net/sched/sch_cbq.c

   1 /*
   2  * net/sched/sch_cbq.c  Class-Based Queueing discipline.
   3  *
   4  *              This program is free software; you can redistribute it and/or
   5  *              modify it under the terms of the GNU General Public License
   6  *              as published by the Free Software Foundation; either version
   7  *              2 of the License, or (at your option) any later version.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  */
  12
  13 #include <linux/module.h>
  14 #include <linux/slab.h>
  15 #include <linux/types.h>
  16 #include <linux/kernel.h>
  17 #include <linux/string.h>
  18 #include <linux/errno.h>
  19 #include <linux/skbuff.h>
  20 #include <net/netlink.h>
  21 #include <net/pkt_sched.h>
  22
  23
  24 /*      Class-Based Queueing (CBQ) algorithm.
  25         =======================================
  26
  27         Sources: [1] Sally Floyd and Van Jacobson, "Link-sharing and Resource
  28                  Management Models for Packet Networks",
  29                  IEEE/ACM Transactions on Networking, Vol.3, No.4, 1995
  30
  31                  [2] Sally Floyd, "Notes on CBQ and Guaranteed Service", 1995
  32
  33                  [3] Sally Floyd, "Notes on Class-Based Queueing: Setting
  34                  Parameters", 1996
  35
  36                  [4] Sally Floyd and Michael Speer, "Experimental Results
  37                  for Class-Based Queueing", 1998, not published.
  38
  39         -----------------------------------------------------------------------
  40
  41         Algorithm skeleton was taken from NS simulator cbq.cc.
  42         If someone wants to check this code against the LBL version,
  43         he should take into account that ONLY the skeleton was borrowed,
  44         the implementation is different. Particularly:
  45
  46         --- The WRR algorithm is different. Our version looks more
  47         reasonable (I hope) and works when quanta are allowed to be
  48         less than MTU, which is always the case when real time classes
  49         have small rates. Note, that the statement of [3] is
  50         incomplete, delay may actually be estimated even if class
  51         per-round allotment is less than MTU. Namely, if per-round
  52         allotment is W*r_i, and r_1+...+r_k = r < 1
  53
  54         delay_i <= ([MTU/(W*r_i)]*W*r + W*r + k*MTU)/B
  55
  56         In the worst case we have IntServ estimate with D = W*r+k*MTU
  57         and C = MTU*r. The proof (if correct at all) is trivial.
  58
  59
  60         --- It seems that cbq-2.0 is not very accurate. At least, I cannot
  61         interpret some places, which look like wrong translations
  62         from NS. Anyone is advised to find these differences
  63         and explain to me, why I am wrong 8).
  64
  65         --- Linux has no EOI event, so that we cannot estimate true class
  66         idle time. Workaround is to consider the next dequeue event
  67         as sign that previous packet is finished. This is wrong because of
  68         internal device queueing, but on a permanently loaded link it is true.
  69         Moreover, combined with clock integrator, this scheme looks
  70         very close to an ideal solution.  */
  71
  72 struct cbq_sched_data;
  73
  74
  75 struct cbq_class {
  76         struct Qdisc_class_common common;
  77         struct cbq_class        *next_alive;    /* next class with backlog in this priority band */
  78
  79 /* Parameters */
  80         unsigned char           priority;       /* class priority */
  81         unsigned char           priority2;      /* priority to be used after overlimit */
  82         unsigned char           ewma_log;       /* time constant for idle time calculation */
  83         unsigned char           ovl_strategy;
  84 #ifdef CONFIG_NET_CLS_ACT
  85         unsigned char           police;
  86 #endif
  87
  88         u32                     defmap;
  89
  90         /* Link-sharing scheduler parameters */
  91         long                    maxidle;        /* Class parameters: see below. */
  92         long                    offtime;
  93         long                    minidle;
  94         u32                     avpkt;
  95         struct qdisc_rate_table *R_tab;
  96
  97         /* Overlimit strategy parameters */
  98         void                    (*overlimit)(struct cbq_class *cl);
  99         psched_tdiff_t          penalty;
 100
 101         /* General scheduler (WRR) parameters */
 102         long                    allot;
 103         long                    quantum;        /* Allotment per WRR round */
 104         long                    weight;         /* Relative allotment: see below */
 105
 106         struct Qdisc            *qdisc;         /* Ptr to CBQ discipline */
 107         struct cbq_class        *split;         /* Ptr to split node */
 108         struct cbq_class        *share;         /* Ptr to LS parent in the class tree */
 109         struct cbq_class        *tparent;       /* Ptr to tree parent in the class tree */
 110         struct cbq_class        *borrow;        /* NULL if class is bandwidth limited;
 111                                                    parent otherwise */
 112         struct cbq_class        *sibling;       /* Sibling chain */
 113         struct cbq_class        *children;      /* Pointer to children chain */
 114
 115         struct Qdisc            *q;             /* Elementary queueing discipline */
 116
 117
 118 /* Variables */
 119         unsigned char           cpriority;      /* Effective priority */
 120         unsigned char           delayed;
 121         unsigned char           level;          /* level of the class in hierarchy:
 122                                                    0 for leaf classes, and maximal
 123                                                    level of children + 1 for nodes.
 124                                                  */
 125
 126         psched_time_t           last;           /* Last end of service */
 127         psched_time_t           undertime;
 128         long                    avgidle;
 129         long                    deficit;        /* Saved deficit for WRR */
 130         psched_time_t           penalized;
 131         struct gnet_stats_basic_packed bstats;
 132         struct gnet_stats_queue qstats;
 133         struct gnet_stats_rate_est rate_est;
 134         struct tc_cbq_xstats    xstats;
 135
 136         struct tcf_proto        *filter_list;
 137
 138         int                     refcnt;
 139         int                     filters;
 140
 141         struct cbq_class        *defaults[TC_PRIO_MAX + 1];
 142 };
 143
 144 struct cbq_sched_data {
 145         struct Qdisc_class_hash clhash;                 /* Hash table of all classes */
 146         int                     nclasses[TC_CBQ_MAXPRIO + 1];
 147         unsigned int            quanta[TC_CBQ_MAXPRIO + 1];
 148
 149         struct cbq_class        link;
 150
 151         unsigned int            activemask;
 152         struct cbq_class        *active[TC_CBQ_MAXPRIO + 1];    /* List of all classes
 153                                                                    with backlog */
 154
 155 #ifdef CONFIG_NET_CLS_ACT
 156         struct cbq_class        *rx_class;
 157 #endif
 158         struct cbq_class        *tx_class;
 159         struct cbq_class        *tx_borrowed;
 160         int                     tx_len;
 161         psched_time_t           now;            /* Cached timestamp */
 162         psched_time_t           now_rt;         /* Cached real time */
 163         unsigned int            pmask;
 164
 165         struct hrtimer          delay_timer;
 166         struct qdisc_watchdog   watchdog;       /* Watchdog timer,
 167                                                    started when CBQ has
 168                                                    backlog, but cannot
 169                                                    transmit just now */
 170         psched_tdiff_t          wd_expires;
 171         int                     toplevel;
 172         u32                     hgenerator;
 173 };
 174
 175
 176 #define L2T(cl, len)    qdisc_l2t((cl)->R_tab, len)
 177
 178 static inline struct cbq_class *
 179 cbq_class_lookup(struct cbq_sched_data *q, u32 classid)
 180 {
 181         struct Qdisc_class_common *clc;
 182
 183         clc = qdisc_class_find(&q->clhash, classid);
 184         if (clc == NULL)
 185                 return NULL;
 186         return container_of(clc, struct cbq_class, common);
 187 }
 188
 189 #ifdef CONFIG_NET_CLS_ACT
 190
 191 static struct cbq_class *
 192 cbq_reclassify(struct sk_buff *skb, struct cbq_class *this)
 193 {
 194         struct cbq_class *cl;
 195
 196         for (cl = this->tparent; cl; cl = cl->tparent) {
 197                 struct cbq_class *new = cl->defaults[TC_PRIO_BESTEFFORT];
 198
 199                 if (new != NULL && new != this)
 200                         return new;
 201         }
 202         return NULL;
 203 }
 204
 205 #endif
 206
 207 /* Classify packet. The procedure is pretty complicated, but
 208  * it allows us to combine link sharing and priority scheduling
 209  * transparently.
 210  *
 211  * Namely, you can put link sharing rules (f.e. route based) at root of CBQ,
 212  * so that it resolves to split nodes. Then packets are classified
 213  * by logical priority, or a more specific classifier may be attached
 214  * to the split node.
 215  */
 216
 217 static struct cbq_class *
 218 cbq_classify(struct sk_buff *skb, struct Qdisc *sch, int *qerr)
 219 {
 220         struct cbq_sched_data *q = qdisc_priv(sch);
 221         struct cbq_class *head = &q->link;
 222         struct cbq_class **defmap;
 223         struct cbq_class *cl = NULL;
 224         u32 prio = skb->priority;
 225         struct tcf_result res;
 226
 227         /*
 228          *  Step 1. If skb->priority points to one of our classes, use it.
 229          */
 230         if (TC_H_MAJ(prio ^ sch->handle) == 0 &&
 231             (cl = cbq_class_lookup(q, prio)) != NULL)
 232                 return cl;
 233
 234         *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
 235         for (;;) {
 236                 int result = 0;
 237                 defmap = head->defaults;
 238
 239                 /*
 240                  * Step 2+n. Apply classifier.
