[linux-2.6-block.git] / net / sched / sch_codel.c

// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
/*
 * Codel - The Controlled-Delay Active Queue Management algorithm
 *
 *  Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
 *  Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
 *
 *  Implemented on linux by :
 *  Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
 *  Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
 */

#include <linux/module.h>
#include <linux/slab.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/skbuff.h>
#include <linux/prefetch.h>
#include <net/pkt_sched.h>
#include <net/codel.h>
#include <net/codel_impl.h>
#include <net/codel_qdisc.h>


#define DEFAULT_CODEL_LIMIT 1000

struct codel_sched_data {
	struct codel_params	params;
	struct codel_vars	vars;
	struct codel_stats	stats;
	u32			drop_overlimit;
};

/* This is the specific function called from codel_dequeue()
 * to dequeue a packet from queue. Note: backlog is handled in
 * codel, we dont need to reduce it here.
 */
static struct sk_buff *dequeue_func(struct codel_vars *vars, void *ctx)
{
	struct Qdisc *sch = ctx;
	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);

	if (skb) {
		sch->qstats.backlog -= qdisc_pkt_len(skb);
		prefetch(&skb->end); /* we'll need skb_shinfo() */
	}
	return skb;
}

static void drop_func(struct sk_buff *skb, void *ctx)
{
	struct Qdisc *sch = ctx;

	kfree_skb(skb);
	qdisc_qstats_drop(sch);
}

static struct sk_buff *codel_qdisc_dequeue(struct Qdisc *sch)
{
	struct codel_sched_data *q = qdisc_priv(sch);
	struct sk_buff *skb;

	skb = codel_dequeue(sch, &sch->qstats.backlog, &q->params, &q->vars,
			    &q->stats, qdisc_pkt_len, codel_get_enqueue_time,
			    drop_func, dequeue_func);

	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
	 * or HTB crashes. Defer it for next round.
	 */
	if (q->stats.drop_count && sch->q.qlen) {
		qdisc_tree_reduce_backlog(sch, q->stats.drop_count, q->stats.drop_len);
		q->stats.drop_count = 0;
		q->stats.drop_len = 0;
	}
	if (skb)
		qdisc_bstats_update(sch, skb);
	return skb;
}

static int codel_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
			       struct sk_buff **to_free)
{
	struct codel_sched_data *q;

	if (likely(qdisc_qlen(sch) < sch->limit)) {
		codel_set_enqueue_time(skb);
		return qdisc_enqueue_tail(skb, sch);
	}
	q = qdisc_priv(sch);
	q->drop_overlimit++;
	return qdisc_drop(skb, sch, to_free);
}

static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
	[TCA_CODEL_TARGET]	= { .type = NLA_U32 },
	[TCA_CODEL_LIMIT]	= { .type = NLA_U32 },
	[TCA_CODEL_INTERVAL]	= { .type = NLA_U32 },
	[TCA_CODEL_ECN]		= { .type = NLA_U32 },
	[TCA_CODEL_CE_THRESHOLD]= { .type = NLA_U32 },
};

static int codel_change(struct Qdisc *sch, struct nlattr *opt,
			struct netlink_ext_ack *extack)
{
	struct codel_sched_data *q = qdisc_priv(sch);
	struct nlattr *tb[TCA_CODEL_MAX + 1];
	unsigned int qlen, dropped = 0;
	int err;

	err = nla_parse_nested_deprecated(tb, TCA_CODEL_MAX, opt,
					  codel_policy, NULL);
	if (err < 0)
		return err;

	sch_tree_lock(sch);

	if (tb[TCA_CODEL_TARGET]) {
		u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);

		WRITE_ONCE(q->params.target,
			   ((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT);
	}

	if (tb[TCA_CODEL_CE_THRESHOLD]) {
		u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);

		WRITE_ONCE(q->params.ce_threshold,
			   (val * NSEC_PER_USEC) >> CODEL_SHIFT);
	}

	if (tb[TCA_CODEL_INTERVAL]) {
		u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);

