1 // SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause 2 3 /* COMMON Applications Kept Enhanced (CAKE) discipline 4 * 5 * Copyright (C) 2014-2018 Jonathan Morton <chromatix99@gmail.com> 6 * Copyright (C) 2015-2018 Toke Høiland-Jørgensen <toke@toke.dk> 7 * Copyright (C) 2014-2018 Dave Täht <dave.taht@gmail.com> 8 * Copyright (C) 2015-2018 Sebastian Moeller <moeller0@gmx.de> 9 * (C) 2015-2018 Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk> 10 * Copyright (C) 2017-2018 Ryan Mounce <ryan@mounce.com.au> 11 * 12 * The CAKE Principles: 13 * (or, how to have your cake and eat it too) 14 * 15 * This is a combination of several shaping, AQM and FQ techniques into one 16 * easy-to-use package: 17 * 18 * - An overall bandwidth shaper, to move the bottleneck away from dumb CPE 19 * equipment and bloated MACs. This operates in deficit mode (as in sch_fq), 20 * eliminating the need for any sort of burst parameter (eg. token bucket 21 * depth). Burst support is limited to that necessary to overcome scheduling 22 * latency. 23 * 24 * - A Diffserv-aware priority queue, giving more priority to certain classes, 25 * up to a specified fraction of bandwidth. Above that bandwidth threshold, 26 * the priority is reduced to avoid starving other tins. 27 * 28 * - Each priority tin has a separate Flow Queue system, to isolate traffic 29 * flows from each other. This prevents a burst on one flow from increasing 30 * the delay to another. Flows are distributed to queues using a 31 * set-associative hash function. 32 * 33 * - Each queue is actively managed by Cobalt, which is a combination of the 34 * Codel and Blue AQM algorithms. This serves flows fairly, and signals 35 * congestion early via ECN (if available) and/or packet drops, to keep 36 * latency low. The codel parameters are auto-tuned based on the bandwidth 37 * setting, as is necessary at low bandwidths. 38 * 39 * The configuration parameters are kept deliberately simple for ease of use. 40 * Everything has sane defaults. Complete generality of configuration is *not* 41 * a goal. 42 * 43 * The priority queue operates according to a weighted DRR scheme, combined with 44 * a bandwidth tracker which reuses the shaper logic to detect which side of the 45 * bandwidth sharing threshold the tin is operating. This determines whether a 46 * priority-based weight (high) or a bandwidth-based weight (low) is used for 47 * that tin in the current pass. 48 * 49 * This qdisc was inspired by Eric Dumazet's fq_codel code, which he kindly 50 * granted us permission to leverage. 51 */ 52 53 #include <linux/module.h> 54 #include <linux/types.h> 55 #include <linux/kernel.h> 56 #include <linux/jiffies.h> 57 #include <linux/string.h> 58 #include <linux/in.h> 59 #include <linux/errno.h> 60 #include <linux/init.h> 61 #include <linux/skbuff.h> 62 #include <linux/jhash.h> 63 #include <linux/slab.h> 64 #include <linux/vmalloc.h> 65 #include <linux/reciprocal_div.h> 66 #include <net/netlink.h> 67 #include <linux/if_vlan.h> 68 #include <net/pkt_sched.h> 69 #include <net/pkt_cls.h> 70 #include <net/tcp.h> 71 #include <net/flow_dissector.h> 72 73 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 74 #include <net/netfilter/nf_conntrack_core.h> 75 #endif 76 77 #define CAKE_SET_WAYS (8) 78 #define CAKE_MAX_TINS (8) 79 #define CAKE_QUEUES (1024) 80 #define CAKE_FLOW_MASK 63 81 #define CAKE_FLOW_NAT_FLAG 64 82 83 /* struct cobalt_params - contains codel and blue parameters 84 * @interval: codel initial drop rate 85 * @target: maximum persistent sojourn time & blue update rate 86 * @mtu_time: serialisation delay of maximum-size packet 87 * @p_inc: increment of blue drop probability (0.32 fxp) 88 * @p_dec: decrement of blue drop probability (0.32 fxp) 89 */ 90 struct cobalt_params { 91 u64 interval; 92 u64 target; 93 u64 mtu_time; 94 u32 p_inc; 95 u32 p_dec; 96 }; 97 98 /* struct cobalt_vars - contains codel and blue variables 99 * @count: codel dropping frequency 100 * @rec_inv_sqrt: reciprocal value of sqrt(count) >> 1 101 * @drop_next: time to drop next packet, or when we dropped last 102 * @blue_timer: Blue time to next drop 103 * @p_drop: BLUE drop probability (0.32 fxp) 104 * @dropping: set if in dropping state 105 * @ecn_marked: set if marked 106 */ 107 struct cobalt_vars { 108 u32 count; 109 u32 rec_inv_sqrt; 110 ktime_t drop_next; 111 ktime_t blue_timer; 112 u32 p_drop; 113 bool dropping; 114 bool ecn_marked; 115 }; 116 117 enum { 118 CAKE_SET_NONE = 0, 119 CAKE_SET_SPARSE, 120 CAKE_SET_SPARSE_WAIT, /* counted in SPARSE, actually in BULK */ 121 CAKE_SET_BULK, 122 CAKE_SET_DECAYING 123 }; 124 125 struct cake_flow { 126 /* this stuff is all needed per-flow at dequeue time */ 127 struct sk_buff *head; 128 struct sk_buff *tail; 129 struct list_head flowchain; 130 s32 deficit; 131 u32 dropped; 132 struct cobalt_vars cvars; 133 u16 srchost; /* index into cake_host table */ 134 u16 dsthost; 135 u8 set; 136 }; /* please try to keep this structure <= 64 bytes */ 137 138 struct cake_host { 139 u32 srchost_tag; 140 u32 dsthost_tag; 141 u16 srchost_bulk_flow_count; 142 u16 dsthost_bulk_flow_count; 143 }; 144 145 struct cake_heap_entry { 146 u16 t:3, b:10; 147 }; 148 149 struct cake_tin_data { 150 struct cake_flow flows[CAKE_QUEUES]; 151 u32 backlogs[CAKE_QUEUES]; 152 u32 tags[CAKE_QUEUES]; /* for set association */ 153 u16 overflow_idx[CAKE_QUEUES]; 154 struct cake_host hosts[CAKE_QUEUES]; /* for triple isolation */ 155 u16 flow_quantum; 156 157 struct cobalt_params cparams; 158 u32 drop_overlimit; 159 u16 bulk_flow_count; 160 u16 sparse_flow_count; 161 u16 decaying_flow_count; 162 u16 unresponsive_flow_count; 163 164 u32 max_skblen; 165 166 struct list_head new_flows; 167 struct list_head old_flows; 168 struct list_head decaying_flows; 169 170 /* time_next = time_this + ((len * rate_ns) >> rate_shft) */ 171 ktime_t time_next_packet; 172 u64 tin_rate_ns; 173 u64 tin_rate_bps; 174 u16 tin_rate_shft; 175 176 u16 tin_quantum; 177 s32 tin_deficit; 178 u32 tin_backlog; 179 u32 tin_dropped; 180 u32 tin_ecn_mark; 181 182 u32 packets; 183 u64 bytes; 184 185 u32 ack_drops; 186 187 /* moving averages */ 188 u64 avge_delay; 189 u64 peak_delay; 190 u64 base_delay; 191 192 /* hash function stats */ 193 u32 way_directs; 194 u32 way_hits; 195 u32 way_misses; 196 u32 way_collisions; 197 }; /* number of tins is small, so size of this struct doesn't matter much */ 198 199 struct cake_sched_data { 200 struct tcf_proto __rcu *filter_list; /* optional external classifier */ 201 struct tcf_block *block; 202 struct cake_tin_data *tins; 203 204 struct cake_heap_entry overflow_heap[CAKE_QUEUES * CAKE_MAX_TINS]; 205 u16 overflow_timeout; 206 207 u16 tin_cnt; 208 u8 tin_mode; 209 u8 flow_mode; 210 u8 ack_filter; 211 u8 atm_mode; 212 213 u32 fwmark_mask; 214 u16 fwmark_shft; 215 216 /* time_next = time_this + ((len * rate_ns) >> rate_shft) */ 217 u16 rate_shft; 218 ktime_t time_next_packet; 219 ktime_t failsafe_next_packet; 220 u64 rate_ns; 221 u64 rate_bps; 222 u16 rate_flags; 223 s16 rate_overhead; 224 u16 rate_mpu; 225 u64 interval; 226 u64 target; 227 228 /* resource tracking */ 229 u32 buffer_used; 230 u32 buffer_max_used; 231 u32 buffer_limit; 232 u32 buffer_config_limit; 233 234 /* indices for dequeue */ 235 u16 cur_tin; 236 u16 cur_flow; 237 238 struct qdisc_watchdog watchdog; 239 const u8 *tin_index; 240 const u8 *tin_order; 241 242 /* bandwidth capacity estimate */ 243 ktime_t last_packet_time; 244 ktime_t avg_window_begin; 245 u64 avg_packet_interval; 246 u64 avg_window_bytes; 247 u64 avg_peak_bandwidth; 248 ktime_t last_reconfig_time; 249 250 /* packet length stats */ 251 u32 avg_netoff; 252 u16 max_netlen; 253 u16 max_adjlen; 254 u16 min_netlen; 255 u16 min_adjlen; 256 }; 257 258 enum { 259 CAKE_FLAG_OVERHEAD = BIT(0), 260 CAKE_FLAG_AUTORATE_INGRESS = BIT(1), 261 CAKE_FLAG_INGRESS = BIT(2), 262 CAKE_FLAG_WASH = BIT(3), 263 CAKE_FLAG_SPLIT_GSO = BIT(4) 264 }; 265 266 /* COBALT operates the Codel and BLUE algorithms in parallel, in order to 267 * obtain the best features of each. Codel is excellent on flows which 268 * respond to congestion signals in a TCP-like way. BLUE is more effective on 269 * unresponsive flows. 270 */ 271 272 struct cobalt_skb_cb { 273 ktime_t enqueue_time; 274 u32 adjusted_len; 275 }; 276 277 static u64 us_to_ns(u64 us) 278 { 279 return us * NSEC_PER_USEC; 280 } 281 282 static struct cobalt_skb_cb *get_cobalt_cb(const struct sk_buff *skb) 283 { 284 qdisc_cb_private_validate(skb, sizeof(struct cobalt_skb_cb)); 285 return (struct cobalt_skb_cb *)qdisc_skb_cb(skb)->data; 286 } 287 288 static ktime_t cobalt_get_enqueue_time(const struct sk_buff *skb) 289 { 290 return get_cobalt_cb(skb)->enqueue_time; 291 } 292 293 static void cobalt_set_enqueue_time(struct sk_buff *skb, 294 ktime_t now) 295 { 296 get_cobalt_cb(skb)->enqueue_time = now; 297 } 298 299 static u16 quantum_div[CAKE_QUEUES + 1] = {0}; 300 301 /* Diffserv lookup tables */ 302 303 static const u8 precedence[] = { 304 0, 0, 0, 0, 0, 0, 0, 0, 305 1, 1, 1, 1, 1, 1, 1, 1, 306 2, 2, 2, 2, 2, 2, 2, 2, 307 3, 3, 3, 3, 3, 3, 3, 3, 308 4, 4, 4, 4, 4, 4, 4, 4, 309 5, 5, 5, 5, 5, 5, 5, 5, 310 6, 6, 6, 6, 6, 6, 6, 6, 311 7, 7, 7, 7, 7, 7, 7, 7, 312 }; 313 314 static const u8 diffserv8[] = { 315 2, 0, 1, 2, 4, 2, 2, 2, 316 1, 2, 1, 2, 1, 2, 1, 2, 317 5, 2, 4, 2, 4, 2, 4, 2, 318 3, 2, 3, 2, 3, 2, 3, 2, 319 6, 2, 3, 2, 3, 2, 3, 2, 320 6, 2, 2, 2, 6, 2, 6, 2, 321 7, 2, 2, 2, 2, 2, 2, 2, 322 7, 2, 2, 2, 2, 2, 2, 2, 323 }; 324 325 static const u8 diffserv4[] = { 326 0, 1, 0, 0, 2, 0, 0, 0, 327 1, 0, 0, 0, 0, 0, 0, 0, 328 2, 0, 2, 0, 2, 0, 2, 0, 329 2, 0, 2, 0, 2, 0, 2, 0, 330 3, 0, 2, 0, 2, 0, 2, 0, 331 3, 0, 0, 0, 3, 0, 3, 0, 332 3, 0, 0, 0, 0, 0, 0, 0, 333 3, 0, 0, 0, 0, 0, 0, 0, 334 }; 335 336 static const u8 diffserv3[] = { 337 0, 1, 0, 0, 2, 0, 0, 0, 338 1, 0, 0, 0, 0, 0, 0, 0, 339 0, 0, 0, 0, 0, 0, 0, 0, 340 0, 0, 0, 0, 0, 0, 0, 0, 341 0, 0, 0, 0, 0, 0, 0, 0, 342 0, 0, 0, 0, 2, 0, 2, 0, 343 2, 0, 0, 0, 0, 0, 0, 0, 344 2, 0, 0, 0, 0, 0, 0, 0, 345 }; 346 347 static const u8 besteffort[] = { 348 0, 0, 0, 0, 0, 0, 0, 0, 349 0, 0, 0, 0, 0, 0, 0, 0, 350 0, 0, 0, 0, 0, 0, 0, 0, 351 0, 0, 0, 0, 0, 0, 0, 0, 352 0, 0, 0, 0, 0, 0, 0, 0, 353 0, 0, 0, 0, 0, 0, 0, 0, 354 0, 0, 0, 0, 0, 0, 0, 0, 355 0, 0, 0, 0, 0, 0, 0, 0, 356 }; 357 358 /* tin priority order for stats dumping */ 359 360 static const u8 normal_order[] = {0, 1, 2, 3, 4, 5, 6, 7}; 361 static const u8 bulk_order[] = {1, 0, 2, 3}; 362 363 #define REC_INV_SQRT_CACHE (16) 364 static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0}; 365 366 /* http://en.wikipedia.org/wiki/Methods_of_computing_square_roots 367 * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2) 368 * 369 * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32 370 */ 371 372 static void cobalt_newton_step(struct cobalt_vars *vars) 373 { 374 u32 invsqrt, invsqrt2; 375 u64 val; 376 377 invsqrt = vars->rec_inv_sqrt; 378 invsqrt2 = ((u64)invsqrt * invsqrt) >> 32; 379 val = (3LL << 32) - ((u64)vars->count * invsqrt2); 380 381 val >>= 2; /* avoid overflow in following multiply */ 382 val = (val * invsqrt) >> (32 - 2 + 1); 383 384 vars->rec_inv_sqrt = val; 385 } 386 387 static void cobalt_invsqrt(struct cobalt_vars *vars) 388 { 389 if (vars->count < REC_INV_SQRT_CACHE) 390 vars->rec_inv_sqrt = cobalt_rec_inv_sqrt_cache[vars->count]; 391 else 392 cobalt_newton_step(vars); 393 } 394 395 /* There is a big difference in timing between the accurate values placed in 396 * the cache and the approximations given by a single Newton step for small 397 * count values, particularly when stepping from count 1 to 2 or vice versa. 