Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | ||
1da177e4 | 15 | #include <linux/module.h> |
1da177e4 LT |
16 | #include <linux/types.h> |
17 | #include <linux/kernel.h> | |
1da177e4 | 18 | #include <linux/string.h> |
1da177e4 | 19 | #include <linux/errno.h> |
1da177e4 | 20 | #include <linux/skbuff.h> |
0ba48053 | 21 | #include <net/netlink.h> |
b757c933 | 22 | #include <net/sch_generic.h> |
1da177e4 LT |
23 | #include <net/pkt_sched.h> |
24 | ||
25 | ||
26 | /* Simple Token Bucket Filter. | |
27 | ======================================= | |
28 | ||
29 | SOURCE. | |
30 | ------- | |
31 | ||
32 | None. | |
33 | ||
34 | Description. | |
35 | ------------ | |
36 | ||
37 | A data flow obeys TBF with rate R and depth B, if for any | |
38 | time interval t_i...t_f the number of transmitted bits | |
39 | does not exceed B + R*(t_f-t_i). | |
40 | ||
41 | Packetized version of this definition: | |
42 | The sequence of packets of sizes s_i served at moments t_i | |
43 | obeys TBF, if for any i<=k: | |
44 | ||
45 | s_i+....+s_k <= B + R*(t_k - t_i) | |
46 | ||
47 | Algorithm. | |
48 | ---------- | |
49 | ||
50 | Let N(t_i) be B/R initially and N(t) grow continuously with time as: | |
51 | ||
52 | N(t+delta) = min{B/R, N(t) + delta} | |
53 | ||
54 | If the first packet in queue has length S, it may be | |
55 | transmitted only at the time t_* when S/R <= N(t_*), | |
56 | and in this case N(t) jumps: | |
57 | ||
58 | N(t_* + 0) = N(t_* - 0) - S/R. | |
59 | ||
60 | ||
61 | ||
62 | Actually, QoS requires two TBF to be applied to a data stream. | |
63 | One of them controls steady state burst size, another | |
64 | one with rate P (peak rate) and depth M (equal to link MTU) | |
65 | limits bursts at a smaller time scale. | |
66 | ||
67 | It is easy to see that P>R, and B>M. If P is infinity, this double | |
68 | TBF is equivalent to a single one. | |
69 | ||
70 | When TBF works in reshaping mode, latency is estimated as: | |
71 | ||
72 | lat = max ((L-B)/R, (L-M)/P) | |
73 | ||
74 | ||
75 | NOTES. | |
76 | ------ | |
77 | ||
78 | If TBF throttles, it starts a watchdog timer, which will wake it up | |
79 | when it is ready to transmit. | |
80 | Note that the minimal timer resolution is 1/HZ. | |
81 | If no new packets arrive during this period, | |
82 | or if the device is not awaken by EOI for some previous packet, | |
83 | TBF can stop its activity for 1/HZ. | |
84 | ||
85 | ||
86 | This means, that with depth B, the maximal rate is | |
87 | ||
88 | R_crit = B*HZ | |
89 | ||
90 | F.e. for 10Mbit ethernet and HZ=100 the minimal allowed B is ~10Kbytes. | |
91 | ||
92 | Note that the peak rate TBF is much more tough: with MTU 1500 | |
93 | P_crit = 150Kbytes/sec. So, if you need greater peak | |
94 | rates, use alpha with HZ=1000 :-) | |
95 | ||
96 | With classful TBF, limit is just kept for backwards compatibility. | |
