed lists linkage, change mode
524 * and add class to the wait event queue if appropriate. New mode should
525 * be different from old one and cl->pq_key has to be valid if changing
526 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree).
529 htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, long *diff)
531 enum htb_cmode new_mode = htb_class_mode(cl, diff);
533 if (new_mode == cl->cmode)
536 if (cl->prio_activity) { /* not necessary: speed optimization */
537 if (cl->cmode != HTB_CANT_SEND)
538 htb_deactivate_prios(q, cl);
539 cl->cmode = new_mode;
540 if (new_mode != HTB_CANT_SEND)
541 htb_activate_prios(q, cl);
543 cl->cmode = new_mode;
547 * htb_activate - inserts leaf cl into appropriate active feeds
549 * Routine learns (new) priority of leaf and activates feed chain
550 * for the prio. It can be called on already active leaf safely.
551 * It also adds leaf into droplist.
553 static inline void htb_activate(struct htb_sched *q, struct htb_class *cl)
555 BUG_TRAP(!cl->level && cl->un.leaf.q && cl->un.leaf.q->q.qlen);
557 if (!cl->prio_activity) {
558 cl->prio_activity = 1 << (cl->un.leaf.aprio = cl->un.leaf.prio);
559 htb_activate_prios(q, cl);
560 list_add_tail(&cl->un.leaf.drop_list,
561 q->drops + cl->un.leaf.aprio);
566 * htb_deactivate - remove leaf cl from active feeds
568 * Make sure that leaf is active. In the other words it can't be called
569 * with non-active leaf. It also removes class from the drop list.
571 static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl)
573 BUG_TRAP(cl->prio_activity);
575 htb_deactivate_prios(q, cl);
576 cl->prio_activity = 0;
577 list_del_init(&cl->un.leaf.drop_list);
580 static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch)
583 struct htb_sched *q = qdisc_priv(sch);
584 struct htb_class *cl = htb_classify(skb, sch, &ret);
586 if (cl == HTB_DIRECT) {
587 /* enqueue to helper queue */
588 if (q->direct_queue.qlen < q->direct_qlen) {
589 __skb_queue_tail(&q->direct_queue, skb);
594 return NET_XMIT_DROP;
596 #ifdef CONFIG_NET_CLS_ACT
598 if (ret == NET_XMIT_BYPASS)
603 } else if (cl->un.leaf.q->enqueue(skb, cl->un.leaf.q) !=
607 return NET_XMIT_DROP;
609 cl->bstats.packets++;
610 cl->bstats.bytes += skb->len;
615 sch->bstats.packets++;
616 sch->bstats.bytes += skb->len;
617 return NET_XMIT_SUCCESS;
620 /* TODO: requeuing packet charges it to policers again !! */
621 static int htb_requeue(struct sk_buff *skb, struct Qdisc *sch)
623 struct htb_sched *q = qdisc_priv(sch);
624 int ret = NET_XMIT_SUCCESS;
625 struct htb_class *cl = htb_classify(skb, sch, &ret);
626 struct sk_buff *tskb;
628 if (cl == HTB_DIRECT || !cl) {
629 /* enqueue to helper queue */
630 if (q->direct_queue.qlen < q->direct_qlen && cl) {
631 __skb_queue_head(&q->direct_queue, skb);
633 __skb_queue_head(&q->direct_queue, skb);
634 tskb = __skb_dequeue_tail(&q->direct_queue);
639 } else if (cl->un.leaf.q->ops->requeue(skb, cl->un.leaf.q) !=
643 return NET_XMIT_DROP;
648 sch->qstats.requeues++;
649 return NET_XMIT_SUCCESS;
653 * htb_charge_class - charges amount "bytes" to leaf and ancestors
655 * Routine assumes that packet "bytes" long was dequeued from leaf cl
656 * borrowing from "level". It accounts bytes to ceil leaky bucket for
657 * leaf and all ancestors and to rate bucket for ancestors at levels
658 * "level" and higher. It also handles possible change of mode resulting
659 * from the update. Note that mode can also increase here (MAY_BORROW to
660 * CAN_SEND) because we can use more precise clock that event queue here.
