1 /* 2 * net/sched/sch_htb.c Hierarchical token bucket, feed tree version 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: Martin Devera, <devik@cdi.cz> 10 * 11 * Credits (in time order) for older HTB versions: 12 * Stef Coene <stef.coene@docum.org> 13 * HTB support at LARTC mailing list 14 * Ondrej Kraus, <krauso@barr.cz> 15 * found missing INIT_QDISC(htb) 16 * Vladimir Smelhaus, Aamer Akhter, Bert Hubert 17 * helped a lot to locate nasty class stall bug 18 * Andi Kleen, Jamal Hadi, Bert Hubert 19 * code review and helpful comments on shaping 20 * Tomasz Wrona, <tw@eter.tym.pl> 21 * created test case so that I was able to fix nasty bug 22 * Wilfried Weissmann 23 * spotted bug in dequeue code and helped with fix 24 * Jiri Fojtasek 25 * fixed requeue routine 26 * and many others. thanks. 27 */ 28 #include <linux/module.h> 29 #include <linux/moduleparam.h> 30 #include <linux/types.h> 31 #include <linux/kernel.h> 32 #include <linux/string.h> 33 #include <linux/errno.h> 34 #include <linux/skbuff.h> 35 #include <linux/list.h> 36 #include <linux/compiler.h> 37 #include <linux/rbtree.h> 38 #include <linux/workqueue.h> 39 #include <linux/slab.h> 40 #include <net/netlink.h> 41 #include <net/sch_generic.h> 42 #include <net/pkt_sched.h> 43 #include <net/pkt_cls.h> 44 45 /* HTB algorithm. 46 Author: devik@cdi.cz 47 ======================================================================== 48 HTB is like TBF with multiple classes. It is also similar to CBQ because 49 it allows to assign priority to each class in hierarchy. 50 In fact it is another implementation of Floyd's formal sharing. 51 52 Levels: 53 Each class is assigned level. Leaf has ALWAYS level 0 and root 54 classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level 55 one less than their parent. 56 */ 57 58 static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */ 59 #define HTB_VER 0x30011 /* major must be matched with number suplied by TC as version */ 60 61 #if HTB_VER >> 16 != TC_HTB_PROTOVER 62 #error "Mismatched sch_htb.c and pkt_sch.h" 63 #endif 64 65 /* Module parameter and sysfs export */ 66 module_param (htb_hysteresis, int, 0640); 67 MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate"); 68 69 static int htb_rate_est = 0; /* htb classes have a default rate estimator */ 70 module_param(htb_rate_est, int, 0640); 71 MODULE_PARM_DESC(htb_rate_est, "setup a default rate estimator (4sec 16sec) for htb classes"); 72 73 /* used internaly to keep status of single class */ 74 enum htb_cmode { 75 HTB_CANT_SEND, /* class can't send and can't borrow */ 76 HTB_MAY_BORROW, /* class can't send but may borrow */ 77 HTB_CAN_SEND /* class can send */ 78 }; 79 80 struct htb_prio { 81 union { 82 struct rb_root row; 83 struct rb_root feed; 84 }; 85 struct rb_node *ptr; 86 /* When class changes from state 1->2 and disconnects from 87 * parent's feed then we lost ptr value and start from the 88 * first child again. Here we store classid of the 89 * last valid ptr (used when ptr is NULL). 90 */ 91 u32 last_ptr_id; 92 }; 93 94 /* interior & leaf nodes; props specific to leaves are marked L: 95 * To reduce false sharing, place mostly read fields at beginning, 96 * and mostly written ones at the end. 97 */ 98 struct htb_class { 99 struct Qdisc_class_common common; 100 struct psched_ratecfg rate; 101 struct psched_ratecfg ceil; 102 s64 buffer, cbuffer;/* token bucket depth/rate */ 103 s64 mbuffer; /* max wait time */ 104 u32 prio; /* these two are used only by leaves... */ 105 int quantum; /* but stored for parent-to-leaf return */ 106 107 struct tcf_proto __rcu *filter_list; /* class attached filters */ 108 int filter_cnt; 109 int refcnt; /* usage count of this class */ 110 111 int level; /* our level (see above) */ 112 unsigned int children; 113 struct htb_class *parent; /* parent class */ 114 115 struct net_rate_estimator __rcu *rate_est; 116 117 /* 118 * Written often fields 119 */ 120 struct gnet_stats_basic_packed bstats; 121 struct tc_htb_xstats xstats; /* our special stats */ 122 123 /* token bucket parameters */ 124 s64 tokens, ctokens;/* current number of tokens */ 125 s64 t_c; /* checkpoint time */ 126 127 union { 128 struct htb_class_leaf { 129 struct list_head drop_list; 130 int deficit[TC_HTB_MAXDEPTH]; 131 struct Qdisc *q; 132 } leaf; 133 struct htb_class_inner { 134 struct htb_prio clprio[TC_HTB_NUMPRIO]; 135 } inner; 136 } un; 137 s64 pq_key; 138 139 int prio_activity; /* for which prios are we active */ 140 enum htb_cmode cmode; /* current mode of the class */ 141 struct rb_node pq_node; /* node for event queue */ 142 struct rb_node node[TC_HTB_NUMPRIO]; /* node for self or feed tree */ 143 144 unsigned int drops ____cacheline_aligned_in_smp; 145 }; 146 147 struct htb_level { 148 struct rb_root wait_pq; 149 struct htb_prio hprio[TC_HTB_NUMPRIO]; 150 }; 151 152 struct htb_sched { 153 struct Qdisc_class_hash clhash; 154 int defcls; /* class where unclassified flows go to */ 155 int rate2quantum; /* quant = rate / rate2quantum */ 156 157 /* filters for qdisc itself */ 158 struct tcf_proto __rcu *filter_list; 159 160 #define HTB_WARN_TOOMANYEVENTS 0x1 161 unsigned int warned; /* only one warning */ 162 int direct_qlen; 163 struct work_struct work; 164 165 /* non shaped skbs; let them go directly thru */ 166 struct qdisc_skb_head direct_queue; 167 long direct_pkts; 168 169 struct qdisc_watchdog watchdog; 170 171 s64 now; /* cached dequeue time */ 172 struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */ 173 174 /* time of nearest event per level (row) */ 175 s64 near_ev_cache[TC_HTB_MAXDEPTH]; 176 177 int row_mask[TC_HTB_MAXDEPTH]; 178 179 struct htb_level hlevel[TC_HTB_MAXDEPTH]; 180 }; 181 182 /* find class in global hash table using given handle */ 183 static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch) 184 { 185 struct htb_sched *q = qdisc_priv(sch); 186 struct Qdisc_class_common *clc; 187 188 clc = qdisc_class_find(&q->clhash, handle); 189 if (clc == NULL) 190 return NULL; 191 return container_of(clc, struct htb_class, common); 192 } 193 194 /** 195 * htb_classify - classify a packet into class 196 * 197 * It returns NULL if the packet should be dropped or -1 if the packet 198 * should be passed directly thru. In all other cases leaf class is returned. 