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. 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/mm.h> 17 #include <linux/module.h> 18 #include <linux/slab.h> 19 #include <linux/types.h> 20 #include <linux/kernel.h> 21 #include <linux/errno.h> 22 #include <linux/skbuff.h> 23 #include <linux/vmalloc.h> 24 #include <linux/rtnetlink.h> 25 #include <linux/reciprocal_div.h> 26 #include <linux/rbtree.h> 27 28 #include <net/netlink.h> 29 #include <net/pkt_sched.h> 30 #include <net/inet_ecn.h> 31 32 #define VERSION "1.3" 33 34 /* Network Emulation Queuing algorithm. 35 ==================================== 36 37 Sources: [1] Mark Carson, Darrin Santay, "NIST Net - A Linux-based 38 Network Emulation Tool 39 [2] Luigi Rizzo, DummyNet for FreeBSD 40 41 ---------------------------------------------------------------- 42 43 This started out as a simple way to delay outgoing packets to 44 test TCP but has grown to include most of the functionality 45 of a full blown network emulator like NISTnet. It can delay 46 packets and add random jitter (and correlation). The random 47 distribution can be loaded from a table as well to provide 48 normal, Pareto, or experimental curves. Packet loss, 49 duplication, and reordering can also be emulated. 50 51 This qdisc does not do classification that can be handled in 52 layering other disciplines. It does not need to do bandwidth 53 control either since that can be handled by using token 54 bucket or other rate control. 55 56 Correlated Loss Generator models 57 58 Added generation of correlated loss according to the 59 "Gilbert-Elliot" model, a 4-state markov model. 60 61 References: 62 [1] NetemCLG Home http://netgroup.uniroma2.it/NetemCLG 63 [2] S. Salsano, F. Ludovici, A. Ordine, "Definition of a general 64 and intuitive loss model for packet networks and its implementation 65 in the Netem module in the Linux kernel", available in [1] 66 67 Authors: Stefano Salsano <stefano.salsano at uniroma2.it 68 Fabio Ludovici <fabio.ludovici at yahoo.it> 69 */ 70 71 struct netem_sched_data { 72 /* internal t(ime)fifo qdisc uses t_root and sch->limit */ 73 struct rb_root t_root; 74 75 /* optional qdisc for classful handling (NULL at netem init) */ 76 struct Qdisc *qdisc; 77 78 struct qdisc_watchdog watchdog; 79 80 psched_tdiff_t latency; 81 psched_tdiff_t jitter; 82 83 u32 loss; 84 u32 ecn; 85 u32 limit; 86 u32 counter; 87 u32 gap; 88 u32 duplicate; 89 u32 reorder; 90 u32 corrupt; 91 u64 rate; 92 s32 packet_overhead; 93 u32 cell_size; 94 struct reciprocal_value cell_size_reciprocal; 95 s32 cell_overhead; 96 97 struct crndstate { 98 u32 last; 99 u32 rho; 100 } delay_cor, loss_cor, dup_cor, reorder_cor, corrupt_cor; 101 102 struct disttable { 103 u32 size; 104 s16 table[0]; 105 } *delay_dist; 106 107 enum { 108 CLG_RANDOM, 109 CLG_4_STATES, 110 CLG_GILB_ELL, 111 } loss_model; 112 113 enum { 114 TX_IN_GAP_PERIOD = 1, 115 TX_IN_BURST_PERIOD, 116 LOST_IN_GAP_PERIOD, 117 LOST_IN_BURST_PERIOD, 118 } _4_state_model; 119 120 /* Correlated Loss Generation models */ 121 struct clgstate { 122 /* state of the Markov chain */ 123 u8 state; 124 125 /* 4-states and Gilbert-Elliot models */ 126 u32 a1; /* p13 for 4-states or p for GE */ 127 u32 a2; /* p31 for 4-states or r for GE */ 128 u32 a3; /* p32 for 4-states or h for GE */ 129 u32 a4; /* p14 for 4-states or 1-k for GE */ 130 u32 a5; /* p23 used only in 4-states */ 131 } clg; 132 133 }; 134 135 /* Time stamp put into socket buffer control block 136 * Only valid when skbs are in our internal t(ime)fifo queue. 137 */ 138 struct netem_skb_cb { 139 psched_time_t time_to_send; 140 ktime_t tstamp_save; 141 }; 142 143 /* Because space in skb->cb[] is tight, netem overloads skb->next/prev/tstamp 144 * to hold a rb_node structure. 