661 * In such case we remove class from event queue first.
663 static void htb_charge_class(struct htb_sched *q, struct htb_class *cl,
664 int level, int bytes)
667 enum htb_cmode old_mode;
669 #define HTB_ACCNT(T,B,R) toks = diff + cl->T; \
670 if (toks > cl->B) toks = cl->B; \
671 toks -= L2T(cl, cl->R, bytes); \
672 if (toks <= -cl->mbuffer) toks = 1-cl->mbuffer; \
676 diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer);
677 if (cl->level >= level) {
678 if (cl->level == level)
680 HTB_ACCNT(tokens, buffer, rate);
682 cl->xstats.borrows++;
683 cl->tokens += diff; /* we moved t_c; update tokens */
685 HTB_ACCNT(ctokens, cbuffer, ceil);
688 old_mode = cl->cmode;
690 htb_change_class_mode(q, cl, &diff);
691 if (old_mode != cl->cmode) {
692 if (old_mode != HTB_CAN_SEND)
693 htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level);
694 if (cl->cmode != HTB_CAN_SEND)
695 htb_add_to_wait_tree(q, cl, diff);
698 /* update byte stats except for leaves which are already updated */
700 cl->bstats.bytes += bytes;
701 cl->bstats.packets++;
708 * htb_do_events - make mode changes to classes at the level
710 * Scans event queue for pending events and applies them. Returns time of
711 * next pending event (0 for no event in pq).
712 * Note: Applied are events whose have cl->pq_key <= q->now.
714 static psched_time_t htb_do_events(struct htb_sched *q, int level)
718 for (i = 0; i < 500; i++) {
719 struct htb_class *cl;
721 struct rb_node *p = rb_first(&q->wait_pq[level]);
726 cl = rb_entry(p, struct htb_class, pq_node);
727 if (cl->pq_key > q->now)
730 htb_safe_rb_erase(p, q->wait_pq + level);
731 diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer);
732 htb_change_class_mode(q, cl, &diff);
733 if (cl->cmode != HTB_CAN_SEND)
734 htb_add_to_wait_tree(q, cl, diff);
737 printk(KERN_WARNING "htb: too many events !\n");
738 return q->now + PSCHED_TICKS_PER_SEC / 10;
741 /* Returns class->node+prio from id-tree where classe's id is >= id. NULL
742 is no such one exists. */
743 static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n,
746 struct rb_node *r = NULL;
748 struct htb_class *cl =
749 rb_entry(n, struct htb_class, node[prio]);
750 if (id == cl->classid)
753 if (id > cl->classid) {
764 * htb_lookup_leaf - returns next leaf class in DRR order
766 * Find leaf where current feed pointers points to.
