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/module.h>
  22 #include <linux/types.h>
  23 #include <linux/kernel.h>
  24 #include <linux/skbuff.h>
  25 #include <net/pkt_sched.h>
  26 #include <net/red.h>
  27
  28 #define GRED_DEF_PRIO (MAX_DPs / 2)
  29 #define GRED_VQ_MASK (MAX_DPs - 1)
  30
  31 struct gred_sched_data;
  32 struct gred_sched;
  33
  34 struct gred_sched_data
  35 {
  36         u32             limit;          /* HARD maximal queue length    */
  37         u32             DP;             /* the drop pramaters */
  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_stats stats;
  45 };
  46
  47 enum {
  48         GRED_WRED_MODE = 1,
  49         GRED_RIO_MODE,
  50 };
  51
  52 struct gred_sched
  53 {
  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_parms 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(struct gred_sched *table,
 130                                       struct gred_sched_data *q)
 131 {
 132         q->parms.qavg = table->wred_set.qavg;
 133         q->parms.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->parms.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                 if ((q = t->tab[dp]) == NULL) {
 163                         /* Pass through packets not assigned to a DP
 164                          * if no default DP has been configured. This
 165                          * allows for DP flows to be left untouched.
 166                          */
 167                         if (skb_queue_len(&sch->q) < qdisc_dev(sch)->tx_queue_len)
 168                                 return qdisc_enqueue_tail(skb, sch);
 169                         else
 170                                 goto drop;
 171                 }
 172
 173                 /* fix tc_index? --could be controvesial but needed for
 174                    requeueing */
 175                 skb->tc_index = (skb->tc_index & ~GRED_VQ_MASK) | dp;
 176         }
 177
 178         /* sum up all the qaves of prios <= to ours to get the new qave */
 179         if (!gred_wred_mode(t) && gred_rio_mode(t)) {
 180                 int i;
 181
 182                 for (i = 0; i < t->DPs; i++) {
 183                         if (t->tab[i] && t->tab[i]->prio < q->prio &&
 184                             !red_is_idling(&t->tab[i]->parms))
 185                                 qavg +=t->tab[i]->parms.qavg;
 186                 }
 187
 188         }
 189
 190         q->packetsin++;
 191         q->bytesin += qdisc_pkt_len(skb);
 192
 193         if (gred_wred_mode(t))
 194                 gred_load_wred_set(t, q);
 195
 196         q->parms.qavg = red_calc_qavg(&q->parms, gred_backlog(t, q, sch));
 197
 198         if (red_is_idling(&q->parms))
 199                 red_end_of_idle_period(&q->parms);
 200
 201         if (gred_wred_mode(t))
 202                 gred_store_wred_set(t, q);
 203
 204         switch (red_action(&q->parms, q->parms.qavg + qavg)) {
 205                 case RED_DONT_MARK:
 206                         break;
 207
 208                 case RED_PROB_MARK:
 209                         sch->qstats.overlimits++;
 210                         if (!gred_use_ecn(t) || !INET_ECN_set_ce(skb)) {
 211                                 q->stats.prob_drop++;
 212                                 goto congestion_drop;
 213                         }
 214
 215                         q->stats.prob_mark++;
 216                         break;
 217
 218                 case RED_HARD_MARK:
 219                         sch->qstats.overlimits++;
 220                         if (gred_use_harddrop(t) || !gred_use_ecn(t) ||
 221                             !INET_ECN_set_ce(skb)) {
 222                                 q->stats.forced_drop++;
 223                                 goto congestion_drop;
 224                         }
 225                         q->stats.forced_mark++;
 226                         break;
 227         }
 228
 229         if (q->backlog + qdisc_pkt_len(skb) <= q->limit) {
 230                 q->backlog += qdisc_pkt_len(skb);
 231                 return qdisc_enqueue_tail(skb, sch);
 232         }
 233
 234         q->stats.pdrop++;
 235 drop:
 236         return qdisc_drop(skb, sch);
 237
 238 congestion_drop:
 239         qdisc_drop(skb, sch);
 240         return NET_XMIT_CN;
 241 }
 242
 243 static int gred_requeue(struct sk_buff *skb, struct Qdisc* sch)
 244 {
 245         struct gred_sched *t = qdisc_priv(sch);
 246         struct gred_sched_data *q;
 247         u16 dp = tc_index_to_dp(skb);
 248