398 * Above 16, a single Newton step gives sufficient accuracy in either 399 * direction, given the precision stored. 400 * 401 * The magnitude of the error when stepping up to count 2 is such as to give 402 * the value that *should* have been produced at count 4. 403 */ 404 405 static void cobalt_cache_init(void) 406 { 407 struct cobalt_vars v; 408 409 memset(&v, 0, sizeof(v)); 410 v.rec_inv_sqrt = ~0U; 411 cobalt_rec_inv_sqrt_cache[0] = v.rec_inv_sqrt; 412 413 for (v.count = 1; v.count < REC_INV_SQRT_CACHE; v.count++) { 414 cobalt_newton_step(&v); 415 cobalt_newton_step(&v); 416 cobalt_newton_step(&v); 417 cobalt_newton_step(&v); 418 419 cobalt_rec_inv_sqrt_cache[v.count] = v.rec_inv_sqrt; 420 } 421 } 422 423 static void cobalt_vars_init(struct cobalt_vars *vars) 424 { 425 memset(vars, 0, sizeof(*vars)); 426 427 if (!cobalt_rec_inv_sqrt_cache[0]) { 428 cobalt_cache_init(); 429 cobalt_rec_inv_sqrt_cache[0] = ~0; 430 } 431 } 432 433 /* CoDel control_law is t + interval/sqrt(count) 434 * We maintain in rec_inv_sqrt the reciprocal value of sqrt(count) to avoid 435 * both sqrt() and divide operation. 436 */ 437 static ktime_t cobalt_control(ktime_t t, 438 u64 interval, 439 u32 rec_inv_sqrt) 440 { 441 return ktime_add_ns(t, reciprocal_scale(interval, 442 rec_inv_sqrt)); 443 } 444 445 /* Call this when a packet had to be dropped due to queue overflow. Returns 446 * true if the BLUE state was quiescent before but active after this call. 447 */ 448 static bool cobalt_queue_full(struct cobalt_vars *vars, 449 struct cobalt_params *p, 450 ktime_t now) 451 { 452 bool up = false; 453 454 if (ktime_to_ns(ktime_sub(now, vars->blue_timer)) > p->target) { 455 up = !vars->p_drop; 456 vars->p_drop += p->p_inc; 457 if (vars->p_drop < p->p_inc) 458 vars->p_drop = ~0; 459 vars->blue_timer = now; 460 } 461 vars->dropping = true; 462 vars->drop_next = now; 463 if (!vars->count) 464 vars->count = 1; 465 466 return up; 467 } 468 469 /* Call this when the queue was serviced but turned out to be empty. Returns 470 * true if the BLUE state was active before but quiescent after this call. 471 */ 472 static bool cobalt_queue_empty(struct cobalt_vars *vars, 473 struct cobalt_params *p, 474 ktime_t now) 475 { 476 bool down = false; 477 478 if (vars->p_drop && 479 ktime_to_ns(ktime_sub(now, vars->blue_timer)) > p->target) { 480 if (vars->p_drop < p->p_dec) 481 vars->p_drop = 0; 482 else 483 vars->p_drop -= p->p_dec; 484 vars->blue_timer = now; 485 down = !vars->p_drop; 486 } 487 vars->dropping = false; 488 489 if (vars->count && ktime_to_ns(ktime_sub(now, vars->drop_next)) >= 0) { 490 vars->count--; 491 cobalt_invsqrt(vars); 492 vars->drop_next = cobalt_control(vars->drop_next, 493 p->interval, 494 vars->rec_inv_sqrt); 495 } 496 497 return down; 498 } 499 500 /* Call this with a freshly dequeued packet for possible congestion marking. 501 * Returns true as an instruction to drop the packet, false for delivery. 502 */ 503 static bool cobalt_should_drop(struct cobalt_vars *vars, 504 struct cobalt_params *p, 505 ktime_t now, 506 struct sk_buff *skb, 507 u32 bulk_flows) 508 { 509 bool next_due, over_target, drop = false; 510 ktime_t schedule; 511 u64 sojourn; 512 513 /* The 'schedule' variable records, in its sign, whether 'now' is before or 514 * after 'drop_next'. This allows 'drop_next' to be updated before the next 515 * scheduling decision is actually branched, without destroying that 516 * information. Similarly, the first 'schedule' value calculated is preserved 517 * in the boolean 'next_due'. 518 * 519 * As for 'drop_next', we take advantage of the fact that 'interval' is both 520 * the delay between first exceeding 'target' and the first signalling event, 521 * *and* the scaling factor for the signalling frequency. It's therefore very 522 * natural to use a single mechanism for both purposes, and eliminates a 523 * significant amount of reference Codel's spaghetti code. To help with this, 524 * both the '0' and '1' entries in the invsqrt cache are 0xFFFFFFFF, as close 525 * as possible to 1.0 in fixed-point. 526 */ 527 528 sojourn = ktime_to_ns(ktime_sub(now, cobalt_get_enqueue_time(skb))); 529 schedule = ktime_sub(now, vars->drop_next); 530 over_target = sojourn > p->target && 531 sojourn > p->mtu_time * bulk_flows * 2 && 532 sojourn > p->mtu_time * 4; 533 next_due = vars->count && ktime_to_ns(schedule) >= 0; 534 535 vars->ecn_marked = false; 536 537 if (over_target) { 538 if (!vars->dropping) { 539 vars->dropping = true; 540 vars->drop_next = cobalt_control(now, 541 p->interval, 542 vars->rec_inv_sqrt); 543 } 544 if (!vars->count) 545 vars->count = 1; 546 } else if (vars->dropping) { 547 vars->dropping = false; 548 } 549 550 if (next_due && vars->dropping) { 551 /* Use ECN mark if possible, otherwise drop */ 552 drop = !(vars->ecn_marked = INET_ECN_set_ce(skb)); 553 554 vars->count++; 555 if (!vars->count) 556 vars->count--; 557 cobalt_invsqrt(vars); 558 vars->drop_next = cobalt_control(vars->drop_next, 559 p->interval, 560 vars->rec_inv_sqrt); 561 schedule = ktime_sub(now, vars->drop_next); 562 } else { 563 while (next_due) { 564 vars->count--; 565 cobalt_invsqrt(vars); 566 vars->drop_next = cobalt_control(vars->drop_next, 567 p->interval, 568 vars->rec_inv_sqrt); 569 schedule = ktime_sub(now, vars->drop_next); 570 next_due = vars->count && ktime_to_ns(schedule) >= 0; 571 } 572 } 573 574 /* Simple BLUE implementation. Lack of ECN is deliberate. */ 575 if (vars->p_drop) 576 drop |= (prandom_u32() < vars->p_drop); 577 578 /* Overload the drop_next field as an activity timeout */ 579 if (!vars->count) 580 vars->drop_next = ktime_add_ns(now, p->interval); 581 else if (ktime_to_ns(schedule) > 0 && !drop) 582 vars->drop_next = now; 583 584 return drop; 585 } 586 587 static bool cake_update_flowkeys(struct flow_keys *keys, 588 const struct sk_buff *skb) 589 { 590 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 591 struct nf_conntrack_tuple tuple = {}; 592 bool rev = !skb->_nfct, upd = false; 593 __be32 ip; 594 595 if (skb_protocol(skb, true) != htons(ETH_P_IP)) 596 return false; 597 598 if (!nf_ct_get_tuple_skb(&tuple, skb)) 599 return false; 600 601 ip = rev ? tuple.dst.u3.ip : tuple.src.u3.ip; 602 if (ip != keys->addrs.v4addrs.src) { 603 keys->addrs.v4addrs.src = ip; 604 upd = true; 605 } 606 ip = rev ? tuple.src.u3.ip : tuple.dst.u3.ip; 607 if (ip != keys->addrs.v4addrs.dst) { 608 keys->addrs.v4addrs.dst = ip; 609 upd = true; 610 } 611 612 if (keys->ports.ports) { 613 __be16 port; 614 615 port = rev ? tuple.dst.u.all : tuple.src.u.all; 616 if (port != keys->ports.src) { 617 keys->ports.src = port; 618 upd = true; 619 } 620 port = rev ? tuple.src.u.all : tuple.dst.u.all; 621 if (port != keys->ports.dst) { 622 port = keys->ports.dst; 623 upd = true; 624 } 625 } 626 return upd; 627 #else 628 return false; 629 #endif 630 } 631 632 /* Cake has several subtle multiple bit settings. In these cases you 633 * would be matching triple isolate mode as well. 634 */ 635 636 static bool cake_dsrc(int flow_mode) 637 { 638 return (flow_mode & CAKE_FLOW_DUAL_SRC) == CAKE_FLOW_DUAL_SRC; 639 } 640 641 static bool cake_ddst(int flow_mode) 642 { 643 return (flow_mode & CAKE_FLOW_DUAL_DST) == CAKE_FLOW_DUAL_DST; 644 } 645 646 static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb, 647 int flow_mode, u16 flow_override, u16 host_override) 648 { 649 bool hash_flows = (!flow_override && !!(flow_mode & CAKE_FLOW_FLOWS)); 650 bool hash_hosts = (!host_override && !!(flow_mode & CAKE_FLOW_HOSTS)); 651 bool nat_enabled = !!(flow_mode & CAKE_FLOW_NAT_FLAG); 652 u32 flow_hash = 0, srchost_hash = 0, dsthost_hash = 0; 653 u16 reduced_hash, srchost_idx, dsthost_idx; 654 struct flow_keys keys, host_keys; 655 bool use_skbhash = skb->l4_hash; 656 657 if (unlikely(flow_mode == CAKE_FLOW_NONE)) 658 return 0; 659 660 /* If both overrides are set, or we can use the SKB hash and nat mode is 661 * disabled, we can skip packet dissection entirely. If nat mode is 662 * enabled there's another check below after doing the conntrack lookup. 663 */ 664 if ((!hash_flows || (use_skbhash && !nat_enabled)) && !hash_hosts) 665 goto skip_hash; 666 667 skb_flow_dissect_flow_keys(skb, &keys, 668 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL); 669 670 /* Don't use the SKB hash if we change the lookup keys from conntrack */ 671 if (nat_enabled && cake_update_flowkeys(&keys, skb)) 672 use_skbhash = false; 673 674 /* If we can still use the SKB hash and don't need the host hash, we can 675 * skip the rest of the hashing procedure 676 */ 677 if (use_skbhash && !hash_hosts) 678 goto skip_hash; 679 680 /* flow_hash_from_keys() sorts the addresses by value, so we have 681 * to preserve their order in a separate data structure to treat 682 * src and dst host addresses as independently selectable. 683 */ 684 host_keys = keys; 685 host_keys.ports.ports = 0; 686 host_keys.basic.ip_proto = 0; 687 host_keys.keyid.keyid = 0; 688 host_keys.tags.flow_label = 0; 689 690 switch (host_keys.control.addr_type) { 691 case FLOW_DISSECTOR_KEY_IPV4_ADDRS: 692 host_keys.addrs.v4addrs.src = 0; 693 dsthost_hash = flow_hash_from_keys(&host_keys); 694 host_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src; 695 host_keys.addrs.v4addrs.dst = 0; 696 srchost_hash = flow_hash_from_keys(&host_keys); 697 break; 698 699 case FLOW_DISSECTOR_KEY_IPV6_ADDRS: 700 memset(&host_keys.addrs.v6addrs.src, 0, 701 sizeof(host_keys.addrs.v6addrs.src)); 702 dsthost_hash = flow_hash_from_keys(&host_keys); 703 host_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src; 704 memset(&host_keys.addrs.v6addrs.dst, 0, 705 sizeof(host_keys.addrs.v6addrs.dst)); 706 srchost_hash = flow_hash_from_keys(&host_keys); 707 break; 708 709 default: 710 dsthost_hash = 0; 711 srchost_hash = 0; 712 } 713 714 /* This *must* be after the above switch, since as a 715 * side-effect it sorts the src and dst addresses. 716 */ 717 if (hash_flows && !use_skbhash) 718 flow_hash = flow_hash_from_keys(&keys); 719 720 skip_hash: 721 if (flow_override) 722 flow_hash = flow_override - 1; 723 else if (use_skbhash) 724 flow_hash = skb->hash; 725 if (host_override) { 726 dsthost_hash = host_override - 1; 727 srchost_hash = host_override - 1; 728 } 729 730 if (!