net/sched/sch_choke.c

   1 /*
   2  * net/sched/sch_choke.c        CHOKE scheduler
   3  *
   4  * Copyright (c) 2011 Stephen Hemminger <shemminger@vyatta.com>
   5  * Copyright (c) 2011 Eric Dumazet <eric.dumazet@gmail.com>
   6  *
   7  * This program is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU General Public License
   9  * version 2 as published by the Free Software Foundation.
  10  *
  11  */
  12
  13 #include <linux/module.h>
  14 #include <linux/types.h>
  15 #include <linux/kernel.h>
  16 #include <linux/skbuff.h>
  17 #include <linux/reciprocal_div.h>
  18 #include <linux/vmalloc.h>
  19 #include <net/pkt_sched.h>
  20 #include <net/inet_ecn.h>
  21 #include <net/red.h>
  22 #include <net/flow_keys.h>
  23
  24 /*
  25    CHOKe stateless AQM for fair bandwidth allocation
  26    =================================================
  27
  28    CHOKe (CHOose and Keep for responsive flows, CHOose and Kill for
  29    unresponsive flows) is a variant of RED that penalizes misbehaving flows but
  30    maintains no flow state. The difference from RED is an additional step
  31    during the enqueuing process. If average queue size is over the
  32    low threshold (qmin), a packet is chosen at random from the queue.
  33    If both the new and chosen packet are from the same flow, both
  34    are dropped. Unlike RED, CHOKe is not really a "classful" qdisc because it
  35    needs to access packets in queue randomly. It has a minimal class
  36    interface to allow overriding the builtin flow classifier with
  37    filters.
  38
  39    Source:
  40    R. Pan, B. Prabhakar, and K. Psounis, "CHOKe, A Stateless
  41    Active Queue Management Scheme for Approximating Fair Bandwidth Allocation",
  42    IEEE INFOCOM, 2000.
  43
  44    A. Tang, J. Wang, S. Low, "Understanding CHOKe: Throughput and Spatial
  45    Characteristics", IEEE/ACM Transactions on Networking, 2004
  46
  47  */
  48
  49 /* Upper bound on size of sk_buff table (packets) */
  50 #define CHOKE_MAX_QUEUE (128*1024 - 1)
  51
  52 struct choke_sched_data {
  53 /* Parameters */
  54         u32              limit;
  55         unsigned char    flags;
  56
  57         struct red_parms parms;
  58
  59 /* Variables */
  60         struct red_vars  vars;
  61         struct tcf_proto *filter_list;
  62         struct {
  63                 u32     prob_drop;      /* Early probability drops */
  64                 u32     prob_mark;      /* Early probability marks */
  65                 u32     forced_drop;    /* Forced drops, qavg > max_thresh */
  66                 u32     forced_mark;    /* Forced marks, qavg > max_thresh */
  67                 u32     pdrop;          /* Drops due to queue limits */
  68                 u32     other;          /* Drops due to drop() calls */
  69                 u32     matched;        /* Drops to flow match */
  70         } stats;
  71
  72         unsigned int     head;
  73         unsigned int     tail;
  74
  75         unsigned int     tab_mask; /* size - 1 */
  76
  77         struct sk_buff **tab;
  78 };
  79
  80 /* deliver a random number between 0 and N - 1 */
  81 static u32 random_N(unsigned int N)
  82 {
  83         return reciprocal_divide(random32(), N);
  84 }
  85
  86 /* number of elements in queue including holes */
  87 static unsigned int choke_len(const struct choke_sched_data *q)
  88 {
  89         return (q->tail - q->head) & q->tab_mask;
  90 }
  91
  92 /* Is ECN parameter configured */
  93 static int use_ecn(const struct choke_sched_data *q)
  94 {
  95         return q->flags & TC_RED_ECN;
  96 }
  97
  98 /* Should packets over max just be dropped (versus marked) */
  99 static int use_harddrop(const struct choke_sched_data *q)
 100 {
 101         return q->flags & TC_RED_HARDDROP;
