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/vmalloc.h>
  18 #include <net/pkt_sched.h>
  19 #include <net/inet_ecn.h>
  20 #include <net/red.h>
  21 #include <net/flow_dissector.h>
  22
  23 /*
  24    CHOKe stateless AQM for fair bandwidth allocation
  25    =================================================
  26
  27    CHOKe (CHOose and Keep for responsive flows, CHOose and Kill for
  28    unresponsive flows) is a variant of RED that penalizes misbehaving flows but
  29    maintains no flow state. The difference from RED is an additional step
  30    during the enqueuing process. If average queue size is over the
  31    low threshold (qmin), a packet is chosen at random from the queue.
  32    If both the new and chosen packet are from the same flow, both
  33    are dropped. Unlike RED, CHOKe is not really a "classful" qdisc because it
  34    needs to access packets in queue randomly. It has a minimal class
  35    interface to allow overriding the builtin flow classifier with
  36    filters.
  37
  38    Source:
  39    R. Pan, B. Prabhakar, and K. Psounis, "CHOKe, A Stateless
  40    Active Queue Management Scheme for Approximating Fair Bandwidth Allocation",
  41    IEEE INFOCOM, 2000.
  42
  43    A. Tang, J. Wang, S. Low, "Understanding CHOKe: Throughput and Spatial
  44    Characteristics", IEEE/ACM Transactions on Networking, 2004
  45
  46  */
  47
  48 /* Upper bound on size of sk_buff table (packets) */
  49 #define CHOKE_MAX_QUEUE (128*1024 - 1)
  50
  51 struct choke_sched_data {
  52 /* Parameters */
  53         u32              limit;
  54         unsigned char    flags;
  55
  56         struct red_parms parms;
  57
  58 /* Variables */
  59         struct red_vars  vars;
  60         struct tcf_proto __rcu *filter_list;
  61         struct {
  62                 u32     prob_drop;      /* Early probability drops */
  63                 u32     prob_mark;      /* Early probability marks */
  64                 u32     forced_drop;    /* Forced drops, qavg > max_thresh */
  65                 u32     forced_mark;    /* Forced marks, qavg > max_thresh */
  66                 u32     pdrop;          /* Drops due to queue limits */
  67                 u32     other;          /* Drops due to drop() calls */
  68                 u32     matched;        /* Drops to flow match */
  69         } stats;
  70
  71         unsigned int     head;
  72         unsigned int     tail;
  73
  74         unsigned int     tab_mask; /* size - 1 */
  75
  76         struct sk_buff **tab;
  77 };
  78
  79 /* number of elements in queue including holes */
  80 static unsigned int choke_len(const struct choke_sched_data *q)
  81 {
  82         return (q->tail - q->head) & q->tab_mask;
  83 }
  84
  85 /* Is ECN parameter configured */
  86 static int use_ecn(const struct choke_sched_data *q)
  87 {
  88         return q->flags & TC_RED_ECN;
  89 }
  90
  91 /* Should packets over max just be dropped (versus marked) */
  92 static int use_harddrop(const struct choke_sched_data *q)
  93 {
  94         return q->flags & TC_RED_HARDDROP;
  95 }
  96
  97 /* Move head pointer forward to skip over holes */
  98 static void choke_zap_head_holes(struct choke_sched_data *q)
  99 {
 100         do {
 101                 q->head = (q->head + 1) & q->tab_mask;
 102                 if (q->head == q->tail)
 103                         break;
 104         } while (q->tab[q->head] == NULL);
 105 }
 106
 107 /* Move tail pointer backwards to reuse holes */
 108 static void choke_zap_tail_holes(struct choke_sched_data *q)
 109 {
 110         do {
 111                 q->tail = (q->tail - 1) & q->tab_mask;
 112                 if (q->head == q->tail)
 113                         break;
 114         } while (q->tab[q->tail] == NULL);
 115 }
 116
