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/pkt_sched.h> 42 43 /* HTB algorithm. 44 Author: devik@cdi.cz 45 ======================================================================== 46 HTB is like TBF with multiple classes. It is also similar to CBQ because 47 it allows to assign priority to each class in hierarchy. 48 In fact it is another implementation of Floyd's formal sharing. 49 50 Levels: 51 Each class is assigned level. Leaf has ALWAYS level 0 and root 52 classes have level TC_HTB_MAXDEPTH-1. Interior nodes has level 53 one less than their parent. 54 */ 55 56 static int htb_hysteresis __read_mostly = 0; /* whether to use mode hysteresis for speedup */ 57 #define HTB_VER 0x30011 /* major must be matched with number suplied by TC as version */ 58 59 #if HTB_VER >> 16 != TC_HTB_PROTOVER 60 #error "Mismatched sch_htb.c and pkt_sch.h" 61 #endif 62 63 /* Module parameter and sysfs export */ 64 module_param (htb_hysteresis, int, 0640); 65 MODULE_PARM_DESC(htb_hysteresis, "Hysteresis mode, less CPU load, less accurate"); 66 67 /* used internaly to keep status of single class */ 68 enum htb_cmode { 69 HTB_CANT_SEND, /* class can't send and can't borrow */ 70 HTB_MAY_BORROW, /* class can't send but may borrow */ 71 HTB_CAN_SEND /* class can send */ 72 }; 73 74 /* interior & leaf nodes; props specific to leaves are marked L: */ 75 struct htb_class { 76 struct Qdisc_class_common common; 77 /* general class parameters */ 78 struct gnet_stats_basic_packed bstats; 79 struct gnet_stats_queue qstats; 80 struct gnet_stats_rate_est rate_est; 81 struct tc_htb_xstats xstats; /* our special stats */ 82 int refcnt; /* usage count of this class */ 83 84 /* topology */ 85 int level; /* our level (see above) */ 86 unsigned int children; 87 struct htb_class *parent; /* parent class */ 88 89 int prio; /* these two are used only by leaves... */ 90 int quantum; /* but stored for parent-to-leaf return */ 91 92 union { 93 struct htb_class_leaf { 94 struct Qdisc *q; 95 int deficit[TC_HTB_MAXDEPTH]; 96 struct list_head drop_list; 97 } leaf; 98 struct htb_class_inner { 99 struct rb_root feed[TC_HTB_NUMPRIO]; /* feed trees */ 100 struct rb_node *ptr[TC_HTB_NUMPRIO]; /* current class ptr */ 101 /* When class changes from state 1->2 and disconnects from 102 parent's feed then we lost ptr value and start from the 103 first child again. Here we store classid of the 104 last valid ptr (used when ptr is NULL). */ 105 u32 last_ptr_id[TC_HTB_NUMPRIO]; 106 } inner; 107 } un; 108 struct rb_node node[TC_HTB_NUMPRIO]; /* node for self or feed tree */ 109 struct rb_node pq_node; /* node for event queue */ 110 psched_time_t pq_key; 111 112 int prio_activity; /* for which prios are we active */ 113 enum htb_cmode cmode; /* current mode of the class */ 114 115 /* class attached filters */ 116 struct tcf_proto *filter_list; 117 int filter_cnt; 118 119 /* token bucket parameters */ 120 struct qdisc_rate_table *rate; /* rate table of the class itself */ 121 struct qdisc_rate_table *ceil; /* ceiling rate (limits borrows too) */ 122 long buffer, cbuffer; /* token bucket depth/rate */ 123 psched_tdiff_t mbuffer; /* max wait time */ 124 long tokens, ctokens; /* current number of tokens */ 125 psched_time_t t_c; /* checkpoint time */ 126 }; 127 128 struct htb_sched { 129 struct Qdisc_class_hash clhash; 130 struct list_head drops[TC_HTB_NUMPRIO];/* active leaves (for drops) */ 131 132 /* self list - roots of self generating tree */ 133 struct rb_root row[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO]; 134 int row_mask[TC_HTB_MAXDEPTH]; 135 struct rb_node *ptr[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO]; 136 u32 last_ptr_id[TC_HTB_MAXDEPTH][TC_HTB_NUMPRIO]; 137 138 /* self wait list - roots of wait PQs per row */ 139 struct rb_root wait_pq[TC_HTB_MAXDEPTH]; 140 141 /* time of nearest event per level (row) */ 142 psched_time_t near_ev_cache[TC_HTB_MAXDEPTH]; 143 144 int defcls; /* class where unclassified flows go to */ 145 146 /* filters for qdisc itself */ 147 struct tcf_proto *filter_list; 148 149 int rate2quantum; /* quant = rate / rate2quantum */ 150 psched_time_t now; /* cached dequeue time */ 151 struct qdisc_watchdog watchdog; 152 153 /* non shaped skbs; let them go directly thru */ 154 struct sk_buff_head direct_queue; 155 int direct_qlen; /* max qlen of above */ 156 157 long direct_pkts; 158 159 #define HTB_WARN_TOOMANYEVENTS 0x1 160 unsigned int warned; /* only one warning */ 161 struct work_struct work; 162 }; 163 164 /* find class in global hash table using given handle */ 165 static inline struct htb_class *htb_find(u32 handle, struct Qdisc *sch) 166 { 167 struct htb_sched *q = qdisc_priv(sch); 168 struct Qdisc_class_common *clc; 169 170 clc = qdisc_class_find(&q->clhash, handle); 171 if (clc == NULL) 172 return NULL; 173 return container_of(clc, struct htb_class, common); 174 } 175 176 /** 177 * htb_classify - classify a packet into class 178 * 179 * It returns NULL if the packet should be dropped or -1 if the packet 180 * should be passed directly thru. In all other cases leaf class is returned. 181 * We allow direct class selection by classid in priority. The we examine 182 * filters in qdisc and in inner nodes (if higher filter points to the inner 183 * node). If we end up with classid MAJOR:0 we enqueue the skb into special 184 * internal fifo (direct). These packets then go directly thru. If we still 185 * have no valid leaf we try to use MAJOR:default leaf. It still unsuccessfull 186 * then finish and return direct queue. 