1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * net/sched/sch_fq.c Fair Queue Packet Scheduler (per flow pacing)
5 * Copyright (C) 2013-2015 Eric Dumazet <edumazet@google.com>
7 * Meant to be mostly used for locally generated traffic :
8 * Fast classification depends on skb->sk being set before reaching us.
9 * If not, (router workload), we use rxhash as fallback, with 32 bits wide hash.
10 * All packets belonging to a socket are considered as a 'flow'.
12 * Flows are dynamically allocated and stored in a hash table of RB trees
13 * They are also part of one Round Robin 'queues' (new or old flows)
15 * Burst avoidance (aka pacing) capability :
17 * Transport (eg TCP) can set in sk->sk_pacing_rate a rate, enqueue a
18 * bunch of packets, and this packet scheduler adds delay between
19 * packets to respect rate limitation.
22 * - lookup one RB tree (out of 1024 or more) to find the flow.
23 * If non existent flow, create it, add it to the tree.
24 * Add skb to the per flow list of skb (fifo).
25 * - Use a special fifo for high prio packets
27 * dequeue() : serves flows in Round Robin
28 * Note : When a flow becomes empty, we do not immediately remove it from
29 * rb trees, for performance reasons (its expected to send additional packets,
30 * or SLAB cache will reuse socket for another flow)
33 #include <linux/module.h>
34 #include <linux/types.h>
35 #include <linux/kernel.h>
36 #include <linux/jiffies.h>
37 #include <linux/string.h>
39 #include <linux/errno.h>
40 #include <linux/init.h>
41 #include <linux/skbuff.h>
42 #include <linux/slab.h>
43 #include <linux/rbtree.h>
44 #include <linux/hash.h>
45 #include <linux/prefetch.h>
46 #include <linux/vmalloc.h>
47 #include <net/netlink.h>
48 #include <net/pkt_sched.h>
50 #include <net/tcp_states.h>
57 static inline struct fq_skb_cb *fq_skb_cb(struct sk_buff *skb)
59 qdisc_cb_private_validate(skb, sizeof(struct fq_skb_cb));
60 return (struct fq_skb_cb *)qdisc_skb_cb(skb)->data;
64 * Per flow structure, dynamically allocated.
65 * If packets have monotically increasing time_to_send, they are placed in O(1)
66 * in linear list (head,tail), otherwise are placed in a rbtree (t_root).
69 /* First cache line : used in fq_gc(), fq_enqueue(), fq_dequeue() */
70 struct rb_root t_root;
71 struct sk_buff *head; /* list of skbs for this flow : first skb */
73 struct sk_buff *tail; /* last skb in the list */
74 unsigned long age; /* (jiffies | 1UL) when flow was emptied, for gc */
76 struct rb_node fq_node; /* anchor in fq_root[] trees */
78 u32 socket_hash; /* sk_hash */
79 int qlen; /* number of packets in flow queue */
81 /* Second cache line, used in fq_dequeue() */
83 /* 32bit hole on 64bit arches */
85 struct fq_flow *next; /* next pointer in RR lists */
87 struct rb_node rate_node; /* anchor in q->delayed tree */
89 } ____cacheline_aligned_in_smp;
92 struct fq_flow *first;
96 struct fq_sched_data {
97 struct fq_flow_head new_flows;
99 struct fq_flow_head old_flows;
101 struct rb_root delayed; /* for rate limited flows */
102 u64 time_next_delayed_flow;
103 u64 ktime_cache; /* copy of last ktime_get_ns() */
104 unsigned long unthrottle_latency_ns;
106 struct fq_flow internal; /* for non classified or high prio packets */
109 u32 flow_refill_delay;
110 u32 flow_plimit; /* max packets per flow */
111 unsigned long flow_max_rate; /* optional max rate per flow */
113 u64 horizon; /* horizon in ns */
114 u32 orphan_mask; /* mask for orphaned skb */
115 u32 low_rate_threshold;
116 struct rb_root *fq_root;
125 u64 stat_internal_packets;
128 u64 stat_horizon_drops;
129 u64 stat_horizon_caps;
130 u64 stat_flows_plimit;
131 u64 stat_pkts_too_long;
132 u64 stat_allocation_errors;
134 u32 timer_slack; /* hrtimer slack in ns */
135 struct qdisc_watchdog watchdog;
139 * f->tail and f->age share the same location.
