[linux-2.6-block.git] / net / ipv4 / tcp_cong.c

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

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

int sysctl_tcp_max_ssthresh = 0;

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("TCP %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("TCP %s already registered\n", ca->name);
		ret = -EEXIST;
	} else {
		list_add_tail_rcu(&ca->list, &tcp_cong_list);
		pr_info("TCP %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_init_congestion_control(struct sock *sk)
{
	struct inet_connection_sock *icsk = inet_csk(sk);
	struct tcp_congestion_ops *ca;

	/* if no choice made yet assign the current value set as default */
	if (icsk->icsk_ca_ops == &tcp_init_congestion_ops) {
		rcu_read_lock();
		list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
			if (try_module_get(ca->owner)) {
				icsk->icsk_ca_ops = ca;
				break;
			}

			/* fallback to next available */
		}
		rcu_read_unlock();
	}

	if (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) || capable(CAP_NET_ADMIN)))
		err = -EPERM;

	else if (!try_module_get(ca->owner))
		err = -EBUSY;

	else {
		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);
	}
 out:
	rcu_read_unlock();
	return err;
}

/* RFC2861 Check whether we are limited by application or congestion window
 * This is the inverse of cwnd check in tcp_tso_should_defer
 */
int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	u32 left;

	if (in_flight >= tp->snd_cwnd)
		return 1;

	left = tp->snd_cwnd - in_flight;
	if (sk_can_gso(sk) &&
	    left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd &&
	    left * tp->mss_cache < sk->sk_gso_max_size)
		return 1;
	return left <= tcp_max_tso_deferred_mss(tp);
}
EXPORT_SYMBOL_GPL(tcp_is_cwnd_limited);

/*
 * Slow start is used when congestion window is less than slow start
 * threshold. This version implements the basic RFC2581 version
 * and optionally supports:
 * 	RFC3742 Limited Slow Start  	  - growth limited to max_ssthresh
 *	RFC3465 Appropriate Byte Counting - growth limited by bytes acknowledged
 */
void tcp_slow_start(struct tcp_sock *tp)
{
	int cnt; /* increase in packets */

	/* RFC3465: ABC Slow start
	 * Increase only after a full MSS of bytes is acked
	 *
	 * TCP sender SHOULD increase cwnd by the number of
	 * previously unacknowledged bytes ACKed by each incoming
	 * acknowledgment, provided the increase is not more than L
	 */
	if (sysctl_tcp_abc && tp->bytes_acked < tp->mss_cache)
		return;

	if (sysctl_tcp_max_ssthresh > 0 && tp->snd_cwnd > sysctl_tcp_max_ssthresh)
		cnt = sysctl_tcp_max_ssthresh >> 1;	/* limited slow start */
	else
		cnt = tp->snd_cwnd;			/* exponential increase */

	/* RFC3465: ABC
	 * We MAY increase by 2 if discovered delayed ack
	 */
	if (sysctl_tcp_abc > 1 && tp->bytes_acked >= 2*tp->mss_cache)
		cnt <<= 1;
	tp->bytes_acked = 0;

	tp->snd_cwnd_cnt += cnt;
	while (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
		tp->snd_cwnd_cnt -= tp->snd_cwnd;
		if (tp->snd_cwnd < tp->snd_cwnd_clamp)
			tp->snd_cwnd++;
	}
}
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 in_flight)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (!tcp_is_cwnd_limited(sk, in_flight))
		return;

	/* In "safe" area, increase. */
	if (tp->snd_cwnd <= tp->snd_ssthresh)
		tcp_slow_start(tp);

	/* In dangerous area, increase slowly. */
	else if (sysctl_tcp_abc) {
		/* RFC3465: Appropriate Byte Count
		 * increase once for each full cwnd acked
		 */
		if (tp->bytes_acked >= tp->snd_cwnd*tp->mss_cache) {
			tp->bytes_acked -= tp->snd_cwnd*tp->mss_cache;
			if (tp->snd_cwnd < tp->snd_cwnd_clamp)
				tp->snd_cwnd++;
		}
	} 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);