768 static struct htb_class *htb_lookup_leaf(struct rb_root *tree, int prio,
769 struct rb_node **pptr, u32 * pid)
773 struct rb_node *root;
774 struct rb_node **pptr;
776 } stk[TC_HTB_MAXDEPTH], *sp = stk;
778 BUG_TRAP(tree->rb_node);
779 sp->root = tree->rb_node;
783 for (i = 0; i < 65535; i++) {
784 if (!*sp->pptr && *sp->pid) {
785 /* ptr was invalidated but id is valid - try to recover
786 the original or next ptr */
788 htb_id_find_next_upper(prio, sp->root, *sp->pid);
790 *sp->pid = 0; /* ptr is valid now so that remove this hint as it
791 can become out of date quickly */
792 if (!*sp->pptr) { /* we are at right end; rewind & go up */
793 *sp->pptr = sp->root;
794 while ((*sp->pptr)->rb_left)
795 *sp->pptr = (*sp->pptr)->rb_left;
801 htb_next_rb_node(sp->pptr);
804 struct htb_class *cl;
805 cl = rb_entry(*sp->pptr, struct htb_class, node[prio]);
808 (++sp)->root = cl->un.inner.feed[prio].rb_node;
809 sp->pptr = cl->un.inner.ptr + prio;
810 sp->pid = cl->un.inner.last_ptr_id + prio;
817 /* dequeues packet at given priority and level; call only if
818 you are sure that there is active class at prio/level */
819 static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, int prio,
822 struct sk_buff *skb = NULL;
823 struct htb_class *cl, *start;
824 /* look initial class up in the row */
825 start = cl = htb_lookup_leaf(q->row[level] + prio, prio,
826 q->ptr[level] + prio,
827 q->last_ptr_id[level] + prio);
835 /* class can be empty - it is unlikely but can be true if leaf
836 qdisc drops packets in enqueue routine or if someone used
837 graft operation on the leaf since last dequeue;
838 simply deactivate and skip such class */
839 if (unlikely(cl->un.leaf.q->q.qlen == 0)) {
840 struct htb_class *next;
841 htb_deactivate(q, cl);
843 /* row/level might become empty */
844 if ((q->row_mask[level] & (1 << prio)) == 0)
847 next = htb_lookup_leaf(q->row[level] + prio,
848 prio, q->ptr[level] + prio,
849 q->last_ptr_id[level] + prio);
851 if (cl == start) /* fix start if we just deleted it */
857 skb = cl->un.leaf.q->dequeue(cl->un.leaf.q);
858 if (likely(skb != NULL))
862 "htb: class %X isn't work conserving ?!\n",
867 htb_next_rb_node((level ? cl->parent->un.inner.ptr : q->
869 cl = htb_lookup_leaf(q->row[level] + prio, prio,
870 q->ptr[level] + prio,
871 q->last_ptr_id[level] + prio);
873 } while (cl != start);
875 if (likely(skb != NULL)) {
876 if ((cl->un.leaf.deficit[level] -= skb->len) < 0) {
877 cl->un.leaf.deficit[level] += cl->un.leaf.quantum;
878 htb_next_rb_node((level ? cl->parent->un.inner.ptr : q->
881 /* this used to be after charge_class but this constelation
882 gives us slightly better performance */
883 if (!cl->un.leaf.q->q.qlen)
884 htb_deactivate(q, cl);
885 htb_charge_class(q, cl, level, skb->len);
890 static struct sk_buff *htb_dequeue(struct Qdisc *sch)
892 struct sk_buff *skb = NULL;
893 struct htb_sched *q = qdisc_priv(sch);
895 psched_time_t next_event;
897 /* try to dequeue direct packets as high prio (!) to minimize cpu work */
898 skb = __skb_dequeue(&q->direct_queue);
900 sch->flags &= ~TCQ_F_THROTTLED;
907 q->now = psched_get_time();
909 next_event = q->now + 5 * PSCHED_TICKS_PER_SEC;