199 * We allow direct class selection by classid in priority. The we examine 200 * filters in qdisc and in inner nodes (if higher filter points to the inner 201 * node). If we end up with classid MAJOR:0 we enqueue the skb into special 202 * internal fifo (direct). These packets then go directly thru. If we still 203 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessful 204 * then finish and return direct queue. 205 */ 206 #define HTB_DIRECT ((struct htb_class *)-1L) 207 208 static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch, 209 int *qerr) 210 { 211 struct htb_sched *q = qdisc_priv(sch); 212 struct htb_class *cl; 213 struct tcf_result res; 214 struct tcf_proto *tcf; 215 int result; 216 217 /* allow to select class by setting skb->priority to valid classid; 218 * note that nfmark can be used too by attaching filter fw with no 219 * rules in it 220 */ 221 if (skb->priority == sch->handle) 222 return HTB_DIRECT; /* X:0 (direct flow) selected */ 223 cl = htb_find(skb->priority, sch); 224 if (cl) { 225 if (cl->level == 0) 226 return cl; 227 /* Start with inner filter chain if a non-leaf class is selected */ 228 tcf = rcu_dereference_bh(cl->filter_list); 229 } else { 230 tcf = rcu_dereference_bh(q->filter_list); 231 } 232 233 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 234 while (tcf && (result = tc_classify(skb, tcf, &res, false)) >= 0) { 235 #ifdef CONFIG_NET_CLS_ACT 236 switch (result) { 237 case TC_ACT_QUEUED: 238 case TC_ACT_STOLEN: 239 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; 240 case TC_ACT_SHOT: 241 return NULL; 242 } 243 #endif 244 cl = (void *)res.class; 245 if (!cl) { 246 if (res.classid == sch->handle) 247 return HTB_DIRECT; /* X:0 (direct flow) */ 248 cl = htb_find(res.classid, sch); 249 if (!cl) 250 break; /* filter selected invalid classid */ 251 } 252 if (!cl->level) 253 return cl; /* we hit leaf; return it */ 254 255 /* we have got inner class; apply inner filter chain */ 256 tcf = rcu_dereference_bh(cl->filter_list); 257 } 258 /* classification failed; try to use default class */ 259 cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch); 260 if (!cl || cl->level) 261 return HTB_DIRECT; /* bad default .. this is safe bet */ 262 return cl; 263 } 264 265 /** 266 * htb_add_to_id_tree - adds class to the round robin list 267 * 268 * Routine adds class to the list (actually tree) sorted by classid. 269 * Make sure that class is not already on such list for given prio. 270 */ 271 static void htb_add_to_id_tree(struct rb_root *root, 272 struct htb_class *cl, int prio) 273 { 274 struct rb_node **p = &root->rb_node, *parent = NULL; 275 276 while (*p) { 277 struct htb_class *c; 278 parent = *p; 279 c = rb_entry(parent, struct htb_class, node[prio]); 280 281 if (cl->common.classid > c->common.classid) 282 p = &parent->rb_right; 283 else 284 p = &parent->rb_left; 285 } 286 rb_link_node(&cl->node[prio], parent, p); 287 rb_insert_color(&cl->node[prio], root); 288 } 289 290 /** 291 * htb_add_to_wait_tree - adds class to the event queue with delay 292 * 293 * The class is added to priority event queue to indicate that class will 294 * change its mode in cl->pq_key microseconds. Make sure that class is not 295 * already in the queue. 296 */ 297 static void htb_add_to_wait_tree(struct htb_sched *q, 298 struct htb_class *cl, s64 delay) 299 { 300 struct rb_node **p = &q->hlevel[cl->level].wait_pq.rb_node, *parent = NULL; 301 302 cl->pq_key = q->now + delay; 303 if (cl->pq_key == q->now) 304 cl->pq_key++; 305 306 /* update the nearest event cache */ 307 if (q->near_ev_cache[cl->level] > cl->pq_key) 308 q->near_ev_cache[cl->level] = cl->pq_key; 309 310 while (*p) { 311 struct htb_class *c; 312 parent = *p; 313 c = rb_entry(parent, struct htb_class, pq_node); 314 if (cl->pq_key >= c->pq_key) 315 p = &parent->rb_right; 316 else 317 p = &parent->rb_left; 318 } 319 rb_link_node(&cl->pq_node, parent, p); 320 rb_insert_color(&cl->pq_node, &q->hlevel[cl->level].wait_pq); 321 } 322 323 /** 324 * htb_next_rb_node - finds next node in binary tree 325 * 326 * When we are past last key we return NULL. 327 * Average complexity is 2 steps per call. 328 */ 329 static inline void htb_next_rb_node(struct rb_node **n) 330 { 331 *n = rb_next(*n); 332 } 333 334 /** 335 * htb_add_class_to_row - add class to its row 336 * 337 * The class is added to row at priorities marked in mask. 338 * It does nothing if mask == 0. 339 */ 340 static inline void htb_add_class_to_row(struct htb_sched *q, 341 struct htb_class *cl, int mask) 342 { 343 q->row_mask[cl->level] |= mask; 344 while (mask) { 345 int prio = ffz(~mask); 346 mask &= ~(1 << prio); 347 htb_add_to_id_tree(&q->hlevel[cl->level].hprio[prio].row, cl, prio); 348 } 349 } 350 351 /* If this triggers, it is a bug in this code, but it need not be fatal */ 352 static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root) 353 { 354 if (RB_EMPTY_NODE(rb)) { 355 WARN_ON(1); 356 } else { 357 rb_erase(rb, root); 358 RB_CLEAR_NODE(rb); 359 } 360 } 361 362 363 /** 364 * htb_remove_class_from_row - removes class from its row 365 * 366 * The class is removed from row at priorities marked in mask. 367 * It does nothing if mask == 0. 368 */ 369 static inline void htb_remove_class_from_row(struct htb_sched *q, 370 struct htb_class *cl, int mask) 371 { 372 int m = 0; 373 struct htb_level *hlevel = &q->hlevel[cl->level]; 374 375 while (mask) { 376 int prio = ffz(~mask); 377 struct htb_prio *hprio = &hlevel->hprio[prio]; 378 379 mask &= ~(1 << prio); 380 if (hprio->ptr == cl->node + prio) 381 htb_next_rb_node(&hprio->ptr); 382 383 htb_safe_rb_erase(cl->node + prio, &hprio->row); 384 if (!hprio->row.rb_node) 385 m |= 1 << prio; 386 } 387 q->row_mask[cl->level] &= ~m; 388 } 389 390 /** 391 * htb_activate_prios - creates active classe's feed chain 392 * 393 * The class is connected to ancestors and/or appropriate rows 394 * for priorities it is participating on. cl->cmode must be new 395 * (activated) mode. It does nothing if cl->prio_activity == 0. 