/* Lower bound on congestion window with halving. */
u32 tcp_reno_min_cwnd(const struct sock *sk)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	return tp->snd_ssthresh/2;
}
EXPORT_SYMBOL_GPL(tcp_reno_min_cwnd);

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,
	.min_cwnd	= tcp_reno_min_cwnd,
};

/* Initial congestion control used (until SYN)
 * really reno under another name so we can tell difference
 * during tcp_set_default_congestion_control
 */
struct tcp_congestion_ops tcp_init_congestion_ops  = {
	.name		= "",
	.owner		= THIS_MODULE,
	.ssthresh	= tcp_reno_ssthresh,
	.cong_avoid	= tcp_reno_cong_avoid,
	.min_cwnd	= tcp_reno_min_cwnd,
};
EXPORT_SYMBOL_GPL(tcp_init_congestion_ops);
Commit	Line	Data
317a76f9 SH	1	/*
	2	* Plugable TCP congestion control support and newReno
	3	* congestion control.
	4	* Based on ideas from I/O scheduler suport and Web100.
	5	*
	6	* Copyright (C) 2005 Stephen Hemminger <shemminger@osdl.org>
	7	*/
	8
317a76f9 SH	9	#include <linux/module.h>
	10	#include <linux/mm.h>
	11	#include <linux/types.h>
	12	#include <linux/list.h>
5a0e3ad6	13	#include <linux/gfp.h>
317a76f9 SH	14	#include <net/tcp.h>
317a76f9 SH	15
886236c1 JH	16	int sysctl_tcp_max_ssthresh = 0;
886236c1 JH	17
317a76f9 SH	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) {
058bd4d2	44	pr_err("TCP %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)) {
058bd4d2	50	pr_notice("TCP %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);
058bd4d2	54	pr_info("TCP %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. */
6687e988	77	void tcp_init_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
4d4d3d1e SH	82	/* if no choice made yet assign the current value set as default */
	83	if (icsk->icsk_ca_ops == &tcp_init_congestion_ops) {
	84	rcu_read_lock();
	85	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	86	if (try_module_get(ca->owner)) {
	87	icsk->icsk_ca_ops = ca;
	88	break;
	89	}
5f8ef48d	90
4d4d3d1e	91	/* fallback to next available */
317a76f9	92	}
4d4d3d1e	93	rcu_read_unlock();
317a76f9	94	}
317a76f9	95
6687e988 ACM	96	if (icsk->icsk_ca_ops->init)
6687e988 ACM	97	icsk->icsk_ca_ops->init(sk);
317a76f9 SH	98	}
	99
	100	/* Manage refcounts on socket close. */
6687e988	101	void tcp_cleanup_congestion_control(struct sock *sk)
317a76f9	102	{
6687e988 ACM	103	struct inet_connection_sock *icsk = inet_csk(sk);
	104
	105	if (icsk->icsk_ca_ops->release)
	106	icsk->icsk_ca_ops->release(sk);
	107	module_put(icsk->icsk_ca_ops->owner);
317a76f9 SH	108	}
	109
	110	/* Used by sysctl to change default congestion control */
	111	int tcp_set_default_congestion_control(const char *name)
	112	{
	113	struct tcp_congestion_ops *ca;
	114	int ret = -ENOENT;
	115
	116	spin_lock(&tcp_cong_list_lock);
	117	ca = tcp_ca_find(name);
95a5afca	118	#ifdef CONFIG_MODULES
a8f80e8f	119	if (!ca && capable(CAP_NET_ADMIN)) {
317a76f9 SH	120	spin_unlock(&tcp_cong_list_lock);
	121
	122	request_module("tcp_%s", name);
	123	spin_lock(&tcp_cong_list_lock);
	124	ca = tcp_ca_find(name);
	125	}
	126	#endif
	127
	128	if (ca) {
164891aa	129	ca->flags \|= TCP_CONG_NON_RESTRICTED; /* default is always allowed */
317a76f9 SH	130	list_move(&ca->list, &tcp_cong_list);
