[linux-block.git] / net / ipv4 / tcp_cong.c

/*
 * Pluggable TCP congestion control support and newReno
 * congestion control.
 * Based on ideas from I/O scheduler support and Web100.
 *
 * Copyright (C) 2005 Stephen Hemminger <shemminger@osdl.org>
 */

#define pr_fmt(fmt) "TCP: " fmt

#include <linux/module.h>
#include <linux/mm.h>
#include <linux/types.h>
#include <linux/list.h>
#include <linux/gfp.h>
#include <net/tcp.h>

static DEFINE_SPINLOCK(tcp_cong_list_lock);
static LIST_HEAD(tcp_cong_list);

/* Simple linear search, don't expect many entries! */
static struct tcp_congestion_ops *tcp_ca_find(const char *name)
{
	struct tcp_congestion_ops *e;

	list_for_each_entry_rcu(e, &tcp_cong_list, list) {
		if (strcmp(e->name, name) == 0)
			return e;
	}

	return NULL;
}

/*
 * Attach new congestion control algorithm to the list
 * of available options.
 */
int tcp_register_congestion_control(struct tcp_congestion_ops *ca)
{
	int ret = 0;

	/* all algorithms must implement ssthresh and cong_avoid ops */
	if (!ca->ssthresh || !ca->cong_avoid) {
		pr_err("%s does not implement required ops\n", ca->name);
		return -EINVAL;
	}

	spin_lock(&tcp_cong_list_lock);
	if (tcp_ca_find(ca->name)) {
		pr_notice("%s already registered\n", ca->name);
		ret = -EEXIST;
	} else {
		list_add_tail_rcu(&ca->list, &tcp_cong_list);
		pr_info("%s registered\n", ca->name);
	}
	spin_unlock(&tcp_cong_list_lock);

	return ret;
}
EXPORT_SYMBOL_GPL(tcp_register_congestion_control);

/*
 * Remove congestion control algorithm, called from
 * the module's remove function.  Module ref counts are used
 * to ensure that this can't be done till all sockets using
 * that method are closed.
 */
void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca)
{
	spin_lock(&tcp_cong_list_lock);
	list_del_rcu(&ca->list);
	spin_unlock(&tcp_cong_list_lock);
}
EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);

/* Assign choice of congestion control. */
void tcp_assign_congestion_control(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct tcp_congestion_ops *ca;

	rcu_read_lock();
	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
		if (likely(try_module_get(ca->owner))) {
			icsk->icsk_ca_ops = ca;
			goto out;
		}
		/* Fallback to next available. The last really
		 * guaranteed fallback is Reno from this list.
		 */
	}
out:
	rcu_read_unlock();

	/* Clear out private data before diag gets it and
	 * the ca has not been initialized.
	 */
	if (ca->get_info)
		memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv));
}

void tcp_init_congestion_control(struct sock *sk)
{
	const struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->init)
		icsk->icsk_ca_ops->init(sk);
}

static void tcp_reinit_congestion_control(struct sock *sk,
					  const struct tcp_congestion_ops *ca)
{
	struct inet_connection_sock *icsk = inet_csk(sk);

	tcp_cleanup_congestion_control(sk);
	icsk->icsk_ca_ops = ca;

	if (sk->sk_state != TCP_CLOSE && icsk->icsk_ca_ops->init)
		icsk->icsk_ca_ops->init(sk);
}

/* Manage refcounts on socket close. */
void tcp_cleanup_congestion_control(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->release)
		icsk->icsk_ca_ops->release(sk);
	module_put(icsk->icsk_ca_ops->owner);
}

/* Used by sysctl to change default congestion control */
int tcp_set_default_congestion_control(const char *name)
{
	struct tcp_congestion_ops *ca;
	int ret = -ENOENT;

	spin_lock(&tcp_cong_list_lock);
	ca = tcp_ca_find(name);
#ifdef CONFIG_MODULES
	if (!ca && capable(CAP_NET_ADMIN)) {
		spin_unlock(&tcp_cong_list_lock);

		request_module("tcp_%s", name);
		spin_lock(&tcp_cong_list_lock);
		ca = tcp_ca_find(name);
	}
#endif

	if (ca) {
		ca->flags |= TCP_CONG_NON_RESTRICTED;	/* default is always allowed */
		list_move(&ca->list, &tcp_cong_list);
		ret = 0;
	}
	spin_unlock(&tcp_cong_list_lock);

	return ret;
}

/* Set default value from kernel configuration at bootup */
static int __init tcp_congestion_default(void)
{
	return tcp_set_default_congestion_control(CONFIG_DEFAULT_TCP_CONG);
}
late_initcall(tcp_congestion_default);