(flow_mode & CAKE_FLOW_FLOWS)) { 731 if (flow_mode & CAKE_FLOW_SRC_IP) 732 flow_hash ^= srchost_hash; 733 734 if (flow_mode & CAKE_FLOW_DST_IP) 735 flow_hash ^= dsthost_hash; 736 } 737 738 reduced_hash = flow_hash % CAKE_QUEUES; 739 740 /* set-associative hashing */ 741 /* fast path if no hash collision (direct lookup succeeds) */ 742 if (likely(q->tags[reduced_hash] == flow_hash && 743 q->flows[reduced_hash].set)) { 744 q->way_directs++; 745 } else { 746 u32 inner_hash = reduced_hash % CAKE_SET_WAYS; 747 u32 outer_hash = reduced_hash - inner_hash; 748 bool allocate_src = false; 749 bool allocate_dst = false; 750 u32 i, k; 751 752 /* check if any active queue in the set is reserved for 753 * this flow. 754 */ 755 for (i = 0, k = inner_hash; i < CAKE_SET_WAYS; 756 i++, k = (k + 1) % CAKE_SET_WAYS) { 757 if (q->tags[outer_hash + k] == flow_hash) { 758 if (i) 759 q->way_hits++; 760 761 if (!q->flows[outer_hash + k].set) { 762 /* need to increment host refcnts */ 763 allocate_src = cake_dsrc(flow_mode); 764 allocate_dst = cake_ddst(flow_mode); 765 } 766 767 goto found; 768 } 769 } 770 771 /* no queue is reserved for this flow, look for an 772 * empty one. 773 */ 774 for (i = 0; i < CAKE_SET_WAYS; 775 i++, k = (k + 1) % CAKE_SET_WAYS) { 776 if (!q->flows[outer_hash + k].set) { 777 q->way_misses++; 778 allocate_src = cake_dsrc(flow_mode); 779 allocate_dst = cake_ddst(flow_mode); 780 goto found; 781 } 782 } 783 784 /* With no empty queues, default to the original 785 * queue, accept the collision, update the host tags. 786 */ 787 q->way_collisions++; 788 if (q->flows[outer_hash + k].set == CAKE_SET_BULK) { 789 q->hosts[q->flows[reduced_hash].srchost].srchost_bulk_flow_count--; 790 q->hosts[q->flows[reduced_hash].dsthost].dsthost_bulk_flow_count--; 791 } 792 allocate_src = cake_dsrc(flow_mode); 793 allocate_dst = cake_ddst(flow_mode); 794 found: 795 /* reserve queue for future packets in same flow */ 796 reduced_hash = outer_hash + k; 797 q->tags[reduced_hash] = flow_hash; 798 799 if (allocate_src) { 800 srchost_idx = srchost_hash % CAKE_QUEUES; 801 inner_hash = srchost_idx % CAKE_SET_WAYS; 802 outer_hash = srchost_idx - inner_hash; 803 for (i = 0, k = inner_hash; i < CAKE_SET_WAYS; 804 i++, k = (k + 1) % CAKE_SET_WAYS) { 805 if (q->hosts[outer_hash + k].srchost_tag == 806 srchost_hash) 807 goto found_src; 808 } 809 for (i = 0; i < CAKE_SET_WAYS; 810 i++, k = (k + 1) % CAKE_SET_WAYS) { 811 if (!q->hosts[outer_hash + k].srchost_bulk_flow_count) 812 break; 813 } 814 q->hosts[outer_hash + k].srchost_tag = srchost_hash; 815 found_src: 816 srchost_idx = outer_hash + k; 817 if (q->flows[reduced_hash].set == CAKE_SET_BULK) 818 q->hosts[srchost_idx].srchost_bulk_flow_count++; 819 q->flows[reduced_hash].srchost = srchost_idx; 820 } 821 822 if (allocate_dst) { 823 dsthost_idx = dsthost_hash % CAKE_QUEUES; 824 inner_hash = dsthost_idx % CAKE_SET_WAYS; 825 outer_hash = dsthost_idx - inner_hash; 826 for (i = 0, k = inner_hash; i < CAKE_SET_WAYS; 827 i++, k = (k + 1) % CAKE_SET_WAYS) { 828 if (q->hosts[outer_hash + k].dsthost_tag == 829 dsthost_hash) 830 goto found_dst; 831 } 832 for (i = 0; i < CAKE_SET_WAYS; 833 i++, k = (k + 1) % CAKE_SET_WAYS) { 834 if (!q->hosts[outer_hash + k].dsthost_bulk_flow_count) 835 break; 836 } 837 q->hosts[outer_hash + k].dsthost_tag = dsthost_hash; 838 found_dst: 839 dsthost_idx = outer_hash + k; 840 if (q->flows[reduced_hash].set == CAKE_SET_BULK) 841 q->hosts[dsthost_idx].dsthost_bulk_flow_count++; 842 q->flows[reduced_hash].dsthost = dsthost_idx; 843 } 844 } 845 846 return reduced_hash; 847 } 848 849 /* helper functions : might be changed when/if skb use a standard list_head */ 850 /* remove one skb from head of slot queue */ 851 852 static struct sk_buff *dequeue_head(struct cake_flow *flow) 853 { 854 struct sk_buff *skb = flow->head; 855 856 if (skb) { 857 flow->head = skb->next; 858 skb_mark_not_on_list(skb); 859 } 860 861 return skb; 862 } 863 864 /* add skb to flow queue (tail add) */ 865 866 static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb) 867 { 868 if (!flow->head) 869 flow->head = skb; 870 else 871 flow->tail->next = skb; 872 flow->tail = skb; 873 skb->next = NULL; 874 } 875 876 static struct iphdr *cake_get_iphdr(const struct sk_buff *skb, 877 struct ipv6hdr *buf) 878 { 879 unsigned int offset = skb_network_offset(skb); 880 struct iphdr *iph; 881 882 iph = skb_header_pointer(skb, offset, sizeof(struct iphdr), buf); 883 884 if (!iph) 885 return NULL; 886 887 if (iph->version == 4 && iph->protocol == IPPROTO_IPV6) 888 return skb_header_pointer(skb, offset + iph->ihl * 4, 889 sizeof(struct ipv6hdr), buf); 890 891 else if (iph->version == 4) 892 return iph; 893 894 else if (iph->version == 6) 895 return skb_header_pointer(skb, offset, sizeof(struct ipv6hdr), 896 buf); 897 898 return NULL; 899 } 900 901 static struct tcphdr *cake_get_tcphdr(const struct sk_buff *skb, 902 void *buf, unsigned int bufsize) 903 { 904 unsigned int offset = skb_network_offset(skb); 905 const struct ipv6hdr *ipv6h; 906 const struct tcphdr *tcph; 907 const struct iphdr *iph; 908 struct ipv6hdr _ipv6h; 909 struct tcphdr _tcph; 910 911 ipv6h = skb_header_pointer(skb, offset, sizeof(_ipv6h), &_ipv6h); 912 913 if (!ipv6h) 914 return NULL; 915 916 if (ipv6h->version == 4) { 917 iph = (struct iphdr *)ipv6h; 918 offset += iph->ihl * 4; 919 920 /* special-case 6in4 tunnelling, as that is a common way to get 921 * v6 connectivity in the home 922 */ 923 if (iph->protocol == IPPROTO_IPV6) { 924 ipv6h = skb_header_pointer(skb, offset, 925 sizeof(_ipv6h), &_ipv6h); 926 927 if (!ipv6h || ipv6h->nexthdr != IPPROTO_TCP) 928 return NULL; 929 930 offset += sizeof(struct ipv6hdr); 931 932 } else if (iph->protocol != IPPROTO_TCP) { 933 return NULL; 934 } 935 936 } else if (ipv6h->version == 6) { 937 if (ipv6h->nexthdr != IPPROTO_TCP) 938 return NULL; 939 940 offset += sizeof(struct ipv6hdr); 941 } else { 942 return NULL; 943 } 944 945 tcph = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph); 946 if (!tcph) 947 return NULL; 948 949 return skb_header_pointer(skb, offset, 950 min(__tcp_hdrlen(tcph), bufsize), buf); 951 } 952 953 static const void *cake_get_tcpopt(const struct tcphdr *tcph, 954 int code, int *oplen) 955 { 956 /* inspired by tcp_parse_options in tcp_input.c */ 957 int length = __tcp_hdrlen(tcph) - sizeof(struct tcphdr); 958 const u8 *ptr = (const u8 *)(tcph + 1); 959 960 while (length > 0) { 961 int opcode = *ptr++; 962 int opsize; 963 964 if (opcode == TCPOPT_EOL) 965 break; 966 if (opcode == TCPOPT_NOP) { 967 length--; 968 continue; 969 } 970 opsize = *ptr++; 971 if (opsize < 2 || opsize > length) 972 break; 973 974 if (opcode == code) { 975 *oplen = opsize; 976 return ptr; 977 } 978 979 ptr += opsize - 2; 980 length -= opsize; 981 } 982 983 return NULL; 984 } 985 986 /* Compare two SACK sequences. A sequence is considered greater if it SACKs more 987 * bytes than the other. In the case where both sequences ACKs bytes that the 988 * other doesn't, A is considered greater. DSACKs in A also makes A be 989 * considered greater. 990 * 991 * @return -1, 0 or 1 as normal compare functions 992 */ 993 static int cake_tcph_sack_compare(const struct tcphdr *tcph_a, 994 const struct tcphdr *tcph_b) 995 { 996 const struct tcp_sack_block_wire *sack_a, *sack_b; 997 u32 ack_seq_a = ntohl(tcph_a->ack_seq); 998 u32 bytes_a = 0, bytes_b = 0; 999 int oplen_a, oplen_b; 1000 bool first = true; 1001 1002 sack_a = cake_get_tcpopt(tcph_a, TCPOPT_SACK, &oplen_a); 1003 sack_b = cake_get_tcpopt(tcph_b, TCPOPT_SACK, &oplen_b); 1004 1005 /* pointers point to option contents */ 1006 oplen_a -= TCPOLEN_SACK_BASE; 1007 oplen_b -= TCPOLEN_SACK_BASE; 1008 1009 if (sack_a && oplen_a >= sizeof(*sack_a) && 1010 (!sack_b || oplen_b < sizeof(*sack_b))) 1011 return -1; 1012 else if (sack_b && oplen_b >= sizeof(*sack_b) && 1013 (!sack_a || oplen_a < sizeof(*sack_a))) 1014 return 1; 1015 else if ((!sack_a || oplen_a < sizeof(*sack_a)) && 1016 (!sack_b || oplen_b < sizeof(*sack_b))) 1017 return 0; 1018 1019 while (oplen_a >= sizeof(*sack_a)) { 1020 const struct tcp_sack_block_wire *sack_tmp = sack_b; 1021 u32 start_a = get_unaligned_be32(&sack_a->start_seq); 1022 u32 end_a = get_unaligned_be32(&sack_a->end_seq); 1023 int oplen_tmp = oplen_b; 1024 bool found = false; 1025 1026 /* DSACK; always considered greater to prevent dropping */ 1027 if (before(start_a, ack_seq_a)) 1028 return -1; 1029 1030 bytes_a += end_a - start_a; 1031 1032 while (oplen_tmp >= sizeof(*sack_tmp)) { 1033 u32 start_b = get_unaligned_be32(&sack_tmp->start_seq); 1034 u32 end_b = get_unaligned_be32(&sack_tmp->end_seq); 1035 1036 /* first time through we count the total size */ 1037 if (first) 1038 bytes_b += end_b - start_b; 1039 1040 if (!after(start_b, start_a) && !before(end_b, end_a)) { 1041 found = true; 1042 if (!first) 1043 break; 1044 } 1045 oplen_tmp -= sizeof(*sack_tmp); 1046 sack_tmp++; 1047 } 1048 1049 if (!found) 1050 return -1; 1051 1052 oplen_a -= sizeof(*sack_a); 1053 sack_a++; 1054 first = false; 1055 } 1056 1057 /* If we made it this far, all ranges SACKed by A are covered by B, so 1058 * either the SACKs are equal, or B SACKs more bytes. 1059 */ 1060 return bytes_b > bytes_a ? 1 : 0; 1061 } 1062 1063 static void cake_tcph_get_tstamp(const struct tcphdr *tcph, 1064 u32 *tsval, u32 *tsecr) 1065 { 1066 const u8 *ptr; 1067 int opsize; 1068 1069 ptr = cake_get_tcpopt(tcph, TCPOPT_TIMESTAMP, &opsize); 1070 1071 if (ptr && opsize == TCPOLEN_TIMESTAMP) { 1072 *tsval = get_unaligned_be32(ptr); 1073 *tsecr = get_unaligned_be32(ptr + 4); 1074 } 1075 } 1076 1077 static bool cake_tcph_may_drop(const struct tcphdr *tcph, 1078 u32 tstamp_new, u32 tsecr_new) 1079 { 1080 /* inspired by tcp_parse_options in tcp_input.c */ 1081 int length = __tcp_hdrlen(tcph) - sizeof(struct tcphdr); 1082 const u8 *ptr = (const u8 *)(tcph + 1); 1083 u32 tstamp, tsecr; 1084 1085 /* 3 reserved flags must be unset to avoid future breakage 1086 * ACK must be set 1087 * ECE/CWR are handled separately 1088 * All other flags URG/PSH/RST/SYN/FIN must be unset 1089 * 0x0FFF0000 = all TCP flags (confirm ACK=1, others zero) 1090 * 0x00C00000 = CWR/ECE (handled separately) 1091 * 0x0F3F0000 = 0x0FFF0000 & ~0x00C00000 1092 */ 1093 if (((tcp_flag_word(tcph) & 1094 cpu_to_be32(0x0F3F0000)) != TCP_FLAG_ACK)) 1095 return false; 1096 1097 while (length > 0) { 1098 int opcode = *ptr++; 1099 int opsize; 1100 1101 if (opcode == TCPOPT_EOL) 1102 break; 1103 if (opcode == TCPOPT_NOP) { 1104 length--; 1105 continue; 1106 } 1107 opsize = *ptr++; 1108 if (opsize < 2 || opsize > length) 1109 break; 1110 1111 switch (opcode) { 1112 case TCPOPT_MD5SIG: /* doesn't influence state */ 1113 break; 1114 1115 case TCPOPT_SACK: /* stricter checking performed later */ 1116 if (opsize % 8 != 2) 1117 return false; 1118 break; 1119 1120 case TCPOPT_TIMESTAMP: 1121 /* only drop timestamps lower than new */ 1122 if (opsize != TCPOLEN_TIMESTAMP) 1123 return false; 1124 tstamp = get_unaligned_be32(ptr); 1125 tsecr = get_unaligned_be32(ptr + 4); 1126 if (after(tstamp, tstamp_new) || 1127 after(tsecr, tsecr_new)) 1128 return false; 1129 break; 1130 1131 case TCPOPT_MSS: /* these should only be set on SYN */ 1132 case TCPOPT_WINDOW: 1133 case TCPOPT_SACK_PERM: 1134 case TCPOPT_FASTOPEN: 1135 case TCPOPT_EXP: 1136 default: /* don't drop if any unknown options are present */ 1137 return false; 1138 } 1139 1140 ptr += opsize - 2; 1141 length -= opsize; 1142 } 1143 1144 return true; 1145 } 1146 1147 static struct sk_buff *cake_ack_filter(struct cake_sched_data *q, 1148 struct cake_flow *flow) 1149 { 1150 bool aggressive = q->ack_filter == CAKE_ACK_AGGRESSIVE; 1151 struct sk_buff *elig_ack = NULL, *elig_ack_prev = NULL; 1152 struct sk_buff *skb_check, *skb_prev = NULL; 1153 const struct ipv6hdr *ipv6h, *ipv6h_check; 1154 unsigned char _tcph[64], _tcph_check[64]; 1155 const struct tcphdr *tcph, *tcph_check; 1156 const struct iphdr *iph, *iph_check; 1157 struct ipv6hdr _iph, _iph_check; 1158 const struct sk_buff *skb; 1159 int seglen, num_found = 0; 1160 u32 tstamp = 0, tsecr = 0; 1161 __be32 elig_flags = 0; 1162 int sack_comp; 1163 1164 /* no other possible ACKs to filter */ 1165 if (flow->head == flow->tail) 1166 return NULL; 1167 1168 skb = flow->tail; 1169 tcph = cake_get_tcphdr(skb, _tcph, sizeof(_tcph)); 1170 iph = cake_get_iphdr(skb, &_iph); 1171 if (!tcph) 1172 return NULL; 1173 1174 cake_tcph_get_tstamp(tcph, &tstamp, &tsecr); 1175 1176 /* the 'triggering' packet need only have the ACK flag set. 1177 * also check that SYN is not set, as there won't be any previous ACKs. 1178 */ 1179 if ((tcp_flag_word(tcph) & 1180 (TCP_FLAG_ACK | TCP_FLAG_SYN)) != TCP_FLAG_ACK) 1181 return NULL; 1182 1183 /* the 'triggering' ACK is at the tail of the queue, we have already 1184 * returned if it is the only packet in the flow. loop through the rest 1185 * of the queue looking for pure ACKs with the same 5-tuple as the 1186 * triggering one. 1187 */ 1188 for (skb_check = flow->head; 1189 skb_check && skb_check != skb; 1190 skb_prev = skb_check, skb_check = skb_check->next) { 1191 iph_check = cake_get_iphdr(skb_check, &_iph_check); 1192 tcph_check = cake_get_tcphdr(skb_check, &_tcph_check, 1193 sizeof(_tcph_check)); 1194 1195 /* only TCP packets with matching 5-tuple are eligible, and only 1196 * drop safe headers 1197 */ 1198 if (!tcph_check || iph->version != iph_check->version || 1199 tcph_check->source != tcph->source || 1200 tcph_check->dest != tcph->dest) 1201 continue; 1202 1203 if (iph_check->version == 4) { 1204 if (iph_check->saddr != iph->saddr || 1205 iph_check->daddr != iph->daddr) 1206 continue; 1207 1208 seglen = ntohs(iph_check->tot_len) - 1209 (4 * iph_check->ihl); 1210 } else if (iph_check->version == 6) { 1211 ipv6h = (struct ipv6hdr *)iph; 1212 ipv6h_check = (struct ipv6hdr *)iph_check; 1213 1214 if (ipv6_addr_cmp(&ipv6h_check->saddr, &ipv6h->saddr) || 1215 ipv6_addr_cmp(&ipv6h_check->daddr, &ipv6h->daddr)) 1216 continue; 1217 1218 seglen = ntohs(ipv6h_check->payload_len); 1219 } else { 1220 WARN_ON(1); /* shouldn't happen */ 1221 continue; 1222 } 1223 1224 /* If the ECE/CWR flags changed from the previous eligible 1225 * packet in the same flow, we should no longer be dropping that 1226 * previous packet as this would lose information. 1227 */ 1228 if (elig_ack && (tcp_flag_word(tcph_check) & 1229 (TCP_FLAG_ECE | TCP_FLAG_CWR)) != elig_flags) { 1230 elig_ack = NULL; 1231 elig_ack_prev = NULL; 1232 num_found--; 1233 } 1234 1235 /* Check TCP options and flags, don't drop ACKs with segment 1236 * data, and don't drop ACKs with a higher cumulative ACK 1237 * counter than the triggering packet. Check ACK seqno here to 1238 * avoid parsing SACK options of packets we are going to exclude 1239 * anyway. 1240 */ 1241 if (!cake_tcph_may_drop(tcph_check, tstamp, tsecr) || 1242 (seglen - __tcp_hdrlen(tcph_check)) != 0 || 1243 after(ntohl(tcph_check->ack_seq), ntohl(tcph->ack_seq))) 1244 continue; 1245 1246 /* Check SACK options. The triggering packet must SACK more data 1247 * than the ACK under consideration, or SACK the same range but 1248 * have a larger cumulative ACK counter. The latter is a 1249 * pathological case, but is contained in the following check 1250 * anyway, just to be safe. 1251 */ 1252 sack_comp = cake_tcph_sack_compare(tcph_check, tcph); 1253 1254 if (sack_comp < 0 || 1255 (ntohl(tcph_check->ack_seq) == ntohl(tcph->ack_seq) && 1256 sack_comp == 0)) 1257 continue; 1258 1259 /* At this point we have found an eligible pure ACK to drop; if 1260 * we are in aggressive mode, we are done. Otherwise, keep 1261 * searching unless this is the second eligible ACK we 1262 * found. 1263 * 1264 * Since we want to drop ACK closest to the head of the queue, 1265 * save the first eligible ACK we find, even if we need to loop 1266 * again. 1267 */ 1268 if (!elig_ack) { 1269 elig_ack = skb_check; 1270 elig_ack_prev = skb_prev; 1271 elig_flags = (tcp_flag_word(tcph_check) 1272 & (TCP_FLAG_ECE | TCP_FLAG_CWR)); 1273 } 1274 1275 if (num_found++ > 0) 1276 goto found; 1277 } 1278 1279 /* We made it through the queue without finding two eligible ACKs . If 1280 * we found a single eligible ACK we can drop it in aggressive mode if 1281 * we can guarantee that this does not interfere with ECN flag 1282 * information. We ensure this by dropping it only if the enqueued 1283 * packet is consecutive with the eligible ACK, and their flags match. 1284 */ 1285 if (elig_ack && aggressive && elig_ack->next == skb && 1286 (elig_flags == (tcp_flag_word(tcph) & 1287 (TCP_FLAG_ECE | TCP_FLAG_CWR)))) 1288 goto found; 1289 1290 return NULL; 1291 1292 found: 1293 if (elig_ack_prev) 1294 elig_ack_prev->next = elig_ack->next; 1295 else 1296 flow->head = elig_ack->next; 1297 1298 skb_mark_not_on_list(elig_ack); 1299 1300 return elig_ack; 1301 } 1302 1303 static u64 cake_ewma(u64 avg, u64 sample, u32 shift) 1304 { 1305 avg -= avg >> shift; 1306 avg += sample >> shift; 1307 return avg; 1308 } 1309 1310 static u32 cake_calc_overhead(struct cake_sched_data *q, u32 len, u32 off) 1311 { 1312 if (q->rate_flags & CAKE_FLAG_OVERHEAD) 1313 len -= off; 1314 1315 if (q->max_netlen < len) 1316 q->max_netlen = len; 1317 if (q->min_netlen > len) 1318 q->min_netlen = len; 1319 1320 len += q->rate_overhead; 1321 1322 if (len < q->rate_mpu) 1323 len = q->rate_mpu; 1324 1325 if (q->atm_mode == CAKE_ATM_ATM) { 1326 len += 47; 1327 len /= 48; 1328 len *= 53; 1329 } else if (q->atm_mode == CAKE_ATM_PTM) { 1330 /* Add one byte per 64 bytes or part thereof. 1331 * This is conservative and easier to calculate than the 1332 * precise value. 1333 */ 1334 len += (len + 63) / 64; 1335 } 1336 1337 if (q->max_adjlen < len) 1338 q->max_adjlen = len; 1339 if (q->min_adjlen > len) 1340 q->min_adjlen = len; 1341 1342 return len; 1343 } 1344 1345 static u32 cake_overhead(struct cake_sched_data *q, const struct sk_buff *skb) 1346 { 1347 const struct skb_shared_info *shinfo = skb_shinfo(skb); 1348 unsigned int hdr_len, last_len = 0; 1349 u32 off = skb_network_offset(skb); 1350 u32 len = qdisc_pkt_len(skb); 1351 u16 segs = 1; 1352 1353 q->avg_netoff = cake_ewma(q->avg_netoff, off << 16, 8); 1354 1355 if (!shinfo->gso_size) 1356 return cake_calc_overhead(q, len, off); 1357 1358 /* borrowed from qdisc_pkt_len_init() */ 1359 hdr_len = skb_transport_header(skb) - skb_mac_header(skb); 1360 1361 /* + transport layer */ 1362 if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | 1363 SKB_GSO_TCPV6))) { 1364 const struct tcphdr *th; 1365 struct tcphdr _tcphdr; 1366 1367 th = skb_header_pointer(skb, skb_transport_offset(skb), 1368 sizeof(_tcphdr), &_tcphdr); 1369 if (likely(th)) 1370 hdr_len += __tcp_hdrlen(th); 1371 } else { 1372 struct udphdr _udphdr; 1373 1374 if (skb_header_pointer(skb, skb_transport_offset(skb), 1375 sizeof(_udphdr), &_udphdr)) 1376 hdr_len += sizeof(struct udphdr); 1377 } 1378 1379 if (unlikely(shinfo->gso_type & SKB_GSO_DODGY)) 1380 segs = DIV_ROUND_UP(skb->len - hdr_len, 1381 shinfo->gso_size); 1382 else 1383 segs = shinfo->gso_segs; 1384 1385 len = shinfo->gso_size + hdr_len; 1386 last_len = skb->len - shinfo->gso_size * (segs - 1); 1387 1388 return (cake_calc_overhead(q, len, off) * (segs - 1) + 1389 cake_calc_overhead(q, last_len, off)); 1390 } 1391 1392 static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j) 1393 { 1394 struct cake_heap_entry ii = q->overflow_heap[i]; 1395 struct cake_heap_entry jj = q->overflow_heap[j]; 1396 1397 q->overflow_heap[i] = jj; 1398 q->overflow_heap[j] = ii; 1399 1400 q->tins[ii.t].overflow_idx[ii.b] = j; 1401 q->tins[jj.t].overflow_idx[jj.b] = i; 1402 } 1403 1404 static u32 cake_heap_get_backlog(const struct cake_sched_data *q, u16 i) 1405 { 1406 struct cake_heap_entry ii = q->overflow_heap[i]; 1407 1408 return q->tins[ii.t].backlogs[ii.b]; 1409 } 1410 1411 static void cake_heapify(struct cake_sched_data *q, u16 i) 1412 { 1413 static const u32 a = CAKE_MAX_TINS * CAKE_QUEUES; 1414 u32 mb = cake_heap_get_backlog(q, i); 1415 u32 m = i; 1416 1417 while (m < a) { 1418 u32 l = m + m + 1; 1419 u32 r = l + 1; 1420 1421 if (l < a) { 1422 u32 lb = cake_heap_get_backlog(q, l); 1423 1424 if (lb > mb) { 1425 m = l; 1426 mb = lb; 1427 } 1428 } 1429 1430 if (r < a) { 1431 u32 rb = cake_heap_get_backlog(q, r); 1432 1433 if (rb > mb) { 1434 m = r; 1435 mb = rb; 1436 } 1437 } 1438 1439 if (m != i) { 1440 cake_heap_swap(q, i, m); 1441 i = m; 1442 } else { 1443 break; 1444 } 1445 } 1446 } 1447 1448 static void cake_heapify_up(struct cake_sched_data *q, u16 i) 1449 { 1450 while (i > 0 && i < CAKE_MAX_TINS * CAKE_QUEUES) { 1451 u16 p = (i - 1) >> 1; 1452 u32 ib = cake_heap_get_backlog(q, i); 1453 u32 pb = cake_heap_get_backlog(q, p); 1454 1455 if (ib > pb) { 1456 cake_heap_swap(q, i, p); 1457 i = p; 1458 } else { 1459 break; 1460 } 1461 } 1462 } 1463 1464 static int cake_advance_shaper(struct cake_sched_data *q, 1465 struct cake_tin_data *b, 1466 struct sk_buff *skb, 1467 ktime_t now, bool drop) 1468 { 1469 u32 len = get_cobalt_cb(skb)->adjusted_len; 1470 1471 /* charge packet bandwidth to this tin 1472 * and to the global shaper. 1473 */ 1474 if (q->rate_ns) { 1475 u64 tin_dur = (len * b->tin_rate_ns) >> b->tin_rate_shft; 1476 u64 global_dur = (len * q->rate_ns) >> q->rate_shft; 1477 u64 failsafe_dur = global_dur + (global_dur >> 1); 1478 1479 if (ktime_before(b->time_next_packet, now)) 1480 b->time_next_packet = ktime_add_ns(b->time_next_packet, 1481 tin_dur); 1482 1483 else if (ktime_before(b->time_next_packet, 1484 ktime_add_ns(now, tin_dur))) 1485 b->time_next_packet = ktime_add_ns(now, tin_dur); 1486 1487 q->time_next_packet = ktime_add_ns(q->time_next_packet, 1488 global_dur); 1489 if (!drop) 1490 q->failsafe_next_packet = \ 1491 ktime_add_ns(q->failsafe_next_packet, 1492 failsafe_dur); 1493 } 1494 return len; 1495 } 1496 1497 static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free) 1498 { 1499 struct cake_sched_data *q = qdisc_priv(sch); 1500 ktime_t now = ktime_get(); 1501 u32 idx = 0, tin = 0, len; 1502 struct cake_heap_entry qq; 1503 struct cake_tin_data *b; 1504 struct cake_flow *flow; 1505 struct sk_buff *skb; 1506 1507 if (!q->overflow_timeout) { 1508 int i; 1509 /* Build fresh max-heap */ 1510 for (i = CAKE_MAX_TINS * CAKE_QUEUES / 2; i >= 0; i--) 1511 cake_heapify(q, i); 1512 } 1513 q->overflow_timeout = 65535; 1514 1515 /* select longest queue for pruning */ 1516 qq = q->overflow_heap[0]; 1517 tin = qq.t; 1518 idx = qq.b; 1519 1520 b = &q->tins[tin]; 1521 flow = &b->flows[idx]; 1522 skb = dequeue_head(flow); 1523 if (unlikely(!skb)) { 1524 /* heap has gone wrong, rebuild it next time */ 1525 q->overflow_timeout = 0; 1526 return idx + (tin << 16); 1527 } 1528 1529 if (cobalt_queue_full(&flow->cvars, &b->cparams, now)) 1530 b->unresponsive_flow_count++; 1531 1532 len = qdisc_pkt_len(skb); 1533 q->buffer_used -= skb->truesize; 1534 b->backlogs[idx] -= len; 