kernel/net/sched/sch_fq.c

   1 /*
   2  * net/sched/sch_fq.c Fair Queue Packet Scheduler (per flow pacing)
   3  *
   4  *  Copyright (C) 2013-2015 Eric Dumazet <edumazet@google.com>
   5  *
   6  *      This program is free software; you can redistribute it and/or
   7  *      modify it under the terms of the GNU General Public License
   8  *      as published by the Free Software Foundation; either version
   9  *      2 of the License, or (at your option) any later version.
  10  *
  11  *  Meant to be mostly used for locally generated traffic :
  12  *  Fast classification depends on skb->sk being set before reaching us.
  13  *  If not, (router workload), we use rxhash as fallback, with 32 bits wide hash.
  14  *  All packets belonging to a socket are considered as a 'flow'.
  15  *
  16  *  Flows are dynamically allocated and stored in a hash table of RB trees
  17  *  They are also part of one Round Robin 'queues' (new or old flows)
  18  *
  19  *  Burst avoidance (aka pacing) capability :
  20  *
  21  *  Transport (eg TCP) can set in sk->sk_pacing_rate a rate, enqueue a
  22  *  bunch of packets, and this packet scheduler adds delay between
  23  *  packets to respect rate limitation.
  24  *
  25  *  enqueue() :
  26  *   - lookup one RB tree (out of 1024 or more) to find the flow.
  27  *     If non existent flow, create it, add it to the tree.
  28  *     Add skb to the per flow list of skb (fifo).
  29  *   - Use a special fifo for high prio packets
  30  *
  31  *  dequeue() : serves flows in Round Robin
  32  *  Note : When a flow becomes empty, we do not immediately remove it from
  33  *  rb trees, for performance reasons (its expected to send additional packets,
  34  *  or SLAB cache will reuse socket for another flow)
  35  */
  36
  37 #include <linux/module.h>
  38 #include <linux/types.h>
  39 #include <linux/kernel.h>
  40 #include <linux/jiffies.h>
  41 #include <linux/string.h>
  42 #include <linux/in.h>
  43 #include <linux/errno.h>
  44 #include <linux/init.h>
  45 #include <linux/skbuff.h>
  46 #include <linux/slab.h>
  47 #include <linux/rbtree.h>
  48 #include <linux/hash.h>
  49 #include <linux/prefetch.h>
  50 #include <linux/vmalloc.h>
  51 #include <net/netlink.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/sock.h>
  54 #include <net/tcp_states.h>
  55 #include <net/tcp.h>
  56
  57 /*
  58  * Per flow structure, dynamically allocated
  59  */
  60 struct fq_flow {
  61         struct sk_buff  *head;          /* list of skbs for this flow : first skb */
  62         union {
  63                 struct sk_buff *tail;   /* last skb in the list */
  64                 unsigned long  age;     /* jiffies when flow was emptied, for gc */
  65         };
  66         struct rb_node  fq_node;        /* anchor in fq_root[] trees */
  67         struct sock     *sk;
  68         int             qlen;           /* number of packets in flow queue */
  69         int             credit;
  70         u32             socket_hash;    /* sk_hash */
  71         struct fq_flow *next;           /* next pointer in RR lists, or &detached */
  72
  73         struct rb_node  rate_node;      /* anchor in q->delayed tree */
  74         u64             time_next_packet;
  75 };
  76
  77 struct fq_flow_head {
  78         struct fq_flow *first;
  79         struct fq_flow *last;
  80 };
  81
  82 struct fq_sched_data {
  83         struct fq_flow_head new_flows;
  84
  85         struct fq_flow_head old_flows;
  86
  87         struct rb_root  delayed;        /* for rate limited flows */
  88         u64             time_next_delayed_flow;
  89
  90         struct fq_flow  internal;       /* for non classified or high prio packets */
  91         u32             quantum;
  92         u32             initial_quantum;
  93         u32             flow_refill_delay;
  94         u32             flow_max_rate;  /* optional max rate per flow */
  95         u32             flow_plimit;    /* max packets per flow */
  96         u32             orphan_mask;    /* mask for orphaned skb */
  97         struct rb_root  *fq_root;
  98         u8              rate_enable;
  99         u8              fq_trees_log;
 100
 101         u32             flows;
