[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / ipv4 / tcp_cong.c

/*
 * Plugable 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>

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("%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_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();
	}

	tcp_sk(sk)->prior_ssthresh = 0;
	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) ||
		   ns_capable(sock_net(sk)->user_ns, 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
 */
bool 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 true;

	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 &&
	    left < sk->sk_gso_max_segs)
		return true;
	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 */
	unsigned int delta = 0;
	u32 snd_cwnd = tp->snd_cwnd;

	if (unlikely(!snd_cwnd)) {
		pr_err_once("snd_cwnd is nul, please report this bug.\n");
		snd_cwnd = 1U;
	}

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

	tp->snd_cwnd_cnt += cnt;
	while (tp->snd_cwnd_cnt >= snd_cwnd) {
		tp->snd_cwnd_cnt -= snd_cwnd;
		delta++;
	}
	tp->snd_cwnd = min(snd_cwnd + delta, 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 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
		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.
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>
317a76f9 SH	17
886236c1 JH	18	int sysctl_tcp_max_ssthresh = 0;
886236c1 JH	19
317a76f9 SH	20	static DEFINE_SPINLOCK(tcp_cong_list_lock);
	21	static LIST_HEAD(tcp_cong_list);
	22
	23	/* Simple linear search, don't expect many entries! */
	24	static struct tcp_congestion_ops tcp_ca_find(const char name)
	25	{
	26	struct tcp_congestion_ops *e;
	27
5f8ef48d	28	list_for_each_entry_rcu(e, &tcp_cong_list, list) {
317a76f9 SH	29	if (strcmp(e->name, name) == 0)
	30	return e;
	31	}
	32
	33	return NULL;
	34	}
	35
	36	/*
d08df601	37	* Attach new congestion control algorithm to the list
317a76f9 SH	38	* of available options.
	39	*/
	40	int tcp_register_congestion_control(struct tcp_congestion_ops *ca)
	41	{
	42	int ret = 0;
	43
	44	/* all algorithms must implement ssthresh and cong_avoid ops */
72dc5b92	45	if (!ca->ssthresh \|\| !ca->cong_avoid) {
afd46503	46	pr_err("%s does not implement required ops\n", ca->name);
317a76f9 SH	47	return -EINVAL;
	48	}
	49
	50	spin_lock(&tcp_cong_list_lock);
	51	if (tcp_ca_find(ca->name)) {
afd46503	52	pr_notice("%s already registered\n", ca->name);
317a76f9 SH	53	ret = -EEXIST;
317a76f9 SH	54	} else {
3d2573f7	55	list_add_tail_rcu(&ca->list, &tcp_cong_list);
afd46503	56	pr_info("%s registered\n", ca->name);
317a76f9 SH	57	}
	58	spin_unlock(&tcp_cong_list_lock);
	59
	60	return ret;
	61	}
	62	EXPORT_SYMBOL_GPL(tcp_register_congestion_control);
	63
	64	/*
	65	* Remove congestion control algorithm, called from
	66	* the module's remove function. Module ref counts are used
	67	* to ensure that this can't be done till all sockets using
	68	* that method are closed.
	69	*/
	70	void tcp_unregister_congestion_control(struct tcp_congestion_ops *ca)
	71	{
	72	spin_lock(&tcp_cong_list_lock);
	73	list_del_rcu(&ca->list);
	74	spin_unlock(&tcp_cong_list_lock);
	75	}
	76	EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);
	77
	78	/* Assign choice of congestion control. */
6687e988	79	void tcp_init_congestion_control(struct sock *sk)
317a76f9	80	{
6687e988	81	struct inet_connection_sock *icsk = inet_csk(sk);
317a76f9 SH	82	struct tcp_congestion_ops *ca;
317a76f9 SH	83
4d4d3d1e SH	84	/* if no choice made yet assign the current value set as default */
	85	if (icsk->icsk_ca_ops == &tcp_init_congestion_ops) {