/* Build string with list of available congestion control values */
void tcp_get_available_congestion_control(char *buf, size_t maxlen)
{
	struct tcp_congestion_ops *ca;
	size_t offs = 0;

	rcu_read_lock();
	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
		offs += snprintf(buf + offs, maxlen - offs,
				 "%s%s",
				 offs == 0 ? "" : " ", ca->name);
	}
	rcu_read_unlock();
}

/* Get current default congestion control */
void tcp_get_default_congestion_control(char *name)
{
	struct tcp_congestion_ops *ca;
	/* We will always have reno... */
	BUG_ON(list_empty(&tcp_cong_list));

	rcu_read_lock();
	ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
	strncpy(name, ca->name, TCP_CA_NAME_MAX);
	rcu_read_unlock();
}

/* Built list of non-restricted congestion control values */
void tcp_get_allowed_congestion_control(char *buf, size_t maxlen)
{
	struct tcp_congestion_ops *ca;
	size_t offs = 0;

	*buf = '\0';
	rcu_read_lock();
	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
		if (!(ca->flags & TCP_CONG_NON_RESTRICTED))
			continue;
		offs += snprintf(buf + offs, maxlen - offs,
				 "%s%s",
				 offs == 0 ? "" : " ", ca->name);
	}
	rcu_read_unlock();
}

/* Change list of non-restricted congestion control */
int tcp_set_allowed_congestion_control(char *val)
{
	struct tcp_congestion_ops *ca;
	char *saved_clone, *clone, *name;
	int ret = 0;

	saved_clone = clone = kstrdup(val, GFP_USER);
	if (!clone)
		return -ENOMEM;

	spin_lock(&tcp_cong_list_lock);
	/* pass 1 check for bad entries */
	while ((name = strsep(&clone, " ")) && *name) {
		ca = tcp_ca_find(name);
		if (!ca) {
			ret = -ENOENT;
			goto out;
		}
	}

	/* pass 2 clear old values */
	list_for_each_entry_rcu(ca, &tcp_cong_list, list)
		ca->flags &= ~TCP_CONG_NON_RESTRICTED;

	/* pass 3 mark as allowed */
	while ((name = strsep(&val, " ")) && *name) {
		ca = tcp_ca_find(name);
		WARN_ON(!ca);
		if (ca)
			ca->flags |= TCP_CONG_NON_RESTRICTED;
	}
out:
	spin_unlock(&tcp_cong_list_lock);
	kfree(saved_clone);

	return ret;
}

/* Change congestion control for socket */
int tcp_set_congestion_control(struct sock *sk, const char *name)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct tcp_congestion_ops *ca;
	int err = 0;

	rcu_read_lock();
	ca = tcp_ca_find(name);

	/* no change asking for existing value */
	if (ca == icsk->icsk_ca_ops)
		goto out;

#ifdef CONFIG_MODULES
	/* not found attempt to autoload module */
	if (!ca && capable(CAP_NET_ADMIN)) {
		rcu_read_unlock();
		request_module("tcp_%s", name);
		rcu_read_lock();
		ca = tcp_ca_find(name);
	}
#endif
	if (!ca)
		err = -ENOENT;
	else if (!((ca->flags & TCP_CONG_NON_RESTRICTED) ||
		   ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)))
		err = -EPERM;
	else if (!try_module_get(ca->owner))
		err = -EBUSY;
	else
		tcp_reinit_congestion_control(sk, ca);
 out:
	rcu_read_unlock();
	return err;
}