911 for (level = 0; level < TC_HTB_MAXDEPTH; level++) {
912 /* common case optimization - skip event handler quickly */
916 if (q->now >= q->near_ev_cache[level]) {
917 event = htb_do_events(q, level);
919 event = q->now + PSCHED_TICKS_PER_SEC;
920 q->near_ev_cache[level] = event;
922 event = q->near_ev_cache[level];
924 if (event && next_event > event)
927 m = ~q->row_mask[level];
928 while (m != (int)(-1)) {
931 skb = htb_dequeue_tree(q, prio, level);
932 if (likely(skb != NULL)) {
934 sch->flags &= ~TCQ_F_THROTTLED;
939 sch->qstats.overlimits++;
940 qdisc_watchdog_schedule(&q->watchdog, next_event);
945 /* try to drop from each class (by prio) until one succeed */
946 static unsigned int htb_drop(struct Qdisc *sch)
948 struct htb_sched *q = qdisc_priv(sch);
951 for (prio = TC_HTB_NUMPRIO - 1; prio >= 0; prio--) {
953 list_for_each(p, q->drops + prio) {
954 struct htb_class *cl = list_entry(p, struct htb_class,
957 if (cl->un.leaf.q->ops->drop &&
958 (len = cl->un.leaf.q->ops->drop(cl->un.leaf.q))) {
960 if (!cl->un.leaf.q->q.qlen)
961 htb_deactivate(q, cl);
969 /* reset all classes */
970 /* always caled under BH & queue lock */
971 static void htb_reset(struct Qdisc *sch)
973 struct htb_sched *q = qdisc_priv(sch);
976 for (i = 0; i < HTB_HSIZE; i++) {
977 struct hlist_node *p;
978 struct htb_class *cl;
980 hlist_for_each_entry(cl, p, q->hash + i, hlist) {
982 memset(&cl->un.inner, 0, sizeof(cl->un.inner));
985 qdisc_reset(cl->un.leaf.q);
986 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
988 cl->prio_activity = 0;
989 cl->cmode = HTB_CAN_SEND;
993 qdisc_watchdog_cancel(&q->watchdog);
994 __skb_queue_purge(&q->direct_queue);
996 memset(q->row, 0, sizeof(q->row));
997 memset(q->row_mask, 0, sizeof(q->row_mask));
998 memset(q->wait_pq, 0, sizeof(q->wait_pq));
999 memset(q->ptr, 0, sizeof(q->ptr));
1000 for (i = 0; i < TC_HTB_NUMPRIO; i++)
1001 INIT_LIST_HEAD(q->drops + i);
1004 static int htb_init(struct Qdisc *sch, struct rtattr *opt)
1006 struct htb_sched *q = qdisc_priv(sch);
1007 struct rtattr *tb[TCA_HTB_INIT];
1008 struct tc_htb_glob *gopt;
1010 if (!opt || rtattr_parse_nested(tb, TCA_HTB_INIT, opt) ||
1011 tb[TCA_HTB_INIT - 1] == NULL ||
1012 RTA_PAYLOAD(tb[TCA_HTB_INIT - 1]) < sizeof(*gopt)) {
1013 printk(KERN_ERR "HTB: hey probably you have bad tc tool ?\n");
1016 gopt = RTA_DATA(tb[TCA_HTB_INIT - 1]);
1017 if (gopt->version != HTB_VER >> 16) {
1019 "HTB: need tc/htb version %d (minor is %d), you have %d\n",
1020 HTB_VER >> 16, HTB_VER & 0xffff, gopt->version);
1024 INIT_LIST_HEAD(&q->root);
1025 for (i = 0; i < HTB_HSIZE; i++)
1026 INIT_HLIST_HEAD(q->hash + i);
1027 for (i = 0; i < TC_HTB_NUMPRIO; i++)
1028 INIT_LIST_HEAD(q->drops + i);
1030 qdisc_watchdog_init(&q->watchdog, sch);
1031 skb_queue_head_init(&q->direct_queue);
1033 q->direct_qlen = sch->dev->tx_queue_len;
1034 if (q->direct_qlen < 2) /* some devices have zero tx_queue_len */
1037 if ((q->rate2quantum = gopt->rate2quantum) < 1)
1038 q->rate2quantum = 1;
1039 q->defcls = gopt->defcls;
1044 static int htb_dump(struct Qdisc *sch, struct sk_buff *skb)
1046 struct htb_sched *q = qdisc_priv(sch);
1047 unsigned char *b = skb_tail_pointer(skb);
1049 struct tc_htb_glob gopt;