97 | It is passed to the default bfifo qdisc - if the inner qdisc is | |
98 | changed the limit is not effective anymore. | |
99 | */ | |
100 | ||
cc7ec456 | 101 | struct tbf_sched_data { |
1da177e4 LT |
102 | /* Parameters */ |
103 | u32 limit; /* Maximal length of backlog: bytes */ | |
b757c933 JP |
104 | s64 buffer; /* Token bucket depth/rate: MUST BE >= MTU/B */ |
105 | s64 mtu; | |
1da177e4 | 106 | u32 max_size; |
b757c933 JP |
107 | struct psched_ratecfg rate; |
108 | struct psched_ratecfg peak; | |
109 | bool peak_present; | |
1da177e4 LT |
110 | |
111 | /* Variables */ | |
b757c933 JP |
112 | s64 tokens; /* Current number of B tokens */ |
113 | s64 ptokens; /* Current number of P tokens */ | |
114 | s64 t_c; /* Time check-point */ | |
1da177e4 | 115 | struct Qdisc *qdisc; /* Inner qdisc, default - bfifo queue */ |
f7f593e3 | 116 | struct qdisc_watchdog watchdog; /* Watchdog timer */ |
1da177e4 LT |
117 | }; |
118 | ||
e43ac79a ED |
119 | |
120 | /* GSO packet is too big, segment it so that tbf can transmit | |
121 | * each segment in time | |
122 | */ | |
123 | static int tbf_segment(struct sk_buff *skb, struct Qdisc *sch) | |
124 | { | |
125 | struct tbf_sched_data *q = qdisc_priv(sch); | |
126 | struct sk_buff *segs, *nskb; | |
127 | netdev_features_t features = netif_skb_features(skb); | |
128 | int ret, nb; | |
129 | ||
130 | segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK); | |
131 | ||
132 | if (IS_ERR_OR_NULL(segs)) | |
133 | return qdisc_reshape_fail(skb, sch); | |
134 | ||
135 | nb = 0; | |
136 | while (segs) { | |
137 | nskb = segs->next; | |
138 | segs->next = NULL; | |
139 | if (likely(segs->len <= q->max_size)) { | |
140 | qdisc_skb_cb(segs)->pkt_len = segs->len; | |
141 | ret = qdisc_enqueue(segs, q->qdisc); | |
142 | } else { | |
143 | ret = qdisc_reshape_fail(skb, sch); | |
144 | } | |
145 | if (ret != NET_XMIT_SUCCESS) { | |
146 | if (net_xmit_drop_count(ret)) | |
147 | sch->qstats.drops++; | |
148 | } else { | |
149 | nb++; | |
150 | } | |
151 | segs = nskb; | |
152 | } | |
153 | sch->q.qlen += nb; | |
154 | if (nb > 1) | |
155 | qdisc_tree_decrease_qlen(sch, 1 - nb); | |
156 | consume_skb(skb); | |
157 | return nb > 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP; | |
158 | } | |
159 | ||
cc7ec456 | 160 | static int tbf_enqueue(struct sk_buff *skb, struct Qdisc *sch) |
1da177e4 LT |
161 | { |
162 | struct tbf_sched_data *q = qdisc_priv(sch); | |
163 | int ret; | |
164 | ||
e43ac79a ED |
165 | if (qdisc_pkt_len(skb) > q->max_size) { |
166 | if (skb_is_gso(skb)) | |
167 | return tbf_segment(skb, sch); | |
69747650 | 168 | return qdisc_reshape_fail(skb, sch); |
e43ac79a | 169 | } |
5f86173b | 170 | ret = qdisc_enqueue(skb, q->qdisc); |
9871e50e | 171 | if (ret != NET_XMIT_SUCCESS) { |
378a2f09 JP |
172 | if (net_xmit_drop_count(ret)) |
173 | sch->qstats.drops++; | |