140 * We can use the low order bit to differentiate if this location points
141 * to a sk_buff or contains a jiffies value, if we force this value to be odd.
142 * This assumes f->tail low order bit must be 0 since alignof(struct sk_buff) >= 2
144 static void fq_flow_set_detached(struct fq_flow *f)
146 f->age = jiffies | 1UL;
149 static bool fq_flow_is_detached(const struct fq_flow *f)
151 return !!(f->age & 1UL);
154 /* special value to mark a throttled flow (not on old/new list) */
155 static struct fq_flow throttled;
157 static bool fq_flow_is_throttled(const struct fq_flow *f)
159 return f->next == &throttled;
162 static void fq_flow_add_tail(struct fq_flow_head *head, struct fq_flow *flow)
165 head->last->next = flow;
172 static void fq_flow_unset_throttled(struct fq_sched_data *q, struct fq_flow *f)
174 rb_erase(&f->rate_node, &q->delayed);
175 q->throttled_flows--;
176 fq_flow_add_tail(&q->old_flows, f);
179 static void fq_flow_set_throttled(struct fq_sched_data *q, struct fq_flow *f)
181 struct rb_node **p = &q->delayed.rb_node, *parent = NULL;
187 aux = rb_entry(parent, struct fq_flow, rate_node);
188 if (f->time_next_packet >= aux->time_next_packet)
189 p = &parent->rb_right;
191 p = &parent->rb_left;
193 rb_link_node(&f->rate_node, parent, p);
194 rb_insert_color(&f->rate_node, &q->delayed);
195 q->throttled_flows++;
198 f->next = &throttled;
199 if (q->time_next_delayed_flow > f->time_next_packet)
200 q->time_next_delayed_flow = f->time_next_packet;
204 static struct kmem_cache *fq_flow_cachep __read_mostly;
207 /* limit number of collected flows per round */
209 #define FQ_GC_AGE (3*HZ)
211 static bool fq_gc_candidate(const struct fq_flow *f)
213 return fq_flow_is_detached(f) &&
214 time_after(jiffies, f->age + FQ_GC_AGE);
217 static void fq_gc(struct fq_sched_data *q,
218 struct rb_root *root,
221 struct rb_node **p, *parent;
222 void *tofree[FQ_GC_MAX];
231 f = rb_entry(parent, struct fq_flow, fq_node);
235 if (fq_gc_candidate(f)) {
237 if (fcnt == FQ_GC_MAX)
242 p = &parent->rb_right;
244 p = &parent->rb_left;
250 for (i = fcnt; i > 0; ) {
252 rb_erase(&f->fq_node, root);
255 q->inactive_flows -= fcnt;
256 q->stat_gc_flows += fcnt;
258 kmem_cache_free_bulk(fq_flow_cachep, fcnt, tofree);
261 static struct fq_flow *fq_classify(struct sk_buff *skb, struct fq_sched_data *q)
263 struct rb_node **p, *parent;
264 struct sock *sk = skb->sk;
265 struct rb_root *root;
268 /* warning: no starvation prevention... */
269 if (unlikely((skb->priority & TC_PRIO_MAX) == TC_PRIO_CONTROL))
272 /* SYNACK messages are attached to a TCP_NEW_SYN_RECV request socket
273 * or a listener (SYNCOOKIE mode)
274 * 1) request sockets are not full blown,
275 * they do not contain sk_pacing_rate
276 * 2) They are not part of a 'flow' yet
277 * 3) We do not want to rate limit them (eg SYNFLOOD attack),
278 * especially if the listener set SO_MAX_PACING_RATE
279 * 4) We pretend they are orphaned
281 if (!sk || sk_listener(sk)) {
282 unsigned long hash = skb_get_hash(skb) & q->orphan_mask;
284 /* By forcing low order bit to 1, we make sure to not
285 * collide with a local flow (socket pointers are word aligned)
287 sk = (struct sock *)((hash << 1) | 1UL);
289 } else if (sk->sk_state == TCP_CLOSE) {
290 unsigned long hash = skb_get_hash(skb) & q->orphan_mask;
292 * Sockets in TCP_CLOSE are non connected.
