1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * net/sched/sch_tbf.c Token Bucket Filter queue. 4 * 5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 6 * Dmitry Torokhov <dtor@mail.ru> - allow attaching inner qdiscs - 7 * original idea by Martin Devera 8 */ 9 10 #include <linux/module.h> 11 #include <linux/types.h> 12 #include <linux/kernel.h> 13 #include <linux/string.h> 14 #include <linux/errno.h> 15 #include <linux/skbuff.h> 16 #include <net/netlink.h> 17 #include <net/sch_generic.h> 18 #include <net/pkt_cls.h> 19 #include <net/pkt_sched.h> 20 21 22 /* Simple Token Bucket Filter. 23 ======================================= 24 25 SOURCE. 26 ------- 27 28 None. 29 30 Description. 31 ------------ 32 33 A data flow obeys TBF with rate R and depth B, if for any 34 time interval t_i...t_f the number of transmitted bits 35 does not exceed B + R*(t_f-t_i). 36 37 Packetized version of this definition: 38 The sequence of packets of sizes s_i served at moments t_i 39 obeys TBF, if for any i<=k: 40 41 s_i+....+s_k <= B + R*(t_k - t_i) 42 43 Algorithm. 44 ---------- 45 46 Let N(t_i) be B/R initially and N(t) grow continuously with time as: 47 48 N(t+delta) = min{B/R, N(t) + delta} 49 50 If the first packet in queue has length S, it may be 51 transmitted only at the time t_* when S/R <= N(t_*), 52 and in this case N(t) jumps: 53 54 N(t_* + 0) = N(t_* - 0) - S/R. 55 56 57 58 Actually, QoS requires two TBF to be applied to a data stream. 59 One of them controls steady state burst size, another 60 one with rate P (peak rate) and depth M (equal to link MTU) 61 limits bursts at a smaller time scale. 62 63 It is easy to see that P>R, and B>M. If P is infinity, this double 64 TBF is equivalent to a single one. 65 66 When TBF works in reshaping mode, latency is estimated as: 67 68 lat = max ((L-B)/R, (L-M)/P) 69 70 71 NOTES. 72 ------ 73 74 If TBF throttles, it starts a watchdog timer, which will wake it up 75 when it is ready to transmit. 76 Note that the minimal timer resolution is 1/HZ. 77 If no new packets arrive during this period, 78 or if the device is not awaken by EOI for some previous packet, 79 TBF can stop its activity for 1/HZ. 80 81 82 This means, that with depth B, the maximal rate is 83 84 R_crit = B*HZ 85 86 F.e. for 10Mbit ethernet and HZ=100 the minimal allowed B is ~10Kbytes. 87 88 Note that the peak rate TBF is much more tough: with MTU 1500 89 P_crit = 150Kbytes/sec. So, if you need greater peak 90 rates, use alpha with HZ=1000 :-) 91 92 With classful TBF, limit is just kept for backwards compatibility. 93 It is passed to the default bfifo qdisc - if the inner qdisc is 94 changed the limit is not effective anymore. 