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/pkt_cls.h>
  20 #include <net/inet_ecn.h>
  21 #include <net/red.h>
  22 #include <net/flow_dissector.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 {
  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 /*
 155  * Compare flow of two packets
 156  *  Returns true only if source and destination address and port match.
 157  *          false for special cases
 158  */
 159 static bool choke_match_flow(struct sk_buff *skb1,
 160                              struct sk_buff *skb2)
 161 {
 162         struct flow_keys temp;
 163
 164         if (skb1->protocol != skb2->protocol)
 165                 return false;
 166
 167         if (!choke_skb_cb(skb1)->keys_valid) {
 168                 choke_skb_cb(skb1)->keys_valid = 1;
 169                 skb_flow_dissect_flow_keys(skb1, &temp, 0);
 170                 make_flow_keys_digest(&choke_skb_cb(skb1)->keys, &temp);
 171         }
 172
 173         if (!choke_skb_cb(skb2)->keys_valid) {
 174                 choke_skb_cb(skb2)->keys_valid = 1;
 175                 skb_flow_dissect_flow_keys(skb2, &temp, 0);
 176                 make_flow_keys_digest(&choke_skb_cb(skb2)->keys, &temp);
 177         }
 178
 179         return !memcmp(&choke_skb_cb(skb1)->keys,
 180                        &choke_skb_cb(skb2)->keys,
 181                        sizeof(choke_skb_cb(skb1)->keys));
 182 }
 183
 184 /*
 185  * Select a packet at random from queue
 186  * HACK: since queue can have holes from previous deletion; retry several
 187  *   times to find a random skb but then just give up and return the head
 188  * Will return NULL if queue is empty (q->head == q->tail)
 189  */
 190 static struct sk_buff *choke_peek_random(const struct choke_sched_data *q,
 191                                          unsigned int *pidx)
 192 {
 193         struct sk_buff *skb;
 194         int retrys = 3;
 195
 196         do {
 197                 *pidx = (q->head + prandom_u32_max(choke_len(q))) & q->tab_mask;
 198                 skb = q->tab[*pidx];
 199                 if (skb)
 200                         return skb;
 201         } while (--retrys > 0);
 202
 203         return q->tab[*pidx = q->head];
 204 }
 205
 206 /*
 207  * Compare new packet with random packet in queue
 208  * returns true if matched and sets *pidx
 209  */
 210 static bool choke_match_random(const struct choke_sched_data *q,
 211                                struct sk_buff *nskb,
 212                                unsigned int *pidx)
 213 {
 214         struct sk_buff *oskb;
 215
 216         if (q->head == q->tail)
 217                 return false;
 218
 219         oskb = choke_peek_random(q, pidx);
 220         return choke_match_flow(oskb, nskb);
 221 }
 222
 223 static int choke_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 224                          struct sk_buff **to_free)
 225 {
 226         struct choke_sched_data *q = qdisc_priv(sch);
 227         const struct red_parms *p = &q->parms;
 228
 229         choke_skb_cb(skb)->keys_valid = 0;
 230         /* Compute average queue usage (see RED) */
 231         q->vars.qavg = red_calc_qavg(p, &q->vars, sch->q.qlen);
 232         if (red_is_idling(&q->vars))
 233                 red_end_of_idle_period(&q->vars);
 234
 235         /* Is queue small? */
 236         if (q->vars.qavg <= p->qth_min)
 237                 q->vars.qcount = -1;
 238         else {
 239                 unsigned int idx;
 240
 241                 /* Draw a packet at random from queue and compare flow */
 242                 if (choke_match_random(q, skb, &idx)) {
 243                         q->stats.matched++;
 244                         choke_drop_by_idx(sch, idx, to_free);
 245                         goto congestion_drop;
 246                 }
 247
 248                 /* Queue is large, always mark/drop */
 249                 if (q->vars.qavg > p->qth_max) {
 250                         q->vars.qcount = -1;
 251
 252                         qdisc_qstats_overlimit(sch);
 253                         if (use_harddrop(q) || !use_ecn(q) ||
 254                             !INET_ECN_set_ce(skb)) {
 255                                 q->stats.forced_drop++;
 256                                 goto congestion_drop;
 257                         }
 258
 259                         q->stats.forced_mark++;
 260                 } else if (++q->vars.qcount) {
 261                         if (red_mark_probability(p, &q->vars, q->vars.qavg)) {
 262                                 q->vars.qcount = 0;
 263                                 q->vars.qR = red_random(p);
 264
 265                                 qdisc_qstats_overlimit(sch);