	131	ret = 0;
	132	}
	133	spin_unlock(&tcp_cong_list_lock);
	134
	135	return ret;
	136	}
	137
b1736a71 SH	138	/* Set default value from kernel configuration at bootup */
	139	static int __init tcp_congestion_default(void)
	140	{
	141	return tcp_set_default_congestion_control(CONFIG_DEFAULT_TCP_CONG);
	142	}
	143	late_initcall(tcp_congestion_default);
	144
	145
3ff825b2 SH	146	/* Build string with list of available congestion control values */
	147	void tcp_get_available_congestion_control(char *buf, size_t maxlen)
	148	{
	149	struct tcp_congestion_ops *ca;
	150	size_t offs = 0;
	151
	152	rcu_read_lock();
	153	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	154	offs += snprintf(buf + offs, maxlen - offs,
	155	"%s%s",
	156	offs == 0 ? "" : " ", ca->name);
	157
	158	}
	159	rcu_read_unlock();
	160	}
	161
317a76f9 SH	162	/* Get current default congestion control */
	163	void tcp_get_default_congestion_control(char *name)
	164	{
	165	struct tcp_congestion_ops *ca;
	166	/* We will always have reno... */
	167	BUG_ON(list_empty(&tcp_cong_list));
	168
	169	rcu_read_lock();
	170	ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
	171	strncpy(name, ca->name, TCP_CA_NAME_MAX);
	172	rcu_read_unlock();
	173	}
	174
ce7bc3bf SH	175	/* Built list of non-restricted congestion control values */
	176	void tcp_get_allowed_congestion_control(char *buf, size_t maxlen)
	177	{
	178	struct tcp_congestion_ops *ca;
	179	size_t offs = 0;
	180
	181	*buf = '\0';
	182	rcu_read_lock();
	183	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
164891aa	184	if (!(ca->flags & TCP_CONG_NON_RESTRICTED))
ce7bc3bf SH	185	continue;
	186	offs += snprintf(buf + offs, maxlen - offs,
	187	"%s%s",
	188	offs == 0 ? "" : " ", ca->name);
	189
	190	}
	191	rcu_read_unlock();
	192	}
	193
	194	/* Change list of non-restricted congestion control */
	195	int tcp_set_allowed_congestion_control(char *val)
	196	{
	197	struct tcp_congestion_ops *ca;
c34186ed	198	char saved_clone, clone, *name;
ce7bc3bf SH	199	int ret = 0;
ce7bc3bf SH	200
c34186ed	201	saved_clone = clone = kstrdup(val, GFP_USER);
ce7bc3bf SH	202	if (!clone)
	203	return -ENOMEM;
	204
	205	spin_lock(&tcp_cong_list_lock);
	206	/* pass 1 check for bad entries */
	207	while ((name = strsep(&clone, " ")) && *name) {
	208	ca = tcp_ca_find(name);
	209	if (!ca) {
	210	ret = -ENOENT;
	211	goto out;
	212	}
	213	}
	214
164891aa	215	/* pass 2 clear old values */
ce7bc3bf	216	list_for_each_entry_rcu(ca, &tcp_cong_list, list)
164891aa	217	ca->flags &= ~TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	218
	219	/* pass 3 mark as allowed */
	220	while ((name = strsep(&val, " ")) && *name) {
	221	ca = tcp_ca_find(name);
	222	WARN_ON(!ca);
	223	if (ca)
164891aa	224	ca->flags \|= TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	225	}
	226	out:
	227	spin_unlock(&tcp_cong_list_lock);
c34186ed	228	kfree(saved_clone);
ce7bc3bf SH	229
	230	return ret;
	231	}
	232
	233
5f8ef48d	234	/* Change congestion control for socket */
6687e988	235	int tcp_set_congestion_control(struct sock sk, const char name)
5f8ef48d	236	{
6687e988	237	struct inet_connection_sock *icsk = inet_csk(sk);
5f8ef48d SH	238	struct tcp_congestion_ops *ca;
