xref: /openbmc/linux/net/sched/sch_fq.c (revision a06c488d)
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 TCP_NEW_SYN_RECV request socket
228 	 * or a listener (SYNCOOKIE mode)
229 	 * 1) request sockets are not full blown,
230 	 *    they do not contain sk_pacing_rate
231 	 * 2) They are not part of a 'flow' yet
232 	 * 3) We do not want to rate limit them (eg SYNFLOOD attack),
233 	 *    especially if the listener set SO_MAX_PACING_RATE
234 	 * 4) We pretend they are orphaned
235 	 */
236 	if (!sk || sk_listener(sk)) {
237 		unsigned long hash = skb_get_hash(skb) & q->orphan_mask;
238 
239 		/* By forcing low order bit to 1, we make sure to not
240 		 * collide with a local flow (socket pointers are word aligned)
241 		 */
242 		sk = (struct sock *)((hash << 1) | 1UL);
243 		skb_orphan(skb);
244 	}
245 
246 	root = &q->fq_root[hash_32((u32)(long)sk, q->fq_trees_log)];
247 
248 	if (q->flows >= (2U << q->fq_trees_log) &&
249 	    q->inactive_flows > q->flows/2)
250 		fq_gc(q, root, sk);
251 
252 	p = &root->rb_node;
253 	parent = NULL;
254 	while (*p) {
255 		parent = *p;
256 
257 		f = container_of(parent, struct fq_flow, fq_node);
258 		if (f->sk == sk) {
259 			/* socket might have been reallocated, so check
260 			 * if its sk_hash is the same.
261 			 * It not, we need to refill credit with
262 			 * initial quantum
263 			 */
264 			if (unlikely(skb->sk &&
265 				     f->socket_hash != sk->sk_hash)) {
266 				f->credit = q->initial_quantum;
267 				f->socket_hash = sk->sk_hash;
268 				f->time_next_packet = 0ULL;
269 			}
270 			return f;
271 		}
272 		if (f->sk > sk)
273 			p = &parent->rb_right;
274 		else
275 			p = &parent->rb_left;
276 	}
277 
278 	f = kmem_cache_zalloc(fq_flow_cachep, GFP_ATOMIC | __GFP_NOWARN);
279 	if (unlikely(!f)) {
280 		q->stat_allocation_errors++;
281 		return &q->internal;
282 	}
283 	fq_flow_set_detached(f);
284 	f->sk = sk;
285 	if (skb->sk)
286 		f->socket_hash = sk->sk_hash;
287 	f->credit = q->initial_quantum;
288 
289 	rb_link_node(&f->fq_node, parent, p);
290 	rb_insert_color(&f->fq_node, root);
291 
292 	q->flows++;
293 	q->inactive_flows++;
294 	return f;
295 }
296 
297 
298 /* remove one skb from head of flow queue */
299 static struct sk_buff *fq_dequeue_head(struct Qdisc *sch, struct fq_flow *flow)
300 {
301 	struct sk_buff *skb = flow->head;
302 
303 	if (skb) {
304 		flow->head = skb->next;
305 		skb->next = NULL;
306 		flow->qlen--;
307 		qdisc_qstats_backlog_dec(sch, skb);
308 		sch->q.qlen--;
309 	}
310 	return skb;
311 }
312 
313 /* We might add in the future detection of retransmits
314  * For the time being, just return false
315  */
316 static bool skb_is_retransmit(struct sk_buff *skb)
317 {
318 	return false;
319 }
320 
321 /* add skb to flow queue
322  * flow queue is a linked list, kind of FIFO, except for TCP retransmits
323  * We special case tcp retransmits to be transmitted before other packets.
324  * We rely on fact that TCP retransmits are unlikely, so we do not waste
325  * a separate queue or a pointer.