187 */ 188 #define HTB_DIRECT (struct htb_class*)-1 189 190 static struct htb_class *htb_classify(struct sk_buff *skb, struct Qdisc *sch, 191 int *qerr) 192 { 193 struct htb_sched *q = qdisc_priv(sch); 194 struct htb_class *cl; 195 struct tcf_result res; 196 struct tcf_proto *tcf; 197 int result; 198 199 /* allow to select class by setting skb->priority to valid classid; 200 note that nfmark can be used too by attaching filter fw with no 201 rules in it */ 202 if (skb->priority == sch->handle) 203 return HTB_DIRECT; /* X:0 (direct flow) selected */ 204 if ((cl = htb_find(skb->priority, sch)) != NULL && cl->level == 0) 205 return cl; 206 207 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 208 tcf = q->filter_list; 209 while (tcf && (result = tc_classify(skb, tcf, &res)) >= 0) { 210 #ifdef CONFIG_NET_CLS_ACT 211 switch (result) { 212 case TC_ACT_QUEUED: 213 case TC_ACT_STOLEN: 214 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; 215 case TC_ACT_SHOT: 216 return NULL; 217 } 218 #endif 219 if ((cl = (void *)res.class) == NULL) { 220 if (res.classid == sch->handle) 221 return HTB_DIRECT; /* X:0 (direct flow) */ 222 if ((cl = htb_find(res.classid, sch)) == NULL) 223 break; /* filter selected invalid classid */ 224 } 225 if (!cl->level) 226 return cl; /* we hit leaf; return it */ 227 228 /* we have got inner class; apply inner filter chain */ 229 tcf = cl->filter_list; 230 } 231 /* classification failed; try to use default class */ 232 cl = htb_find(TC_H_MAKE(TC_H_MAJ(sch->handle), q->defcls), sch); 233 if (!cl || cl->level) 234 return HTB_DIRECT; /* bad default .. this is safe bet */ 235 return cl; 236 } 237 238 /** 239 * htb_add_to_id_tree - adds class to the round robin list 240 * 241 * Routine adds class to the list (actually tree) sorted by classid. 242 * Make sure that class is not already on such list for given prio. 243 */ 244 static void htb_add_to_id_tree(struct rb_root *root, 245 struct htb_class *cl, int prio) 246 { 247 struct rb_node **p = &root->rb_node, *parent = NULL; 248 249 while (*p) { 250 struct htb_class *c; 251 parent = *p; 252 c = rb_entry(parent, struct htb_class, node[prio]); 253 254 if (cl->common.classid > c->common.classid) 255 p = &parent->rb_right; 256 else 257 p = &parent->rb_left; 258 } 259 rb_link_node(&cl->node[prio], parent, p); 260 rb_insert_color(&cl->node[prio], root); 261 } 262 263 /** 264 * htb_add_to_wait_tree - adds class to the event queue with delay 265 * 266 * The class is added to priority event queue to indicate that class will 267 * change its mode in cl->pq_key microseconds. Make sure that class is not 268 * already in the queue. 269 */ 270 static void htb_add_to_wait_tree(struct htb_sched *q, 271 struct htb_class *cl, long delay) 272 { 273 struct rb_node **p = &q->wait_pq[cl->level].rb_node, *parent = NULL; 274 275 cl->pq_key = q->now + delay; 276 if (cl->pq_key == q->now) 277 cl->pq_key++; 278 279 /* update the nearest event cache */ 280 if (q->near_ev_cache[cl->level] > cl->pq_key) 281 q->near_ev_cache[cl->level] = cl->pq_key; 282 283 while (*p) { 284 struct htb_class *c; 285 parent = *p; 286 c = rb_entry(parent, struct htb_class, pq_node); 287 if (cl->pq_key >= c->pq_key) 288 p = &parent->rb_right; 289 else 290 p = &parent->rb_left; 291 } 292 rb_link_node(&cl->pq_node, parent, p); 293 rb_insert_color(&cl->pq_node, &q->wait_pq[cl->level]); 294 } 295 296 /** 297 * htb_next_rb_node - finds next node in binary tree 298 * 299 * When we are past last key we return NULL. 300 * Average complexity is 2 steps per call. 301 */ 302 static inline void htb_next_rb_node(struct rb_node **n) 303 { 304 *n = rb_next(*n); 305 } 306 307 /** 308 * htb_add_class_to_row - add class to its row 309 * 310 * The class is added to row at priorities marked in mask. 311 * It does nothing if mask == 0. 312 */ 313 static inline void htb_add_class_to_row(struct htb_sched *q, 314 struct htb_class *cl, int mask) 315 { 316 q->row_mask[cl->level] |= mask; 317 while (mask) { 318 int prio = ffz(~mask); 319 mask &= ~(1 << prio); 320 htb_add_to_id_tree(q->row[cl->level] + prio, cl, prio); 321 } 322 } 323 324 /* If this triggers, it is a bug in this code, but it need not be fatal */ 325 static void htb_safe_rb_erase(struct rb_node *rb, struct rb_root *root) 326 { 327 if (RB_EMPTY_NODE(rb)) { 328 WARN_ON(1); 329 } else { 330 rb_erase(rb, root); 331 RB_CLEAR_NODE(rb); 332 } 333 } 334 335 336 /** 337 * htb_remove_class_from_row - removes class from its row 338 * 339 * The class is removed from row at priorities marked in mask. 340 * It does nothing if mask == 0. 341 */ 342 static inline void htb_remove_class_from_row(struct htb_sched *q, 343 struct htb_class *cl, int mask) 344 { 345 int m = 0; 346 347 while (mask) { 348 int prio = ffz(~mask); 349 350 mask &= ~(1 << prio); 351 if (q->ptr[cl->level][prio] == cl->node + prio) 352 htb_next_rb_node(q->ptr[cl->level] + prio); 353 354 htb_safe_rb_erase(cl->node + prio, q->row[cl->level] + prio); 355 if (!q->row[cl->level][prio].rb_node) 356 m |= 1 << prio; 357 } 358 q->row_mask[cl->level] &= ~m; 359 } 360 361 /** 362 * htb_activate_prios - creates active classe's feed chain 363 * 364 * The class is connected to ancestors and/or appropriate rows 365 * for priorities it is participating on. cl->cmode must be new 366 * (activated) mode. It does nothing if cl->prio_activity == 0. 