396 */ 397 static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl) 398 { 399 struct htb_class *p = cl->parent; 400 long m, mask = cl->prio_activity; 401 402 while (cl->cmode == HTB_MAY_BORROW && p && mask) { 403 m = mask; 404 while (m) { 405 int prio = ffz(~m); 406 m &= ~(1 << prio); 407 408 if (p->un.inner.clprio[prio].feed.rb_node) 409 /* parent already has its feed in use so that 410 * reset bit in mask as parent is already ok 411 */ 412 mask &= ~(1 << prio); 413 414 htb_add_to_id_tree(&p->un.inner.clprio[prio].feed, cl, prio); 415 } 416 p->prio_activity |= mask; 417 cl = p; 418 p = cl->parent; 419 420 } 421 if (cl->cmode == HTB_CAN_SEND && mask) 422 htb_add_class_to_row(q, cl, mask); 423 } 424 425 /** 426 * htb_deactivate_prios - remove class from feed chain 427 * 428 * cl->cmode must represent old mode (before deactivation). It does 429 * nothing if cl->prio_activity == 0. Class is removed from all feed 430 * chains and rows. 431 */ 432 static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl) 433 { 434 struct htb_class *p = cl->parent; 435 long m, mask = cl->prio_activity; 436 437 while (cl->cmode == HTB_MAY_BORROW && p && mask) { 438 m = mask; 439 mask = 0; 440 while (m) { 441 int prio = ffz(~m); 442 m &= ~(1 << prio); 443 444 if (p->un.inner.clprio[prio].ptr == cl->node + prio) { 445 /* we are removing child which is pointed to from 446 * parent feed - forget the pointer but remember 447 * classid 448 */ 449 p->un.inner.clprio[prio].last_ptr_id = cl->common.classid; 450 p->un.inner.clprio[prio].ptr = NULL; 451 } 452 453 htb_safe_rb_erase(cl->node + prio, 454 &p->un.inner.clprio[prio].feed); 455 456 if (!p->un.inner.clprio[prio].feed.rb_node) 457 mask |= 1 << prio; 458 } 459 460 p->prio_activity &= ~mask; 461 cl = p; 462 p = cl->parent; 463 464 } 465 if (cl->cmode == HTB_CAN_SEND && mask) 466 htb_remove_class_from_row(q, cl, mask); 467 } 468 469 static inline s64 htb_lowater(const struct htb_class *cl) 470 { 471 if (htb_hysteresis) 472 return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0; 473 else 474 return 0; 475 } 476 static inline s64 htb_hiwater(const struct htb_class *cl) 477 { 478 if (htb_hysteresis) 479 return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0; 480 else 481 return 0; 482 } 483 484 485 /** 486 * htb_class_mode - computes and returns current class mode 487 * 488 * It computes cl's mode at time cl->t_c+diff and returns it. If mode 489 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference 490 * from now to time when cl will change its state. 491 * Also it is worth to note that class mode doesn't change simply 492 * at cl->{c,}tokens == 0 but there can rather be hysteresis of 493 * 0 .. -cl->{c,}buffer range. It is meant to limit number of 494 * mode transitions per time unit. The speed gain is about 1/6. 495 */ 496 static inline enum htb_cmode 497 htb_class_mode(struct htb_class *cl, s64 *diff) 498 { 499 s64 toks; 500 501 if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) { 502 *diff = -toks; 503 return HTB_CANT_SEND; 504 } 505 506 if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl)) 507 return HTB_CAN_SEND; 508 509 *diff = -toks; 510 return HTB_MAY_BORROW; 511 } 512 513 /** 514 * htb_change_class_mode - changes classe's mode 515 * 516 * This should be the only way how to change classe's mode under normal 517 * cirsumstances. Routine will update feed lists linkage, change mode 518 * and add class to the wait event queue if appropriate. New mode should 519 * be different from old one and cl->pq_key has to be valid if changing 520 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree). 521 */ 522 static void 523 htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, s64 *diff) 524 { 525 enum htb_cmode new_mode = htb_class_mode(cl, diff); 526 527 if (new_mode == cl->cmode) 528 return; 529 530 if (cl->prio_activity) { /* not necessary: speed optimization */ 531 if (cl->cmode != HTB_CANT_SEND) 532 htb_deactivate_prios(q, cl); 533 cl->cmode = new_mode; 534 if (new_mode != HTB_CANT_SEND) 535 htb_activate_prios(q, cl); 536 } else 537 cl->cmode = new_mode; 538 } 539 540 /** 541 * htb_activate - inserts leaf cl into appropriate active feeds 542 * 543 * Routine learns (new) priority of leaf and activates feed chain 544 * for the prio. It can be called on already active leaf safely. 545 * It also adds leaf into droplist. 546 */ 547 static inline void htb_activate(struct htb_sched *q, struct htb_class *cl) 548 { 549 WARN_ON(cl->level || !cl->un.leaf.q || !cl->un.leaf.q->q.qlen); 550 551 if (!cl->prio_activity) { 552 cl->prio_activity = 1 << cl->prio; 553 htb_activate_prios(q, cl); 554 list_add_tail(&cl->un.leaf.drop_list, 555 q->drops + cl->prio); 556 } 557 } 558 559 /** 560 * htb_deactivate - remove leaf cl from active feeds 561 * 562 * Make sure that leaf is active. In the other words it can't be called 563 * with non-active leaf. It also removes class from the drop list. 564 */ 565 static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl) 566 { 567 WARN_ON(!cl->prio_activity); 568 569 htb_deactivate_prios(q, cl); 570 cl->prio_activity = 0; 571 list_del_init(&cl->un.leaf.drop_list); 572 } 573 574 static void htb_enqueue_tail(struct sk_buff *skb, struct Qdisc *sch, 575 struct qdisc_skb_head *qh) 576 { 577 struct sk_buff *last = qh->tail; 578 579 if (last) { 580 skb->next = NULL; 581 last->next = skb; 582 qh->tail = skb; 583 } else { 584 qh->tail = skb; 585 qh->head = skb; 586 } 587 qh->qlen++; 588 } 589 590 static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch, 591 struct sk_buff **to_free) 592 { 593 int uninitialized_var(ret); 