145 * 146 * If struct sk_buff layout is changed, the following checks will complain. 147 */ 148 static struct rb_node *netem_rb_node(struct sk_buff *skb) 149 { 150 BUILD_BUG_ON(offsetof(struct sk_buff, next) != 0); 151 BUILD_BUG_ON(offsetof(struct sk_buff, prev) != 152 offsetof(struct sk_buff, next) + sizeof(skb->next)); 153 BUILD_BUG_ON(offsetof(struct sk_buff, tstamp) != 154 offsetof(struct sk_buff, prev) + sizeof(skb->prev)); 155 BUILD_BUG_ON(sizeof(struct rb_node) > sizeof(skb->next) + 156 sizeof(skb->prev) + 157 sizeof(skb->tstamp)); 158 return (struct rb_node *)&skb->next; 159 } 160 161 static struct sk_buff *netem_rb_to_skb(struct rb_node *rb) 162 { 163 return (struct sk_buff *)rb; 164 } 165 166 static inline struct netem_skb_cb *netem_skb_cb(struct sk_buff *skb) 167 { 168 /* we assume we can use skb next/prev/tstamp as storage for rb_node */ 169 qdisc_cb_private_validate(skb, sizeof(struct netem_skb_cb)); 170 return (struct netem_skb_cb *)qdisc_skb_cb(skb)->data; 171 } 172 173 /* init_crandom - initialize correlated random number generator 174 * Use entropy source for initial seed. 175 */ 176 static void init_crandom(struct crndstate *state, unsigned long rho) 177 { 178 state->rho = rho; 179 state->last = prandom_u32(); 180 } 181 182 /* get_crandom - correlated random number generator 183 * Next number depends on last value. 184 * rho is scaled to avoid floating point. 185 */ 186 static u32 get_crandom(struct crndstate *state) 187 { 188 u64 value, rho; 189 unsigned long answer; 190 191 if (state->rho == 0) /* no correlation */ 192 return prandom_u32(); 193 194 value = prandom_u32(); 195 rho = (u64)state->rho + 1; 196 answer = (value * ((1ull<<32) - rho) + state->last * rho) >> 32; 197 state->last = answer; 198 return answer; 199 } 200 201 /* loss_4state - 4-state model loss generator 202 * Generates losses according to the 4-state Markov chain adopted in 203 * the GI (General and Intuitive) loss model. 204 */ 205 static bool loss_4state(struct netem_sched_data *q) 206 { 207 struct clgstate *clg = &q->clg; 208 u32 rnd = prandom_u32(); 209 210 /* 211 * Makes a comparison between rnd and the transition 212 * probabilities outgoing from the current state, then decides the 213 * next state and if the next packet has to be transmitted or lost. 214 * The four states correspond to: 215 * TX_IN_GAP_PERIOD => successfully transmitted packets within a gap period 216 * LOST_IN_BURST_PERIOD => isolated losses within a gap period 217 * LOST_IN_GAP_PERIOD => lost packets within a burst period 218 * TX_IN_GAP_PERIOD => successfully transmitted packets within a burst period 219 */ 220 switch (clg->state) { 221 case TX_IN_GAP_PERIOD: 222 if (rnd < clg->a4) { 223 clg->state = LOST_IN_BURST_PERIOD; 224 return true; 225 } else if (clg->a4 < rnd && rnd < clg->a1 + clg->a4) { 226 clg->state = LOST_IN_GAP_PERIOD; 227 return true; 228 } else if (clg->a1 + clg->a4 < rnd) { 229 clg->state = TX_IN_GAP_PERIOD; 230 } 231 232 break; 233 case TX_IN_BURST_PERIOD: 234 if (rnd < clg->a5) { 235 clg->state = LOST_IN_GAP_PERIOD; 236 return true; 237 } else { 238 clg->state = TX_IN_BURST_PERIOD; 239 } 240 241 break; 242 case LOST_IN_GAP_PERIOD: 243 if (rnd < clg->a3) 244 clg->state = TX_IN_BURST_PERIOD; 245 else if (clg->a3 < rnd && rnd < clg->a2 + clg->a3) { 246 clg->state = TX_IN_GAP_PERIOD; 247 } else if (clg->a2 + clg->a3 < rnd) { 248 clg->state = LOST_IN_GAP_PERIOD; 249 return true; 250 } 251 break; 252 case LOST_IN_BURST_PERIOD: 253 clg->state = TX_IN_GAP_PERIOD; 254 break; 255 } 256 257 return false; 258 } 259 260 /* loss_gilb_ell - Gilbert-Elliot model loss generator 261 * Generates losses according