326  * head->  [retrans pkt 1]
327  *         [retrans pkt 2]
328  *         [ normal pkt 1]
329  *         [ normal pkt 2]
330  *         [ normal pkt 3]
331  * tail->  [ normal pkt 4]
332  */
333 static void flow_queue_add(struct fq_flow *flow, struct sk_buff *skb)
334 {
335 	struct sk_buff *prev, *head = flow->head;
336 
337 	skb->next = NULL;
338 	if (!head) {
339 		flow->head = skb;
340 		flow->tail = skb;
341 		return;
342 	}
343 	if (likely(!skb_is_retransmit(skb))) {
344 		flow->tail->next = skb;
345 		flow->tail = skb;
346 		return;
347 	}
348 
349 	/* This skb is a tcp retransmit,
350 	 * find the last retrans packet in the queue
351 	 */
352 	prev = NULL;
353 	while (skb_is_retransmit(head)) {
354 		prev = head;
355 		head = head->next;
356 		if (!head)
357 			break;
358 	}
359 	if (!prev) { /* no rtx packet in queue, become the new head */
360 		skb->next = flow->head;
361 		flow->head = skb;
362 	} else {
363 		if (prev == flow->tail)
364 			flow->tail = skb;
365 		else
366 			skb->next = prev->next;
367 		prev->next = skb;
368 	}
369 }
370 
371 static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch)
372 {
373 	struct fq_sched_data *q = qdisc_priv(sch);
374 	struct fq_flow *f;
375 
376 	if (unlikely(sch->q.qlen >= sch->limit))
377 		return qdisc_drop(skb, sch);
378 
379 	f = fq_classify(skb, q);
380 	if (unlikely(f->qlen >= q->flow_plimit && f != &q->internal)) {
381 		q->stat_flows_plimit++;
382 		return qdisc_drop(skb, sch);
383 	}
384 
385 	f->qlen++;
386 	if (skb_is_retransmit(skb))
387 		q->stat_tcp_retrans++;
388 	qdisc_qstats_backlog_inc(sch, skb);
389 	if (fq_flow_is_detached(f)) {
390 		fq_flow_add_tail(&q->new_flows, f);
391 		if (time_after(jiffies, f->age + q->flow_refill_delay))
392 			f->credit = max_t(u32, f->credit, q->quantum);
393 		q->inactive_flows--;
394 	}
395 
396 	/* Note: this overwrites f->age */
397 	flow_queue_add(f, skb);
398 
399 	if (unlikely(f == &q->internal)) {
400 		q->stat_internal_packets++;
401 	}
402 	sch->q.qlen++;
403 
404 	return NET_XMIT_SUCCESS;
405 }
406 
407 static void fq_check_throttled(struct fq_sched_data *q, u64 now)
408 {
409 	struct rb_node *p;
410 
411 	if (q->time_next_delayed_flow > now)
412 		return;
413 
414 	q->time_next_delayed_flow = ~0ULL;
415 	while ((p = rb_first(&q->delayed)) != NULL) {
416 		struct fq_flow *f = container_of(p, struct fq_flow, rate_node);
417 
418 		if (f->time_next_packet > now) {
419 			q->time_next_delayed_flow = f->time_next_packet;
420 			break;
421 		}
422 		rb_erase(p, &q->delayed);
423 		q->throttled_flows--;
424 		fq_flow_add_tail(&q->old_flows, f);
425 	}
426 }
427 
428 static struct sk_buff *fq_dequeue(struct Qdisc *sch)
429 {
430 	struct fq_sched_data *q = qdisc_priv(sch);
431 	u64 now = ktime_get_ns();
432 	struct fq_flow_head *head;
433 	struct sk_buff *skb;
434 	struct fq_flow *f;
435 	u32 rate;
436 
437 	skb = fq_dequeue_head(sch, &q->internal);
438 	if (skb)
439 		goto out;
440 	fq_check_throttled(q, now);