 249         if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 250                 if (net_ratelimit())
 251                         printk(KERN_WARNING "GRED: Unable to relocate VQ 0x%x "
 252                                "for requeue, screwing up backlog.\n",
 253                                tc_index_to_dp(skb));
 254         } else {
 255                 if (red_is_idling(&q->parms))
 256                         red_end_of_idle_period(&q->parms);
 257                 q->backlog += qdisc_pkt_len(skb);
 258         }
 259
 260         return qdisc_requeue(skb, sch);
 261 }
 262
 263 static struct sk_buff *gred_dequeue(struct Qdisc* sch)
 264 {
 265         struct sk_buff *skb;
 266         struct gred_sched *t = qdisc_priv(sch);
 267
 268         skb = qdisc_dequeue_head(sch);
 269
 270         if (skb) {
 271                 struct gred_sched_data *q;
 272                 u16 dp = tc_index_to_dp(skb);
 273
 274                 if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 275                         if (net_ratelimit())
 276                                 printk(KERN_WARNING "GRED: Unable to relocate "
 277                                        "VQ 0x%x after dequeue, screwing up "
 278                                        "backlog.\n", tc_index_to_dp(skb));
 279                 } else {
 280                         q->backlog -= qdisc_pkt_len(skb);
 281
 282                         if (!q->backlog && !gred_wred_mode(t))
 283                                 red_start_of_idle_period(&q->parms);
 284                 }
 285
 286                 return skb;
 287         }
 288
 289         if (gred_wred_mode(t) && !red_is_idling(&t->wred_set))
 290                 red_start_of_idle_period(&t->wred_set);
 291
 292         return NULL;
 293 }
 294
 295 static unsigned int gred_drop(struct Qdisc* sch)
 296 {
 297         struct sk_buff *skb;
 298         struct gred_sched *t = qdisc_priv(sch);
 299
 300         skb = qdisc_dequeue_tail(sch);
 301         if (skb) {
 302                 unsigned int len = qdisc_pkt_len(skb);
 303                 struct gred_sched_data *q;
 304                 u16 dp = tc_index_to_dp(skb);
 305
 306                 if (dp >= t->DPs || (q = t->tab[dp]) == NULL) {
 307                         if (net_ratelimit())
 308                                 printk(KERN_WARNING "GRED: Unable to relocate "
 309                                        "VQ 0x%x while dropping, screwing up "
 310                                        "backlog.\n", tc_index_to_dp(skb));
 311                 } else {
 312                         q->backlog -= len;
 313                         q->stats.other++;
 314
 315                         if (!q->backlog && !gred_wred_mode(t))
 316                                 red_start_of_idle_period(&q->parms);
 317                 }
 318
 319                 qdisc_drop(skb, sch);
 320                 return len;
 321         }
 322
 323         if (gred_wred_mode(t) && !red_is_idling(&t->wred_set))
 324                 red_start_of_idle_period(&t->wred_set);
 325
 326         return 0;
 327
 328 }
 329
 330 static void gred_reset(struct Qdisc* sch)
 331 {
 332         int i;
 333         struct gred_sched *t = qdisc_priv(sch);
 334
 335         qdisc_reset_queue(sch);
 336
 337         for (i = 0; i < t->DPs; i++) {
 338                 struct gred_sched_data *q = t->tab[i];
 339
 340                 if (!q)
 341                         continue;
 342
 343                 red_restart(&q->parms);
 344                 q->backlog = 0;
 345         }
 346 }
 347
 348 static inline void gred_destroy_vq(struct gred_sched_data *q)
 349 {
 350         kfree(q);
 351 }
 352
 353 static inline int gred_change_table_def(struct Qdisc *sch, struct nlattr *dps)
 354 {
 355         struct gred_sched *table = qdisc_priv(sch);
 356         struct tc_gred_sopt *sopt;
 357         int i;
 358
 359         if (dps == NULL)
 360                 return -EINVAL;
 361
 362         sopt = nla_data(dps);
 363
 364         if (sopt->DPs > MAX_DPs || sopt->DPs == 0 || sopt->def_DP >= sopt->DPs)
 365                 return -EINVAL;
 366
 367         sch_tree_lock(sch);
 368         table->DPs = sopt->DPs;
 369         table->def = sopt->def_DP;
 370         table->red_flags = sopt->flags;
 371
 372         /*
 373          * Every entry point to GRED is synchronized with the above code
 374          * and the DP is checked against DPs, i.e. shadowed VQs can no
 375          * longer be found so we can unlock right here.