to the Gilbert-Elliot loss model or 262 * its special cases (Gilbert or Simple Gilbert) 263 * 264 * Makes a comparison between random number and the transition 265 * probabilities outgoing from the current state, then decides the 266 * next state. A second random number is extracted and the comparison 267 * with the loss probability of the current state decides if the next 268 * packet will be transmitted or lost. 269 */ 270 static bool loss_gilb_ell(struct netem_sched_data *q) 271 { 272 struct clgstate *clg = &q->clg; 273 274 switch (clg->state) { 275 case 1: 276 if (prandom_u32() < clg->a1) 277 clg->state = 2; 278 if (prandom_u32() < clg->a4) 279 return true; 280 break; 281 case 2: 282 if (prandom_u32() < clg->a2) 283 clg->state = 1; 284 if (prandom_u32() > clg->a3) 285 return true; 286 } 287 288 return false; 289 } 290 291 static bool loss_event(struct netem_sched_data *q) 292 { 293 switch (q->loss_model) { 294 case CLG_RANDOM: 295 /* Random packet drop 0 => none, ~0 => all */ 296 return q->loss && q->loss >= get_crandom(&q->loss_cor); 297 298 case CLG_4_STATES: 299 /* 4state loss model algorithm (used also for GI model) 300 * Extracts a value from the markov 4 state loss generator, 301 * if it is 1 drops a packet and if needed writes the event in 302 * the kernel logs 303 */ 304 return loss_4state(q); 305 306 case CLG_GILB_ELL: 307 /* Gilbert-Elliot loss model algorithm 308 * Extracts a value from the Gilbert-Elliot loss generator, 309 * if it is 1 drops a packet and if needed writes the event in 310 * the kernel logs 311 */ 312 return loss_gilb_ell(q); 313 } 314 315 return false; /* not reached */ 316 } 317 318 319 /* tabledist - return a pseudo-randomly distributed value with mean mu and 320 * std deviation sigma. Uses table lookup to approximate the desired 321 * distribution, and a uniformly-distributed pseudo-random source. 322 */ 323 static psched_tdiff_t tabledist(psched_tdiff_t mu, psched_tdiff_t sigma, 324 struct crndstate *state, 325 const struct disttable *dist) 326 { 327 psched_tdiff_t x; 328 long t; 329 u32 rnd; 330 331 if (sigma == 0) 332 return mu; 333 334 rnd = get_crandom(state); 335 336 /* default uniform distribution */ 337 if (dist == NULL) 338 return (rnd % (2*sigma)) - sigma + mu; 339 340 t = dist->table[rnd % dist->size]; 341 x = (sigma % NETEM_DIST_SCALE) * t; 342 if (x >= 0) 343 x += NETEM_DIST_SCALE/2; 344 else 345 x -= NETEM_DIST_SCALE/2; 346 347 return x / NETEM_DIST_SCALE + (sigma / NETEM_DIST_SCALE) * t + mu; 348 } 349 350 static psched_time_t packet_len_2_sched_time(unsigned int len, struct netem_sched_data *q) 351 { 352 u64 ticks; 353 354 len += q->packet_overhead; 355 356 if (q->cell_size) { 357 u32 cells = reciprocal_divide(len, q->cell_size_reciprocal); 358 359 if (len > cells * q->cell_size) /* extra cell needed for remainder */ 360 cells++; 361 len = cells * (q->cell_size + q->cell_overhead); 362 } 363 364 ticks = (u64)len * NSEC_PER_SEC; 365 366 do_div(ticks, q->rate); 367 return PSCHED_NS2TICKS(ticks); 368 } 369 370 static void tfifo_reset(struct Qdisc *sch) 371 { 372 struct netem_sched_data *q = qdisc_priv(sch); 373 struct rb_node *p; 374 375 while ((p = rb_first(&q->t_root))) { 376 struct sk_buff *skb = netem_rb_to_skb(p); 377 378 rb_erase(p, &q->t_root); 379 skb->next = NULL; 380 skb->prev = NULL; 381 kfree_skb(skb); 382 } 383 } 384 385 static void tfifo_enqueue(struct sk_buff *nskb, struct Qdisc *sch) 386 { 387 struct netem_sched_data *q = qdisc_priv(sch); 388 psched_time_t tnext = netem_skb_cb(nskb)->time_to_send; 389 struct rb_node **p = &q->t_root.rb_node, *parent = NULL; 390 391 while (*p) { 392 struct sk_buff *skb; 393 394 parent = *p; 395 skb = netem_rb_to_skb(parent); 396 if (tnext >= netem_skb_cb(skb)->time_to_send) 397 p = &parent->rb_right; 398 else 399 p = &parent->rb_left; 400 } 401 rb_link_node(netem_rb_node(nskb), parent, p); 402 rb_insert_color(netem_rb_node(nskb), &q->t_root); 403 sch->q.qlen++; 404 } 405 406 /* 407 * Insert one skb into qdisc. 