	239	int err = 0;
	240
	241	rcu_read_lock();
	242	ca = tcp_ca_find(name);
4d4d3d1e	243
35bfbc94	244	/* no change asking for existing value */
6687e988	245	if (ca == icsk->icsk_ca_ops)
5f8ef48d SH	246	goto out;
5f8ef48d SH	247
95a5afca	248	#ifdef CONFIG_MODULES
35bfbc94	249	/* not found attempt to autoload module */
a8f80e8f	250	if (!ca && capable(CAP_NET_ADMIN)) {
35bfbc94 SH	251	rcu_read_unlock();
	252	request_module("tcp_%s", name);
	253	rcu_read_lock();
	254	ca = tcp_ca_find(name);
	255	}
	256	#endif
5f8ef48d SH	257	if (!ca)
	258	err = -ENOENT;
	259
164891aa	260	else if (!((ca->flags & TCP_CONG_NON_RESTRICTED) \|\| capable(CAP_NET_ADMIN)))
ce7bc3bf SH	261	err = -EPERM;
ce7bc3bf SH	262
5f8ef48d SH	263	else if (!try_module_get(ca->owner))
	264	err = -EBUSY;
	265
	266	else {
6687e988 ACM	267	tcp_cleanup_congestion_control(sk);
6687e988 ACM	268	icsk->icsk_ca_ops = ca;
4d4d3d1e SH	269
4d4d3d1e SH	270	if (sk->sk_state != TCP_CLOSE && icsk->icsk_ca_ops->init)
6687e988	271	icsk->icsk_ca_ops->init(sk);
5f8ef48d SH	272	}
	273	out:
	274	rcu_read_unlock();
	275	return err;
	276	}
	277
cea14e0e IJ	278	/* RFC2861 Check whether we are limited by application or congestion window
	279	* This is the inverse of cwnd check in tcp_tso_should_defer
	280	*/
	281	int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
	282	{
	283	const struct tcp_sock *tp = tcp_sk(sk);
	284	u32 left;
	285
	286	if (in_flight >= tp->snd_cwnd)
	287	return 1;
	288
cea14e0e	289	left = tp->snd_cwnd - in_flight;
ce447eb9	290	if (sk_can_gso(sk) &&
246eb2af JH	291	left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd &&
246eb2af JH	292	left * tp->mss_cache < sk->sk_gso_max_size)
ce447eb9	293	return 1;
6b5a5c0d	294	return left <= tcp_max_tso_deferred_mss(tp);
cea14e0e IJ	295	}
cea14e0e IJ	296	EXPORT_SYMBOL_GPL(tcp_is_cwnd_limited);
40efc6fa SH	297
40efc6fa SH	298	/*
a02ba041 SH	299	* Slow start is used when congestion window is less than slow start
	300	* threshold. This version implements the basic RFC2581 version
	301	* and optionally supports:
	302	* RFC3742 Limited Slow Start - growth limited to max_ssthresh
	303	* RFC3465 Appropriate Byte Counting - growth limited by bytes acknowledged
40efc6fa SH	304	*/
	305	void tcp_slow_start(struct tcp_sock *tp)
	306	{
a02ba041 SH	307	int cnt; /* increase in packets */
	308
	309	/* RFC3465: ABC Slow start
	310	* Increase only after a full MSS of bytes is acked
	311	*
	312	* TCP sender SHOULD increase cwnd by the number of
	313	* previously unacknowledged bytes ACKed by each incoming
	314	* acknowledgment, provided the increase is not more than L
	315	*/
	316	if (sysctl_tcp_abc && tp->bytes_acked < tp->mss_cache)
	317	return;
	318
	319	if (sysctl_tcp_max_ssthresh > 0 && tp->snd_cwnd > sysctl_tcp_max_ssthresh)
	320	cnt = sysctl_tcp_max_ssthresh >> 1; /* limited slow start */
886236c1	321	else
a02ba041	322	cnt = tp->snd_cwnd; /* exponential increase */
886236c1	323
a02ba041 SH	324	/* RFC3465: ABC
	325	* We MAY increase by 2 if discovered delayed ack
	326	*/
886236c1 JH	327	if (sysctl_tcp_abc > 1 && tp->bytes_acked >= 2*tp->mss_cache)
886236c1 JH	328	cnt <<= 1;
40efc6fa SH	329	tp->bytes_acked = 0;
40efc6fa SH	330
