1 /* 2 * net/sched/sch_netem.c Network emulator 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 * Many of the algorithms and ideas for this came from 10 * NIST Net which is not copyrighted. 11 * 12 * Authors: Stephen Hemminger <shemminger@osdl.org> 13 * Catalin(ux aka Dino) BOIE <catab at umbrella dot ro> 14 */ 15 16 #include <linux/config.h> 17 #include <linux/module.h> 18 #include <linux/bitops.h> 19 #include <linux/types.h> 20 #include <linux/kernel.h> 21 #include <linux/errno.h> 22 #include <linux/netdevice.h> 23 #include <linux/skbuff.h> 24 #include <linux/rtnetlink.h> 25 26 #include <net/pkt_sched.h> 27 28 #define VERSION "1.1" 29 30 /* Network Emulation Queuing algorithm. 31 ==================================== 32 33 Sources: [1] Mark Carson, Darrin Santay, "NIST Net - A Linux-based 34 Network Emulation Tool 35 [2] Luigi Rizzo, DummyNet for FreeBSD 36 37 ---------------------------------------------------------------- 38 39 This started out as a simple way to delay outgoing packets to 40 test TCP but has grown to include most of the functionality 41 of a full blown network emulator like NISTnet. It can delay 42 packets and add random jitter (and correlation). The random 43 distribution can be loaded from a table as well to provide 44 normal, Pareto, or experimental curves. Packet loss, 45 duplication, and reordering can also be emulated. 46 47 This qdisc does not do classification that can be handled in 48 layering other disciplines. It does not need to do bandwidth 49 control either since that can be handled by using token 50 bucket or other rate control. 51 52 The simulator is limited by the Linux timer resolution 53 and will create packet bursts on the HZ boundary (1ms). 54 */ 55 56 struct netem_sched_data { 57 struct Qdisc *qdisc; 58 struct timer_list timer; 59 60 u32 latency; 61 u32 loss; 62 u32 limit; 63 u32 counter; 64 u32 gap; 65 u32 jitter; 66 u32 duplicate; 67 u32 reorder; 68 69 struct crndstate { 70 unsigned long last; 71 unsigned long rho; 72 } delay_cor, loss_cor, dup_cor, reorder_cor; 73 74 struct disttable { 75 u32 size; 76 s16 table[0]; 77 } *delay_dist; 78 }; 79 80 /* Time stamp put into socket buffer control block */ 81 struct netem_skb_cb { 82 psched_time_t time_to_send; 83 }; 84 85 /* init_crandom - initialize correlated random number generator 86 * Use entropy source for initial seed. 87 */ 88 static void init_crandom(struct crndstate *state, unsigned long rho) 89 { 90 state->rho = rho; 91 state->last = net_random(); 92 } 93 94 /* get_crandom - correlated random number generator 95 * Next number depends on last value. 96 * rho is scaled to avoid floating point. 97 */ 98 static unsigned long get_crandom(struct crndstate *state) 99 { 100 u64 value, rho; 101 unsigned long answer; 102 103 if (state->rho == 0) /* no correllation */ 104 return net_random(); 105 106 value = net_random(); 107 rho = (u64)state->rho + 1; 108 answer = (value * ((1ull<<32) - rho) + state->last * rho) >> 32; 109 state->last = answer; 110 return answer; 111 } 112 113 /* tabledist - return a pseudo-randomly distributed value with mean mu and 114 * std deviation sigma. Uses table lookup to approximate the desired 115 * distribution, and a uniformly-distributed pseudo-random source. 