293 * Typical use case is UDP sockets, they can send packets
294 * with sendto() to many different destinations.
295 * We probably could use a generic bit advertising
296 * non connected sockets, instead of sk_state == TCP_CLOSE,
299 sk = (struct sock *)((hash << 1) | 1UL);
302 root = &q->fq_root[hash_ptr(sk, q->fq_trees_log)];
304 if (q->flows >= (2U << q->fq_trees_log) &&
305 q->inactive_flows > q->flows/2)
313 f = rb_entry(parent, struct fq_flow, fq_node);
315 /* socket might have been reallocated, so check
316 * if its sk_hash is the same.
317 * It not, we need to refill credit with
320 if (unlikely(skb->sk == sk &&
321 f->socket_hash != sk->sk_hash)) {
322 f->credit = q->initial_quantum;
323 f->socket_hash = sk->sk_hash;
325 smp_store_release(&sk->sk_pacing_status,
327 if (fq_flow_is_throttled(f))
328 fq_flow_unset_throttled(q, f);
329 f->time_next_packet = 0ULL;
334 p = &parent->rb_right;
336 p = &parent->rb_left;
339 f = kmem_cache_zalloc(fq_flow_cachep, GFP_ATOMIC | __GFP_NOWARN);
341 q->stat_allocation_errors++;
344 /* f->t_root is already zeroed after kmem_cache_zalloc() */
346 fq_flow_set_detached(f);
349 f->socket_hash = sk->sk_hash;
351 smp_store_release(&sk->sk_pacing_status,
354 f->credit = q->initial_quantum;
356 rb_link_node(&f->fq_node, parent, p);
357 rb_insert_color(&f->fq_node, root);
364 static struct sk_buff *fq_peek(struct fq_flow *flow)
366 struct sk_buff *skb = skb_rb_first(&flow->t_root);
367 struct sk_buff *head = flow->head;
375 if (fq_skb_cb(skb)->time_to_send < fq_skb_cb(head)->time_to_send)
380 static void fq_erase_head(struct Qdisc *sch, struct fq_flow *flow,
383 if (skb == flow->head) {
384 flow->head = skb->next;
386 rb_erase(&skb->rbnode, &flow->t_root);
387 skb->dev = qdisc_dev(sch);
391 /* Remove one skb from flow queue.
392 * This skb must be the return value of prior fq_peek().
394 static void fq_dequeue_skb(struct Qdisc *sch, struct fq_flow *flow,
397 fq_erase_head(sch, flow, skb);
398 skb_mark_not_on_list(skb);
400 qdisc_qstats_backlog_dec(sch, skb);
404 static void flow_queue_add(struct fq_flow *flow, struct sk_buff *skb)
406 struct rb_node **p, *parent;
407 struct sk_buff *head, *aux;
411 fq_skb_cb(skb)->time_to_send >= fq_skb_cb(flow->tail)->time_to_send) {
415 flow->tail->next = skb;
421 p = &flow->t_root.rb_node;
426 aux = rb_to_skb(parent);
427 if (fq_skb_cb(skb)->time_to_send >= fq_skb_cb(aux)->time_to_send)
428 p = &parent->rb_right;
430 p = &parent->rb_left;
432 rb_link_node(&skb->rbnode, parent, p);
433 rb_insert_color(&skb->rbnode, &flow->t_root);
436 static bool fq_packet_beyond_horizon(const struct sk_buff *skb,
437 const struct fq_sched_data *q)
439 return unlikely((s64)skb->tstamp > (s64)(q->ktime_cache + q->horizon));
442 static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
443 struct sk_buff **to_free)
445 struct fq_sched_data *q = qdisc_priv(sch);
448 if (unlikely(sch->q.qlen >= sch->limit))
449 return qdisc_drop(skb, sch, to_free);
452 fq_skb_cb(skb)->time_to_send = q->ktime_cache = ktime_get_ns();
454 /* Check if packet timestamp is too far in the future.
