1 // SPDX-License-Identifier: GPL-2.0-only 2 /* Flow Queue PIE discipline 3 * 4 * Copyright (C) 2019 Mohit P. Tahiliani <tahiliani@nitk.edu.in> 5 * Copyright (C) 2019 Sachin D. Patil <sdp.sachin@gmail.com> 6 * Copyright (C) 2019 V. Saicharan <vsaicharan1998@gmail.com> 7 * Copyright (C) 2019 Mohit Bhasi <mohitbhasi1998@gmail.com> 8 * Copyright (C) 2019 Leslie Monis <lesliemonis@gmail.com> 9 * Copyright (C) 2019 Gautam Ramakrishnan <gautamramk@gmail.com> 10 */ 11 12 #include <linux/jhash.h> 13 #include <linux/sizes.h> 14 #include <linux/vmalloc.h> 15 #include <net/pkt_cls.h> 16 #include <net/pie.h> 17 18 /* Flow Queue PIE 19 * 20 * Principles: 21 * - Packets are classified on flows. 22 * - This is a Stochastic model (as we use a hash, several flows might 23 * be hashed to the same slot) 24 * - Each flow has a PIE managed queue. 25 * - Flows are linked onto two (Round Robin) lists, 26 * so that new flows have priority on old ones. 27 * - For a given flow, packets are not reordered. 28 * - Drops during enqueue only. 29 * - ECN capability is off by default. 30 * - ECN threshold (if ECN is enabled) is at 10% by default. 31 * - Uses timestamps to calculate queue delay by default. 32 */ 33 34 /** 35 * struct fq_pie_flow - contains data for each flow 36 * @vars: pie vars associated with the flow 37 * @deficit: number of remaining byte credits 38 * @backlog: size of data in the flow 39 * @qlen: number of packets in the flow 40 * @flowchain: flowchain for the flow 41 * @head: first packet in the flow 42 * @tail: last packet in the flow 43 */ 44 struct fq_pie_flow { 45 struct pie_vars vars; 46 s32 deficit; 47 u32 backlog; 48 u32 qlen; 49 struct list_head flowchain; 50 struct sk_buff *head; 51 struct sk_buff *tail; 52 }; 53 54 struct fq_pie_sched_data { 55 struct tcf_proto __rcu *filter_list; /* optional external classifier */ 56 struct tcf_block *block; 57 struct fq_pie_flow *flows; 58 struct Qdisc *sch; 59 struct list_head old_flows; 60 struct list_head new_flows; 61 struct pie_params p_params; 62 u32 ecn_prob; 63 u32 flows_cnt; 64 u32 quantum; 65 u32 memory_limit; 66 u32 new_flow_count; 67 u32 memory_usage; 68 u32 overmemory; 69 struct pie_stats stats; 70 struct timer_list adapt_timer; 71 }; 72 73 static unsigned int fq_pie_hash(const struct fq_pie_sched_data *q, 74 struct sk_buff *skb) 75 { 76 return reciprocal_scale(skb_get_hash(skb), q->flows_cnt); 77 } 78 79 static unsigned int fq_pie_classify(struct sk_buff *skb, struct Qdisc *sch, 80 int *qerr) 81 { 82 struct fq_pie_sched_data *q = qdisc_priv(sch); 83 struct tcf_proto *filter; 84 struct tcf_result res; 85 int result; 86 87 if (TC_H_MAJ(skb->priority) == sch->handle && 88 TC_H_MIN(skb->priority) > 0 && 89 