 102 }
 103
 104 /* Move head pointer forward to skip over holes */
 105 static void choke_zap_head_holes(struct choke_sched_data *q)
 106 {
 107         do {
 108                 q->head = (q->head + 1) & q->tab_mask;
 109                 if (q->head == q->tail)
 110                         break;
 111         } while (q->tab[q->head] == NULL);
 112 }
 113
 114 /* Move tail pointer backwards to reuse holes */
 115 static void choke_zap_tail_holes(struct choke_sched_data *q)
 116 {
 117         do {
 118                 q->tail = (q->tail - 1) & q->tab_mask;
 119                 if (q->head == q->tail)
 120                         break;
 121         } while (q->tab[q->tail] == NULL);
 122 }
 123
 124 /* Drop packet from queue array by creating a "hole" */
 125 static void choke_drop_by_idx(struct Qdisc *sch, unsigned int idx)
 126 {
 127         struct choke_sched_data *q = qdisc_priv(sch);
 128         struct sk_buff *skb = q->tab[idx];
 129
 130         q->tab[idx] = NULL;
 131
 132         if (idx == q->head)
 133                 choke_zap_head_holes(q);
 134         if (idx == q->tail)
 135                 choke_zap_tail_holes(q);
 136
 137         sch->qstats.backlog -= qdisc_pkt_len(skb);
 138         qdisc_drop(skb, sch);
 139         qdisc_tree_decrease_qlen(sch, 1);
 140         --sch->q.qlen;
 141 }
 142
 143 struct choke_skb_cb {
 144         u16                     classid;
 145         u8                      keys_valid;
 146         struct flow_keys        keys;
 147 };
 148
 149 static inline struct choke_skb_cb *choke_skb_cb(const struct sk_buff *skb)
 150 {
 151         qdisc_cb_private_validate(skb, sizeof(struct choke_skb_cb));
 152         return (struct choke_skb_cb *)qdisc_skb_cb(skb)->data;
 153 }
 154
 155 static inline void choke_set_classid(struct sk_buff *skb, u16 classid)
 156 {
 157         choke_skb_cb(skb)->classid = classid;
 158 }
 159
 160 static u16 choke_get_classid(const struct sk_buff *skb)
 161 {
 162         return choke_skb_cb(skb)->classid;
 163 }
 164
 165 /*
 166  * Compare flow of two packets
 167  *  Returns true only if source and destination address and port match.
 168  *          false for special cases
 169  */
 170 static bool choke_match_flow(struct sk_buff *skb1,
 171                              struct sk_buff *skb2)
 172 {
 173         if (skb1->protocol != skb2->protocol)
 174                 return false;
 175
 176         if (!choke_skb_cb(skb1)->keys_valid) {
 177                 choke_skb_cb(skb1)->keys_valid = 1;
 178                 skb_flow_dissect(skb1, &choke_skb_cb(skb1)->keys);
 179         }
 180
 181         if (!choke_skb_cb(skb2)->keys_valid) {
 182                 choke_skb_cb(skb2)->keys_valid = 1;
 183                 skb_flow_dissect(skb2, &choke_skb_cb(skb2)->keys);
 184         }
 185
 186         return !memcmp(&choke_skb_cb(skb1)->keys,
 187                        &choke_skb_cb(skb2)->keys,
 188                        sizeof(struct flow_keys));
 189 }
 190
 191 /*
 192  * Classify flow using either:
 193  *  1. pre-existing classification result in skb
 194  *  2. fast internal classification
 195  *  3. use TC filter based classification
 196  */
 197 static bool choke_classify(struct sk_buff *skb,
 198                            struct Qdisc *sch, int *qerr)
 199
 200 {
 201         struct choke_sched_data *q = qdisc_priv(sch);
 202         struct tcf_result res;
 203         int result;
 204
 205         result = tc_classify(skb, q->filter_list, &res);
 206         if (result >= 0) {
 207 #ifdef CONFIG_NET_CLS_ACT
 208                 switch (result) {
 209                 case TC_ACT_STOLEN:
 210                 case TC_ACT_QUEUED:
 211                         *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 212                 case TC_ACT_SHOT:
 213                         return false;
 214                 }
 215 #endif
 216                 choke_set_classid(skb, TC_H_MIN(res.classid));
 217                 return true;
 218         }
 219
 220         return false;
 221 }
 222
 223 /*
 224  * Select a packet at random from queue
 225  * HACK: since queue can have holes from previous deletion; retry several
 226  *   times to find a random skb but then just give up and return the head
 227  * Will return NULL if queue is empty (q->head == q->tail)
 228  */
 229 static struct sk_buff *choke_peek_random(const struct choke_sched_data *q,
 230                                          unsigned int *pidx)
 231 {
 232         struct sk_buff *skb;
 233         int retrys = 3;
 234
 235         do {
 236                 *pidx = (q->head + random_N(choke_len(q))) & q->tab_mask;
 237                 skb = q->tab[*pidx];
 238                 if (skb)
 239                         return skb;
 240         } while (--retrys > 0);
 241
 242         return q->tab[*pidx = q->head];
 243 }
 244
 245 /*
 246  * Compare new packet with random packet in queue
 247  * returns true if matched and sets *pidx
 248  */
 249 static bool choke_match_random(const struct choke_sched_data *q,
 250                                struct sk_buff *nskb,
 251                                unsigned int *pidx)
 252 {
 253         struct sk_buff *oskb;
 254
 255         if (q->head == q->tail)
 256                 return false;
 257
 258         oskb = choke_peek_random(q, pidx);
 259         if (q->filter_list)
 260                 return choke_get_classid(nskb) == choke_get_classid(oskb);
 261
 262         return choke_match_flow(oskb, nskb);
 263 }
 264
 265 static int choke_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 266 {
 267         struct choke_sched_data *q = qdisc_priv(sch);
 268         const struct red_parms *p = &q->parms;
 269         int ret = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
 270
 271         if (q->filter_list) {
 272                 /* If using external classifiers, get result and record it. */
 273                 if (!choke_classify(skb, sch, &ret))
 274                         goto other_drop;        /* Packet was eaten by filter */
 275         }
 276
 277         choke_skb_cb(skb)->keys_valid = 0;
 278         /* Compute average queue usage (see RED) */
 279         q->vars.qavg = red_calc_qavg(p, &q->vars, sch->q.qlen);
 280         if (red_is_idling(&q->vars))
 281                 red_end_of_idle_period(&q->vars);
 282
 283         /* Is queue small? */
 284         if (q->vars.qavg <= p->qth_min)
 285                 q->vars.qcount = -1;
 286         else {
 287                 unsigned int idx;
 288
 289                 /* Draw a packet at random from queue and compare flow */
 290                 if (choke_match_random(q, skb, &idx)) {
 291                         q->stats.matched++;
 292                         choke_drop_by_idx(sch, idx);
 293                         goto congestion_drop;
 294                 }
 295
 296                 /* Queue is large, always mark/drop */
 297                 if (q->vars.qavg > p->qth_max) {
 298                         q->vars.qcount = -1;
 299
 300                         sch->qstats.overlimits++;
 301                         if (use_harddrop(q) || !use_ecn(q) ||
 302                             !INET_ECN_set_ce(skb)) {
 303                                 q->stats.forced_drop++;
 304                                 goto congestion_drop;
 305                         }
 306
 307                         q->stats.forced_mark++;
 308                 } else if (++q->vars.qcount) {
 309                         if (red_mark_probability(p, &q->vars, q->vars.qavg)) {
 310                                 q->vars.qcount = 0;
