net/sched/sch_gred.c

   1 /*
   2  * net/sched/sch_gred.c Generic Random Early Detection queue.
   3  *
   4  *
   5  *              This program is free software; you can redistribute it and/or
   6  *              modify it under the terms of the GNU General Public License
   7  *              as published by the Free Software Foundation; either version
   8  *              2 of the License, or (at your option) any later version.
   9  *
  10  * Authors:    J Hadi Salim (hadi@cyberus.ca) 1998-2002
  11  *
  12  *             991129: -  Bug fix with grio mode
  13  *                     - a better sing. AvgQ mode with Grio(WRED)
  14  *                     - A finer grained VQ dequeue based on sugestion
  15  *                       from Ren Liu
  16  *                     - More error checks
  17  *
  18  *  For all the glorious comments look at include/net/red.h
  19  */
  20
  21 #include <linux/slab.h>
  22 #include <linux/module.h>
  23 #include <linux/types.h>
  24 #include <linux/kernel.h>
  25 #include <linux/skbuff.h>
  26 #include <net/pkt_sched.h>
  27 #include <net/red.h>
  28
  29 #define GRED_DEF_PRIO (MAX_DPs / 2)
  30 #define GRED_VQ_MASK (MAX_DPs - 1)
  31
  32 struct gred_sched_data;
  33 struct gred_sched;
  34
  35 struct gred_sched_data {
  36         u32             limit;          /* HARD maximal queue length    */
  37         u32             DP;             /* the drop parameters */
  38         u32             bytesin;        /* bytes seen on virtualQ so far*/
  39         u32             packetsin;      /* packets seen on virtualQ so far*/
  40         u32             backlog;        /* bytes on the virtualQ */
  41         u8              prio;           /* the prio of this vq */
  42
  43         struct red_parms parms;
  44         struct red_vars  vars;
  45         struct red_stats stats;
  46 };
  47
  48 enum {
  49         GRED_WRED_MODE = 1,
  50         GRED_RIO_MODE,
  51 };
  52
  53 struct gred_sched {
  54         struct gred_sched_data *tab[MAX_DPs];
  55         unsigned long   flags;
  56         u32             red_flags;
  57         u32             DPs;
  58         u32             def;
  59         struct red_vars wred_set;
  60 };
  61
  62 static inline int gred_wred_mode(struct gred_sched *table)
  63 {
  64         return test_bit(GRED_WRED_MODE, &table->flags);
  65 }
  66
  67 static inline void gred_enable_wred_mode(struct gred_sched *table)
  68 {
  69         __set_bit(GRED_WRED_MODE, &table->flags);
  70 }
  71
  72 static inline void gred_disable_wred_mode(struct gred_sched *table)
  73 {
  74         __clear_bit(GRED_WRED_MODE, &table->flags);
  75 }
  76
  77 static inline int gred_rio_mode(struct gred_sched *table)
  78 {
  79         return test_bit(GRED_RIO_MODE, &table->flags);
  80 }
  81
  82 static inline void gred_enable_rio_mode(struct gred_sched *table)
  83 {
  84         __set_bit(GRED_RIO_MODE, &table->flags);
  85 }
  86
  87 static inline void gred_disable_rio_mode(struct gred_sched *table)
  88 {
  89         __clear_bit(GRED_RIO_MODE, &table->flags);
  90 }
  91
  92 static inline int gred_wred_mode_check(struct Qdisc *sch)
  93 {
  94         struct gred_sched *table = qdisc_priv(sch);
  95         int i;
  96
  97         /* Really ugly O(n^2) but shouldn't be necessary too frequent. */
  98         for (i = 0; i < table->DPs; i++) {
  99                 struct gred_sched_data *q = table->tab[i];
 100                 int n;
 101
 102                 if (q == NULL)
 103                         continue;
 104
 105                 for (n = 0; n < table->DPs; n++)
 106                         if (table->tab[n] && table->tab[n] != q &&
 107                             table->tab[n]->prio == q->prio)
 108                                 return 1;
 109         }
 110
 111         return 0;
 112 }
 113
 114 static inline unsigned int gred_backlog(struct gred_sched *table,
 115                                         struct gred_sched_data *q,
 116                                         struct Qdisc *sch)
 117 {
 118         if (gred_wred_mode(table))
 119                 return sch->qstats.backlog;
 120         else
 121                 return q->backlog;
 122 }
 123
 124 static inline u16 tc_index_to_dp(struct sk_buff *skb)
 125 {
 126         return skb->tc_index & GRED_VQ_MASK;
 127 }
 128
 129 static inline void gred_load_wred_set(const struct gred_sched *table,
 130                                       struct gred_sched_data *q)
 131 {
 132         q->vars.qavg = table->wred_set.qavg;
 133         q->vars.qidlestart = table->wred_set.qidlestart;
 134 }
 135
 136 static inline void gred_store_wred_set(struct gred_sched *table,
 137                                        struct gred_sched_data *q)
 138 {
 139         table->wred_set.qavg = q->vars.qavg;
 140 }
 141
 142 static inline int gred_use_ecn(struct gred_sched *t)
 143 {
 144         return t->red_flags & TC_RED_ECN;
 145 }
 146
 147 static inline int gred_use_harddrop(struct gred_sched *t)
 148 {
 149         return t->red_flags & TC_RED_HARDDROP;
 150 }
 151
 152 static int gred_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 153 {
 154         struct gred_sched_data *q = NULL;
 155         struct gred_sched *t = qdisc_priv(sch);
 156         unsigned long qavg = 0;
 157         u16 dp = tc_index_to_dp(skb);
 158
 159         if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 160                 dp = t->def;
 161
 162                 q = t->tab[dp];
 163                 if (!q) {
 164                         /* Pass through packets not assigned to a DP
 165                          * if no default DP has been configured. This
 166                          * allows for DP flows to be left untouched.