594 struct htb_sched *q = qdisc_priv(sch); 595 struct htb_class *cl = htb_classify(skb, sch, &ret); 596 597 if (cl == HTB_DIRECT) { 598 /* enqueue to helper queue */ 599 if (q->direct_queue.qlen < q->direct_qlen) { 600 htb_enqueue_tail(skb, sch, &q->direct_queue); 601 q->direct_pkts++; 602 } else { 603 return qdisc_drop(skb, sch, to_free); 604 } 605 #ifdef CONFIG_NET_CLS_ACT 606 } else if (!cl) { 607 if (ret & __NET_XMIT_BYPASS) 608 qdisc_qstats_drop(sch); 609 __qdisc_drop(skb, to_free); 610 return ret; 611 #endif 612 } else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q, 613 to_free)) != NET_XMIT_SUCCESS) { 614 if (net_xmit_drop_count(ret)) { 615 qdisc_qstats_drop(sch); 616 cl->drops++; 617 } 618 return ret; 619 } else { 620 htb_activate(q, cl); 621 } 622 623 qdisc_qstats_backlog_inc(sch, skb); 624 sch->q.qlen++; 625 return NET_XMIT_SUCCESS; 626 } 627 628 static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, s64 diff) 629 { 630 s64 toks = diff + cl->tokens; 631 632 if (toks > cl->buffer) 633 toks = cl->buffer; 634 toks -= (s64) psched_l2t_ns(&cl->rate, bytes); 635 if (toks <= -cl->mbuffer) 636 toks = 1 - cl->mbuffer; 637 638 cl->tokens = toks; 639 } 640 641 static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, s64 diff) 642 { 643 s64 toks = diff + cl->ctokens; 644 645 if (toks > cl->cbuffer) 646 toks = cl->cbuffer; 647 toks -= (s64) psched_l2t_ns(&cl->ceil, bytes); 648 if (toks <= -cl->mbuffer) 649 toks = 1 - cl->mbuffer; 650 651 cl->ctokens = toks; 652 } 653 654 /** 655 * htb_charge_class - charges amount "bytes" to leaf and ancestors 656 * 657 * Routine assumes that packet "bytes" long was dequeued from leaf cl 658 * borrowing from "level". It accounts bytes to ceil leaky bucket for 659 * leaf and all ancestors and to rate bucket for ancestors at levels 660 * "level" and higher. It also handles possible change of mode resulting 661 * from the update. Note that mode can also increase here (MAY_BORROW to 662 * CAN_SEND) because we can use more precise clock that event queue here. 663 * In such case we remove class from event queue first. 664 */ 665 static void htb_charge_class(struct htb_sched *q, struct htb_class *cl, 666 int level, struct sk_buff *skb) 667 { 668 int bytes = qdisc_pkt_len(skb); 669 enum htb_cmode old_mode; 670 s64 diff; 671 672 while (cl) { 673 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer); 674 if (cl->level >= level) { 675 if (cl->level == level) 676 cl->xstats.lends++; 677 htb_accnt_tokens(cl, bytes, diff); 678 } else { 679 cl->xstats.borrows++; 680 cl->tokens += diff; /* we moved t_c; update tokens */ 681 } 682 htb_accnt_ctokens(cl, bytes, diff); 683 cl->t_c = q->now; 684 685 old_mode = cl->cmode; 686 diff = 0; 687 htb_change_class_mode(q, cl, &diff); 688 if (old_mode != cl->cmode) { 689 if (old_mode != HTB_CAN_SEND) 690 htb_safe_rb_erase(&cl->pq_node, &q->hlevel[cl->level].wait_pq); 691 if (cl->cmode != HTB_CAN_SEND) 692 htb_add_to_wait_tree(q, cl, diff); 693 } 694 695 /* update basic stats except for leaves which are already updated */ 696 if (cl->level) 697 bstats_update(&cl->bstats, skb); 698 699 cl = cl->parent; 700 } 701 } 702 703 /** 704 * htb_do_events - make mode changes to classes at the level 705 * 706 * Scans event queue for pending events and applies them. Returns time of 707 * next pending event (0 for no event in pq, q->now for too many events). 708 * Note: Applied are events whose have cl->pq_key <= q->now. 709 */ 710 static s64 htb_do_events(struct htb_sched *q, const int level, 711 unsigned long start) 712 { 713 /* don't run for longer than 2 jiffies; 2 is used instead of 714 * 1 to simplify things when jiffy is going to be incremented 715 * too soon 716 */ 717 unsigned long stop_at = start + 2; 718 struct rb_root *wait_pq = &q->hlevel[level].wait_pq; 719 720 while (time_before(jiffies, stop_at)) { 721 struct htb_class *cl; 722 s64 diff; 723 struct rb_node *p = rb_first(wait_pq); 724 725 if (!p) 726 return 0; 727 728 cl = rb_entry(p, struct htb_class, pq_node); 729 if (cl->pq_key > q->now) 730 return cl->pq_key; 731 732 htb_safe_rb_erase(p, wait_pq); 733 diff = min_t(s64, q->now - cl->t_c, cl->mbuffer); 734 htb_change_class_mode(q, cl, &diff); 735 if (cl->cmode != HTB_CAN_SEND) 736 htb_add_to_wait_tree(q, cl, diff); 737 } 738 739 /* too much load - let's continue after a break for scheduling */ 740 if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) { 741 pr_warn("htb: too many events!\n"); 742 q->warned |= HTB_WARN_TOOMANYEVENTS; 743 } 744 745 return q->now; 746 } 747 748 /* Returns class->node+prio from id-tree where classe's id is >= id. NULL 749 * is no such one exists. 750 */ 751 static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n, 752 u32 id) 753 { 754 struct rb_node *r = NULL; 755 while (n) { 756 struct htb_class *cl = 757 rb_entry(n, struct htb_class, node[prio]); 758 759 if (id > cl->common.classid) { 760 n = n->rb_right; 761 } else if (id < cl->common.classid) { 762 r = n; 763 n = n->rb_left; 764 } else { 765 return n; 766 } 767 } 768 return r; 769 } 770 771 /** 772 * htb_lookup_leaf - returns next leaf class in DRR order 773 * 774 * Find leaf where current feed pointers points to. 775 */ 776 static struct htb_class *htb_lookup_leaf(struct htb_prio *hprio, const int prio) 777 { 778 int i; 779 struct { 780 struct rb_node *root; 781 struct rb_node **pptr; 782 u32 *pid; 783 } stk[TC_HTB_MAXDEPTH], *sp = stk; 784 785 BUG_ON(!hprio->row.rb_node); 786 sp->root = hprio->row.rb_node; 787 sp->pptr = &hprio->ptr; 788 sp->pid = &hprio->last_ptr_id; 789 790 for (i = 0; i < 65535; i++) { 791 if (!*sp->pptr && *sp->pid) { 792 /* ptr was invalidated but id is valid - try to recover 793 * the original or next ptr 794 */ 795 *sp->pptr = 796 htb_id_find_next_upper(prio, sp->root, *sp->pid); 797 } 798 *sp->pid = 0; /* ptr is valid now so that remove this hint as it 799 * can become out of date quickly 800 */ 801 if (!