1535 b->tin_backlog -= len; 1536 sch->qstats.backlog -= len; 1537 qdisc_tree_reduce_backlog(sch, 1, len); 1538 1539 flow->dropped++; 1540 b->tin_dropped++; 1541 sch->qstats.drops++; 1542 1543 if (q->rate_flags & CAKE_FLAG_INGRESS) 1544 cake_advance_shaper(q, b, skb, now, true); 1545 1546 __qdisc_drop(skb, to_free); 1547 sch->q.qlen--; 1548 1549 cake_heapify(q, 0); 1550 1551 return idx + (tin << 16); 1552 } 1553 1554 static u8 cake_handle_diffserv(struct sk_buff *skb, bool wash) 1555 { 1556 const int offset = skb_network_offset(skb); 1557 u16 *buf, buf_; 1558 u8 dscp; 1559 1560 switch (skb_protocol(skb, true)) { 1561 case htons(ETH_P_IP): 1562 buf = skb_header_pointer(skb, offset, sizeof(buf_), &buf_); 1563 if (unlikely(!buf)) 1564 return 0; 1565 1566 /* ToS is in the second byte of iphdr */ 1567 dscp = ipv4_get_dsfield((struct iphdr *)buf) >> 2; 1568 1569 if (wash && dscp) { 1570 const int wlen = offset + sizeof(struct iphdr); 1571 1572 if (!pskb_may_pull(skb, wlen) || 1573 skb_try_make_writable(skb, wlen)) 1574 return 0; 1575 1576 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, 0); 1577 } 1578 1579 return dscp; 1580 1581 case htons(ETH_P_IPV6): 1582 buf = skb_header_pointer(skb, offset, sizeof(buf_), &buf_); 1583 if (unlikely(!buf)) 1584 return 0; 1585 1586 /* Traffic class is in the first and second bytes of ipv6hdr */ 1587 dscp = ipv6_get_dsfield((struct ipv6hdr *)buf) >> 2; 1588 1589 if (wash && dscp) { 1590 const int wlen = offset + sizeof(struct ipv6hdr); 1591 1592 if (!pskb_may_pull(skb, wlen) || 1593 skb_try_make_writable(skb, wlen)) 1594 return 0; 1595 1596 ipv6_change_dsfield(ipv6_hdr(skb), INET_ECN_MASK, 0); 1597 } 1598 1599 return dscp; 1600 1601 case htons(ETH_P_ARP): 1602 return 0x38; /* CS7 - Net Control */ 1603 1604 default: 1605 /* If there is no Diffserv field, treat as best-effort */ 1606 return 0; 1607 } 1608 } 1609 1610 static struct cake_tin_data *cake_select_tin(struct Qdisc *sch, 1611 struct sk_buff *skb) 1612 { 1613 struct cake_sched_data *q = qdisc_priv(sch); 1614 u32 tin, mark; 1615 bool wash; 1616 u8 dscp; 1617 1618 /* Tin selection: Default to diffserv-based selection, allow overriding 1619 * using firewall marks or skb->priority. Call DSCP parsing early if 1620 * wash is enabled, otherwise defer to below to skip unneeded parsing. 1621 */ 1622 mark = (skb->mark & q->fwmark_mask) >> q->fwmark_shft; 1623 wash = !!(q->rate_flags & CAKE_FLAG_WASH); 1624 if (wash) 1625 dscp = cake_handle_diffserv(skb, wash); 1626 1627 if (q->tin_mode == CAKE_DIFFSERV_BESTEFFORT) 1628 tin = 0; 1629 1630 else if (mark && mark <= q->tin_cnt) 1631 tin = q->tin_order[mark - 1]; 1632 1633 else if (TC_H_MAJ(skb->priority) == sch->handle && 1634 TC_H_MIN(skb->priority) > 0 && 1635 TC_H_MIN(skb->priority) <= q->tin_cnt) 1636 tin = q->tin_order[TC_H_MIN(skb->priority) - 1]; 1637 1638 else { 1639 if (!wash) 1640 dscp = cake_handle_diffserv(skb, wash); 1641 tin = q->tin_index[dscp]; 1642 1643 if (unlikely(tin >= q->tin_cnt)) 1644 tin = 0; 1645 } 1646 1647 return &q->tins[tin]; 1648 } 1649 1650 static u32 cake_classify(struct Qdisc *sch, struct cake_tin_data **t, 1651 struct sk_buff *skb, int flow_mode, int *qerr) 1652 { 1653 struct cake_sched_data *q = qdisc_priv(sch); 1654 struct tcf_proto *filter; 1655 struct tcf_result res; 1656 u16 flow = 0, host = 0; 1657 int result; 1658 1659 filter = rcu_dereference_bh(q->filter_list); 1660 if (!filter) 1661 goto hash; 1662 1663 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS; 1664 result = tcf_classify(skb, filter, &res, false); 1665 1666 if (result >= 0) { 1667 #ifdef CONFIG_NET_CLS_ACT 1668 switch (result) { 1669 case TC_ACT_STOLEN: 1670 case TC_ACT_QUEUED: 1671 case TC_ACT_TRAP: 1672 *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN; 1673 /* fall through */ 1674 case TC_ACT_SHOT: 1675 return 0; 1676 } 1677 #endif 1678 if (TC_H_MIN(res.classid) <= CAKE_QUEUES) 1679 flow = TC_H_MIN(res.classid); 1680 if (TC_H_MAJ(res.classid) <= (CAKE_QUEUES << 16)) 1681 host = TC_H_MAJ(res.classid) >> 16; 1682 } 1683 hash: 1684 *t = cake_select_tin(sch, skb); 1685 return cake_hash(*t, skb, flow_mode, flow, host) + 1; 1686 } 1687 1688 static void cake_reconfigure(struct Qdisc *sch); 1689 1690 static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch, 1691 struct sk_buff **to_free) 1692 { 1693 struct cake_sched_data *q = qdisc_priv(sch); 1694 int len = qdisc_pkt_len(skb); 1695 int uninitialized_var(ret); 1696 struct sk_buff *ack = NULL; 1697 ktime_t now = ktime_get(); 1698 struct cake_tin_data *b; 1699 struct cake_flow *flow; 1700 u32 idx; 1701 1702 /* choose flow to insert into */ 1703 idx = cake_classify(sch, &b, skb, q->flow_mode, &ret); 1704 if (idx == 0) { 1705 if (ret & __NET_XMIT_BYPASS) 1706 qdisc_qstats_drop(sch); 1707 __qdisc_drop(skb, to_free); 1708 return ret; 1709 } 1710 idx--; 1711 flow = &b->flows[idx]; 1712 1713 /* ensure shaper state isn't stale */ 1714 if (!b->tin_backlog) { 1715 if (ktime_before(b->time_next_packet, now)) 1716 b->time_next_packet = now; 1717 1718 if (!sch->q.qlen) { 1719 if (ktime_before(q->time_next_packet, now)) { 1720 q->failsafe_next_packet = now; 1721 q->time_next_packet = now; 1722 } else if (ktime_after(q->time_next_packet, now) && 1723 ktime_after(q->failsafe_next_packet, now)) { 1724 u64 next = \ 1725 min(ktime_to_ns(q->time_next_packet), 1726 ktime_to_ns( 1727 q->failsafe_next_packet)); 1728 sch->qstats.overlimits++; 1729 qdisc_watchdog_schedule_ns(&q->watchdog, next); 1730 } 1731 } 1732 } 1733 1734 if (unlikely(len > b->max_skblen)) 1735 b->max_skblen = len; 1736 1737 if (skb_is_gso(skb) && q->rate_flags & CAKE_FLAG_SPLIT_GSO) { 1738 struct sk_buff *segs, *nskb; 1739 netdev_features_t features = netif_skb_features(skb); 1740 unsigned int slen = 0, numsegs = 0; 1741 1742 segs = skb_gso_segment(skb, features & ~NETIF_F_GSO_MASK); 1743 if (IS_ERR_OR_NULL(segs)) 1744 return qdisc_drop(skb, sch, to_free); 1745 1746 skb_list_walk_safe(segs, segs, nskb) { 1747 skb_mark_not_on_list(segs); 1748 qdisc_skb_cb(segs)->pkt_len = segs->len; 1749 cobalt_set_enqueue_time(segs, now); 1750 get_cobalt_cb(segs)->adjusted_len = cake_overhead(q, 1751 segs); 1752 flow_queue_add(flow, segs); 1753 1754 sch->q.qlen++; 1755 numsegs++; 1756 slen += segs->len; 1757 q->buffer_used += segs->truesize; 1758 b->packets++; 1759 } 1760 1761 /* stats */ 1762 b->bytes += slen; 1763 b->backlogs[idx] += slen; 1764 b->tin_backlog += slen; 1765 sch->qstats.backlog += slen; 1766 q->avg_window_bytes += slen; 1767 1768 qdisc_tree_reduce_backlog(sch, 1-numsegs, len-slen); 1769 consume_skb(skb); 1770 } else { 1771 /* not splitting */ 1772 cobalt_set_enqueue_time(skb, now); 1773 get_cobalt_cb(skb)->adjusted_len = cake_overhead(q, skb); 1774 flow_queue_add(flow, skb); 1775 1776 if (q->ack_filter) 1777 ack = cake_ack_filter(q, flow); 1778 1779 if (ack) { 1780 b->ack_drops++; 1781 sch->qstats.drops++; 1782 b->bytes += qdisc_pkt_len(ack); 1783 len -= qdisc_pkt_len(ack); 1784 q->buffer_used += skb->truesize - ack->truesize; 1785 if (q->rate_flags & CAKE_FLAG_INGRESS) 1786 cake_advance_shaper(q, b, ack, now, true); 1787 1788 qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack)); 1789 consume_skb(ack); 1790 } else { 1791 sch->q.qlen++; 1792 q->buffer_used += skb->truesize; 1793 } 1794 1795 /* stats */ 1796 b->packets++; 1797 b->bytes += len; 1798 b->backlogs[idx] += len; 1799 b->tin_backlog += len; 1800 sch->qstats.backlog += len; 1801 q->avg_window_bytes += len; 1802 } 1803 1804 if (q->overflow_timeout) 1805 cake_heapify_up(q, b->overflow_idx[idx]); 1806 1807 /* incoming bandwidth capacity estimate */ 1808 if (q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS) { 1809 u64 packet_interval = \ 1810 ktime_to_ns(ktime_sub(now, q->last_packet_time)); 1811 1812 if (packet_interval > NSEC_PER_SEC) 1813 packet_interval = NSEC_PER_SEC; 1814 1815 /* filter out short-term bursts, eg. wifi aggregation */ 1816 q->avg_packet_interval = \ 1817 cake_ewma(q->avg_packet_interval, 1818 packet_interval, 1819 (packet_interval > q->avg_packet_interval ? 1820 2 : 8)); 1821 1822 q->last_packet_time = now; 1823 1824 if (packet_interval > q->avg_packet_interval) { 1825 u64 window_interval = \ 1826 ktime_to_ns(ktime_sub(now, 1827 q->avg_window_begin)); 1828 u64 b = q->avg_window_bytes * (u64)NSEC_PER_SEC; 1829 1830 b = div64_u64(b, window_interval); 1831 q->avg_peak_bandwidth = 1832 cake_ewma(q->avg_peak_bandwidth, b, 1833 b > q->avg_peak_bandwidth ? 2 : 8); 1834 q->avg_window_bytes = 0; 1835 q->avg_window_begin = now; 1836 1837 if (ktime_after(now, 1838 ktime_add_ms(q->last_reconfig_time, 1839 250))) { 1840 q->rate_bps = (q->avg_peak_bandwidth * 15) >> 4; 1841 cake_reconfigure(sch); 1842 } 1843 } 1844 } else { 1845 q->avg_window_bytes = 0; 1846 q->last_packet_time = now; 1847 } 1848 1849 /* flowchain */ 1850 if (!flow->set || flow->set == CAKE_SET_DECAYING) { 1851 struct cake_host *srchost = &b->hosts[flow->srchost]; 1852 struct cake_host *dsthost = &b->hosts[flow->dsthost]; 1853 u16 host_load = 1; 1854 1855 if (!flow->set) { 1856 list_add_tail(&flow->flowchain, &b->new_flows); 1857 } else { 1858 b->decaying_flow_count--; 1859 list_move_tail(&flow->flowchain, &b->new_flows); 1860 } 1861 flow->set = CAKE_SET_SPARSE; 1862 b->sparse_flow_count++; 1863 1864 if (cake_dsrc(q->flow_mode)) 1865 host_load = max(host_load, srchost->srchost_bulk_flow_count); 1866 1867 if (cake_ddst(q->flow_mode)) 1868 host_load = max(host_load, dsthost->dsthost_bulk_flow_count); 1869 1870 flow->deficit = (b->flow_quantum * 1871 quantum_div[host_load]) >> 16; 1872 } else if (flow->set == CAKE_SET_SPARSE_WAIT) { 1873 struct cake_host *srchost = &b->hosts[flow->srchost]; 1874 struct cake_host *dsthost = &b->hosts[flow->dsthost]; 1875 1876 /* this flow was empty, accounted as a sparse flow, but actually 1877 * in the bulk rotation. 1878 */ 1879 flow->set = CAKE_SET_BULK; 1880 b->sparse_flow_count--; 1881 b->bulk_flow_count++; 1882 1883 if (cake_dsrc(q->flow_mode)) 1884 srchost->srchost_bulk_flow_count++; 1885 1886 if (cake_ddst(q->flow_mode)) 1887 dsthost->dsthost_bulk_flow_count++; 1888 1889 } 1890 1891 if (q->buffer_used > q->buffer_max_used) 1892 q->buffer_max_used = q->buffer_used; 1893 1894 if (q->buffer_used > q->buffer_limit) { 1895 u32 dropped = 0; 1896 1897 while (q->buffer_used > q->buffer_limit) { 1898 dropped++; 1899 cake_drop(sch, to_free); 1900 } 1901 b->drop_overlimit += dropped; 1902 } 1903 return NET_XMIT_SUCCESS; 1904 } 1905 1906 static struct sk_buff *cake_dequeue_one(struct Qdisc *sch) 1907 { 1908 struct cake_sched_data *q = qdisc_priv(sch); 1909 struct cake_tin_data *b = &q->tins[q->cur_tin]; 1910 struct cake_flow *flow = &b->flows[q->cur_flow]; 1911 struct sk_buff *skb = NULL; 1912 u32 len; 1913 1914 if (flow->head) { 1915 skb = dequeue_head(flow); 1916 len = qdisc_pkt_len(skb); 1917 b->backlogs[q->cur_flow] -= len; 1918 b->tin_backlog -= len; 1919 sch->qstats.backlog -= len; 1920 q->buffer_used -= skb->truesize; 1921 sch->q.qlen--; 1922 1923 if (q->overflow_timeout) 1924 cake_heapify(q, b->overflow_idx[q->cur_flow]); 1925 } 1926 return skb; 1927 } 1928 1929 /* Discard leftover packets from a tin no longer in use. */ 1930 static void cake_clear_tin(struct Qdisc *sch, u16 tin) 1931 { 1932 struct cake_sched_data *q = qdisc_priv(sch); 1933 struct sk_buff *skb; 1934 1935 q->cur_tin = tin; 1936 for (q->cur_flow = 0; q->cur_flow < CAKE_QUEUES; q->cur_flow++) 1937 while (!!