	86	rcu_read_lock();
	87	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	88	if (try_module_get(ca->owner)) {
	89	icsk->icsk_ca_ops = ca;
	90	break;
	91	}
5f8ef48d	92
4d4d3d1e	93	/* fallback to next available */
317a76f9	94	}
4d4d3d1e	95	rcu_read_unlock();
317a76f9	96	}
317a76f9	97
946272f8	98	tcp_sk(sk)->prior_ssthresh = 0;
6687e988 ACM	99	if (icsk->icsk_ca_ops->init)
6687e988 ACM	100	icsk->icsk_ca_ops->init(sk);
317a76f9 SH	101	}
	102
	103	/* Manage refcounts on socket close. */
6687e988	104	void tcp_cleanup_congestion_control(struct sock *sk)
317a76f9	105	{
6687e988 ACM	106	struct inet_connection_sock *icsk = inet_csk(sk);
	107
	108	if (icsk->icsk_ca_ops->release)
	109	icsk->icsk_ca_ops->release(sk);
	110	module_put(icsk->icsk_ca_ops->owner);
317a76f9 SH	111	}
	112
	113	/* Used by sysctl to change default congestion control */
	114	int tcp_set_default_congestion_control(const char *name)
	115	{
	116	struct tcp_congestion_ops *ca;
	117	int ret = -ENOENT;
	118
	119	spin_lock(&tcp_cong_list_lock);
	120	ca = tcp_ca_find(name);
95a5afca	121	#ifdef CONFIG_MODULES
a8f80e8f	122	if (!ca && capable(CAP_NET_ADMIN)) {
317a76f9 SH	123	spin_unlock(&tcp_cong_list_lock);
	124
	125	request_module("tcp_%s", name);
	126	spin_lock(&tcp_cong_list_lock);
	127	ca = tcp_ca_find(name);
	128	}
	129	#endif
	130
	131	if (ca) {
164891aa	132	ca->flags \|= TCP_CONG_NON_RESTRICTED; /* default is always allowed */
317a76f9 SH	133	list_move(&ca->list, &tcp_cong_list);
	134	ret = 0;
	135	}
	136	spin_unlock(&tcp_cong_list_lock);
	137
	138	return ret;
	139	}
	140
b1736a71 SH	141	/* Set default value from kernel configuration at bootup */
	142	static int __init tcp_congestion_default(void)
	143	{
	144	return tcp_set_default_congestion_control(CONFIG_DEFAULT_TCP_CONG);
	145	}
	146	late_initcall(tcp_congestion_default);
	147
	148
3ff825b2 SH	149	/* Build string with list of available congestion control values */
	150	void tcp_get_available_congestion_control(char *buf, size_t maxlen)
	151	{
	152	struct tcp_congestion_ops *ca;
	153	size_t offs = 0;
	154
	155	rcu_read_lock();
	156	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
	157	offs += snprintf(buf + offs, maxlen - offs,
	158	"%s%s",
	159	offs == 0 ? "" : " ", ca->name);
	160
	161	}
	162	rcu_read_unlock();
	163	}
	164
317a76f9 SH	165	/* Get current default congestion control */
	166	void tcp_get_default_congestion_control(char *name)
	167	{
	168	struct tcp_congestion_ops *ca;
	169	/* We will always have reno... */
	170	BUG_ON(list_empty(&tcp_cong_list));
	171
	172	rcu_read_lock();
	173	ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
	174	strncpy(name, ca->name, TCP_CA_NAME_MAX);
	175	rcu_read_unlock();
	176	}
	177
ce7bc3bf SH	178	/* Built list of non-restricted congestion control values */
	179	void tcp_get_allowed_congestion_control(char *buf, size_t maxlen)
	180	{
	181	struct tcp_congestion_ops *ca;
	182	size_t offs = 0;
	183
	184	*buf = '\0';
	185	rcu_read_lock();
	186	list_for_each_entry_rcu(ca, &tcp_cong_list, list) {
164891aa	187	if (!(ca->flags & TCP_CONG_NON_RESTRICTED))
ce7bc3bf SH	188	continue;
	189	offs += snprintf(buf + offs, maxlen - offs,
	190	"%s%s",
	191	offs == 0 ? "" : " ", ca->name);
	192
	193	}
	194	rcu_read_unlock();
	195	}
	196
	197	/* Change list of non-restricted congestion control */
	198	int tcp_set_allowed_congestion_control(char *val)
	199	{
	200	struct tcp_congestion_ops *ca;
c34186ed	201	char saved_clone, clone, *name;
ce7bc3bf SH	202	int ret = 0;
ce7bc3bf SH	203
c34186ed	204	saved_clone = clone = kstrdup(val, GFP_USER);