 376          */
 377         sch_tree_unlock(sch);
 378
 379         if (sopt->grio) {
 380                 gred_enable_rio_mode(table);
 381                 gred_disable_wred_mode(table);
 382                 if (gred_wred_mode_check(sch))
 383                         gred_enable_wred_mode(table);
 384         } else {
 385                 gred_disable_rio_mode(table);
 386                 gred_disable_wred_mode(table);
 387         }
 388
 389         for (i = table->DPs; i < MAX_DPs; i++) {
 390                 if (table->tab[i]) {
 391                         printk(KERN_WARNING "GRED: Warning: Destroying "
 392                                "shadowed VQ 0x%x\n", i);
 393                         gred_destroy_vq(table->tab[i]);
 394                         table->tab[i] = NULL;
 395                 }
 396         }
 397
 398         return 0;
 399 }
 400
 401 static inline int gred_change_vq(struct Qdisc *sch, int dp,
 402                                  struct tc_gred_qopt *ctl, int prio, u8 *stab)
 403 {
 404         struct gred_sched *table = qdisc_priv(sch);
 405         struct gred_sched_data *q;
 406
 407         if (table->tab[dp] == NULL) {
 408                 table->tab[dp] = kzalloc(sizeof(*q), GFP_KERNEL);
 409                 if (table->tab[dp] == NULL)
 410                         return -ENOMEM;
 411         }
 412
 413         q = table->tab[dp];
 414         q->DP = dp;
 415         q->prio = prio;
 416         q->limit = ctl->limit;
 417
 418         if (q->backlog == 0)
 419                 red_end_of_idle_period(&q->parms);
 420
 421         red_set_parms(&q->parms,
 422                       ctl->qth_min, ctl->qth_max, ctl->Wlog, ctl->Plog,
 423                       ctl->Scell_log, stab);
 424
 425         return 0;
 426 }
 427
 428 static const struct nla_policy gred_policy[TCA_GRED_MAX + 1] = {
 429         [TCA_GRED_PARMS]        = { .len = sizeof(struct tc_gred_qopt) },
 430         [TCA_GRED_STAB]         = { .len = 256 },
 431         [TCA_GRED_DPS]          = { .len = sizeof(struct tc_gred_sopt) },
 432 };
 433
 434 static int gred_change(struct Qdisc *sch, struct nlattr *opt)
 435 {
 436         struct gred_sched *table = qdisc_priv(sch);
 437         struct tc_gred_qopt *ctl;
 438         struct nlattr *tb[TCA_GRED_MAX + 1];
 439         int err, prio = GRED_DEF_PRIO;
 440         u8 *stab;
 441
 442         if (opt == NULL)
 443                 return -EINVAL;
 444
 445         err = nla_parse_nested(tb, TCA_GRED_MAX, opt, gred_policy);
 446         if (err < 0)
 447                 return err;
 448
 449         if (tb[TCA_GRED_PARMS] == NULL && tb[TCA_GRED_STAB] == NULL)
 450                 return gred_change_table_def(sch, opt);
 451
 452         if (tb[TCA_GRED_PARMS] == NULL ||
 453             tb[TCA_GRED_STAB] == NULL)
 454                 return -EINVAL;
 455
 456         err = -EINVAL;
 457         ctl = nla_data(tb[TCA_GRED_PARMS]);
 458         stab = nla_data(tb[TCA_GRED_STAB]);
 459
 460         if (ctl->DP >= table->DPs)
 461                 goto errout;
 462
 463         if (gred_rio_mode(table)) {
 464                 if (ctl->prio == 0) {
 465                         int def_prio = GRED_DEF_PRIO;
 466
 467                         if (table->tab[table->def])
 468                                 def_prio = table->tab[table->def]->prio;
 469
 470                         printk(KERN_DEBUG "GRED: DP %u does not have a prio "
 471                                "setting default to %d\n", ctl->DP, def_prio);
 472
 473                         prio = def_prio;
 474                 } else
 475                         prio = ctl->prio;
 476         }
 477
 478         sch_tree_lock(sch);
 479
 480         err = gred_change_vq(sch, ctl->DP, ctl, prio, stab);
 481         if (err < 0)
 482                 goto errout_locked;
 483
 484         if (gred_rio_mode(table)) {
 485                 gred_disable_wred_mode(table);
 486                 if (gred_wred_mode_check(sch))
 487                         gred_enable_wred_mode(table);
 488         }
 489
 490         err = 0;
 491
 492 errout_locked:
 493         sch_tree_unlock(sch);
 494 errout:
 495         return err;
 496 }
 497
 498 static int gred_init(struct Qdisc *sch, struct nlattr *opt)
 499 {
 500         struct nlattr *tb[TCA_GRED_MAX + 1];
 501         int err;
 502
 503         if (opt == NULL)
 504                 return -EINVAL;
 505
 506         err = nla_parse_nested(tb, TCA_GRED_MAX, opt, gred_policy);
