blob: d9a5d298d755539443fa2a618ff70ba4863126da [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * net/sched/sch_tbf.c Token Bucket Filter queue.
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: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * Dmitry Torokhov <dtor@mail.ru> - allow attaching inner qdiscs -
11 * original idea by Martin Devera
12 *
13 */
14
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/module.h>
16#include <asm/uaccess.h>
17#include <asm/system.h>
18#include <linux/bitops.h>
19#include <linux/types.h>
20#include <linux/kernel.h>
21#include <linux/jiffies.h>
22#include <linux/string.h>
23#include <linux/mm.h>
24#include <linux/socket.h>
25#include <linux/sockios.h>
26#include <linux/in.h>
27#include <linux/errno.h>
28#include <linux/interrupt.h>
29#include <linux/if_ether.h>
30#include <linux/inet.h>
31#include <linux/netdevice.h>
32#include <linux/etherdevice.h>
33#include <linux/notifier.h>
34#include <net/ip.h>
35#include <net/route.h>
36#include <linux/skbuff.h>
37#include <net/sock.h>
38#include <net/pkt_sched.h>
39
40
41/* Simple Token Bucket Filter.
42 =======================================
43
44 SOURCE.
45 -------
46
47 None.
48
49 Description.
50 ------------
51
52 A data flow obeys TBF with rate R and depth B, if for any
53 time interval t_i...t_f the number of transmitted bits
54 does not exceed B + R*(t_f-t_i).
55
56 Packetized version of this definition:
57 The sequence of packets of sizes s_i served at moments t_i
58 obeys TBF, if for any i<=k:
59
60 s_i+....+s_k <= B + R*(t_k - t_i)
61
62 Algorithm.
63 ----------
64
65 Let N(t_i) be B/R initially and N(t) grow continuously with time as:
66
67 N(t+delta) = min{B/R, N(t) + delta}
68
69 If the first packet in queue has length S, it may be
70 transmitted only at the time t_* when S/R <= N(t_*),
71 and in this case N(t) jumps:
72
73 N(t_* + 0) = N(t_* - 0) - S/R.
74
75
76
77 Actually, QoS requires two TBF to be applied to a data stream.
78 One of them controls steady state burst size, another
79 one with rate P (peak rate) and depth M (equal to link MTU)
80 limits bursts at a smaller time scale.
81
82 It is easy to see that P>R, and B>M. If P is infinity, this double
83 TBF is equivalent to a single one.
84
85 When TBF works in reshaping mode, latency is estimated as:
86
87 lat = max ((L-B)/R, (L-M)/P)
88
89
90 NOTES.
91 ------
92
93 If TBF throttles, it starts a watchdog timer, which will wake it up
94 when it is ready to transmit.
95 Note that the minimal timer resolution is 1/HZ.
96 If no new packets arrive during this period,
97 or if the device is not awaken by EOI for some previous packet,
98 TBF can stop its activity for 1/HZ.
99
100
101 This means, that with depth B, the maximal rate is
102
103 R_crit = B*HZ
104
105 F.e. for 10Mbit ethernet and HZ=100 the minimal allowed B is ~10Kbytes.
106
107 Note that the peak rate TBF is much more tough: with MTU 1500
108 P_crit = 150Kbytes/sec. So, if you need greater peak
109 rates, use alpha with HZ=1000 :-)
110
111 With classful TBF, limit is just kept for backwards compatibility.
112 It is passed to the default bfifo qdisc - if the inner qdisc is
113 changed the limit is not effective anymore.