 241                  */
 242                 if (!head->filter_list ||
 243                     (result = tc_classify_compat(skb, head->filter_list, &res)) < 0)
 244                         goto fallback;
 245
 246                 cl = (void *)res.class;
 247                 if (!cl) {
 248                         if (TC_H_MAJ(res.classid))
 249                                 cl = cbq_class_lookup(q, res.classid);
 250                         else if ((cl = defmap[res.classid & TC_PRIO_MAX]) == NULL)
 251                                 cl = defmap[TC_PRIO_BESTEFFORT];
 252
 253                         if (cl == NULL)
 254                                 goto fallback;
 255                 }
 256                 if (cl->level >= head->level)
 257                         goto fallback;
 258 #ifdef CONFIG_NET_CLS_ACT
 259                 switch (result) {
 260                 case TC_ACT_QUEUED:
 261                 case TC_ACT_STOLEN:
 262                         *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
 263                 case TC_ACT_SHOT:
 264                         return NULL;
 265                 case TC_ACT_RECLASSIFY:
 266                         return cbq_reclassify(skb, cl);
 267                 }
 268 #endif
 269                 if (cl->level == 0)
 270                         return cl;
 271
 272                 /*
 273                  * Step 3+n. If classifier selected a link sharing class,
 274                  *         apply agency specific classifier.
 275                  *         Repeat this procdure until we hit a leaf node.
 276                  */
 277                 head = cl;
 278         }
 279
 280 fallback:
 281         cl = head;
 282
 283         /*
 284          * Step 4. No success...
 285          */
 286         if (TC_H_MAJ(prio) == 0 &&
 287             !(cl = head->defaults[prio & TC_PRIO_MAX]) &&
 288             !(cl = head->defaults[TC_PRIO_BESTEFFORT]))
 289                 return head;
 290
 291         return cl;
 292 }
 293
 294 /*
 295  * A packet has just been enqueued on the empty class.
 296  * cbq_activate_class adds it to the tail of active class list
 297  * of its priority band.
 298  */
 299
 300 static inline void cbq_activate_class(struct cbq_class *cl)
 301 {
 302         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 303         int prio = cl->cpriority;
 304         struct cbq_class *cl_tail;
 305
 306         cl_tail = q->active[prio];
 307         q->active[prio] = cl;
 308
 309         if (cl_tail != NULL) {
 310                 cl->next_alive = cl_tail->next_alive;
 311                 cl_tail->next_alive = cl;
 312         } else {
 313                 cl->next_alive = cl;
 314                 q->activemask |= (1<<prio);
 315         }
 316 }
 317
 318 /*
 319  * Unlink class from active chain.
 320  * Note that this same procedure is done directly in cbq_dequeue*
 321  * during round-robin procedure.
 322  */
 323
 324 static void cbq_deactivate_class(struct cbq_class *this)
 325 {
 326         struct cbq_sched_data *q = qdisc_priv(this->qdisc);
 327         int prio = this->cpriority;
 328         struct cbq_class *cl;
 329         struct cbq_class *cl_prev = q->active[prio];
 330
 331         do {
 332                 cl = cl_prev->next_alive;
 333                 if (cl == this) {
 334                         cl_prev->next_alive = cl->next_alive;
 335                         cl->next_alive = NULL;
 336
 337                         if (cl == q->active[prio]) {
 338                                 q->active[prio] = cl_prev;
 339                                 if (cl == q->active[prio]) {
 340                                         q->active[prio] = NULL;
 341                                         q->activemask &= ~(1<<prio);
 342                                         return;
 343                                 }
 344                         }
 345                         return;
 346                 }
 347         } while ((cl_prev = cl) != q->active[prio]);
 348 }
 349
 350 static void
 351 cbq_mark_toplevel(struct cbq_sched_data *q, struct cbq_class *cl)
 352 {
 353         int toplevel = q->toplevel;
 354
 355         if (toplevel > cl->level && !(qdisc_is_throttled(cl->q))) {
 356                 psched_time_t now;
 357                 psched_tdiff_t incr;
 358
 359                 now = psched_get_time();
 360                 incr = now - q->now_rt;
 361                 now = q->now + incr;
 362
 363                 do {
 364                         if (cl->undertime < now) {
 365                                 q->toplevel = cl->level;
 366                                 return;
 367                         }
 368                 } while ((cl = cl->borrow) != NULL && toplevel > cl->level);
 369         }
 370 }
 371
 372 static int
 373 cbq_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 374 {
 375         struct cbq_sched_data *q = qdisc_priv(sch);
 376         int uninitialized_var(ret);
 377         struct cbq_class *cl = cbq_classify(skb, sch, &ret);
 378
 379 #ifdef CONFIG_NET_CLS_ACT
 380         q->rx_class = cl;
 381 #endif
 382         if (cl == NULL) {
 383                 if (ret & __NET_XMIT_BYPASS)
 384                         sch->qstats.drops++;
 385                 kfree_skb(skb);
 386                 return ret;
 387         }
 388
 389 #ifdef CONFIG_NET_CLS_ACT
 390         cl->q->__parent = sch;
 391 #endif
 392         ret = qdisc_enqueue(skb, cl->q);
 393         if (ret == NET_XMIT_SUCCESS) {
 394                 sch->q.qlen++;
 395                 cbq_mark_toplevel(q, cl);
 396                 if (!cl->next_alive)
 397                         cbq_activate_class(cl);
 398                 return ret;
 399         }
 400
 401         if (net_xmit_drop_count(ret)) {
 402                 sch->qstats.drops++;
 403                 cbq_mark_toplevel(q, cl);
 404                 cl->qstats.drops++;
 405         }
 406         return ret;
 407 }
 408
 409 /* Overlimit actions */
 410
 411 /* TC_CBQ_OVL_CLASSIC: (default) penalize leaf class by adding offtime */
 412
 413 static void cbq_ovl_classic(struct cbq_class *cl)
 414 {
 415         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 416         psched_tdiff_t delay = cl->undertime - q->now;
 417
 418         if (!cl->delayed) {
 419                 delay += cl->offtime;
 420
 421                 /*
 422                  * Class goes to sleep, so that it will have no
 423                  * chance to work avgidle. Let's forgive it 8)
 424                  *
 425                  * BTW cbq-2.0 has a crap in this
 426                  * place, apparently they forgot to shift it by cl->ewma_log.
 427                  */
 428                 if (cl->avgidle < 0)
 429                         delay -= (-cl->avgidle) - ((-cl->avgidle) >> cl->ewma_log);
 430                 if (cl->avgidle < cl->minidle)
 431                         cl->avgidle = cl->minidle;
 432                 if (delay <= 0)
 433                         delay = 1;
 434                 cl->undertime = q->now + delay;
 435
 436                 cl->xstats.overactions++;
 437                 cl->delayed = 1;
 438         }
 439         if (q->wd_expires == 0 || q->wd_expires > delay)
 440                 q->wd_expires = delay;
 441
 442         /* Dirty work! We must schedule wakeups based on
 443          * real available rate, rather than leaf rate,
 444          * which may be tiny (even zero).
 445          */
 446         if (q->toplevel == TC_CBQ_MAXLEVEL) {
 447                 struct cbq_class *b;
 448                 psched_tdiff_t base_delay = q->wd_expires;
 449
 450                 for (b = cl->borrow; b; b = b->borrow) {
 451                         delay = b->undertime - q->now;
 452                         if (delay < base_delay) {
 453                                 if (delay <= 0)
 454                                         delay = 1;
 455                                 base_delay = delay;
 456                         }
 457                 }
 458
 459                 q->wd_expires = base_delay;
 460         }
 461 }
 462
 463 /* TC_CBQ_OVL_RCLASSIC: penalize by offtime classes in hierarchy, when
 464  * they go overlimit
 465  */
 466
 467 static void cbq_ovl_rclassic(struct cbq_class *cl)
 468 {
 469         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 470         struct cbq_class *this = cl;
 471
 472         do {
 473                 if (cl->level > q->toplevel) {
 474                         cl = NULL;
 475                         break;
 476                 }
 477         } while ((cl = cl->borrow) != NULL);
 478
 479         if (cl == NULL)
 480                 cl = this;
 481         cbq_ovl_classic(cl);
 482 }
 483
 484 /* TC_CBQ_OVL_DELAY: delay until it will go to underlimit */
 485
 486 static void cbq_ovl_delay(struct cbq_class *cl)
 487 {
 488         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 489         psched_tdiff_t delay = cl->undertime - q->now;
 490
 491         if (test_bit(__QDISC_STATE_DEACTIVATED,
 492                      &qdisc_root_sleeping(cl->qdisc)->state))
 493                 return;
 494
 495         if (!cl->delayed) {
 496                 psched_time_t sched = q->now;
 497                 ktime_t expires;
 498
 499                 delay += cl->offtime;
 500                 if (cl->avgidle < 0)
 501                         delay -= (-cl->avgidle) - ((-cl->avgidle) >> cl->ewma_log);
 502                 if (cl->avgidle < cl->minidle)
 503                         cl->avgidle = cl->minidle;
 504                 cl->undertime = q->now + delay;
 505
 506                 if (delay > 0) {
 507                         sched += delay + cl->penalty;
 508                         cl->penalized = sched;
 509                         cl->cpriority = TC_CBQ_MAXPRIO;
 510                         q->pmask |= (1<<TC_CBQ_MAXPRIO);
 511
 512                         expires = ktime_set(0, 0);
 513                         expires = ktime_add_ns(expires, PSCHED_TICKS2NS(sched));
 514                         if (hrtimer_try_to_cancel(&q->delay_timer) &&
 515                             ktime_to_ns(ktime_sub(
 516                                         hrtimer_get_expires(&q->delay_timer),
 517                                         expires)) > 0)
 518                                 hrtimer_set_expires(&q->delay_timer, expires);
 519                         hrtimer_restart(&q->delay_timer);
 520                         cl->delayed = 1;
 521                         cl->xstats.overactions++;
 522                         return;