		WRITE_ONCE(q->params.interval,
			   ((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT);
	}

	if (tb[TCA_CODEL_LIMIT])
		WRITE_ONCE(sch->limit,
			   nla_get_u32(tb[TCA_CODEL_LIMIT]));

	if (tb[TCA_CODEL_ECN])
		WRITE_ONCE(q->params.ecn,
			   !!nla_get_u32(tb[TCA_CODEL_ECN]));

	qlen = sch->q.qlen;
	while (sch->q.qlen > sch->limit) {
		struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);

		dropped += qdisc_pkt_len(skb);
		qdisc_qstats_backlog_dec(sch, skb);
		rtnl_qdisc_drop(skb, sch);
	}
	qdisc_tree_reduce_backlog(sch, qlen - sch->q.qlen, dropped);

	sch_tree_unlock(sch);
	return 0;
}

static int codel_init(struct Qdisc *sch, struct nlattr *opt,
		      struct netlink_ext_ack *extack)
{
	struct codel_sched_data *q = qdisc_priv(sch);

	sch->limit = DEFAULT_CODEL_LIMIT;

	codel_params_init(&q->params);
	codel_vars_init(&q->vars);
	codel_stats_init(&q->stats);
	q->params.mtu = psched_mtu(qdisc_dev(sch));

	if (opt) {
		int err = codel_change(sch, opt, extack);

		if (err)
			return err;
	}

	if (sch->limit >= 1)
		sch->flags |= TCQ_F_CAN_BYPASS;
	else
		sch->flags &= ~TCQ_F_CAN_BYPASS;

	return 0;
}

static int codel_dump(struct Qdisc *sch, struct sk_buff *skb)
{
	struct codel_sched_data *q = qdisc_priv(sch);
	codel_time_t ce_threshold;
	struct nlattr *opts;

	opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
	if (opts == NULL)
		goto nla_put_failure;

	if (nla_put_u32(skb, TCA_CODEL_TARGET,
			codel_time_to_us(READ_ONCE(q->params.target))) ||
	    nla_put_u32(skb, TCA_CODEL_LIMIT,
			READ_ONCE(sch->limit)) ||
	    nla_put_u32(skb, TCA_CODEL_INTERVAL,
			codel_time_to_us(READ_ONCE(q->params.interval))) ||
	    nla_put_u32(skb, TCA_CODEL_ECN,
			READ_ONCE(q->params.ecn)))
		goto nla_put_failure;
	ce_threshold = READ_ONCE(q->params.ce_threshold);
	if (ce_threshold != CODEL_DISABLED_THRESHOLD &&
	    nla_put_u32(skb, TCA_CODEL_CE_THRESHOLD,
			codel_time_to_us(ce_threshold)))
		goto nla_put_failure;
	return nla_nest_end(skb, opts);

nla_put_failure:
	nla_nest_cancel(skb, opts);
	return -1;
}

static int codel_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
{
	const struct codel_sched_data *q = qdisc_priv(sch);
	struct tc_codel_xstats st = {
		.maxpacket	= q->stats.maxpacket,
		.count		= q->vars.count,
		.lastcount	= q->vars.lastcount,
		.drop_overlimit = q->drop_overlimit,
		.ldelay		= codel_time_to_us(q->vars.ldelay),
		.dropping	= q->vars.dropping,
		.ecn_mark	= q->stats.ecn_mark,
		.ce_mark	= q->stats.ce_mark,
	};

	if (q->vars.dropping) {
		codel_tdiff_t delta = q->vars.drop_next - codel_get_time();

		if (delta >= 0)
			st.drop_next = codel_time_to_us(delta);
		else
			st.drop_next = -codel_time_to_us(-delta);
	}

	return gnet_stats_copy_app(d, &st, sizeof(st));
}

static void codel_reset(struct Qdisc *sch)
{
	struct codel_sched_data *q = qdisc_priv(sch);

	qdisc_reset_queue(sch);
	codel_vars_init(&q->vars);
}

static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
	.id		=	"codel",
	.priv_size	=	sizeof(struct codel_sched_data),

	.enqueue	=	codel_qdisc_enqueue,
	.dequeue	=	codel_qdisc_dequeue,
	.peek		=	qdisc_peek_dequeued,
	.init		=	codel_init,
	.reset		=	codel_reset,
	.change 	=	codel_change,
	.dump		=	codel_dump,
	.dump_stats	=	codel_dump_stats,
	.owner		=	THIS_MODULE,
};
MODULE_ALIAS_NET_SCH("codel");

static int __init codel_module_init(void)
{
	return register_qdisc(&codel_qdisc_ops);
}

static void __exit codel_module_exit(void)
{
	unregister_qdisc(&codel_qdisc_ops);
}

module_init(codel_module_init)
module_exit(codel_module_exit)