/* Slow start is used when congestion window is no greater than the slow start
 * threshold. We base on RFC2581 and also handle stretch ACKs properly.
 * We do not implement RFC3465 Appropriate Byte Counting (ABC) per se but
 * something better;) a packet is only considered (s)acked in its entirety to
 * defend the ACK attacks described in the RFC. Slow start processes a stretch
 * ACK of degree N as if N acks of degree 1 are received back to back except
 * ABC caps N to 2. Slow start exits when cwnd grows over ssthresh and
 * returns the leftover acks to adjust cwnd in congestion avoidance mode.
 */
void tcp_slow_start(struct tcp_sock *tp, u32 acked)
{
	u32 cwnd = tp->snd_cwnd + acked;

	if (cwnd > tp->snd_ssthresh)
		cwnd = tp->snd_ssthresh + 1;
	tp->snd_cwnd = min(cwnd, tp->snd_cwnd_clamp);
}
EXPORT_SYMBOL_GPL(tcp_slow_start);

/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w) */
void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w)
{
	if (tp->snd_cwnd_cnt >= w) {
		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
			tp->snd_cwnd++;
		tp->snd_cwnd_cnt = 0;
	} else {
		tp->snd_cwnd_cnt++;
	}
}
EXPORT_SYMBOL_GPL(tcp_cong_avoid_ai);

/*
 * TCP Reno congestion control
 * This is special case used for fallback as well.
 */
/* This is Jacobson's slow start and congestion avoidance.
 * SIGCOMM '88, p. 328.
 */
void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (!tcp_is_cwnd_limited(sk))
		return;

	/* In "safe" area, increase. */
	if (tp->snd_cwnd <= tp->snd_ssthresh)
		tcp_slow_start(tp, acked);
	/* In dangerous area, increase slowly. */
	else
		tcp_cong_avoid_ai(tp, tp->snd_cwnd);
}
EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);

/* Slow start threshold is half the congestion window (min 2) */
u32 tcp_reno_ssthresh(struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);

	return max(tp->snd_cwnd >> 1U, 2U);
}
EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);