95 */ 96 97 struct tbf_sched_data { 98 /* Parameters */ 99 u32 limit; /* Maximal length of backlog: bytes */ 100 u32 max_size; 101 s64 buffer; /* Token bucket depth/rate: MUST BE >= MTU/B */ 102 s64 mtu; 103 struct psched_ratecfg rate; 104 struct psched_ratecfg peak; 105 106 /* Variables */ 107 s64 tokens; /* Current number of B tokens */ 108 s64 ptokens; /* Current number of P tokens */ 109 s64 t_c; /* Time check-point */ 110 struct Qdisc *qdisc; /* Inner qdisc, default - bfifo queue */ 111 struct qdisc_watchdog watchdog; /* Watchdog timer */ 112 }; 113 114 115 /* Time to Length, convert time in ns to length in bytes 116 * to determinate how many bytes can be sent in given time. 117 */ 118 static u64 psched_ns_t2l(const struct psched_ratecfg *r, 119 u64 time_in_ns) 120 { 121 /* The formula is : 122 * len = (time_in_ns * r->rate_bytes_ps) / NSEC_PER_SEC 123 */ 124 u64 len = time_in_ns * r->rate_bytes_ps; 125 126 do_div(len, NSEC_PER_SEC); 127 128 if (unlikely(r->linklayer == TC_LINKLAYER_ATM)) { 129 do_div(len, 53); 130 len = len * 48; 131 } 132 133 if (len > r->overhead) 134 len -= r->overhead; 135 else 136 len = 0; 137 138 return len; 139 } 140 141 static void tbf_offload_change(struct Qdisc *sch) 142 { 143 struct tbf_sched_data *q = qdisc_priv(sch); 144 struct net_device *dev = qdisc_dev(sch); 145 struct tc_tbf_qopt_offload qopt; 146 147 if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc) 148 return; 149 150 qopt.command = TC_TBF_REPLACE; 151 qopt.handle = sch->handle; 152 qopt.parent = sch->parent; 153 qopt.replace_params.rate = q->rate; 154 qopt.replace_params.max_size = q->max_size; 155 qopt.replace_params.qstats = &sch->qstats; 156 157 dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_QDISC_TBF, &qopt); 158 } 159 160 static void tbf_offload_destroy(struct Qdisc *sch) 161 { 162 struct net_device *dev = qdisc_dev(sch); 163 struct tc_tbf_qopt_offload qopt; 164 165 if (!tc_can_offload(dev) || !dev->netdev_ops->ndo_setup_tc) 166 return; 167 168 qopt.command = TC_TBF_DESTROY; 169 qopt.handle = sch->handle; 170 qopt.parent = sch->parent; 171 dev->netdev_ops->ndo_setup_tc(dev, TC_SETUP_QDISC_TBF, &qopt); 172 } 173 174 static int tbf_offload_dump(struct Qdisc *sch) 175 { 176 struct tc_tbf_qopt_offload qopt; 177 178 qopt.command = TC_TBF_STATS; 179 qopt.handle = sch->handle; 180 qopt.parent = sch->parent; 181 qopt.stats.bstats = &sch->bstats; 182 qopt.stats.qstats = &sch->qstats; 183 184 return qdisc_offload_dump_helper(sch, TC_SETUP_QDISC_TBF, &qopt); 185 } 186 187 static void tbf_offload_graft(struct Qdisc *sch, struct Qdisc *new, 188 struct Qdisc *old, struct netlink_ext_ack *extack) 189 { 190 struct tc_tbf_qopt_offload graft_offload = { 191 .handle = sch->handle, 192 .parent = sch->parent, 193 .child_handle = new->handle, 194 .command = TC_TBF_GRAFT, 195 }; 196 197 qdisc_offload_graft_helper(qdisc_dev(sch), sch, new, old, 198 TC_SETUP_QDISC_TBF, &graft_offload, extack); 199 } 200 201 /* GSO packet is too big, segment it so that tbf can transmit 202 * each