116 */ 117 static long tabledist(unsigned long mu, long sigma, 118 struct crndstate *state, const struct disttable *dist) 119 { 120 long t, x; 121 unsigned long rnd; 122 123 if (sigma == 0) 124 return mu; 125 126 rnd = get_crandom(state); 127 128 /* default uniform distribution */ 129 if (dist == NULL) 130 return (rnd % (2*sigma)) - sigma + mu; 131 132 t = dist->table[rnd % dist->size]; 133 x = (sigma % NETEM_DIST_SCALE) * t; 134 if (x >= 0) 135 x += NETEM_DIST_SCALE/2; 136 else 137 x -= NETEM_DIST_SCALE/2; 138 139 return x / NETEM_DIST_SCALE + (sigma / NETEM_DIST_SCALE) * t + mu; 140 } 141 142 /* 143 * Insert one skb into qdisc. 144 * Note: parent depends on return value to account for queue length. 145 * NET_XMIT_DROP: queue length didn't change. 146 * NET_XMIT_SUCCESS: one skb was queued. 147 */ 148 static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch) 149 { 150 struct netem_sched_data *q = qdisc_priv(sch); 151 struct netem_skb_cb *cb = (struct netem_skb_cb *)skb->cb; 152 struct sk_buff *skb2; 153 int ret; 154 int count = 1; 155 156 pr_debug("netem_enqueue skb=%p\n", skb); 157 158 /* Random duplication */ 159 if (q->duplicate && q->duplicate >= get_crandom(&q->dup_cor)) 160 ++count; 161 162 /* Random packet drop 0 => none, ~0 => all */ 163 if (q->loss && q->loss >= get_crandom(&q->loss_cor)) 164 --count; 165 166 if (count == 0) { 167 sch->qstats.drops++; 168 kfree_skb(skb); 169 return NET_XMIT_DROP; 170 } 171 172 /* 173 * If we need to duplicate packet, then re-insert at top of the 174 * qdisc tree, since parent queuer expects that only one 175 * skb will be queued. 176 */ 177 if (count > 1 && (skb2 = skb_clone(skb, GFP_ATOMIC)) != NULL) { 178 struct Qdisc *rootq = sch->dev->qdisc; 179 u32 dupsave = q->duplicate; /* prevent duplicating a dup... */ 180 q->duplicate = 0; 181 182 rootq->enqueue(skb2, rootq); 183 q->duplicate = dupsave; 184 } 185 186 if (q->gap == 0 /* not doing reordering */ 187 || q->counter < q->gap /* inside last reordering gap */ 188 || q->reorder < get_crandom(&q->reorder_cor)) { 189 psched_time_t now; 190 psched_tdiff_t delay; 191 192 delay = tabledist(q->latency, q->jitter, 193 &q->delay_cor, q->delay_dist); 194 195 PSCHED_GET_TIME(now); 196 PSCHED_TADD2(now, delay, cb->time_to_send); 197 ++q->counter; 198 ret = q->qdisc->enqueue(skb, q->qdisc); 199 } else { 200 /* 201 * Do re-ordering by putting one out of N packets at the front 202 * of the queue. 