 167                          */
 168                         if (skb_queue_len(&sch->q) < qdisc_dev(sch)->tx_queue_len)
 169                                 return qdisc_enqueue_tail(skb, sch);
 170                         else
 171                                 goto drop;
 172                 }
 173
 174                 /* fix tc_index? --could be controversial but needed for
 175                    requeueing */
 176                 skb->tc_index = (skb->tc_index & ~GRED_VQ_MASK) | dp;
 177         }
 178
 179         /* sum up all the qaves of prios <= to ours to get the new qave */
 180         if (!gred_wred_mode(t) && gred_rio_mode(t)) {
 181                 int i;
 182
 183                 for (i = 0; i < t->DPs; i++) {
 184                         if (t->tab[i] && t->tab[i]->prio < q->prio &&
 185                             !red_is_idling(&t->tab[i]->vars))
 186                                 qavg += t->tab[i]->vars.qavg;
 187                 }
 188
 189         }
 190
 191         q->packetsin++;
 192         q->bytesin += qdisc_pkt_len(skb);
 193
 194         if (gred_wred_mode(t))
 195                 gred_load_wred_set(t, q);
 196
 197         q->vars.qavg = red_calc_qavg(&q->parms,
 198                                      &q->vars,
 199                                      gred_backlog(t, q, sch));
 200
 201         if (red_is_idling(&q->vars))
 202                 red_end_of_idle_period(&q->vars);
 203
 204         if (gred_wred_mode(t))
 205                 gred_store_wred_set(t, q);
 206
 207         switch (red_action(&q->parms, &q->vars, q->vars.qavg + qavg)) {
 208         case RED_DONT_MARK:
 209                 break;
 210
 211         case RED_PROB_MARK:
 212                 sch->qstats.overlimits++;
 213                 if (!gred_use_ecn(t) || !INET_ECN_set_ce(skb)) {
 214                         q->stats.prob_drop++;
 215                         goto congestion_drop;
 216                 }
 217
 218                 q->stats.prob_mark++;
 219                 break;
 220
 221         case RED_HARD_MARK:
 222                 sch->qstats.overlimits++;
 223                 if (gred_use_harddrop(t) || !gred_use_ecn(t) ||
 224                     !INET_ECN_set_ce(skb)) {
 225                         q->stats.forced_drop++;
 226                         goto congestion_drop;
 227                 }
 228                 q->stats.forced_mark++;
 229                 break;
 230         }
 231
 232         if (q->backlog + qdisc_pkt_len(skb) <= q->limit) {
 233                 q->backlog += qdisc_pkt_len(skb);
 234                 return qdisc_enqueue_tail(skb, sch);
 235         }
 236
 237         q->stats.pdrop++;
 238 drop:
 239         return qdisc_drop(skb, sch);
 240
 241 congestion_drop:
 242         qdisc_drop(skb, sch);
 243         return NET_XMIT_CN;
 244 }
 245
 246 static struct sk_buff *gred_dequeue(struct Qdisc *sch)
 247 {
 248         struct sk_buff *skb;
 249         struct gred_sched *t = qdisc_priv(sch);
 250
 251         skb = qdisc_dequeue_head(sch);
 252
 253         if (skb) {
 254                 struct gred_sched_data *q;
 255                 u16 dp = tc_index_to_dp(skb);
 256
 257                 if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 258                         net_warn_ratelimited("GRED: Unable to relocate VQ 0x%x after dequeue, screwing up backlog\n",