MODULE_DESCRIPTION("Controlled Delay queue discipline");
MODULE_AUTHOR("Dave Taht");
MODULE_AUTHOR("Eric Dumazet");
MODULE_LICENSE("Dual BSD/GPL");
Commit	Line	Data
32c7eec2	1	// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
76e3cc12 ED	2	/*
	3	* Codel - The Controlled-Delay Active Queue Management algorithm
	4	*
	5	* Copyright (C) 2011-2012 Kathleen Nichols <nichols@pollere.com>
	6	* Copyright (C) 2011-2012 Van Jacobson <van@pollere.net>
	7	*
	8	* Implemented on linux by :
	9	* Copyright (C) 2012 Michael D. Taht <dave.taht@bufferbloat.net>
80ba92fa	10	* Copyright (C) 2012,2015 Eric Dumazet <edumazet@google.com>
76e3cc12 ED	11	*/
	12
	13	#include <linux/module.h>
	14	#include <linux/slab.h>
	15	#include <linux/types.h>
	16	#include <linux/kernel.h>
	17	#include <linux/errno.h>
	18	#include <linux/skbuff.h>
ce5b4b97	19	#include <linux/prefetch.h>
76e3cc12 ED	20	#include <net/pkt_sched.h>
76e3cc12 ED	21	#include <net/codel.h>
d068ca2a MK	22	#include <net/codel_impl.h>
d068ca2a MK	23	#include <net/codel_qdisc.h>
76e3cc12 ED	24
	25
	26	#define DEFAULT_CODEL_LIMIT 1000
	27
	28	struct codel_sched_data {
	29	struct codel_params params;
	30	struct codel_vars vars;
	31	struct codel_stats stats;
	32	u32 drop_overlimit;
	33	};
	34
	35	/* This is the specific function called from codel_dequeue()
	36	* to dequeue a packet from queue. Note: backlog is handled in
	37	* codel, we dont need to reduce it here.
	38	*/
79bdc4c8	39	static struct sk_buff dequeue_func(struct codel_vars vars, void *ctx)
76e3cc12	40	{
79bdc4c8	41	struct Qdisc *sch = ctx;
ed760cb8	42	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
76e3cc12	43
051c7b39	44	if (skb) {
79bdc4c8	45	sch->qstats.backlog -= qdisc_pkt_len(skb);
051c7b39 JJB	46	prefetch(&skb->end); /* we'll need skb_shinfo() */
051c7b39 JJB	47	}
76e3cc12 ED	48	return skb;
	49	}
	50
79bdc4c8 MK	51	static void drop_func(struct sk_buff skb, void ctx)
	52	{
	53	struct Qdisc *sch = ctx;
	54
520ac30f ED	55	kfree_skb(skb);
520ac30f ED	56	qdisc_qstats_drop(sch);
79bdc4c8 MK	57	}
79bdc4c8 MK	58
76e3cc12 ED	59	static struct sk_buff codel_qdisc_dequeue(struct Qdisc sch)
	60	{
	61	struct codel_sched_data *q = qdisc_priv(sch);
	62	struct sk_buff *skb;
	63
79bdc4c8 MK	64	skb = codel_dequeue(sch, &sch->qstats.backlog, &q->params, &q->vars,
	65	&q->stats, qdisc_pkt_len, codel_get_enqueue_time,
	66	drop_func, dequeue_func);
865ec552	67
2ccccf5f	68	/* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
76e3cc12 ED	69	* or HTB crashes. Defer it for next round.
	70	*/
	71	if (q->stats.drop_count && sch->q.qlen) {
2ccccf5f	72	qdisc_tree_reduce_backlog(sch, q->stats.drop_count, q->stats.drop_len);
76e3cc12	73	q->stats.drop_count = 0;
2ccccf5f	74	q->stats.drop_len = 0;
76e3cc12 ED	75	}
	76	if (skb)
	77	qdisc_bstats_update(sch, skb);
	78	return skb;
	79	}
	80
520ac30f ED	81	static int codel_qdisc_enqueue(struct sk_buff skb, struct Qdisc sch,
520ac30f ED	82	struct sk_buff **to_free)
76e3cc12 ED	83	{
	84	struct codel_sched_data *q;
	85
	86	if (likely(qdisc_qlen(sch) < sch->limit)) {
	87	codel_set_enqueue_time(skb);
	88	return qdisc_enqueue_tail(skb, sch);
	89	}