 117 /* Drop packet from queue array by creating a "hole" */
 118 static void choke_drop_by_idx(struct Qdisc *sch, unsigned int idx,
 119                               struct sk_buff **to_free)
 120 {
 121         struct choke_sched_data *q = qdisc_priv(sch);
 122         struct sk_buff *skb = q->tab[idx];
 123
 124         q->tab[idx] = NULL;
 125
 126         if (idx == q->head)
 127                 choke_zap_head_holes(q);
 128         if (idx == q->tail)
 129                 choke_zap_tail_holes(q);
 130
 131         qdisc_qstats_backlog_dec(sch, skb);
 132         qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
 133         qdisc_drop(skb, sch, to_free);
 134         --sch->q.qlen;
 135 }
 136
 137 struct choke_skb_cb {
 138         u16                     classid;
 139         u8                      keys_valid;
 140         struct                  flow_keys_digest keys;
 141 };
 142
 143 static inline struct choke_skb_cb *choke_skb_cb(const struct sk_buff *skb)
 144 {
 145         qdisc_cb_private_validate(skb, sizeof(struct choke_skb_cb));
 146         return (struct choke_skb_cb *)qdisc_skb_cb(skb)->data;
 147 }
 148
 149 static inline void choke_set_classid(struct sk_buff *skb, u16 classid)
 150 {
 151         choke_skb_cb(skb)->classid = classid;
 152 }
 153
 154 static u16 choke_get_classid(const struct sk_buff *skb)
 155 {
 156         return choke_skb_cb(skb)->classid;
 157 }
 158
 159 /*
 160  * Compare flow of two packets
 161  *  Returns true only if source and destination address and port match.
 162  *          false for special cases
 163  */
 164 static bool choke_match_flow(struct sk_buff *skb1,
 165                              struct sk_buff *skb2)
 166 {
 167         struct flow_keys temp;
 168
 169         if (skb1->protocol != skb2->protocol)
 170                 return false;
 171
 172         if (!choke_skb_cb(skb1)->keys_valid) {
 173                 choke_skb_cb(skb1)->keys_valid = 1;
 174                 skb_flow_dissect_flow_keys(skb1, &temp, 0);
 175                 make_flow_keys_digest(&choke_skb_cb(skb1)->keys, &temp);
 176         }
 177
 178         if (!choke_skb_cb(skb2)->keys_valid) {
 179                 choke_skb_cb(skb2)->keys_valid = 1;
 180                 skb_flow_dissect_flow_keys(skb2, &temp, 0);
 181                 make_flow_keys_digest(&choke_skb_cb(skb2)->keys, &temp);
 182         }
 183
 184         return !memcmp(&choke_skb_cb(skb1)->keys,
 185                        &choke_skb_cb(skb2)->keys,
 186                        sizeof(choke_skb_cb(skb1)->keys));
 187 }
 188
 189 /*
 190  * Classify flow using either:
 191  *  1. pre-existing classification result in skb
 192  *  2. fast internal classification
 193  *  3. use TC filter based classification
 194  */
 195 static bool choke_classify(struct sk_buff *skb,
 196                            struct Qdisc *sch, int *qerr)
 197
 198 {
 199         struct choke_sched_data *q = qdisc_priv(sch);
 200         struct tcf_result res;
 201         struct tcf_proto *fl;
 202         int result;
 203
 204         fl = rcu_dereference_bh(q->filter_list);
 205         result = tc_classify(skb, fl, &res, false);
 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 + prandom_u32_max(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 (rcu_access_pointer(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                          struct sk_buff **to_free)
 267 {
 268         int ret = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
 269         struct choke_sched_data *q = qdisc_priv(sch);
 270         const struct red_parms *p = &q->parms;
 271
 272         if (rcu_access_pointer(q->filter_list)) {
 273                 /* If using external classifiers, get result and record it. */
 274                 if (!choke_classify(skb, sch, &ret))