1050 spin_lock_bh(&sch->dev->queue_lock);
1051 gopt.direct_pkts = q->direct_pkts;
1053 gopt.version = HTB_VER;
1054 gopt.rate2quantum = q->rate2quantum;
1055 gopt.defcls = q->defcls;
1057 rta = (struct rtattr *)b;
1058 RTA_PUT(skb, TCA_OPTIONS, 0, NULL);
1059 RTA_PUT(skb, TCA_HTB_INIT, sizeof(gopt), &gopt);
1060 rta->rta_len = skb_tail_pointer(skb) - b;
1061 spin_unlock_bh(&sch->dev->queue_lock);
1064 spin_unlock_bh(&sch->dev->queue_lock);
1065 nlmsg_trim(skb, skb_tail_pointer(skb));
1069 static int htb_dump_class(struct Qdisc *sch, unsigned long arg,
1070 struct sk_buff *skb, struct tcmsg *tcm)
1072 struct htb_class *cl = (struct htb_class *)arg;
1073 unsigned char *b = skb_tail_pointer(skb);
1075 struct tc_htb_opt opt;
1077 spin_lock_bh(&sch->dev->queue_lock);
1078 tcm->tcm_parent = cl->parent ? cl->parent->classid : TC_H_ROOT;
1079 tcm->tcm_handle = cl->classid;
1080 if (!cl->level && cl->un.leaf.q)
1081 tcm->tcm_info = cl->un.leaf.q->handle;
1083 rta = (struct rtattr *)b;
1084 RTA_PUT(skb, TCA_OPTIONS, 0, NULL);
1086 memset(&opt, 0, sizeof(opt));
1088 opt.rate = cl->rate->rate;
1089 opt.buffer = cl->buffer;
1090 opt.ceil = cl->ceil->rate;
1091 opt.cbuffer = cl->cbuffer;
1092 opt.quantum = cl->un.leaf.quantum;
1093 opt.prio = cl->un.leaf.prio;
1094 opt.level = cl->level;
1095 RTA_PUT(skb, TCA_HTB_PARMS, sizeof(opt), &opt);
1096 rta->rta_len = skb_tail_pointer(skb) - b;
1097 spin_unlock_bh(&sch->dev->queue_lock);
1100 spin_unlock_bh(&sch->dev->queue_lock);
1106 htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
1108 struct htb_class *cl = (struct htb_class *)arg;
1110 if (!cl->level && cl->un.leaf.q)
1111 cl->qstats.qlen = cl->un.leaf.q->q.qlen;
1112 cl->xstats.tokens = cl->tokens;
1113 cl->xstats.ctokens = cl->ctokens;
1115 if (gnet_stats_copy_basic(d, &cl->bstats) < 0 ||
1116 gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 ||
1117 gnet_stats_copy_queue(d, &cl->qstats) < 0)
1120 return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
1123 static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
1126 struct htb_class *cl = (struct htb_class *)arg;
1128 if (cl && !cl->level) {
1130 (new = qdisc_create_dflt(sch->dev, &pfifo_qdisc_ops,
1135 if ((*old = xchg(&cl->un.leaf.q, new)) != NULL) {
1136 qdisc_tree_decrease_qlen(*old, (*old)->q.qlen);
1139 sch_tree_unlock(sch);
1145 static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg)
1147 struct htb_class *cl = (struct htb_class *)arg;
1148 return (cl && !cl->level) ? cl->un.leaf.q : NULL;
1151 static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg)
1153 struct htb_class *cl = (struct htb_class *)arg;
1155 if (cl->un.leaf.q->q.qlen == 0)
1156 htb_deactivate(qdisc_priv(sch), cl);
1159 static unsigned long htb_get(struct Qdisc *sch, u32 classid)
1161 struct htb_class *cl = htb_find(classid, sch);
1164 return (unsigned long)cl;
1167 static inline int htb_parent_last_child(struct htb_class *cl)
1170 /* the root class */
1173 if (!(cl->parent->children.next == &cl->sibling &&
1174 cl->parent->children.prev == &cl->sibling))
1175 /* not the last child */
1181 static void htb_parent_to_leaf(struct htb_class *cl, struct Qdisc *new_q)
1183 struct htb_class *parent = cl->parent;