TC_H_MIN(skb->priority) <= q->flows_cnt) 90 return TC_H_MIN(skb->priority); 91 92 filter = rcu_dereference_bh(q->filter_list); 93 if (!filter) 94 return fq_pie_hash(q, skb) + 1; 95 96 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 97 result = tcf_classify(skb, NULL, filter, &res, false); 98 if (result >= 0) { 99 #ifdef CONFIG_NET_CLS_ACT 100 switch (result) { 101 case TC_ACT_STOLEN: 102 case TC_ACT_QUEUED: 103 case TC_ACT_TRAP: 104 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; 105 fallthrough; 106 case TC_ACT_SHOT: 107 return 0; 108 } 109 #endif 110 if (TC_H_MIN(res.classid) <= q->flows_cnt) 111 return TC_H_MIN(res.classid); 112 } 113 return 0; 114 } 115 116 /* add skb to flow queue (tail add) */ 117 static inline void flow_queue_add(struct fq_pie_flow *flow, 118 struct sk_buff *skb) 119 { 120 if (!flow->head) 121 flow->head = skb; 122 else 123 flow->tail->next = skb; 124 flow->tail = skb; 125 skb->next = NULL; 126 } 127 128 static int fq_pie_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch, 129 struct sk_buff **to_free) 130 { 131 struct fq_pie_sched_data *q = qdisc_priv(sch); 132 struct fq_pie_flow *sel_flow; 133 int ret; 134 u8 memory_limited = false; 135 u8 enqueue = false; 136 u32 pkt_len; 137 u32 idx; 138 139 /* Classifies packet into corresponding flow */ 140 idx = fq_pie_classify(skb, sch, &ret); 141 if (idx == 0) { 142 if (ret & __NET_XMIT_BYPASS) 143 qdisc_qstats_drop(sch); 144 __qdisc_drop(skb, to_free); 145 return ret; 146 } 147 idx--; 148 149 sel_flow = &q->flows[idx]; 150 /* Checks whether adding a new packet would exceed memory limit */ 151 get_pie_cb(skb)->mem_usage = skb->truesize; 152 memory_limited = q->memory_usage > q->memory_limit + skb->truesize; 153 154 /* Checks if the qdisc is full */ 155 if (unlikely(qdisc_qlen(sch) >= sch->limit)) { 156 q->stats.overlimit++; 157 goto out; 158 } else if (unlikely(memory_limited)) { 159 q->overmemory++; 160 } 161 162 if (!pie_drop_early(sch, &q->p_params, &sel_flow->vars, 163 sel_flow->backlog, skb->len)) { 164 enqueue = true; 165 } else if (q->p_params.ecn && 166 sel_flow->vars.prob <= (MAX_PROB / 100) * q->ecn_prob && 167 INET_ECN_set_ce(skb)) { 168 /* If packet is ecn capable, mark it if drop probability 169 * is lower than the parameter ecn_prob, else drop it. 170 */ 171 q->stats.ecn_mark++; 172 enqueue = true; 173 } 174 if (enqueue) { 175 /* Set enqueue time only when dq_rate_estimator is disabled. */ 176 if (!q->p_params.dq_rate_estimator) 177 pie_set_enqueue_time(skb); 178 179 pkt_len = qdisc_pkt_len(skb); 180 q->stats.packets_in++; 181 q->memory_usage += skb->truesize; 182 sch->qstats.backlog += pkt_len; 183 sch->q.qlen++; 184 flow_queue_add(sel_flow, skb); 185 if (list_empty(&sel_flow->flowchain)) { 186 list_add_tail(&sel_flow->flowchain, &q->new_flows); 187 q->new_flow_count++; 188 sel_flow->deficit = q->quantum; 189 sel_flow->qlen = 0; 190 sel_flow->backlog = 0; 191 } 192 sel_flow->qlen++; 193 sel_flow->backlog += pkt_len; 194 return NET_XMIT_SUCCESS; 195 } 196 out: 197 q->stats.dropped++; 198 sel_flow->vars.accu_prob = 0; 199 __qdisc_drop(skb, to_free); 200 qdisc_qstats_drop(sch); 201 return NET_XMIT_CN; 202 } 203 204 static const struct nla_policy fq_pie_policy[TCA_FQ_PIE_MAX + 1] = { 205 [TCA_FQ_PIE_LIMIT] = {.type = NLA_U32}, 206 [TCA_FQ_PIE_FLOWS] = {.type = NLA_U32}, 207 [TCA_FQ_PIE_TARGET] = {.type = NLA_U32}, 208 [TCA_FQ_PIE_TUPDATE] = {.type = NLA_U32}, 209 [TCA_FQ_PIE_ALPHA] = {.type = NLA_U32}, 210 [TCA_FQ_PIE_BETA] = {.type = NLA_U32}, 211 [TCA_FQ_PIE_QUANTUM] = {.type = NLA_U32}, 212 [TCA_FQ_PIE_MEMORY_LIMIT] = {.type = NLA_U32}, 213 [TCA_FQ_PIE_ECN_PROB] = {.type = NLA_U32}, 214 [TCA_FQ_PIE_ECN] = {.type = NLA_U32}, 215 [TCA_FQ_PIE_BYTEMODE] = {.type = NLA_U32}, 216 [TCA_FQ_PIE_DQ_RATE_ESTIMATOR] = {.type = NLA_U32}, 217 }; 218 219 static inline struct sk_buff *dequeue_head(struct fq_pie_flow *flow) 220 { 221 struct sk_buff *skb = flow->head; 222 223 flow->head = skb->next; 224 skb->next = NULL; 225 return skb; 226 } 227 228 static struct sk_buff *fq_pie_qdisc_dequeue(struct Qdisc *sch) 229 { 230 struct fq_pie_sched_data *q = qdisc_priv(sch); 231 struct sk_buff *skb = NULL; 232 struct fq_pie_flow *flow; 233 struct list_head *head; 234 u32 pkt_len; 235 236 begin: 237 head = &q->new_flows; 238 if (list_empty(head)) { 239 head = &q->old_flows; 240 if (list_empty(head)) 241 return NULL; 242 } 243 244 flow = list_first_entry(head, struct fq_pie_flow, flowchain); 245 /* Flow has exhausted all its credits */ 246 if (flow->deficit <= 0) { 247 flow->deficit += q->quantum; 248 list_move_tail(&flow->flowchain, &q->old_flows); 249 goto begin; 250 } 251 252 if (flow->head) { 253 skb = dequeue_head(flow); 254 pkt_len = qdisc_pkt_len(skb); 255 sch->qstats.backlog -= pkt_len; 256 sch->q.qlen--; 257 qdisc_bstats_update(sch, skb); 258 } 259 260 if (!skb) { 261 /* force a pass through old_flows to prevent starvation */ 262 if (head == &q->new_flows && !list_empty(&q->old_flows)) 263 list_move_tail(&flow->flowchain, &q->old_flows); 264 else 265 list_del_init(&flow->flowchain); 266 goto begin; 267 } 268 269 flow->qlen--; 270 flow->deficit -= pkt_len; 271 flow->backlog -= pkt_len; 272 q->memory_usage -= get_pie_cb(skb)->mem_usage; 273 pie_process_dequeue(skb, &q->p_params, &flow->vars, flow->backlog); 274 return skb; 275 } 276 277 static int fq_pie_change(struct Qdisc *sch, struct nlattr *opt, 278 struct netlink_ext_ack *extack) 279 { 280 struct fq_pie_sched_data *q = qdisc_priv(sch); 281 struct nlattr *tb[TCA_FQ_PIE_MAX + 1]; 282 unsigned int len_dropped = 0; 283 unsigned int num_dropped = 0; 284 int err; 285 286 if (!opt) 287 return -EINVAL; 288 289 err = nla_parse_nested(tb, TCA_FQ_PIE_MAX, opt, fq_pie_policy, extack); 290 if (err < 0) 291 return err; 292 293 sch_tree_lock(sch); 294 if (tb[TCA_FQ_PIE_LIMIT]) { 295 u32 limit = nla_get_u32(tb[TCA_FQ_PIE_LIMIT]); 296 297 q->p_params.limit = limit; 298 sch->limit = limit; 299 } 300 if (tb[TCA_FQ_PIE_FLOWS]) { 301 if (q->flows) { 302 NL_SET_ERR_MSG_MOD(extack, 303 "Number of flows cannot be changed"); 304 goto flow_error; 305 } 306 q->flows_cnt = nla_get_u32(tb[TCA_FQ_PIE_FLOWS]); 307 if (!q->flows_cnt || q->flows_cnt > 65536) { 308 NL_SET_ERR_MSG_MOD(extack, 309 "Number of flows must range in [1..65536]"); 310 goto flow_error; 311 } 312 } 313 314 /* convert from microseconds to pschedtime */ 315 if (tb[TCA_FQ_PIE_TARGET]) { 316 /* target is in us */ 317 u32 target = nla_get_u32(tb[TCA_FQ_PIE_TARGET]); 318 319 /* convert to pschedtime */ 320 q->p_params.target = 321 PSCHED_NS2TICKS((u64)target * NSEC_PER_USEC); 322 } 323 324 /* tupdate is in jiffies */ 325 if (tb[TCA_FQ_PIE_TUPDATE]) 326 q->p_params.tupdate = 327 usecs_to_jiffies(nla_get_u32(tb[TCA_FQ_PIE_TUPDATE])); 328 329 if (tb[TCA_FQ_PIE_ALPHA]) 330 q->p_params.alpha = nla_get_u32(tb[TCA_FQ_PIE_ALPHA]); 331 332 if (tb[TCA_FQ_PIE_BETA]) 333 q->p_params.beta = nla_get_u32(tb[TCA_FQ_PIE_BETA]); 334 335 if (tb[TCA_FQ_PIE_QUANTUM]) 336 q->quantum = nla_get_u32(tb[TCA_FQ_PIE_QUANTUM]); 337 338 if (tb[TCA_FQ_PIE_MEMORY_LIMIT]) 339 q->memory_limit = nla_get_u32(tb[TCA_FQ_PIE_MEMORY_LIMIT]); 340 341 if (tb[TCA_FQ_PIE_ECN_PROB]) 342 q->ecn_prob = nla_get_u32(tb[TCA_FQ_PIE_ECN_PROB]); 343 344 if (tb[TCA_FQ_PIE_ECN]) 345 q->p_params.ecn = nla_get_u32(tb[TCA_FQ_PIE_ECN]); 346 347 if (tb[TCA_FQ_PIE_BYTEMODE]) 348 q->p_params.bytemode = nla_get_u32(tb[TCA_FQ_PIE_BYTEMODE]); 349 350 if (tb[TCA_FQ_PIE_DQ_RATE_ESTIMATOR]) 351 q->p_params.dq_rate_estimator = 352 nla_get_u32(tb[TCA_FQ_PIE_DQ_RATE_ESTIMATOR]); 353 354 /* Drop excess packets if new limit is lower */ 355 while (sch->q.qlen > sch->limit) { 356 struct sk_buff *skb = fq_pie_qdisc_dequeue(sch); 357 358 len_dropped += qdisc_pkt_len(skb); 359 num_dropped += 1; 360 rtnl_kfree_skbs(skb, skb); 361 } 362 qdisc_tree_reduce_backlog(sch, num_dropped, len_dropped); 363 364 sch_tree_unlock(sch); 365 return 0; 366 367 flow_error: 368 sch_tree_unlock(sch); 369 return -EINVAL; 370 } 371 372 static void fq_pie_timer(struct timer_list *t) 373 { 374 struct fq_pie_sched_data *q = from_timer(q, t, adapt_timer); 375 struct Qdisc *sch = q->sch; 