(skb = cake_dequeue_one(sch))) 1938 kfree_skb(skb); 1939 } 1940 1941 static struct sk_buff *cake_dequeue(struct Qdisc *sch) 1942 { 1943 struct cake_sched_data *q = qdisc_priv(sch); 1944 struct cake_tin_data *b = &q->tins[q->cur_tin]; 1945 struct cake_host *srchost, *dsthost; 1946 ktime_t now = ktime_get(); 1947 struct cake_flow *flow; 1948 struct list_head *head; 1949 bool first_flow = true; 1950 struct sk_buff *skb; 1951 u16 host_load; 1952 u64 delay; 1953 u32 len; 1954 1955 begin: 1956 if (!sch->q.qlen) 1957 return NULL; 1958 1959 /* global hard shaper */ 1960 if (ktime_after(q->time_next_packet, now) && 1961 ktime_after(q->failsafe_next_packet, now)) { 1962 u64 next = min(ktime_to_ns(q->time_next_packet), 1963 ktime_to_ns(q->failsafe_next_packet)); 1964 1965 sch->qstats.overlimits++; 1966 qdisc_watchdog_schedule_ns(&q->watchdog, next); 1967 return NULL; 1968 } 1969 1970 /* Choose a class to work on. */ 1971 if (!q->rate_ns) { 1972 /* In unlimited mode, can't rely on shaper timings, just balance 1973 * with DRR 1974 */ 1975 bool wrapped = false, empty = true; 1976 1977 while (b->tin_deficit < 0 || 1978 !(b->sparse_flow_count + b->bulk_flow_count)) { 1979 if (b->tin_deficit <= 0) 1980 b->tin_deficit += b->tin_quantum; 1981 if (b->sparse_flow_count + b->bulk_flow_count) 1982 empty = false; 1983 1984 q->cur_tin++; 1985 b++; 1986 if (q->cur_tin >= q->tin_cnt) { 1987 q->cur_tin = 0; 1988 b = q->tins; 1989 1990 if (wrapped) { 1991 /* It's possible for q->qlen to be 1992 * nonzero when we actually have no 1993 * packets anywhere. 1994 */ 1995 if (empty) 1996 return NULL; 1997 } else { 1998 wrapped = true; 1999 } 2000 } 2001 } 2002 } else { 2003 /* In shaped mode, choose: 2004 * - Highest-priority tin with queue and meeting schedule, or 2005 * - The earliest-scheduled tin with queue. 2006 */ 2007 ktime_t best_time = KTIME_MAX; 2008 int tin, best_tin = 0; 2009 2010 for (tin = 0; tin < q->tin_cnt; tin++) { 2011 b = q->tins + tin; 2012 if ((b->sparse_flow_count + b->bulk_flow_count) > 0) { 2013 ktime_t time_to_pkt = \ 2014 ktime_sub(b->time_next_packet, now); 2015 2016 if (ktime_to_ns(time_to_pkt) <= 0 || 2017 ktime_compare(time_to_pkt, 2018 best_time) <= 0) { 2019 best_time = time_to_pkt; 2020 best_tin = tin; 2021 } 2022 } 2023 } 2024 2025 q->cur_tin = best_tin; 2026 b = q->tins + best_tin; 2027 2028 /* No point in going further if no packets to deliver. */ 2029 if (unlikely(!(b->sparse_flow_count + b->bulk_flow_count))) 2030 return NULL; 2031 } 2032 2033 retry: 2034 /* service this class */ 2035 head = &b->decaying_flows; 2036 if (!first_flow || list_empty(head)) { 2037 head = &b->new_flows; 2038 if (list_empty(head)) { 2039 head = &b->old_flows; 2040 if (unlikely(list_empty(head))) { 2041 head = &b->decaying_flows; 2042 if (unlikely(list_empty(head))) 2043 goto begin; 2044 } 2045 } 2046 } 2047 flow = list_first_entry(head, struct cake_flow, flowchain); 2048 q->cur_flow = flow - b->flows; 2049 first_flow = false; 2050 2051 /* triple isolation (modified DRR++) */ 2052 srchost = &b->hosts[flow->srchost]; 2053 dsthost = &b->hosts[flow->dsthost]; 2054 host_load = 1; 2055 2056 /* flow isolation (DRR++) */ 2057 if (flow->deficit <= 0) { 2058 /* Keep all flows with deficits out of the sparse and decaying 2059 * rotations. No non-empty flow can go into the decaying 2060 * rotation, so they can't get deficits 2061 */ 2062 if (flow->set == CAKE_SET_SPARSE) { 2063 if (flow->head) { 2064 b->sparse_flow_count--; 2065 b->bulk_flow_count++; 2066 2067 if (cake_dsrc(q->flow_mode)) 2068 srchost->srchost_bulk_flow_count++; 2069 2070 if (cake_ddst(q->flow_mode)) 2071 dsthost->dsthost_bulk_flow_count++; 2072 2073 flow->set = CAKE_SET_BULK; 2074 } else { 2075 /* we've moved it to the bulk rotation for 2076 * correct deficit accounting but we still want 2077 * to count it as a sparse flow, not a bulk one. 2078 */ 2079 flow->set = CAKE_SET_SPARSE_WAIT; 2080 } 2081 } 2082 2083 if (cake_dsrc(q->flow_mode)) 2084 host_load = max(host_load, srchost->srchost_bulk_flow_count); 2085 2086 if (cake_ddst(q->flow_mode)) 2087 host_load = max(host_load, dsthost->dsthost_bulk_flow_count); 2088 2089 WARN_ON(host_load > CAKE_QUEUES); 2090 2091 /* The shifted prandom_u32() is a way to apply dithering to 2092 * avoid accumulating roundoff errors 2093 */ 2094 flow->deficit += (b->flow_quantum * quantum_div[host_load] + 2095 (prandom_u32() >> 16)) >> 16; 2096 list_move_tail(&flow->flowchain, &b->old_flows); 2097 2098 goto retry; 2099 } 2100 2101 /* Retrieve a packet via the AQM */ 2102 while (1) { 2103 skb = cake_dequeue_one(sch); 2104 if (!skb) { 2105 /* this queue was actually empty */ 2106 if (cobalt_queue_empty(&flow->cvars, &b->cparams, now)) 2107 b->unresponsive_flow_count--; 2108 2109 if (flow->cvars.p_drop || flow->cvars.count || 2110 ktime_before(now, flow->cvars.drop_next)) { 2111 /* keep in the flowchain until the state has 2112 * decayed to rest 2113 */ 2114 list_move_tail(&flow->flowchain, 2115 &b->decaying_flows); 2116 if (flow->set == CAKE_SET_BULK) { 2117 b->bulk_flow_count--; 2118 2119 if (cake_dsrc(q->flow_mode)) 2120 srchost->srchost_bulk_flow_count--; 2121 2122 if (cake_ddst(q->flow_mode)) 2123 dsthost->dsthost_bulk_flow_count--; 2124 2125 b->decaying_flow_count++; 2126 } else if (flow->set == CAKE_SET_SPARSE || 2127 flow->set == CAKE_SET_SPARSE_WAIT) { 2128 b->sparse_flow_count--; 2129 b->decaying_flow_count++; 2130 } 2131 flow->set = CAKE_SET_DECAYING; 2132 } else { 2133 /* remove empty queue from the flowchain */ 2134 list_del_init(&flow->flowchain); 2135 if (flow->set == CAKE_SET_SPARSE || 2136 flow->set == CAKE_SET_SPARSE_WAIT) 2137 b->sparse_flow_count--; 2138 else if (flow->set == CAKE_SET_BULK) { 2139 b->bulk_flow_count--; 2140 2141 if (cake_dsrc(q->flow_mode)) 2142 srchost->srchost_bulk_flow_count--; 2143 2144 if (cake_ddst(q->flow_mode)) 2145 dsthost->dsthost_bulk_flow_count--; 2146 2147 } else 2148 b->decaying_flow_count--; 2149 2150 flow->set = CAKE_SET_NONE; 2151 } 2152 goto begin; 2153 } 2154 2155 /* Last packet in queue may be marked, shouldn't be dropped */ 2156 if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb, 2157 (b->bulk_flow_count * 2158 !!(q->rate_flags & 2159 CAKE_FLAG_INGRESS))) || 2160 !flow->head) 2161 break; 2162 2163 /* drop this packet, get another one */ 2164 if (q->rate_flags & CAKE_FLAG_INGRESS) { 2165 len = cake_advance_shaper(q, b, skb, 2166 now, true); 2167 flow->deficit -= len; 2168 b->tin_deficit -= len; 2169 } 2170 flow->dropped++; 2171 b->tin_dropped++; 2172 qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb)); 2173 qdisc_qstats_drop(sch); 2174 kfree_skb(skb); 2175 if (q->rate_flags & CAKE_FLAG_INGRESS) 2176 goto retry; 2177 } 2178 2179 b->tin_ecn_mark += !!flow->cvars.ecn_marked; 2180 qdisc_bstats_update(sch, skb); 2181 2182 /* collect delay stats */ 2183 delay = ktime_to_ns(ktime_sub(now, cobalt_get_enqueue_time(skb))); 2184 b->avge_delay = cake_ewma(b->avge_delay, delay, 8); 2185 b->peak_delay = cake_ewma(b->peak_delay, delay, 2186 delay > b->peak_delay ? 2 : 8); 2187 b->base_delay = cake_ewma(b->base_delay, delay, 2188 delay < b->base_delay ? 2 : 8); 2189 2190 len = cake_advance_shaper(q, b, skb, now, false); 2191 flow->deficit -= len; 2192 b->tin_deficit -= len; 2193 2194 if (ktime_after(q->time_next_packet, now) && sch->q.qlen) { 2195 u64 next = min(ktime_to_ns(q->time_next_packet), 2196 ktime_to_ns(q->failsafe_next_packet)); 2197 2198 qdisc_watchdog_schedule_ns(&q->watchdog, next); 2199 } else if (!sch->q.qlen) { 2200 int i; 2201 2202 for (i = 0; i < q->tin_cnt; i++) { 2203 if (q->tins[i].decaying_flow_count) { 2204 ktime_t next = \ 2205 ktime_add_ns(now, 2206 q->tins[i].cparams.target); 2207 2208 qdisc_watchdog_schedule_ns(&q->watchdog, 2209 ktime_to_ns(next)); 2210 break; 2211 } 2212 } 2213 } 2214 2215 if (q->overflow_timeout) 2216 q->overflow_timeout--; 2217 2218 return skb; 2219 } 2220 2221 static void cake_reset(struct Qdisc *sch) 2222 { 2223 u32 c; 2224 2225 for (c = 0; c < CAKE_MAX_TINS; c++) 2226 cake_clear_tin(sch, c); 2227 } 2228 2229 static const struct nla_policy cake_policy[TCA_CAKE_MAX + 1] = { 2230 [TCA_CAKE_BASE_RATE64] = { .type = NLA_U64 }, 2231 [TCA_CAKE_DIFFSERV_MODE] = { .type = NLA_U32 }, 2232 [TCA_CAKE_ATM] = { .type = NLA_U32 }, 2233 [TCA_CAKE_FLOW_MODE] = { .type = NLA_U32 }, 2234 [TCA_CAKE_OVERHEAD] = { .type = NLA_S32 }, 2235 [TCA_CAKE_RTT] = { .type = NLA_U32 }, 2236 [TCA_CAKE_TARGET] = { .type = NLA_U32 }, 2237 [TCA_CAKE_AUTORATE] = { .type = NLA_U32 }, 2238 [TCA_CAKE_MEMORY] = { .type = NLA_U32 }, 2239 [TCA_CAKE_NAT] = { .type = NLA_U32 }, 2240 [TCA_CAKE_RAW] = { .type = NLA_U32 }, 2241 [TCA_CAKE_WASH] = { .type = NLA_U32 }, 2242 [TCA_CAKE_MPU] = { .type = NLA_U32 }, 2243 [TCA_CAKE_INGRESS] = { .type = NLA_U32 }, 2244 [TCA_CAKE_ACK_FILTER] = { .type = NLA_U32 }, 2245 [TCA_CAKE_SPLIT_GSO] = { .type = NLA_U32 }, 2246 [TCA_CAKE_FWMARK] = { .type = NLA_U32 }, 2247 }; 2248 2249 static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu, 2250 u64 target_ns, u64 rtt_est_ns) 2251 { 2252 /* convert byte-rate into time-per-byte 2253 * so it will always unwedge in reasonable time. 2254 */ 2255 static const u64 MIN_RATE = 64; 2256 u32 byte_target = mtu; 2257 u64 byte_target_ns; 2258 u8 rate_shft = 0; 2259 u64 rate_ns = 0; 2260 2261 b->flow_quantum = 1514; 2262 if (rate) { 2263 b->flow_quantum = max(min(rate >> 12, 1514ULL), 300ULL); 2264 rate_shft = 34; 2265 rate_ns = ((u64)NSEC_PER_SEC) << rate_shft; 2266 rate_ns = div64_u64(rate_ns, max(MIN_RATE, rate)); 2267 while (!!(rate_ns >> 34)) { 2268 rate_ns >>= 1; 2269 rate_shft--; 2270 } 2271 } /* else unlimited, ie. zero delay */ 2272 2273 b->tin_rate_bps = rate; 2274 b->tin_rate_ns = rate_ns; 2275 b->tin_rate_shft = rate_shft; 2276 2277 byte_target_ns = (byte_target * rate_ns) >> rate_shft; 2278 2279 b->cparams.target = max((byte_target_ns * 3) / 2, target_ns); 2280 b->cparams.interval = max(rtt_est_ns + 2281 b->cparams.target - target_ns, 2282 b->cparams.target * 2); 2283 b->cparams.mtu_time = byte_target_ns; 2284 b->cparams.p_inc = 1 << 24; /* 1/256 */ 2285 b->cparams.p_dec = 1 << 20; /* 1/4096 */ 2286 } 2287 2288 static int cake_config_besteffort(struct Qdisc *sch) 2289 { 2290 struct cake_sched_data *q = qdisc_priv(sch); 2291 struct cake_tin_data *b = &q->tins[0]; 2292 u32 mtu = psched_mtu(qdisc_dev(sch)); 2293 u64 rate = q->rate_bps; 2294 2295 q->tin_cnt = 1; 2296 2297 q->tin_index = besteffort; 2298 q->tin_order = normal_order; 2299 2300 cake_set_rate(b, rate, mtu, 2301 us_to_ns(q->target), us_to_ns(q->interval)); 2302 b->tin_quantum = 65535; 2303 2304 return 0; 2305 } 2306 2307 static int cake_config_precedence(struct Qdisc *sch) 2308 { 2309 /* convert high-level (user visible) parameters into internal format */ 2310 struct cake_sched_data *q = qdisc_priv(sch); 2311 u32 mtu = psched_mtu(qdisc_dev(sch)); 2312 u64 rate = q->rate_bps; 2313 u32 quantum = 256; 2314 u32 i; 2315 2316 q->tin_cnt = 8; 2317 q->tin_index = precedence; 2318 q->tin_order = normal_order; 2319 2320 for (i = 0; i < q->tin_cnt; i++) { 2321 struct cake_tin_data *b = &q->tins[i]; 2322 2323 cake_set_rate(b, rate, mtu, us_to_ns(q->target), 2324 us_to_ns(q->interval)); 2325 2326 b->tin_quantum = max_t(u16, 1U, quantum); 2327 2328 /* calculate next class's parameters */ 2329 rate *= 7; 2330 rate >>= 3; 2331 2332 quantum *= 7; 2333 quantum >>= 3; 2334 } 2335 2336 return 0; 2337 } 2338 2339 /* List of known Diffserv codepoints: 2340 * 2341 * Least Effort (CS1) 2342 * Best Effort (CS0) 2343 * Max Reliability & LLT "Lo" (TOS1) 2344 * Max Throughput (TOS2) 2345 * Min Delay (TOS4) 2346 * LLT "La" (TOS5) 2347 * Assured Forwarding 1 (AF1x) - x3 2348 * Assured Forwarding 2 (AF2x) - x3 2349 * Assured Forwarding 3 (AF3x) - x3 2350 * Assured Forwarding 4 (AF4x) - x3 2351 * Precedence Class 2 (CS2) 2352 * Precedence Class 3 (CS3) 2353 * Precedence Class 4 (CS4) 2354 * Precedence Class 5 (CS5) 2355 * Precedence Class 6 (CS6) 2356 * Precedence Class 7 (CS7) 2357 * Voice Admit (VA) 2358 * Expedited Forwarding (EF) 2359 2360 * Total 25 codepoints. 