203 */ 204 PSCHED_GET_TIME(cb->time_to_send); 205 q->counter = 0; 206 ret = q->qdisc->ops->requeue(skb, q->qdisc); 207 } 208 209 if (likely(ret == NET_XMIT_SUCCESS)) { 210 sch->q.qlen++; 211 sch->bstats.bytes += skb->len; 212 sch->bstats.packets++; 213 } else 214 sch->qstats.drops++; 215 216 pr_debug("netem: enqueue ret %d\n", ret); 217 return ret; 218 } 219 220 /* Requeue packets but don't change time stamp */ 221 static int netem_requeue(struct sk_buff *skb, struct Qdisc *sch) 222 { 223 struct netem_sched_data *q = qdisc_priv(sch); 224 int ret; 225 226 if ((ret = q->qdisc->ops->requeue(skb, q->qdisc)) == 0) { 227 sch->q.qlen++; 228 sch->qstats.requeues++; 229 } 230 231 return ret; 232 } 233 234 static unsigned int netem_drop(struct Qdisc* sch) 235 { 236 struct netem_sched_data *q = qdisc_priv(sch); 237 unsigned int len; 238 239 if ((len = q->qdisc->ops->drop(q->qdisc)) != 0) { 240 sch->q.qlen--; 241 sch->qstats.drops++; 242 } 243 return len; 244 } 245 246 static struct sk_buff *netem_dequeue(struct Qdisc *sch) 247 { 248 struct netem_sched_data *q = qdisc_priv(sch); 249 struct sk_buff *skb; 250 251 skb = q->qdisc->dequeue(q->qdisc); 252 if (skb) { 253 const struct netem_skb_cb *cb 254 = (const struct netem_skb_cb *)skb->cb; 255 psched_time_t now; 256 257 /* if more time remaining? */ 258 PSCHED_GET_TIME(now); 259 260 if (PSCHED_TLESS(cb->time_to_send, now)) { 261 pr_debug("netem_dequeue: return skb=%p\n", skb); 262 sch->q.qlen--; 263 sch->flags &= ~TCQ_F_THROTTLED; 264 return skb; 265 } else { 266 psched_tdiff_t delay = PSCHED_TDIFF(cb->time_to_send, now); 267 268 if (q->qdisc->ops->requeue(skb, q->qdisc) != NET_XMIT_SUCCESS) { 269 sch->qstats.drops++; 270 271 /* After this qlen is confused */ 272 printk(KERN_ERR "netem: queue discpline %s could not requeue\n", 273 q->qdisc->ops->id); 274 275 sch->q.qlen--; 276 } 277 278 mod_timer(&q->timer, jiffies + PSCHED_US2JIFFIE(delay)); 279 sch->flags |= TCQ_F_THROTTLED; 280 } 281 } 282 283 return NULL; 284 } 285 286 static void netem_watchdog(unsigned long arg) 287 { 288 struct Qdisc *sch = (struct Qdisc *)arg; 289 290 pr_debug("netem_watchdog qlen=%d\n", sch->q.qlen); 291 sch->flags &= ~TCQ_F_THROTTLED; 292 netif_schedule(sch->dev); 293 } 294 295 static void netem_reset(struct Qdisc *sch) 296 { 297 struct netem_sched_data *q = qdisc_priv(sch); 298 299 qdisc_reset(q->qdisc); 300 sch->q.qlen = 0; 301 sch->flags &= ~TCQ_F_THROTTLED; 302 del_timer_sync(&q->timer); 303 } 304 305 /* Pass size change message down to embedded FIFO */ 306 static int set_fifo_limit(struct Qdisc *q, int limit) 307 { 308 struct rtattr *rta; 309 int ret = -ENOMEM; 310 311 /* Hack to avoid sending change message to non-FIFO */ 312 if (strncmp(q->ops->id + 1, "fifo", 4) != 0) 313 return 0; 314 315 rta = kmalloc(RTA_LENGTH(sizeof(struct tc_fifo_qopt)), GFP_KERNEL); 316 if (rta) { 317 rta->rta_type = RTM_NEWQDISC; 318 rta->rta_len = RTA_LENGTH(sizeof(struct tc_fifo_qopt)); 319 ((struct tc_fifo_qopt *)RTA_DATA(rta))->limit = limit; 320 321 ret = q->ops->change(q, rta); 322 kfree(rta); 323 } 324 return ret; 325 } 326 327 /* 328 * Distribution data is a variable size payload containing 329 * signed 16 bit values. 