1da177e4 LT |
174 | return ret; |
175 | } | |
176 | ||
177 | sch->q.qlen++; | |
9871e50e | 178 | return NET_XMIT_SUCCESS; |
1da177e4 LT |
179 | } |
180 | ||
cc7ec456 | 181 | static unsigned int tbf_drop(struct Qdisc *sch) |
1da177e4 LT |
182 | { |
183 | struct tbf_sched_data *q = qdisc_priv(sch); | |
6d037a26 | 184 | unsigned int len = 0; |
1da177e4 | 185 | |
6d037a26 | 186 | if (q->qdisc->ops->drop && (len = q->qdisc->ops->drop(q->qdisc)) != 0) { |
1da177e4 LT |
187 | sch->q.qlen--; |
188 | sch->qstats.drops++; | |
189 | } | |
190 | return len; | |
191 | } | |
192 | ||
cc7ec456 | 193 | static struct sk_buff *tbf_dequeue(struct Qdisc *sch) |
1da177e4 LT |
194 | { |
195 | struct tbf_sched_data *q = qdisc_priv(sch); | |
196 | struct sk_buff *skb; | |
197 | ||
03c05f0d | 198 | skb = q->qdisc->ops->peek(q->qdisc); |
1da177e4 LT |
199 | |
200 | if (skb) { | |
b757c933 JP |
201 | s64 now; |
202 | s64 toks; | |
203 | s64 ptoks = 0; | |
0abf77e5 | 204 | unsigned int len = qdisc_pkt_len(skb); |
1da177e4 | 205 | |
b757c933 JP |
206 | now = ktime_to_ns(ktime_get()); |
207 | toks = min_t(s64, now - q->t_c, q->buffer); | |
1da177e4 | 208 | |
b757c933 | 209 | if (q->peak_present) { |
1da177e4 | 210 | ptoks = toks + q->ptokens; |
b757c933 | 211 | if (ptoks > q->mtu) |
1da177e4 | 212 | ptoks = q->mtu; |
b757c933 | 213 | ptoks -= (s64) psched_l2t_ns(&q->peak, len); |
1da177e4 LT |
214 | } |
215 | toks += q->tokens; | |
b757c933 | 216 | if (toks > q->buffer) |
1da177e4 | 217 | toks = q->buffer; |
b757c933 | 218 | toks -= (s64) psched_l2t_ns(&q->rate, len); |
1da177e4 LT |
219 | |
220 | if ((toks|ptoks) >= 0) { | |
77be155c | 221 | skb = qdisc_dequeue_peeked(q->qdisc); |
03c05f0d JP |
222 | if (unlikely(!skb)) |
223 | return NULL; | |
224 | ||
1da177e4 LT |
225 | q->t_c = now; |
226 | q->tokens = toks; | |
227 | q->ptokens = ptoks; | |
228 | sch->q.qlen--; | |
fd245a4a | 229 | qdisc_unthrottled(sch); |
9190b3b3 | 230 | qdisc_bstats_update(sch, skb); |
1da177e4 LT |
231 | return skb; |
232 | } | |
233 | ||
b757c933 JP |
234 | qdisc_watchdog_schedule_ns(&q->watchdog, |
235 | now + max_t(long, -toks, -ptoks)); | |
1da177e4 LT |
236 | |
237 | /* Maybe we have a shorter packet in the queue, | |
238 | which can be sent now. It sounds cool, | |
239 | but, however, this is wrong in principle. | |
240 | We MUST NOT reorder packets under these circumstances. | |
241 | ||
242 | Really, if we split the flow into independent | |
243 | subflows, it would be a very good solution. | |
244 | This is the main idea of all FQ algorithms | |
245 | (cf. CSZ, HPFQ, HFSC) | |
246 | */ | |
247 | ||
1da177e4 LT |
248 | sch->qstats.overlimits++; |
249 | } | |
250 | return NULL; | |
251 | } | |
252 | ||
cc7ec456 | 253 | static void tbf_reset(struct Qdisc *sch) |
1da177e4 LT |
254 | { |
255 | struct tbf_sched_data *q = qdisc_priv(sch); | |
256 | ||