2361 */ 2362 2363 /* List of traffic classes in RFC 4594: 2364 * (roughly descending order of contended priority) 2365 * (roughly ascending order of uncontended throughput) 2366 * 2367 * Network Control (CS6,CS7) - routing traffic 2368 * Telephony (EF,VA) - aka. VoIP streams 2369 * Signalling (CS5) - VoIP setup 2370 * Multimedia Conferencing (AF4x) - aka. video calls 2371 * Realtime Interactive (CS4) - eg. games 2372 * Multimedia Streaming (AF3x) - eg. YouTube, NetFlix, Twitch 2373 * Broadcast Video (CS3) 2374 * Low Latency Data (AF2x,TOS4) - eg. database 2375 * Ops, Admin, Management (CS2,TOS1) - eg. ssh 2376 * Standard Service (CS0 & unrecognised codepoints) 2377 * High Throughput Data (AF1x,TOS2) - eg. web traffic 2378 * Low Priority Data (CS1) - eg. BitTorrent 2379 2380 * Total 12 traffic classes. 2381 */ 2382 2383 static int cake_config_diffserv8(struct Qdisc *sch) 2384 { 2385 /* Pruned list of traffic classes for typical applications: 2386 * 2387 * Network Control (CS6, CS7) 2388 * Minimum Latency (EF, VA, CS5, CS4) 2389 * Interactive Shell (CS2, TOS1) 2390 * Low Latency Transactions (AF2x, TOS4) 2391 * Video Streaming (AF4x, AF3x, CS3) 2392 * Bog Standard (CS0 etc.) 2393 * High Throughput (AF1x, TOS2) 2394 * Background Traffic (CS1) 2395 * 2396 * Total 8 traffic classes. 2397 */ 2398 2399 struct cake_sched_data *q = qdisc_priv(sch); 2400 u32 mtu = psched_mtu(qdisc_dev(sch)); 2401 u64 rate = q->rate_bps; 2402 u32 quantum = 256; 2403 u32 i; 2404 2405 q->tin_cnt = 8; 2406 2407 /* codepoint to class mapping */ 2408 q->tin_index = diffserv8; 2409 q->tin_order = normal_order; 2410 2411 /* class characteristics */ 2412 for (i = 0; i < q->tin_cnt; i++) { 2413 struct cake_tin_data *b = &q->tins[i]; 2414 2415 cake_set_rate(b, rate, mtu, us_to_ns(q->target), 2416 us_to_ns(q->interval)); 2417 2418 b->tin_quantum = max_t(u16, 1U, quantum); 2419 2420 /* calculate next class's parameters */ 2421 rate *= 7; 2422 rate >>= 3; 2423 2424 quantum *= 7; 2425 quantum >>= 3; 2426 } 2427 2428 return 0; 2429 } 2430 2431 static int cake_config_diffserv4(struct Qdisc *sch) 2432 { 2433 /* Further pruned list of traffic classes for four-class system: 2434 * 2435 * Latency Sensitive (CS7, CS6, EF, VA, CS5, CS4) 2436 * Streaming Media (AF4x, AF3x, CS3, AF2x, TOS4, CS2, TOS1) 2437 * Best Effort (CS0, AF1x, TOS2, and those not specified) 2438 * Background Traffic (CS1) 2439 * 2440 * Total 4 traffic classes. 2441 */ 2442 2443 struct cake_sched_data *q = qdisc_priv(sch); 2444 u32 mtu = psched_mtu(qdisc_dev(sch)); 2445 u64 rate = q->rate_bps; 2446 u32 quantum = 1024; 2447 2448 q->tin_cnt = 4; 2449 2450 /* codepoint to class mapping */ 2451 q->tin_index = diffserv4; 2452 q->tin_order = bulk_order; 2453 2454 /* class characteristics */ 2455 cake_set_rate(&q->tins[0], rate, mtu, 2456 us_to_ns(q->target), us_to_ns(q->interval)); 2457 cake_set_rate(&q->tins[1], rate >> 4, mtu, 2458 us_to_ns(q->target), us_to_ns(q->interval)); 2459 cake_set_rate(&q->tins[2], rate >> 1, mtu, 2460 us_to_ns(q->target), us_to_ns(q->interval)); 2461 cake_set_rate(&q->tins[3], rate >> 2, mtu, 2462 us_to_ns(q->target), us_to_ns(q->interval)); 2463 2464 /* bandwidth-sharing weights */ 2465 q->tins[0].tin_quantum = quantum; 2466 q->tins[1].tin_quantum = quantum >> 4; 2467 q->tins[2].tin_quantum = quantum >> 1; 2468 q->tins[3].tin_quantum = quantum >> 2; 2469 2470 return 0; 2471 } 2472 2473 static int cake_config_diffserv3(struct Qdisc *sch) 2474 { 2475 /* Simplified Diffserv structure with 3 tins. 2476 * Low Priority (CS1) 2477 * Best Effort 2478 * Latency Sensitive (TOS4, VA, EF, CS6, CS7) 2479 */ 2480 struct cake_sched_data *q = qdisc_priv(sch); 2481 u32 mtu = psched_mtu(qdisc_dev(sch)); 2482 u64 rate = q->rate_bps; 2483 u32 quantum = 1024; 2484 2485 q->tin_cnt = 3; 2486 2487 /* codepoint to class mapping */ 2488 q->tin_index = diffserv3; 2489 q->tin_order = bulk_order; 2490 2491 /* class characteristics */ 2492 cake_set_rate(&q->tins[0], rate, mtu, 2493 us_to_ns(q->target), us_to_ns(q->interval)); 2494 cake_set_rate(&q->tins[1], rate >> 4, mtu, 2495 us_to_ns(q->target), us_to_ns(q->interval)); 2496 cake_set_rate(&q->tins[2], rate >> 2, mtu, 2497 us_to_ns(q->target), us_to_ns(q->interval)); 2498 2499 /* bandwidth-sharing weights */ 2500 q->tins[0].tin_quantum = quantum; 2501 q->tins[1].tin_quantum = quantum >> 4; 2502 q->tins[2].tin_quantum = quantum >> 2; 2503 2504 return 0; 2505 } 2506 2507 static void cake_reconfigure(struct Qdisc *sch) 2508 { 2509 struct cake_sched_data *q = qdisc_priv(sch); 2510 int c, ft; 2511 2512 switch (q->tin_mode) { 2513 case CAKE_DIFFSERV_BESTEFFORT: 2514 ft = cake_config_besteffort(sch); 2515 break; 2516 2517 case CAKE_DIFFSERV_PRECEDENCE: 2518 ft = cake_config_precedence(sch); 2519 break; 2520 2521 case CAKE_DIFFSERV_DIFFSERV8: 2522 ft = cake_config_diffserv8(sch); 2523 break; 2524 2525 case CAKE_DIFFSERV_DIFFSERV4: 2526 ft = cake_config_diffserv4(sch); 2527 break; 2528 2529 case CAKE_DIFFSERV_DIFFSERV3: 2530 default: 2531 ft = cake_config_diffserv3(sch); 2532 break; 2533 } 2534 2535 for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) { 2536 cake_clear_tin(sch, c); 2537 q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time; 2538 } 2539 2540 q->rate_ns = q->tins[ft].tin_rate_ns; 2541 q->rate_shft = q->tins[ft].tin_rate_shft; 2542 2543 if (q->buffer_config_limit) { 2544 q->buffer_limit = q->buffer_config_limit; 2545 } else if (q->rate_bps) { 2546 u64 t = q->rate_bps * q->interval; 2547 2548 do_div(t, USEC_PER_SEC / 4); 2549 q->buffer_limit = max_t(u32, t, 4U << 20); 2550 } else { 2551 q->buffer_limit = ~0; 2552 } 2553 2554 sch->flags &= ~TCQ_F_CAN_BYPASS; 2555 2556 q->buffer_limit = min(q->buffer_limit, 2557 max(sch->limit * psched_mtu(qdisc_dev(sch)), 2558 q->buffer_config_limit)); 2559 } 2560 2561 static int cake_change(struct Qdisc *sch, struct nlattr *opt, 2562 struct netlink_ext_ack *extack) 2563 { 2564 struct cake_sched_data *q = qdisc_priv(sch); 2565 struct nlattr *tb[TCA_CAKE_MAX + 1]; 2566 int err; 2567 2568 if (!opt) 2569 return -EINVAL; 2570 2571 err = nla_parse_nested_deprecated(tb, TCA_CAKE_MAX, opt, cake_policy, 2572 extack); 2573 if (err < 0) 2574 return err; 2575 2576 if (tb[TCA_CAKE_NAT]) { 2577 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 2578 q->flow_mode &= ~CAKE_FLOW_NAT_FLAG; 2579 q->flow_mode |= CAKE_FLOW_NAT_FLAG * 2580 !!nla_get_u32(tb[TCA_CAKE_NAT]); 2581 #else 2582 NL_SET_ERR_MSG_ATTR(extack, tb[TCA_CAKE_NAT], 2583 "No conntrack support in kernel"); 2584 return -EOPNOTSUPP; 2585 #endif 2586 } 2587 2588 if (tb[TCA_CAKE_BASE_RATE64]) 2589 q->rate_bps = nla_get_u64(tb[TCA_CAKE_BASE_RATE64]); 2590 2591 if (tb[TCA_CAKE_DIFFSERV_MODE]) 2592 q->tin_mode = nla_get_u32(tb[TCA_CAKE_DIFFSERV_MODE]); 2593 2594 if (tb[TCA_CAKE_WASH]) { 2595 if (!!nla_get_u32(tb[TCA_CAKE_WASH])) 2596 q->rate_flags |= CAKE_FLAG_WASH; 2597 else 2598 q->rate_flags &= ~CAKE_FLAG_WASH; 2599 } 2600 2601 if (tb[TCA_CAKE_FLOW_MODE]) 2602 q->flow_mode = ((q->flow_mode & CAKE_FLOW_NAT_FLAG) | 2603 (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) & 2604 CAKE_FLOW_MASK)); 2605 2606 if (tb[TCA_CAKE_ATM]) 2607 q->atm_mode = nla_get_u32(tb[TCA_CAKE_ATM]); 2608 2609 if (tb[TCA_CAKE_OVERHEAD]) { 2610 q->rate_overhead = nla_get_s32(tb[TCA_CAKE_OVERHEAD]); 2611 q->rate_flags |= CAKE_FLAG_OVERHEAD; 2612 2613 q->max_netlen = 0; 2614 q->max_adjlen = 0; 2615 q->min_netlen = ~0; 2616 q->min_adjlen = ~0; 2617 } 2618 2619 if (tb[TCA_CAKE_RAW]) { 2620 q->rate_flags &= ~CAKE_FLAG_OVERHEAD; 2621 2622 q->max_netlen = 0; 2623 q->max_adjlen = 0; 2624 q->min_netlen = ~0; 2625 q->min_adjlen = ~0; 2626 } 2627 2628 if (tb[TCA_CAKE_MPU]) 2629 q->rate_mpu = nla_get_u32(tb[TCA_CAKE_MPU]); 2630 2631 if (tb[TCA_CAKE_RTT]) { 2632 q->interval = nla_get_u32(tb[TCA_CAKE_RTT]); 2633 2634 if (!q->interval) 2635 q->interval = 1; 2636 } 2637 2638 if (tb[TCA_CAKE_TARGET]) { 2639 q->target = nla_get_u32(tb[TCA_CAKE_TARGET]); 2640 2641 if (!q->target) 2642 q->target = 1; 2643 } 2644 2645 if (tb[TCA_CAKE_AUTORATE]) { 2646 if (!!nla_get_u32(tb[TCA_CAKE_AUTORATE])) 2647 q->rate_flags |= CAKE_FLAG_AUTORATE_INGRESS; 2648 else 2649 q->rate_flags &= ~CAKE_FLAG_AUTORATE_INGRESS; 2650 } 2651 2652 if (tb[TCA_CAKE_INGRESS]) { 2653 if (!!nla_get_u32(tb[TCA_CAKE_INGRESS])) 2654 q->rate_flags |= CAKE_FLAG_INGRESS; 2655 else 2656 q->rate_flags &= ~CAKE_FLAG_INGRESS; 2657 } 2658 2659 if (tb[TCA_CAKE_ACK_FILTER]) 2660 q->ack_filter = nla_get_u32(tb[TCA_CAKE_ACK_FILTER]); 2661 2662 if (tb[TCA_CAKE_MEMORY]) 2663 q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]); 2664 2665 if (tb[TCA_CAKE_SPLIT_GSO]) { 2666 if (!!nla_get_u32(tb[TCA_CAKE_SPLIT_GSO])) 2667 q->rate_flags |= CAKE_FLAG_SPLIT_GSO; 2668 else 2669 q->rate_flags &= ~CAKE_FLAG_SPLIT_GSO; 2670 } 2671 2672 if (tb[TCA_CAKE_FWMARK]) { 2673 q->fwmark_mask = nla_get_u32(tb[TCA_CAKE_FWMARK]); 2674 q->fwmark_shft = q->fwmark_mask ? __ffs(q->fwmark_mask) : 0; 2675 } 2676 2677 if (q->tins) { 2678 sch_tree_lock(sch); 2679 cake_reconfigure(sch); 2680 sch_tree_unlock(sch); 2681 } 2682 2683 return 0; 2684 } 2685 2686 static void cake_destroy(struct Qdisc *sch) 2687 { 2688 struct cake_sched_data *q = qdisc_priv(sch); 2689 2690 qdisc_watchdog_cancel(&q->watchdog); 2691 tcf_block_put(q->block); 2692 kvfree(q->tins); 2693 } 2694 2695 static int cake_init(struct Qdisc *sch, struct nlattr *opt, 2696 struct netlink_ext_ack *extack) 2697 { 2698 struct cake_sched_data *q = qdisc_priv(sch); 2699 int i, j, err; 2700 2701 sch->limit = 10240; 2702 q->tin_mode = CAKE_DIFFSERV_DIFFSERV3; 2703 q->flow_mode = CAKE_FLOW_TRIPLE; 2704 2705 q->rate_bps = 0; /* unlimited by default */ 2706 2707 q->interval = 100000; /* 100ms default */ 2708 q->target = 5000; /* 5ms: codel RFC argues 2709 * for 5 to 10% of interval 2710 */ 2711 q->rate_flags |= CAKE_FLAG_SPLIT_GSO; 2712 q->cur_tin = 0; 2713 q->cur_flow = 0; 2714 2715 qdisc_watchdog_init(&q->watchdog, sch); 2716 2717 if (opt) { 2718 err = cake_change(sch, opt, extack); 2719 2720 if (err) 2721 return err; 2722 } 2723 2724 err = tcf_block_get(&q->block, &q->filter_list, sch, extack); 2725 if (err) 2726 