 102         u32             inactive_flows;
 103         u32             throttled_flows;
 104
 105         u64             stat_gc_flows;
 106         u64             stat_internal_packets;
 107         u64             stat_tcp_retrans;
 108         u64             stat_throttled;
 109         u64             stat_flows_plimit;
 110         u64             stat_pkts_too_long;
 111         u64             stat_allocation_errors;
 112         struct qdisc_watchdog watchdog;
 113 };
 114
 115 /* special value to mark a detached flow (not on old/new list) */
 116 static struct fq_flow detached, throttled;
 117
 118 static void fq_flow_set_detached(struct fq_flow *f)
 119 {
 120         f->next = &detached;
 121         f->age = jiffies;
 122 }
 123
 124 static bool fq_flow_is_detached(const struct fq_flow *f)
 125 {
 126         return f->next == &detached;
 127 }
 128
 129 static void fq_flow_set_throttled(struct fq_sched_data *q, struct fq_flow *f)
 130 {
 131         struct rb_node **p = &q->delayed.rb_node, *parent = NULL;
 132
 133         while (*p) {
 134                 struct fq_flow *aux;
 135
 136                 parent = *p;
 137                 aux = container_of(parent, struct fq_flow, rate_node);
 138                 if (f->time_next_packet >= aux->time_next_packet)
 139                         p = &parent->rb_right;
 140                 else
 141                         p = &parent->rb_left;
 142         }
 143         rb_link_node(&f->rate_node, parent, p);
 144         rb_insert_color(&f->rate_node, &q->delayed);
 145         q->throttled_flows++;
 146         q->stat_throttled++;
 147
 148         f->next = &throttled;
 149         if (q->time_next_delayed_flow > f->time_next_packet)
 150                 q->time_next_delayed_flow = f->time_next_packet;
 151 }
 152
 153
 154 static struct kmem_cache *fq_flow_cachep __read_mostly;
 155
 156 static void fq_flow_add_tail(struct fq_flow_head *head, struct fq_flow *flow)
 157 {
 158         if (head->first)
 159                 head->last->next = flow;
 160         else
 161                 head->first = flow;
 162         head->last = flow;
 163         flow->next = NULL;
 164 }
 165
 166 /* limit number of collected flows per round */
 167 #define FQ_GC_MAX 8
 168 #define FQ_GC_AGE (3*HZ)
 169
 170 static bool fq_gc_candidate(const struct fq_flow *f)
 171 {
 172         return fq_flow_is_detached(f) &&
 173                time_after(jiffies, f->age + FQ_GC_AGE);
 174 }
 175
 176 static void fq_gc(struct fq_sched_data *q,
 177                   struct rb_root *root,
 178                   struct sock *sk)
 179 {
 180         struct fq_flow *f, *tofree[FQ_GC_MAX];
 181         struct rb_node **p, *parent;
 182         int fcnt = 0;
 183
 184         p = &root->rb_node;
 185         parent = NULL;
 186         while (*p) {
 187                 parent = *p;
 188
 189                 f = container_of(parent, struct fq_flow, fq_node);
 190                 if (f->sk == sk)
 191                         break;
 192
 193                 if (fq_gc_candidate(f)) {
 194                         tofree[fcnt++] = f;
 195                         if (fcnt == FQ_GC_MAX)
 196                                 break;
 197                 }
 198
 199                 if (f->sk > sk)
 200                         p = &parent->rb_right;
 201                 else
 202                         p = &parent->rb_left;
 203         }
 204
 205         q->flows -= fcnt;
 206         q->inactive_flows -= fcnt;
 207         q->stat_gc_flows += fcnt;
 208         while (fcnt) {
 209                 struct fq_flow *f = tofree[--fcnt];
 210
 211                 rb_erase(&f->fq_node, root);
 212                 kmem_cache_free(fq_flow_cachep, f);
 213         }
 214 }
 215
 216 static struct fq_flow *fq_classify(struct sk_buff *skb, struct fq_sched_data *q)
 217 {
 218         struct rb_node **p, *parent;
 219         struct sock *sk = skb->sk;
 220         struct rb_root *root;
 221         struct fq_flow *f;
 222
 223         /* warning: no starvation prevention... */
 224         if (unlikely((skb->priority & TC_PRIO_MAX) == TC_PRIO_CONTROL))
 225                 return &q->internal;
 226
 227         /* SYNACK messages are attached to a listener socket.