114*/
115
116struct tbf_sched_data
117{
118/* Parameters */
119 u32 limit; /* Maximal length of backlog: bytes */
120 u32 buffer; /* Token bucket depth/rate: MUST BE >= MTU/B */
121 u32 mtu;
122 u32 max_size;
123 struct qdisc_rate_table *R_tab;
124 struct qdisc_rate_table *P_tab;
125
126/* Variables */
127 long tokens; /* Current number of B tokens */
128 long ptokens; /* Current number of P tokens */
129 psched_time_t t_c; /* Time check-point */
130 struct timer_list wd_timer; /* Watchdog timer */
131 struct Qdisc *qdisc; /* Inner qdisc, default - bfifo queue */
132};
133
134#define L2T(q,L) ((q)->R_tab->data[(L)>>(q)->R_tab->rate.cell_log])
135#define L2T_P(q,L) ((q)->P_tab->data[(L)>>(q)->P_tab->rate.cell_log])
136
137static int tbf_enqueue(struct sk_buff *skb, struct Qdisc* sch)
138{
139 struct tbf_sched_data *q = qdisc_priv(sch);
140 int ret;
141
142 if (skb->len > q->max_size) {
143 sch->qstats.drops++;
144#ifdef CONFIG_NET_CLS_POLICE
145 if (sch->reshape_fail == NULL || sch->reshape_fail(skb, sch))
146#endif
147 kfree_skb(skb);
148
149 return NET_XMIT_DROP;
150 }
151
152 if ((ret = q->qdisc->enqueue(skb, q->qdisc)) != 0) {
153 sch->qstats.drops++;
154 return ret;
155 }
156
157 sch->q.qlen++;
158 sch->bstats.bytes += skb->len;
159 sch->bstats.packets++;
160 return 0;
161}
162
163static int tbf_requeue(struct sk_buff *skb, struct Qdisc* sch)
164{
165 struct tbf_sched_data *q = qdisc_priv(sch);
166 int ret;
167
168 if ((ret = q->qdisc->ops->requeue(skb, q->qdisc)) == 0) {
169 sch->q.qlen++;
170 sch->qstats.requeues++;
171 }
172
173 return ret;
174}
175
176static unsigned int tbf_drop(struct Qdisc* sch)
177{
178 struct tbf_sched_data *q = qdisc_priv(sch);
Patrick McHardy6d037a22006-03-20 19:00:49 -0800179 unsigned int len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
Patrick McHardy6d037a22006-03-20 19:00:49 -0800181 if (q->qdisc->ops->drop && (len = q->qdisc->ops->drop(q->qdisc)) != 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 sch->q.qlen--;
183 sch->qstats.drops++;
184 }
185 return len;
186}
187
188static void tbf_watchdog(unsigned long arg)
189{
190 struct Qdisc *sch = (struct Qdisc*)arg;
191
192 sch->flags &= ~TCQ_F_THROTTLED;
193 netif_schedule(sch->dev);
194}
195
196static struct sk_buff *tbf_dequeue(struct Qdisc* sch)
197{
198 struct tbf_sched_data *q = qdisc_priv(sch);
199 struct sk_buff *skb;
200
201 skb = q->qdisc->dequeue(q->qdisc);
202
203 if (skb) {
204 psched_time_t now;
205 long toks, delay;
206 long ptoks = 0;
207 unsigned int len = skb->len;
208
209 PSCHED_GET_TIME(now);
210
211 toks = PSCHED_TDIFF_SAFE(now, q->t_c, q->buffer);
212
213 if (q->P_tab) {
214 ptoks = toks + q->ptokens;
215 if (ptoks > (long)q->mtu)
216 ptoks = q->mtu;
217 ptoks -= L2T_P(q, len);
218 }
219 toks += q->tokens;
220 if (toks > (long)q->buffer)
221 toks = q->buffer;
222 toks -= L2T(q, len);
223
224 if ((toks|ptoks) >= 0) {
225 q->t_c = now;
226 q->tokens = toks;
227 q->ptokens = ptoks;
228 sch->q.qlen--;
229 sch->flags &= ~TCQ_F_THROTTLED;
230 return skb;
231 }
232
233 delay = PSCHED_US2JIFFIE(max_t(long, -toks, -ptoks));
234
235 if (delay == 0)
236 delay = 1;
237
238 mod_timer(&q->wd_timer, jiffies+delay);
239
240 /* Maybe we have a shorter packet in the queue,
241 which can be sent now. It sounds cool,
242 but, however, this is wrong in principle.
243 We MUST NOT reorder packets under these circumstances.
244
245 Really, if we split the flow into independent
246 subflows, it would be a very good solution.