886236c1 JH	331	tp->snd_cwnd_cnt += cnt;
	332	while (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
	333	tp->snd_cwnd_cnt -= tp->snd_cwnd;
	334	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
	335	tp->snd_cwnd++;
	336	}
40efc6fa SH	337	}
	338	EXPORT_SYMBOL_GPL(tcp_slow_start);
	339
758ce5c8 IJ	340	/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w) */
	341	void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w)
	342	{
	343	if (tp->snd_cwnd_cnt >= w) {
	344	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
	345	tp->snd_cwnd++;
	346	tp->snd_cwnd_cnt = 0;
	347	} else {
	348	tp->snd_cwnd_cnt++;
	349	}
	350	}
	351	EXPORT_SYMBOL_GPL(tcp_cong_avoid_ai);
	352
317a76f9 SH	353	/*
	354	* TCP Reno congestion control
	355	* This is special case used for fallback as well.
	356	*/
	357	/* This is Jacobson's slow start and congestion avoidance.
	358	* SIGCOMM '88, p. 328.
	359	*/
c3a05c60	360	void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
317a76f9	361	{
6687e988 ACM	362	struct tcp_sock *tp = tcp_sk(sk);
6687e988 ACM	363
f4805ede	364	if (!tcp_is_cwnd_limited(sk, in_flight))
317a76f9 SH	365	return;
317a76f9 SH	366
7faffa1c	367	/* In "safe" area, increase. */
e905a9ed	368	if (tp->snd_cwnd <= tp->snd_ssthresh)
7faffa1c	369	tcp_slow_start(tp);
9772efb9	370
e905a9ed	371	/* In dangerous area, increase slowly. */
9772efb9	372	else if (sysctl_tcp_abc) {
e905a9ed YH	373	/* RFC3465: Appropriate Byte Count
	374	* increase once for each full cwnd acked
	375	*/
	376	if (tp->bytes_acked >= tp->snd_cwnd*tp->mss_cache) {
	377	tp->bytes_acked -= tp->snd_cwnd*tp->mss_cache;
	378	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
	379	tp->snd_cwnd++;
	380	}
	381	} else {
758ce5c8	382	tcp_cong_avoid_ai(tp, tp->snd_cwnd);
e905a9ed	383	}
317a76f9 SH	384	}
	385	EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);
	386
	387	/* Slow start threshold is half the congestion window (min 2) */
6687e988	388	u32 tcp_reno_ssthresh(struct sock *sk)
317a76f9	389	{
6687e988	390	const struct tcp_sock *tp = tcp_sk(sk);
317a76f9 SH	391	return max(tp->snd_cwnd >> 1U, 2U);
	392	}
	393	EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);
	394
72dc5b92 SH	395	/* Lower bound on congestion window with halving. */
72dc5b92 SH	396	u32 tcp_reno_min_cwnd(const struct sock *sk)
317a76f9	397	{
6687e988	398	const struct tcp_sock *tp = tcp_sk(sk);
317a76f9 SH	399	return tp->snd_ssthresh/2;
	400	}
	401	EXPORT_SYMBOL_GPL(tcp_reno_min_cwnd);
	402
	403	struct tcp_congestion_ops tcp_reno = {
164891aa	404	.flags = TCP_CONG_NON_RESTRICTED,
317a76f9 SH	405	.name = "reno",
	406	.owner = THIS_MODULE,
	407	.ssthresh = tcp_reno_ssthresh,
	408	.cong_avoid = tcp_reno_cong_avoid,
	409	.min_cwnd = tcp_reno_min_cwnd,
	410	};
	411
5f8ef48d SH	412	/* Initial congestion control used (until SYN)
	413	* really reno under another name so we can tell difference
	414	* during tcp_set_default_congestion_control
	415	*/
	416	struct tcp_congestion_ops tcp_init_congestion_ops = {
	417	.name = "",
	418	.owner = THIS_MODULE,
	419	.ssthresh = tcp_reno_ssthresh,
	420	.cong_avoid = tcp_reno_cong_avoid,
	421	.min_cwnd = tcp_reno_min_cwnd,
	422	};
	423	EXPORT_SYMBOL_GPL(tcp_init_congestion_ops);