	90	q = qdisc_priv(sch);
	91	q->drop_overlimit++;
520ac30f	92	return qdisc_drop(skb, sch, to_free);
76e3cc12 ED	93	}
	94
	95	static const struct nla_policy codel_policy[TCA_CODEL_MAX + 1] = {
	96	[TCA_CODEL_TARGET] = { .type = NLA_U32 },
	97	[TCA_CODEL_LIMIT] = { .type = NLA_U32 },
	98	[TCA_CODEL_INTERVAL] = { .type = NLA_U32 },
	99	[TCA_CODEL_ECN] = { .type = NLA_U32 },
80ba92fa	100	[TCA_CODEL_CE_THRESHOLD]= { .type = NLA_U32 },
76e3cc12 ED	101	};
76e3cc12 ED	102
2030721c AA	103	static int codel_change(struct Qdisc sch, struct nlattr opt,
2030721c AA	104	struct netlink_ext_ack *extack)
76e3cc12 ED	105	{
	106	struct codel_sched_data *q = qdisc_priv(sch);
	107	struct nlattr *tb[TCA_CODEL_MAX + 1];
2ccccf5f	108	unsigned int qlen, dropped = 0;
76e3cc12 ED	109	int err;
76e3cc12 ED	110
8cb08174 JB	111	err = nla_parse_nested_deprecated(tb, TCA_CODEL_MAX, opt,
8cb08174 JB	112	codel_policy, NULL);
76e3cc12 ED	113	if (err < 0)
	114	return err;
	115
	116	sch_tree_lock(sch);
	117
	118	if (tb[TCA_CODEL_TARGET]) {
	119	u32 target = nla_get_u32(tb[TCA_CODEL_TARGET]);
	120
c45bd26c ED	121	WRITE_ONCE(q->params.target,
c45bd26c ED	122	((u64)target * NSEC_PER_USEC) >> CODEL_SHIFT);
76e3cc12 ED	123	}
76e3cc12 ED	124
80ba92fa ED	125	if (tb[TCA_CODEL_CE_THRESHOLD]) {
	126	u64 val = nla_get_u32(tb[TCA_CODEL_CE_THRESHOLD]);
	127
c45bd26c ED	128	WRITE_ONCE(q->params.ce_threshold,
c45bd26c ED	129	(val * NSEC_PER_USEC) >> CODEL_SHIFT);
80ba92fa ED	130	}
80ba92fa ED	131
76e3cc12 ED	132	if (tb[TCA_CODEL_INTERVAL]) {
	133	u32 interval = nla_get_u32(tb[TCA_CODEL_INTERVAL]);
	134
c45bd26c ED	135	WRITE_ONCE(q->params.interval,
c45bd26c ED	136	((u64)interval * NSEC_PER_USEC) >> CODEL_SHIFT);
76e3cc12 ED	137	}
	138
	139	if (tb[TCA_CODEL_LIMIT])
c45bd26c ED	140	WRITE_ONCE(sch->limit,
c45bd26c ED	141	nla_get_u32(tb[TCA_CODEL_LIMIT]));
76e3cc12 ED	142
76e3cc12 ED	143	if (tb[TCA_CODEL_ECN])
c45bd26c ED	144	WRITE_ONCE(q->params.ecn,
c45bd26c ED	145	!!nla_get_u32(tb[TCA_CODEL_ECN]));
76e3cc12 ED	146
	147	qlen = sch->q.qlen;
	148	while (sch->q.qlen > sch->limit) {
ed760cb8	149	struct sk_buff *skb = __qdisc_dequeue_head(&sch->q);
76e3cc12	150
2ccccf5f	151	dropped += qdisc_pkt_len(skb);
25331d6c	152	qdisc_qstats_backlog_dec(sch, skb);
b3d7e2b2	153	rtnl_qdisc_drop(skb, sch);
76e3cc12	154	}
2ccccf5f	155	qdisc_tree_reduce_backlog(sch, qlen - sch->q.qlen, dropped);
76e3cc12 ED	156
	157	sch_tree_unlock(sch);
	158	return 0;
	159	}
	160
e63d7dfd AA	161	static int codel_init(struct Qdisc sch, struct nlattr opt,
e63d7dfd AA	162	struct netlink_ext_ack *extack)
76e3cc12 ED	163	{
	164	struct codel_sched_data *q = qdisc_priv(sch);
	165
	166	sch->limit = DEFAULT_CODEL_LIMIT;
	167
79bdc4c8	168	codel_params_init(&q->params);
76e3cc12 ED	169	codel_vars_init(&q->vars);
76e3cc12 ED	170	codel_stats_init(&q->stats);
79bdc4c8	171	q->params.mtu = psched_mtu(qdisc_dev(sch));
76e3cc12 ED	172
76e3cc12 ED	173	if (opt) {
2030721c	174	int err = codel_change(sch, opt, extack);
76e3cc12 ED	175
	176	if (err)
	177	return err;
	178	}
	179
	180	if (sch->limit >= 1)
	181	sch->flags \|= TCQ_F_CAN_BYPASS;
	182	else
	183	sch->flags &= ~TCQ_F_CAN_BYPASS;
	184