257 | qdisc_reset(q->qdisc); | |
258 | sch->q.qlen = 0; | |
b757c933 | 259 | q->t_c = ktime_to_ns(ktime_get()); |
1da177e4 LT |
260 | q->tokens = q->buffer; |
261 | q->ptokens = q->mtu; | |
f7f593e3 | 262 | qdisc_watchdog_cancel(&q->watchdog); |
1da177e4 LT |
263 | } |
264 | ||
27a3421e PM |
265 | static const struct nla_policy tbf_policy[TCA_TBF_MAX + 1] = { |
266 | [TCA_TBF_PARMS] = { .len = sizeof(struct tc_tbf_qopt) }, | |
267 | [TCA_TBF_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, | |
268 | [TCA_TBF_PTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, | |
269 | }; | |
270 | ||
cc7ec456 | 271 | static int tbf_change(struct Qdisc *sch, struct nlattr *opt) |
1da177e4 | 272 | { |
cee63723 | 273 | int err; |
1da177e4 | 274 | struct tbf_sched_data *q = qdisc_priv(sch); |
1e90474c | 275 | struct nlattr *tb[TCA_TBF_PTAB + 1]; |
1da177e4 LT |
276 | struct tc_tbf_qopt *qopt; |
277 | struct qdisc_rate_table *rtab = NULL; | |
278 | struct qdisc_rate_table *ptab = NULL; | |
279 | struct Qdisc *child = NULL; | |
cc7ec456 | 280 | int max_size, n; |
1da177e4 | 281 | |
27a3421e | 282 | err = nla_parse_nested(tb, TCA_TBF_PTAB, opt, tbf_policy); |
cee63723 PM |
283 | if (err < 0) |
284 | return err; | |
285 | ||
286 | err = -EINVAL; | |
27a3421e | 287 | if (tb[TCA_TBF_PARMS] == NULL) |
1da177e4 LT |
288 | goto done; |
289 | ||
1e90474c PM |
290 | qopt = nla_data(tb[TCA_TBF_PARMS]); |
291 | rtab = qdisc_get_rtab(&qopt->rate, tb[TCA_TBF_RTAB]); | |
1da177e4 LT |
292 | if (rtab == NULL) |
293 | goto done; | |
294 | ||
295 | if (qopt->peakrate.rate) { | |
296 | if (qopt->peakrate.rate > qopt->rate.rate) | |
1e90474c | 297 | ptab = qdisc_get_rtab(&qopt->peakrate, tb[TCA_TBF_PTAB]); |
1da177e4 LT |
298 | if (ptab == NULL) |
299 | goto done; | |
300 | } | |
301 | ||
302 | for (n = 0; n < 256; n++) | |
cc7ec456 ED |
303 | if (rtab->data[n] > qopt->buffer) |
304 | break; | |
305 | max_size = (n << qopt->rate.cell_log) - 1; | |
1da177e4 LT |
306 | if (ptab) { |
307 | int size; | |
308 | ||
309 | for (n = 0; n < 256; n++) | |
cc7ec456 ED |
310 | if (ptab->data[n] > qopt->mtu) |
311 | break; | |
312 | size = (n << qopt->peakrate.cell_log) - 1; | |
313 | if (size < max_size) | |
314 | max_size = size; | |
1da177e4 LT |
315 | } |
316 | if (max_size < 0) | |
317 | goto done; | |
318 | ||
f0cd1508 | 319 | if (q->qdisc != &noop_qdisc) { |
320 | err = fifo_set_limit(q->qdisc, qopt->limit); | |
321 | if (err) | |
322 | goto done; | |
323 | } else if (qopt->limit > 0) { | |
fb0305ce PM |
324 | child = fifo_create_dflt(sch, &bfifo_qdisc_ops, qopt->limit); |
325 | if (IS_ERR(child)) { | |
326 | err = PTR_ERR(child); | |
1da177e4 | 327 | goto done; |
fb0305ce | 328 | } |
1da177e4 LT |
329 | } |
330 | ||
331 | sch_tree_lock(sch); | |
5e50da01 PM |
332 | if (child) { |
333 | qdisc_tree_decrease_qlen(q->qdisc, q->qdisc->q.qlen); | |