455 * Try first if our cached value, to avoid ktime_get_ns()
456 * cost in most cases.
458 if (fq_packet_beyond_horizon(skb, q)) {
459 /* Refresh our cache and check another time */
460 q->ktime_cache = ktime_get_ns();
461 if (fq_packet_beyond_horizon(skb, q)) {
462 if (q->horizon_drop) {
463 q->stat_horizon_drops++;
464 return qdisc_drop(skb, sch, to_free);
466 q->stat_horizon_caps++;
467 skb->tstamp = q->ktime_cache + q->horizon;
470 fq_skb_cb(skb)->time_to_send = skb->tstamp;
473 f = fq_classify(skb, q);
474 if (unlikely(f->qlen >= q->flow_plimit && f != &q->internal)) {
475 q->stat_flows_plimit++;
476 return qdisc_drop(skb, sch, to_free);
480 qdisc_qstats_backlog_inc(sch, skb);
481 if (fq_flow_is_detached(f)) {
482 fq_flow_add_tail(&q->new_flows, f);
483 if (time_after(jiffies, f->age + q->flow_refill_delay))
484 f->credit = max_t(u32, f->credit, q->quantum);
488 /* Note: this overwrites f->age */
489 flow_queue_add(f, skb);
491 if (unlikely(f == &q->internal)) {
492 q->stat_internal_packets++;
496 return NET_XMIT_SUCCESS;
499 static void fq_check_throttled(struct fq_sched_data *q, u64 now)
501 unsigned long sample;
504 if (q->time_next_delayed_flow > now)
507 /* Update unthrottle latency EWMA.
508 * This is cheap and can help diagnosing timer/latency problems.
510 sample = (unsigned long)(now - q->time_next_delayed_flow);
511 q->unthrottle_latency_ns -= q->unthrottle_latency_ns >> 3;
512 q->unthrottle_latency_ns += sample >> 3;
514 q->time_next_delayed_flow = ~0ULL;
515 while ((p = rb_first(&q->delayed)) != NULL) {
516 struct fq_flow *f = rb_entry(p, struct fq_flow, rate_node);
518 if (f->time_next_packet > now) {
519 q->time_next_delayed_flow = f->time_next_packet;
522 fq_flow_unset_throttled(q, f);
526 static struct sk_buff *fq_dequeue(struct Qdisc *sch)
528 struct fq_sched_data *q = qdisc_priv(sch);
529 struct fq_flow_head *head;
539 skb = fq_peek(&q->internal);
541 fq_dequeue_skb(sch, &q->internal, skb);
545 q->ktime_cache = now = ktime_get_ns();
546 fq_check_throttled(q, now);
548 head = &q->new_flows;
550 head = &q->old_flows;
552 if (q->time_next_delayed_flow != ~0ULL)
553 qdisc_watchdog_schedule_range_ns(&q->watchdog,
554 q->time_next_delayed_flow,
561 if (f->credit <= 0) {
562 f->credit += q->quantum;
563 head->first = f->next;
564 fq_flow_add_tail(&q->old_flows, f);
570 u64 time_next_packet = max_t(u64, fq_skb_cb(skb)->time_to_send,
571 f->time_next_packet);
573 if (now < time_next_packet) {
574 head->first = f->next;
575 f->time_next_packet = time_next_packet;
576 fq_flow_set_throttled(q, f);
580 if ((s64)(now - time_next_packet - q->ce_threshold) > 0) {
581 INET_ECN_set_ce(skb);
584 fq_dequeue_skb(sch, f, skb);
586 head->first = f->next;
587 /* force a pass through old_flows to prevent starvation */
588 if ((head == &q->new_flows) && q->old_flows.first) {
589 fq_flow_add_tail(&q->old_flows, f);
591 fq_flow_set_detached(f);
596 plen = qdisc_pkt_len(skb);
602 rate = q->flow_max_rate;
604 /* If EDT time was provided for this skb, we need to
605 * update f->time_next_packet only if this qdisc enforces
610 rate = min(skb->sk->sk_pacing_rate, rate);
612 if (rate <= q->low_rate_threshold) {
615 plen = max(plen, q->quantum);
621 u64 len = (u64)plen * NSEC_PER_SEC;
624 len = div64_ul(len, rate);
625 /* Since socket rate can change later,
626 * clamp the delay to 1 second.