376 spinlock_t *root_lock; /* to lock qdisc for probability calculations */ 377 u32 idx; 378 379 root_lock = qdisc_lock(qdisc_root_sleeping(sch)); 380 spin_lock(root_lock); 381 382 for (idx = 0; idx < q->flows_cnt; idx++) 383 pie_calculate_probability(&q->p_params, &q->flows[idx].vars, 384 q->flows[idx].backlog); 385 386 /* reset the timer to fire after 'tupdate' jiffies. */ 387 if (q->p_params.tupdate) 388 mod_timer(&q->adapt_timer, jiffies + q->p_params.tupdate); 389 390 spin_unlock(root_lock); 391 } 392 393 static int fq_pie_init(struct Qdisc *sch, struct nlattr *opt, 394 struct netlink_ext_ack *extack) 395 { 396 struct fq_pie_sched_data *q = qdisc_priv(sch); 397 int err; 398 u32 idx; 399 400 pie_params_init(&q->p_params); 401 sch->limit = 10 * 1024; 402 q->p_params.limit = sch->limit; 403 q->quantum = psched_mtu(qdisc_dev(sch)); 404 q->sch = sch; 405 q->ecn_prob = 10; 406 q->flows_cnt = 1024; 407 q->memory_limit = SZ_32M; 408 409 INIT_LIST_HEAD(&q->new_flows); 410 INIT_LIST_HEAD(&q->old_flows); 411 timer_setup(&q->adapt_timer, fq_pie_timer, 0); 412 413 if (opt) { 414 err = fq_pie_change(sch, opt, extack); 415 416 if (err) 417 return err; 418 } 419 420 err = tcf_block_get(&q->block, &q->filter_list, sch, extack); 421 if (err) 422 goto init_failure; 423 424 q->flows = kvcalloc(q->flows_cnt, sizeof(struct fq_pie_flow), 425 GFP_KERNEL); 426 if (!q->flows) { 427 err = -ENOMEM; 428 goto init_failure; 429 } 430 for (idx = 0; idx < q->flows_cnt; idx++) { 431 struct fq_pie_flow *flow = q->flows + idx; 432 433 INIT_LIST_HEAD(&flow->flowchain); 434 pie_vars_init(&flow->vars); 435 } 436 437 mod_timer(&q->adapt_timer, jiffies + HZ / 2); 438 439 return 0; 440 441 init_failure: 442 q->flows_cnt = 0; 443 444 return err; 445 } 446 447 static int fq_pie_dump(struct Qdisc *sch, struct sk_buff *skb) 448 { 449 struct fq_pie_sched_data *q = qdisc_priv(sch); 450 struct nlattr *opts; 451 452 opts = nla_nest_start(skb, TCA_OPTIONS); 453 if (!opts) 454 return -EMSGSIZE; 455 456 /* convert target from pschedtime to us */ 457 if (nla_put_u32(skb, TCA_FQ_PIE_LIMIT, sch->limit) || 458 nla_put_u32(skb, TCA_FQ_PIE_FLOWS, q->flows_cnt) || 459 nla_put_u32(skb, TCA_FQ_PIE_TARGET, 460 ((u32)PSCHED_TICKS2NS(q->p_params.target)) / 461 NSEC_PER_USEC) || 462 nla_put_u32(skb, TCA_FQ_PIE_TUPDATE, 463 jiffies_to_usecs(q->p_params.tupdate)) || 464 nla_put_u32(skb, TCA_FQ_PIE_ALPHA, q->p_params.alpha) || 465 nla_put_u32(skb, TCA_FQ_PIE_BETA, q->p_params.beta) || 466 nla_put_u32(skb, TCA_FQ_PIE_QUANTUM, q->quantum) || 467 nla_put_u32(skb, TCA_FQ_PIE_MEMORY_LIMIT, q->memory_limit) || 468 nla_put_u32(skb, TCA_FQ_PIE_ECN_PROB, q->ecn_prob) || 469 nla_put_u32(skb, TCA_FQ_PIE_ECN, q->p_params.ecn) || 