367 */ 368 static void htb_activate_prios(struct htb_sched *q, struct htb_class *cl) 369 { 370 struct htb_class *p = cl->parent; 371 long m, mask = cl->prio_activity; 372 373 while (cl->cmode == HTB_MAY_BORROW && p && mask) { 374 m = mask; 375 while (m) { 376 int prio = ffz(~m); 377 m &= ~(1 << prio); 378 379 if (p->un.inner.feed[prio].rb_node) 380 /* parent already has its feed in use so that 381 reset bit in mask as parent is already ok */ 382 mask &= ~(1 << prio); 383 384 htb_add_to_id_tree(p->un.inner.feed + prio, cl, prio); 385 } 386 p->prio_activity |= mask; 387 cl = p; 388 p = cl->parent; 389 390 } 391 if (cl->cmode == HTB_CAN_SEND && mask) 392 htb_add_class_to_row(q, cl, mask); 393 } 394 395 /** 396 * htb_deactivate_prios - remove class from feed chain 397 * 398 * cl->cmode must represent old mode (before deactivation). It does 399 * nothing if cl->prio_activity == 0. Class is removed from all feed 400 * chains and rows. 401 */ 402 static void htb_deactivate_prios(struct htb_sched *q, struct htb_class *cl) 403 { 404 struct htb_class *p = cl->parent; 405 long m, mask = cl->prio_activity; 406 407 while (cl->cmode == HTB_MAY_BORROW && p && mask) { 408 m = mask; 409 mask = 0; 410 while (m) { 411 int prio = ffz(~m); 412 m &= ~(1 << prio); 413 414 if (p->un.inner.ptr[prio] == cl->node + prio) { 415 /* we are removing child which is pointed to from 416 parent feed - forget the pointer but remember 417 classid */ 418 p->un.inner.last_ptr_id[prio] = cl->common.classid; 419 p->un.inner.ptr[prio] = NULL; 420 } 421 422 htb_safe_rb_erase(cl->node + prio, p->un.inner.feed + prio); 423 424 if (!p->un.inner.feed[prio].rb_node) 425 mask |= 1 << prio; 426 } 427 428 p->prio_activity &= ~mask; 429 cl = p; 430 p = cl->parent; 431 432 } 433 if (cl->cmode == HTB_CAN_SEND && mask) 434 htb_remove_class_from_row(q, cl, mask); 435 } 436 437 static inline long htb_lowater(const struct htb_class *cl) 438 { 439 if (htb_hysteresis) 440 return cl->cmode != HTB_CANT_SEND ? -cl->cbuffer : 0; 441 else 442 return 0; 443 } 444 static inline long htb_hiwater(const struct htb_class *cl) 445 { 446 if (htb_hysteresis) 447 return cl->cmode == HTB_CAN_SEND ? -cl->buffer : 0; 448 else 449 return 0; 450 } 451 452 453 /** 454 * htb_class_mode - computes and returns current class mode 455 * 456 * It computes cl's mode at time cl->t_c+diff and returns it. If mode 457 * is not HTB_CAN_SEND then cl->pq_key is updated to time difference 458 * from now to time when cl will change its state. 459 * Also it is worth to note that class mode doesn't change simply 460 * at cl->{c,}tokens == 0 but there can rather be hysteresis of 461 * 0 .. -cl->{c,}buffer range. It is meant to limit number of 462 * mode transitions per time unit. The speed gain is about 1/6. 463 */ 464 static inline enum htb_cmode 465 htb_class_mode(struct htb_class *cl, long *diff) 466 { 467 long toks; 468 469 if ((toks = (cl->ctokens + *diff)) < htb_lowater(cl)) { 470 *diff = -toks; 471 return HTB_CANT_SEND; 472 } 473 474 if ((toks = (cl->tokens + *diff)) >= htb_hiwater(cl)) 475 return HTB_CAN_SEND; 476 477 *diff = -toks; 478 return HTB_MAY_BORROW; 479 } 480 481 /** 482 * htb_change_class_mode - changes classe's mode 483 * 484 * This should be the only way how to change classe's mode under normal 485 * cirsumstances. Routine will update feed lists linkage, change mode 486 * and add class to the wait event queue if appropriate. New mode should 487 * be different from old one and cl->pq_key has to be valid if changing 488 * to mode other than HTB_CAN_SEND (see htb_add_to_wait_tree). 489 */ 490 static void 491 htb_change_class_mode(struct htb_sched *q, struct htb_class *cl, long *diff) 492 { 493 enum htb_cmode new_mode = htb_class_mode(cl, diff); 494 495 if (new_mode == cl->cmode) 496 return; 497 498 if (cl->prio_activity) { /* not necessary: speed optimization */ 499 if (cl->cmode != HTB_CANT_SEND) 500 htb_deactivate_prios(q, cl); 501 cl->cmode = new_mode; 502 if (new_mode != HTB_CANT_SEND) 503 htb_activate_prios(q, cl); 504 } else 505 cl->cmode = new_mode; 506 } 507 508 /** 509 * htb_activate - inserts leaf cl into appropriate active feeds 510 * 511 * Routine learns (new) priority of leaf and activates feed chain 512 * for the prio. It can be called on already active leaf safely. 513 * It also adds leaf into droplist. 514 */ 515 static inline void htb_activate(struct htb_sched *q, struct htb_class *cl) 516 { 517 WARN_ON(cl->level || !cl->un.leaf.q || !cl->un.leaf.q->q.qlen); 518 519 if (!cl->prio_activity) { 520 cl->prio_activity = 1 << cl->prio; 521 htb_activate_prios(q, cl); 522 list_add_tail(&cl->un.leaf.drop_list, 523 q->drops + cl->prio); 524 } 525 } 526 527 /** 528 * htb_deactivate - remove leaf cl from active feeds 529 * 530 * Make sure that leaf is active. In the other words it can't be called 531 * with non-active leaf. It also removes class from the drop list. 532 */ 533 static inline void htb_deactivate(struct htb_sched *q, struct htb_class *cl) 534 { 535 WARN_ON(!cl->prio_activity); 536 537 htb_deactivate_prios(q, cl); 538 cl->prio_activity = 0; 539 list_del_init(&cl->un.leaf.drop_list); 540 } 541 542 static int htb_enqueue(struct sk_buff *skb, struct Qdisc *sch) 543 { 544 int uninitialized_var(ret); 545 struct htb_sched *q = qdisc_priv(sch); 546 struct htb_class *cl = htb_classify(skb, sch, &ret); 547 548 if (cl == HTB_DIRECT) { 549 /* enqueue to helper queue */ 550 if (q->direct_queue.qlen < q->direct_qlen) { 551 __skb_queue_tail(&q->direct_queue, skb); 552 q->direct_pkts++; 553 } else { 554 kfree_skb(skb); 555 sch->qstats.drops++; 556 return NET_XMIT_DROP; 557 } 558 #ifdef CONFIG_NET_CLS_ACT 559 } else if (!cl) { 560 if (ret & __NET_XMIT_BYPASS) 561 sch->qstats.drops++; 562 kfree_skb(skb); 563 return ret; 564 #endif 565 } else if ((ret = qdisc_enqueue(skb, cl->un.leaf.q)) != NET_XMIT_SUCCESS) { 566 if (net_xmit_drop_count(ret)) { 567 sch->qstats.drops++; 568 cl->qstats.drops++; 569 } 570 return ret; 571 } else { 572 cl->bstats.packets += 573 skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1; 574 cl->bstats.bytes += qdisc_pkt_len(skb); 575 htb_activate(q, cl); 576 } 577 578 sch->q.qlen++; 579 sch->bstats.packets += skb_is_gso(skb)?skb_shinfo(skb)->gso_segs:1; 580 sch->bstats.bytes += qdisc_pkt_len(skb); 581 return NET_XMIT_SUCCESS; 582 } 583 