247 This is the main idea of all FQ algorithms
248 (cf. CSZ, HPFQ, HFSC)
249 */
250
251 if (q->qdisc->ops->requeue(skb, q->qdisc) != NET_XMIT_SUCCESS) {
252 /* When requeue fails skb is dropped */
253 sch->q.qlen--;
254 sch->qstats.drops++;
255 }
256
257 sch->flags |= TCQ_F_THROTTLED;
258 sch->qstats.overlimits++;
259 }
260 return NULL;
261}
262
263static void tbf_reset(struct Qdisc* sch)
264{
265 struct tbf_sched_data *q = qdisc_priv(sch);
266
267 qdisc_reset(q->qdisc);
268 sch->q.qlen = 0;
269 PSCHED_GET_TIME(q->t_c);
270 q->tokens = q->buffer;
271 q->ptokens = q->mtu;
272 sch->flags &= ~TCQ_F_THROTTLED;
273 del_timer(&q->wd_timer);
274}
275
276static struct Qdisc *tbf_create_dflt_qdisc(struct net_device *dev, u32 limit)
277{
278 struct Qdisc *q = qdisc_create_dflt(dev, &bfifo_qdisc_ops);
279 struct rtattr *rta;
280 int ret;
281
282 if (q) {
283 rta = kmalloc(RTA_LENGTH(sizeof(struct tc_fifo_qopt)), GFP_KERNEL);
284 if (rta) {
285 rta->rta_type = RTM_NEWQDISC;
286 rta->rta_len = RTA_LENGTH(sizeof(struct tc_fifo_qopt));
287 ((struct tc_fifo_qopt *)RTA_DATA(rta))->limit = limit;
288
289 ret = q->ops->change(q, rta);
290 kfree(rta);
291
292 if (ret == 0)
293 return q;
294 }
295 qdisc_destroy(q);
296 }
297
298 return NULL;
299}
300
301static int tbf_change(struct Qdisc* sch, struct rtattr *opt)
302{
303 int err = -EINVAL;
304 struct tbf_sched_data *q = qdisc_priv(sch);
305 struct rtattr *tb[TCA_TBF_PTAB];
306 struct tc_tbf_qopt *qopt;
307 struct qdisc_rate_table *rtab = NULL;
308 struct qdisc_rate_table *ptab = NULL;
309 struct Qdisc *child = NULL;
310 int max_size,n;
311
312 if (rtattr_parse_nested(tb, TCA_TBF_PTAB, opt) ||
313 tb[TCA_TBF_PARMS-1] == NULL ||
314 RTA_PAYLOAD(tb[TCA_TBF_PARMS-1]) < sizeof(*qopt))
315 goto done;
316
317 qopt = RTA_DATA(tb[TCA_TBF_PARMS-1]);
318 rtab = qdisc_get_rtab(&qopt->rate, tb[TCA_TBF_RTAB-1]);
319 if (rtab == NULL)
320 goto done;
321
322 if (qopt->peakrate.rate) {
323 if (qopt->peakrate.rate > qopt->rate.rate)
324 ptab = qdisc_get_rtab(&qopt->peakrate, tb[TCA_TBF_PTAB-1]);
325 if (ptab == NULL)
326 goto done;
327 }
328
329 for (n = 0; n < 256; n++)
330 if (rtab->data[n] > qopt->buffer) break;
331 max_size = (n << qopt->rate.cell_log)-1;
332 if (ptab) {
333 int size;
334
335 for (n = 0; n < 256; n++)
336 if (ptab->data[n] > qopt->mtu) break;
337 size = (n << qopt->peakrate.cell_log)-1;
338 if (size < max_size) max_size = size;
339 }
340 if (max_size < 0)
341 goto done;
342
Patrick McHardy053cfed2006-03-20 19:01:21 -0800343 if (qopt->limit > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 if ((child = tbf_create_dflt_qdisc(sch->dev, qopt->limit)) == NULL)
345 goto done;
346 }
347
348 sch_tree_lock(sch);
Patrick McHardy053cfed2006-03-20 19:01:21 -0800349 if (child)
350 qdisc_destroy(xchg(&q->qdisc, child));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 q->limit = qopt->limit;
352 q->mtu = qopt->mtu;
353 q->max_size = max_size;
354 q->buffer = qopt->buffer;
355 q->tokens = q->buffer;
356 q->ptokens = q->mtu;
357 rtab = xchg(&q->R_tab, rtab);
358 ptab = xchg(&q->P_tab, ptab);
359 sch_tree_unlock(sch);
360 err = 0;
361done:
362 if (rtab)
363 qdisc_put_rtab(rtab);
364 if (ptab)
365 qdisc_put_rtab(ptab);
366 return err;
367}
368
369static int tbf_init(struct Qdisc* sch, struct rtattr *opt)
370{
371 struct tbf_sched_data *q = qdisc_priv(sch);
372
373 if (opt == NULL)
374 return -EINVAL;
375
376 PSCHED_GET_TIME(q->t_c);
377 init_timer(&q->wd_timer);
378 q->wd_timer.function = tbf_watchdog;
379 q->wd_timer.data = (unsigned long)sch;
380
381 q->qdisc = &noop_qdisc;
382
383 return tbf_change(sch, opt);