408 * Note: parent depends on return value to account for queue length. 409 * NET_XMIT_DROP: queue length didn't change. 410 * NET_XMIT_SUCCESS: one skb was queued. 411 */ 412 static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch) 413 { 414 struct netem_sched_data *q = qdisc_priv(sch); 415 /* We don't fill cb now as skb_unshare() may invalidate it */ 416 struct netem_skb_cb *cb; 417 struct sk_buff *skb2; 418 int count = 1; 419 420 /* Random duplication */ 421 if (q->duplicate && q->duplicate >= get_crandom(&q->dup_cor)) 422 ++count; 423 424 /* Drop packet? */ 425 if (loss_event(q)) { 426 if (q->ecn && INET_ECN_set_ce(skb)) 427 sch->qstats.drops++; /* mark packet */ 428 else 429 --count; 430 } 431 if (count == 0) { 432 sch->qstats.drops++; 433 kfree_skb(skb); 434 return NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 435 } 436 437 /* If a delay is expected, orphan the skb. (orphaning usually takes 438 * place at TX completion time, so _before_ the link transit delay) 439 */ 440 if (q->latency || q->jitter) 441 skb_orphan_partial(skb); 442 443 /* 444 * If we need to duplicate packet, then re-insert at top of the 445 * qdisc tree, since parent queuer expects that only one 446 * skb will be queued. 447 */ 448 if (count > 1 && (skb2 = skb_clone(skb, GFP_ATOMIC)) != NULL) { 449 struct Qdisc *rootq = qdisc_root(sch); 450 u32 dupsave = q->duplicate; /* prevent duplicating a dup... */ 451 q->duplicate = 0; 452 453 qdisc_enqueue_root(skb2, rootq); 454 q->duplicate = dupsave; 455 } 456 457 /* 458 * Randomized packet corruption. 459 * Make copy if needed since we are modifying 460 * If packet is going to be hardware checksummed, then 461 * do it now in software before we mangle it. 462 */ 463 if (q->corrupt && q->corrupt >= get_crandom(&q->corrupt_cor)) { 464 if (!(skb = skb_unshare(skb, GFP_ATOMIC)) || 465 (skb->ip_summed == CHECKSUM_PARTIAL && 466 skb_checksum_help(skb))) 467 return qdisc_drop(skb, sch); 468 469 skb->data[prandom_u32() % skb_headlen(skb)] ^= 470 1<<(prandom_u32() % 8); 471 } 472 473 if (unlikely(skb_queue_len(&sch->q) >= sch->limit)) 474 return qdisc_reshape_fail(skb, sch); 475 476 sch->qstats.backlog += qdisc_pkt_len(skb); 477 478 cb = netem_skb_cb(skb); 479 if (q->gap == 0 || /* not doing reordering */ 480 q->counter < q->gap - 1 || /* inside last reordering gap */ 481 q->reorder < get_crandom(&q->reorder_cor)) { 482 psched_time_t now; 483 psched_tdiff_t delay; 484 485 delay = tabledist(q->latency, q->jitter, 486 &q->delay_cor, q->delay_dist); 487 488 now = psched_get_time(); 489 490 if (q->rate) { 491 struct sk_buff *last; 492 493 if (!skb_queue_empty(&sch->q)) 494 last = skb_peek_tail(&sch->q); 495 else 496 last = netem_rb_to_skb(rb_last(&q->t_root)); 497 if (last) { 498 /* 499 * Last packet in queue is reference point (now), 500 * calculate this time bonus and subtract 501 * from delay. 502 */ 503 delay -= netem_skb_cb(last)->time_to_send - now; 504 delay = max_t(psched_tdiff_t, 0, delay); 505 now = netem_skb_cb(last)->time_to_send; 506 } 507 508 delay += packet_len_2_sched_time(qdisc_pkt_len(skb), q); 509 } 510 511 cb->time_to_send = now + delay; 512 cb->tstamp_save = skb->tstamp; 513 ++q->counter; 514 tfifo_enqueue(skb, sch); 515 } else { 516 /* 517 * Do re-ordering by putting one out of N packets at the front 518 * of the queue. 