 523                 }
 524                 delay = 1;
 525         }
 526         if (q->wd_expires == 0 || q->wd_expires > delay)
 527                 q->wd_expires = delay;
 528 }
 529
 530 /* TC_CBQ_OVL_LOWPRIO: penalize class by lowering its priority band */
 531
 532 static void cbq_ovl_lowprio(struct cbq_class *cl)
 533 {
 534         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 535
 536         cl->penalized = q->now + cl->penalty;
 537
 538         if (cl->cpriority != cl->priority2) {
 539                 cl->cpriority = cl->priority2;
 540                 q->pmask |= (1<<cl->cpriority);
 541                 cl->xstats.overactions++;
 542         }
 543         cbq_ovl_classic(cl);
 544 }
 545
 546 /* TC_CBQ_OVL_DROP: penalize class by dropping */
 547
 548 static void cbq_ovl_drop(struct cbq_class *cl)
 549 {
 550         if (cl->q->ops->drop)
 551                 if (cl->q->ops->drop(cl->q))
 552                         cl->qdisc->q.qlen--;
 553         cl->xstats.overactions++;
 554         cbq_ovl_classic(cl);
 555 }
 556
 557 static psched_tdiff_t cbq_undelay_prio(struct cbq_sched_data *q, int prio,
 558                                        psched_time_t now)
 559 {
 560         struct cbq_class *cl;
 561         struct cbq_class *cl_prev = q->active[prio];
 562         psched_time_t sched = now;
 563
 564         if (cl_prev == NULL)
 565                 return 0;
 566
 567         do {
 568                 cl = cl_prev->next_alive;
 569                 if (now - cl->penalized > 0) {
 570                         cl_prev->next_alive = cl->next_alive;
 571                         cl->next_alive = NULL;
 572                         cl->cpriority = cl->priority;
 573                         cl->delayed = 0;
 574                         cbq_activate_class(cl);
 575
 576                         if (cl == q->active[prio]) {
 577                                 q->active[prio] = cl_prev;
 578                                 if (cl == q->active[prio]) {
 579                                         q->active[prio] = NULL;
 580                                         return 0;
 581                                 }
 582                         }
 583
 584                         cl = cl_prev->next_alive;
 585                 } else if (sched - cl->penalized > 0)
 586                         sched = cl->penalized;
 587         } while ((cl_prev = cl) != q->active[prio]);
 588
 589         return sched - now;
 590 }
 591
 592 static enum hrtimer_restart cbq_undelay(struct hrtimer *timer)
 593 {
 594         struct cbq_sched_data *q = container_of(timer, struct cbq_sched_data,
 595                                                 delay_timer);
 596         struct Qdisc *sch = q->watchdog.qdisc;
 597         psched_time_t now;
 598         psched_tdiff_t delay = 0;
 599         unsigned int pmask;
 600
 601         now = psched_get_time();
 602
 603         pmask = q->pmask;
 604         q->pmask = 0;
 605
 606         while (pmask) {
 607                 int prio = ffz(~pmask);
 608                 psched_tdiff_t tmp;
 609
 610                 pmask &= ~(1<<prio);
 611
 612                 tmp = cbq_undelay_prio(q, prio, now);
 613                 if (tmp > 0) {
 614                         q->pmask |= 1<<prio;
 615                         if (tmp < delay || delay == 0)
 616                                 delay = tmp;
 617                 }
 618         }
 619
 620         if (delay) {
 621                 ktime_t time;
 622
 623                 time = ktime_set(0, 0);
 624                 time = ktime_add_ns(time, PSCHED_TICKS2NS(now + delay));
 625                 hrtimer_start(&q->delay_timer, time, HRTIMER_MODE_ABS);
 626         }
 627
 628         qdisc_unthrottled(sch);
 629         __netif_schedule(qdisc_root(sch));
 630         return HRTIMER_NORESTART;
 631 }
 632
 633 #ifdef CONFIG_NET_CLS_ACT
 634 static int cbq_reshape_fail(struct sk_buff *skb, struct Qdisc *child)
 635 {
 636         struct Qdisc *sch = child->__parent;
 637         struct cbq_sched_data *q = qdisc_priv(sch);
 638         struct cbq_class *cl = q->rx_class;
 639
 640         q->rx_class = NULL;
 641
 642         if (cl && (cl = cbq_reclassify(skb, cl)) != NULL) {
 643                 int ret;
 644
 645                 cbq_mark_toplevel(q, cl);
 646
 647                 q->rx_class = cl;
 648                 cl->q->__parent = sch;
 649
 650                 ret = qdisc_enqueue(skb, cl->q);
 651                 if (ret == NET_XMIT_SUCCESS) {
 652                         sch->q.qlen++;
 653                         if (!cl->next_alive)
 654                                 cbq_activate_class(cl);
 655                         return 0;
 656                 }
 657                 if (net_xmit_drop_count(ret))
 658                         sch->qstats.drops++;
 659                 return 0;
 660         }
 661
 662         sch->qstats.drops++;
 663         return -1;
 664 }
 665 #endif
 666
 667 /*
 668  * It is mission critical procedure.
 669  *
 670  * We "regenerate" toplevel cutoff, if transmitting class
 671  * has backlog and it is not regulated. It is not part of
 672  * original CBQ description, but looks more reasonable.
 673  * Probably, it is wrong. This question needs further investigation.
 674  */
 675
 676 static inline void
 677 cbq_update_toplevel(struct cbq_sched_data *q, struct cbq_class *cl,
 678                     struct cbq_class *borrowed)
 679 {
 680         if (cl && q->toplevel >= borrowed->level) {
 681                 if (cl->q->q.qlen > 1) {
 682                         do {
 683                                 if (borrowed->undertime == PSCHED_PASTPERFECT) {
 684                                         q->toplevel = borrowed->level;
 685                                         return;
 686                                 }
 687                         } while ((borrowed = borrowed->borrow) != NULL);
 688                 }
 689 #if 0
 690         /* It is not necessary now. Uncommenting it
 691            will save CPU cycles, but decrease fairness.
 692          */
 693                 q->toplevel = TC_CBQ_MAXLEVEL;
 694 #endif
 695         }
 696 }
 697
 698 static void
 699 cbq_update(struct cbq_sched_data *q)
 700 {
 701         struct cbq_class *this = q->tx_class;
 702         struct cbq_class *cl = this;
 703         int len = q->tx_len;
 704
 705         q->tx_class = NULL;
 706
 707         for ( ; cl; cl = cl->share) {
 708                 long avgidle = cl->avgidle;
 709                 long idle;
 710
 711                 cl->bstats.packets++;
 712                 cl->bstats.bytes += len;
 713
 714                 /*
 715                  * (now - last) is total time between packet right edges.
 716                  * (last_pktlen/rate) is "virtual" busy time, so that
 717                  *
 718                  *      idle = (now - last) - last_pktlen/rate
 719                  */
 720
 721                 idle = q->now - cl->last;
 722                 if ((unsigned long)idle > 128*1024*1024) {
 723                         avgidle = cl->maxidle;
 724                 } else {
 725                         idle -= L2T(cl, len);
 726
 727                 /* true_avgidle := (1-W)*true_avgidle + W*idle,
 728                  * where W=2^{-ewma_log}. But cl->avgidle is scaled:
 729                  * cl->avgidle == true_avgidle/W,
 730                  * hence:
 731                  */
 732                         avgidle += idle - (avgidle>>cl->ewma_log);
 733                 }
 734
 735                 if (avgidle <= 0) {
 736                         /* Overlimit or at-limit */
 737
 738                         if (avgidle < cl->minidle)
 739                                 avgidle = cl->minidle;
 740
 741                         cl->avgidle = avgidle;
 742
 743                         /* Calculate expected time, when this class
 744                          * will be allowed to send.
 745                          * It will occur, when:
 746                          * (1-W)*true_avgidle + W*delay = 0, i.e.
 747                          * idle = (1/W - 1)*(-true_avgidle)
 748                          * or
 749                          * idle = (1 - W)*(-cl->avgidle);
 750                          */
 751                         idle = (-avgidle) - ((-avgidle) >> cl->ewma_log);
 752
 753                         /*
 754                          * That is not all.
 755                          * To maintain the rate allocated to the class,
 756                          * we add to undertime virtual clock,
 757                          * necessary to complete transmitted packet.
 758                          * (len/phys_bandwidth has been already passed
 759                          * to the moment of cbq_update)
 760                          */
 761
 762                         idle -= L2T(&q->link, len);
 763                         idle += L2T(cl, len);
 764
 765                         cl->undertime = q->now + idle;
 766                 } else {
 767                         /* Underlimit */
 768
 769                         cl->undertime = PSCHED_PASTPERFECT;
 770                         if (avgidle > cl->maxidle)
 771                                 cl->avgidle = cl->maxidle;
 772                         else
 773                                 cl->avgidle = avgidle;
 774                 }
 775                 cl->last = q->now;
 776         }
 777
 778         cbq_update_toplevel(q, this, q->tx_borrowed);
 779 }
 780
 781 static inline struct cbq_class *
 782 cbq_under_limit(struct cbq_class *cl)
 783 {
 784         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
 785         struct cbq_class *this_cl = cl;
 786
 787         if (cl->tparent == NULL)
 788                 return cl;
 789
 790         if (cl->undertime == PSCHED_PASTPERFECT || q->now >= cl->undertime) {
 791                 cl->delayed = 0;
 792                 return cl;
 793         }
 794
 795         do {
 796                 /* It is very suspicious place. Now overlimit
 797                  * action is generated for not bounded classes
 798                  * only if link is completely congested.
 799                  * Though it is in agree with ancestor-only paradigm,
 800                  * it looks very stupid. Particularly,
 801                  * it means that this chunk of code will either
 802                  * never be called or result in strong amplification
 803                  * of burstiness. Dangerous, silly, and, however,
 804                  * no another solution exists.