	185	return 0;
	186	}
	187
	188	static int codel_dump(struct Qdisc sch, struct sk_buff skb)
	189	{
	190	struct codel_sched_data *q = qdisc_priv(sch);
c45bd26c	191	codel_time_t ce_threshold;
76e3cc12 ED	192	struct nlattr *opts;
76e3cc12 ED	193
ae0be8de	194	opts = nla_nest_start_noflag(skb, TCA_OPTIONS);
76e3cc12 ED	195	if (opts == NULL)
	196	goto nla_put_failure;
	197
	198	if (nla_put_u32(skb, TCA_CODEL_TARGET,
c45bd26c	199	codel_time_to_us(READ_ONCE(q->params.target))) \|\|
76e3cc12	200	nla_put_u32(skb, TCA_CODEL_LIMIT,
c45bd26c	201	READ_ONCE(sch->limit)) \|\|
76e3cc12	202	nla_put_u32(skb, TCA_CODEL_INTERVAL,
c45bd26c	203	codel_time_to_us(READ_ONCE(q->params.interval))) \|\|
76e3cc12	204	nla_put_u32(skb, TCA_CODEL_ECN,
c45bd26c	205	READ_ONCE(q->params.ecn)))
76e3cc12	206	goto nla_put_failure;
c45bd26c ED	207	ce_threshold = READ_ONCE(q->params.ce_threshold);
c45bd26c ED	208	if (ce_threshold != CODEL_DISABLED_THRESHOLD &&
80ba92fa	209	nla_put_u32(skb, TCA_CODEL_CE_THRESHOLD,
c45bd26c	210	codel_time_to_us(ce_threshold)))
80ba92fa	211	goto nla_put_failure;
76e3cc12 ED	212	return nla_nest_end(skb, opts);
	213
	214	nla_put_failure:
	215	nla_nest_cancel(skb, opts);
	216	return -1;
	217	}
	218
	219	static int codel_dump_stats(struct Qdisc sch, struct gnet_dump d)
	220	{
	221	const struct codel_sched_data *q = qdisc_priv(sch);
	222	struct tc_codel_xstats st = {
	223	.maxpacket = q->stats.maxpacket,
	224	.count = q->vars.count,
	225	.lastcount = q->vars.lastcount,
	226	.drop_overlimit = q->drop_overlimit,
	227	.ldelay = codel_time_to_us(q->vars.ldelay),
	228	.dropping = q->vars.dropping,
	229	.ecn_mark = q->stats.ecn_mark,
80ba92fa	230	.ce_mark = q->stats.ce_mark,
76e3cc12 ED	231	};
	232
	233	if (q->vars.dropping) {
	234	codel_tdiff_t delta = q->vars.drop_next - codel_get_time();
	235
	236	if (delta >= 0)
	237	st.drop_next = codel_time_to_us(delta);
	238	else
	239	st.drop_next = -codel_time_to_us(-delta);
	240	}
	241
	242	return gnet_stats_copy_app(d, &st, sizeof(st));
	243	}
	244
	245	static void codel_reset(struct Qdisc *sch)
	246	{
	247	struct codel_sched_data *q = qdisc_priv(sch);
	248
	249	qdisc_reset_queue(sch);
	250	codel_vars_init(&q->vars);
	251	}
	252
	253	static struct Qdisc_ops codel_qdisc_ops __read_mostly = {
	254	.id = "codel",
	255	.priv_size = sizeof(struct codel_sched_data),
	256
	257	.enqueue = codel_qdisc_enqueue,
	258	.dequeue = codel_qdisc_dequeue,
	259	.peek = qdisc_peek_dequeued,
	260	.init = codel_init,
	261	.reset = codel_reset,
	262	.change = codel_change,
	263	.dump = codel_dump,
	264	.dump_stats = codel_dump_stats,
	265	.owner = THIS_MODULE,
	266	};
241a94ab	267	MODULE_ALIAS_NET_SCH("codel");
76e3cc12 ED	268
	269	static int __init codel_module_init(void)
	270	{
	271	return register_qdisc(&codel_qdisc_ops);
	272	}
	273
	274	static void __exit codel_module_exit(void)
	275	{
	276	unregister_qdisc(&codel_qdisc_ops);
	277	}
	278
	279	module_init(codel_module_init)
	280	module_exit(codel_module_exit)
	281
	282	MODULE_DESCRIPTION("Controlled Delay queue discipline");
	283	MODULE_AUTHOR("Dave Taht");
	284	MODULE_AUTHOR("Eric Dumazet");
	285	MODULE_LICENSE("Dual BSD/GPL");