struct tcp_congestion_ops tcp_reno = {
	.flags		= TCP_CONG_NON_RESTRICTED,
	.name		= "reno",
	.owner		= THIS_MODULE,
	.ssthresh	= tcp_reno_ssthresh,
	.cong_avoid	= tcp_reno_cong_avoid,
};
Commit	Line	Data
317a76f9	1	/*
b92022f3	2	* Pluggable TCP congestion control support and newReno
317a76f9	3	* congestion control.
02582e9b	4	* Based on ideas from I/O scheduler support and Web100.
317a76f9 SH	5	*
	6	* Copyright (C) 2005 Stephen Hemminger <shemminger@osdl.org>
	7	*/
	8
afd46503 JP	9	#define pr_fmt(fmt) "TCP: " fmt
afd46503 JP	10
317a76f9 SH	11	#include <linux/module.h>
	12	#include <linux/mm.h>
	13	#include <linux/types.h>
	14	#include <linux/list.h>
5a0e3ad6	15	#include <linux/gfp.h>
317a76f9 SH	16	#include <net/tcp.h>
	17
	18	static DEFINE_SPINLOCK(tcp_cong_list_lock);
	19	static LIST_HEAD(tcp_cong_list);
	20
	21	/* Simple linear search, don't expect many entries! */
	22	static struct tcp_congestion_ops tcp_ca_find(const char name)
	23	{
	24	struct tcp_congestion_ops *e;
	25
5f8ef48d	26	list_for_each_entry_rcu(e, &tcp_cong_list, list) {
317a76f9 SH	27	if (strcmp(e->name, name) == 0)
	28	return e;
	29	}
	30
	31	return NULL;
	32	}
	33
	34	/*
d08df601	35	* Attach new congestion control algorithm to the list
317a76f9 SH	36	* of available options.
	37	*/
	38	int tcp_register_congestion_control(struct tcp_congestion_ops *ca)
	39	{
	40	int ret = 0;
	41
	42	/* all algorithms must implement ssthresh and cong_avoid ops */
72dc5b92	43	if (!ca->ssthresh \|\| !ca->cong_avoid) {
afd46503	44	pr_err("%s does not implement required ops\n", ca->name);
317a76f9 SH	45	return -EINVAL;
	46	}
	47
	48	spin_lock(&tcp_cong_list_lock);
	49	if (tcp_ca_find(ca->name)) {
afd46503	50	pr_notice("%s already registered\n", ca->name);
317a76f9 SH	51	ret = -EEXIST;
317a76f9 SH	52	} else {
3d2573f7	53	list_add_tail_rcu(&ca->list, &tcp_cong_list);
afd46503	54	pr_info("%s registered\n", ca->name);
317a76f9 SH	55	}
	56	spin_unlock(&tcp_cong_list_lock);
	57
	58	return ret;
	59	}
	60	EXPORT_SYMBOL_GPL(tcp_register_congestion_control);
	61
	62	/*
	63	* Remove congestion control algorithm, called from
	64	* the module's remove function. Module ref counts are used
	65	* to ensure that this can't be done till all sockets using
	66	* that method are closed.
	67	*/
	68	void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca)
	69	{
	70	spin_lock(&tcp_cong_list_lock);
	71	list_del_rcu(&ca->list);
	72	spin_unlock(&tcp_cong_list_lock);
	73	}
	74	EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);
	75
	76	/* Assign choice of congestion control. */
55d8694f	77	void tcp_assign_congestion_control(struct sock *sk)
317a76f9	78	{
6687e988	79	struct inet_connection_sock *icsk = inet_csk(sk);
317a76f9 SH	80	struct tcp_congestion_ops *ca;
317a76f9 SH	81
55d8694f FW	82	rcu_read_lock();
	83	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	84	if (likely(try_module_get(ca->owner))) {
	85	icsk->icsk_ca_ops = ca;
	86	goto out;
317a76f9	87	}
55d8694f FW	88	/* Fallback to next available. The last really
	89	* guaranteed fallback is Reno from this list.
	90	*/
317a76f9	91	}
55d8694f FW	92	out:
	93	rcu_read_unlock();
	94
	95	/* Clear out private data before diag gets it and
	96	* the ca has not been initialized.
	97	*/
	98	if (ca->get_info)
	99	memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv));
	100	}
	101
	102	void tcp_init_congestion_control(struct sock *sk)
	103	{
	104	const struct inet_connection_sock *icsk = inet_csk(sk);
317a76f9	105
6687e988 ACM	106	if (icsk->icsk_ca_ops->init)
6687e988 ACM	107	icsk->icsk_ca_ops->init(sk);
317a76f9 SH	108	}
317a76f9 SH	109
29ba4fff DB	110	static void tcp_reinit_congestion_control(struct sock *sk,
	111	const struct tcp_congestion_ops *ca)
	112	{
	113	struct inet_connection_sock *icsk = inet_csk(sk);
	114
	115	tcp_cleanup_congestion_control(sk);
	116	icsk->icsk_ca_ops = ca;
	117
	118	if (sk->sk_state != TCP_CLOSE && icsk->icsk_ca_ops->init)
	119	icsk->icsk_ca_ops->init(sk);
	120	}
	121