 805                  */
 806                 cl = cl->borrow;
 807                 if (!cl) {
 808                         this_cl->qstats.overlimits++;
 809                         this_cl->overlimit(this_cl);
 810                         return NULL;
 811                 }
 812                 if (cl->level > q->toplevel)
 813                         return NULL;
 814         } while (cl->undertime != PSCHED_PASTPERFECT && q->now < cl->undertime);
 815
 816         cl->delayed = 0;
 817         return cl;
 818 }
 819
 820 static inline struct sk_buff *
 821 cbq_dequeue_prio(struct Qdisc *sch, int prio)
 822 {
 823         struct cbq_sched_data *q = qdisc_priv(sch);
 824         struct cbq_class *cl_tail, *cl_prev, *cl;
 825         struct sk_buff *skb;
 826         int deficit;
 827
 828         cl_tail = cl_prev = q->active[prio];
 829         cl = cl_prev->next_alive;
 830
 831         do {
 832                 deficit = 0;
 833
 834                 /* Start round */
 835                 do {
 836                         struct cbq_class *borrow = cl;
 837
 838                         if (cl->q->q.qlen &&
 839                             (borrow = cbq_under_limit(cl)) == NULL)
 840                                 goto skip_class;
 841
 842                         if (cl->deficit <= 0) {
 843                                 /* Class exhausted its allotment per
 844                                  * this round. Switch to the next one.
 845                                  */
 846                                 deficit = 1;
 847                                 cl->deficit += cl->quantum;
 848                                 goto next_class;
 849                         }
 850
 851                         skb = cl->q->dequeue(cl->q);
 852
 853                         /* Class did not give us any skb :-(
 854                          * It could occur even if cl->q->q.qlen != 0
 855                          * f.e. if cl->q == "tbf"
 856                          */
 857                         if (skb == NULL)
 858                                 goto skip_class;
 859
 860                         cl->deficit -= qdisc_pkt_len(skb);
 861                         q->tx_class = cl;
 862                         q->tx_borrowed = borrow;
 863                         if (borrow != cl) {
 864 #ifndef CBQ_XSTATS_BORROWS_BYTES
 865                                 borrow->xstats.borrows++;
 866                                 cl->xstats.borrows++;
 867 #else
 868                                 borrow->xstats.borrows += qdisc_pkt_len(skb);
 869                                 cl->xstats.borrows += qdisc_pkt_len(skb);
 870 #endif
 871                         }
 872                         q->tx_len = qdisc_pkt_len(skb);
 873
 874                         if (cl->deficit <= 0) {
 875                                 q->active[prio] = cl;
 876                                 cl = cl->next_alive;
 877                                 cl->deficit += cl->quantum;
 878                         }
 879                         return skb;
 880
 881 skip_class:
 882                         if (cl->q->q.qlen == 0 || prio != cl->cpriority) {
 883                                 /* Class is empty or penalized.
 884                                  * Unlink it from active chain.
 885                                  */
 886                                 cl_prev->next_alive = cl->next_alive;
 887                                 cl->next_alive = NULL;
 888
 889                                 /* Did cl_tail point to it? */
 890                                 if (cl == cl_tail) {
 891                                         /* Repair it! */
 892                                         cl_tail = cl_prev;
 893
 894                                         /* Was it the last class in this band? */
 895                                         if (cl == cl_tail) {
 896                                                 /* Kill the band! */
 897                                                 q->active[prio] = NULL;
 898                                                 q->activemask &= ~(1<<prio);
 899                                                 if (cl->q->q.qlen)
 900                                                         cbq_activate_class(cl);
 901                                                 return NULL;
 902                                         }
 903
 904                                         q->active[prio] = cl_tail;
 905                                 }
 906                                 if (cl->q->q.qlen)
 907                                         cbq_activate_class(cl);
 908
 909                                 cl = cl_prev;
 910                         }
 911
 912 next_class:
 913                         cl_prev = cl;
 914                         cl = cl->next_alive;
 915                 } while (cl_prev != cl_tail);
 916         } while (deficit);
 917
 918         q->active[prio] = cl_prev;
 919
 920         return NULL;
 921 }
 922
 923 static inline struct sk_buff *
 924 cbq_dequeue_1(struct Qdisc *sch)
 925 {
 926         struct cbq_sched_data *q = qdisc_priv(sch);
 927         struct sk_buff *skb;
 928         unsigned int activemask;
 929
 930         activemask = q->activemask & 0xFF;
 931         while (activemask) {
 932                 int prio = ffz(~activemask);
 933                 activemask &= ~(1<<prio);
 934                 skb = cbq_dequeue_prio(sch, prio);
 935                 if (skb)
 936                         return skb;
 937         }
 938         return NULL;
 939 }
 940
 941 static struct sk_buff *
 942 cbq_dequeue(struct Qdisc *sch)
 943 {
 944         struct sk_buff *skb;
 945         struct cbq_sched_data *q = qdisc_priv(sch);
 946         psched_time_t now;
 947         psched_tdiff_t incr;
 948
 949         now = psched_get_time();
 950         incr = now - q->now_rt;
 951
 952         if (q->tx_class) {
 953                 psched_tdiff_t incr2;
 954                 /* Time integrator. We calculate EOS time
 955                  * by adding expected packet transmission time.
 956                  * If real time is greater, we warp artificial clock,
 957                  * so that:
 958                  *
 959                  * cbq_time = max(real_time, work);
 960                  */
 961                 incr2 = L2T(&q->link, q->tx_len);
 962                 q->now += incr2;
 963                 cbq_update(q);
 964                 if ((incr -= incr2) < 0)
 965                         incr = 0;
 966         }
 967         q->now += incr;
 968         q->now_rt = now;
 969
 970         for (;;) {
 971                 q->wd_expires = 0;
 972
 973                 skb = cbq_dequeue_1(sch);
 974                 if (skb) {
 975                         qdisc_bstats_update(sch, skb);
 976                         sch->q.qlen--;
 977                         qdisc_unthrottled(sch);
 978                         return skb;
 979                 }
 980
 981                 /* All the classes are overlimit.
 982                  *
 983                  * It is possible, if:
 984                  *
 985                  * 1. Scheduler is empty.
 986                  * 2. Toplevel cutoff inhibited borrowing.
 987                  * 3. Root class is overlimit.
 988                  *
 989                  * Reset 2d and 3d conditions and retry.
 990                  *
 991                  * Note, that NS and cbq-2.0 are buggy, peeking
 992                  * an arbitrary class is appropriate for ancestor-only
 993                  * sharing, but not for toplevel algorithm.
 994                  *
 995                  * Our version is better, but slower, because it requires
 996                  * two passes, but it is unavoidable with top-level sharing.
 997                  */
 998
 999                 if (q->toplevel == TC_CBQ_MAXLEVEL &&
1000                     q->link.undertime == PSCHED_PASTPERFECT)
1001                         break;
1002
1003                 q->toplevel = TC_CBQ_MAXLEVEL;
1004                 q->link.undertime = PSCHED_PASTPERFECT;
1005         }
1006
1007         /* No packets in scheduler or nobody wants to give them to us :-(
1008          * Sigh... start watchdog timer in the last case.
1009          */
1010
1011         if (sch->q.qlen) {
1012                 sch->qstats.overlimits++;
1013                 if (q->wd_expires)
1014                         qdisc_watchdog_schedule(&q->watchdog,
1015                                                 now + q->wd_expires);
1016         }
1017         return NULL;
1018 }
1019
1020 /* CBQ class maintanance routines */
1021
1022 static void cbq_adjust_levels(struct cbq_class *this)
1023 {
1024         if (this == NULL)
1025                 return;
1026
1027         do {
1028                 int level = 0;
1029                 struct cbq_class *cl;
1030
1031                 cl = this->children;
1032                 if (cl) {
1033                         do {
1034                                 if (cl->level > level)
1035                                         level = cl->level;
1036                         } while ((cl = cl->sibling) != this->children);
1037                 }
1038                 this->level = level + 1;
1039         } while ((this = this->tparent) != NULL);
1040 }
1041
1042 static void cbq_normalize_quanta(struct cbq_sched_data *q, int prio)
1043 {
1044         struct cbq_class *cl;
1045         struct hlist_node *n;
1046         unsigned int h;
1047
1048         if (q->quanta[prio] == 0)
1049                 return;
1050
1051         for (h = 0; h < q->clhash.hashsize; h++) {
1052                 hlist_for_each_entry(cl, n, &q->clhash.hash[h], common.hnode) {
1053                         /* BUGGGG... Beware! This expression suffer of
1054                          * arithmetic overflows!