 311                                 q->vars.qR = red_random(p);
 312
 313                                 sch->qstats.overlimits++;
 314                                 if (!use_ecn(q) || !INET_ECN_set_ce(skb)) {
 315                                         q->stats.prob_drop++;
 316                                         goto congestion_drop;
 317                                 }
 318
 319                                 q->stats.prob_mark++;
 320                         }
 321                 } else
 322                         q->vars.qR = red_random(p);
 323         }
 324
 325         /* Admit new packet */
 326         if (sch->q.qlen < q->limit) {
 327                 q->tab[q->tail] = skb;
 328                 q->tail = (q->tail + 1) & q->tab_mask;
 329                 ++sch->q.qlen;
 330                 sch->qstats.backlog += qdisc_pkt_len(skb);
 331                 return NET_XMIT_SUCCESS;
 332         }
 333
 334         q->stats.pdrop++;
 335         sch->qstats.drops++;
 336         kfree_skb(skb);
 337         return NET_XMIT_DROP;
 338
 339  congestion_drop:
 340         qdisc_drop(skb, sch);
 341         return NET_XMIT_CN;
 342
 343  other_drop:
 344         if (ret & __NET_XMIT_BYPASS)
 345                 sch->qstats.drops++;
 346         kfree_skb(skb);
 347         return ret;
 348 }
 349
 350 static struct sk_buff *choke_dequeue(struct Qdisc *sch)
 351 {
 352         struct choke_sched_data *q = qdisc_priv(sch);
 353         struct sk_buff *skb;
 354
 355         if (q->head == q->tail) {
 356                 if (!red_is_idling(&q->vars))
 357                         red_start_of_idle_period(&q->vars);
 358                 return NULL;
 359         }
 360
 361         skb = q->tab[q->head];
 362         q->tab[q->head] = NULL;
 363         choke_zap_head_holes(q);
 364         --sch->q.qlen;
 365         sch->qstats.backlog -= qdisc_pkt_len(skb);
 366         qdisc_bstats_update(sch, skb);
 367
 368         return skb;
 369 }
 370
 371 static unsigned int choke_drop(struct Qdisc *sch)
 372 {
 373         struct choke_sched_data *q = qdisc_priv(sch);
 374         unsigned int len;
 375
 376         len = qdisc_queue_drop(sch);
 377         if (len > 0)
 378                 q->stats.other++;
 379         else {
 380                 if (!red_is_idling(&q->vars))
 381                         red_start_of_idle_period(&q->vars);
 382         }
 383
 384         return len;
 385 }
 386
 387 static void choke_reset(struct Qdisc *sch)
 388 {
 389         struct choke_sched_data *q = qdisc_priv(sch);
 390
 391         red_restart(&q->vars);
 392 }
 393
 394 static const struct nla_policy choke_policy[TCA_CHOKE_MAX + 1] = {
 395         [TCA_CHOKE_PARMS]       = { .len = sizeof(struct tc_red_qopt) },
 396         [TCA_CHOKE_STAB]        = { .len = RED_STAB_SIZE },
 397         [TCA_CHOKE_MAX_P]       = { .type = NLA_U32 },
 398 };
 399
 400
 401 static void choke_free(void *addr)
 402 {
 403         if (addr) {
 404                 if (is_vmalloc_addr(addr))
 405                         vfree(addr);
 406                 else
 407                         kfree(addr);
 408         }
 409 }
 410
 411 static int choke_change(struct Qdisc *sch, struct nlattr *opt)
 412 {
 413         struct choke_sched_data *q = qdisc_priv(sch);
 414         struct nlattr *tb[TCA_CHOKE_MAX + 1];
 415         const struct tc_red_qopt *ctl;
 416         int err;
 417         struct sk_buff **old = NULL;
 418         unsigned int mask;
 419         u32 max_P;
 420
 421         if (opt == NULL)
 422                 return -EINVAL;
 423
 424         err = nla_parse_nested(tb, TCA_CHOKE_MAX, opt, choke_policy);
 425         if (err < 0)
 426                 return err;
 427
 428         if (tb[TCA_CHOKE_PARMS] == NULL ||