return err; 2727 2728 quantum_div[0] = ~0; 2729 for (i = 1; i <= CAKE_QUEUES; i++) 2730 quantum_div[i] = 65535 / i; 2731 2732 q->tins = kvcalloc(CAKE_MAX_TINS, sizeof(struct cake_tin_data), 2733 GFP_KERNEL); 2734 if (!q->tins) 2735 goto nomem; 2736 2737 for (i = 0; i < CAKE_MAX_TINS; i++) { 2738 struct cake_tin_data *b = q->tins + i; 2739 2740 INIT_LIST_HEAD(&b->new_flows); 2741 INIT_LIST_HEAD(&b->old_flows); 2742 INIT_LIST_HEAD(&b->decaying_flows); 2743 b->sparse_flow_count = 0; 2744 b->bulk_flow_count = 0; 2745 b->decaying_flow_count = 0; 2746 2747 for (j = 0; j < CAKE_QUEUES; j++) { 2748 struct cake_flow *flow = b->flows + j; 2749 u32 k = j * CAKE_MAX_TINS + i; 2750 2751 INIT_LIST_HEAD(&flow->flowchain); 2752 cobalt_vars_init(&flow->cvars); 2753 2754 q->overflow_heap[k].t = i; 2755 q->overflow_heap[k].b = j; 2756 b->overflow_idx[j] = k; 2757 } 2758 } 2759 2760 cake_reconfigure(sch); 2761 q->avg_peak_bandwidth = q->rate_bps; 2762 q->min_netlen = ~0; 2763 q->min_adjlen = ~0; 2764 return 0; 2765 2766 nomem: 2767 cake_destroy(sch); 2768 return -ENOMEM; 2769 } 2770 2771 static int cake_dump(struct Qdisc *sch, struct sk_buff *skb) 2772 { 2773 struct cake_sched_data *q = qdisc_priv(sch); 2774 struct nlattr *opts; 2775 2776 opts = nla_nest_start_noflag(skb, TCA_OPTIONS); 2777 if (!opts) 2778 goto nla_put_failure; 2779 2780 if (nla_put_u64_64bit(skb, TCA_CAKE_BASE_RATE64, q->rate_bps, 2781 TCA_CAKE_PAD)) 2782 goto nla_put_failure; 2783 2784 if (nla_put_u32(skb, TCA_CAKE_FLOW_MODE, 2785 q->flow_mode & CAKE_FLOW_MASK)) 2786 goto nla_put_failure; 2787 2788 if (nla_put_u32(skb, TCA_CAKE_RTT, q->interval)) 2789 goto nla_put_failure; 2790 2791 if (nla_put_u32(skb, TCA_CAKE_TARGET, q->target)) 2792 goto nla_put_failure; 2793 2794 if (nla_put_u32(skb, TCA_CAKE_MEMORY, q->buffer_config_limit)) 2795 goto nla_put_failure; 2796 2797 if (nla_put_u32(skb, TCA_CAKE_AUTORATE, 2798 !!(q->rate_flags & CAKE_FLAG_AUTORATE_INGRESS))) 2799 goto nla_put_failure; 2800 2801 if (nla_put_u32(skb, TCA_CAKE_INGRESS, 2802 !!(q->rate_flags & CAKE_FLAG_INGRESS))) 2803 goto nla_put_failure; 2804 2805 if (nla_put_u32(skb, TCA_CAKE_ACK_FILTER, q->ack_filter)) 2806 goto nla_put_failure; 2807 2808 if (nla_put_u32(skb, TCA_CAKE_NAT, 2809 !!(q->flow_mode & CAKE_FLOW_NAT_FLAG))) 2810 goto nla_put_failure; 2811 2812 if (nla_put_u32(skb, TCA_CAKE_DIFFSERV_MODE, q->tin_mode)) 2813 goto nla_put_failure; 2814 2815 if (nla_put_u32(skb, TCA_CAKE_WASH, 2816 !!(q->rate_flags & CAKE_FLAG_WASH))) 2817 goto nla_put_failure; 2818 2819 if (nla_put_u32(skb, TCA_CAKE_OVERHEAD, q->rate_overhead)) 2820 goto nla_put_failure; 2821 2822 if (!(q->rate_flags & CAKE_FLAG_OVERHEAD)) 2823 if (nla_put_u32(skb, TCA_CAKE_RAW, 0)) 2824 goto nla_put_failure; 2825 2826 if (nla_put_u32(skb, TCA_CAKE_ATM, q->atm_mode)) 2827 goto nla_put_failure; 2828 2829 if (nla_put_u32(skb, TCA_CAKE_MPU, q->rate_mpu)) 2830 goto nla_put_failure; 2831 2832 if (nla_put_u32(skb, TCA_CAKE_SPLIT_GSO, 2833 !!(q->rate_flags & CAKE_FLAG_SPLIT_GSO))) 2834 goto nla_put_failure; 2835 2836 if (nla_put_u32(skb, TCA_CAKE_FWMARK, q->fwmark_mask)) 2837 goto nla_put_failure; 2838 2839 return nla_nest_end(skb, opts); 2840 2841 nla_put_failure: 2842 return -1; 2843 } 2844 2845 static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d) 2846 { 2847 struct nlattr *stats = nla_nest_start_noflag(d->skb, TCA_STATS_APP); 2848 struct cake_sched_data *q = qdisc_priv(sch); 2849 struct nlattr *tstats, *ts; 2850 int i; 2851 2852 if (!stats) 2853 return -1; 2854 2855 #define PUT_STAT_U32(attr, data) do { \ 2856 if (nla_put_u32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \ 2857 goto nla_put_failure; \ 2858 } while (0) 2859 #define PUT_STAT_U64(attr, data) do { \ 2860 if (nla_put_u64_64bit(d->skb, TCA_CAKE_STATS_ ## attr, \ 2861 data, TCA_CAKE_STATS_PAD)) \ 2862 goto nla_put_failure; \ 2863 } while (0) 2864 2865 PUT_STAT_U64(CAPACITY_ESTIMATE64, q->avg_peak_bandwidth); 2866 PUT_STAT_U32(MEMORY_LIMIT, q->buffer_limit); 2867 PUT_STAT_U32(MEMORY_USED, q->buffer_max_used); 2868 PUT_STAT_U32(AVG_NETOFF, ((q->avg_netoff + 0x8000) >> 16)); 2869 PUT_STAT_U32(MAX_NETLEN, q->max_netlen); 2870 PUT_STAT_U32(MAX_ADJLEN, q->max_adjlen); 2871 PUT_STAT_U32(MIN_NETLEN, q->min_netlen); 2872 PUT_STAT_U32(MIN_ADJLEN, q->min_adjlen); 2873 2874 #undef PUT_STAT_U32 2875 #undef PUT_STAT_U64 2876 2877 tstats = nla_nest_start_noflag(d->skb, TCA_CAKE_STATS_TIN_STATS); 2878 if (!tstats) 2879 goto nla_put_failure; 2880 2881 #define PUT_TSTAT_U32(attr, data) do { \ 2882 if (nla_put_u32(d->skb, TCA_CAKE_TIN_STATS_ ## attr, data)) \ 2883 goto nla_put_failure; \ 2884 } while (0) 2885 #define PUT_TSTAT_U64(attr, data) do { \ 2886 if (nla_put_u64_64bit(d->skb, TCA_CAKE_TIN_STATS_ ## attr, \ 2887 data, TCA_CAKE_TIN_STATS_PAD)) \ 2888 goto nla_put_failure; \ 2889 } while (0) 2890 2891 for (i = 0; i < q->tin_cnt; i++) { 2892 struct cake_tin_data *b = &q->tins[q->tin_order[i]]; 2893 2894 ts = nla_nest_start_noflag(d->skb, i + 1); 2895 if (!ts) 2896 goto nla_put_failure; 2897 2898 PUT_TSTAT_U64(THRESHOLD_RATE64, b->tin_rate_bps); 2899 PUT_TSTAT_U64(SENT_BYTES64, b->bytes); 2900 PUT_TSTAT_U32(BACKLOG_BYTES, b->tin_backlog); 2901 2902 PUT_TSTAT_U32(TARGET_US, 2903 ktime_to_us(ns_to_ktime(b->cparams.target))); 2904 PUT_TSTAT_U32(INTERVAL_US, 2905 ktime_to_us(ns_to_ktime(b->cparams.interval))); 2906 2907 PUT_TSTAT_U32(SENT_PACKETS, b->packets); 2908 PUT_TSTAT_U32(DROPPED_PACKETS, b->tin_dropped); 2909 PUT_TSTAT_U32(ECN_MARKED_PACKETS, b->tin_ecn_mark); 2910 PUT_TSTAT_U32(ACKS_DROPPED_PACKETS, b->ack_drops); 2911 2912 PUT_TSTAT_U32(PEAK_DELAY_US, 2913 ktime_to_us(ns_to_ktime(b->peak_delay))); 2914 PUT_TSTAT_U32(AVG_DELAY_US, 2915 ktime_to_us(ns_to_ktime(b->avge_delay))); 2916 PUT_TSTAT_U32(BASE_DELAY_US, 2917 ktime_to_us(ns_to_ktime(b->base_delay))); 2918 2919 PUT_TSTAT_U32(WAY_INDIRECT_HITS, b->way_hits); 2920 PUT_TSTAT_U32(WAY_MISSES, b->way_misses); 2921 PUT_TSTAT_U32(WAY_COLLISIONS, b->way_collisions); 2922 2923 PUT_TSTAT_U32(SPARSE_FLOWS, b->sparse_flow_count + 2924 b->decaying_flow_count); 2925 PUT_TSTAT_U32(BULK_FLOWS, b->bulk_flow_count); 2926 PUT_TSTAT_U32(UNRESPONSIVE_FLOWS, b->unresponsive_flow_count); 2927 PUT_TSTAT_U32(MAX_SKBLEN, b->max_skblen); 2928 2929 PUT_TSTAT_U32(FLOW_QUANTUM, b->flow_quantum); 2930 nla_nest_end(d->skb, ts); 2931 } 2932 2933 #undef PUT_TSTAT_U32 2934 #undef PUT_TSTAT_U64 2935 2936 nla_nest_end(d->skb, tstats); 2937 return nla_nest_end(d->skb, stats); 2938 2939 nla_put_failure: 2940 nla_nest_cancel(d->skb, stats); 2941 return -1; 2942 } 2943 2944 static struct Qdisc *cake_leaf(struct Qdisc *sch, unsigned long arg) 2945 { 2946 return NULL; 2947 } 2948 2949 static unsigned long cake_find(struct Qdisc *sch, u32 classid) 2950 { 2951 return 0; 2952 } 2953 2954 static unsigned long cake_bind(struct Qdisc *sch, unsigned long parent, 2955 u32 classid) 2956 { 2957 return 0; 2958 } 2959 2960 static void cake_unbind(struct Qdisc *q, unsigned long cl) 2961 { 2962 } 2963 2964 static struct tcf_block *cake_tcf_block(struct Qdisc *sch, unsigned long cl, 2965 struct netlink_ext_ack *extack) 2966 { 2967 struct cake_sched_data *q = qdisc_priv(sch); 2968 2969 if (cl) 2970 return NULL; 2971 return q->block; 2972 } 2973 2974 static int cake_dump_class(struct Qdisc *sch, unsigned long cl, 2975 struct sk_buff *skb, struct tcmsg *tcm) 2976 { 2977 tcm->tcm_handle |= TC_H_MIN(cl); 2978 return 0; 2979 } 2980 2981 static int cake_dump_class_stats(struct Qdisc *sch, unsigned long cl, 2982 struct gnet_dump *d) 2983 { 2984 struct cake_sched_data *q = qdisc_priv(sch); 2985 const struct cake_flow *flow = NULL; 2986 struct gnet_stats_queue qs = { 0 }; 2987 struct nlattr *stats; 2988 u32 idx = cl - 1; 2989 2990 if (idx < CAKE_QUEUES * q->tin_cnt) { 2991 const struct cake_tin_data *b = \ 2992 &q->tins[q->tin_order[idx / CAKE_QUEUES]]; 2993 const struct sk_buff *skb; 2994 2995 flow = &b->flows[idx % CAKE_QUEUES]; 2996 2997 if (flow->head) { 2998 sch_tree_lock(sch); 2999 skb = flow->head; 3000 while (skb) { 3001 qs.qlen++; 3002 skb = skb->next; 3003 } 3004 sch_tree_unlock(sch); 3005 } 3006 qs.backlog = b->backlogs[idx % CAKE_QUEUES]; 3007 qs.drops = flow->dropped; 3008 } 3009 if (gnet_stats_copy_queue(d, NULL, &qs, qs.qlen) < 0) 3010 return -1; 3011 if (flow) { 3012 ktime_t now = ktime_get(); 3013 3014 stats = nla_nest_start_noflag(d->skb, TCA_STATS_APP); 3015 if (!stats) 3016 return -1; 3017 3018 #define PUT_STAT_U32(attr, data) do { \ 3019 if (nla_put_u32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \ 3020 goto nla_put_failure; \ 3021 } while (0) 3022 #define PUT_STAT_S32(attr, data) do { \ 3023 if (nla_put_s32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \ 3024 goto nla_put_failure; \ 3025 } while (0) 3026 3027 PUT_STAT_S32(DEFICIT, flow->deficit); 3028 PUT_STAT_U32(DROPPING, flow->cvars.dropping); 3029 PUT_STAT_U32(COBALT_COUNT, flow->cvars.count); 3030 PUT_STAT_U32(P_DROP, flow->cvars.p_drop); 3031 if (flow->cvars.p_drop) { 3032 PUT_STAT_S32(BLUE_TIMER_US, 3033 ktime_to_us( 3034 ktime_sub(now, 3035 flow->cvars.blue_timer))); 3036 } 3037 if (flow->cvars.dropping) { 3038 PUT_STAT_S32(DROP_NEXT_US, 3039 ktime_to_us( 3040 ktime_sub(now, 3041 flow->cvars.drop_next))); 3042 } 3043 3044 if (nla_nest_end(d->skb, stats) < 0) 3045 return -1; 3046 } 3047 3048 return 0; 3049 3050 nla_put_failure: 3051 nla_nest_cancel(d->skb, stats); 3052 return -1; 3053 } 3054 3055 static void cake_walk(struct Qdisc *sch, struct qdisc_walker *arg) 3056 { 3057 struct cake_sched_data *q = qdisc_priv(sch); 3058 unsigned int i, j; 3059 3060 if (arg->stop) 3061 return; 3062 3063 for (i = 0; i < q->tin_cnt; i++) { 3064 struct cake_tin_data *b = &q->tins[q->tin_order[i]]; 3065 3066 for (j = 0; j < CAKE_QUEUES; j++) { 3067 if (list_empty(&b->flows[j].flowchain) || 3068 arg->count < arg->skip) { 3069 arg->count++; 3070 continue; 3071 } 3072 if (arg->fn(sch, i * CAKE_QUEUES + j + 1, arg) < 0) { 3073 arg->stop = 1; 3074 break; 3075 } 3076 arg->count++; 3077 } 3078 } 3079 } 3080 3081 static const struct Qdisc_class_ops cake_class_ops = { 3082 .leaf = cake_leaf, 3083 .find = cake_find, 3084 .tcf_block = cake_tcf_block, 3085 .bind_tcf = cake_bind, 3086 .unbind_tcf = cake_unbind, 3087 .dump = cake_dump_class, 3088 .dump_stats = cake_dump_class_stats, 3089 .walk = cake_walk, 3090 }; 3091 3092 static struct Qdisc_ops cake_qdisc_ops __read_mostly = { 3093 .cl_ops = &cake_class_ops, 3094 .id = "cake", 3095 .priv_size = sizeof(struct cake_sched_data), 3096 .enqueue = cake_enqueue, 3097 .dequeue = cake_dequeue, 3098 .peek = qdisc_peek_dequeued, 3099 .init = cake_init, 3100 .reset = cake_reset, 3101 .destroy = cake_destroy, 3102 .change = cake_change, 3103 .dump = cake_dump, 3104 .dump_stats = cake_dump_stats, 3105 .owner = THIS_MODULE, 3106 }; 3107 3108 static int __init cake_module_init(void) 3109 { 3110 return register_qdisc(&cake_qdisc_ops); 3111 } 3112 3113 static void __exit cake_module_exit(void) 3114 { 3115 unregister_qdisc(&cake_qdisc_ops); 3116 } 3117 3118 module_init(cake_module_init) 3119 module_exit(cake_module_exit) 3120 MODULE_AUTHOR("Jonathan Morton"); 3121 MODULE_LICENSE("Dual BSD/GPL"); 3122 MODULE_DESCRIPTION("The CAKE shaper."); 3123