519 */ 520 cb->time_to_send = psched_get_time(); 521 q->counter = 0; 522 523 __skb_queue_head(&sch->q, skb); 524 sch->qstats.requeues++; 525 } 526 527 return NET_XMIT_SUCCESS; 528 } 529 530 static unsigned int netem_drop(struct Qdisc *sch) 531 { 532 struct netem_sched_data *q = qdisc_priv(sch); 533 unsigned int len; 534 535 len = qdisc_queue_drop(sch); 536 537 if (!len) { 538 struct rb_node *p = rb_first(&q->t_root); 539 540 if (p) { 541 struct sk_buff *skb = netem_rb_to_skb(p); 542 543 rb_erase(p, &q->t_root); 544 sch->q.qlen--; 545 skb->next = NULL; 546 skb->prev = NULL; 547 len = qdisc_pkt_len(skb); 548 sch->qstats.backlog -= len; 549 kfree_skb(skb); 550 } 551 } 552 if (!len && q->qdisc && q->qdisc->ops->drop) 553 len = q->qdisc->ops->drop(q->qdisc); 554 if (len) 555 sch->qstats.drops++; 556 557 return len; 558 } 559 560 static struct sk_buff *netem_dequeue(struct Qdisc *sch) 561 { 562 struct netem_sched_data *q = qdisc_priv(sch); 563 struct sk_buff *skb; 564 struct rb_node *p; 565 566 if (qdisc_is_throttled(sch)) 567 return NULL; 568 569 tfifo_dequeue: 570 skb = __skb_dequeue(&sch->q); 571 if (skb) { 572 deliver: 573 sch->qstats.backlog -= qdisc_pkt_len(skb); 574 qdisc_unthrottled(sch); 575 qdisc_bstats_update(sch, skb); 576 return skb; 577 } 578 p = rb_first(&q->t_root); 579 if (p) { 580 psched_time_t time_to_send; 581 582 skb = netem_rb_to_skb(p); 583 584 /* if more time remaining? */ 585 time_to_send = netem_skb_cb(skb)->time_to_send; 586 if (time_to_send <= psched_get_time()) { 587 rb_erase(p, &q->t_root); 588 589 sch->q.qlen--; 590 skb->next = NULL; 591 skb->prev = NULL; 592 skb->tstamp = netem_skb_cb(skb)->tstamp_save; 593 594 #ifdef CONFIG_NET_CLS_ACT 595 /* 596 * If it's at ingress let's pretend the delay is 597 * from the network (tstamp will be updated). 598 */ 599 if (G_TC_FROM(skb->tc_verd) & AT_INGRESS) 600 skb->tstamp.tv64 = 0; 601 #endif 602 603 if (q->qdisc) { 604 int err = qdisc_enqueue(skb, q->qdisc); 605 606 if (unlikely(err != NET_XMIT_SUCCESS)) { 607 if (net_xmit_drop_count(err)) { 608 sch->qstats.drops++; 609 qdisc_tree_decrease_qlen(sch, 1); 610 } 611 } 612 goto tfifo_dequeue; 613 } 614 goto deliver; 615 } 616 617 if (q->qdisc) { 618 skb = q->qdisc->ops->dequeue(q->qdisc); 619 if (skb) 620 goto deliver; 621 } 622 qdisc_watchdog_schedule(&q->watchdog, time_to_send); 623 } 624 625 if (q->qdisc) { 626 skb = q->qdisc->ops->dequeue(q->qdisc); 627 if (skb) 628 goto deliver; 629 } 630 return NULL; 631 } 632 633 static void netem_reset(struct Qdisc *sch) 634 { 635 struct netem_sched_data *q = qdisc_priv(sch); 636 637 qdisc_reset_queue(sch); 638 tfifo_reset(sch); 639 if (q->qdisc) 640 qdisc_reset(q->qdisc); 641 qdisc_watchdog_cancel(&q->watchdog); 642 } 643 644 static void dist_free(struct disttable *d) 645 { 646 if (d) { 647 if (is_vmalloc_addr(d)) 648 vfree(d); 649 else 650 kfree(d); 651 } 652 } 653 654 /* 655 * Distribution data is a variable size payload containing 656 * signed 16 bit values. 657 */ 658 static int get_dist_table(struct Qdisc *sch, const struct nlattr *attr) 659 { 660 struct netem_sched_data *q = qdisc_priv(sch); 661 size_t n = nla_len(attr)/sizeof(__s16); 662 const __s16 *data = nla_data(attr); 663 spinlock_t *root_lock; 664 struct disttable *d; 665 int i; 666 size_t s; 667 668 if (n > NETEM_DIST_MAX) 669 return -EINVAL; 670 671 s = sizeof(struct disttable) + n * sizeof(s16); 672 d = kmalloc(s, GFP_KERNEL | __GFP_NOWARN); 673 if (!d) 674 d = vmalloc(s); 675 if (!d) 676 return -ENOMEM; 677 678 d->size = n; 679 for (i = 0; i < n; i++) 680 d->table[i] = data[i]; 681 682 root_lock = qdisc_root_sleeping_lock(sch); 683 684 spin_lock_bh(root_lock); 685 swap(q->delay_dist, d); 686 spin_unlock_bh(root_lock); 687 688 dist_free(d); 689 return 0; 690 } 691 692 static void get_correlation(struct Qdisc *sch, const struct nlattr *attr) 693 { 694 struct netem_sched_data *q = qdisc_priv(sch); 695 const struct tc_netem_corr *c = nla_data(attr); 696 697 init_crandom(&q->delay_cor, c->delay_corr); 698 init_crandom(&q->loss_cor, c->loss_corr); 699 init_crandom(&q->dup_cor, c->dup_corr); 700 } 701 702 static void get_reorder(struct Qdisc *sch, const struct nlattr *attr) 703 { 704 struct netem_sched_data *q = qdisc_priv(sch); 705 const struct tc_netem_reorder *r = nla_data(attr); 706 707 q->reorder = r->probability; 708 init_crandom(&q->reorder_cor, r->correlation); 709 } 710 711 static void get_corrupt(struct Qdisc *sch, const struct nlattr *attr) 712 { 713 struct netem_sched_data *q = qdisc_priv(sch); 714 const struct tc_netem_corrupt *r = nla_data(attr); 715 716 q->corrupt = r->probability; 717 init_crandom(&q->corrupt_cor, r->correlation); 718 } 719 720 static void get_rate(struct Qdisc *sch, const struct nlattr *attr) 721 { 722 struct netem_sched_data *q = qdisc_priv(sch); 723 const struct tc_netem_rate *r = nla_data(attr); 724 725 q->rate = r->rate; 726 q->packet_overhead = r->packet_overhead; 727 q->cell_size = r->cell_size; 728 q->cell_overhead = r->cell_overhead; 729 if (q->cell_size) 730 q->cell_size_reciprocal = reciprocal_value(q->cell_size); 731 else 732 q->cell_size_reciprocal = (struct reciprocal_value) { 0 }; 733 } 734 735 static int get_loss_clg(struct Qdisc *sch, const struct nlattr *attr) 736 { 737 struct netem_sched_data *q = qdisc_priv(sch); 738 const struct nlattr *la; 739 int rem; 740 741 nla_for_each_nested(la, attr, rem) { 742 u16 type = nla_type(la); 743 744 switch (type) { 745 case NETEM_LOSS_GI: { 746 const struct tc_netem_gimodel *gi = nla_data(la); 747 748 if (nla_len(la) < sizeof(struct tc_netem_gimodel)) { 749 pr_info("netem: incorrect gi model size\n"); 750 return -EINVAL; 751 } 752 753 q->loss_model = CLG_4_STATES; 754 755 q->clg.state = 1; 756 q->clg.a1 = gi->p13; 757 q->clg.a2 = gi->p31; 758 q->clg.a3 = gi->p32; 759 q->clg.a4 = gi->p14; 760 q->clg.a5 = gi->p23; 761 break; 762 } 763 764 case NETEM_LOSS_GE: { 765 const struct tc_netem_gemodel *ge = nla_data(la); 766 767 if (nla_len(la) < sizeof(struct tc_netem_gemodel)) { 768 pr_info("netem: incorrect ge model size\n"); 769 return -EINVAL; 770 } 771 772 q->loss_model = CLG_GILB_ELL; 773 q->clg.state = 1; 774 q->clg.a1 = ge->p; 775 q->clg.a2 = ge->r; 776 q->clg.a3 = ge->h; 777 q->clg.a4 = ge->k1; 778 break; 779 } 780 781 default: 782 pr_info("netem: unknown loss type %u\n", type); 783 return -EINVAL; 784 } 785 } 786 787 return 0; 788 } 789 790 static const struct nla_policy netem_policy[TCA_NETEM_MAX + 1] = { 791 [TCA_NETEM_CORR] = { .len = sizeof(struct tc_netem_corr) }, 792 [TCA_NETEM_REORDER] = { .len = sizeof(struct tc_netem_reorder) }, 793 [TCA_NETEM_CORRUPT] = { .len = sizeof(struct tc_netem_corrupt) }, 794 [TCA_NETEM_RATE] = { .len = sizeof(struct tc_netem_rate) }, 795 [TCA_NETEM_LOSS] = { .type = NLA_NESTED }, 796 [TCA_NETEM_ECN] = { .type = NLA_U32 }, 797 [TCA_NETEM_RATE64] = { .type = NLA_U64 }, 798 }; 799 800 static int parse_attr(struct nlattr *tb[], int maxtype, struct nlattr *nla, 801 const struct nla_policy *policy, int len) 802 { 803 int nested_len = nla_len(nla) - NLA_ALIGN(len); 804 805 if (nested_len < 0) { 806 pr_info("netem: invalid attributes len %d\n", nested_len); 807 return -EINVAL; 808 } 809 810 if (nested_len >= nla_attr_size(0)) 811 return nla_parse(tb, maxtype, nla_data(nla) + NLA_ALIGN(len), 812 nested_len, policy); 813 814 memset(tb, 0, sizeof(struct nlattr *) * (maxtype + 1)); 815 return 0; 816 } 817 818 /* Parse netlink message to set options */ 819 static int netem_change(struct Qdisc *sch, struct nlattr *opt) 820 { 821 struct