627 * Really, providers of too big packets should be fixed !
629 if (unlikely(len > NSEC_PER_SEC)) {
631 q->stat_pkts_too_long++;
633 /* Account for schedule/timers drifts.
634 * f->time_next_packet was set when prior packet was sent,
635 * and current time (@now) can be too late by tens of us.
637 if (f->time_next_packet)
638 len -= min(len/2, now - f->time_next_packet);
639 f->time_next_packet = now + len;
642 qdisc_bstats_update(sch, skb);
646 static void fq_flow_purge(struct fq_flow *flow)
648 struct rb_node *p = rb_first(&flow->t_root);
651 struct sk_buff *skb = rb_to_skb(p);
654 rb_erase(&skb->rbnode, &flow->t_root);
655 rtnl_kfree_skbs(skb, skb);
657 rtnl_kfree_skbs(flow->head, flow->tail);
662 static void fq_reset(struct Qdisc *sch)
664 struct fq_sched_data *q = qdisc_priv(sch);
665 struct rb_root *root;
671 sch->qstats.backlog = 0;
673 fq_flow_purge(&q->internal);
678 for (idx = 0; idx < (1U << q->fq_trees_log); idx++) {
679 root = &q->fq_root[idx];
680 while ((p = rb_first(root)) != NULL) {
681 f = rb_entry(p, struct fq_flow, fq_node);
686 kmem_cache_free(fq_flow_cachep, f);
689 q->new_flows.first = NULL;
690 q->old_flows.first = NULL;
691 q->delayed = RB_ROOT;
693 q->inactive_flows = 0;
694 q->throttled_flows = 0;
697 static void fq_rehash(struct fq_sched_data *q,
698 struct rb_root *old_array, u32 old_log,
699 struct rb_root *new_array, u32 new_log)
701 struct rb_node *op, **np, *parent;
702 struct rb_root *oroot, *nroot;
703 struct fq_flow *of, *nf;
707 for (idx = 0; idx < (1U << old_log); idx++) {
708 oroot = &old_array[idx];
709 while ((op = rb_first(oroot)) != NULL) {
711 of = rb_entry(op, struct fq_flow, fq_node);
712 if (fq_gc_candidate(of)) {
714 kmem_cache_free(fq_flow_cachep, of);
717 nroot = &new_array[hash_ptr(of->sk, new_log)];
719 np = &nroot->rb_node;
724 nf = rb_entry(parent, struct fq_flow, fq_node);
725 BUG_ON(nf->sk == of->sk);
728 np = &parent->rb_right;
730 np = &parent->rb_left;
733 rb_link_node(&of->fq_node, parent, np);
734 rb_insert_color(&of->fq_node, nroot);
738 q->inactive_flows -= fcnt;
739 q->stat_gc_flows += fcnt;
742 static void fq_free(void *addr)
747 static int fq_resize(struct Qdisc *sch, u32 log)
749 struct fq_sched_data *q = qdisc_priv(sch);
750 struct rb_root *array;
754 if (q->fq_root && log == q->fq_trees_log)
757 /* If XPS was setup, we can allocate memory on right NUMA node */
758 array = kvmalloc_node(sizeof(struct rb_root) << log, GFP_KERNEL | __GFP_RETRY_MAYFAIL,
759 netdev_queue_numa_node_read(sch->dev_queue));
763 for (idx = 0; idx < (1U << log); idx++)
764 array[idx] = RB_ROOT;