*sp->pptr) { /* we are at right end; rewind & go up */ 802 *sp->pptr = sp->root; 803 while ((*sp->pptr)->rb_left) 804 *sp->pptr = (*sp->pptr)->rb_left; 805 if (sp > stk) { 806 sp--; 807 if (!*sp->pptr) { 808 WARN_ON(1); 809 return NULL; 810 } 811 htb_next_rb_node(sp->pptr); 812 } 813 } else { 814 struct htb_class *cl; 815 struct htb_prio *clp; 816 817 cl = rb_entry(*sp->pptr, struct htb_class, node[prio]); 818 if (!cl->level) 819 return cl; 820 clp = &cl->un.inner.clprio[prio]; 821 (++sp)->root = clp->feed.rb_node; 822 sp->pptr = &clp->ptr; 823 sp->pid = &clp->last_ptr_id; 824 } 825 } 826 WARN_ON(1); 827 return NULL; 828 } 829 830 /* dequeues packet at given priority and level; call only if 831 * you are sure that there is active class at prio/level 832 */ 833 static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, const int prio, 834 const int level) 835 { 836 struct sk_buff *skb = NULL; 837 struct htb_class *cl, *start; 838 struct htb_level *hlevel = &q->hlevel[level]; 839 struct htb_prio *hprio = &hlevel->hprio[prio]; 840 841 /* look initial class up in the row */ 842 start = cl = htb_lookup_leaf(hprio, prio); 843 844 do { 845 next: 846 if (unlikely(!cl)) 847 return NULL; 848 849 /* class can be empty - it is unlikely but can be true if leaf 850 * qdisc drops packets in enqueue routine or if someone used 851 * graft operation on the leaf since last dequeue; 852 * simply deactivate and skip such class 853 */ 854 if (unlikely(cl->un.leaf.q->q.qlen == 0)) { 855 struct htb_class *next; 856 htb_deactivate(q, cl); 857 858 /* row/level might become empty */ 859 if ((q->row_mask[level] & (1 << prio)) == 0) 860 return NULL; 861 862 next = htb_lookup_leaf(hprio, prio); 863 864 if (cl == start) /* fix start if we just deleted it */ 865 start = next; 866 cl = next; 867 goto next; 868 } 869 870 skb = cl->un.leaf.q->dequeue(cl->un.leaf.q); 871 if (likely(skb != NULL)) 872 break; 873 874 qdisc_warn_nonwc("htb", cl->un.leaf.q); 875 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr: 876 &q->hlevel[0].hprio[prio].ptr); 877 cl = htb_lookup_leaf(hprio, prio); 878 879 } while (cl != start); 880 881 if (likely(skb != NULL)) { 882 bstats_update(&cl->bstats, skb); 883 cl->un.leaf.deficit[level] -= qdisc_pkt_len(skb); 884 if (cl->un.leaf.deficit[level] < 0) { 885 cl->un.leaf.deficit[level] += cl->quantum; 886 htb_next_rb_node(level ? &cl->parent->un.inner.clprio[prio].ptr : 887 &q->hlevel[0].hprio[prio].ptr); 888 } 889 /* this used to be after charge_class but this constelation 890 * gives us slightly better performance 891 */ 892 if (!cl->un.leaf.q->q.qlen) 893 htb_deactivate(q, cl); 894 htb_charge_class(q, cl, level, skb); 895 } 896 return skb; 897 } 898 899 static struct sk_buff *htb_dequeue(struct Qdisc *sch) 900 { 901 struct sk_buff *skb; 902 struct htb_sched *q = qdisc_priv(sch); 903 int level; 904 s64 next_event; 905 unsigned long start_at; 906 907 /* try to dequeue direct packets as high prio (!) to minimize cpu work */ 908 skb = __qdisc_dequeue_head(&q->direct_queue); 909 if (skb != NULL) { 910 ok: 911 qdisc_bstats_update(sch, skb); 912 qdisc_qstats_backlog_dec(sch, skb); 913 sch->q.qlen--; 914 return skb; 915 } 916 917 if (!sch->q.qlen) 918 goto fin; 919 q->now = ktime_get_ns(); 920 start_at = jiffies; 921 922 next_event = q->now + 5LLU * NSEC_PER_SEC; 923 924 for (level = 0; level < TC_HTB_MAXDEPTH; level++) { 925 /* common case optimization - skip event handler quickly */ 926 int m; 927 s64 event = q->near_ev_cache[level]; 928 929 if (q->now >= event) { 930 event = htb_do_events(q, level, start_at); 931 if (!event) 932 event = q->now + NSEC_PER_SEC; 933 q->near_ev_cache[level] = event; 934 } 935 936 if (next_event > event) 937 next_event = event; 938 939 m = ~q->row_mask[level]; 940 while (m != (int)(-1)) { 941 int prio = ffz(m); 942 943 m |= 1 << prio; 944 skb = htb_dequeue_tree(q, prio, level); 945 if (likely(skb != NULL)) 946 goto ok; 947 } 948 } 949 qdisc_qstats_overlimit(sch); 950 if (likely(next_event > q->now)) 951 qdisc_watchdog_schedule_ns(&q->watchdog, next_event); 952 else 953 schedule_work(&q->work); 954 fin: 955 return skb; 956 } 957 958 /* reset all classes */ 959 /* always caled under BH & queue lock */ 960 static void htb_reset(struct Qdisc *sch) 961 { 962 struct htb_sched *q = qdisc_priv(sch); 963 struct htb_class *cl; 964 unsigned int i; 965 966 for (i = 0; i < q->clhash.hashsize; i++) { 967 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) { 968 if (cl->level) 969 memset(&cl->un.inner, 0, sizeof(cl->un.inner)); 970 else { 971 if (cl->un.leaf.q) 972 qdisc_reset(cl->un.leaf.q); 973 INIT_LIST_HEAD(&cl->un.leaf.drop_list); 974 } 975 cl->prio_activity = 0; 976 cl->cmode = HTB_CAN_SEND; 977 } 978 } 979 qdisc_watchdog_cancel(&q->watchdog); 980 __qdisc_reset_queue(&q->direct_queue); 981 sch->q.qlen = 0; 982 sch->qstats.backlog = 0; 983 memset(q->hlevel, 0, sizeof(q->hlevel)); 984 memset(q->row_mask, 0, sizeof(q->row_mask)); 985 for (i = 0; i < TC_HTB_NUMPRIO; i++) 986 INIT_LIST_HEAD(q->drops + i); 987 } 988 989 static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = { 990 [TCA_HTB_PARMS] = { .len = sizeof(struct tc_htb_opt) }, 991 [TCA_HTB_INIT] = { .len = sizeof(struct tc_htb_glob) }, 992 [TCA_HTB_CTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 993 [TCA_HTB_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 994 [TCA_HTB_DIRECT_QLEN] = { .type = NLA_U32 }, 995 [TCA_HTB_RATE64] = { .type = NLA_U64 }, 996 [TCA_HTB_CEIL64] = { .type = NLA_U64 }, 997 }; 998 999 static void htb_work_func(struct work_struct *work) 1000 { 1001 struct htb_sched *q = container_of(work, struct htb_sched, work); 1002 struct Qdisc *sch = q->watchdog.qdisc; 1003 1004 rcu_read_lock(); 1005 __netif_schedule(qdisc_root(sch)); 1006 rcu_read_unlock(); 1007 } 1008 1009 static int htb_init(struct Qdisc *sch, struct nlattr *opt) 1010 { 1011 struct htb_sched *q = qdisc_priv(sch); 1012 struct nlattr *tb[TCA_HTB_MAX + 1]; 1013 struct tc_htb_glob *gopt; 1014 int err; 1015 int i; 1016 1017 if (!opt) 