 266                                 if (!use_ecn(q) || !INET_ECN_set_ce(skb)) {
 267                                         q->stats.prob_drop++;
 268                                         goto congestion_drop;
 269                                 }
 270
 271                                 q->stats.prob_mark++;
 272                         }
 273                 } else
 274                         q->vars.qR = red_random(p);
 275         }
 276
 277         /* Admit new packet */
 278         if (sch->q.qlen < q->limit) {
 279                 q->tab[q->tail] = skb;
 280                 q->tail = (q->tail + 1) & q->tab_mask;
 281                 ++sch->q.qlen;
 282                 qdisc_qstats_backlog_inc(sch, skb);
 283                 return NET_XMIT_SUCCESS;
 284         }
 285
 286         q->stats.pdrop++;
 287         return qdisc_drop(skb, sch, to_free);
 288
 289 congestion_drop:
 290         qdisc_drop(skb, sch, to_free);
 291         return NET_XMIT_CN;
 292 }
 293
 294 static struct sk_buff *choke_dequeue(struct Qdisc *sch)
 295 {
 296         struct choke_sched_data *q = qdisc_priv(sch);
 297         struct sk_buff *skb;
 298
 299         if (q->head == q->tail) {
 300                 if (!red_is_idling(&q->vars))
 301                         red_start_of_idle_period(&q->vars);
 302                 return NULL;
 303         }
 304
 305         skb = q->tab[q->head];
 306         q->tab[q->head] = NULL;
 307         choke_zap_head_holes(q);
 308         --sch->q.qlen;
 309         qdisc_qstats_backlog_dec(sch, skb);
 310         qdisc_bstats_update(sch, skb);
 311
 312         return skb;
 313 }
 314
 315 static void choke_reset(struct Qdisc *sch)
 316 {
 317         struct choke_sched_data *q = qdisc_priv(sch);
 318
 319         while (q->head != q->tail) {
 320                 struct sk_buff *skb = q->tab[q->head];
 321
 322                 q->head = (q->head + 1) & q->tab_mask;
 323                 if (!skb)
 324                         continue;
 325                 rtnl_qdisc_drop(skb, sch);
 326         }
 327
 328         sch->q.qlen = 0;
 329         sch->qstats.backlog = 0;
 330         memset(q->tab, 0, (q->tab_mask + 1) * sizeof(struct sk_buff *));
 331         q->head = q->tail = 0;
 332         red_restart(&q->vars);
 333 }
 334
 335 static const struct nla_policy choke_policy[TCA_CHOKE_MAX + 1] = {
 336         [TCA_CHOKE_PARMS]       = { .len = sizeof(struct tc_red_qopt) },
 337         [TCA_CHOKE_STAB]        = { .len = RED_STAB_SIZE },
 338         [TCA_CHOKE_MAX_P]       = { .type = NLA_U32 },
 339 };
 340
 341
 342 static void choke_free(void *addr)
 343 {
 344         kvfree(addr);
 345 }
 346
 347 static int choke_change(struct Qdisc *sch, struct nlattr *opt,
 348                         struct netlink_ext_ack *extack)
 349 {
 350         struct choke_sched_data *q = qdisc_priv(sch);
 351         struct nlattr *tb[TCA_CHOKE_MAX + 1];
 352         const struct tc_red_qopt *ctl;
 353         int err;
 354         struct sk_buff **old = NULL;
 355         unsigned int mask;
 356         u32 max_P;
 357
 358         if (opt == NULL)
 359                 return -EINVAL;
 360
 361         err = nla_parse_nested(tb, TCA_CHOKE_MAX, opt, choke_policy, NULL);
 362         if (err < 0)
 363                 return err;
 364
 365         if (tb[TCA_CHOKE_PARMS] == NULL ||
 366             tb[TCA_CHOKE_STAB] == NULL)
 367                 return -EINVAL;
 368
 369         max_P = tb[TCA_CHOKE_MAX_P] ? nla_get_u32(tb[TCA_CHOKE_MAX_P]) : 0;
 370
 371         ctl = nla_data(tb[TCA_CHOKE_PARMS]);
 372
 373         if (!red_check_params(ctl->qth_min, ctl->qth_max, ctl->Wlog))
 374                 return -EINVAL;
 375
 376         if (ctl->limit > CHOKE_MAX_QUEUE)
 377                 return -EINVAL;
 378
 379         mask = roundup_pow_of_two(ctl->limit + 1) - 1;
 380         if (mask != q->tab_mask) {
 381                 struct sk_buff **ntab;
 382
 383                 ntab = kvmalloc_array((mask + 1), sizeof(struct sk_buff *), GFP_KERNEL | __GFP_ZERO);
 384                 if (!ntab)
 385                         return -ENOMEM;
 386
 387                 sch_tree_lock(sch);
 388                 old = q->tab;
 389                 if (old) {
 390                         unsigned int oqlen = sch->q.qlen, tail = 0;
 391                         unsigned dropped = 0;
 392
 393                         while (q->head != q->tail) {
 394                                 struct sk_buff *skb = q->tab[q->head];
 395
 396                                 q->head = (q->head + 1) & q->tab_mask;
 397                                 if (!skb)
 398                                         continue;