768 old_fq_root = q->fq_root;
770 fq_rehash(q, old_fq_root, q->fq_trees_log, array, log);
773 q->fq_trees_log = log;
775 sch_tree_unlock(sch);
777 fq_free(old_fq_root);
782 static const struct nla_policy fq_policy[TCA_FQ_MAX + 1] = {
783 [TCA_FQ_UNSPEC] = { .strict_start_type = TCA_FQ_TIMER_SLACK },
785 [TCA_FQ_PLIMIT] = { .type = NLA_U32 },
786 [TCA_FQ_FLOW_PLIMIT] = { .type = NLA_U32 },
787 [TCA_FQ_QUANTUM] = { .type = NLA_U32 },
788 [TCA_FQ_INITIAL_QUANTUM] = { .type = NLA_U32 },
789 [TCA_FQ_RATE_ENABLE] = { .type = NLA_U32 },
790 [TCA_FQ_FLOW_DEFAULT_RATE] = { .type = NLA_U32 },
791 [TCA_FQ_FLOW_MAX_RATE] = { .type = NLA_U32 },
792 [TCA_FQ_BUCKETS_LOG] = { .type = NLA_U32 },
793 [TCA_FQ_FLOW_REFILL_DELAY] = { .type = NLA_U32 },
794 [TCA_FQ_ORPHAN_MASK] = { .type = NLA_U32 },
795 [TCA_FQ_LOW_RATE_THRESHOLD] = { .type = NLA_U32 },
796 [TCA_FQ_CE_THRESHOLD] = { .type = NLA_U32 },
797 [TCA_FQ_TIMER_SLACK] = { .type = NLA_U32 },
798 [TCA_FQ_HORIZON] = { .type = NLA_U32 },
799 [TCA_FQ_HORIZON_DROP] = { .type = NLA_U8 },
802 static int fq_change(struct Qdisc *sch, struct nlattr *opt,
803 struct netlink_ext_ack *extack)
805 struct fq_sched_data *q = qdisc_priv(sch);
806 struct nlattr *tb[TCA_FQ_MAX + 1];
807 int err, drop_count = 0;
808 unsigned drop_len = 0;
814 err = nla_parse_nested_deprecated(tb, TCA_FQ_MAX, opt, fq_policy,
821 fq_log = q->fq_trees_log;
823 if (tb[TCA_FQ_BUCKETS_LOG]) {
824 u32 nval = nla_get_u32(tb[TCA_FQ_BUCKETS_LOG]);
826 if (nval >= 1 && nval <= ilog2(256*1024))
831 if (tb[TCA_FQ_PLIMIT])
832 sch->limit = nla_get_u32(tb[TCA_FQ_PLIMIT]);
834 if (tb[TCA_FQ_FLOW_PLIMIT])
835 q->flow_plimit = nla_get_u32(tb[TCA_FQ_FLOW_PLIMIT]);
837 if (tb[TCA_FQ_QUANTUM]) {
838 u32 quantum = nla_get_u32(tb[TCA_FQ_QUANTUM]);
840 if (quantum > 0 && quantum <= (1 << 20)) {
841 q->quantum = quantum;
843 NL_SET_ERR_MSG_MOD(extack, "invalid quantum");
848 if (tb[TCA_FQ_INITIAL_QUANTUM])
849 q->initial_quantum = nla_get_u32(tb[TCA_FQ_INITIAL_QUANTUM]);
851 if (tb[TCA_FQ_FLOW_DEFAULT_RATE])
852 pr_warn_ratelimited("sch_fq: defrate %u ignored.\n",
853 nla_get_u32(tb[TCA_FQ_FLOW_DEFAULT_RATE]));
855 if (tb[TCA_FQ_FLOW_MAX_RATE]) {
856 u32 rate = nla_get_u32(tb[TCA_FQ_FLOW_MAX_RATE]);
858 q->flow_max_rate = (rate == ~0U) ? ~0UL : rate;
860 if (tb[TCA_FQ_LOW_RATE_THRESHOLD])
861 q->low_rate_threshold =
862 nla_get_u32(tb[TCA_FQ_LOW_RATE_THRESHOLD]);
864 if (tb[TCA_FQ_RATE_ENABLE]) {
865 u32 enable = nla_get_u32(tb[TCA_FQ_RATE_ENABLE]);
868 q->rate_enable = enable;
873 if (tb[TCA_FQ_FLOW_REFILL_DELAY]) {
874 u32 usecs_delay = nla_get_u32(tb[TCA_FQ_FLOW_REFILL_DELAY]) ;