 259                                              tc_index_to_dp(skb));
 260                 } else {
 261                         q->backlog -= qdisc_pkt_len(skb);
 262
 263                         if (!q->backlog && !gred_wred_mode(t))
 264                                 red_start_of_idle_period(&q->vars);
 265                 }
 266
 267                 return skb;
 268         }
 269
 270         if (gred_wred_mode(t) && !red_is_idling(&t->wred_set))
 271                 red_start_of_idle_period(&t->wred_set);
 272
 273         return NULL;
 274 }
 275
 276 static unsigned int gred_drop(struct Qdisc *sch)
 277 {
 278         struct sk_buff *skb;
 279         struct gred_sched *t = qdisc_priv(sch);
 280
 281         skb = qdisc_dequeue_tail(sch);
 282         if (skb) {
 283                 unsigned int len = qdisc_pkt_len(skb);
 284                 struct gred_sched_data *q;
 285                 u16 dp = tc_index_to_dp(skb);
 286
 287                 if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 288                         net_warn_ratelimited("GRED: Unable to relocate VQ 0x%x while dropping, screwing up backlog\n",
 289                                              tc_index_to_dp(skb));
 290                 } else {
 291                         q->backlog -= len;
 292                         q->stats.other++;
 293
 294                         if (!q->backlog && !gred_wred_mode(t))
 295                                 red_start_of_idle_period(&q->vars);
 296                 }
 297
 298                 qdisc_drop(skb, sch);
 299                 return len;
 300         }
 301
 302         if (gred_wred_mode(t) && !red_is_idling(&t->wred_set))
 303                 red_start_of_idle_period(&t->wred_set);
 304
 305         return 0;
 306
 307 }
 308
 309 static void gred_reset(struct Qdisc *sch)
 310 {
 311         int i;
 312         struct gred_sched *t = qdisc_priv(sch);
 313
 314         qdisc_reset_queue(sch);
 315
 316         for (i = 0; i < t->DPs; i++) {
 317                 struct gred_sched_data *q = t->tab[i];
 318
 319                 if (!q)
 320                         continue;
 321
 322                 red_restart(&q->vars);
 323                 q->backlog = 0;
 324         }
 325 }
 326
 327 static inline void gred_destroy_vq(struct gred_sched_data *q)
 328 {
 329         kfree(q);
 330 }
 331
 332 static inline int gred_change_table_def(struct Qdisc *sch, struct nlattr *dps)
 333 {
 334         struct gred_sched *table = qdisc_priv(sch);
 335         struct tc_gred_sopt *sopt;
 336         int i;
 337
 338         if (dps == NULL)
 339                 return -EINVAL;
 340
 341         sopt = nla_data(dps);
 342
 343         if (sopt->DPs > MAX_DPs || sopt->DPs == 0 || sopt->def_DP >= sopt->DPs)
 344                 return -EINVAL;
 345
 346         sch_tree_lock(sch);
 347         table->DPs = sopt->DPs;
 348         table->def = sopt->def_DP;
 349         table->red_flags = sopt->flags;
 350
 351         /*
 352          * Every entry point to GRED is synchronized with the above code
 353          * and the DP is checked against DPs, i.e. shadowed VQs can no
 354          * longer be found so we can unlock right here.