1185 BUG_TRAP(!cl->level && cl->un.leaf.q && !cl->prio_activity);
1188 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1189 INIT_LIST_HEAD(&parent->un.leaf.drop_list);
1190 parent->un.leaf.q = new_q ? new_q : &noop_qdisc;
1191 parent->un.leaf.quantum = parent->quantum;
1192 parent->un.leaf.prio = parent->prio;
1193 parent->tokens = parent->buffer;
1194 parent->ctokens = parent->cbuffer;
1195 parent->t_c = psched_get_time();
1196 parent->cmode = HTB_CAN_SEND;
1199 static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl)
1201 struct htb_sched *q = qdisc_priv(sch);
1204 BUG_TRAP(cl->un.leaf.q);
1205 qdisc_destroy(cl->un.leaf.q);
1207 gen_kill_estimator(&cl->bstats, &cl->rate_est);
1208 qdisc_put_rtab(cl->rate);
1209 qdisc_put_rtab(cl->ceil);
1211 tcf_destroy_chain(cl->filter_list);
1213 while (!list_empty(&cl->children))
1214 htb_destroy_class(sch, list_entry(cl->children.next,
1215 struct htb_class, sibling));
1217 /* note: this delete may happen twice (see htb_delete) */
1218 hlist_del_init(&cl->hlist);
1219 list_del(&cl->sibling);
1221 if (cl->prio_activity)
1222 htb_deactivate(q, cl);
1224 if (cl->cmode != HTB_CAN_SEND)
1225 htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level);
1230 /* always caled under BH & queue lock */
1231 static void htb_destroy(struct Qdisc *sch)
1233 struct htb_sched *q = qdisc_priv(sch);
1235 qdisc_watchdog_cancel(&q->watchdog);
1236 /* This line used to be after htb_destroy_class call below
1237 and surprisingly it worked in 2.4. But it must precede it
1238 because filter need its target class alive to be able to call
1239 unbind_filter on it (without Oops). */
1240 tcf_destroy_chain(q->filter_list);
1242 while (!list_empty(&q->root))
1243 htb_destroy_class(sch, list_entry(q->root.next,
1244 struct htb_class, sibling));
1246 __skb_queue_purge(&q->direct_queue);
1249 static int htb_delete(struct Qdisc *sch, unsigned long arg)
1251 struct htb_sched *q = qdisc_priv(sch);
1252 struct htb_class *cl = (struct htb_class *)arg;
1254 struct Qdisc *new_q = NULL;
1257 // TODO: why don't allow to delete subtree ? references ? does
1258 // tc subsys quarantee us that in htb_destroy it holds no class
1259 // refs so that we can remove children safely there ?
1260 if (!list_empty(&cl->children) || cl->filter_cnt)
1263 if (!cl->level && htb_parent_last_child(cl)) {
1264 new_q = qdisc_create_dflt(sch->dev, &pfifo_qdisc_ops,
1265 cl->parent->classid);
1272 qlen = cl->un.leaf.q->q.qlen;
1273 qdisc_reset(cl->un.leaf.q);
1274 qdisc_tree_decrease_qlen(cl->un.leaf.q, qlen);
1277 /* delete from hash and active; remainder in destroy_class */
1278 hlist_del_init(&cl->hlist);
1280 if (cl->prio_activity)
1281 htb_deactivate(q, cl);
1284 htb_parent_to_leaf(cl, new_q);
1286 if (--cl->refcnt == 0)
1287 htb_destroy_class(sch, cl);
1289 sch_tree_unlock(sch);
1293 static void htb_put(struct Qdisc *sch, unsigned long arg)
1295 struct htb_class *cl = (struct htb_class *)arg;
1297 if (--cl->refcnt == 0)
1298 htb_destroy_class(sch, cl);
1301 static int htb_change_class(struct Qdisc *sch, u32 classid,
1302 u32 parentid, struct rtattr **tca,
1306 struct htb_sched *q = qdisc_priv(sch);
1307 struct htb_class *cl = (struct htb_class *)*arg, *parent;