330 */ 331 static int get_dist_table(struct Qdisc *sch, const struct rtattr *attr) 332 { 333 struct netem_sched_data *q = qdisc_priv(sch); 334 unsigned long n = RTA_PAYLOAD(attr)/sizeof(__s16); 335 const __s16 *data = RTA_DATA(attr); 336 struct disttable *d; 337 int i; 338 339 if (n > 65536) 340 return -EINVAL; 341 342 d = kmalloc(sizeof(*d) + n*sizeof(d->table[0]), GFP_KERNEL); 343 if (!d) 344 return -ENOMEM; 345 346 d->size = n; 347 for (i = 0; i < n; i++) 348 d->table[i] = data[i]; 349 350 spin_lock_bh(&sch->dev->queue_lock); 351 d = xchg(&q->delay_dist, d); 352 spin_unlock_bh(&sch->dev->queue_lock); 353 354 kfree(d); 355 return 0; 356 } 357 358 static int get_correlation(struct Qdisc *sch, const struct rtattr *attr) 359 { 360 struct netem_sched_data *q = qdisc_priv(sch); 361 const struct tc_netem_corr *c = RTA_DATA(attr); 362 363 if (RTA_PAYLOAD(attr) != sizeof(*c)) 364 return -EINVAL; 365 366 init_crandom(&q->delay_cor, c->delay_corr); 367 init_crandom(&q->loss_cor, c->loss_corr); 368 init_crandom(&q->dup_cor, c->dup_corr); 369 return 0; 370 } 371 372 static int get_reorder(struct Qdisc *sch, const struct rtattr *attr) 373 { 374 struct netem_sched_data *q = qdisc_priv(sch); 375 const struct tc_netem_reorder *r = RTA_DATA(attr); 376 377 if (RTA_PAYLOAD(attr) != sizeof(*r)) 378 return -EINVAL; 379 380 q->reorder = r->probability; 381 init_crandom(&q->reorder_cor, r->correlation); 382 return 0; 383 } 384 385 static int netem_change(struct Qdisc *sch, struct rtattr *opt) 386 { 387 struct netem_sched_data *q = qdisc_priv(sch); 388 struct tc_netem_qopt *qopt; 389 int ret; 390 391 if (opt == NULL || RTA_PAYLOAD(opt) < sizeof(*qopt)) 392 return -EINVAL; 393 394 qopt = RTA_DATA(opt); 395 ret = set_fifo_limit(q->qdisc, qopt->limit); 396 if (ret) { 397 pr_debug("netem: can't set fifo limit\n"); 398 return ret; 399 } 400 401 q->latency = qopt->latency; 402 q->jitter = qopt->jitter; 403 q->limit = qopt->limit; 404 q->gap = qopt->gap; 405 q->counter = 0; 406 q->loss = qopt->loss; 407 q->duplicate = qopt->duplicate; 408 409 /* for compatiablity with earlier versions. 410 * if gap is set, need to assume 100% probablity 411 */ 412 q->reorder = ~0; 413 414 /* Handle nested options after initial queue options. 415 * Should have put all options in nested format but too late now. 416 */ 417 if (RTA_PAYLOAD(opt) > sizeof(*qopt)) { 418 struct rtattr *tb[TCA_NETEM_MAX]; 419 if (rtattr_parse(tb, TCA_NETEM_MAX, 420 RTA_DATA(opt) + sizeof(*qopt), 421 RTA_PAYLOAD(opt) - sizeof(*qopt))) 422 return -EINVAL; 423 424 if (tb[TCA_NETEM_CORR-1]) { 425 ret = get_correlation(sch, tb[TCA_NETEM_CORR-1]); 426 if (ret) 427 return ret; 428 } 429 430 if (tb[TCA_NETEM_DELAY_DIST-1]) { 431 ret = get_dist_table(sch, tb[TCA_NETEM_DELAY_DIST-1]); 432 if (ret) 433 return ret; 434 } 435 if (tb[TCA_NETEM_REORDER-1]) { 436 ret = get_reorder(sch, tb[TCA_NETEM_REORDER-1]); 437 if (ret) 438 return ret; 439 } 440 } 441 442 443 return 0; 444 } 445 446 /* 447 * Special case version of FIFO queue for use by netem. 