b94c8afc PM |
334 | qdisc_destroy(q->qdisc); |
335 | q->qdisc = child; | |
5e50da01 | 336 | } |
1da177e4 | 337 | q->limit = qopt->limit; |
b757c933 | 338 | q->mtu = PSCHED_TICKS2NS(qopt->mtu); |
1da177e4 | 339 | q->max_size = max_size; |
b757c933 | 340 | q->buffer = PSCHED_TICKS2NS(qopt->buffer); |
1da177e4 LT |
341 | q->tokens = q->buffer; |
342 | q->ptokens = q->mtu; | |
b94c8afc | 343 | |
01cb71d2 | 344 | psched_ratecfg_precompute(&q->rate, &rtab->rate); |
b757c933 | 345 | if (ptab) { |
01cb71d2 | 346 | psched_ratecfg_precompute(&q->peak, &ptab->rate); |
b757c933 JP |
347 | q->peak_present = true; |
348 | } else { | |
349 | q->peak_present = false; | |
350 | } | |
b94c8afc | 351 | |
1da177e4 LT |
352 | sch_tree_unlock(sch); |
353 | err = 0; | |
354 | done: | |
355 | if (rtab) | |
356 | qdisc_put_rtab(rtab); | |
357 | if (ptab) | |
358 | qdisc_put_rtab(ptab); | |
359 | return err; | |
360 | } | |
361 | ||
cc7ec456 | 362 | static int tbf_init(struct Qdisc *sch, struct nlattr *opt) |
1da177e4 LT |
363 | { |
364 | struct tbf_sched_data *q = qdisc_priv(sch); | |
365 | ||
366 | if (opt == NULL) | |
367 | return -EINVAL; | |
368 | ||
b757c933 | 369 | q->t_c = ktime_to_ns(ktime_get()); |
f7f593e3 | 370 | qdisc_watchdog_init(&q->watchdog, sch); |
1da177e4 LT |
371 | q->qdisc = &noop_qdisc; |
372 | ||
373 | return tbf_change(sch, opt); | |
374 | } | |
375 | ||
376 | static void tbf_destroy(struct Qdisc *sch) | |
377 | { | |
378 | struct tbf_sched_data *q = qdisc_priv(sch); | |
379 | ||
f7f593e3 | 380 | qdisc_watchdog_cancel(&q->watchdog); |
1da177e4 LT |
381 | qdisc_destroy(q->qdisc); |
382 | } | |
383 | ||
384 | static int tbf_dump(struct Qdisc *sch, struct sk_buff *skb) | |
385 | { | |
386 | struct tbf_sched_data *q = qdisc_priv(sch); | |
4b3550ef | 387 | struct nlattr *nest; |
1da177e4 LT |
388 | struct tc_tbf_qopt opt; |
389 | ||
b0460e44 | 390 | sch->qstats.backlog = q->qdisc->qstats.backlog; |
4b3550ef PM |
391 | nest = nla_nest_start(skb, TCA_OPTIONS); |
392 | if (nest == NULL) | |
393 | goto nla_put_failure; | |
1da177e4 LT |
394 | |
395 | opt.limit = q->limit; | |
01cb71d2 | 396 | psched_ratecfg_getrate(&opt.rate, &q->rate); |
b757c933 | 397 | if (q->peak_present) |
01cb71d2 | 398 | psched_ratecfg_getrate(&opt.peakrate, &q->peak); |
1da177e4 LT |
399 | else |
400 | memset(&opt.peakrate, 0, sizeof(opt.peakrate)); | |
b757c933 JP |
401 | opt.mtu = PSCHED_NS2TICKS(q->mtu); |
402 | opt.buffer = PSCHED_NS2TICKS(q->buffer); | |
1b34ec43 DM |
403 | if (nla_put(skb, TCA_TBF_PARMS, sizeof(opt), &opt)) |
404 | goto nla_put_failure; | |
1da177e4 | 405 | |
4b3550ef | 406 | nla_nest_end(skb, nest); |
1da177e4 LT |
407 | return skb->len; |
408 | ||
1e90474c | 409 | nla_put_failure: |
4b3550ef | 410 | nla_nest_cancel(skb, nest); |
1da177e4 LT |
411 | return -1; |
412 | } | |
413 | ||
414 | static int tbf_dump_class(struct Qdisc *sch, unsigned long cl, | |