 399                                 if (tail < mask) {
 400                                         ntab[tail++] = skb;
 401                                         continue;
 402                                 }
 403                                 dropped += qdisc_pkt_len(skb);
 404                                 qdisc_qstats_backlog_dec(sch, skb);
 405                                 --sch->q.qlen;
 406                                 rtnl_qdisc_drop(skb, sch);
 407                         }
 408                         qdisc_tree_reduce_backlog(sch, oqlen - sch->q.qlen, dropped);
 409                         q->head = 0;
 410                         q->tail = tail;
 411                 }
 412
 413                 q->tab_mask = mask;
 414                 q->tab = ntab;
 415         } else
 416                 sch_tree_lock(sch);
 417
 418         q->flags = ctl->flags;
 419         q->limit = ctl->limit;
 420
 421         red_set_parms(&q->parms, ctl->qth_min, ctl->qth_max, ctl->Wlog,
 422                       ctl->Plog, ctl->Scell_log,
 423                       nla_data(tb[TCA_CHOKE_STAB]),
 424                       max_P);
 425         red_set_vars(&q->vars);
 426
 427         if (q->head == q->tail)
 428                 red_end_of_idle_period(&q->vars);
 429
 430         sch_tree_unlock(sch);
 431         choke_free(old);
 432         return 0;
 433 }
 434
 435 static int choke_init(struct Qdisc *sch, struct nlattr *opt,
 436                       struct netlink_ext_ack *extack)
 437 {
 438         return choke_change(sch, opt, extack);
 439 }
 440
 441 static int choke_dump(struct Qdisc *sch, struct sk_buff *skb)
 442 {
 443         struct choke_sched_data *q = qdisc_priv(sch);
 444         struct nlattr *opts = NULL;
 445         struct tc_red_qopt opt = {
 446                 .limit          = q->limit,
 447                 .flags          = q->flags,
 448                 .qth_min        = q->parms.qth_min >> q->parms.Wlog,
 449                 .qth_max        = q->parms.qth_max >> q->parms.Wlog,
 450                 .Wlog           = q->parms.Wlog,
 451                 .Plog           = q->parms.Plog,
 452                 .Scell_log      = q->parms.Scell_log,
 453         };
 454
 455         opts = nla_nest_start(skb, TCA_OPTIONS);
 456         if (opts == NULL)
 457                 goto nla_put_failure;
 458
 459         if (nla_put(skb, TCA_CHOKE_PARMS, sizeof(opt), &opt) ||
 460             nla_put_u32(skb, TCA_CHOKE_MAX_P, q->parms.max_P))
 461                 goto nla_put_failure;
 462         return nla_nest_end(skb, opts);
 463
 464 nla_put_failure:
 465         nla_nest_cancel(skb, opts);
 466         return -EMSGSIZE;
 467 }
 468
 469 static int choke_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 470 {
 471         struct choke_sched_data *q = qdisc_priv(sch);
 472         struct tc_choke_xstats st = {
 473                 .early  = q->stats.prob_drop + q->stats.forced_drop,
 474                 .marked = q->stats.prob_mark + q->stats.forced_mark,
 475                 .pdrop  = q->stats.pdrop,
 476                 .other  = q->stats.other,
 477                 .matched = q->stats.matched,
 478         };
 479
 480         return gnet_stats_copy_app(d, &st, sizeof(st));
 481 }
 482
 483 static void choke_destroy(struct Qdisc *sch)
 484 {
 485         struct choke_sched_data *q = qdisc_priv(sch);
 486
 487         choke_free(q->tab);
 488 }
 489
 490 static struct sk_buff *choke_peek_head(struct Qdisc *sch)
 491 {
 492         struct choke_sched_data *q = qdisc_priv(sch);
 493
 494         return (q->head != q->tail) ? q->tab[q->head] : NULL;
 495 }
 496
 497 static struct Qdisc_ops choke_qdisc_ops __read_mostly = {
 498         .id             =       "choke",
 499         .priv_size      =       sizeof(struct choke_sched_data),
 500
 501         .enqueue        =       choke_enqueue,
 502         .dequeue        =       choke_dequeue,
 503         .peek           =       choke_peek_head,
 504         .init           =       choke_init,
 505         .destroy        =       choke_destroy,
 506         .reset          =       choke_reset,
 507         .change         =       choke_change,
 508         .dump           =       choke_dump,
 509         .dump_stats     =       choke_dump_stats,
 510         .owner          =       THIS_MODULE,
 511 };
 512
 513 static int __init choke_module_init(void)
 514 {
 515         return register_qdisc(&choke_qdisc_ops);
 516 }
 517
 518 static void __exit choke_module_exit(void)
 519 {
 520         unregister_qdisc(&choke_qdisc_ops);
 521 }
 522
 523 module_init(choke_module_init)
 524 module_exit(choke_module_exit)
 525
 526 MODULE_LICENSE("GPL");