 228          * 1) They are not part of a 'flow' yet
 229          * 2) We do not want to rate limit them (eg SYNFLOOD attack),
 230          *    especially if the listener set SO_MAX_PACING_RATE
 231          * 3) We pretend they are orphaned
 232          */
 233         if (!sk || sk->sk_state == TCP_LISTEN) {
 234                 unsigned long hash = skb_get_hash(skb) & q->orphan_mask;
 235
 236                 /* By forcing low order bit to 1, we make sure to not
 237                  * collide with a local flow (socket pointers are word aligned)
 238                  */
 239                 sk = (struct sock *)((hash << 1) | 1UL);
 240                 skb_orphan(skb);
 241         }
 242
 243         root = &q->fq_root[hash_32((u32)(long)sk, q->fq_trees_log)];
 244
 245         if (q->flows >= (2U << q->fq_trees_log) &&
 246             q->inactive_flows > q->flows/2)
 247                 fq_gc(q, root, sk);
 248
 249         p = &root->rb_node;
 250         parent = NULL;
 251         while (*p) {
 252                 parent = *p;
 253
 254                 f = container_of(parent, struct fq_flow, fq_node);
 255                 if (f->sk == sk) {
 256                         /* socket might have been reallocated, so check
 257                          * if its sk_hash is the same.
 258                          * It not, we need to refill credit with
 259                          * initial quantum
 260                          */
 261                         if (unlikely(skb->sk &&
 262                                      f->socket_hash != sk->sk_hash)) {
 263                                 f->credit = q->initial_quantum;
 264                                 f->socket_hash = sk->sk_hash;
 265                                 f->time_next_packet = 0ULL;
 266                         }
 267                         return f;
 268                 }
 269                 if (f->sk > sk)
 270                         p = &parent->rb_right;
 271                 else
 272                         p = &parent->rb_left;
 273         }
 274
 275         f = kmem_cache_zalloc(fq_flow_cachep, GFP_ATOMIC | __GFP_NOWARN);
 276         if (unlikely(!f)) {
 277                 q->stat_allocation_errors++;
 278                 return &q->internal;
 279         }
 280         fq_flow_set_detached(f);
 281         f->sk = sk;
 282         if (skb->sk)
 283                 f->socket_hash = sk->sk_hash;
 284         f->credit = q->initial_quantum;
 285
 286         rb_link_node(&f->fq_node, parent, p);
 287         rb_insert_color(&f->fq_node, root);
 288
 289         q->flows++;
 290         q->inactive_flows++;
 291         return f;
 292 }
 293
 294
 295 /* remove one skb from head of flow queue */
 296 static struct sk_buff *fq_dequeue_head(struct Qdisc *sch, struct fq_flow *flow)
 297 {
 298         struct sk_buff *skb = flow->head;
 299
 300         if (skb) {
 301                 flow->head = skb->next;
 302                 skb->next = NULL;
 303                 flow->qlen--;
 304                 qdisc_qstats_backlog_dec(sch, skb);
 305                 sch->q.qlen--;
 306         }
 307         return skb;
 308 }
 309
 310 /* We might add in the future detection of retransmits
 311  * For the time being, just return false
 312  */
 313 static bool skb_is_retransmit(struct sk_buff *skb)
 314 {
 315         return false;
 316 }
 317
 318 /* add skb to flow queue
 319  * flow queue is a linked list, kind of FIFO, except for TCP retransmits
 320  * We special case tcp retransmits to be transmitted before other packets.
 321  * We rely on fact that TCP retransmits are unlikely, so we do not waste
 322  * a separate queue or a pointer.