448 * It queues in order based on timestamps in skb's 449 */ 450 struct fifo_sched_data { 451 u32 limit; 452 }; 453 454 static int tfifo_enqueue(struct sk_buff *nskb, struct Qdisc *sch) 455 { 456 struct fifo_sched_data *q = qdisc_priv(sch); 457 struct sk_buff_head *list = &sch->q; 458 const struct netem_skb_cb *ncb 459 = (const struct netem_skb_cb *)nskb->cb; 460 struct sk_buff *skb; 461 462 if (likely(skb_queue_len(list) < q->limit)) { 463 skb_queue_reverse_walk(list, skb) { 464 const struct netem_skb_cb *cb 465 = (const struct netem_skb_cb *)skb->cb; 466 467 if (!PSCHED_TLESS(ncb->time_to_send, cb->time_to_send)) 468 break; 469 } 470 471 __skb_queue_after(list, skb, nskb); 472 473 sch->qstats.backlog += nskb->len; 474 sch->bstats.bytes += nskb->len; 475 sch->bstats.packets++; 476 477 return NET_XMIT_SUCCESS; 478 } 479 480 return qdisc_drop(nskb, sch); 481 } 482 483 static int tfifo_init(struct Qdisc *sch, struct rtattr *opt) 484 { 485 struct fifo_sched_data *q = qdisc_priv(sch); 486 487 if (opt) { 488 struct tc_fifo_qopt *ctl = RTA_DATA(opt); 489 if (RTA_PAYLOAD(opt) < sizeof(*ctl)) 490 return -EINVAL; 491 492 q->limit = ctl->limit; 493 } else 494 q->limit = max_t(u32, sch->dev->tx_queue_len, 1); 495 496 return 0; 497 } 498 499 static int tfifo_dump(struct Qdisc *sch, struct sk_buff *skb) 500 { 501 struct fifo_sched_data *q = qdisc_priv(sch); 502 struct tc_fifo_qopt opt = { .limit = q->limit }; 503 504 RTA_PUT(skb, TCA_OPTIONS, sizeof(opt), &opt); 505 return skb->len; 506 507 rtattr_failure: 508 return -1; 509 } 510 511 static struct Qdisc_ops tfifo_qdisc_ops = { 512 .id = "tfifo", 513 .priv_size = sizeof(struct fifo_sched_data), 514 .enqueue = tfifo_enqueue, 515 .dequeue = qdisc_dequeue_head, 516 .requeue = qdisc_requeue, 517 .drop = qdisc_queue_drop, 518 .init = tfifo_init, 519 .reset = qdisc_reset_queue, 520 .change = tfifo_init, 521 .dump = tfifo_dump, 522 }; 523 524 static int netem_init(struct Qdisc *sch, struct rtattr *opt) 525 { 526 struct netem_sched_data *q = qdisc_priv(sch); 527 int ret; 528 529 if (!opt) 530 return -EINVAL; 531 532 init_timer(&q->timer); 533 q->timer.function = netem_watchdog; 534 q->timer.data = (unsigned long) sch; 535 536 q->qdisc = qdisc_create_dflt(sch->dev, &tfifo_qdisc_ops); 537 if (!q->qdisc) { 538 pr_debug("netem: qdisc create failed\n"); 539 return -ENOMEM; 540 } 541 542 ret = netem_change(sch, opt); 543 if (ret) { 544 pr_debug("netem: change failed\n"); 545 qdisc_destroy(q->qdisc); 546 } 547 return ret; 548 } 549 550 static void netem_destroy(struct Qdisc *sch) 551 { 552 struct netem_sched_data *q = qdisc_priv(sch); 553 554 del_timer_sync(&q->timer); 555 qdisc_destroy(q->qdisc); 556 kfree(q->delay_dist); 557 } 558 559 static int netem_dump(struct Qdisc *sch, struct sk_buff *skb) 560 { 561 const struct netem_sched_data *q = qdisc_priv(sch); 562 unsigned char *b = skb->tail; 563 struct rtattr *rta = (struct rtattr *) b; 564 struct tc_netem_qopt qopt; 565 struct tc_netem_corr cor; 566 struct tc_netem_reorder reorder; 567 568 qopt.latency = q->latency; 569 qopt.jitter = q->jitter; 570 qopt.limit = q->limit; 571 qopt.loss = q->loss; 572 qopt.gap = q->gap; 573 qopt.duplicate = q->duplicate; 574 RTA_PUT(skb, TCA_OPTIONS, sizeof(qopt), &qopt); 575 576 cor.delay_corr = q->delay_cor.rho; 577 cor.loss_corr = q->loss_cor.rho; 578 cor.dup_corr = q->dup_cor.rho; 579 RTA_PUT(skb, TCA_NETEM_CORR, sizeof(cor), &cor); 580 581 