441 begin:
442 	head = &q->new_flows;
443 	if (!head->first) {
444 		head = &q->old_flows;
445 		if (!head->first) {
446 			if (q->time_next_delayed_flow != ~0ULL)
447 				qdisc_watchdog_schedule_ns(&q->watchdog,
448 							   q->time_next_delayed_flow,
449 							   false);
450 			return NULL;
451 		}
452 	}
453 	f = head->first;
454 
455 	if (f->credit <= 0) {
456 		f->credit += q->quantum;
457 		head->first = f->next;
458 		fq_flow_add_tail(&q->old_flows, f);
459 		goto begin;
460 	}
461 
462 	skb = f->head;
463 	if (unlikely(skb && now < f->time_next_packet &&
464 		     !skb_is_tcp_pure_ack(skb))) {
465 		head->first = f->next;
466 		fq_flow_set_throttled(q, f);
467 		goto begin;
468 	}
469 
470 	skb = fq_dequeue_head(sch, f);
471 	if (!skb) {
472 		head->first = f->next;
473 		/* force a pass through old_flows to prevent starvation */
474 		if ((head == &q->new_flows) && q->old_flows.first) {
475 			fq_flow_add_tail(&q->old_flows, f);
476 		} else {
477 			fq_flow_set_detached(f);
478 			q->inactive_flows++;
479 		}
480 		goto begin;
481 	}
482 	prefetch(&skb->end);
483 	f->credit -= qdisc_pkt_len(skb);
484 
485 	if (f->credit > 0 || !q->rate_enable)
486 		goto out;
487 
488 	/* Do not pace locally generated ack packets */
489 	if (skb_is_tcp_pure_ack(skb))
490 		goto out;
491 
492 	rate = q->flow_max_rate;
493 	if (skb->sk)
494 		rate = min(skb->sk->sk_pacing_rate, rate);
495 
496 	if (rate != ~0U) {
497 		u32 plen = max(qdisc_pkt_len(skb), q->quantum);
498 		u64 len = (u64)plen * NSEC_PER_SEC;
499 
500 		if (likely(rate))
501 			do_div(len, rate);
502 		/* Since socket rate can change later,
503 		 * clamp the delay to 1 second.
504 		 * Really, providers of too big packets should be fixed !
505 		 */
506 		if (unlikely(len > NSEC_PER_SEC)) {
507 			len = NSEC_PER_SEC;
508 			q->stat_pkts_too_long++;
509 		}
510 
511 		f->time_next_packet = now + len;
512 	}
513 out:
514 	qdisc_bstats_update(sch, skb);
515 	return skb;
516 }
517 
518 static void fq_reset(struct Qdisc *sch)
519 {
520 	struct fq_sched_data *q = qdisc_priv(sch);
521 	struct rb_root *root;
522 	struct sk_buff *skb;
523 	struct rb_node *p;
524 	struct fq_flow *f;
525 	unsigned int idx;
526 
527 	while ((skb = fq_dequeue_head(sch, &q->internal)) != NULL)
528 		kfree_skb(skb);
529 
530 	if (!q->fq_root)
531 		return;
532 
533 	for (idx = 0; idx < (1U << q->fq_trees_log); idx++) {
534 		root = &q->fq_root[idx];
535 		while ((p = rb_first(root)) != NULL) {
536 			f = container_of(p, struct fq_flow, fq_node);
537 			rb_erase(p, root);
538 
539 			while ((skb = fq_dequeue_head(sch, f)) != NULL)
540 				kfree_skb(skb);
541 
542 			kmem_cache_free(fq_flow_cachep, f);
543 		}
544 	}
545 	q->new_flows.first	= NULL;
546 	q->old_flows.first	= NULL;
547 	q->delayed		= RB_ROOT;
548 	q->flows		= 0;
549 	q->inactive_flows	= 0;
550 	q->throttled_flows	= 0;
551 }
552 
553 static void fq_rehash(struct fq_sched_data *q,
554 		      struct rb_root *old_array, u32 old_log,