 323  * head->  [retrans pkt 1]
 324  *         [retrans pkt 2]
 325  *         [ normal pkt 1]
 326  *         [ normal pkt 2]
 327  *         [ normal pkt 3]
 328  * tail->  [ normal pkt 4]
 329  */
 330 static void flow_queue_add(struct fq_flow *flow, struct sk_buff *skb)
 331 {
 332         struct sk_buff *prev, *head = flow->head;
 333
 334         skb->next = NULL;
 335         if (!head) {
 336                 flow->head = skb;
 337                 flow->tail = skb;
 338                 return;
 339         }
 340         if (likely(!skb_is_retransmit(skb))) {
 341                 flow->tail->next = skb;
 342                 flow->tail = skb;
 343                 return;
 344         }
 345
 346         /* This skb is a tcp retransmit,
 347          * find the last retrans packet in the queue
 348          */
 349         prev = NULL;
 350         while (skb_is_retransmit(head)) {
 351                 prev = head;
 352                 head = head->next;
 353                 if (!head)
 354                         break;
 355         }
 356         if (!prev) { /* no rtx packet in queue, become the new head */
 357                 skb->next = flow->head;
 358                 flow->head = skb;
 359         } else {
 360                 if (prev == flow->tail)
 361                         flow->tail = skb;
 362                 else
 363                         skb->next = prev->next;
 364                 prev->next = skb;
 365         }
 366 }
 367
 368 static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch)
 369 {
 370         struct fq_sched_data *q = qdisc_priv(sch);
 371         struct fq_flow *f;
 372
 373         if (unlikely(sch->q.qlen >= sch->limit))
 374                 return qdisc_drop(skb, sch);
 375
 376         f = fq_classify(skb, q);
 377         if (unlikely(f->qlen >= q->flow_plimit && f != &q->internal)) {
 378                 q->stat_flows_plimit++;
 379                 return qdisc_drop(skb, sch);
 380         }
 381
 382         f->qlen++;
 383         if (skb_is_retransmit(skb))
 384                 q->stat_tcp_retrans++;
 385         qdisc_qstats_backlog_inc(sch, skb);
 386         if (fq_flow_is_detached(f)) {
 387                 fq_flow_add_tail(&q->new_flows, f);
 388                 if (time_after(jiffies, f->age + q->flow_refill_delay))
 389                         f->credit = max_t(u32, f->credit, q->quantum);
 390                 q->inactive_flows--;
 391         }
 392
 393         /* Note: this overwrites f->age */
 394         flow_queue_add(f, skb);
 395
 396         if (unlikely(f == &q->internal)) {
 397                 q->stat_internal_packets++;
 398         }
 399         sch->q.qlen++;
 400
 401         return NET_XMIT_SUCCESS;
 402 }
 403
 404 static void fq_check_throttled(struct fq_sched_data *q, u64 now)
 405 {
 406         struct rb_node *p;
 407
 408         if (q->time_next_delayed_flow > now)
 409                 return;
 410
 411         q->time_next_delayed_flow = ~0ULL;
 412         while ((p = rb_first(&q->delayed)) != NULL) {
 413                 struct fq_flow *f = container_of(p, struct fq_flow, rate_node);
 414
 415                 if (f->time_next_packet > now) {
 416                         q->time_next_delayed_flow = f->time_next_packet;
 417                         break;
 418                 }
 419                 rb_erase(p, &q->delayed);
 420                 q->throttled_flows--;
 421                 fq_flow_add_tail(&q->old_flows, f);
 422         }
 423 }
 424
 425 static struct sk_buff *fq_dequeue(struct Qdisc *sch)
 426 {
 427         struct fq_sched_data *q = qdisc_priv(sch);
 428         u64 now = ktime_get_ns();
 429         struct fq_flow_head *head;
 430         struct sk_buff *skb;
 431         struct fq_flow *f;
 432         u32 rate;
 433
 434         skb = fq_dequeue_head(sch, &q->internal);
 435         if (skb)
 436                 goto out;
 437         fq_check_throttled(q, now);
 438 begin:
 439         head = &q->new_flows;
 440         if (!head->first) {
 441                 head = &q->old_flows;