 507         if (err < 0)
 508                 return err;
 509
 510         if (tb[TCA_GRED_PARMS] || tb[TCA_GRED_STAB])
 511                 return -EINVAL;
 512
 513         return gred_change_table_def(sch, tb[TCA_GRED_DPS]);
 514 }
 515
 516 static int gred_dump(struct Qdisc *sch, struct sk_buff *skb)
 517 {
 518         struct gred_sched *table = qdisc_priv(sch);
 519         struct nlattr *parms, *opts = NULL;
 520         int i;
 521         struct tc_gred_sopt sopt = {
 522                 .DPs    = table->DPs,
 523                 .def_DP = table->def,
 524                 .grio   = gred_rio_mode(table),
 525                 .flags  = table->red_flags,
 526         };
 527
 528         opts = nla_nest_start(skb, TCA_OPTIONS);
 529         if (opts == NULL)
 530                 goto nla_put_failure;
 531         NLA_PUT(skb, TCA_GRED_DPS, sizeof(sopt), &sopt);
 532         parms = nla_nest_start(skb, TCA_GRED_PARMS);
 533         if (parms == NULL)
 534                 goto nla_put_failure;
 535
 536         for (i = 0; i < MAX_DPs; i++) {
 537                 struct gred_sched_data *q = table->tab[i];
 538                 struct tc_gred_qopt opt;
 539
 540                 memset(&opt, 0, sizeof(opt));
 541
 542                 if (!q) {
 543                         /* hack -- fix at some point with proper message
 544                            This is how we indicate to tc that there is no VQ
 545                            at this DP */
 546
 547                         opt.DP = MAX_DPs + i;
 548                         goto append_opt;
 549                 }
 550
 551                 opt.limit       = q->limit;
 552                 opt.DP          = q->DP;
 553                 opt.backlog     = q->backlog;
 554                 opt.prio        = q->prio;
 555                 opt.qth_min     = q->parms.qth_min >> q->parms.Wlog;
 556                 opt.qth_max     = q->parms.qth_max >> q->parms.Wlog;
 557                 opt.Wlog        = q->parms.Wlog;
 558                 opt.Plog        = q->parms.Plog;
 559                 opt.Scell_log   = q->parms.Scell_log;
 560                 opt.other       = q->stats.other;
 561                 opt.early       = q->stats.prob_drop;
 562                 opt.forced      = q->stats.forced_drop;
 563                 opt.pdrop       = q->stats.pdrop;
 564                 opt.packets     = q->packetsin;
 565                 opt.bytesin     = q->bytesin;
 566
 567                 if (gred_wred_mode(table)) {
 568                         q->parms.qidlestart =
 569                                 table->tab[table->def]->parms.qidlestart;
 570                         q->parms.qavg = table->tab[table->def]->parms.qavg;
 571                 }
 572
 573                 opt.qave = red_calc_qavg(&q->parms, q->parms.qavg);
 574
 575 append_opt:
 576                 if (nla_append(skb, sizeof(opt), &opt) < 0)
 577                         goto nla_put_failure;
 578         }
 579
 580         nla_nest_end(skb, parms);
 581
 582         return nla_nest_end(skb, opts);
 583
 584 nla_put_failure:
 585         nla_nest_cancel(skb, opts);
 586         return -EMSGSIZE;
 587 }
 588
 589 static void gred_destroy(struct Qdisc *sch)
 590 {
 591         struct gred_sched *table = qdisc_priv(sch);
 592         int i;
 593
 594         for (i = 0; i < table->DPs; i++) {
 595                 if (table->tab[i])
 596                         gred_destroy_vq(table->tab[i]);
 597         }
 598 }
 599
 600 static struct Qdisc_ops gred_qdisc_ops __read_mostly = {
 601         .id             =       "gred",
 602         .priv_size      =       sizeof(struct gred_sched),
 603         .enqueue        =       gred_enqueue,
 604         .dequeue        =       gred_dequeue,
 605         .requeue        =       gred_requeue,
 606         .drop           =       gred_drop,
 607         .init           =       gred_init,
 608         .reset          =       gred_reset,
 609         .destroy        =       gred_destroy,
 610         .change         =       gred_change,
 611         .dump           =       gred_dump,
 612         .owner          =       THIS_MODULE,
 613 };
 614
 615 static int __init gred_module_init(void)
 616 {
 617         return register_qdisc(&gred_qdisc_ops);
 618 }
 619
 620 static void __exit gred_module_exit(void)
 621 {
 622         unregister_qdisc(&gred_qdisc_ops);
 623 }
 624
 625 module_init(gred_module_init)
 626 module_exit(gred_module_exit)
 627
 628 MODULE_LICENSE("GPL");