317a76f9	122	/* Manage refcounts on socket close. */
6687e988	123	void tcp_cleanup_congestion_control(struct sock *sk)
317a76f9	124	{
6687e988 ACM	125	struct inet_connection_sock *icsk = inet_csk(sk);
	126
	127	if (icsk->icsk_ca_ops->release)
	128	icsk->icsk_ca_ops->release(sk);
	129	module_put(icsk->icsk_ca_ops->owner);
317a76f9 SH	130	}
	131
	132	/* Used by sysctl to change default congestion control */
	133	int tcp_set_default_congestion_control(const char *name)
	134	{
	135	struct tcp_congestion_ops *ca;
	136	int ret = -ENOENT;
	137
	138	spin_lock(&tcp_cong_list_lock);
	139	ca = tcp_ca_find(name);
95a5afca	140	#ifdef CONFIG_MODULES
a8f80e8f	141	if (!ca && capable(CAP_NET_ADMIN)) {
317a76f9 SH	142	spin_unlock(&tcp_cong_list_lock);
	143
	144	request_module("tcp_%s", name);
	145	spin_lock(&tcp_cong_list_lock);
	146	ca = tcp_ca_find(name);
	147	}
	148	#endif
	149
	150	if (ca) {
164891aa	151	ca->flags \|= TCP_CONG_NON_RESTRICTED; /* default is always allowed */
317a76f9 SH	152	list_move(&ca->list, &tcp_cong_list);
	153	ret = 0;
	154	}
	155	spin_unlock(&tcp_cong_list_lock);
	156
	157	return ret;
	158	}
	159
b1736a71 SH	160	/* Set default value from kernel configuration at bootup */
	161	static int __init tcp_congestion_default(void)
	162	{
	163	return tcp_set_default_congestion_control(CONFIG_DEFAULT_TCP_CONG);
	164	}
	165	late_initcall(tcp_congestion_default);
	166
3ff825b2 SH	167	/* Build string with list of available congestion control values */
	168	void tcp_get_available_congestion_control(char *buf, size_t maxlen)
	169	{
	170	struct tcp_congestion_ops *ca;
	171	size_t offs = 0;
	172
	173	rcu_read_lock();
	174	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	175	offs += snprintf(buf + offs, maxlen - offs,
	176	"%s%s",
	177	offs == 0 ? "" : " ", ca->name);
3ff825b2 SH	178	}
	179	rcu_read_unlock();
	180	}
	181
317a76f9 SH	182	/* Get current default congestion control */
	183	void tcp_get_default_congestion_control(char *name)
	184	{
	185	struct tcp_congestion_ops *ca;
	186	/* We will always have reno... */
	187	BUG_ON(list_empty(&tcp_cong_list));
	188
	189	rcu_read_lock();
	190	ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
	191	strncpy(name, ca->name, TCP_CA_NAME_MAX);
	192	rcu_read_unlock();
	193	}
	194
ce7bc3bf SH	195	/* Built list of non-restricted congestion control values */
	196	void tcp_get_allowed_congestion_control(char *buf, size_t maxlen)
	197	{
	198	struct tcp_congestion_ops *ca;
	199	size_t offs = 0;
	200
	201	*buf = '\0';
	202	rcu_read_lock();
	203	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
164891aa	204	if (!(ca->flags & TCP_CONG_NON_RESTRICTED))
ce7bc3bf SH	205	continue;
	206	offs += snprintf(buf + offs, maxlen - offs,
	207	"%s%s",
	208	offs == 0 ? "" : " ", ca->name);
ce7bc3bf SH	209	}
	210	rcu_read_unlock();
	211	}
	212
	213	/* Change list of non-restricted congestion control */
	214	int tcp_set_allowed_congestion_control(char *val)
	215	{
	216	struct tcp_congestion_ops *ca;
c34186ed	217	char saved_clone, clone, *name;
ce7bc3bf SH	218	int ret = 0;
ce7bc3bf SH	219
c34186ed	220	saved_clone = clone = kstrdup(val, GFP_USER);
ce7bc3bf SH	221	if (!clone)
	222	return -ENOMEM;
	223
	224	spin_lock(&tcp_cong_list_lock);
	225	/* pass 1 check for bad entries */
	226	while ((name = strsep(&clone, " ")) && *name) {
	227	ca = tcp_ca_find(name);
	228	if (!ca) {
	229	ret = -ENOENT;
	230	goto out;
	231	}
	232	}
	233
164891aa	234	/* pass 2 clear old values */
ce7bc3bf	235	list_for_each_entry_rcu(ca, &tcp_cong_list, list)
164891aa	236	ca->flags &= ~TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	237
	238	/* pass 3 mark as allowed */
	239	while ((name = strsep(&val, " ")) && *name) {
	240	ca = tcp_ca_find(name);
	241	WARN_ON(!ca);
	242	if (ca)
164891aa	243	ca->flags \|= TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	244	}
	245	out:
	246	spin_unlock(&tcp_cong_list_lock);
c34186ed	247	kfree(saved_clone);
ce7bc3bf SH	248
	249	return ret;
	250	}
	251
5f8ef48d	252	/* Change congestion control for socket */