 355          */
 356         sch_tree_unlock(sch);
 357
 358         if (sopt->grio) {
 359                 gred_enable_rio_mode(table);
 360                 gred_disable_wred_mode(table);
 361                 if (gred_wred_mode_check(sch))
 362                         gred_enable_wred_mode(table);
 363         } else {
 364                 gred_disable_rio_mode(table);
 365                 gred_disable_wred_mode(table);
 366         }
 367
 368         for (i = table->DPs; i < MAX_DPs; i++) {
 369                 if (table->tab[i]) {
 370                         pr_warning("GRED: Warning: Destroying "
 371                                    "shadowed VQ 0x%x\n", i);
 372                         gred_destroy_vq(table->tab[i]);
 373                         table->tab[i] = NULL;
 374                 }
 375         }
 376
 377         return 0;
 378 }
 379
 380 static inline int gred_change_vq(struct Qdisc *sch, int dp,
 381                                  struct tc_gred_qopt *ctl, int prio,
 382                                  u8 *stab, u32 max_P,
 383                                  struct gred_sched_data **prealloc)
 384 {
 385         struct gred_sched *table = qdisc_priv(sch);
 386         struct gred_sched_data *q = table->tab[dp];
 387
 388         if (!q) {
 389                 table->tab[dp] = q = *prealloc;
 390                 *prealloc = NULL;
 391                 if (!q)
 392                         return -ENOMEM;
 393         }
 394
 395         q->DP = dp;
 396         q->prio = prio;
 397         q->limit = ctl->limit;
 398
 399         if (q->backlog == 0)
 400                 red_end_of_idle_period(&q->vars);
 401
 402         red_set_parms(&q->parms,
 403                       ctl->qth_min, ctl->qth_max, ctl->Wlog, ctl->Plog,
 404                       ctl->Scell_log, stab, max_P);
 405         red_set_vars(&q->vars);
 406         return 0;
 407 }
 408
 409 static const struct nla_policy gred_policy[TCA_GRED_MAX + 1] = {
 410         [TCA_GRED_PARMS]        = { .len = sizeof(struct tc_gred_qopt) },
 411         [TCA_GRED_STAB]         = { .len = 256 },
 412         [TCA_GRED_DPS]          = { .len = sizeof(struct tc_gred_sopt) },
 413         [TCA_GRED_MAX_P]        = { .type = NLA_U32 },
 414 };
 415
 416 static int gred_change(struct Qdisc *sch, struct nlattr *opt)
 417 {
 418         struct gred_sched *table = qdisc_priv(sch);
 419         struct tc_gred_qopt *ctl;
 420         struct nlattr *tb[TCA_GRED_MAX + 1];
 421         int err, prio = GRED_DEF_PRIO;
 422         u8 *stab;
 423         u32 max_P;
 424         struct gred_sched_data *prealloc;
 425
 426         if (opt == NULL)
 427                 return -EINVAL;
 428
 429         err = nla_parse_nested(tb, TCA_GRED_MAX, opt, gred_policy);
 430         if (err < 0)
 431                 return err;
 432
 433         if (tb[TCA_GRED_PARMS] == NULL && tb[TCA_GRED_STAB] == NULL)
 434                 return gred_change_table_def(sch, opt);
 435
 436         if (tb[TCA_GRED_PARMS] == NULL ||
 437             tb[TCA_GRED_STAB] == NULL)
 438                 return -EINVAL;
 439
 440         max_P = tb[TCA_GRED_MAX_P] ? nla_get_u32(tb[TCA_GRED_MAX_P]) : 0;
 441
 442         err = -EINVAL;
 443         ctl = nla_data(tb[TCA_GRED_PARMS]);
 444         stab = nla_data(tb[TCA_GRED_STAB]);
 445
 446         if (ctl->DP >= table->DPs)
 447                 goto errout;
 448
 449         if (gred_rio_mode(table)) {
 450                 if (ctl->prio == 0) {
 451                         int def_prio = GRED_DEF_PRIO;
 452
 453                         if (table->tab[table->def])
 454                                 def_prio = table->tab[table->def]->prio;
 455
 456                         printk(KERN_DEBUG "GRED: DP %u does not have a prio "
 457                                "setting default to %d\n", ctl->DP, def_prio);
 458
 459                         prio = def_prio;
 460                 } else
 461                         prio = ctl->prio;
 462         }
 463
 464         prealloc = kzalloc(sizeof(*prealloc), GFP_KERNEL);
 465         sch_tree_lock(sch);
 466
 467         err = gred_change_vq(sch, ctl->DP, ctl, prio, stab, max_P, &prealloc);
 468         if (err < 0)
 469                 goto errout_locked;
 470
 471         if (gred_rio_mode(table)) {
 472                 gred_disable_wred_mode(table);
 473                 if (gred_wred_mode_check(sch))
 474                         gred_enable_wred_mode(table);
 475         }
 476
 477         err = 0;
 478
 479 errout_locked:
 480         sch_tree_unlock(sch);
 481         kfree(prealloc);
 482 errout:
 483         return err;
 484 }
 485
 486 static int gred_init(struct Qdisc *sch, struct nlattr *opt)
 487 {
 488         struct nlattr *tb[TCA_GRED_MAX + 1];
 489         int err;
 490
 491         if (opt == NULL)