1308 struct rtattr *opt = tca[TCA_OPTIONS - 1];
1309 struct qdisc_rate_table *rtab = NULL, *ctab = NULL;
1310 struct rtattr *tb[TCA_HTB_RTAB];
1311 struct tc_htb_opt *hopt;
1313 /* extract all subattrs from opt attr */
1314 if (!opt || rtattr_parse_nested(tb, TCA_HTB_RTAB, opt) ||
1315 tb[TCA_HTB_PARMS - 1] == NULL ||
1316 RTA_PAYLOAD(tb[TCA_HTB_PARMS - 1]) < sizeof(*hopt))
1319 parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch);
1321 hopt = RTA_DATA(tb[TCA_HTB_PARMS - 1]);
1323 rtab = qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB - 1]);
1324 ctab = qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB - 1]);
1328 if (!cl) { /* new class */
1329 struct Qdisc *new_q;
1333 struct gnet_estimator opt;
1336 .rta_len = RTA_LENGTH(sizeof(est.opt)),
1337 .rta_type = TCA_RATE,
1340 /* 4s interval, 16s averaging constant */
1346 /* check for valid classid */
1347 if (!classid || TC_H_MAJ(classid ^ sch->handle)
1348 || htb_find(classid, sch))
1351 /* check maximal depth */
1352 if (parent && parent->parent && parent->parent->level < 2) {
1353 printk(KERN_ERR "htb: tree is too deep\n");
1357 if ((cl = kzalloc(sizeof(*cl), GFP_KERNEL)) == NULL)
1360 gen_new_estimator(&cl->bstats, &cl->rate_est,
1361 &sch->dev->queue_lock,
1362 tca[TCA_RATE-1] ? : &est.rta);
1364 INIT_LIST_HEAD(&cl->sibling);
1365 INIT_HLIST_NODE(&cl->hlist);
1366 INIT_LIST_HEAD(&cl->children);
1367 INIT_LIST_HEAD(&cl->un.leaf.drop_list);
1368 RB_CLEAR_NODE(&cl->pq_node);
1370 for (prio = 0; prio < TC_HTB_NUMPRIO; prio++)
1371 RB_CLEAR_NODE(&cl->node[prio]);
1373 /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL)
1374 so that can't be used inside of sch_tree_lock
1375 -- thanks to Karlis Peisenieks */
1376 new_q = qdisc_create_dflt(sch->dev, &pfifo_qdisc_ops, classid);
1378 if (parent && !parent->level) {
1379 unsigned int qlen = parent->un.leaf.q->q.qlen;
1381 /* turn parent into inner node */
1382 qdisc_reset(parent->un.leaf.q);
1383 qdisc_tree_decrease_qlen(parent->un.leaf.q, qlen);
1384 qdisc_destroy(parent->un.leaf.q);
1385 if (parent->prio_activity)
1386 htb_deactivate(q, parent);
1388 /* remove from evt list because of level change */
1389 if (parent->cmode != HTB_CAN_SEND) {
1390 htb_safe_rb_erase(&parent->pq_node, q->wait_pq);
1391 parent->cmode = HTB_CAN_SEND;
1393 parent->level = (parent->parent ? parent->parent->level
1394 : TC_HTB_MAXDEPTH) - 1;
1395 memset(&parent->un.inner, 0, sizeof(parent->un.inner));
1397 /* leaf (we) needs elementary qdisc */
1398 cl->un.leaf.q = new_q ? new_q : &noop_qdisc;
1400 cl->classid = classid;
1401 cl->parent = parent;
1403 /* set class to be in HTB_CAN_SEND state */
1404 cl->tokens = hopt->buffer;
1405 cl->ctokens = hopt->cbuffer;
1406 cl->mbuffer = 60 * PSCHED_TICKS_PER_SEC; /* 1min */
1407 cl->t_c = psched_get_time();
1408 cl->cmode = HTB_CAN_SEND;
1410 /* attach to the hash list and parent's family */
1411 hlist_add_head(&cl->hlist, q->hash + htb_hash(classid));
1412 list_add_tail(&cl->sibling,
1413 parent ? &parent->children : &q->root);
1415 if (tca[TCA_RATE-1])
1416 gen_replace_estimator(&cl->bstats, &cl->rate_est,
1417 &sch->dev->queue_lock,
1422 /* it used to be a nasty bug here, we have to check that node