reorder.probability = q->reorder; 582 reorder.correlation = q->reorder_cor.rho; 583 RTA_PUT(skb, TCA_NETEM_REORDER, sizeof(reorder), &reorder); 584 585 rta->rta_len = skb->tail - b; 586 587 return skb->len; 588 589 rtattr_failure: 590 skb_trim(skb, b - skb->data); 591 return -1; 592 } 593 594 static int netem_dump_class(struct Qdisc *sch, unsigned long cl, 595 struct sk_buff *skb, struct tcmsg *tcm) 596 { 597 struct netem_sched_data *q = qdisc_priv(sch); 598 599 if (cl != 1) /* only one class */ 600 return -ENOENT; 601 602 tcm->tcm_handle |= TC_H_MIN(1); 603 tcm->tcm_info = q->qdisc->handle; 604 605 return 0; 606 } 607 608 static int netem_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 609 struct Qdisc **old) 610 { 611 struct netem_sched_data *q = qdisc_priv(sch); 612 613 if (new == NULL) 614 new = &noop_qdisc; 615 616 sch_tree_lock(sch); 617 *old = xchg(&q->qdisc, new); 618 qdisc_reset(*old); 619 sch->q.qlen = 0; 620 sch_tree_unlock(sch); 621 622 return 0; 623 } 624 625 static struct Qdisc *netem_leaf(struct Qdisc *sch, unsigned long arg) 626 { 627 struct netem_sched_data *q = qdisc_priv(sch); 628 return q->qdisc; 629 } 630 631 static unsigned long netem_get(struct Qdisc *sch, u32 classid) 632 { 633 return 1; 634 } 635 636 static void netem_put(struct Qdisc *sch, unsigned long arg) 637 { 638 } 639 640 static int netem_change_class(struct Qdisc *sch, u32 classid, u32 parentid, 641 struct rtattr **tca, unsigned long *arg) 642 { 643 return -ENOSYS; 644 } 645 646 static int netem_delete(struct Qdisc *sch, unsigned long arg) 647 { 648 return -ENOSYS; 649 } 650 651 static void netem_walk(struct Qdisc *sch, struct qdisc_walker *walker) 652 { 653 if (!walker->stop) { 654 if (walker->count >= walker->skip) 655 if (walker->fn(sch, 1, walker) < 0) { 656 walker->stop = 1; 657 return; 658 } 659 walker->count++; 660 } 661 } 662 663 static struct tcf_proto **netem_find_tcf(struct Qdisc *sch, unsigned long cl) 664 { 665 return NULL; 666 } 667 668 static struct Qdisc_class_ops netem_class_ops = { 669 .graft = netem_graft, 670 .leaf = netem_leaf, 671 .get = netem_get, 672 .put = netem_put, 673 .change = netem_change_class, 674 .delete = netem_delete, 675 .walk = netem_walk, 676 .tcf_chain = netem_find_tcf, 677 .dump = netem_dump_class, 678 }; 679 680 static struct Qdisc_ops netem_qdisc_ops = { 681 .id = "netem", 682 .cl_ops = &netem_class_ops, 683 .priv_size = sizeof(struct netem_sched_data), 684 .enqueue = netem_enqueue, 685 .dequeue = netem_dequeue, 686 .requeue = netem_requeue, 687 .drop = netem_drop, 688 .init = netem_init, 689 .reset = netem_reset, 690 .destroy = netem_destroy, 691 .change = netem_change, 692 .dump = netem_dump, 693 .owner = THIS_MODULE, 694 }; 695 696 697 static int __init netem_module_init(void) 698 { 699 pr_info("netem: version " VERSION "\n"); 700 return register_qdisc(&netem_qdisc_ops); 701 } 702 static void __exit netem_module_exit(void) 703 { 704 unregister_qdisc(&netem_qdisc_ops); 705 } 706 module_init(netem_module_init) 707 module_exit(netem_module_exit) 708 MODULE_LICENSE("GPL"); 709