415 | struct sk_buff *skb, struct tcmsg *tcm) | |
416 | { | |
417 | struct tbf_sched_data *q = qdisc_priv(sch); | |
418 | ||
1da177e4 LT |
419 | tcm->tcm_handle |= TC_H_MIN(1); |
420 | tcm->tcm_info = q->qdisc->handle; | |
421 | ||
422 | return 0; | |
423 | } | |
424 | ||
425 | static int tbf_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, | |
426 | struct Qdisc **old) | |
427 | { | |
428 | struct tbf_sched_data *q = qdisc_priv(sch); | |
429 | ||
430 | if (new == NULL) | |
431 | new = &noop_qdisc; | |
432 | ||
433 | sch_tree_lock(sch); | |
b94c8afc PM |
434 | *old = q->qdisc; |
435 | q->qdisc = new; | |
5e50da01 | 436 | qdisc_tree_decrease_qlen(*old, (*old)->q.qlen); |
1da177e4 | 437 | qdisc_reset(*old); |
1da177e4 LT |
438 | sch_tree_unlock(sch); |
439 | ||
440 | return 0; | |
441 | } | |
442 | ||
443 | static struct Qdisc *tbf_leaf(struct Qdisc *sch, unsigned long arg) | |
444 | { | |
445 | struct tbf_sched_data *q = qdisc_priv(sch); | |
446 | return q->qdisc; | |
447 | } | |
448 | ||
449 | static unsigned long tbf_get(struct Qdisc *sch, u32 classid) | |
450 | { | |
451 | return 1; | |
452 | } | |
453 | ||
454 | static void tbf_put(struct Qdisc *sch, unsigned long arg) | |
455 | { | |
456 | } | |
457 | ||
1da177e4 LT |
458 | static void tbf_walk(struct Qdisc *sch, struct qdisc_walker *walker) |
459 | { | |
460 | if (!walker->stop) { | |
461 | if (walker->count >= walker->skip) | |
462 | if (walker->fn(sch, 1, walker) < 0) { | |
463 | walker->stop = 1; | |
464 | return; | |
465 | } | |
466 | walker->count++; | |
467 | } | |
468 | } | |
469 | ||
cc7ec456 | 470 | static const struct Qdisc_class_ops tbf_class_ops = { |
1da177e4 LT |
471 | .graft = tbf_graft, |
472 | .leaf = tbf_leaf, | |
473 | .get = tbf_get, | |
474 | .put = tbf_put, | |
1da177e4 | 475 | .walk = tbf_walk, |
1da177e4 LT |
476 | .dump = tbf_dump_class, |
477 | }; | |
478 | ||
20fea08b | 479 | static struct Qdisc_ops tbf_qdisc_ops __read_mostly = { |
1da177e4 LT |
480 | .next = NULL, |
481 | .cl_ops = &tbf_class_ops, | |
482 | .id = "tbf", | |
483 | .priv_size = sizeof(struct tbf_sched_data), | |
484 | .enqueue = tbf_enqueue, | |
485 | .dequeue = tbf_dequeue, | |
77be155c | 486 | .peek = qdisc_peek_dequeued, |
1da177e4 LT |
487 | .drop = tbf_drop, |
488 | .init = tbf_init, | |
489 | .reset = tbf_reset, | |
490 | .destroy = tbf_destroy, | |
491 | .change = tbf_change, | |
492 | .dump = tbf_dump, | |
493 | .owner = THIS_MODULE, | |
494 | }; | |
495 | ||
496 | static int __init tbf_module_init(void) | |
497 | { | |
498 | return register_qdisc(&tbf_qdisc_ops); | |
499 | } | |
500 | ||
501 | static void __exit tbf_module_exit(void) | |
502 | { | |
503 | unregister_qdisc(&tbf_qdisc_ops); | |
504 | } | |
505 | module_init(tbf_module_init) | |
506 | module_exit(tbf_module_exit) | |
507 | MODULE_LICENSE("GPL"); |