584 static inline void htb_accnt_tokens(struct htb_class *cl, int bytes, long diff) 585 { 586 long toks = diff + cl->tokens; 587 588 if (toks > cl->buffer) 589 toks = cl->buffer; 590 toks -= (long) qdisc_l2t(cl->rate, bytes); 591 if (toks <= -cl->mbuffer) 592 toks = 1 - cl->mbuffer; 593 594 cl->tokens = toks; 595 } 596 597 static inline void htb_accnt_ctokens(struct htb_class *cl, int bytes, long diff) 598 { 599 long toks = diff + cl->ctokens; 600 601 if (toks > cl->cbuffer) 602 toks = cl->cbuffer; 603 toks -= (long) qdisc_l2t(cl->ceil, bytes); 604 if (toks <= -cl->mbuffer) 605 toks = 1 - cl->mbuffer; 606 607 cl->ctokens = toks; 608 } 609 610 /** 611 * htb_charge_class - charges amount "bytes" to leaf and ancestors 612 * 613 * Routine assumes that packet "bytes" long was dequeued from leaf cl 614 * borrowing from "level". It accounts bytes to ceil leaky bucket for 615 * leaf and all ancestors and to rate bucket for ancestors at levels 616 * "level" and higher. It also handles possible change of mode resulting 617 * from the update. Note that mode can also increase here (MAY_BORROW to 618 * CAN_SEND) because we can use more precise clock that event queue here. 619 * In such case we remove class from event queue first. 620 */ 621 static void htb_charge_class(struct htb_sched *q, struct htb_class *cl, 622 int level, struct sk_buff *skb) 623 { 624 int bytes = qdisc_pkt_len(skb); 625 enum htb_cmode old_mode; 626 long diff; 627 628 while (cl) { 629 diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer); 630 if (cl->level >= level) { 631 if (cl->level == level) 632 cl->xstats.lends++; 633 htb_accnt_tokens(cl, bytes, diff); 634 } else { 635 cl->xstats.borrows++; 636 cl->tokens += diff; /* we moved t_c; update tokens */ 637 } 638 htb_accnt_ctokens(cl, bytes, diff); 639 cl->t_c = q->now; 640 641 old_mode = cl->cmode; 642 diff = 0; 643 htb_change_class_mode(q, cl, &diff); 644 if (old_mode != cl->cmode) { 645 if (old_mode != HTB_CAN_SEND) 646 htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level); 647 if (cl->cmode != HTB_CAN_SEND) 648 htb_add_to_wait_tree(q, cl, diff); 649 } 650 651 /* update byte stats except for leaves which are already updated */ 652 if (cl->level) { 653 cl->bstats.bytes += bytes; 654 cl->bstats.packets += skb_is_gso(skb)? 655 skb_shinfo(skb)->gso_segs:1; 656 } 657 cl = cl->parent; 658 } 659 } 660 661 /** 662 * htb_do_events - make mode changes to classes at the level 663 * 664 * Scans event queue for pending events and applies them. Returns time of 665 * next pending event (0 for no event in pq, q->now for too many events). 666 * Note: Applied are events whose have cl->pq_key <= q->now. 667 */ 668 static psched_time_t htb_do_events(struct htb_sched *q, int level, 669 unsigned long start) 670 { 671 /* don't run for longer than 2 jiffies; 2 is used instead of 672 1 to simplify things when jiffy is going to be incremented 673 too soon */ 674 unsigned long stop_at = start + 2; 675 while (time_before(jiffies, stop_at)) { 676 struct htb_class *cl; 677 long diff; 678 struct rb_node *p = rb_first(&q->wait_pq[level]); 679 680 if (!p) 681 return 0; 682 683 cl = rb_entry(p, struct htb_class, pq_node); 684 if (cl->pq_key > q->now) 685 return cl->pq_key; 686 687 htb_safe_rb_erase(p, q->wait_pq + level); 688 diff = psched_tdiff_bounded(q->now, cl->t_c, cl->mbuffer); 689 htb_change_class_mode(q, cl, &diff); 690 if (cl->cmode != HTB_CAN_SEND) 691 htb_add_to_wait_tree(q, cl, diff); 692 } 693 694 /* too much load - let's continue after a break for scheduling */ 695 if (!(q->warned & HTB_WARN_TOOMANYEVENTS)) { 696 printk(KERN_WARNING "htb: too many events!\n"); 697 q->warned |= HTB_WARN_TOOMANYEVENTS; 698 } 699 700 return q->now; 701 } 702 703 /* Returns class->node+prio from id-tree where classe's id is >= id. NULL 704 is no such one exists. */ 705 static struct rb_node *htb_id_find_next_upper(int prio, struct rb_node *n, 706 u32 id) 707 { 708 struct rb_node *r = NULL; 709 while (n) { 710 struct htb_class *cl = 711 rb_entry(n, struct htb_class, node[prio]); 712 713 if (id > cl->common.classid) { 714 n = n->rb_right; 715 } else if (id < cl->common.classid) { 716 r = n; 717 n = n->rb_left; 718 } else { 719 return n; 720 } 721 } 722 return r; 723 } 724 725 /** 726 * htb_lookup_leaf - returns next leaf class in DRR order 727 * 728 * Find leaf where current feed pointers points to. 729 */ 730 static struct htb_class *htb_lookup_leaf(struct rb_root *tree, int prio, 731 struct rb_node **pptr, u32 * pid) 732 { 733 int i; 734 struct { 735 struct rb_node *root; 736 struct rb_node **pptr; 737 u32 *pid; 738 } stk[TC_HTB_MAXDEPTH], *sp = stk; 739 740 BUG_ON(!tree->rb_node); 741 sp->root = tree->rb_node; 742 sp->pptr = pptr; 743 sp->pid = pid; 744 745 for (i = 0; i < 65535; i++) { 746 if (!*sp->pptr && *sp->pid) { 747 /* ptr was invalidated but id is valid - try to recover 748 the original or next ptr */ 749 *sp->pptr = 750 htb_id_find_next_upper(prio, sp->root, *sp->pid); 751 } 752 *sp->pid = 0; /* ptr is valid now so that remove this hint as it 753 can become out of date quickly */ 754 if (!*sp->pptr) { /* we are at right end; rewind & go up */ 755 *sp->pptr = sp->root; 756 while ((*sp->pptr)->rb_left) 757 *sp->pptr = (*sp->pptr)->rb_left; 758 if (sp > stk) { 759 sp--; 760 if (!