netem_sched_data *q = qdisc_priv(sch); 822 struct nlattr *tb[TCA_NETEM_MAX + 1]; 823 struct tc_netem_qopt *qopt; 824 int ret; 825 826 if (opt == NULL) 827 return -EINVAL; 828 829 qopt = nla_data(opt); 830 ret = parse_attr(tb, TCA_NETEM_MAX, opt, netem_policy, sizeof(*qopt)); 831 if (ret < 0) 832 return ret; 833 834 sch->limit = qopt->limit; 835 836 q->latency = qopt->latency; 837 q->jitter = qopt->jitter; 838 q->limit = qopt->limit; 839 q->gap = qopt->gap; 840 q->counter = 0; 841 q->loss = qopt->loss; 842 q->duplicate = qopt->duplicate; 843 844 /* for compatibility with earlier versions. 845 * if gap is set, need to assume 100% probability 846 */ 847 if (q->gap) 848 q->reorder = ~0; 849 850 if (tb[TCA_NETEM_CORR]) 851 get_correlation(sch, tb[TCA_NETEM_CORR]); 852 853 if (tb[TCA_NETEM_DELAY_DIST]) { 854 ret = get_dist_table(sch, tb[TCA_NETEM_DELAY_DIST]); 855 if (ret) 856 return ret; 857 } 858 859 if (tb[TCA_NETEM_REORDER]) 860 get_reorder(sch, tb[TCA_NETEM_REORDER]); 861 862 if (tb[TCA_NETEM_CORRUPT]) 863 get_corrupt(sch, tb[TCA_NETEM_CORRUPT]); 864 865 if (tb[TCA_NETEM_RATE]) 866 get_rate(sch, tb[TCA_NETEM_RATE]); 867 868 if (tb[TCA_NETEM_RATE64]) 869 q->rate = max_t(u64, q->rate, 870 nla_get_u64(tb[TCA_NETEM_RATE64])); 871 872 if (tb[TCA_NETEM_ECN]) 873 q->ecn = nla_get_u32(tb[TCA_NETEM_ECN]); 874 875 q->loss_model = CLG_RANDOM; 876 if (tb[TCA_NETEM_LOSS]) 877 ret = get_loss_clg(sch, tb[TCA_NETEM_LOSS]); 878 879 return ret; 880 } 881 882 static int netem_init(struct Qdisc *sch, struct nlattr *opt) 883 { 884 struct netem_sched_data *q = qdisc_priv(sch); 885 int ret; 886 887 if (!opt) 888 return -EINVAL; 889 890 qdisc_watchdog_init(&q->watchdog, sch); 891 892 q->loss_model = CLG_RANDOM; 893 ret = netem_change(sch, opt); 894 if (ret) 895 pr_info("netem: change failed\n"); 896 return ret; 897 } 898 899 static void netem_destroy(struct Qdisc *sch) 900 { 901 struct netem_sched_data *q = qdisc_priv(sch); 902 903 qdisc_watchdog_cancel(&q->watchdog); 904 if (q->qdisc) 905 qdisc_destroy(q->qdisc); 906 dist_free(q->delay_dist); 907 } 908 909 static int dump_loss_model(const struct netem_sched_data *q, 910 struct sk_buff *skb) 911 { 912 struct nlattr *nest; 913 914 nest = nla_nest_start(skb, TCA_NETEM_LOSS); 915 if (nest == NULL) 916 goto nla_put_failure; 917 918 switch (q->loss_model) { 919 case CLG_RANDOM: 920 /* legacy loss model */ 921 nla_nest_cancel(skb, nest); 922 return 0; /* no data */ 923 924 case CLG_4_STATES: { 925 struct tc_netem_gimodel gi = { 926 .p13 = q->clg.a1, 927 .p31 = q->clg.a2, 928 .p32 = q->clg.a3, 929 .p14 = q->clg.a4, 930 .p23 = q->clg.a5, 931 }; 932 933 if (nla_put(skb, NETEM_LOSS_GI, sizeof(gi), &gi)) 934 goto nla_put_failure; 935 break; 936 } 937 case CLG_GILB_ELL: { 938 struct tc_netem_gemodel ge = { 939 .p = q->clg.a1, 940 .r = q->clg.a2, 941 .h = q->clg.a3, 942 .k1 = q->clg.a4, 943 }; 944 945 if (nla_put(skb, NETEM_LOSS_GE, sizeof(ge), &ge)) 946 goto nla_put_failure; 947 break; 948 } 949 } 950 951 nla_nest_end(skb, nest); 952 return 0; 953 954 nla_put_failure: 955 nla_nest_cancel(skb, nest); 956 return -1; 957 } 958 959 static int netem_dump(struct Qdisc *sch, struct sk_buff *skb) 960 { 961 const struct netem_sched_data *q = qdisc_priv(sch); 962 struct nlattr *nla = (struct nlattr *) skb_tail_pointer(skb); 963 struct tc_netem_qopt qopt; 964 struct tc_netem_corr cor; 965 struct tc_netem_reorder reorder; 966 struct tc_netem_corrupt corrupt; 967 struct tc_netem_rate rate; 968 969 qopt.latency = q->latency; 970 qopt.jitter = q->jitter; 971 qopt.limit = q->limit; 972 qopt.loss = q->loss; 973 qopt.gap = q->gap; 974 qopt.duplicate = q->duplicate; 975 if (nla_put(skb, TCA_OPTIONS, sizeof(qopt), &qopt)) 976 goto