1423 is really leaf before changing cl->un.leaf ! */
1425 cl->un.leaf.quantum = rtab->rate.rate / q->rate2quantum;
1426 if (!hopt->quantum && cl->un.leaf.quantum < 1000) {
1428 "HTB: quantum of class %X is small. Consider r2q change.\n",
1430 cl->un.leaf.quantum = 1000;
1432 if (!hopt->quantum && cl->un.leaf.quantum > 200000) {
1434 "HTB: quantum of class %X is big. Consider r2q change.\n",
1436 cl->un.leaf.quantum = 200000;
1439 cl->un.leaf.quantum = hopt->quantum;
1440 if ((cl->un.leaf.prio = hopt->prio) >= TC_HTB_NUMPRIO)
1441 cl->un.leaf.prio = TC_HTB_NUMPRIO - 1;
1443 /* backup for htb_parent_to_leaf */
1444 cl->quantum = cl->un.leaf.quantum;
1445 cl->prio = cl->un.leaf.prio;
1448 cl->buffer = hopt->buffer;
1449 cl->cbuffer = hopt->cbuffer;
1451 qdisc_put_rtab(cl->rate);
1454 qdisc_put_rtab(cl->ceil);
1456 sch_tree_unlock(sch);
1458 *arg = (unsigned long)cl;
1463 qdisc_put_rtab(rtab);
1465 qdisc_put_rtab(ctab);
1469 static struct tcf_proto **htb_find_tcf(struct Qdisc *sch, unsigned long arg)
1471 struct htb_sched *q = qdisc_priv(sch);
1472 struct htb_class *cl = (struct htb_class *)arg;
1473 struct tcf_proto **fl = cl ? &cl->filter_list : &q->filter_list;
1478 static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent,
1481 struct htb_sched *q = qdisc_priv(sch);
1482 struct htb_class *cl = htb_find(classid, sch);
1484 /*if (cl && !cl->level) return 0;
1485 The line above used to be there to prevent attaching filters to
1486 leaves. But at least tc_index filter uses this just to get class
1487 for other reasons so that we have to allow for it.
1489 19.6.2002 As Werner explained it is ok - bind filter is just
1490 another way to "lock" the class - unlike "get" this lock can
1491 be broken by class during destroy IIUC.
1497 return (unsigned long)cl;
1500 static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg)
1502 struct htb_sched *q = qdisc_priv(sch);
1503 struct htb_class *cl = (struct htb_class *)arg;
1511 static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg)
1513 struct htb_sched *q = qdisc_priv(sch);
1519 for (i = 0; i < HTB_HSIZE; i++) {
1520 struct hlist_node *p;
1521 struct htb_class *cl;
1523 hlist_for_each_entry(cl, p, q->hash + i, hlist) {
1524 if (arg->count < arg->skip) {
1528 if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
1537 static struct Qdisc_class_ops htb_class_ops = {
1540 .qlen_notify = htb_qlen_notify,
1543 .change = htb_change_class,
1544 .delete = htb_delete,
1546 .tcf_chain = htb_find_tcf,
1547 .bind_tcf = htb_bind_filter,
1548 .unbind_tcf = htb_unbind_filter,
1549 .dump = htb_dump_class,
1550 .dump_stats = htb_dump_class_stats,
1553 static struct Qdisc_ops htb_qdisc_ops = {
1555 .cl_ops = &htb_class_ops,
1557 .priv_size = sizeof(struct htb_sched),
1558 .enqueue = htb_enqueue,
1559 .dequeue = htb_dequeue,
1560 .requeue = htb_requeue,
1564 .destroy = htb_destroy,
1565 .change = NULL /* htb_change */,
1567 .owner = THIS_MODULE,
1570 static int __init htb_module_init(void)
1572 return register_qdisc(&htb_qdisc_ops);
1574 static void __exit htb_module_exit(void)
1576 unregister_qdisc(&htb_qdisc_ops);
1579 module_init(htb_module_init)
1580 module_exit(htb_module_exit)
1581 MODULE_LICENSE("GPL");