*sp->pptr) { 761 WARN_ON(1); 762 return NULL; 763 } 764 htb_next_rb_node(sp->pptr); 765 } 766 } else { 767 struct htb_class *cl; 768 cl = rb_entry(*sp->pptr, struct htb_class, node[prio]); 769 if (!cl->level) 770 return cl; 771 (++sp)->root = cl->un.inner.feed[prio].rb_node; 772 sp->pptr = cl->un.inner.ptr + prio; 773 sp->pid = cl->un.inner.last_ptr_id + prio; 774 } 775 } 776 WARN_ON(1); 777 return NULL; 778 } 779 780 /* dequeues packet at given priority and level; call only if 781 you are sure that there is active class at prio/level */ 782 static struct sk_buff *htb_dequeue_tree(struct htb_sched *q, int prio, 783 int level) 784 { 785 struct sk_buff *skb = NULL; 786 struct htb_class *cl, *start; 787 /* look initial class up in the row */ 788 start = cl = htb_lookup_leaf(q->row[level] + prio, prio, 789 q->ptr[level] + prio, 790 q->last_ptr_id[level] + prio); 791 792 do { 793 next: 794 if (unlikely(!cl)) 795 return NULL; 796 797 /* class can be empty - it is unlikely but can be true if leaf 798 qdisc drops packets in enqueue routine or if someone used 799 graft operation on the leaf since last dequeue; 800 simply deactivate and skip such class */ 801 if (unlikely(cl->un.leaf.q->q.qlen == 0)) { 802 struct htb_class *next; 803 htb_deactivate(q, cl); 804 805 /* row/level might become empty */ 806 if ((q->row_mask[level] & (1 << prio)) == 0) 807 return NULL; 808 809 next = htb_lookup_leaf(q->row[level] + prio, 810 prio, q->ptr[level] + prio, 811 q->last_ptr_id[level] + prio); 812 813 if (cl == start) /* fix start if we just deleted it */ 814 start = next; 815 cl = next; 816 goto next; 817 } 818 819 skb = cl->un.leaf.q->dequeue(cl->un.leaf.q); 820 if (likely(skb != NULL)) 821 break; 822 823 qdisc_warn_nonwc("htb", cl->un.leaf.q); 824 htb_next_rb_node((level ? cl->parent->un.inner.ptr : q-> 825 ptr[0]) + prio); 826 cl = htb_lookup_leaf(q->row[level] + prio, prio, 827 q->ptr[level] + prio, 828 q->last_ptr_id[level] + prio); 829 830 } while (cl != start); 831 832 if (likely(skb != NULL)) { 833 cl->un.leaf.deficit[level] -= qdisc_pkt_len(skb); 834 if (cl->un.leaf.deficit[level] < 0) { 835 cl->un.leaf.deficit[level] += cl->quantum; 836 htb_next_rb_node((level ? cl->parent->un.inner.ptr : q-> 837 ptr[0]) + prio); 838 } 839 /* this used to be after charge_class but this constelation 840 gives us slightly better performance */ 841 if (!cl->un.leaf.q->q.qlen) 842 htb_deactivate(q, cl); 843 htb_charge_class(q, cl, level, skb); 844 } 845 return skb; 846 } 847 848 static struct sk_buff *htb_dequeue(struct Qdisc *sch) 849 { 850 struct sk_buff *skb = NULL; 851 struct htb_sched *q = qdisc_priv(sch); 852 int level; 853 psched_time_t next_event; 854 unsigned long start_at; 855 856 /* try to dequeue direct packets as high prio (!) to minimize cpu work */ 857 skb = __skb_dequeue(&q->direct_queue); 858 if (skb != NULL) { 859 sch->flags &= ~TCQ_F_THROTTLED; 860 sch->q.qlen--; 861 return skb; 862 } 863 864 if (!sch->q.qlen) 865 goto fin; 866 q->now = psched_get_time(); 867 start_at = jiffies; 868 869 next_event = q->now + 5 * PSCHED_TICKS_PER_SEC; 870 871 for (level = 0; level < TC_HTB_MAXDEPTH; level++) { 872 /* common case optimization - skip event handler quickly */ 873 int m; 874 psched_time_t event; 875 876 if (q->now >= q->near_ev_cache[level]) { 877 event = htb_do_events(q, level, start_at); 878 if (!event) 879 event = q->now + PSCHED_TICKS_PER_SEC; 880 q->near_ev_cache[level] = event; 881 } else 882 event = q->near_ev_cache[level]; 883 884 if (next_event > event) 885 next_event = event; 886 887 m = ~q->row_mask[level]; 888 while (m != (int)(-1)) { 889 int prio = ffz(m); 890 m |= 1 << prio; 891 skb = htb_dequeue_tree(q, prio, level); 892 if (likely(skb != NULL)) { 893 sch->q.qlen--; 894 sch->flags &= ~TCQ_F_THROTTLED; 895 goto fin; 896 } 897 } 898 } 899 sch->qstats.overlimits++; 900 if (likely(next_event > q->now)) 901 qdisc_watchdog_schedule(&q->watchdog, next_event); 902 else 903 schedule_work(&q->work); 904 fin: 905 return skb; 906 } 907 908 /* try to drop from each class (by prio) until one succeed */ 909 static unsigned int htb_drop(struct Qdisc *sch) 910 { 911 struct htb_sched *q = qdisc_priv(sch); 912 int prio; 913 914 for (prio = TC_HTB_NUMPRIO - 1; prio >= 0; prio--) { 915 struct list_head *p; 916 list_for_each(p, q->drops + prio) { 917 struct htb_class *cl = list_entry(p, struct htb_class, 918 un.leaf.drop_list); 919 unsigned int len; 920 if (cl->un.leaf.q->ops->drop && 921 (len = cl->un.leaf.q->ops->drop(cl->un.leaf.q))) { 922 sch->q.qlen--; 923 if (!cl->un.leaf.q->q.qlen) 924 htb_deactivate(q, cl); 925 return len; 926 } 927 } 928 } 929 return 0; 930 } 931 932 /* reset all classes */ 933 /* always caled under BH & queue lock */ 934 static void htb_reset(struct Qdisc *sch) 935 { 936 struct htb_sched *q = qdisc_priv(sch); 937 struct htb_class *cl; 938 struct hlist_node *n; 939 unsigned int i; 940 941 for (i = 0; i < q->clhash.hashsize; i++) { 942 hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) { 943 if (cl->level) 944 memset(&cl->un.inner, 0, sizeof(cl->un.inner)); 945 else { 946 if (cl->un.leaf.q) 947 qdisc_reset(cl->un.leaf.q); 948 INIT_LIST_HEAD(&cl->un.leaf.drop_list); 949 } 950 cl->prio_activity = 0; 951 cl->cmode = HTB_CAN_SEND; 952 953 } 954 } 955 qdisc_watchdog_cancel(&q->watchdog); 956 __skb_queue_purge(&q->direct_queue); 957 sch->q.qlen = 0; 958 memset(q->row, 0, sizeof(q->row)); 959 memset(q->row_mask, 0, sizeof(q->row_mask)); 960 memset(q->wait_pq, 0, sizeof(q->wait_pq)); 961 memset(q->ptr, 0, sizeof(q->ptr)); 962 for (i = 0; i < TC_HTB_NUMPRIO; i++) 963 INIT_LIST_HEAD(q->drops + i); 964 } 965 966 static const struct nla_policy htb_policy[TCA_HTB_MAX + 1] = { 967 [TCA_HTB_PARMS] = { .len = sizeof(struct tc_htb_opt) }, 968 [TCA_HTB_INIT] = { .len = sizeof(struct tc_htb_glob) }, 969 [TCA_HTB_CTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 970 [TCA_HTB_RTAB] = { .type = NLA_BINARY, .len = TC_RTAB_SIZE }, 971 }; 972 973 static void htb_work_func(struct work_struct *work) 974 { 975 struct htb_sched *q = container_of(work, struct htb_sched, work); 976 struct Qdisc *sch = q->watchdog.qdisc; 977 978 __netif_schedule(qdisc_root(sch)); 979 } 980 981 static int htb_init(struct Qdisc *sch, struct nlattr *opt) 982 { 983 struct htb_sched *q = qdisc_priv(sch); 984 struct nlattr *tb[TCA_HTB_INIT + 1]; 985 struct tc_htb_glob *gopt; 986 int err; 987 int i; 988 989 if (!opt) 990 return -EINVAL; 991 992 err = nla_parse_nested(tb, TCA_HTB_INIT, opt, htb_policy); 993 if (err < 0) 994 return err; 995 996 if (tb[TCA_HTB_INIT] == NULL) { 997 printk(KERN_ERR "HTB: hey probably you have bad tc tool ?