470 nla_put_u32(skb, TCA_FQ_PIE_BYTEMODE, q->p_params.bytemode) || 471 nla_put_u32(skb, TCA_FQ_PIE_DQ_RATE_ESTIMATOR, 472 q->p_params.dq_rate_estimator)) 473 goto nla_put_failure; 474 475 return nla_nest_end(skb, opts); 476 477 nla_put_failure: 478 nla_nest_cancel(skb, opts); 479 return -EMSGSIZE; 480 } 481 482 static int fq_pie_dump_stats(struct Qdisc *sch, struct gnet_dump *d) 483 { 484 struct fq_pie_sched_data *q = qdisc_priv(sch); 485 struct tc_fq_pie_xstats st = { 486 .packets_in = q->stats.packets_in, 487 .overlimit = q->stats.overlimit, 488 .overmemory = q->overmemory, 489 .dropped = q->stats.dropped, 490 .ecn_mark = q->stats.ecn_mark, 491 .new_flow_count = q->new_flow_count, 492 .memory_usage = q->memory_usage, 493 }; 494 struct list_head *pos; 495 496 sch_tree_lock(sch); 497 list_for_each(pos, &q->new_flows) 498 st.new_flows_len++; 499 500 list_for_each(pos, &q->old_flows) 501 st.old_flows_len++; 502 sch_tree_unlock(sch); 503 504 return gnet_stats_copy_app(d, &st, sizeof(st)); 505 } 506 507 static void fq_pie_reset(struct Qdisc *sch) 508 { 509 struct fq_pie_sched_data *q = qdisc_priv(sch); 510 u32 idx; 511 512 INIT_LIST_HEAD(&q->new_flows); 513 INIT_LIST_HEAD(&q->old_flows); 514 for (idx = 0; idx < q->flows_cnt; idx++) { 515 struct fq_pie_flow *flow = q->flows + idx; 516 517 /* Removes all packets from flow */ 518 rtnl_kfree_skbs(flow->head, flow->tail); 519 flow->head = NULL; 520 521 INIT_LIST_HEAD(&flow->flowchain); 522 pie_vars_init(&flow->vars); 523 } 524 525 sch->q.qlen = 0; 526 sch->qstats.backlog = 0; 527 } 528 529 static void fq_pie_destroy(struct Qdisc *sch) 530 { 531 struct fq_pie_sched_data *q = qdisc_priv(sch); 532 533 tcf_block_put(q->block); 534 q->p_params.tupdate = 0; 535 del_timer_sync(&q->adapt_timer); 536 kvfree(q->flows); 537 } 538 539 static struct Qdisc_ops fq_pie_qdisc_ops __read_mostly = { 540 .id = "fq_pie", 541 .priv_size = sizeof(struct fq_pie_sched_data), 542 .enqueue = fq_pie_qdisc_enqueue, 543 .dequeue = fq_pie_qdisc_dequeue, 544 .peek = qdisc_peek_dequeued, 545 .init = fq_pie_init, 546 .destroy = fq_pie_destroy, 547 .reset = fq_pie_reset, 548 .change = fq_pie_change, 549 .dump = fq_pie_dump, 550 .dump_stats = fq_pie_dump_stats, 551 .owner = THIS_MODULE, 552 }; 553 554 static int __init fq_pie_module_init(void) 555 { 556 return register_qdisc(&fq_pie_qdisc_ops); 557 } 558 559 static void __exit fq_pie_module_exit(void) 560 { 561 unregister_qdisc(&fq_pie_qdisc_ops); 562 } 563 564 module_init(fq_pie_module_init); 565 module_exit(fq_pie_module_exit); 566 567 MODULE_DESCRIPTION("Flow Queue Proportional Integral controller Enhanced (FQ-PIE)"); 568 MODULE_AUTHOR("Mohit P. Tahiliani"); 569 MODULE_LICENSE("GPL"); 570