 442                 if (!head->first) {
 443                         if (q->time_next_delayed_flow != ~0ULL)
 444                                 qdisc_watchdog_schedule_ns(&q->watchdog,
 445                                                            q->time_next_delayed_flow,
 446                                                            false);
 447                         return NULL;
 448                 }
 449         }
 450         f = head->first;
 451
 452         if (f->credit <= 0) {
 453                 f->credit += q->quantum;
 454                 head->first = f->next;
 455                 fq_flow_add_tail(&q->old_flows, f);
 456                 goto begin;
 457         }
 458
 459         skb = f->head;
 460         if (unlikely(skb && now < f->time_next_packet &&
 461                      !skb_is_tcp_pure_ack(skb))) {
 462                 head->first = f->next;
 463                 fq_flow_set_throttled(q, f);
 464                 goto begin;
 465         }
 466
 467         skb = fq_dequeue_head(sch, f);
 468         if (!skb) {
 469                 head->first = f->next;
 470                 /* force a pass through old_flows to prevent starvation */
 471                 if ((head == &q->new_flows) && q->old_flows.first) {
 472                         fq_flow_add_tail(&q->old_flows, f);
 473                 } else {
 474                         fq_flow_set_detached(f);
 475                         q->inactive_flows++;
 476                 }
 477                 goto begin;
 478         }
 479         prefetch(&skb->end);
 480         f->credit -= qdisc_pkt_len(skb);
 481
 482         if (f->credit > 0 || !q->rate_enable)
 483                 goto out;
 484
 485         /* Do not pace locally generated ack packets */
 486         if (skb_is_tcp_pure_ack(skb))
 487                 goto out;
 488
 489         rate = q->flow_max_rate;
 490         if (skb->sk)
 491                 rate = min(skb->sk->sk_pacing_rate, rate);
 492
 493         if (rate != ~0U) {
 494                 u32 plen = max(qdisc_pkt_len(skb), q->quantum);
 495                 u64 len = (u64)plen * NSEC_PER_SEC;
 496
 497                 if (likely(rate))
 498                         do_div(len, rate);
 499                 /* Since socket rate can change later,
 500                  * clamp the delay to 1 second.
 501                  * Really, providers of too big packets should be fixed !
 502                  */
 503                 if (unlikely(len > NSEC_PER_SEC)) {
 504                         len = NSEC_PER_SEC;
 505                         q->stat_pkts_too_long++;
 506                 }
 507
 508                 f->time_next_packet = now + len;
 509         }
 510 out:
 511         qdisc_bstats_update(sch, skb);
 512         return skb;
 513 }
 514
 515 static void fq_reset(struct Qdisc *sch)
 516 {
 517         struct fq_sched_data *q = qdisc_priv(sch);
 518         struct rb_root *root;
 519         struct sk_buff *skb;
 520         struct rb_node *p;
 521         struct fq_flow *f;
 522         unsigned int idx;
 523
 524         while ((skb = fq_dequeue_head(sch, &q->internal)) != NULL)
 525                 kfree_skb(skb);
 526
 527         if (!q->fq_root)
 528                 return;
 529
 530         for (idx = 0; idx < (1U << q->fq_trees_log); idx++) {
 531                 root = &q->fq_root[idx];
 532                 while ((p = rb_first(root)) != NULL) {
 533                         f = container_of(p, struct fq_flow, fq_node);
 534                         rb_erase(p, root);
 535
 536                         while ((skb = fq_dequeue_head(sch, f)) != NULL)
 537                                 kfree_skb(skb);
 538
 539                         kmem_cache_free(fq_flow_cachep, f);
 540                 }
 541         }
 542         q->new_flows.first      = NULL;
 543         q->old_flows.first      = NULL;
 544         q->delayed              = RB_ROOT;
 545         q->flows                = 0;
 546         q->inactive_flows       = 0;
 547         q->throttled_flows      = 0;
 548 }
 549