 275                         goto other_drop;        /* Packet was eaten by filter */
 276         }
 277
 278         choke_skb_cb(skb)->keys_valid = 0;
 279         /* Compute average queue usage (see RED) */
 280         q->vars.qavg = red_calc_qavg(p, &q->vars, sch->q.qlen);
 281         if (red_is_idling(&q->vars))
 282                 red_end_of_idle_period(&q->vars);
 283
 284         /* Is queue small? */
 285         if (q->vars.qavg <= p->qth_min)
 286                 q->vars.qcount = -1;
 287         else {
 288                 unsigned int idx;
 289
 290                 /* Draw a packet at random from queue and compare flow */
 291                 if (choke_match_random(q, skb, &idx)) {
 292                         q->stats.matched++;
 293                         choke_drop_by_idx(sch, idx, to_free);
 294                         goto congestion_drop;
 295                 }
 296
 297                 /* Queue is large, always mark/drop */
 298                 if (q->vars.qavg > p->qth_max) {
 299                         q->vars.qcount = -1;
 300
 301                         qdisc_qstats_overlimit(sch);
 302                         if (use_harddrop(q) || !use_ecn(q) ||
 303                             !INET_ECN_set_ce(skb)) {
 304                                 q->stats.forced_drop++;
 305                                 goto congestion_drop;
 306                         }
 307
 308                         q->stats.forced_mark++;
 309                 } else if (++q->vars.qcount) {
 310                         if (red_mark_probability(p, &q->vars, q->vars.qavg)) {
 311                                 q->vars.qcount = 0;
 312                                 q->vars.qR = red_random(p);
 313
 314                                 qdisc_qstats_overlimit(sch);
 315                                 if (!use_ecn(q) || !INET_ECN_set_ce(skb)) {
 316                                         q->stats.prob_drop++;
 317                                         goto congestion_drop;
 318                                 }
 319
 320                                 q->stats.prob_mark++;
 321                         }
 322                 } else
 323                         q->vars.qR = red_random(p);
 324         }
 325
 326         /* Admit new packet */
 327         if (sch->q.qlen < q->limit) {
 328                 q->tab[q->tail] = skb;
 329                 q->tail = (q->tail + 1) & q->tab_mask;
 330                 ++sch->q.qlen;
 331                 qdisc_qstats_backlog_inc(sch, skb);
 332                 return NET_XMIT_SUCCESS;
 333         }
 334
 335         q->stats.pdrop++;
 336         return qdisc_drop(skb, sch, to_free);
 337
 338 congestion_drop:
 339         qdisc_drop(skb, sch, to_free);
 340         return NET_XMIT_CN;
 341
 342 other_drop:
 343         if (ret & __NET_XMIT_BYPASS)
 344                 qdisc_qstats_drop(sch);
 345         __qdisc_drop(skb, to_free);
 346         return ret;
 347 }
 348
 349 static struct sk_buff *choke_dequeue(struct Qdisc *sch)
 350 {
 351         struct choke_sched_data *q = qdisc_priv(sch);
 352         struct sk_buff *skb;
 353
 354         if (q->head == q->tail) {
 355                 if (!red_is_idling(&q->vars))
 356                         red_start_of_idle_period(&q->vars);
 357                 return NULL;
 358         }
 359
 360         skb = q->tab[q->head];
 361         q->tab[q->head] = NULL;
 362         choke_zap_head_holes(q);
 363         --sch->q.qlen;
 364         qdisc_qstats_backlog_dec(sch, skb);
 365         qdisc_bstats_update(sch, skb);
 366
 367         return skb;
 368 }
 369
 370 static void choke_reset(struct Qdisc *sch)
 371 {
 372         struct choke_sched_data *q = qdisc_priv(sch);
 373
 374         while (q->head != q->tail) {
 375                 struct sk_buff *skb = q->tab[q->head];