 429             tb[TCA_CHOKE_STAB] == NULL)
 430                 return -EINVAL;
 431
 432         max_P = tb[TCA_CHOKE_MAX_P] ? nla_get_u32(tb[TCA_CHOKE_MAX_P]) : 0;
 433
 434         ctl = nla_data(tb[TCA_CHOKE_PARMS]);
 435
 436         if (ctl->limit > CHOKE_MAX_QUEUE)
 437                 return -EINVAL;
 438
 439         mask = roundup_pow_of_two(ctl->limit + 1) - 1;
 440         if (mask != q->tab_mask) {
 441                 struct sk_buff **ntab;
 442
 443                 ntab = kcalloc(mask + 1, sizeof(struct sk_buff *), GFP_KERNEL);
 444                 if (!ntab)
 445                         ntab = vzalloc((mask + 1) * sizeof(struct sk_buff *));
 446                 if (!ntab)
 447                         return -ENOMEM;
 448
 449                 sch_tree_lock(sch);
 450                 old = q->tab;
 451                 if (old) {
 452                         unsigned int oqlen = sch->q.qlen, tail = 0;
 453
 454                         while (q->head != q->tail) {
 455                                 struct sk_buff *skb = q->tab[q->head];
 456
 457                                 q->head = (q->head + 1) & q->tab_mask;
 458                                 if (!skb)
 459                                         continue;
 460                                 if (tail < mask) {
 461                                         ntab[tail++] = skb;
 462                                         continue;
 463                                 }
 464                                 sch->qstats.backlog -= qdisc_pkt_len(skb);
 465                                 --sch->q.qlen;
 466                                 qdisc_drop(skb, sch);
 467                         }
 468                         qdisc_tree_decrease_qlen(sch, oqlen - sch->q.qlen);
 469                         q->head = 0;
 470                         q->tail = tail;
 471                 }
 472
 473                 q->tab_mask = mask;
 474                 q->tab = ntab;
 475         } else
 476                 sch_tree_lock(sch);
 477
 478         q->flags = ctl->flags;
 479         q->limit = ctl->limit;
 480
 481         red_set_parms(&q->parms, ctl->qth_min, ctl->qth_max, ctl->Wlog,
 482                       ctl->Plog, ctl->Scell_log,
 483                       nla_data(tb[TCA_CHOKE_STAB]),
 484                       max_P);
 485         red_set_vars(&q->vars);
 486
 487         if (q->head == q->tail)
 488                 red_end_of_idle_period(&q->vars);
 489
 490         sch_tree_unlock(sch);
 491         choke_free(old);
 492         return 0;
 493 }
 494
 495 static int choke_init(struct Qdisc *sch, struct nlattr *opt)
 496 {
 497         return choke_change(sch, opt);
 498 }
 499
 500 static int choke_dump(struct Qdisc *sch, struct sk_buff *skb)
 501 {
 502         struct choke_sched_data *q = qdisc_priv(sch);
 503         struct nlattr *opts = NULL;
 504         struct tc_red_qopt opt = {
 505                 .limit          = q->limit,
 506                 .flags          = q->flags,
 507                 .qth_min        = q->parms.qth_min >> q->parms.Wlog,
 508                 .qth_max        = q->parms.qth_max >> q->parms.Wlog,
 509                 .Wlog           = q->parms.Wlog,
 510                 .Plog           = q->parms.Plog,
 511                 .Scell_log      = q->parms.Scell_log,
 512         };
 513
 514         opts = nla_nest_start(skb, TCA_OPTIONS);
 515         if (opts == NULL)
 516                 goto nla_put_failure;
 517
 518         NLA_PUT(skb, TCA_CHOKE_PARMS, sizeof(opt), &opt);
 519         NLA_PUT_U32(skb, TCA_CHOKE_MAX_P, q->parms.max_P);
 520         return nla_nest_end(skb, opts);
 521
 522 nla_put_failure:
 523         nla_nest_cancel(skb, opts);
 524         return -EMSGSIZE;
 525 }
 526
 527 static int choke_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 528 {