 492                 return -EINVAL;
 493
 494         err = nla_parse_nested(tb, TCA_GRED_MAX, opt, gred_policy);
 495         if (err < 0)
 496                 return err;
 497
 498         if (tb[TCA_GRED_PARMS] || tb[TCA_GRED_STAB])
 499                 return -EINVAL;
 500
 501         return gred_change_table_def(sch, tb[TCA_GRED_DPS]);
 502 }
 503
 504 static int gred_dump(struct Qdisc *sch, struct sk_buff *skb)
 505 {
 506         struct gred_sched *table = qdisc_priv(sch);
 507         struct nlattr *parms, *opts = NULL;
 508         int i;
 509         u32 max_p[MAX_DPs];
 510         struct tc_gred_sopt sopt = {
 511                 .DPs    = table->DPs,
 512                 .def_DP = table->def,
 513                 .grio   = gred_rio_mode(table),
 514                 .flags  = table->red_flags,
 515         };
 516
 517         opts = nla_nest_start(skb, TCA_OPTIONS);
 518         if (opts == NULL)
 519                 goto nla_put_failure;
 520         if (nla_put(skb, TCA_GRED_DPS, sizeof(sopt), &sopt))
 521                 goto nla_put_failure;
 522
 523         for (i = 0; i < MAX_DPs; i++) {
 524                 struct gred_sched_data *q = table->tab[i];
 525
 526                 max_p[i] = q ? q->parms.max_P : 0;
 527         }
 528         if (nla_put(skb, TCA_GRED_MAX_P, sizeof(max_p), max_p))
 529                 goto nla_put_failure;
 530
 531         parms = nla_nest_start(skb, TCA_GRED_PARMS);
 532         if (parms == NULL)
 533                 goto nla_put_failure;
 534
 535         for (i = 0; i < MAX_DPs; i++) {
 536                 struct gred_sched_data *q = table->tab[i];
 537                 struct tc_gred_qopt opt;
 538
 539                 memset(&opt, 0, sizeof(opt));
 540
 541                 if (!q) {
 542                         /* hack -- fix at some point with proper message
 543                            This is how we indicate to tc that there is no VQ
 544                            at this DP */
 545
 546                         opt.DP = MAX_DPs + i;
 547                         goto append_opt;
 548                 }
 549
 550                 opt.limit       = q->limit;
 551                 opt.DP          = q->DP;
 552                 opt.backlog     = q->backlog;
 553                 opt.prio        = q->prio;
 554                 opt.qth_min     = q->parms.qth_min >> q->parms.Wlog;
 555                 opt.qth_max     = q->parms.qth_max >> q->parms.Wlog;
 556                 opt.Wlog        = q->parms.Wlog;
 557                 opt.Plog        = q->parms.Plog;
 558                 opt.Scell_log   = q->parms.Scell_log;
 559                 opt.other       = q->stats.other;
 560                 opt.early       = q->stats.prob_drop;
 561                 opt.forced      = q->stats.forced_drop;
 562                 opt.pdrop       = q->stats.pdrop;
 563                 opt.packets     = q->packetsin;
 564                 opt.bytesin     = q->bytesin;
 565
 566                 if (gred_wred_mode(table))
 567                         gred_load_wred_set(table, q);
 568
 569                 opt.qave = red_calc_qavg(&q->parms, &q->vars, q->vars.qavg);
 570
 571 append_opt:
 572                 if (nla_append(skb, sizeof(opt), &opt) < 0)
 573                         goto nla_put_failure;
 574         }
 575
 576         nla_nest_end(skb, parms);
 577
 578         return nla_nest_end(skb, opts);
 579
 580 nla_put_failure:
 581         nla_nest_cancel(skb, opts);
 582         return -EMSGSIZE;
 583 }
 584
 585 static void gred_destroy(struct Qdisc *sch)
 586 {
 587         struct gred_sched *table = qdisc_priv(sch);
 588         int i;
 589
 590         for (i = 0; i < table->DPs; i++) {
 591                 if (table->tab[i])
 592                         gred_destroy_vq(table->tab[i]);
 593         }
 594 }
 595
 596 static struct Qdisc_ops gred_qdisc_ops __read_mostly = {
 597         .id             =       "gred",
 598         .priv_size      =       sizeof(struct gred_sched),
 599         .enqueue        =       gred_enqueue,
 600         .dequeue        =       gred_dequeue,
 601         .peek           =       qdisc_peek_head,
 602         .drop           =       gred_drop,
 603         .init           =       gred_init,
 604         .reset          =       gred_reset,
 605         .destroy        =       gred_destroy,
 606         .change         =       gred_change,
 607         .dump           =       gred_dump,
 608         .owner          =       THIS_MODULE,
 609 };
 610
 611 static int __init gred_module_init(void)
 612 {
 613         return register_qdisc(&gred_qdisc_ops);
 614 }
 615
 616 static void __exit gred_module_exit(void)
 617 {
 618         unregister_qdisc(&gred_qdisc_ops);
 619 }
 620
 621 module_init(gred_module_init)
 622 module_exit(gred_module_exit)
 623
 624 MODULE_LICENSE("GPL");