segment in time 203 */ 204 static int tbf_segment(struct sk_buff *skb, struct Qdisc *sch, 205 struct sk_buff **to_free) 206 { 207 struct tbf_sched_data *q = qdisc_priv(sch); 208 struct sk_buff *segs, *nskb; 209 netdev_features_t features = netif_skb_features(skb); 210 unsigned int len = 0, prev_len = qdisc_pkt_len(skb); 211 int ret, nb; 212 213 segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK); 214 215 if (IS_ERR_OR_NULL(segs)) 216 return qdisc_drop(skb, sch, to_free); 217 218 nb = 0; 219 skb_list_walk_safe(segs, segs, nskb) { 220 skb_mark_not_on_list(segs); 221 qdisc_skb_cb(segs)->pkt_len = segs->len; 222 len += segs->len; 223 ret = qdisc_enqueue(segs, q->qdisc, to_free); 224 if (ret != NET_XMIT_SUCCESS) { 225 if (net_xmit_drop_count(ret)) 226 qdisc_qstats_drop(sch); 227 } else { 228 nb++; 229 } 230 } 231 sch->q.qlen += nb; 232 if (nb > 1) 233 qdisc_tree_reduce_backlog(sch, 1 - nb, prev_len - len); 234 consume_skb(skb); 235 return nb > 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP; 236 } 237 238 static int tbf_enqueue(struct sk_buff *skb, struct Qdisc *sch, 239 struct sk_buff **to_free) 240 { 241 struct tbf_sched_data *q = qdisc_priv(sch); 242 unsigned int len = qdisc_pkt_len(skb); 243 int ret; 244 245 if (qdisc_pkt_len(skb) > q->max_size) { 246 if (skb_is_gso(skb) && 247 skb_gso_validate_mac_len(skb, q->max_size)) 248 return tbf_segment(skb, sch, to_free); 249 return qdisc_drop(skb, sch, to_free); 250 } 251 ret = qdisc_enqueue(skb, q->qdisc, to_free); 252 if (ret != NET_XMIT_SUCCESS) { 253 if (net_xmit_drop_count(ret)) 254 qdisc_qstats_drop(sch); 255 return ret; 256 } 257 258 sch->qstats.backlog += len; 259 sch->q.qlen++; 260 return NET_XMIT_SUCCESS; 261 } 262 263 static bool tbf_peak_present(const struct tbf_sched_data *q) 264 { 265 return q->peak.rate_bytes_ps; 266 } 267 268 static struct sk_buff *tbf_dequeue(struct Qdisc *sch) 269 { 270 struct tbf_sched_data *q = qdisc_priv(sch); 271 struct sk_buff *skb; 272 273 skb = q->qdisc->ops->peek(q->qdisc); 274 275 if (skb) { 276 s64 now; 277 s64 toks; 278 s64 ptoks = 0; 279 unsigned int len = qdisc_pkt_len(skb); 280 281 now = ktime_get_ns(); 282 toks = min_t(s64, now - q->t_c, q->buffer); 283 284 if (tbf_peak_present(q)) { 285 ptoks = toks + q->ptokens; 286 if (ptoks > q->mtu) 287 ptoks = q->mtu; 288 ptoks -= (s64) psched_l2t_ns(&q->peak, len); 289 } 290 toks += q->tokens; 291 if (toks > q->buffer) 292 toks = q->buffer; 293 toks -= (s64) psched_l2t_ns(&q->rate, len); 294 295 if ((toks|ptoks) >= 0) { 296 skb = qdisc_dequeue_peeked(q->qdisc); 297 if (unlikely(!skb)) 298 return NULL; 299 300 q->t_c = now; 301 q->tokens = toks; 302 q->ptokens = ptoks; 303 qdisc_qstats_backlog_dec(sch, skb); 304 sch->q.qlen--; 305 qdisc_bstats_update(sch, skb); 306 return skb; 307 } 308 309 qdisc_watchdog_schedule_ns(&q->watchdog, 310 now + max_t(long, -toks, -ptoks)); 311 312 /* Maybe we have a shorter packet in the queue, 313 which can be sent now. It sounds cool, 314 but, however, this is wrong in principle. 