555 		      struct rb_root *new_array, u32 new_log)
556 {
557 	struct rb_node *op, **np, *parent;
558 	struct rb_root *oroot, *nroot;
559 	struct fq_flow *of, *nf;
560 	int fcnt = 0;
561 	u32 idx;
562 
563 	for (idx = 0; idx < (1U << old_log); idx++) {
564 		oroot = &old_array[idx];
565 		while ((op = rb_first(oroot)) != NULL) {
566 			rb_erase(op, oroot);
567 			of = container_of(op, struct fq_flow, fq_node);
568 			if (fq_gc_candidate(of)) {
569 				fcnt++;
570 				kmem_cache_free(fq_flow_cachep, of);
571 				continue;
572 			}
573 			nroot = &new_array[hash_32((u32)(long)of->sk, new_log)];
574 
575 			np = &nroot->rb_node;
576 			parent = NULL;
577 			while (*np) {
578 				parent = *np;
579 
580 				nf = container_of(parent, struct fq_flow, fq_node);
581 				BUG_ON(nf->sk == of->sk);
582 
583 				if (nf->sk > of->sk)
584 					np = &parent->rb_right;
585 				else
586 					np = &parent->rb_left;
587 			}
588 
589 			rb_link_node(&of->fq_node, parent, np);
590 			rb_insert_color(&of->fq_node, nroot);
591 		}
592 	}
593 	q->flows -= fcnt;
594 	q->inactive_flows -= fcnt;
595 	q->stat_gc_flows += fcnt;
596 }
597 
598 static void *fq_alloc_node(size_t sz, int node)
599 {
600 	void *ptr;
601 
602 	ptr = kmalloc_node(sz, GFP_KERNEL | __GFP_REPEAT | __GFP_NOWARN, node);
603 	if (!ptr)
604 		ptr = vmalloc_node(sz, node);
605 	return ptr;
606 }
607 
608 static void fq_free(void *addr)
609 {
610 	kvfree(addr);
611 }
612 
613 static int fq_resize(struct Qdisc *sch, u32 log)
614 {
615 	struct fq_sched_data *q = qdisc_priv(sch);
616 	struct rb_root *array;
617 	void *old_fq_root;
618 	u32 idx;
619 
620 	if (q->fq_root && log == q->fq_trees_log)
621 		return 0;
622 
623 	/* If XPS was setup, we can allocate memory on right NUMA node */
624 	array = fq_alloc_node(sizeof(struct rb_root) << log,
625 			      netdev_queue_numa_node_read(sch->dev_queue));
626 	if (!array)
627 		return -ENOMEM;
628 
629 	for (idx = 0; idx < (1U << log); idx++)
630 		array[idx] = RB_ROOT;
631 
632 	sch_tree_lock(sch);
633 
634 	old_fq_root = q->fq_root;
635 	if (old_fq_root)
636 		fq_rehash(q, old_fq_root, q->fq_trees_log, array, log);
637 
638 	q->fq_root = array;
639 	q->fq_trees_log = log;
640 
641 	sch_tree_unlock(sch);
642 
643 	fq_free(old_fq_root);
644 
645 	return 0;
646 }
647 
648 static const struct nla_policy fq_policy[TCA_FQ_MAX + 1] = {
649 	[TCA_FQ_PLIMIT]			= { .type = NLA_U32 },
650 	[TCA_FQ_FLOW_PLIMIT]		= { .type = NLA_U32 },
651 	[TCA_FQ_QUANTUM]		= { .type = NLA_U32 },
652 	[TCA_FQ_INITIAL_QUANTUM]	= { .type = NLA_U32 },
653 	[TCA_FQ_RATE_ENABLE]		= { .type = NLA_U32 },
654 	[TCA_FQ_FLOW_DEFAULT_RATE]	= { .type = NLA_U32 },
655 	[TCA_FQ_FLOW_MAX_RATE]		= { .type = NLA_U32 },
656 	[TCA_FQ_BUCKETS_LOG]		= { .type = NLA_U32 },
657 	[TCA_FQ_FLOW_REFILL_DELAY]	= { .type = NLA_U32 },
658 };
659 
660 static int fq_change(struct Qdisc *sch, struct nlattr *opt)
661 {
662 	struct fq_sched_data *q = qdisc_priv(sch);
663 	struct nlattr *tb[TCA_FQ_MAX + 1];