1018 return -EINVAL; 1019 1020 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy, NULL); 1021 if (err < 0) 1022 return err; 1023 1024 if (!tb[TCA_HTB_INIT]) 1025 return -EINVAL; 1026 1027 gopt = nla_data(tb[TCA_HTB_INIT]); 1028 if (gopt->version != HTB_VER >> 16) 1029 return -EINVAL; 1030 1031 err = qdisc_class_hash_init(&q->clhash); 1032 if (err < 0) 1033 return err; 1034 for (i = 0; i < TC_HTB_NUMPRIO; i++) 1035 INIT_LIST_HEAD(q->drops + i); 1036 1037 qdisc_watchdog_init(&q->watchdog, sch); 1038 INIT_WORK(&q->work, htb_work_func); 1039 qdisc_skb_head_init(&q->direct_queue); 1040 1041 if (tb[TCA_HTB_DIRECT_QLEN]) 1042 q->direct_qlen = nla_get_u32(tb[TCA_HTB_DIRECT_QLEN]); 1043 else 1044 q->direct_qlen = qdisc_dev(sch)->tx_queue_len; 1045 1046 if ((q->rate2quantum = gopt->rate2quantum) < 1) 1047 q->rate2quantum = 1; 1048 q->defcls = gopt->defcls; 1049 1050 return 0; 1051 } 1052 1053 static int htb_dump(struct Qdisc *sch, struct sk_buff *skb) 1054 { 1055 struct htb_sched *q = qdisc_priv(sch); 1056 struct nlattr *nest; 1057 struct tc_htb_glob gopt; 1058 1059 /* Its safe to not acquire qdisc lock. As we hold RTNL, 1060 * no change can happen on the qdisc parameters. 1061 */ 1062 1063 gopt.direct_pkts = q->direct_pkts; 1064 gopt.version = HTB_VER; 1065 gopt.rate2quantum = q->rate2quantum; 1066 gopt.defcls = q->defcls; 1067 gopt.debug = 0; 1068 1069 nest = nla_nest_start(skb, TCA_OPTIONS); 1070 if (nest == NULL) 1071 goto nla_put_failure; 1072 if (nla_put(skb, TCA_HTB_INIT, sizeof(gopt), &gopt) || 1073 nla_put_u32(skb, TCA_HTB_DIRECT_QLEN, q->direct_qlen)) 1074 goto nla_put_failure; 1075 1076 return nla_nest_end(skb, nest); 1077 1078 nla_put_failure: 1079 nla_nest_cancel(skb, nest); 1080 return -1; 1081 } 1082 1083 static int htb_dump_class(struct Qdisc *sch, unsigned long arg, 1084 struct sk_buff *skb, struct tcmsg *tcm) 1085 { 1086 struct htb_class *cl = (struct htb_class *)arg; 1087 struct nlattr *nest; 1088 struct tc_htb_opt opt; 1089 1090 /* Its safe to not acquire qdisc lock. As we hold RTNL, 1091 * no change can happen on the class parameters. 1092 */ 1093 tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT; 1094 tcm->tcm_handle = cl->common.classid; 1095 if (!cl->level && cl->un.leaf.q) 1096 tcm->tcm_info = cl->un.leaf.q->handle; 1097 1098 nest = nla_nest_start(skb, TCA_OPTIONS); 1099 if (nest == NULL) 1100 goto nla_put_failure; 1101 1102 memset(&opt, 0, sizeof(opt)); 1103 1104 psched_ratecfg_getrate(&opt.rate, &cl->rate); 1105 opt.buffer = PSCHED_NS2TICKS(cl->buffer); 1106 psched_ratecfg_getrate(&opt.ceil, &cl->ceil); 1107 opt.cbuffer = PSCHED_NS2TICKS(cl->cbuffer); 1108 opt.quantum = cl->quantum; 1109 opt.prio = cl->prio; 1110 opt.level = cl->level; 1111 if (nla_put(skb, TCA_HTB_PARMS, sizeof(opt), &opt)) 1112 goto nla_put_failure; 1113 if ((cl->rate.rate_bytes_ps >= (1ULL << 32)) && 1114 nla_put_u64_64bit(skb, TCA_HTB_RATE64, cl->rate.rate_bytes_ps, 1115 TCA_HTB_PAD)) 1116 goto nla_put_failure; 1117 if ((cl->ceil.rate_bytes_ps >= (1ULL << 32)) && 1118 nla_put_u64_64bit(skb, TCA_HTB_CEIL64, cl->ceil.rate_bytes_ps, 1119 TCA_HTB_PAD)) 1120 goto nla_put_failure; 1121 1122 return nla_nest_end(skb, nest); 1123 1124 nla_put_failure: 1125 nla_nest_cancel(skb, nest); 1126 return -1; 1127 } 1128 1129 static int 1130 htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d) 1131 { 1132 struct htb_class *cl = (struct htb_class *)arg; 1133 struct gnet_stats_queue qs = { 1134 .drops = cl->drops, 1135 }; 1136 __u32 qlen = 0; 1137 1138 if (!cl->level && cl->un.leaf.q) { 1139 qlen = cl->un.leaf.q->q.qlen; 1140 qs.backlog = cl->un.leaf.q->qstats.backlog; 1141 } 1142 cl->xstats.tokens = clamp_t(s64, PSCHED_NS2TICKS(cl->tokens), 1143 INT_MIN, INT_MAX); 1144 cl->xstats.ctokens = clamp_t(s64, PSCHED_NS2TICKS(cl->ctokens), 1145 INT_MIN, INT_MAX); 1146 1147 if (gnet_stats_copy_basic(qdisc_root_sleeping_running(sch), 1148 d, NULL, &cl->bstats) < 0 || 1149 gnet_stats_copy_rate_est(d, &cl->rate_est) < 0 || 1150 gnet_stats_copy_queue(d, NULL, &qs, qlen) < 0) 1151 return -1; 1152 1153 return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats)); 1154 } 1155 1156 static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 1157 struct Qdisc **old) 1158 { 1159 struct htb_class *cl = (struct htb_class *)arg; 1160 1161 if (cl->level) 1162 return -EINVAL; 1163 if (new == NULL && 1164 (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, 1165 cl->common.classid)) == NULL) 1166 return -ENOBUFS; 1167 1168 *old = qdisc_replace(sch, new, &cl->un.leaf.q); 1169 return 0; 1170 } 1171 1172 static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg) 1173 { 1174 struct htb_class *cl = (struct htb_class *)arg; 1175 return !cl->level ? cl->un.leaf.q : NULL; 1176 } 1177 1178 static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg) 1179 { 1180 struct htb_class *cl = (struct htb_class *)arg; 1181 1182 if (cl->un.leaf.q->q.qlen == 0) 1183 htb_deactivate(qdisc_priv(sch), cl); 1184 } 1185 1186 static unsigned long htb_get(struct Qdisc *sch, u32 classid) 1187 { 1188 struct htb_class *cl = htb_find(classid, sch); 1189 if (cl) 1190 cl->refcnt++; 1191 return (unsigned long)cl; 1192 } 1193 1194 static inline int htb_parent_last_child(struct htb_class *cl) 1195 { 1196 if (!cl->parent) 1197 /* the root class */ 1198 return 0; 1199 if (cl->parent->children > 1) 1200 /* not the last child */ 1201 return 0; 1202 return 1; 1203 } 1204 1205 static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl, 1206 struct Qdisc *new_q) 1207 { 1208 struct htb_class *parent = cl->parent; 1209 1210 WARN_ON(cl->level || !cl->un.leaf.q || cl->prio_activity); 1211 1212 if (parent->cmode != HTB_CAN_SEND) 1213 htb_safe_rb_erase(&parent->pq_node, 1214 &q->hlevel[parent->level].wait_pq); 1215 1216 parent->level = 0; 1217 memset(&parent->un.inner, 0, sizeof(parent->un.inner)); 1218 INIT_LIST_HEAD(&parent->un.leaf.drop_list); 1219 parent->un.leaf.q = new_q ? new_q : &noop_qdisc; 1220 parent->tokens = parent->buffer; 1221 parent->ctokens = parent->cbuffer; 1222 parent->t_c = ktime_get_ns(); 1223 parent->cmode = HTB_CAN_SEND; 1224 } 1225 1226 static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl) 1227 { 1228 if (!cl->level) { 1229 WARN_ON(!cl->un.leaf.q); 1230 qdisc_destroy(cl->un.leaf.q); 1231 } 1232 gen_kill_estimator(&cl->rate_est); 1233 tcf_destroy_chain(&cl->filter_list); 1234 kfree(cl); 1235 } 1236 1237 static void htb_destroy(struct Qdisc *sch) 1238 { 1239 struct htb_sched *q = qdisc_priv(sch); 1240 struct hlist_node *next; 1241 struct htb_class *cl; 1242 unsigned int i; 1243 1244 cancel_work_sync(&q->work); 1245 qdisc_watchdog_cancel(&q->watchdog); 1246 /* This line used to be after htb_destroy_class call below 1247 * and surprisingly it worked in 2.4. But it must precede it 1248 * because filter need its target class alive to be able to call 1249 * unbind_filter on it (without Oops). 1250 */ 1251 tcf_destroy_chain(&q->filter_list); 1252 1253 for (i = 0; i < q->clhash.hashsize; i++) { 1254 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) 1255 tcf_destroy_chain(&cl->filter_list); 1256 } 1257 for (i = 0; i < q->clhash.hashsize; i++) { 1258 hlist_for_each_entry_safe(cl, next, &q->clhash.hash[i], 1259 common.hnode) 1260 htb_destroy_class(sch, cl); 1261 } 1262 qdisc_class_hash_destroy(&q->clhash); 1263 __qdisc_reset_queue(&q->direct_queue); 1264 } 1265 1266 static int htb_delete(struct Qdisc *sch, unsigned long arg) 1267 { 1268 struct htb_sched *q = qdisc_priv(sch); 1269 struct htb_class *cl = (struct htb_class *)arg; 1270 struct Qdisc *new_q = NULL; 1271 int last_child = 0; 1272 1273 /* TODO: why don't allow to delete subtree ? references ? does 1274 * tc subsys guarantee us that in htb_destroy it holds no class 1275 * refs so that we can remove children safely there ? 1276 */ 1277 if (cl->children || cl->filter_cnt) 1278 return -EBUSY; 1279 1280 if (!cl->level && htb_parent_last_child(cl)) { 1281 new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, 1282 cl->parent->common.classid); 1283 last_child = 1; 1284 } 1285 1286 sch_tree_lock(sch); 1287 1288 if (!cl->level) { 1289 unsigned int qlen = cl->un.leaf.q->q.qlen; 1290 unsigned int backlog = cl->un.leaf.q->qstats.backlog; 1291 1292 qdisc_reset(cl->un.leaf.q); 1293 qdisc_tree_reduce_backlog(cl->un.leaf.q, qlen, backlog); 1294 } 1295 1296 /* delete from hash and active; remainder in destroy_class */ 1297 qdisc_class_hash_remove(&q->clhash, &cl->common); 1298 if (cl->parent) 1299 cl->parent->children--; 1300 1301 if (cl->prio_activity) 1302 htb_deactivate(q, cl); 1303 1304 if (cl->cmode != HTB_CAN_SEND) 1305 htb_safe_rb_erase(&cl->pq_node, 1306 &q->hlevel[cl->level].wait_pq); 1307 1308 if (last_child) 1309 htb_parent_to_leaf(q, cl, new_q); 1310 1311 BUG_ON(--cl->refcnt == 0); 1312 /* 1313 * This shouldn't happen: we "hold" one cops->get() when called 1314 * from tc_ctl_tclass; the destroy method is done from cops->put(). 1315 */ 1316 1317 sch_tree_unlock(sch); 1318 return 0; 1319 } 1320 1321 static void htb_put(struct Qdisc *sch, unsigned long arg) 1322 { 1323 struct htb_class *cl = (struct htb_class *)arg; 1324 1325 if (--cl->refcnt == 0) 1326 htb_destroy_class(sch, cl); 1327 } 1328 1329 static int htb_change_class(struct Qdisc *sch, u32 classid, 1330 u32 parentid, struct nlattr **tca, 1331 unsigned long *arg) 1332 { 1333 int err = -EINVAL; 1334 struct htb_sched *q = qdisc_priv(sch); 1335 struct htb_class *cl = (struct htb_class *)*arg, *parent; 1336 struct nlattr *opt = tca[TCA_OPTIONS]; 1337 struct nlattr *tb[TCA_HTB_MAX + 1]; 1338 struct tc_htb_opt *hopt; 1339 u64 rate64, ceil64; 1340 1341 /* extract all subattrs from opt attr */ 1342 if (!opt) 1343 goto failure; 1344 1345 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy, NULL); 1346 if (err < 0) 1347 goto failure; 1348 1349 err = -EINVAL; 1350 if (tb[TCA_HTB_PARMS] == NULL) 1351 goto failure; 1352 1353 parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch); 1354 1355 hopt = nla_data(tb[TCA_HTB_PARMS]); 1356 if (!hopt->rate.rate || !hopt->ceil.rate) 1357 goto failure; 1358 1359 /* Keeping backward compatible with rate_table based iproute2 tc */ 1360 if (hopt->rate.linklayer == TC_LINKLAYER_UNAWARE) 1361 qdisc_put_rtab(qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB])); 1362 1363 if (hopt->ceil.linklayer == TC_LINKLAYER_UNAWARE) 1364 qdisc_put_rtab(qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB])); 1365 1366 if (!cl) { /* new class */ 1367 struct Qdisc *new_q; 1368 int prio; 1369 struct { 1370 struct nlattr nla; 1371 struct gnet_estimator opt; 1372 } est = { 1373 .nla = { 1374 .nla_len = nla_attr_size(sizeof(est.opt)), 1375 .nla_type = TCA_RATE, 1376 }, 1377 .opt = { 1378 /* 4s interval, 16s averaging constant */ 1379 .interval = 2, 1380 .ewma_log = 2, 1381 }, 1382 }; 1383 1384 /* check for valid classid */ 1385 if (!classid || TC_H_MAJ(classid ^ sch->handle) || 1386 htb_find(classid, sch)) 1387 goto failure; 1388 1389 /* check maximal depth */ 1390 if (parent && parent->parent && parent->parent->level < 2) { 1391 pr_err("htb: tree is too deep\n"); 1392 goto failure; 1393 } 1394 err = -ENOBUFS; 1395 cl = kzalloc(sizeof(*cl), GFP_KERNEL); 1396 if (!cl) 1397 goto failure; 1398 1399 if (htb_rate_est || tca[TCA_RATE]) { 1400 err = gen_new_estimator(&cl->bstats, NULL, 1401 &cl->rate_est, 1402 NULL, 1403 qdisc_root_sleeping_running(sch), 1404 tca[TCA_RATE] ? : &est.nla); 1405 if (err) { 1406 kfree(cl); 1407 goto failure; 1408 } 1409 } 1410 1411 cl->refcnt = 1; 1412 cl->children = 0; 1413 INIT_LIST_HEAD(&cl->un.leaf.drop_list); 1414 RB_CLEAR_NODE(&cl->pq_node); 1415 1416 for (prio = 0; prio < TC_HTB_NUMPRIO; prio++) 1417 RB_CLEAR_NODE(&cl->node[prio]); 1418 1419 /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL) 1420 * so that can't be used inside of sch_tree_lock 1421 * -- thanks to Karlis Peisenieks 1422 */ 1423 new_q = qdisc_create_dflt(sch->dev_queue, 1424 &pfifo_qdisc_ops, classid); 