315 We MUST NOT reorder packets under these circumstances. 316 317 Really, if we split the flow into independent 318 subflows, it would be a very good solution. 319 This is the main idea of all FQ algorithms 320 (cf. CSZ, HPFQ, HFSC) 321 */ 322 323 qdisc_qstats_overlimit(sch); 324 } 325 return NULL; 326 } 327 328 static void tbf_reset(struct Qdisc *sch) 329 { 330 struct tbf_sched_data *q = qdisc_priv(sch); 331 332 qdisc_reset(q->qdisc); 333 sch->qstats.backlog = 0; 334 sch->q.qlen = 0; 335 q->t_c = ktime_get_ns(); 336 q->tokens = q->buffer; 337 q->ptokens = q->mtu; 338 qdisc_watchdog_cancel(&q->watchdog); 339 } 340 341 static const struct nla_policy tbf_policy[TCA_TBF_MAX + 1] = { 342 [TCA_TBF_PARMS] = { .len = sizeof(struct tc_tbf_qopt) }, 343 [TCA_TBF_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 344 [TCA_TBF_PTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 345 [TCA_TBF_RATE64] = { .type = NLA_U64 }, 346 [TCA_TBF_PRATE64] = { .type = NLA_U64 }, 347 [TCA_TBF_BURST] = { .type = NLA_U32 }, 348 [TCA_TBF_PBURST] = { .type = NLA_U32 }, 349 }; 350 351 static int tbf_change(struct Qdisc *sch, struct nlattr *opt, 352 struct netlink_ext_ack *extack) 353 { 354 int err; 355 struct tbf_sched_data *q = qdisc_priv(sch); 356 struct nlattr *tb[TCA_TBF_MAX + 1]; 357 struct tc_tbf_qopt *qopt; 358 struct Qdisc *child = NULL; 359 struct Qdisc *old = NULL; 360 struct psched_ratecfg rate; 361 struct psched_ratecfg peak; 362 u64 max_size; 363 s64 buffer, mtu; 364 u64 rate64 = 0, prate64 = 0; 365 366 err = nla_parse_nested_deprecated(tb, TCA_TBF_MAX, opt, tbf_policy, 367 NULL); 368 if (err < 0) 369 return err; 370 371 err = -EINVAL; 372 if (tb[TCA_TBF_PARMS] == NULL) 373 goto done; 374 375 qopt = nla_data(tb[TCA_TBF_PARMS]); 376 if (qopt->rate.linklayer == TC_LINKLAYER_UNAWARE) 377 qdisc_put_rtab(qdisc_get_rtab(&qopt->rate, 378 tb[TCA_TBF_RTAB], 379 NULL)); 380 381 if (qopt->peakrate.linklayer == TC_LINKLAYER_UNAWARE) 382 qdisc_put_rtab(qdisc_get_rtab(&qopt->peakrate, 383 tb[TCA_TBF_PTAB], 384 NULL)); 385 386 buffer = min_t(u64, PSCHED_TICKS2NS(qopt->buffer), ~0U); 387 mtu = min_t(u64, PSCHED_TICKS2NS(qopt->mtu), ~0U); 388 389 if (tb[TCA_TBF_RATE64]) 390 rate64 = nla_get_u64(tb[TCA_TBF_RATE64]); 391 psched_ratecfg_precompute(&rate, &qopt->rate, rate64); 392 393 if (tb[TCA_TBF_BURST]) { 394 max_size = nla_get_u32(tb[TCA_TBF_BURST]); 395 buffer = psched_l2t_ns(&rate, max_size); 396 } else { 397 max_size = min_t(u64, psched_ns_t2l(&rate, buffer), ~0U); 398 } 399 400 if (qopt->peakrate.rate) { 401 if (tb[TCA_TBF_PRATE64]) 402 prate64 = nla_get_u64(tb[TCA_TBF_PRATE64]); 403 psched_ratecfg_precompute(&peak, &qopt->peakrate, prate64); 404 if (peak.rate_bytes_ps <= rate.rate_bytes_ps) { 405 pr_warn_ratelimited("sch_tbf: peakrate %llu is lower than or equals to rate %llu !