664 	int err, drop_count = 0;
665 	u32 fq_log;
666 
667 	if (!opt)
668 		return -EINVAL;
669 
670 	err = nla_parse_nested(tb, TCA_FQ_MAX, opt, fq_policy);
671 	if (err < 0)
672 		return err;
673 
674 	sch_tree_lock(sch);
675 
676 	fq_log = q->fq_trees_log;
677 
678 	if (tb[TCA_FQ_BUCKETS_LOG]) {
679 		u32 nval = nla_get_u32(tb[TCA_FQ_BUCKETS_LOG]);
680 
681 		if (nval >= 1 && nval <= ilog2(256*1024))
682 			fq_log = nval;
683 		else
684 			err = -EINVAL;
685 	}
686 	if (tb[TCA_FQ_PLIMIT])
687 		sch->limit = nla_get_u32(tb[TCA_FQ_PLIMIT]);
688 
689 	if (tb[TCA_FQ_FLOW_PLIMIT])
690 		q->flow_plimit = nla_get_u32(tb[TCA_FQ_FLOW_PLIMIT]);
691 
692 	if (tb[TCA_FQ_QUANTUM]) {
693 		u32 quantum = nla_get_u32(tb[TCA_FQ_QUANTUM]);
694 
695 		if (quantum > 0)
696 			q->quantum = quantum;
697 		else
698 			err = -EINVAL;
699 	}
700 
701 	if (tb[TCA_FQ_INITIAL_QUANTUM])
702 		q->initial_quantum = nla_get_u32(tb[TCA_FQ_INITIAL_QUANTUM]);
703 
704 	if (tb[TCA_FQ_FLOW_DEFAULT_RATE])
705 		pr_warn_ratelimited("sch_fq: defrate %u ignored.\n",
706 				    nla_get_u32(tb[TCA_FQ_FLOW_DEFAULT_RATE]));
707 
708 	if (tb[TCA_FQ_FLOW_MAX_RATE])
709 		q->flow_max_rate = nla_get_u32(tb[TCA_FQ_FLOW_MAX_RATE]);
710 
711 	if (tb[TCA_FQ_RATE_ENABLE]) {
712 		u32 enable = nla_get_u32(tb[TCA_FQ_RATE_ENABLE]);
713 
714 		if (enable <= 1)
715 			q->rate_enable = enable;
716 		else
717 			err = -EINVAL;
718 	}
719 
720 	if (tb[TCA_FQ_FLOW_REFILL_DELAY]) {
721 		u32 usecs_delay = nla_get_u32(tb[TCA_FQ_FLOW_REFILL_DELAY]) ;
722 
723 		q->flow_refill_delay = usecs_to_jiffies(usecs_delay);
724 	}
725 
726 	if (tb[TCA_FQ_ORPHAN_MASK])
727 		q->orphan_mask = nla_get_u32(tb[TCA_FQ_ORPHAN_MASK]);
728 
729 	if (!err) {
730 		sch_tree_unlock(sch);
731 		err = fq_resize(sch, fq_log);
732 		sch_tree_lock(sch);
733 	}
734 	while (sch->q.qlen > sch->limit) {
735 		struct sk_buff *skb = fq_dequeue(sch);
736 
737 		if (!skb)
738 			break;
739 		kfree_skb(skb);
740 		drop_count++;
741 	}
742 	qdisc_tree_decrease_qlen(sch, drop_count);
743 
744 	sch_tree_unlock(sch);
745 	return err;
746 }
747 
748 static void fq_destroy(struct Qdisc *sch)
749 {
750 	struct fq_sched_data *q = qdisc_priv(sch);
751 
752 	fq_reset(sch);
753 	fq_free(q->fq_root);
754 	qdisc_watchdog_cancel(&q->watchdog);
755 }
756 
757 static int fq_init(struct Qdisc *sch, struct nlattr *opt)
758 {
759 	struct fq_sched_data *q = qdisc_priv(sch);
760 	int err;
761 
762 	sch->limit		= 10000;
763 	q->flow_plimit		= 100;
764 	q->quantum		= 2 * psched_mtu(qdisc_dev(sch));
765 	q->initial_quantum	= 10 * psched_mtu(qdisc_dev(sch));
766 	q->flow_refill_delay	= msecs_to_jiffies(40);
767 	q->flow_max_rate	= ~0U;
768 	q->rate_enable		= 1;
769 	q->new_flows.first	= NULL;
770 	q->old_flows.first	= NULL;
771 	q->delayed		= RB_ROOT;
772 	q->fq_root		= NULL;
773 	q->fq_trees_log		= ilog2(1024);
774 	q->orphan_mask		= 1024 - 1;
775 	qdisc_watchdog_init(&q->watchdog, sch);
776 
777 	if (opt)
778 		err = fq_change(sch, opt);