876 q->flow_refill_delay = usecs_to_jiffies(usecs_delay);
879 if (tb[TCA_FQ_ORPHAN_MASK])
880 q->orphan_mask = nla_get_u32(tb[TCA_FQ_ORPHAN_MASK]);
882 if (tb[TCA_FQ_CE_THRESHOLD])
883 q->ce_threshold = (u64)NSEC_PER_USEC *
884 nla_get_u32(tb[TCA_FQ_CE_THRESHOLD]);
886 if (tb[TCA_FQ_TIMER_SLACK])
887 q->timer_slack = nla_get_u32(tb[TCA_FQ_TIMER_SLACK]);
889 if (tb[TCA_FQ_HORIZON])
890 q->horizon = (u64)NSEC_PER_USEC *
891 nla_get_u32(tb[TCA_FQ_HORIZON]);
893 if (tb[TCA_FQ_HORIZON_DROP])
894 q->horizon_drop = nla_get_u8(tb[TCA_FQ_HORIZON_DROP]);
898 sch_tree_unlock(sch);
899 err = fq_resize(sch, fq_log);
902 while (sch->q.qlen > sch->limit) {
903 struct sk_buff *skb = fq_dequeue(sch);
907 drop_len += qdisc_pkt_len(skb);
908 rtnl_kfree_skbs(skb, skb);
911 qdisc_tree_reduce_backlog(sch, drop_count, drop_len);
913 sch_tree_unlock(sch);
917 static void fq_destroy(struct Qdisc *sch)
919 struct fq_sched_data *q = qdisc_priv(sch);
923 qdisc_watchdog_cancel(&q->watchdog);
926 static int fq_init(struct Qdisc *sch, struct nlattr *opt,
927 struct netlink_ext_ack *extack)
929 struct fq_sched_data *q = qdisc_priv(sch);
933 q->flow_plimit = 100;
934 q->quantum = 2 * psched_mtu(qdisc_dev(sch));
935 q->initial_quantum = 10 * psched_mtu(qdisc_dev(sch));
936 q->flow_refill_delay = msecs_to_jiffies(40);
937 q->flow_max_rate = ~0UL;
938 q->time_next_delayed_flow = ~0ULL;
940 q->new_flows.first = NULL;
941 q->old_flows.first = NULL;
942 q->delayed = RB_ROOT;
944 q->fq_trees_log = ilog2(1024);
945 q->orphan_mask = 1024 - 1;
946 q->low_rate_threshold = 550000 / 8;
948 q->timer_slack = 10 * NSEC_PER_USEC; /* 10 usec of hrtimer slack */
950 q->horizon = 10ULL * NSEC_PER_SEC; /* 10 seconds */
951 q->horizon_drop = 1; /* by default, drop packets beyond horizon */
953 /* Default ce_threshold of 4294 seconds */
954 q->ce_threshold = (u64)NSEC_PER_USEC * ~0U;
956 qdisc_watchdog_init_clockid(&q->watchdog, sch, CLOCK_MONOTONIC);
959 err = fq_change(sch, opt, extack);
961 err = fq_resize(sch, q->fq_trees_log);
966 static int fq_dump(struct Qdisc *sch, struct sk_buff *skb)
968 struct fq_sched_data *q = qdisc_priv(sch);
969 u64 ce_threshold = q->ce_threshold;
970 u64 horizon = q->horizon;
973 opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
975 goto nla_put_failure;
977 /* TCA_FQ_FLOW_DEFAULT_RATE is not used anymore */
979 do_div(ce_threshold, NSEC_PER_USEC);
980 do_div(horizon, NSEC_PER_USEC);
982 if (nla_put_u32(skb, TCA_FQ_PLIMIT, sch->limit) ||
983 nla_put_u32(skb, TCA_FQ_FLOW_PLIMIT, q->flow_plimit) ||
984 nla_put_u32(skb, TCA_FQ_QUANTUM, q->quantum) ||