 550 static void fq_rehash(struct fq_sched_data *q,
 551                       struct rb_root *old_array, u32 old_log,
 552                       struct rb_root *new_array, u32 new_log)
 553 {
 554         struct rb_node *op, **np, *parent;
 555         struct rb_root *oroot, *nroot;
 556         struct fq_flow *of, *nf;
 557         int fcnt = 0;
 558         u32 idx;
 559
 560         for (idx = 0; idx < (1U << old_log); idx++) {
 561                 oroot = &old_array[idx];
 562                 while ((op = rb_first(oroot)) != NULL) {
 563                         rb_erase(op, oroot);
 564                         of = container_of(op, struct fq_flow, fq_node);
 565                         if (fq_gc_candidate(of)) {
 566                                 fcnt++;
 567                                 kmem_cache_free(fq_flow_cachep, of);
 568                                 continue;
 569                         }
 570                         nroot = &new_array[hash_32((u32)(long)of->sk, new_log)];
 571
 572                         np = &nroot->rb_node;
 573                         parent = NULL;
 574                         while (*np) {
 575                                 parent = *np;
 576
 577                                 nf = container_of(parent, struct fq_flow, fq_node);
 578                                 BUG_ON(nf->sk == of->sk);
 579
 580                                 if (nf->sk > of->sk)
 581                                         np = &parent->rb_right;
 582                                 else
 583                                         np = &parent->rb_left;
 584                         }
 585
 586                         rb_link_node(&of->fq_node, parent, np);
 587                         rb_insert_color(&of->fq_node, nroot);
 588                 }
 589         }
 590         q->flows -= fcnt;
 591         q->inactive_flows -= fcnt;
 592         q->stat_gc_flows += fcnt;
 593 }
 594
 595 static void *fq_alloc_node(size_t sz, int node)
 596 {
 597         void *ptr;
 598
 599         ptr = kmalloc_node(sz, GFP_KERNEL | __GFP_REPEAT | __GFP_NOWARN, node);
 600         if (!ptr)
 601                 ptr = vmalloc_node(sz, node);
 602         return ptr;
 603 }
 604
 605 static void fq_free(void *addr)
 606 {
 607         kvfree(addr);
 608 }
 609
 610 static int fq_resize(struct Qdisc *sch, u32 log)
 611 {
 612         struct fq_sched_data *q = qdisc_priv(sch);
 613         struct rb_root *array;
 614         void *old_fq_root;
 615         u32 idx;
 616
 617         if (q->fq_root && log == q->fq_trees_log)
 618                 return 0;
 619
 620         /* If XPS was setup, we can allocate memory on right NUMA node */
 621         array = fq_alloc_node(sizeof(struct rb_root) << log,
 622                               netdev_queue_numa_node_read(sch->dev_queue));
 623         if (!array)
 624                 return -ENOMEM;
 625
 626         for (idx = 0; idx < (1U << log); idx++)
 627                 array[idx] = RB_ROOT;
 628
 629         sch_tree_lock(sch);
 630
 631         old_fq_root = q->fq_root;
 632         if (old_fq_root)
 633                 fq_rehash(q, old_fq_root, q->fq_trees_log, array, log);
 634
 635         q->fq_root = array;
 636         q->fq_trees_log = log;
 637
 638         sch_tree_unlock(sch);
 639
 640         fq_free(old_fq_root);
 641
 642         return 0;
 643 }
 644
 645 static const struct nla_policy fq_policy[TCA_FQ_MAX + 1] = {
 646         [TCA_FQ_PLIMIT]                 = { .type = NLA_U32 },
 647         [TCA_FQ_FLOW_PLIMIT]            = { .type = NLA_U32 },
 648         [TCA_FQ_QUANTUM]                = { .type = NLA_U32 },
 649         [TCA_FQ_INITIAL_QUANTUM]        = { .type = NLA_U32 },
 650         [TCA_FQ_RATE_ENABLE]            = { .type = NLA_U32 },
 651         [TCA_FQ_FLOW_DEFAULT_RATE]      = { .type = NLA_U32 },
 652         [TCA_FQ_FLOW_MAX_RATE]          = { .type = NLA_U32 },
 653         [TCA_FQ_BUCKETS_LOG]            = { .type = NLA_U32 },
 654         [TCA_FQ_FLOW_REFILL_DELAY]      = { .type = NLA_U32 },
 655 };
 656