1055                          */
1056                         if (cl->priority == prio) {
1057                                 cl->quantum = (cl->weight*cl->allot*q->nclasses[prio])/
1058                                         q->quanta[prio];
1059                         }
1060                         if (cl->quantum <= 0 || cl->quantum>32*qdisc_dev(cl->qdisc)->mtu) {
1061                                 pr_warning("CBQ: class %08x has bad quantum==%ld, repaired.\n",
1062                                            cl->common.classid, cl->quantum);
1063                                 cl->quantum = qdisc_dev(cl->qdisc)->mtu/2 + 1;
1064                         }
1065                 }
1066         }
1067 }
1068
1069 static void cbq_sync_defmap(struct cbq_class *cl)
1070 {
1071         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
1072         struct cbq_class *split = cl->split;
1073         unsigned int h;
1074         int i;
1075
1076         if (split == NULL)
1077                 return;
1078
1079         for (i = 0; i <= TC_PRIO_MAX; i++) {
1080                 if (split->defaults[i] == cl && !(cl->defmap & (1<<i)))
1081                         split->defaults[i] = NULL;
1082         }
1083
1084         for (i = 0; i <= TC_PRIO_MAX; i++) {
1085                 int level = split->level;
1086
1087                 if (split->defaults[i])
1088                         continue;
1089
1090                 for (h = 0; h < q->clhash.hashsize; h++) {
1091                         struct hlist_node *n;
1092                         struct cbq_class *c;
1093
1094                         hlist_for_each_entry(c, n, &q->clhash.hash[h],
1095                                              common.hnode) {
1096                                 if (c->split == split && c->level < level &&
1097                                     c->defmap & (1<<i)) {
1098                                         split->defaults[i] = c;
1099                                         level = c->level;
1100                                 }
1101                         }
1102                 }
1103         }
1104 }
1105
1106 static void cbq_change_defmap(struct cbq_class *cl, u32 splitid, u32 def, u32 mask)
1107 {
1108         struct cbq_class *split = NULL;
1109
1110         if (splitid == 0) {
1111                 split = cl->split;
1112                 if (!split)
1113                         return;
1114                 splitid = split->common.classid;
1115         }
1116
1117         if (split == NULL || split->common.classid != splitid) {
1118                 for (split = cl->tparent; split; split = split->tparent)
1119                         if (split->common.classid == splitid)
1120                                 break;
1121         }
1122
1123         if (split == NULL)
1124                 return;
1125
1126         if (cl->split != split) {
1127                 cl->defmap = 0;
1128                 cbq_sync_defmap(cl);
1129                 cl->split = split;
1130                 cl->defmap = def & mask;
1131         } else
1132                 cl->defmap = (cl->defmap & ~mask) | (def & mask);
1133
1134         cbq_sync_defmap(cl);
1135 }
1136
1137 static void cbq_unlink_class(struct cbq_class *this)
1138 {
1139         struct cbq_class *cl, **clp;
1140         struct cbq_sched_data *q = qdisc_priv(this->qdisc);
1141
1142         qdisc_class_hash_remove(&q->clhash, &this->common);
1143
1144         if (this->tparent) {
1145                 clp = &this->sibling;
1146                 cl = *clp;
1147                 do {
1148                         if (cl == this) {
1149                                 *clp = cl->sibling;
1150                                 break;
1151                         }
1152                         clp = &cl->sibling;
1153                 } while ((cl = *clp) != this->sibling);
1154
1155                 if (this->tparent->children == this) {
1156                         this->tparent->children = this->sibling;
1157                         if (this->sibling == this)
1158                                 this->tparent->children = NULL;
1159                 }
1160         } else {
1161                 WARN_ON(this->sibling != this);
1162         }
1163 }
1164
1165 static void cbq_link_class(struct cbq_class *this)
1166 {
1167         struct cbq_sched_data *q = qdisc_priv(this->qdisc);
1168         struct cbq_class *parent = this->tparent;
1169
1170         this->sibling = this;
1171         qdisc_class_hash_insert(&q->clhash, &this->common);
1172
1173         if (parent == NULL)
1174                 return;
1175
1176         if (parent->children == NULL) {
1177                 parent->children = this;
1178         } else {
1179                 this->sibling = parent->children->sibling;
1180                 parent->children->sibling = this;
1181         }
1182 }
1183
1184 static unsigned int cbq_drop(struct Qdisc *sch)
1185 {
1186         struct cbq_sched_data *q = qdisc_priv(sch);
1187         struct cbq_class *cl, *cl_head;
1188         int prio;
1189         unsigned int len;
1190
1191         for (prio = TC_CBQ_MAXPRIO; prio >= 0; prio--) {
1192                 cl_head = q->active[prio];
1193                 if (!cl_head)
1194                         continue;
1195
1196                 cl = cl_head;
1197                 do {
1198                         if (cl->q->ops->drop && (len = cl->q->ops->drop(cl->q))) {
1199                                 sch->q.qlen--;
1200                                 if (!cl->q->q.qlen)
1201                                         cbq_deactivate_class(cl);
1202                                 return len;
1203                         }
1204                 } while ((cl = cl->next_alive) != cl_head);
1205         }
1206         return 0;
1207 }
1208
1209 static void
1210 cbq_reset(struct Qdisc *sch)
1211 {
1212         struct cbq_sched_data *q = qdisc_priv(sch);
1213         struct cbq_class *cl;
1214         struct hlist_node *n;
1215         int prio;
1216         unsigned int h;
1217
1218         q->activemask = 0;
1219         q->pmask = 0;
1220         q->tx_class = NULL;
1221         q->tx_borrowed = NULL;
1222         qdisc_watchdog_cancel(&q->watchdog);
1223         hrtimer_cancel(&q->delay_timer);
1224         q->toplevel = TC_CBQ_MAXLEVEL;
1225         q->now = psched_get_time();
1226         q->now_rt = q->now;
1227
1228         for (prio = 0; prio <= TC_CBQ_MAXPRIO; prio++)
1229                 q->active[prio] = NULL;
1230
1231         for (h = 0; h < q->clhash.hashsize; h++) {
1232                 hlist_for_each_entry(cl, n, &q->clhash.hash[h], common.hnode) {
1233                         qdisc_reset(cl->q);
1234
1235                         cl->next_alive = NULL;
1236                         cl->undertime = PSCHED_PASTPERFECT;
1237                         cl->avgidle = cl->maxidle;
1238                         cl->deficit = cl->quantum;
1239                         cl->cpriority = cl->priority;
1240                 }
1241         }
1242         sch->q.qlen = 0;
1243 }
1244
1245
1246 static int cbq_set_lss(struct cbq_class *cl, struct tc_cbq_lssopt *lss)
1247 {
1248         if (lss->change & TCF_CBQ_LSS_FLAGS) {
1249                 cl->share = (lss->flags & TCF_CBQ_LSS_ISOLATED) ? NULL : cl->tparent;
1250                 cl->borrow = (lss->flags & TCF_CBQ_LSS_BOUNDED) ? NULL : cl->tparent;
1251         }
1252         if (lss->change & TCF_CBQ_LSS_EWMA)
1253                 cl->ewma_log = lss->ewma_log;
1254         if (lss->change & TCF_CBQ_LSS_AVPKT)
1255                 cl->avpkt = lss->avpkt;
1256         if (lss->change & TCF_CBQ_LSS_MINIDLE)
1257                 cl->minidle = -(long)lss->minidle;
1258         if (lss->change & TCF_CBQ_LSS_MAXIDLE) {
1259                 cl->maxidle = lss->maxidle;
1260                 cl->avgidle = lss->maxidle;
1261         }
1262         if (lss->change & TCF_CBQ_LSS_OFFTIME)
1263                 cl->offtime = lss->offtime;
1264         return 0;
1265 }
1266
1267 static void cbq_rmprio(struct cbq_sched_data *q, struct cbq_class *cl)
1268 {
1269         q->nclasses[cl->priority]--;
1270         q->quanta[cl->priority] -= cl->weight;
1271         cbq_normalize_quanta(q, cl->priority);
1272 }
1273
1274 static void cbq_addprio(struct cbq_sched_data *q, struct cbq_class *cl)
1275 {
1276         q->nclasses[cl->priority]++;
1277         q->quanta[cl->priority] += cl->weight;
1278         cbq_normalize_quanta(q, cl->priority);
1279 }
1280
1281 static int cbq_set_wrr(struct cbq_class *cl, struct tc_cbq_wrropt *wrr)
1282 {
1283         struct cbq_sched_data *q = qdisc_priv(cl->qdisc);
1284
1285         if (wrr->allot)
1286                 cl->allot = wrr->allot;
1287         if (wrr->weight)
1288                 cl->weight = wrr->weight;
1289         if (wrr->priority) {
1290                 cl->priority = wrr->priority - 1;
1291                 cl->cpriority = cl->priority;
1292                 if (cl->priority >= cl->priority2)
1293                         cl->priority2 = TC_CBQ_MAXPRIO - 1;
1294         }
1295
1296         cbq_addprio(q, cl);
1297         return 0;
1298 }
1299
1300 static int cbq_set_overlimit(struct cbq_class *cl, struct tc_cbq_ovl *ovl)
1301 {
1302         switch (ovl->strategy) {
1303         case TC_CBQ_OVL_CLASSIC:
1304                 cl->overlimit = cbq_ovl_classic;
1305                 break;
1306         case TC_CBQ_OVL_DELAY:
1307                 cl->overlimit = cbq_ovl_delay;
1308                 break;
1309         case TC_CBQ_OVL_LOWPRIO:
1310                 if (ovl->priority2 - 1 >= TC_CBQ_MAXPRIO ||
1311                     ovl->priority2 - 1 <= cl->priority)
1312                         return -EINVAL;
1313                 cl->priority2 = ovl->priority2 - 1;
1314                 cl->overlimit = cbq_ovl_lowprio;
1315                 break;
1316         case TC_CBQ_OVL_DROP:
1317                 cl->overlimit = cbq_ovl_drop;
1318                 break;
1319         case TC_CBQ_OVL_RCLASSIC:
1320                 cl->overlimit = cbq_ovl_rclassic;
1321                 break;
1322         default:
1323                 return -EINVAL;
1324         }
1325         cl->penalty = ovl->penalty;
1326         return 0;
1327 }
1328
1329 #ifdef CONFIG_NET_CLS_ACT
1330 static int cbq_set_police(struct cbq_class *cl, struct tc_cbq_police *p)
1331 {
1332         cl->police = p->police;
1333
1334         if (cl->q->handle) {
1335                 if (p->police == TC_POLICE_RECLASSIFY)
1336                         cl->q->reshape_fail = cbq_reshape_fail;
1337                 else
1338                         cl->q->reshape_fail = NULL;
1339         }
1340         return 0;
1341 }
1342 #endif
1343
1344 static int cbq_set_fopt(struct cbq_class *cl, struct tc_cbq_fopt *fopt)
1345 {
1346         cbq_change_defmap(cl, fopt->split, fopt->defmap, fopt->defchange);
1347         return 0;
1348 }
1349
1350 static const struct nla_policy cbq_policy[TCA_CBQ_MAX + 1] = {
1351         [TCA_CBQ_LSSOPT]        = { .len = sizeof(struct tc_cbq_lssopt) },
1352         [TCA_CBQ_WRROPT]        = { .len = sizeof(struct tc_cbq_wrropt) },
1353         [TCA_CBQ_FOPT]          = { .len = sizeof(struct tc_cbq_fopt) },
1354         [TCA_CBQ_OVL_STRATEGY]  = { .len = sizeof(struct tc_cbq_ovl) },
1355         [TCA_CBQ_RATE]          = { .len = sizeof(struct tc_ratespec) },
1356         [TCA_CBQ_RTAB]          = { .type = NLA_BINARY, .len = TC_RTAB_SIZE },
1357         [TCA_CBQ_POLICE]        = { .len = sizeof(struct tc_cbq_police) },
1358 };
1359
1360 static int cbq_init(struct Qdisc *sch, struct nlattr *opt)
1361 {
1362         struct cbq_sched_data *q = qdisc_priv(sch);
1363         struct nlattr *tb[TCA_CBQ_MAX + 1];
1364         struct tc_ratespec *r;
1365         int err;
1366
1367         err = nla_parse_nested(tb, TCA_CBQ_MAX, opt, cbq_policy);
1368         if (err < 0)
1369                 return err;
1370
1371         if (tb[TCA_CBQ_RTAB] == NULL || tb[TCA_CBQ_RATE] == NULL)
1372                 return -EINVAL;
1373
1374         r = nla_data(tb[TCA_CBQ_RATE]);
1375
1376         if ((q->link.R_tab = qdisc_get_rtab(r, tb[TCA_CBQ_RTAB])) == NULL)
1377                 return -EINVAL;
1378
1379         err = qdisc_class_hash_init(&q->clhash);
1380         if (err < 0)
1381                 goto put_rtab;
1382
1383         q->link.refcnt = 1;
1384         q->link.sibling = &q->link;
1385         q->link.common.classid = sch->handle;
1386         q->link.qdisc = sch;
1387         q->link.q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
1388                                       sch->handle);
1389         if (!q->link.q)
1390                 q->link.q = &noop_qdisc;
1391
1392         q->link.priority = TC_CBQ_MAXPRIO - 1;
1393         q->link.priority2 = TC_CBQ_MAXPRIO - 1;
1394         q->link.cpriority = TC_CBQ_MAXPRIO - 1;
1395         q->link.ovl_strategy = TC_CBQ_OVL_CLASSIC;
1396         q->link.overlimit = cbq_ovl_classic;
1397         q->link.allot = psched_mtu(qdisc_dev(sch));
1398         q->link.quantum = q->link.allot;
1399         q->link.weight = q->link.R_tab->rate.rate;
1400
1401         q->link.ewma_log = TC_CBQ_DEF_EWMA;
1402         q->link.avpkt = q->link.allot/2;
1403         q->link.minidle = -0x7FFFFFFF;
1404
1405         qdisc_watchdog_init(&q->watchdog, sch);
1406         hrtimer_init(&q->delay_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
1407         q->delay_timer.function = cbq_undelay;
1408         q->toplevel = TC_CBQ_MAXLEVEL;
1409         q->now = psched_get_time();
1410         q->now_rt = q->now;
1411
1412         cbq_link_class(&q->link);
1413
1414         if (tb[TCA_CBQ_LSSOPT])
1415                 cbq_set_lss(&q->link, nla_data(tb[TCA_CBQ_LSSOPT]));
1416
1417         cbq_addprio(q, &q->link);
1418         return 0;
1419
1420 put_rtab:
1421         qdisc_put_rtab(q->link.R_tab);
1422         return err;
1423 }
1424
1425 static int cbq_dump_rate(struct sk_buff *skb, struct cbq_class *cl)
1426 {
1427         unsigned char *b = skb_tail_pointer(skb);
1428
1429         if (nla_put(skb, TCA_CBQ_RATE, sizeof(cl->R_tab->rate), &cl->R_tab->rate))
1430                 goto nla_put_failure;
1431         return skb->len;
1432
1433 nla_put_failure:
1434         nlmsg_trim(skb, b);
1435         return -1;
1436 }
1437
1438 static int cbq_dump_lss(struct sk_buff *skb, struct cbq_class *cl)
1439 {
1440         unsigned char *b = skb_tail_pointer(skb);
1441         struct tc_cbq_lssopt opt;
1442
1443         opt.flags = 0;
1444         if (cl->borrow == NULL)
1445                 opt.flags |= TCF_CBQ_LSS_BOUNDED;
1446         if (cl->share == NULL)
1447                 opt.flags |= TCF_CBQ_LSS_ISOLATED;
1448         opt.ewma_log = cl->ewma_log;
1449         opt.level = cl->level;
1450         opt.avpkt = cl->avpkt;
1451         opt.maxidle = cl->maxidle;
1452         opt.minidle = (u32)(-cl->minidle);
1453         opt.offtime = cl->offtime;
1454         opt.change = ~0;
1455         if (nla_put(skb, TCA_CBQ_LSSOPT, sizeof(opt), &opt))
1456                 goto nla_put_failure;
1457         return skb->len;
1458
1459 nla_put_failure:
1460         nlmsg_trim(skb, b);
1461         return -1;
1462 }
1463
1464 static int cbq_dump_wrr(struct sk_buff *skb, struct cbq_class *cl)
1465 {
1466         unsigned char *b = skb_tail_pointer(skb);
1467         struct tc_cbq_wrropt opt;
1468
1469         opt.flags = 0;
1470         opt.allot = cl->allot;
1471         opt.priority = cl->priority + 1;
1472         opt.cpriority = cl->cpriority + 1;
1473         opt.weight = cl->weight;
1474         if (nla_put(skb, TCA_CBQ_WRROPT, sizeof(opt), &opt))
1475                 goto nla_put_failure;
1476         return skb->len;
1477
1478 nla_put_failure:
1479         nlmsg_trim(skb, b);
1480         return -1;
1481 }
1482
1483 static int cbq_dump_ovl(struct sk_buff *skb, struct cbq_class *cl)
1484 {
1485         unsigned char *b = skb_tail_pointer(skb);
1486         struct tc_cbq_ovl opt;
1487
1488         opt.strategy = cl->ovl_strategy;
1489         opt.priority2 = cl->priority2 + 1;
1490         opt.pad = 0;
1491         opt.penalty = cl->penalty;
1492         if (nla_put(skb, TCA_CBQ_OVL_STRATEGY, sizeof(opt), &opt))
1493                 goto nla_put_failure;
1494         return skb->len;
1495
1496 nla_put_failure:
1497         nlmsg_trim(skb, b);
1498         return -1;
1499 }
1500
1501 static int cbq_dump_fopt(struct sk_buff *skb, struct cbq_class *cl)
1502 {
1503         unsigned char *b = skb_tail_pointer(skb);
1504         struct tc_cbq_fopt opt;
1505
1506         if (cl->split || cl->defmap) {
1507                 opt.split = cl->split ? cl->split->common.classid : 0;
1508                 opt.defmap = cl->defmap;
1509                 opt.defchange = ~0;
1510                 if (nla_put(skb, TCA_CBQ_FOPT, sizeof(opt), &opt))
1511                         goto nla_put_failure;
1512         }
1513         return skb->len;
1514
1515 nla_put_failure:
1516         nlmsg_trim(skb, b);
1517         return -1;
1518 }
1519
1520 #ifdef CONFIG_NET_CLS_ACT
1521 static int cbq_dump_police(struct sk_buff *skb, struct cbq_class *cl)
1522 {
1523         unsigned char *b = skb_tail_pointer(skb);
1524         struct tc_cbq_police opt;
1525
1526         if (cl->police) {
1527                 opt.police = cl->police;
1528                 opt.__res1 = 0;
1529                 opt.__res2 = 0;
1530                 if (nla_put(skb, TCA_CBQ_POLICE, sizeof(opt), &opt))
1531                         goto nla_put_failure;
1532         }
1533         return skb->len;
1534
1535 nla_put_failure:
1536         nlmsg_trim(skb, b);
1537         return -1;
1538 }
1539 #endif
1540
1541 static int cbq_dump_attr(struct sk_buff *skb, struct cbq_class *cl)
1542 {
1543         if (cbq_dump_lss(skb, cl) < 0 ||
1544             cbq_dump_rate(skb, cl) < 0 ||
1545             cbq_dump_wrr(skb, cl) < 0 ||
1546             cbq_dump_ovl(skb, cl) < 0 ||
1547 #ifdef CONFIG_NET_CLS_ACT
1548             cbq_dump_police(skb, cl) < 0 ||
1549 #endif
1550             cbq_dump_fopt(skb, cl) < 0)
1551                 return -1;
1552         return 0;
1553 }
1554
1555 static int cbq_dump(struct Qdisc *sch, struct sk_buff *skb)
1556 {
1557         struct cbq_sched_data *q = qdisc_priv(sch);
1558         struct nlattr *nest;
1559
1560         nest = nla_nest_start(skb, TCA_OPTIONS);
1561         if (nest == NULL)
1562                 goto nla_put_failure;
1563         if (cbq_dump_attr(skb, &q->link) < 0)
1564                 goto nla_put_failure;
1565         nla_nest_end(skb, nest);