\n"); 998 return -EINVAL; 999 } 1000 gopt = nla_data(tb[TCA_HTB_INIT]); 1001 if (gopt->version != HTB_VER >> 16) { 1002 printk(KERN_ERR 1003 "HTB: need tc/htb version %d (minor is %d), you have %d\n", 1004 HTB_VER >> 16, HTB_VER & 0xffff, gopt->version); 1005 return -EINVAL; 1006 } 1007 1008 err = qdisc_class_hash_init(&q->clhash); 1009 if (err < 0) 1010 return err; 1011 for (i = 0; i < TC_HTB_NUMPRIO; i++) 1012 INIT_LIST_HEAD(q->drops + i); 1013 1014 qdisc_watchdog_init(&q->watchdog, sch); 1015 INIT_WORK(&q->work, htb_work_func); 1016 skb_queue_head_init(&q->direct_queue); 1017 1018 q->direct_qlen = qdisc_dev(sch)->tx_queue_len; 1019 if (q->direct_qlen < 2) /* some devices have zero tx_queue_len */ 1020 q->direct_qlen = 2; 1021 1022 if ((q->rate2quantum = gopt->rate2quantum) < 1) 1023 q->rate2quantum = 1; 1024 q->defcls = gopt->defcls; 1025 1026 return 0; 1027 } 1028 1029 static int htb_dump(struct Qdisc *sch, struct sk_buff *skb) 1030 { 1031 spinlock_t *root_lock = qdisc_root_sleeping_lock(sch); 1032 struct htb_sched *q = qdisc_priv(sch); 1033 struct nlattr *nest; 1034 struct tc_htb_glob gopt; 1035 1036 spin_lock_bh(root_lock); 1037 1038 gopt.direct_pkts = q->direct_pkts; 1039 gopt.version = HTB_VER; 1040 gopt.rate2quantum = q->rate2quantum; 1041 gopt.defcls = q->defcls; 1042 gopt.debug = 0; 1043 1044 nest = nla_nest_start(skb, TCA_OPTIONS); 1045 if (nest == NULL) 1046 goto nla_put_failure; 1047 NLA_PUT(skb, TCA_HTB_INIT, sizeof(gopt), &gopt); 1048 nla_nest_end(skb, nest); 1049 1050 spin_unlock_bh(root_lock); 1051 return skb->len; 1052 1053 nla_put_failure: 1054 spin_unlock_bh(root_lock); 1055 nla_nest_cancel(skb, nest); 1056 return -1; 1057 } 1058 1059 static int htb_dump_class(struct Qdisc *sch, unsigned long arg, 1060 struct sk_buff *skb, struct tcmsg *tcm) 1061 { 1062 struct htb_class *cl = (struct htb_class *)arg; 1063 spinlock_t *root_lock = qdisc_root_sleeping_lock(sch); 1064 struct nlattr *nest; 1065 struct tc_htb_opt opt; 1066 1067 spin_lock_bh(root_lock); 1068 tcm->tcm_parent = cl->parent ? cl->parent->common.classid : TC_H_ROOT; 1069 tcm->tcm_handle = cl->common.classid; 1070 if (!cl->level && cl->un.leaf.q) 1071 tcm->tcm_info = cl->un.leaf.q->handle; 1072 1073 nest = nla_nest_start(skb, TCA_OPTIONS); 1074 if (nest == NULL) 1075 goto nla_put_failure; 1076 1077 memset(&opt, 0, sizeof(opt)); 1078 1079 opt.rate = cl->rate->rate; 1080 opt.buffer = cl->buffer; 1081 opt.ceil = cl->ceil->rate; 1082 opt.cbuffer = cl->cbuffer; 1083 opt.quantum = cl->quantum; 1084 opt.prio = cl->prio; 1085 opt.level = cl->level; 1086 NLA_PUT(skb, TCA_HTB_PARMS, sizeof(opt), &opt); 1087 1088 nla_nest_end(skb, nest); 1089 spin_unlock_bh(root_lock); 1090 return skb->len; 1091 1092 nla_put_failure: 1093 spin_unlock_bh(root_lock); 1094 nla_nest_cancel(skb, nest); 1095 return -1; 1096 } 1097 1098 static int 1099 htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d) 1100 { 1101 struct htb_class *cl = (struct htb_class *)arg; 1102 1103 if (!cl->level && cl->un.leaf.q) 1104 cl->qstats.qlen = cl->un.leaf.q->q.qlen; 1105 cl->xstats.tokens = cl->tokens; 1106 cl->xstats.ctokens = cl->ctokens; 1107 1108 if (gnet_stats_copy_basic(d, &cl->bstats) < 0 || 1109 gnet_stats_copy_rate_est(d, NULL, &cl->rate_est) < 0 || 1110 gnet_stats_copy_queue(d, &cl->qstats) < 0) 1111 return -1; 1112 1113 return gnet_stats_copy_app(d, &cl->xstats, sizeof(cl->xstats)); 1114 } 1115 1116 static int htb_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new, 1117 struct Qdisc **old) 1118 { 1119 struct htb_class *cl = (struct htb_class *)arg; 1120 1121 if (cl->level) 1122 return -EINVAL; 1123 if (new == NULL && 1124 (new = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, 1125 cl->common.classid)) == NULL) 1126 return -ENOBUFS; 1127 1128 sch_tree_lock(sch); 1129 *old = cl->un.leaf.q; 1130 cl->un.leaf.q = new; 1131 if (*old != NULL) { 1132 qdisc_tree_decrease_qlen(*old, (*old)->q.qlen); 1133 qdisc_reset(*old); 1134 } 1135 sch_tree_unlock(sch); 1136 return 0; 1137 } 1138 1139 static struct Qdisc *htb_leaf(struct Qdisc *sch, unsigned long arg) 1140 { 1141 struct htb_class *cl = (struct htb_class *)arg; 1142 return !cl->level ? cl->un.leaf.q : NULL; 1143 } 1144 1145 static void htb_qlen_notify(struct Qdisc *sch, unsigned long arg) 1146 { 1147 struct htb_class *cl = (struct htb_class *)arg; 1148 1149 if (cl->un.leaf.q->q.qlen == 0) 1150 htb_deactivate(qdisc_priv(sch), cl); 1151 } 1152 1153 static unsigned long htb_get(struct Qdisc *sch, u32 classid) 1154 { 1155 struct htb_class *cl = htb_find(classid, sch); 1156 if (cl) 1157 cl->refcnt++; 1158 return (unsigned long)cl; 1159 } 1160 1161 static inline int htb_parent_last_child(struct htb_class *cl) 1162 { 1163 if (!cl->parent) 1164 /* the root class */ 1165 return 0; 1166 if (cl->parent->children > 1) 1167 /* not the last child */ 1168 return 0; 1169 return 1; 1170 } 1171 1172 static void htb_parent_to_leaf(struct htb_sched *q, struct htb_class *cl, 1173 struct Qdisc *new_q) 1174 { 1175 struct htb_class *parent = cl->parent; 1176 1177 WARN_ON(cl->level || !cl->un.leaf.q || cl->prio_activity); 1178 1179 if (parent->cmode != HTB_CAN_SEND) 1180 htb_safe_rb_erase(&parent->pq_node, q->wait_pq + parent->level); 1181 1182 parent->level = 0; 1183 memset(&parent->un.inner, 