 657 static int fq_change(struct Qdisc *sch, struct nlattr *opt)
 658 {
 659         struct fq_sched_data *q = qdisc_priv(sch);
 660         struct nlattr *tb[TCA_FQ_MAX + 1];
 661         int err, drop_count = 0;
 662         u32 fq_log;
 663
 664         if (!opt)
 665                 return -EINVAL;
 666
 667         err = nla_parse_nested(tb, TCA_FQ_MAX, opt, fq_policy);
 668         if (err < 0)
 669                 return err;
 670
 671         sch_tree_lock(sch);
 672
 673         fq_log = q->fq_trees_log;
 674
 675         if (tb[TCA_FQ_BUCKETS_LOG]) {
 676                 u32 nval = nla_get_u32(tb[TCA_FQ_BUCKETS_LOG]);
 677
 678                 if (nval >= 1 && nval <= ilog2(256*1024))
 679                         fq_log = nval;
 680                 else
 681                         err = -EINVAL;
 682         }
 683         if (tb[TCA_FQ_PLIMIT])
 684                 sch->limit = nla_get_u32(tb[TCA_FQ_PLIMIT]);
 685
 686         if (tb[TCA_FQ_FLOW_PLIMIT])
 687                 q->flow_plimit = nla_get_u32(tb[TCA_FQ_FLOW_PLIMIT]);
 688
 689         if (tb[TCA_FQ_QUANTUM]) {
 690                 u32 quantum = nla_get_u32(tb[TCA_FQ_QUANTUM]);
 691
 692                 if (quantum > 0)
 693                         q->quantum = quantum;
 694                 else
 695                         err = -EINVAL;
 696         }
 697
 698         if (tb[TCA_FQ_INITIAL_QUANTUM])
 699                 q->initial_quantum = nla_get_u32(tb[TCA_FQ_INITIAL_QUANTUM]);
 700
 701         if (tb[TCA_FQ_FLOW_DEFAULT_RATE])
 702                 pr_warn_ratelimited("sch_fq: defrate %u ignored.\n",
 703                                     nla_get_u32(tb[TCA_FQ_FLOW_DEFAULT_RATE]));
 704
 705         if (tb[TCA_FQ_FLOW_MAX_RATE])
 706                 q->flow_max_rate = nla_get_u32(tb[TCA_FQ_FLOW_MAX_RATE]);
 707
 708         if (tb[TCA_FQ_RATE_ENABLE]) {
 709                 u32 enable = nla_get_u32(tb[TCA_FQ_RATE_ENABLE]);
 710
 711                 if (enable <= 1)
 712                         q->rate_enable = enable;
 713                 else
 714                         err = -EINVAL;
 715         }
 716
 717         if (tb[TCA_FQ_FLOW_REFILL_DELAY]) {
 718                 u32 usecs_delay = nla_get_u32(tb[TCA_FQ_FLOW_REFILL_DELAY]) ;
 719
 720                 q->flow_refill_delay = usecs_to_jiffies(usecs_delay);
 721         }
 722
 723         if (tb[TCA_FQ_ORPHAN_MASK])
 724                 q->orphan_mask = nla_get_u32(tb[TCA_FQ_ORPHAN_MASK]);
 725
 726         if (!err) {
 727                 sch_tree_unlock(sch);
 728                 err = fq_resize(sch, fq_log);
 729                 sch_tree_lock(sch);
 730         }
 731         while (sch->q.qlen > sch->limit) {
 732                 struct sk_buff *skb = fq_dequeue(sch);
 733
 734                 if (!skb)
 735                         break;
 736                 kfree_skb(skb);
 737                 drop_count++;
 738         }
 739         qdisc_tree_decrease_qlen(sch, drop_count);
 740
 741         sch_tree_unlock(sch);
 742         return err;
 743 }
 744
 745 static void fq_destroy(struct Qdisc *sch)
 746 {
 747         struct fq_sched_data *q = qdisc_priv(sch);
 748
 749         fq_reset(sch);
 750         fq_free(q->fq_root);
 751         qdisc_watchdog_cancel(&q->watchdog);
 752 }
 753
 754 static int fq_init(struct Qdisc *sch, struct nlattr *opt)
 755 {
 756         struct fq_sched_data *q = qdisc_priv(sch);
 757         int err;
 758
 759         sch->limit              = 10000;
 760         q->flow_plimit          = 100;
 761         q->quantum              = 2 * psched_mtu(qdisc_dev(sch));
 762         q->initial_quantum      = 10 * psched_mtu(qdisc_dev(sch));
 763         q->flow_refill_delay    = msecs_to_jiffies(40);
 764         q->flow_max_rate        = ~0U;
 765         q->rate_enable          = 1;
 766         q->new_flows.first      = NULL;
 767         q->old_flows.first      = NULL;
 768         q->delayed              = RB_ROOT;
 769         q->fq_root              = NULL;
 770         q->fq_trees_log         = ilog2(1024);