1425 sch_tree_lock(sch); 1426 if (parent && !parent->level) { 1427 unsigned int qlen = parent->un.leaf.q->q.qlen; 1428 unsigned int backlog = parent->un.leaf.q->qstats.backlog; 1429 1430 /* turn parent into inner node */ 1431 qdisc_reset(parent->un.leaf.q); 1432 qdisc_tree_reduce_backlog(parent->un.leaf.q, qlen, backlog); 1433 qdisc_destroy(parent->un.leaf.q); 1434 if (parent->prio_activity) 1435 htb_deactivate(q, parent); 1436 1437 /* remove from evt list because of level change */ 1438 if (parent->cmode != HTB_CAN_SEND) { 1439 htb_safe_rb_erase(&parent->pq_node, &q->hlevel[0].wait_pq); 1440 parent->cmode = HTB_CAN_SEND; 1441 } 1442 parent->level = (parent->parent ? parent->parent->level 1443 : TC_HTB_MAXDEPTH) - 1; 1444 memset(&parent->un.inner, 0, sizeof(parent->un.inner)); 1445 } 1446 /* leaf (we) needs elementary qdisc */ 1447 cl->un.leaf.q = new_q ? new_q : &noop_qdisc; 1448 1449 cl->common.classid = classid; 1450 cl->parent = parent; 1451 1452 /* set class to be in HTB_CAN_SEND state */ 1453 cl->tokens = PSCHED_TICKS2NS(hopt->buffer); 1454 cl->ctokens = PSCHED_TICKS2NS(hopt->cbuffer); 1455 cl->mbuffer = 60ULL * NSEC_PER_SEC; /* 1min */ 1456 cl->t_c = ktime_get_ns(); 1457 cl->cmode = HTB_CAN_SEND; 1458 1459 /* attach to the hash list and parent's family */ 1460 qdisc_class_hash_insert(&q->clhash, &cl->common); 1461 if (parent) 1462 parent->children++; 1463 if (cl->un.leaf.q != &noop_qdisc) 1464 qdisc_hash_add(cl->un.leaf.q, true); 1465 } else { 1466 if (tca[TCA_RATE]) { 1467 err = gen_replace_estimator(&cl->bstats, NULL, 1468 &cl->rate_est, 1469 NULL, 1470 qdisc_root_sleeping_running(sch), 1471 tca[TCA_RATE]); 1472 if (err) 1473 return err; 1474 } 1475 sch_tree_lock(sch); 1476 } 1477 1478 rate64 = tb[TCA_HTB_RATE64] ? nla_get_u64(tb[TCA_HTB_RATE64]) : 0; 1479 1480 ceil64 = tb[TCA_HTB_CEIL64] ? nla_get_u64(tb[TCA_HTB_CEIL64]) : 0; 1481 1482 psched_ratecfg_precompute(&cl->rate, &hopt->rate, rate64); 1483 psched_ratecfg_precompute(&cl->ceil, &hopt->ceil, ceil64); 1484 1485 /* it used to be a nasty bug here, we have to check that node 1486 * is really leaf before changing cl->un.leaf ! 1487 */ 1488 if (!cl->level) { 1489 u64 quantum = cl->rate.rate_bytes_ps; 1490 1491 do_div(quantum, q->rate2quantum); 1492 cl->quantum = min_t(u64, quantum, INT_MAX); 1493 1494 if (!hopt->quantum && cl->quantum < 1000) { 1495 pr_warn("HTB: quantum of class %X is small. Consider r2q change.\n", 1496 cl->common.classid); 1497 cl->quantum = 1000; 1498 } 1499 if (!hopt->quantum && cl->quantum > 200000) { 1500 pr_warn("HTB: quantum of class %X is big. Consider r2q change.\n", 1501 cl->common.classid); 1502 cl->quantum = 200000; 1503 } 1504 if (hopt->quantum) 1505 cl->quantum = hopt->quantum; 1506 if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO) 1507 cl->prio = TC_HTB_NUMPRIO - 1; 1508 } 1509 1510 cl->buffer = PSCHED_TICKS2NS(hopt->buffer); 1511 cl->cbuffer = PSCHED_TICKS2NS(hopt->cbuffer); 1512 1513 sch_tree_unlock(sch); 1514 1515 qdisc_class_hash_grow(sch, &q->clhash); 1516 1517 *arg = (unsigned long)cl; 1518 return 0; 1519 1520 failure: 1521 return err; 1522 } 1523 1524 static struct tcf_proto __rcu **htb_find_tcf(struct Qdisc *sch, 1525 unsigned long arg) 1526 { 1527 struct htb_sched *q = qdisc_priv(sch); 1528 struct htb_class *cl = (struct htb_class *)arg; 1529 struct tcf_proto __rcu **fl = cl ? &cl->filter_list : &q->filter_list; 1530 1531 return fl; 1532 } 1533 1534 static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent, 1535 u32 classid) 1536 { 1537 struct htb_class *cl = htb_find(classid, sch); 1538 1539 /*if (cl && !cl->level) return 0; 1540 * The line above used to be there to prevent attaching filters to 1541 * leaves. But at least tc_index filter uses this just to get class 1542 * for other reasons so that we have to allow for it. 1543 * ---- 1544 * 19.6.2002 As Werner explained it is ok - bind filter is just 1545 * another way to "lock" the class - unlike "get" this lock can 1546 * be broken by class during destroy IIUC. 1547 */ 1548 if (cl) 1549 cl->filter_cnt++; 1550 return (unsigned long)cl; 1551 } 1552 1553 static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg) 1554 { 1555 struct htb_class *cl = (struct htb_class *)arg; 1556 1557 if (cl) 1558 cl->filter_cnt--; 1559 } 1560 1561 static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg) 1562 { 1563 struct htb_sched *q = qdisc_priv(sch); 1564 struct htb_class *cl; 1565 unsigned int i; 1566 1567 if (arg->stop) 1568 return; 1569 1570 for (i = 0; i < q->clhash.hashsize; i++) { 1571 hlist_for_each_entry(cl, &q->clhash.hash[i], common.hnode) { 1572 if (arg->count < arg->skip) { 1573 arg->count++; 1574 continue; 1575 } 1576 if (arg->fn(sch, (unsigned long)cl, arg) < 0) { 1577 arg->stop = 1; 1578 return; 1579 } 1580 arg->count++; 1581 } 1582 } 1583 } 1584 1585 static const struct Qdisc_class_ops htb_class_ops = { 1586 .graft = htb_graft, 1587 .leaf = htb_leaf, 1588 .qlen_notify = htb_qlen_notify, 1589 .get = htb_get, 1590 .put = htb_put, 1591 .change = htb_change_class, 1592 .delete = htb_delete, 1593 .walk = htb_walk, 1594 .tcf_chain = htb_find_tcf, 1595 .bind_tcf = htb_bind_filter, 1596 .unbind_tcf = htb_unbind_filter, 1597 .dump = htb_dump_class, 1598 .dump_stats = htb_dump_class_stats, 1599 }; 1600 1601 static struct Qdisc_ops htb_qdisc_ops __read_mostly = { 1602 .cl_ops = &htb_class_ops, 1603 .id = "htb", 1604 .priv_size = sizeof(struct htb_sched), 1605 .enqueue = htb_enqueue, 1606 .dequeue = htb_dequeue, 1607 .peek = qdisc_peek_dequeued, 1608 .init = htb_init, 1609 .reset = htb_reset, 1610 .destroy = htb_destroy, 1611 .dump = htb_dump, 1612 .owner = THIS_MODULE, 1613 }; 1614 1615 static int __init htb_module_init(void) 1616 { 1617 return register_qdisc(&htb_qdisc_ops); 1618 } 1619 static void __exit htb_module_exit(void) 1620 { 1621 unregister_qdisc(&htb_qdisc_ops); 1622 } 1623 1624 module_init(htb_module_init) 1625 module_exit(htb_module_exit) 1626 MODULE_LICENSE("GPL"); 1627