1566         return skb->len;
1567
1568 nla_put_failure:
1569         nla_nest_cancel(skb, nest);
1570         return -1;
1571 }
1572
1573 static int
1574 cbq_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
1575 {
1576         struct cbq_sched_data *q = qdisc_priv(sch);
1577
1578         q->link.xstats.avgidle = q->link.avgidle;
1579         return gnet_stats_copy_app(d, &q->link.xstats, sizeof(q->link.xstats));
1580 }
1581
1582 static int
1583 cbq_dump_class(struct Qdisc *sch, unsigned long arg,
1584                struct sk_buff *skb, struct tcmsg *tcm)
1585 {
1586         struct cbq_class *cl = (struct cbq_class *)arg;
1587         struct nlattr *nest;
1588
1589         if (cl->tparent)
1590                 tcm->tcm_parent = cl->tparent->common.classid;
1591         else
1592                 tcm->tcm_parent = TC_H_ROOT;
1593         tcm->tcm_handle = cl->common.classid;
1594         tcm->tcm_info = cl->q->handle;
1595
1596         nest = nla_nest_start(skb, TCA_OPTIONS);
1597         if (nest == NULL)
1598                 goto nla_put_failure;
1599         if (cbq_dump_attr(skb, cl) < 0)
1600                 goto nla_put_failure;
1601         nla_nest_end(skb, nest);
1602         return skb->len;
1603
1604 nla_put_failure:
1605         nla_nest_cancel(skb, nest);
1606         return -1;
1607 }
1608
1609 static int
1610 cbq_dump_class_stats(struct Qdisc *sch, unsigned long arg,
1611         struct gnet_dump *d)
1612 {
1613         struct cbq_sched_data *q = qdisc_priv(sch);
1614         struct cbq_class *cl = (struct cbq_class *)arg;
1615
1616         cl->qstats.qlen = cl->q->q.qlen;
1617         cl->xstats.avgidle = cl->avgidle;
1618         cl->xstats.undertime = 0;
1619
1620         if (cl->undertime != PSCHED_PASTPERFECT)
1621                 cl->xstats.undertime = cl->undertime - q->now;
1622
1623         if (gnet_stats_copy_basic(d, &cl->bstats) < 0 ||
1624             gnet_stats_copy_rate_est(d, &cl->bstats, &cl->rate_est) < 0 ||
1625             gnet_stats_copy_queue(d, &cl->qstats) < 0)
1626                 return -1;
1627
1628         return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats));
1629 }
1630
1631 static int cbq_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
1632                      struct Qdisc **old)
1633 {
1634         struct cbq_class *cl = (struct cbq_class *)arg;
1635
1636         if (new == NULL) {
1637                 new = qdisc_create_dflt(sch->dev_queue,
1638                                         &pfifo_qdisc_ops, cl->common.classid);
1639                 if (new == NULL)
1640                         return -ENOBUFS;
1641         } else {
1642 #ifdef CONFIG_NET_CLS_ACT
1643                 if (cl->police == TC_POLICE_RECLASSIFY)
1644                         new->reshape_fail = cbq_reshape_fail;
1645 #endif
1646         }
1647         sch_tree_lock(sch);
1648         *old = cl->q;
1649         cl->q = new;
1650         qdisc_tree_decrease_qlen(*old, (*old)->q.qlen);
1651         qdisc_reset(*old);
1652         sch_tree_unlock(sch);
1653
1654         return 0;
1655 }
1656
1657 static struct Qdisc *cbq_leaf(struct Qdisc *sch, unsigned long arg)
1658 {
1659         struct cbq_class *cl = (struct cbq_class *)arg;
1660
1661         return cl->q;
1662 }
1663
1664 static void cbq_qlen_notify(struct Qdisc *sch, unsigned long arg)
1665 {
1666         struct cbq_class *cl = (struct cbq_class *)arg;
1667
1668         if (cl->q->q.qlen == 0)
1669                 cbq_deactivate_class(cl);
1670 }
1671
1672 static unsigned long cbq_get(struct Qdisc *sch, u32 classid)
1673 {
1674         struct cbq_sched_data *q = qdisc_priv(sch);
1675         struct cbq_class *cl = cbq_class_lookup(q, classid);
1676
1677         if (cl) {
1678                 cl->refcnt++;
1679                 return (unsigned long)cl;
1680         }
1681         return 0;
1682 }
1683
1684 static void cbq_destroy_class(struct Qdisc *sch, struct cbq_class *cl)
1685 {
1686         struct cbq_sched_data *q = qdisc_priv(sch);
1687
1688         WARN_ON(cl->filters);
1689
1690         tcf_destroy_chain(&cl->filter_list);
1691         qdisc_destroy(cl->q);
1692         qdisc_put_rtab(cl->R_tab);
1693         gen_kill_estimator(&cl->bstats, &cl->rate_est);
1694         if (cl != &q->link)
1695                 kfree(cl);
1696 }
1697
1698 static void cbq_destroy(struct Qdisc *sch)
1699 {
1700         struct cbq_sched_data *q = qdisc_priv(sch);
1701         struct hlist_node *n, *next;
1702         struct cbq_class *cl;
1703         unsigned int h;
1704
1705 #ifdef CONFIG_NET_CLS_ACT
1706         q->rx_class = NULL;
1707 #endif
1708         /*
1709          * Filters must be destroyed first because we don't destroy the
1710          * classes from root to leafs which means that filters can still
1711          * be bound to classes which have been destroyed already. --TGR '04
1712          */
1713         for (h = 0; h < q->clhash.hashsize; h++) {
1714                 hlist_for_each_entry(cl, n, &q->clhash.hash[h], common.hnode)
1715                         tcf_destroy_chain(&cl->filter_list);
1716         }
1717         for (h = 0; h < q->clhash.hashsize; h++) {
1718                 hlist_for_each_entry_safe(cl, n, next, &q->clhash.hash[h],
1719                                           common.hnode)
1720                         cbq_destroy_class(sch, cl);
1721         }
1722         qdisc_class_hash_destroy(&q->clhash);
1723 }
1724
1725 static void cbq_put(struct Qdisc *sch, unsigned long arg)
1726 {
1727         struct cbq_class *cl = (struct cbq_class *)arg;
1728
1729         if (--cl->refcnt == 0) {
1730 #ifdef CONFIG_NET_CLS_ACT
1731                 spinlock_t *root_lock = qdisc_root_sleeping_lock(sch);
1732                 struct cbq_sched_data *q = qdisc_priv(sch);
1733
1734                 spin_lock_bh(root_lock);
1735                 if (q->rx_class == cl)
1736                         q->rx_class = NULL;
1737                 spin_unlock_bh(root_lock);
1738 #endif
1739
1740                 cbq_destroy_class(sch, cl);
1741         }
1742 }
1743
1744 static int
1745 cbq_change_class(struct Qdisc *sch, u32 classid, u32 parentid, struct nlattr **tca,
1746                  unsigned long *arg)
1747 {
1748         int err;
1749         struct cbq_sched_data *q = qdisc_priv(sch);
1750         struct cbq_class *cl = (struct cbq_class *)*arg;
1751         struct nlattr *opt = tca[TCA_OPTIONS];
1752         struct nlattr *tb[TCA_CBQ_MAX + 1];
1753         struct cbq_class *parent;
1754         struct qdisc_rate_table *rtab = NULL;
1755
1756         if (opt == NULL)
1757                 return -EINVAL;
1758
1759         err = nla_parse_nested(tb, TCA_CBQ_MAX, opt, cbq_policy);
1760         if (err < 0)
1761                 return err;
1762
1763         if (cl) {
1764                 /* Check parent */
1765                 if (parentid) {
1766                         if (cl->tparent &&
1767                             cl->tparent->common.classid != parentid)
1768                                 return -EINVAL;
1769                         if (!cl->tparent && parentid != TC_H_ROOT)
1770                                 return -EINVAL;
1771                 }
1772
1773                 if (tb[TCA_CBQ_RATE]) {
1774                         rtab = qdisc_get_rtab(nla_data(tb[TCA_CBQ_RATE]),
1775                                               tb[TCA_CBQ_RTAB]);
1776                         if (rtab == NULL)
1777                                 return -EINVAL;
1778                 }
1779
1780                 if (tca[TCA_RATE]) {
1781                         err = gen_replace_estimator(&cl->bstats, &cl->rate_est,
1782                                                     qdisc_root_sleeping_lock(sch),
1783                                                     tca[TCA_RATE]);
1784                         if (err) {
1785                                 if (rtab)
1786                                         qdisc_put_rtab(rtab);
1787                                 return err;
1788                         }
1789                 }
1790
1791                 /* Change class parameters */
1792                 sch_tree_lock(sch);
1793
1794                 if (cl->next_alive != NULL)
1795                         cbq_deactivate_class(cl);
1796
1797                 if (rtab) {
1798                         qdisc_put_rtab(cl->R_tab);
1799                         cl->R_tab = rtab;
1800                 }
1801
1802                 if (tb[TCA_CBQ_LSSOPT])
1803                         cbq_set_lss(cl, nla_data(tb[TCA_CBQ_LSSOPT]));
1804
1805                 if (tb[TCA_CBQ_WRROPT]) {
1806                         cbq_rmprio(q, cl);
1807                         cbq_set_wrr(cl, nla_data(tb[TCA_CBQ_WRROPT]));
1808                 }
1809
1810                 if (tb[TCA_CBQ_OVL_STRATEGY])
1811                         cbq_set_overlimit(cl, nla_data(tb[TCA_CBQ_OVL_STRATEGY]));
1812
1813 #ifdef CONFIG_NET_CLS_ACT
1814                 if (tb[TCA_CBQ_POLICE])
1815                         cbq_set_police(cl, nla_data(tb[TCA_CBQ_POLICE]));
1816 #endif
1817
1818                 if (tb[TCA_CBQ_FOPT])
1819                         cbq_set_fopt(cl, nla_data(tb[TCA_CBQ_FOPT]));
1820
1821                 if (cl->q->q.qlen)
1822                         cbq_activate_class(cl);
1823