985 nla_put_u32(skb, TCA_FQ_INITIAL_QUANTUM, q->initial_quantum) ||
986 nla_put_u32(skb, TCA_FQ_RATE_ENABLE, q->rate_enable) ||
987 nla_put_u32(skb, TCA_FQ_FLOW_MAX_RATE,
988 min_t(unsigned long, q->flow_max_rate, ~0U)) ||
989 nla_put_u32(skb, TCA_FQ_FLOW_REFILL_DELAY,
990 jiffies_to_usecs(q->flow_refill_delay)) ||
991 nla_put_u32(skb, TCA_FQ_ORPHAN_MASK, q->orphan_mask) ||
992 nla_put_u32(skb, TCA_FQ_LOW_RATE_THRESHOLD,
993 q->low_rate_threshold) ||
994 nla_put_u32(skb, TCA_FQ_CE_THRESHOLD, (u32)ce_threshold) ||
995 nla_put_u32(skb, TCA_FQ_BUCKETS_LOG, q->fq_trees_log) ||
996 nla_put_u32(skb, TCA_FQ_TIMER_SLACK, q->timer_slack) ||
997 nla_put_u32(skb, TCA_FQ_HORIZON, (u32)horizon) ||
998 nla_put_u8(skb, TCA_FQ_HORIZON_DROP, q->horizon_drop))
999 goto nla_put_failure;
1001 return nla_nest_end(skb, opts);
1007 static int fq_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
1009 struct fq_sched_data *q = qdisc_priv(sch);
1010 struct tc_fq_qd_stats st;
1014 st.gc_flows = q->stat_gc_flows;
1015 st.highprio_packets = q->stat_internal_packets;
1017 st.throttled = q->stat_throttled;
1018 st.flows_plimit = q->stat_flows_plimit;
1019 st.pkts_too_long = q->stat_pkts_too_long;
1020 st.allocation_errors = q->stat_allocation_errors;
1021 st.time_next_delayed_flow = q->time_next_delayed_flow + q->timer_slack -
1023 st.flows = q->flows;
1024 st.inactive_flows = q->inactive_flows;
1025 st.throttled_flows = q->throttled_flows;
1026 st.unthrottle_latency_ns = min_t(unsigned long,
1027 q->unthrottle_latency_ns, ~0U);
1028 st.ce_mark = q->stat_ce_mark;
1029 st.horizon_drops = q->stat_horizon_drops;
1030 st.horizon_caps = q->stat_horizon_caps;
1031 sch_tree_unlock(sch);
1033 return gnet_stats_copy_app(d, &st, sizeof(st));
1036 static struct Qdisc_ops fq_qdisc_ops __read_mostly = {
1038 .priv_size = sizeof(struct fq_sched_data),
1040 .enqueue = fq_enqueue,
1041 .dequeue = fq_dequeue,
1042 .peek = qdisc_peek_dequeued,
1045 .destroy = fq_destroy,
1046 .change = fq_change,
1048 .dump_stats = fq_dump_stats,
1049 .owner = THIS_MODULE,
1052 static int __init fq_module_init(void)
1056 fq_flow_cachep = kmem_cache_create("fq_flow_cache",
1057 sizeof(struct fq_flow),
1059 if (!fq_flow_cachep)
1062 ret = register_qdisc(&fq_qdisc_ops);
1064 kmem_cache_destroy(fq_flow_cachep);
1068 static void __exit fq_module_exit(void)
1070 unregister_qdisc(&fq_qdisc_ops);
1071 kmem_cache_destroy(fq_flow_cachep);
1074 module_init(fq_module_init)
1075 module_exit(fq_module_exit)
1076 MODULE_AUTHOR("Eric Dumazet");
1077 MODULE_LICENSE("GPL");
1078 MODULE_DESCRIPTION("Fair Queue Packet Scheduler");