\n", 406 peak.rate_bytes_ps, rate.rate_bytes_ps); 407 err = -EINVAL; 408 goto done; 409 } 410 411 if (tb[TCA_TBF_PBURST]) { 412 u32 pburst = nla_get_u32(tb[TCA_TBF_PBURST]); 413 max_size = min_t(u32, max_size, pburst); 414 mtu = psched_l2t_ns(&peak, pburst); 415 } else { 416 max_size = min_t(u64, max_size, psched_ns_t2l(&peak, mtu)); 417 } 418 } else { 419 memset(&peak, 0, sizeof(peak)); 420 } 421 422 if (max_size < psched_mtu(qdisc_dev(sch))) 423 pr_warn_ratelimited("sch_tbf: burst %llu is lower than device %s mtu (%u) !\n", 424 max_size, qdisc_dev(sch)->name, 425 psched_mtu(qdisc_dev(sch))); 426 427 if (!max_size) { 428 err = -EINVAL; 429 goto done; 430 } 431 432 if (q->qdisc != &noop_qdisc) { 433 err = fifo_set_limit(q->qdisc, qopt->limit); 434 if (err) 435 goto done; 436 } else if (qopt->limit > 0) { 437 child = fifo_create_dflt(sch, &bfifo_qdisc_ops, qopt->limit, 438 extack); 439 if (IS_ERR(child)) { 440 err = PTR_ERR(child); 441 goto done; 442 } 443 444 /* child is fifo, no need to check for noop_qdisc */ 445 qdisc_hash_add(child, true); 446 } 447 448 sch_tree_lock(sch); 449 if (child) { 450 qdisc_tree_flush_backlog(q->qdisc); 451 old = q->qdisc; 452 q->qdisc = child; 453 } 454 q->limit = qopt->limit; 455 if (tb[TCA_TBF_PBURST]) 456 q->mtu = mtu; 457 else 458 q->mtu = PSCHED_TICKS2NS(qopt->mtu); 459 q->max_size = max_size; 460 if (tb[TCA_TBF_BURST]) 461 q->buffer = buffer; 462 else 463 q->buffer = PSCHED_TICKS2NS(qopt->buffer); 464 q->tokens = q->buffer; 465 q->ptokens = q->mtu; 466 467 memcpy(&q->rate, &rate, sizeof(struct psched_ratecfg)); 468 memcpy(&q->peak, &peak, sizeof(struct psched_ratecfg)); 469 470 sch_tree_unlock(sch); 471 qdisc_put(old); 472 err = 0; 473 474 tbf_offload_change(sch); 475 done: 476 return err; 477 } 478 479 static int tbf_init(struct Qdisc *sch, struct nlattr *opt, 480 struct netlink_ext_ack *extack) 481 { 482 struct tbf_sched_data *q = qdisc_priv(sch); 483 484 qdisc_watchdog_init(&q->watchdog, sch); 485 q->qdisc = &noop_qdisc; 486 487 if (!opt) 488 return -EINVAL; 489 490 q->t_c = ktime_get_ns(); 491 492 return tbf_change(sch, opt, extack); 493 } 494 495 static void tbf_destroy(struct Qdisc *sch) 496 { 497 struct tbf_sched_data *q = qdisc_priv(sch); 498 499 qdisc_watchdog_cancel(&q->watchdog); 500 tbf_offload_destroy(sch); 501 qdisc_put(q->qdisc); 502 } 503 504 static int tbf_dump(struct Qdisc *sch, struct sk_buff *skb) 505 { 506 struct tbf_sched_data *q = qdisc_priv(sch); 507 struct nlattr *nest; 508 struct tc_tbf_qopt opt; 509 int err; 510 511 err = tbf_offload_dump(sch); 512 if (err) 513 return err; 514 515 nest = nla_nest_start_noflag(skb, TCA_OPTIONS); 516 if (nest == NULL) 517 goto nla_put_failure; 518 519 opt.limit = q->limit; 520 