1824                 sch_tree_unlock(sch);
1825
1826                 return 0;
1827         }
1828
1829         if (parentid == TC_H_ROOT)
1830                 return -EINVAL;
1831
1832         if (tb[TCA_CBQ_WRROPT] == NULL || tb[TCA_CBQ_RATE] == NULL ||
1833             tb[TCA_CBQ_LSSOPT] == NULL)
1834                 return -EINVAL;
1835
1836         rtab = qdisc_get_rtab(nla_data(tb[TCA_CBQ_RATE]), tb[TCA_CBQ_RTAB]);
1837         if (rtab == NULL)
1838                 return -EINVAL;
1839
1840         if (classid) {
1841                 err = -EINVAL;
1842                 if (TC_H_MAJ(classid ^ sch->handle) ||
1843                     cbq_class_lookup(q, classid))
1844                         goto failure;
1845         } else {
1846                 int i;
1847                 classid = TC_H_MAKE(sch->handle, 0x8000);
1848
1849                 for (i = 0; i < 0x8000; i++) {
1850                         if (++q->hgenerator >= 0x8000)
1851                                 q->hgenerator = 1;
1852                         if (cbq_class_lookup(q, classid|q->hgenerator) == NULL)
1853                                 break;
1854                 }
1855                 err = -ENOSR;
1856                 if (i >= 0x8000)
1857                         goto failure;
1858                 classid = classid|q->hgenerator;
1859         }
1860
1861         parent = &q->link;
1862         if (parentid) {
1863                 parent = cbq_class_lookup(q, parentid);
1864                 err = -EINVAL;
1865                 if (parent == NULL)
1866                         goto failure;
1867         }
1868
1869         err = -ENOBUFS;
1870         cl = kzalloc(sizeof(*cl), GFP_KERNEL);
1871         if (cl == NULL)
1872                 goto failure;
1873
1874         if (tca[TCA_RATE]) {
1875                 err = gen_new_estimator(&cl->bstats, &cl->rate_est,
1876                                         qdisc_root_sleeping_lock(sch),
1877                                         tca[TCA_RATE]);
1878                 if (err) {
1879                         kfree(cl);
1880                         goto failure;
1881                 }
1882         }
1883
1884         cl->R_tab = rtab;
1885         rtab = NULL;
1886         cl->refcnt = 1;
1887         cl->q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, classid);
1888         if (!cl->q)
1889                 cl->q = &noop_qdisc;
1890         cl->common.classid = classid;
1891         cl->tparent = parent;
1892         cl->qdisc = sch;
1893         cl->allot = parent->allot;
1894         cl->quantum = cl->allot;
1895         cl->weight = cl->R_tab->rate.rate;
1896
1897         sch_tree_lock(sch);
1898         cbq_link_class(cl);
1899         cl->borrow = cl->tparent;
1900         if (cl->tparent != &q->link)
1901                 cl->share = cl->tparent;
1902         cbq_adjust_levels(parent);
1903         cl->minidle = -0x7FFFFFFF;
1904         cbq_set_lss(cl, nla_data(tb[TCA_CBQ_LSSOPT]));
1905         cbq_set_wrr(cl, nla_data(tb[TCA_CBQ_WRROPT]));
1906         if (cl->ewma_log == 0)
1907                 cl->ewma_log = q->link.ewma_log;
1908         if (cl->maxidle == 0)
1909                 cl->maxidle = q->link.maxidle;
1910         if (cl->avpkt == 0)
1911                 cl->avpkt = q->link.avpkt;
1912         cl->overlimit = cbq_ovl_classic;
1913         if (tb[TCA_CBQ_OVL_STRATEGY])
1914                 cbq_set_overlimit(cl, nla_data(tb[TCA_CBQ_OVL_STRATEGY]));
1915 #ifdef CONFIG_NET_CLS_ACT
1916         if (tb[TCA_CBQ_POLICE])
1917                 cbq_set_police(cl, nla_data(tb[TCA_CBQ_POLICE]));
1918 #endif
1919         if (tb[TCA_CBQ_FOPT])
1920                 cbq_set_fopt(cl, nla_data(tb[TCA_CBQ_FOPT]));
1921         sch_tree_unlock(sch);
1922
1923         qdisc_class_hash_grow(sch, &q->clhash);
1924
1925         *arg = (unsigned long)cl;
1926         return 0;
1927
1928 failure:
1929         qdisc_put_rtab(rtab);
1930         return err;
1931 }
1932
1933 static int cbq_delete(struct Qdisc *sch, unsigned long arg)
1934 {
1935         struct cbq_sched_data *q = qdisc_priv(sch);
1936         struct cbq_class *cl = (struct cbq_class *)arg;
1937         unsigned int qlen;
1938
1939         if (cl->filters || cl->children || cl == &q->link)
1940                 return -EBUSY;
1941
1942         sch_tree_lock(sch);
1943
1944         qlen = cl->q->q.qlen;
1945         qdisc_reset(cl->q);
1946         qdisc_tree_decrease_qlen(cl->q, qlen);
1947
1948         if (cl->next_alive)
1949                 cbq_deactivate_class(cl);
1950
1951         if (q->tx_borrowed == cl)
1952                 q->tx_borrowed = q->tx_class;
1953         if (q->tx_class == cl) {
1954                 q->tx_class = NULL;
1955                 q->tx_borrowed = NULL;
1956         }
1957 #ifdef CONFIG_NET_CLS_ACT
1958         if (q->rx_class == cl)
1959                 q->rx_class = NULL;
1960 #endif
1961
1962         cbq_unlink_class(cl);
1963         cbq_adjust_levels(cl->tparent);
1964         cl->defmap = 0;
1965         cbq_sync_defmap(cl);
1966
1967         cbq_rmprio(q, cl);
1968         sch_tree_unlock(sch);
1969
1970         BUG_ON(--cl->refcnt == 0);
1971         /*
1972          * This shouldn't happen: we "hold" one cops->get() when called
1973          * from tc_ctl_tclass; the destroy method is done from cops->put().
1974          */
1975
1976         return 0;
1977 }
1978
1979 static struct tcf_proto **cbq_find_tcf(struct Qdisc *sch, unsigned long arg)
1980 {
1981         struct cbq_sched_data *q = qdisc_priv(sch);
1982         struct cbq_class *cl = (struct cbq_class *)arg;
1983
1984         if (cl == NULL)
1985                 cl = &q->link;
1986
1987         return &cl->filter_list;
1988 }
1989
1990 static unsigned long cbq_bind_filter(struct Qdisc *sch, unsigned long parent,
1991                                      u32 classid)
1992 {
1993         struct cbq_sched_data *q = qdisc_priv(sch);
1994         struct cbq_class *p = (struct cbq_class *)parent;
1995         struct cbq_class *cl = cbq_class_lookup(q, classid);
1996
1997         if (cl) {
1998                 if (p && p->level <= cl->level)
1999                         return 0;
2000                 cl->filters++;
2001                 return (unsigned long)cl;
2002         }
2003         return 0;
2004 }
2005
2006 static void cbq_unbind_filter(struct Qdisc *sch, unsigned long arg)
2007 {
2008         struct cbq_class *cl = (struct cbq_class *)arg;
2009
2010         cl->filters--;
2011 }
2012
2013 static void cbq_walk(struct Qdisc *sch, struct qdisc_walker *arg)
2014 {
2015         struct cbq_sched_data *q = qdisc_priv(sch);
2016         struct cbq_class *cl;
2017         struct hlist_node *n;
2018         unsigned int h;
2019
2020         if (arg->stop)
2021                 return;
2022
2023         for (h = 0; h < q->clhash.hashsize; h++) {
2024                 hlist_for_each_entry(cl, n, &q->clhash.hash[h], common.hnode) {
2025                         if (arg->count < arg->skip) {
2026                                 arg->count++;
2027                                 continue;
2028                         }
2029                         if (arg->fn(sch, (unsigned long)cl, arg) < 0) {
2030                                 arg->stop = 1;
2031                                 return;
2032                         }
2033                         arg->count++;
2034                 }
2035         }
2036 }
2037
2038 static const struct Qdisc_class_ops cbq_class_ops = {
2039         .graft          =       cbq_graft,
2040         .leaf           =       cbq_leaf,
2041         .qlen_notify    =       cbq_qlen_notify,
2042         .get            =       cbq_get,
2043         .put            =       cbq_put,
2044         .change         =       cbq_change_class,
2045         .delete         =       cbq_delete,
2046         .walk           =       cbq_walk,
2047         .tcf_chain      =       cbq_find_tcf,
2048         .bind_tcf       =       cbq_bind_filter,
2049         .unbind_tcf     =       cbq_unbind_filter,
2050         .dump           =       cbq_dump_class,
2051         .dump_stats     =       cbq_dump_class_stats,
2052 };
2053
2054 static struct Qdisc_ops cbq_qdisc_ops __read_mostly = {
2055         .next           =       NULL,
2056         .cl_ops         =       &cbq_class_ops,
2057         .id             =       "cbq",
2058         .priv_size      =       sizeof(struct cbq_sched_data),
2059         .enqueue        =       cbq_enqueue,
2060         .dequeue        =       cbq_dequeue,
2061         .peek           =       qdisc_peek_dequeued,
2062         .drop           =       cbq_drop,
2063         .init           =       cbq_init,
2064         .reset          =       cbq_reset,
2065         .destroy        =       cbq_destroy,
2066         .change         =       NULL,
2067         .dump           =       cbq_dump,
2068         .dump_stats     =       cbq_dump_stats,
2069         .owner          =       THIS_MODULE,
2070 };
2071
2072 static int __init cbq_module_init(void)
2073 {
2074         return register_qdisc(&cbq_qdisc_ops);
2075 }
2076 static void __exit cbq_module_exit(void)
2077 {
2078         unregister_qdisc(&cbq_qdisc_ops);
2079 }
2080 module_init(cbq_module_init)
2081 module_exit(cbq_module_exit)
2082 MODULE_LICENSE("GPL");