 376
 377                 q->head = (q->head + 1) & q->tab_mask;
 378                 if (!skb)
 379                         continue;
 380                 rtnl_qdisc_drop(skb, sch);
 381         }
 382
 383         sch->q.qlen = 0;
 384         sch->qstats.backlog = 0;
 385         if (q->tab)
 386                 memset(q->tab, 0, (q->tab_mask + 1) * sizeof(struct sk_buff *));
 387         q->head = q->tail = 0;
 388         red_restart(&q->vars);
 389 }
 390
 391 static const struct nla_policy choke_policy[TCA_CHOKE_MAX + 1] = {
 392         [TCA_CHOKE_PARMS]       = { .len = sizeof(struct tc_red_qopt) },
 393         [TCA_CHOKE_STAB]        = { .len = RED_STAB_SIZE },
 394         [TCA_CHOKE_MAX_P]       = { .type = NLA_U32 },
 395 };
 396
 397
 398 static void choke_free(void *addr)
 399 {
 400         kvfree(addr);
 401 }
 402
 403 static int choke_change(struct Qdisc *sch, struct nlattr *opt)
 404 {
 405         struct choke_sched_data *q = qdisc_priv(sch);
 406         struct nlattr *tb[TCA_CHOKE_MAX + 1];
 407         const struct tc_red_qopt *ctl;
 408         int err;
 409         struct sk_buff **old = NULL;
 410         unsigned int mask;
 411         u32 max_P;
 412
 413         if (opt == NULL)
 414                 return -EINVAL;
 415
 416         err = nla_parse_nested(tb, TCA_CHOKE_MAX, opt, choke_policy);
 417         if (err < 0)
 418                 return err;
 419
 420         if (tb[TCA_CHOKE_PARMS] == NULL ||
 421             tb[TCA_CHOKE_STAB] == NULL)
 422                 return -EINVAL;
 423
 424         max_P = tb[TCA_CHOKE_MAX_P] ? nla_get_u32(tb[TCA_CHOKE_MAX_P]) : 0;
 425
 426         ctl = nla_data(tb[TCA_CHOKE_PARMS]);
 427
 428         if (!red_check_params(ctl->qth_min, ctl->qth_max, ctl->Wlog))
 429                 return -EINVAL;
 430
 431         if (ctl->limit > CHOKE_MAX_QUEUE)
 432                 return -EINVAL;
 433
 434         mask = roundup_pow_of_two(ctl->limit + 1) - 1;
 435         if (mask != q->tab_mask) {
 436                 struct sk_buff **ntab;
 437
 438                 ntab = kcalloc(mask + 1, sizeof(struct sk_buff *),
 439                                GFP_KERNEL | __GFP_NOWARN);
 440                 if (!ntab)
 441                         ntab = vzalloc((mask + 1) * sizeof(struct sk_buff *));
 442                 if (!ntab)
 443                         return -ENOMEM;
 444
 445                 sch_tree_lock(sch);
 446                 old = q->tab;
 447                 if (old) {
 448                         unsigned int oqlen = sch->q.qlen, tail = 0;
 449                         unsigned dropped = 0;
 450
 451                         while (q->head != q->tail) {
 452                                 struct sk_buff *skb = q->tab[q->head];
 453
 454                                 q->head = (q->head + 1) & q->tab_mask;
 455                                 if (!skb)
 456                                         continue;
 457                                 if (tail < mask) {
 458                                         ntab[tail++] = skb;
 459                                         continue;
 460                                 }
 461                                 dropped += qdisc_pkt_len(skb);
 462                                 qdisc_qstats_backlog_dec(sch, skb);
 463                                 --sch->q.qlen;
 464                                 rtnl_qdisc_drop(skb, sch);
 465                         }
 466                         qdisc_tree_reduce_backlog(sch, oqlen - sch->q.qlen, dropped);
 467                         q->head = 0;
 468                         q->tail = tail;
 469                 }
 470
 471                 q->tab_mask = mask;
 472                 q->tab = ntab;
 473         } else
 474                 sch_tree_lock(sch);
 475
 476         q->flags = ctl->flags;