nla_put_failure; 977 978 cor.delay_corr = q->delay_cor.rho; 979 cor.loss_corr = q->loss_cor.rho; 980 cor.dup_corr = q->dup_cor.rho; 981 if (nla_put(skb, TCA_NETEM_CORR, sizeof(cor), &cor)) 982 goto nla_put_failure; 983 984 reorder.probability = q->reorder; 985 reorder.correlation = q->reorder_cor.rho; 986 if (nla_put(skb, TCA_NETEM_REORDER, sizeof(reorder), &reorder)) 987 goto nla_put_failure; 988 989 corrupt.probability = q->corrupt; 990 corrupt.correlation = q->corrupt_cor.rho; 991 if (nla_put(skb, TCA_NETEM_CORRUPT, sizeof(corrupt), &corrupt)) 992 goto nla_put_failure; 993 994 if (q->rate >= (1ULL << 32)) { 995 if (nla_put_u64(skb, TCA_NETEM_RATE64, q->rate)) 996 goto nla_put_failure; 997 rate.rate = ~0U; 998 } else { 999 rate.rate = q->rate; 1000 } 1001 rate.packet_overhead = q->packet_overhead; 1002 rate.cell_size = q->cell_size; 1003 rate.cell_overhead = q->cell_overhead; 1004 if (nla_put(skb, TCA_NETEM_RATE, sizeof(rate), &rate)) 1005 goto nla_put_failure; 1006 1007 if (q->ecn && nla_put_u32(skb, TCA_NETEM_ECN, q->ecn)) 1008 goto nla_put_failure; 1009 1010 if (dump_loss_model(q, skb) != 0) 1011 goto nla_put_failure; 1012 1013 return nla_nest_end(skb, nla); 1014 1015 nla_put_failure: 1016 nlmsg_trim(skb, nla); 1017 return -1; 1018 } 1019 1020 static int netem_dump_class(struct Qdisc *sch, unsigned long cl, 1021 struct sk_buff *skb, struct tcmsg *tcm) 1022 { 1023 struct netem_sched_data *q = qdisc_priv(sch); 1024 1025 if (cl != 1 || !q->qdisc) /* only one class */ 1026 return -ENOENT; 1027 1028 tcm->tcm_handle |= TC_H_MIN(1); 1029 tcm->tcm_info = q->qdisc->handle; 1030 1031 return 0; 1032 } 1033 1034 static int netem_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 1035 struct Qdisc **old) 1036 { 1037 struct netem_sched_data *q = qdisc_priv(sch); 1038 1039 sch_tree_lock(sch); 1040 *old = q->qdisc; 1041 q->qdisc = new; 1042 if (*old) { 1043 qdisc_tree_decrease_qlen(*old, (*old)->q.qlen); 1044 qdisc_reset(*old); 1045 } 1046 sch_tree_unlock(sch); 1047 1048 return 0; 1049 } 1050 1051 static struct Qdisc *netem_leaf(struct Qdisc *sch, unsigned long arg) 1052 { 1053 struct netem_sched_data *q = qdisc_priv(sch); 1054 return q->qdisc; 1055 } 1056 1057 static unsigned long netem_get(struct Qdisc *sch, u32 classid) 1058 { 1059 return 1; 1060 } 1061 1062 static void netem_put(struct Qdisc *sch, unsigned long arg) 1063 { 1064 } 1065 1066 static void netem_walk(struct Qdisc *sch, struct qdisc_walker *walker) 1067 { 1068 if (!walker->stop) { 1069 if (walker->count >= walker->skip) 1070 if (walker->fn(sch, 1, walker) < 0) { 1071 walker->stop = 1; 1072 return; 1073 } 1074 walker->count++; 1075 } 1076 } 1077 1078 static const struct Qdisc_class_ops netem_class_ops = { 1079 .graft = netem_graft, 1080 .leaf = netem_leaf, 1081 .get = netem_get, 1082 .put = netem_put, 1083 .walk = netem_walk, 1084 .dump = netem_dump_class, 1085 }; 1086 1087 static struct Qdisc_ops netem_qdisc_ops __read_mostly = { 1088 .id = "netem", 1089 .cl_ops = &netem_class_ops, 1090 .priv_size = sizeof(struct netem_sched_data), 1091 .enqueue = netem_enqueue, 1092 .dequeue = netem_dequeue, 1093 .peek = qdisc_peek_dequeued, 1094 .drop = netem_drop, 1095 .init = netem_init, 1096 .reset = netem_reset, 1097 .destroy = netem_destroy, 1098 .change = netem_change, 1099 .dump = netem_dump, 1100 .owner = THIS_MODULE, 1101 }; 1102 1103 1104 static int __init netem_module_init(void) 1105 { 1106 pr_info("netem: version " VERSION "\n"); 1107 return register_qdisc(&netem_qdisc_ops); 1108 } 1109 static void __exit netem_module_exit(void) 1110 { 1111 unregister_qdisc(&netem_qdisc_ops); 1112 } 1113 module_init(netem_module_init) 1114 module_exit(netem_module_exit) 1115 MODULE_LICENSE("GPL"); 1116