779 	else
780 		err = fq_resize(sch, q->fq_trees_log);
781 
782 	return err;
783 }
784 
785 static int fq_dump(struct Qdisc *sch, struct sk_buff *skb)
786 {
787 	struct fq_sched_data *q = qdisc_priv(sch);
788 	struct nlattr *opts;
789 
790 	opts = nla_nest_start(skb, TCA_OPTIONS);
791 	if (opts == NULL)
792 		goto nla_put_failure;
793 
794 	/* TCA_FQ_FLOW_DEFAULT_RATE is not used anymore */
795 
796 	if (nla_put_u32(skb, TCA_FQ_PLIMIT, sch->limit) ||
797 	    nla_put_u32(skb, TCA_FQ_FLOW_PLIMIT, q->flow_plimit) ||
798 	    nla_put_u32(skb, TCA_FQ_QUANTUM, q->quantum) ||
799 	    nla_put_u32(skb, TCA_FQ_INITIAL_QUANTUM, q->initial_quantum) ||
800 	    nla_put_u32(skb, TCA_FQ_RATE_ENABLE, q->rate_enable) ||
801 	    nla_put_u32(skb, TCA_FQ_FLOW_MAX_RATE, q->flow_max_rate) ||
802 	    nla_put_u32(skb, TCA_FQ_FLOW_REFILL_DELAY,
803 			jiffies_to_usecs(q->flow_refill_delay)) ||
804 	    nla_put_u32(skb, TCA_FQ_ORPHAN_MASK, q->orphan_mask) ||
805 	    nla_put_u32(skb, TCA_FQ_BUCKETS_LOG, q->fq_trees_log))
806 		goto nla_put_failure;
807 
808 	return nla_nest_end(skb, opts);
809 
810 nla_put_failure:
811 	return -1;
812 }
813 
814 static int fq_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
815 {
816 	struct fq_sched_data *q = qdisc_priv(sch);
817 	u64 now = ktime_get_ns();
818 	struct tc_fq_qd_stats st = {
819 		.gc_flows		= q->stat_gc_flows,
820 		.highprio_packets	= q->stat_internal_packets,
821 		.tcp_retrans		= q->stat_tcp_retrans,
822 		.throttled		= q->stat_throttled,
823 		.flows_plimit		= q->stat_flows_plimit,
824 		.pkts_too_long		= q->stat_pkts_too_long,
825 		.allocation_errors	= q->stat_allocation_errors,
826 		.flows			= q->flows,
827 		.inactive_flows		= q->inactive_flows,
828 		.throttled_flows	= q->throttled_flows,
829 		.time_next_delayed_flow	= q->time_next_delayed_flow - now,
830 	};
831 
832 	return gnet_stats_copy_app(d, &st, sizeof(st));
833 }
834 
835 static struct Qdisc_ops fq_qdisc_ops __read_mostly = {
836 	.id		=	"fq",
837 	.priv_size	=	sizeof(struct fq_sched_data),
838 
839 	.enqueue	=	fq_enqueue,
840 	.dequeue	=	fq_dequeue,
841 	.peek		=	qdisc_peek_dequeued,
842 	.init		=	fq_init,
843 	.reset		=	fq_reset,
844 	.destroy	=	fq_destroy,
845 	.change		=	fq_change,
846 	.dump		=	fq_dump,
847 	.dump_stats	=	fq_dump_stats,
848 	.owner		=	THIS_MODULE,
849 };
850 
851 static int __init fq_module_init(void)
852 {
853 	int ret;
854 
855 	fq_flow_cachep = kmem_cache_create("fq_flow_cache",
856 					   sizeof(struct fq_flow),
857 					   0, 0, NULL);
858 	if (!fq_flow_cachep)
859 		return -ENOMEM;
860 
861 	ret = register_qdisc(&fq_qdisc_ops);
862 	if (ret)
863 		kmem_cache_destroy(fq_flow_cachep);
864 	return ret;
865 }
866 
867 static void __exit fq_module_exit(void)
868 {
869 	unregister_qdisc(&fq_qdisc_ops);
870 	kmem_cache_destroy(fq_flow_cachep);
871 }
872 
873 module_init(fq_module_init)
874 module_exit(fq_module_exit)
875 MODULE_AUTHOR("Eric Dumazet");
876 MODULE_LICENSE("GPL");
877