psched_ratecfg_getrate(&opt.rate, &q->rate); 521 if (tbf_peak_present(q)) 522 psched_ratecfg_getrate(&opt.peakrate, &q->peak); 523 else 524 memset(&opt.peakrate, 0, sizeof(opt.peakrate)); 525 opt.mtu = PSCHED_NS2TICKS(q->mtu); 526 opt.buffer = PSCHED_NS2TICKS(q->buffer); 527 if (nla_put(skb, TCA_TBF_PARMS, sizeof(opt), &opt)) 528 goto nla_put_failure; 529 if (q->rate.rate_bytes_ps >= (1ULL << 32) && 530 nla_put_u64_64bit(skb, TCA_TBF_RATE64, q->rate.rate_bytes_ps, 531 TCA_TBF_PAD)) 532 goto nla_put_failure; 533 if (tbf_peak_present(q) && 534 q->peak.rate_bytes_ps >= (1ULL << 32) && 535 nla_put_u64_64bit(skb, TCA_TBF_PRATE64, q->peak.rate_bytes_ps, 536 TCA_TBF_PAD)) 537 goto nla_put_failure; 538 539 return nla_nest_end(skb, nest); 540 541 nla_put_failure: 542 nla_nest_cancel(skb, nest); 543 return -1; 544 } 545 546 static int tbf_dump_class(struct Qdisc *sch, unsigned long cl, 547 struct sk_buff *skb, struct tcmsg *tcm) 548 { 549 struct tbf_sched_data *q = qdisc_priv(sch); 550 551 tcm->tcm_handle |= TC_H_MIN(1); 552 tcm->tcm_info = q->qdisc->handle; 553 554 return 0; 555 } 556 557 static int tbf_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 558 struct Qdisc **old, struct netlink_ext_ack *extack) 559 { 560 struct tbf_sched_data *q = qdisc_priv(sch); 561 562 if (new == NULL) 563 new = &noop_qdisc; 564 565 *old = qdisc_replace(sch, new, &q->qdisc); 566 567 tbf_offload_graft(sch, new, *old, extack); 568 return 0; 569 } 570 571 static struct Qdisc *tbf_leaf(struct Qdisc *sch, unsigned long arg) 572 { 573 struct tbf_sched_data *q = qdisc_priv(sch); 574 return q->qdisc; 575 } 576 577 static unsigned long tbf_find(struct Qdisc *sch, u32 classid) 578 { 579 return 1; 580 } 581 582 static void tbf_walk(struct Qdisc *sch, struct qdisc_walker *walker) 583 { 584 if (!walker->stop) { 585 if (walker->count >= walker->skip) 586 if (walker->fn(sch, 1, walker) < 0) { 587 walker->stop = 1; 588 return; 589 } 590 walker->count++; 591 } 592 } 593 594 static const struct Qdisc_class_ops tbf_class_ops = { 595 .graft = tbf_graft, 596 .leaf = tbf_leaf, 597 .find = tbf_find, 598 .walk = tbf_walk, 599 .dump = tbf_dump_class, 600 }; 601 602 static struct Qdisc_ops tbf_qdisc_ops __read_mostly = { 603 .next = NULL, 604 .cl_ops = &tbf_class_ops, 605 .id = "tbf", 606 .priv_size = sizeof(struct tbf_sched_data), 607 .enqueue = tbf_enqueue, 608 .dequeue = tbf_dequeue, 609 .peek = qdisc_peek_dequeued, 610 .init = tbf_init, 611 .reset = tbf_reset, 612 .destroy = tbf_destroy, 613 .change = tbf_change, 614 .dump = tbf_dump, 615 .owner = THIS_MODULE, 616 }; 617 618 static int __init tbf_module_init(void) 619 { 620 return register_qdisc(&tbf_qdisc_ops); 621 } 622 623 static void __exit tbf_module_exit(void) 624 { 625 unregister_qdisc(&tbf_qdisc_ops); 626 } 627 module_init(tbf_module_init) 628 module_exit(tbf_module_exit) 629 MODULE_LICENSE("GPL"); 630