 529         struct choke_sched_data *q = qdisc_priv(sch);
 530         struct tc_choke_xstats st = {
 531                 .early  = q->stats.prob_drop + q->stats.forced_drop,
 532                 .marked = q->stats.prob_mark + q->stats.forced_mark,
 533                 .pdrop  = q->stats.pdrop,
 534                 .other  = q->stats.other,
 535                 .matched = q->stats.matched,
 536         };
 537
 538         return gnet_stats_copy_app(d, &st, sizeof(st));
 539 }
 540
 541 static void choke_destroy(struct Qdisc *sch)
 542 {
 543         struct choke_sched_data *q = qdisc_priv(sch);
 544
 545         tcf_destroy_chain(&q->filter_list);
 546         choke_free(q->tab);
 547 }
 548
 549 static struct Qdisc *choke_leaf(struct Qdisc *sch, unsigned long arg)
 550 {
 551         return NULL;
 552 }
 553
 554 static unsigned long choke_get(struct Qdisc *sch, u32 classid)
 555 {
 556         return 0;
 557 }
 558
 559 static void choke_put(struct Qdisc *q, unsigned long cl)
 560 {
 561 }
 562
 563 static unsigned long choke_bind(struct Qdisc *sch, unsigned long parent,
 564                                 u32 classid)
 565 {
 566         return 0;
 567 }
 568
 569 static struct tcf_proto **choke_find_tcf(struct Qdisc *sch, unsigned long cl)
 570 {
 571         struct choke_sched_data *q = qdisc_priv(sch);
 572
 573         if (cl)
 574                 return NULL;
 575         return &q->filter_list;
 576 }
 577
 578 static int choke_dump_class(struct Qdisc *sch, unsigned long cl,
 579                           struct sk_buff *skb, struct tcmsg *tcm)
 580 {
 581         tcm->tcm_handle |= TC_H_MIN(cl);
 582         return 0;
 583 }
 584
 585 static void choke_walk(struct Qdisc *sch, struct qdisc_walker *arg)
 586 {
 587         if (!arg->stop) {
 588                 if (arg->fn(sch, 1, arg) < 0) {
 589                         arg->stop = 1;
 590                         return;
 591                 }
 592                 arg->count++;
 593         }
 594 }
 595
 596 static const struct Qdisc_class_ops choke_class_ops = {
 597         .leaf           =       choke_leaf,
 598         .get            =       choke_get,
 599         .put            =       choke_put,
 600         .tcf_chain      =       choke_find_tcf,
 601         .bind_tcf       =       choke_bind,
 602         .unbind_tcf     =       choke_put,
 603         .dump           =       choke_dump_class,
 604         .walk           =       choke_walk,
 605 };
 606
 607 static struct sk_buff *choke_peek_head(struct Qdisc *sch)
 608 {
 609         struct choke_sched_data *q = qdisc_priv(sch);
 610
 611         return (q->head != q->tail) ? q->tab[q->head] : NULL;
 612 }
 613
 614 static struct Qdisc_ops choke_qdisc_ops __read_mostly = {
 615         .id             =       "choke",
 616         .priv_size      =       sizeof(struct choke_sched_data),
 617
 618         .enqueue        =       choke_enqueue,
 619         .dequeue        =       choke_dequeue,
 620         .peek           =       choke_peek_head,
 621         .drop           =       choke_drop,
 622         .init           =       choke_init,
 623         .destroy        =       choke_destroy,
 624         .reset          =       choke_reset,
 625         .change         =       choke_change,
 626         .dump           =       choke_dump,
 627         .dump_stats     =       choke_dump_stats,
 628         .owner          =       THIS_MODULE,
 629 };
 630
 631 static int __init choke_module_init(void)
 632 {
 633         return register_qdisc(&choke_qdisc_ops);
 634 }
 635
 636 static void __exit choke_module_exit(void)
 637 {
 638         unregister_qdisc(&choke_qdisc_ops);
 639 }
 640
 641 module_init(choke_module_init)
 642 module_exit(choke_module_exit)
 643
 644 MODULE_LICENSE("GPL");