 771         q->orphan_mask          = 1024 - 1;
 772         qdisc_watchdog_init(&q->watchdog, sch);
 773
 774         if (opt)
 775                 err = fq_change(sch, opt);
 776         else
 777                 err = fq_resize(sch, q->fq_trees_log);
 778
 779         return err;
 780 }
 781
 782 static int fq_dump(struct Qdisc *sch, struct sk_buff *skb)
 783 {
 784         struct fq_sched_data *q = qdisc_priv(sch);
 785         struct nlattr *opts;
 786
 787         opts = nla_nest_start(skb, TCA_OPTIONS);
 788         if (opts == NULL)
 789                 goto nla_put_failure;
 790
 791         /* TCA_FQ_FLOW_DEFAULT_RATE is not used anymore */
 792
 793         if (nla_put_u32(skb, TCA_FQ_PLIMIT, sch->limit) ||
 794             nla_put_u32(skb, TCA_FQ_FLOW_PLIMIT, q->flow_plimit) ||
 795             nla_put_u32(skb, TCA_FQ_QUANTUM, q->quantum) ||
 796             nla_put_u32(skb, TCA_FQ_INITIAL_QUANTUM, q->initial_quantum) ||
 797             nla_put_u32(skb, TCA_FQ_RATE_ENABLE, q->rate_enable) ||
 798             nla_put_u32(skb, TCA_FQ_FLOW_MAX_RATE, q->flow_max_rate) ||
 799             nla_put_u32(skb, TCA_FQ_FLOW_REFILL_DELAY,
 800                         jiffies_to_usecs(q->flow_refill_delay)) ||
 801             nla_put_u32(skb, TCA_FQ_ORPHAN_MASK, q->orphan_mask) ||
 802             nla_put_u32(skb, TCA_FQ_BUCKETS_LOG, q->fq_trees_log))
 803                 goto nla_put_failure;
 804
 805         return nla_nest_end(skb, opts);
 806
 807 nla_put_failure:
 808         return -1;
 809 }
 810
 811 static int fq_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 812 {
 813         struct fq_sched_data *q = qdisc_priv(sch);
 814         u64 now = ktime_get_ns();
 815         struct tc_fq_qd_stats st = {
 816                 .gc_flows               = q->stat_gc_flows,
 817                 .highprio_packets       = q->stat_internal_packets,
 818                 .tcp_retrans            = q->stat_tcp_retrans,
 819                 .throttled              = q->stat_throttled,
 820                 .flows_plimit           = q->stat_flows_plimit,
 821                 .pkts_too_long          = q->stat_pkts_too_long,
 822                 .allocation_errors      = q->stat_allocation_errors,
 823                 .flows                  = q->flows,
 824                 .inactive_flows         = q->inactive_flows,
 825                 .throttled_flows        = q->throttled_flows,
 826                 .time_next_delayed_flow = q->time_next_delayed_flow - now,
 827         };
 828
 829         return gnet_stats_copy_app(d, &st, sizeof(st));
 830 }
 831
 832 static struct Qdisc_ops fq_qdisc_ops __read_mostly = {
 833         .id             =       "fq",
 834         .priv_size      =       sizeof(struct fq_sched_data),
 835
 836         .enqueue        =       fq_enqueue,
 837         .dequeue        =       fq_dequeue,
 838         .peek           =       qdisc_peek_dequeued,
 839         .init           =       fq_init,
 840         .reset          =       fq_reset,
 841         .destroy        =       fq_destroy,
 842         .change         =       fq_change,
 843         .dump           =       fq_dump,
 844         .dump_stats     =       fq_dump_stats,
 845         .owner          =       THIS_MODULE,
 846 };
 847
 848 static int __init fq_module_init(void)
 849 {
 850         int ret;
 851
 852         fq_flow_cachep = kmem_cache_create("fq_flow_cache",
 853                                            sizeof(struct fq_flow),
 854                                            0, 0, NULL);
 855         if (!fq_flow_cachep)
 856                 return -ENOMEM;
 857
 858         ret = register_qdisc(&fq_qdisc_ops);
 859         if (ret)
 860                 kmem_cache_destroy(fq_flow_cachep);
 861         return ret;
 862 }
 863
 864 static void __exit fq_module_exit(void)
 865 {
 866         unregister_qdisc(&fq_qdisc_ops);
 867         kmem_cache_destroy(fq_flow_cachep);
 868 }
 869
 870 module_init(fq_module_init)
 871 module_exit(fq_module_exit)
 872 MODULE_AUTHOR("Eric Dumazet");
 873 MODULE_LICENSE("GPL");