0, sizeof(parent->un.inner)); 1184 INIT_LIST_HEAD(&parent->un.leaf.drop_list); 1185 parent->un.leaf.q = new_q ? new_q : &noop_qdisc; 1186 parent->tokens = parent->buffer; 1187 parent->ctokens = parent->cbuffer; 1188 parent->t_c = psched_get_time(); 1189 parent->cmode = HTB_CAN_SEND; 1190 } 1191 1192 static void htb_destroy_class(struct Qdisc *sch, struct htb_class *cl) 1193 { 1194 if (!cl->level) { 1195 WARN_ON(!cl->un.leaf.q); 1196 qdisc_destroy(cl->un.leaf.q); 1197 } 1198 gen_kill_estimator(&cl->bstats, &cl->rate_est); 1199 qdisc_put_rtab(cl->rate); 1200 qdisc_put_rtab(cl->ceil); 1201 1202 tcf_destroy_chain(&cl->filter_list); 1203 kfree(cl); 1204 } 1205 1206 static void htb_destroy(struct Qdisc *sch) 1207 { 1208 struct htb_sched *q = qdisc_priv(sch); 1209 struct hlist_node *n, *next; 1210 struct htb_class *cl; 1211 unsigned int i; 1212 1213 cancel_work_sync(&q->work); 1214 qdisc_watchdog_cancel(&q->watchdog); 1215 /* This line used to be after htb_destroy_class call below 1216 and surprisingly it worked in 2.4. But it must precede it 1217 because filter need its target class alive to be able to call 1218 unbind_filter on it (without Oops). */ 1219 tcf_destroy_chain(&q->filter_list); 1220 1221 for (i = 0; i < q->clhash.hashsize; i++) { 1222 hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) 1223 tcf_destroy_chain(&cl->filter_list); 1224 } 1225 for (i = 0; i < q->clhash.hashsize; i++) { 1226 hlist_for_each_entry_safe(cl, n, next, &q->clhash.hash[i], 1227 common.hnode) 1228 htb_destroy_class(sch, cl); 1229 } 1230 qdisc_class_hash_destroy(&q->clhash); 1231 __skb_queue_purge(&q->direct_queue); 1232 } 1233 1234 static int htb_delete(struct Qdisc *sch, unsigned long arg) 1235 { 1236 struct htb_sched *q = qdisc_priv(sch); 1237 struct htb_class *cl = (struct htb_class *)arg; 1238 unsigned int qlen; 1239 struct Qdisc *new_q = NULL; 1240 int last_child = 0; 1241 1242 // TODO: why don't allow to delete subtree ? references ? does 1243 // tc subsys quarantee us that in htb_destroy it holds no class 1244 // refs so that we can remove children safely there ? 1245 if (cl->children || cl->filter_cnt) 1246 return -EBUSY; 1247 1248 if (!cl->level && htb_parent_last_child(cl)) { 1249 new_q = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops, 1250 cl->parent->common.classid); 1251 last_child = 1; 1252 } 1253 1254 sch_tree_lock(sch); 1255 1256 if (!cl->level) { 1257 qlen = cl->un.leaf.q->q.qlen; 1258 qdisc_reset(cl->un.leaf.q); 1259 qdisc_tree_decrease_qlen(cl->un.leaf.q, qlen); 1260 } 1261 1262 /* delete from hash and active; remainder in destroy_class */ 1263 qdisc_class_hash_remove(&q->clhash, &cl->common); 1264 if (cl->parent) 1265 cl->parent->children--; 1266 1267 if (cl->prio_activity) 1268 htb_deactivate(q, cl); 1269 1270 if (cl->cmode != HTB_CAN_SEND) 1271 htb_safe_rb_erase(&cl->pq_node, q->wait_pq + cl->level); 1272 1273 if (last_child) 1274 htb_parent_to_leaf(q, cl, new_q); 1275 1276 BUG_ON(--cl->refcnt == 0); 1277 /* 1278 * This shouldn't happen: we "hold" one cops->get() when called 1279 * from tc_ctl_tclass; the destroy method is done from cops->put(). 1280 */ 1281 1282 sch_tree_unlock(sch); 1283 return 0; 1284 } 1285 1286 static void htb_put(struct Qdisc *sch, unsigned long arg) 1287 { 1288 struct htb_class *cl = (struct htb_class *)arg; 1289 1290 if (--cl->refcnt == 0) 1291 htb_destroy_class(sch, cl); 1292 } 1293 1294 static int htb_change_class(struct Qdisc *sch, u32 classid, 1295 u32 parentid, struct nlattr **tca, 1296 unsigned long *arg) 1297 { 1298 int err = -EINVAL; 1299 struct htb_sched *q = qdisc_priv(sch); 1300 struct htb_class *cl = (struct htb_class *)*arg, *parent; 1301 struct nlattr *opt = tca[TCA_OPTIONS]; 1302 struct qdisc_rate_table *rtab = NULL, *ctab = NULL; 1303 struct nlattr *tb[__TCA_HTB_MAX]; 1304 struct tc_htb_opt *hopt; 1305 1306 /* extract all subattrs from opt attr */ 1307 if (!opt) 1308 goto failure; 1309 1310 err = nla_parse_nested(tb, TCA_HTB_MAX, opt, htb_policy); 1311 if (err < 0) 1312 goto failure; 1313 1314 err = -EINVAL; 1315 if (tb[TCA_HTB_PARMS] == NULL) 1316 goto failure; 1317 1318 parent = parentid == TC_H_ROOT ? NULL : htb_find(parentid, sch); 1319 1320 hopt = nla_data(tb[TCA_HTB_PARMS]); 1321 1322 rtab = qdisc_get_rtab(&hopt->rate, tb[TCA_HTB_RTAB]); 1323 ctab = qdisc_get_rtab(&hopt->ceil, tb[TCA_HTB_CTAB]); 1324 if (!rtab || !ctab) 1325 goto failure; 1326 1327 if (!cl) { /* new class */ 1328 struct Qdisc *new_q; 1329 int prio; 1330 struct { 1331 struct nlattr nla; 1332 struct gnet_estimator opt; 1333 } est = { 1334 .nla = { 1335 .nla_len = nla_attr_size(sizeof(est.opt)), 1336 .nla_type = TCA_RATE, 1337 }, 1338 .opt = { 1339 /* 4s interval, 16s averaging constant */ 1340 .interval = 2, 1341 .ewma_log = 2, 1342 }, 1343 }; 1344 1345 /* check for valid classid */ 1346 if (!classid || TC_H_MAJ(classid ^ sch->handle) || 1347 htb_find(classid, sch)) 1348 goto failure; 1349 1350 /* check maximal depth */ 1351 if (parent && parent->parent && parent->parent->level < 2) { 1352 printk(KERN_ERR "htb: tree is too deep\n"); 1353 goto failure; 1354 } 1355 err = -ENOBUFS; 1356 if ((cl = kzalloc(sizeof(*cl), GFP_KERNEL)) == NULL) 1357 goto failure; 1358 1359 err = gen_new_estimator(&cl->bstats, &cl->rate_est, 1360 qdisc_root_sleeping_lock(sch), 1361 tca[TCA_RATE] ? : &est.nla); 1362 if (err) { 1363 kfree(cl); 1364 goto failure; 1365 } 1366 1367 cl->refcnt = 1; 1368 cl->children = 0; 1369 INIT_LIST_HEAD(&cl->un.leaf.drop_list); 1370 RB_CLEAR_NODE(&cl->pq_node); 1371 1372 for (prio = 0; prio < TC_HTB_NUMPRIO; prio++) 1373 RB_CLEAR_NODE(&cl->node[prio]); 1374 1375 /* create leaf qdisc early because it uses kmalloc(GFP_KERNEL) 1376 so that can't be used inside of sch_tree_lock 1377 -- thanks to