384}
385
386static void tbf_destroy(struct Qdisc *sch)
387{
388 struct tbf_sched_data *q = qdisc_priv(sch);
389
390 del_timer(&q->wd_timer);
391
392 if (q->P_tab)
393 qdisc_put_rtab(q->P_tab);
394 if (q->R_tab)
395 qdisc_put_rtab(q->R_tab);
396
397 qdisc_destroy(q->qdisc);
398}
399
400static int tbf_dump(struct Qdisc *sch, struct sk_buff *skb)
401{
402 struct tbf_sched_data *q = qdisc_priv(sch);
403 unsigned char *b = skb->tail;
404 struct rtattr *rta;
405 struct tc_tbf_qopt opt;
406
407 rta = (struct rtattr*)b;
408 RTA_PUT(skb, TCA_OPTIONS, 0, NULL);
409
410 opt.limit = q->limit;
411 opt.rate = q->R_tab->rate;
412 if (q->P_tab)
413 opt.peakrate = q->P_tab->rate;
414 else
415 memset(&opt.peakrate, 0, sizeof(opt.peakrate));
416 opt.mtu = q->mtu;
417 opt.buffer = q->buffer;
418 RTA_PUT(skb, TCA_TBF_PARMS, sizeof(opt), &opt);
419 rta->rta_len = skb->tail - b;
420
421 return skb->len;
422
423rtattr_failure:
424 skb_trim(skb, b - skb->data);
425 return -1;
426}
427
428static int tbf_dump_class(struct Qdisc *sch, unsigned long cl,
429 struct sk_buff *skb, struct tcmsg *tcm)
430{
431 struct tbf_sched_data *q = qdisc_priv(sch);
432
433 if (cl != 1) /* only one class */
434 return -ENOENT;
435
436 tcm->tcm_handle |= TC_H_MIN(1);
437 tcm->tcm_info = q->qdisc->handle;
438
439 return 0;
440}
441
442static int tbf_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
443 struct Qdisc **old)
444{
445 struct tbf_sched_data *q = qdisc_priv(sch);
446
447 if (new == NULL)
448 new = &noop_qdisc;
449
450 sch_tree_lock(sch);
451 *old = xchg(&q->qdisc, new);
452 qdisc_reset(*old);
453 sch->q.qlen = 0;
454 sch_tree_unlock(sch);
455
456 return 0;
457}
458
459static struct Qdisc *tbf_leaf(struct Qdisc *sch, unsigned long arg)
460{
461 struct tbf_sched_data *q = qdisc_priv(sch);
462 return q->qdisc;
463}
464
465static unsigned long tbf_get(struct Qdisc *sch, u32 classid)
466{
467 return 1;
468}
469
470static void tbf_put(struct Qdisc *sch, unsigned long arg)
471{
472}
473
474static int tbf_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
475 struct rtattr **tca, unsigned long *arg)
476{
477 return -ENOSYS;
478}
479
480static int tbf_delete(struct Qdisc *sch, unsigned long arg)
481{
482 return -ENOSYS;
483}
484
485static void tbf_walk(struct Qdisc *sch, struct qdisc_walker *walker)
486{
487 if (!walker->stop) {
488 if (walker->count >= walker->skip)
489 if (walker->fn(sch, 1, walker) < 0) {
490 walker->stop = 1;
491 return;
492 }
493 walker->count++;
494 }
495}
496
497static struct tcf_proto **tbf_find_tcf(struct Qdisc *sch, unsigned long cl)
498{
499 return NULL;
500}
501
502static struct Qdisc_class_ops tbf_class_ops =
503{
504 .graft = tbf_graft,
505 .leaf = tbf_leaf,
506 .get = tbf_get,
507 .put = tbf_put,
508 .change = tbf_change_class,
509 .delete = tbf_delete,
510 .walk = tbf_walk,
511 .tcf_chain = tbf_find_tcf,
512 .dump = tbf_dump_class,
513};
514
515static struct Qdisc_ops tbf_qdisc_ops = {
516 .next = NULL,
517 .cl_ops = &tbf_class_ops,
518 .id = "tbf",
519 .priv_size = sizeof(struct tbf_sched_data),
520 .enqueue = tbf_enqueue,
521 .dequeue = tbf_dequeue,
522 .requeue = tbf_requeue,
523 .drop = tbf_drop,
524 .init = tbf_init,
525 .reset = tbf_reset,
526 .destroy = tbf_destroy,
527 .change = tbf_change,
528 .dump = tbf_dump,
529 .owner = THIS_MODULE,
530};
531
532static int __init tbf_module_init(void)
533{
534 return register_qdisc(&tbf_qdisc_ops);
535}
536
537static void __exit tbf_module_exit(void)
538{
539 unregister_qdisc(&tbf_qdisc_ops);
540}
541module_init(tbf_module_init)
542module_exit(tbf_module_exit)
543MODULE_LICENSE("GPL");