6687e988	253	int tcp_set_congestion_control(struct sock sk, const char name)
5f8ef48d	254	{
6687e988	255	struct inet_connection_sock *icsk = inet_csk(sk);
5f8ef48d SH	256	struct tcp_congestion_ops *ca;
	257	int err = 0;
	258
	259	rcu_read_lock();
	260	ca = tcp_ca_find(name);
4d4d3d1e	261
35bfbc94	262	/* no change asking for existing value */
6687e988	263	if (ca == icsk->icsk_ca_ops)
5f8ef48d SH	264	goto out;
5f8ef48d SH	265
95a5afca	266	#ifdef CONFIG_MODULES
35bfbc94	267	/* not found attempt to autoload module */
a8f80e8f	268	if (!ca && capable(CAP_NET_ADMIN)) {
35bfbc94 SH	269	rcu_read_unlock();
	270	request_module("tcp_%s", name);
	271	rcu_read_lock();
	272	ca = tcp_ca_find(name);
	273	}
	274	#endif
5f8ef48d SH	275	if (!ca)
5f8ef48d SH	276	err = -ENOENT;
52e804c6 EB	277	else if (!((ca->flags & TCP_CONG_NON_RESTRICTED) \|\|
52e804c6 EB	278	ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)))
ce7bc3bf	279	err = -EPERM;
5f8ef48d SH	280	else if (!try_module_get(ca->owner))
5f8ef48d SH	281	err = -EBUSY;
29ba4fff DB	282	else
29ba4fff DB	283	tcp_reinit_congestion_control(sk, ca);
5f8ef48d SH	284	out:
	285	rcu_read_unlock();
	286	return err;
	287	}
	288
9f9843a7 YC	289	/* Slow start is used when congestion window is no greater than the slow start
	290	* threshold. We base on RFC2581 and also handle stretch ACKs properly.
	291	* We do not implement RFC3465 Appropriate Byte Counting (ABC) per se but
	292	* something better;) a packet is only considered (s)acked in its entirety to
	293	* defend the ACK attacks described in the RFC. Slow start processes a stretch
	294	* ACK of degree N as if N acks of degree 1 are received back to back except
	295	* ABC caps N to 2. Slow start exits when cwnd grows over ssthresh and
	296	* returns the leftover acks to adjust cwnd in congestion avoidance mode.
40efc6fa	297	*/
a12a601e	298	void tcp_slow_start(struct tcp_sock *tp, u32 acked)
40efc6fa	299	{
9f9843a7	300	u32 cwnd = tp->snd_cwnd + acked;
a02ba041	301
9f9843a7 YC	302	if (cwnd > tp->snd_ssthresh)
9f9843a7 YC	303	cwnd = tp->snd_ssthresh + 1;
9f9843a7	304	tp->snd_cwnd = min(cwnd, tp->snd_cwnd_clamp);
40efc6fa SH	305	}
	306	EXPORT_SYMBOL_GPL(tcp_slow_start);
	307
758ce5c8 IJ	308	/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w) */
	309	void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w)
	310	{
	311	if (tp->snd_cwnd_cnt >= w) {
	312	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
	313	tp->snd_cwnd++;
	314	tp->snd_cwnd_cnt = 0;
	315	} else {
	316	tp->snd_cwnd_cnt++;
	317	}
	318	}
	319	EXPORT_SYMBOL_GPL(tcp_cong_avoid_ai);
	320
317a76f9 SH	321	/*
	322	* TCP Reno congestion control
	323	* This is special case used for fallback as well.
	324	*/
	325	/* This is Jacobson's slow start and congestion avoidance.
	326	* SIGCOMM '88, p. 328.
	327	*/
24901551	328	void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked)
317a76f9	329	{
6687e988 ACM	330	struct tcp_sock *tp = tcp_sk(sk);
6687e988 ACM	331
24901551	332	if (!tcp_is_cwnd_limited(sk))
317a76f9 SH	333	return;
317a76f9 SH	334
7faffa1c	335	/* In "safe" area, increase. */
e905a9ed	336	if (tp->snd_cwnd <= tp->snd_ssthresh)
9f9843a7	337	tcp_slow_start(tp, acked);
e905a9ed	338	/* In dangerous area, increase slowly. */
ca2eb567	339	else
758ce5c8	340	tcp_cong_avoid_ai(tp, tp->snd_cwnd);
317a76f9 SH	341	}
	342	EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);
	343
	344	/* Slow start threshold is half the congestion window (min 2) */
6687e988	345	u32 tcp_reno_ssthresh(struct sock *sk)
317a76f9	346	{
6687e988	347	const struct tcp_sock *tp = tcp_sk(sk);
688d1945	348
317a76f9 SH	349	return max(tp->snd_cwnd >> 1U, 2U);
	350	}
	351	EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);
	352
317a76f9	353	struct tcp_congestion_ops tcp_reno = {
164891aa	354	.flags = TCP_CONG_NON_RESTRICTED,
317a76f9 SH	355	.name = "reno",
	356	.owner = THIS_MODULE,
	357	.ssthresh = tcp_reno_ssthresh,
	358	.cong_avoid = tcp_reno_cong_avoid,
317a76f9	359	};