Karlis Peisenieks */ 1378 new_q = qdisc_create_dflt(sch->dev_queue, 1379 &pfifo_qdisc_ops, classid); 1380 sch_tree_lock(sch); 1381 if (parent && !parent->level) { 1382 unsigned int qlen = parent->un.leaf.q->q.qlen; 1383 1384 /* turn parent into inner node */ 1385 qdisc_reset(parent->un.leaf.q); 1386 qdisc_tree_decrease_qlen(parent->un.leaf.q, qlen); 1387 qdisc_destroy(parent->un.leaf.q); 1388 if (parent->prio_activity) 1389 htb_deactivate(q, parent); 1390 1391 /* remove from evt list because of level change */ 1392 if (parent->cmode != HTB_CAN_SEND) { 1393 htb_safe_rb_erase(&parent->pq_node, q->wait_pq); 1394 parent->cmode = HTB_CAN_SEND; 1395 } 1396 parent->level = (parent->parent ? parent->parent->level 1397 : TC_HTB_MAXDEPTH) - 1; 1398 memset(&parent->un.inner, 0, sizeof(parent->un.inner)); 1399 } 1400 /* leaf (we) needs elementary qdisc */ 1401 cl->un.leaf.q = new_q ? new_q : &noop_qdisc; 1402 1403 cl->common.classid = classid; 1404 cl->parent = parent; 1405 1406 /* set class to be in HTB_CAN_SEND state */ 1407 cl->tokens = hopt->buffer; 1408 cl->ctokens = hopt->cbuffer; 1409 cl->mbuffer = 60 * PSCHED_TICKS_PER_SEC; /* 1min */ 1410 cl->t_c = psched_get_time(); 1411 cl->cmode = HTB_CAN_SEND; 1412 1413 /* attach to the hash list and parent's family */ 1414 qdisc_class_hash_insert(&q->clhash, &cl->common); 1415 if (parent) 1416 parent->children++; 1417 } else { 1418 if (tca[TCA_RATE]) { 1419 err = gen_replace_estimator(&cl->bstats, &cl->rate_est, 1420 qdisc_root_sleeping_lock(sch), 1421 tca[TCA_RATE]); 1422 if (err) 1423 return err; 1424 } 1425 sch_tree_lock(sch); 1426 } 1427 1428 /* it used to be a nasty bug here, we have to check that node 1429 is really leaf before changing cl->un.leaf ! */ 1430 if (!cl->level) { 1431 cl->quantum = rtab->rate.rate / q->rate2quantum; 1432 if (!hopt->quantum && cl->quantum < 1000) { 1433 printk(KERN_WARNING 1434 "HTB: quantum of class %X is small. Consider r2q change.\n", 1435 cl->common.classid); 1436 cl->quantum = 1000; 1437 } 1438 if (!hopt->quantum && cl->quantum > 200000) { 1439 printk(KERN_WARNING 1440 "HTB: quantum of class %X is big. Consider r2q change.\n", 1441 cl->common.classid); 1442 cl->quantum = 200000; 1443 } 1444 if (hopt->quantum) 1445 cl->quantum = hopt->quantum; 1446 if ((cl->prio = hopt->prio) >= TC_HTB_NUMPRIO) 1447 cl->prio = TC_HTB_NUMPRIO - 1; 1448 } 1449 1450 cl->buffer = hopt->buffer; 1451 cl->cbuffer = hopt->cbuffer; 1452 if (cl->rate) 1453 qdisc_put_rtab(cl->rate); 1454 cl->rate = rtab; 1455 if (cl->ceil) 1456 qdisc_put_rtab(cl->ceil); 1457 cl->ceil = ctab; 1458 sch_tree_unlock(sch); 1459 1460 qdisc_class_hash_grow(sch, &q->clhash); 1461 1462 *arg = (unsigned long)cl; 1463 return 0; 1464 1465 failure: 1466 if (rtab) 1467 qdisc_put_rtab(rtab); 1468 if (ctab) 1469 qdisc_put_rtab(ctab); 1470 return err; 1471 } 1472 1473 static struct tcf_proto **htb_find_tcf(struct Qdisc *sch, unsigned long arg) 1474 { 1475 struct htb_sched *q = qdisc_priv(sch); 1476 struct htb_class *cl = (struct htb_class *)arg; 1477 struct tcf_proto **fl = cl ? &cl->filter_list : &q->filter_list; 1478 1479 return fl; 1480 } 1481 1482 static unsigned long htb_bind_filter(struct Qdisc *sch, unsigned long parent, 1483 u32 classid) 1484 { 1485 struct htb_class *cl = htb_find(classid, sch); 1486 1487 /*if (cl && !cl->level) return 0; 1488 The line above used to be there to prevent attaching filters to 1489 leaves. But at least tc_index filter uses this just to get class 1490 for other reasons so that we have to allow for it. 1491 ---- 1492 19.6.2002 As Werner explained it is ok - bind filter is just 1493 another way to "lock" the class - unlike "get" this lock can 1494 be broken by class during destroy IIUC. 1495 */ 1496 if (cl) 1497 cl->filter_cnt++; 1498 return (unsigned long)cl; 1499 } 1500 1501 static void htb_unbind_filter(struct Qdisc *sch, unsigned long arg) 1502 { 1503 struct htb_class *cl = (struct htb_class *)arg; 1504 1505 if (cl) 1506 cl->filter_cnt--; 1507 } 1508 1509 static void htb_walk(struct Qdisc *sch, struct qdisc_walker *arg) 1510 { 1511 struct htb_sched *q = qdisc_priv(sch); 1512 struct htb_class *cl; 1513 struct hlist_node *n; 1514 unsigned int i; 1515 1516 if (arg->stop) 1517 return; 1518 1519 for (i = 0; i < q->clhash.hashsize; i++) { 1520 hlist_for_each_entry(cl, n, &q->clhash.hash[i], common.hnode) { 1521 if (arg->count < arg->skip) { 1522 arg->count++; 1523 continue; 1524 } 1525 if (arg->fn(sch, (unsigned long)cl, arg) < 0) { 1526 arg->stop = 1; 1527 return; 1528 } 1529 arg->count++; 1530 } 1531 } 1532 } 1533 1534 static const struct Qdisc_class_ops htb_class_ops = { 1535 .graft = htb_graft, 1536 .leaf = htb_leaf, 1537 .qlen_notify = htb_qlen_notify, 1538 .get = htb_get, 1539 .put = htb_put, 1540 .change = htb_change_class, 1541 .delete = htb_delete, 1542 .walk = htb_walk, 1543 .tcf_chain = htb_find_tcf, 1544 .bind_tcf = htb_bind_filter, 1545 .unbind_tcf = htb_unbind_filter, 1546 .dump = htb_dump_class, 1547 .dump_stats = htb_dump_class_stats, 1548 }; 1549 1550 static struct Qdisc_ops htb_qdisc_ops __read_mostly = { 1551 .cl_ops = &htb_class_ops, 1552 .id = "htb", 1553 .priv_size = sizeof(struct htb_sched), 1554 .enqueue = htb_enqueue, 1555 .dequeue = htb_dequeue, 1556 .peek = qdisc_peek_dequeued, 1557 .drop = htb_drop, 1558 .init = htb_init, 1559 .reset = htb_reset, 1560 .destroy = htb_destroy, 1561 .dump = htb_dump, 1562 .owner = THIS_MODULE, 1563 }; 1564 1565 static int __init htb_module_init(void) 1566 { 1567 return register_qdisc(&htb_qdisc_ops); 1568 } 1569 static void __exit htb_module_exit(void) 1570 { 1571 unregister_qdisc(&htb_qdisc_ops); 1572 } 1573 1574 module_init(htb_module_init) 1575 module_exit(htb_module_exit) 1576 MODULE_LICENSE("GPL"); 1577