 477         q->limit = ctl->limit;
 478
 479         red_set_parms(&q->parms, ctl->qth_min, ctl->qth_max, ctl->Wlog,
 480                       ctl->Plog, ctl->Scell_log,
 481                       nla_data(tb[TCA_CHOKE_STAB]),
 482                       max_P);
 483         red_set_vars(&q->vars);
 484
 485         if (q->head == q->tail)
 486                 red_end_of_idle_period(&q->vars);
 487
 488         sch_tree_unlock(sch);
 489         choke_free(old);
 490         return 0;
 491 }
 492
 493 static int choke_init(struct Qdisc *sch, struct nlattr *opt)
 494 {
 495         return choke_change(sch, opt);
 496 }
 497
 498 static int choke_dump(struct Qdisc *sch, struct sk_buff *skb)
 499 {
 500         struct choke_sched_data *q = qdisc_priv(sch);
 501         struct nlattr *opts = NULL;
 502         struct tc_red_qopt opt = {
 503                 .limit          = q->limit,
 504                 .flags          = q->flags,
 505                 .qth_min        = q->parms.qth_min >> q->parms.Wlog,
 506                 .qth_max        = q->parms.qth_max >> q->parms.Wlog,
 507                 .Wlog           = q->parms.Wlog,
 508                 .Plog           = q->parms.Plog,
 509                 .Scell_log      = q->parms.Scell_log,
 510         };
 511
 512         opts = nla_nest_start(skb, TCA_OPTIONS);
 513         if (opts == NULL)
 514                 goto nla_put_failure;
 515
 516         if (nla_put(skb, TCA_CHOKE_PARMS, sizeof(opt), &opt) ||
 517             nla_put_u32(skb, TCA_CHOKE_MAX_P, q->parms.max_P))
 518                 goto nla_put_failure;
 519         return nla_nest_end(skb, opts);
 520
 521 nla_put_failure:
 522         nla_nest_cancel(skb, opts);
 523         return -EMSGSIZE;
 524 }
 525
 526 static int choke_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 527 {
 528         struct choke_sched_data *q = qdisc_priv(sch);
 529         struct tc_choke_xstats st = {
 530                 .early  = q->stats.prob_drop + q->stats.forced_drop,
 531                 .marked = q->stats.prob_mark + q->stats.forced_mark,
 532                 .pdrop  = q->stats.pdrop,
 533                 .other  = q->stats.other,
 534                 .matched = q->stats.matched,
 535         };
 536
 537         return gnet_stats_copy_app(d, &st, sizeof(st));
 538 }
 539
 540 static void choke_destroy(struct Qdisc *sch)
 541 {
 542         struct choke_sched_data *q = qdisc_priv(sch);
 543
 544         tcf_destroy_chain(&q->filter_list);
 545         choke_free(q->tab);
 546 }
 547
 548 static struct sk_buff *choke_peek_head(struct Qdisc *sch)
 549 {
 550         struct choke_sched_data *q = qdisc_priv(sch);
 551
 552         return (q->head != q->tail) ? q->tab[q->head] : NULL;
 553 }
 554
 555 static struct Qdisc_ops choke_qdisc_ops __read_mostly = {
 556         .id             =       "choke",
 557         .priv_size      =       sizeof(struct choke_sched_data),
 558
 559         .enqueue        =       choke_enqueue,
 560         .dequeue        =       choke_dequeue,
 561         .peek           =       choke_peek_head,
 562         .init           =       choke_init,
 563         .destroy        =       choke_destroy,
 564         .reset          =       choke_reset,
 565         .change         =       choke_change,
 566         .dump           =       choke_dump,
 567         .dump_stats     =       choke_dump_stats,
 568         .owner          =       THIS_MODULE,
 569 };
 570
 571 static int __init choke_module_init(void)
 572 {
 573         return register_qdisc(&choke_qdisc_ops);
 574 }
 575
 576 static void __exit choke_module_exit(void)
 577 {
 578         unregister_qdisc(&choke_qdisc_ops);
 579 }
 580
 581 module_init(choke_module_init)
 582 module_exit(choke_module_exit)
 583
 584 MODULE_LICENSE("GPL");