ce7bc3bf SH	205	if (!clone)
	206	return -ENOMEM;
	207
	208	spin_lock(&tcp_cong_list_lock);
	209	/* pass 1 check for bad entries */
	210	while ((name = strsep(&clone, " ")) && *name) {
	211	ca = tcp_ca_find(name);
	212	if (!ca) {
	213	ret = -ENOENT;
	214	goto out;
	215	}
	216	}
	217
164891aa	218	/* pass 2 clear old values */
ce7bc3bf	219	list_for_each_entry_rcu(ca, &tcp_cong_list, list)
164891aa	220	ca->flags &= ~TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	221
	222	/* pass 3 mark as allowed */
	223	while ((name = strsep(&val, " ")) && *name) {
	224	ca = tcp_ca_find(name);
	225	WARN_ON(!ca);
	226	if (ca)
164891aa	227	ca->flags \|= TCP_CONG_NON_RESTRICTED;
ce7bc3bf SH	228	}
	229	out:
	230	spin_unlock(&tcp_cong_list_lock);
c34186ed	231	kfree(saved_clone);
ce7bc3bf SH	232
	233	return ret;
	234	}
	235
	236
5f8ef48d	237	/* Change congestion control for socket */
6687e988	238	int tcp_set_congestion_control(struct sock sk, const char name)
5f8ef48d	239	{
6687e988	240	struct inet_connection_sock *icsk = inet_csk(sk);
5f8ef48d SH	241	struct tcp_congestion_ops *ca;
	242	int err = 0;
	243
	244	rcu_read_lock();
	245	ca = tcp_ca_find(name);
4d4d3d1e	246
35bfbc94	247	/* no change asking for existing value */
6687e988	248	if (ca == icsk->icsk_ca_ops)
5f8ef48d SH	249	goto out;
5f8ef48d SH	250
95a5afca	251	#ifdef CONFIG_MODULES
35bfbc94	252	/* not found attempt to autoload module */
a8f80e8f	253	if (!ca && capable(CAP_NET_ADMIN)) {
35bfbc94 SH	254	rcu_read_unlock();
	255	request_module("tcp_%s", name);
	256	rcu_read_lock();
	257	ca = tcp_ca_find(name);
	258	}
	259	#endif
5f8ef48d SH	260	if (!ca)
	261	err = -ENOENT;
	262
52e804c6 EB	263	else if (!((ca->flags & TCP_CONG_NON_RESTRICTED) \|\|
52e804c6 EB	264	ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)))
ce7bc3bf SH	265	err = -EPERM;
ce7bc3bf SH	266
5f8ef48d SH	267	else if (!try_module_get(ca->owner))
	268	err = -EBUSY;
	269
	270	else {
6687e988 ACM	271	tcp_cleanup_congestion_control(sk);
6687e988 ACM	272	icsk->icsk_ca_ops = ca;
4d4d3d1e SH	273
4d4d3d1e SH	274	if (sk->sk_state != TCP_CLOSE && icsk->icsk_ca_ops->init)
6687e988	275	icsk->icsk_ca_ops->init(sk);
5f8ef48d SH	276	}
	277	out:
	278	rcu_read_unlock();
	279	return err;
	280	}
	281
cea14e0e IJ	282	/* RFC2861 Check whether we are limited by application or congestion window
	283	* This is the inverse of cwnd check in tcp_tso_should_defer
	284	*/
a2a385d6	285	bool tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
cea14e0e IJ	286	{
	287	const struct tcp_sock *tp = tcp_sk(sk);
	288	u32 left;
	289
	290	if (in_flight >= tp->snd_cwnd)
a2a385d6	291	return true;
cea14e0e	292
cea14e0e	293	left = tp->snd_cwnd - in_flight;
ce447eb9	294	if (sk_can_gso(sk) &&
246eb2af	295	left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd &&
1485348d BH	296	left * tp->mss_cache < sk->sk_gso_max_size &&
1485348d BH	297	left < sk->sk_gso_max_segs)
a2a385d6	298	return true;
6b5a5c0d	299	return left <= tcp_max_tso_deferred_mss(tp);
cea14e0e IJ	300	}
cea14e0e IJ	301	EXPORT_SYMBOL_GPL(tcp_is_cwnd_limited);
40efc6fa SH	302
40efc6fa SH	303	/*
a02ba041 SH	304	* Slow start is used when congestion window is less than slow start
	305	* threshold. This version implements the basic RFC2581 version
	306	* and optionally supports:
	307	* RFC3742 Limited Slow Start - growth limited to max_ssthresh
	308	* RFC3465 Appropriate Byte Counting - growth limited by bytes acknowledged
40efc6fa SH	309	*/
	310	void tcp_slow_start(struct tcp_sock *tp)
	311	{
a02ba041	312	int cnt; /* increase in packets */
9dc27415	313	unsigned int delta = 0;
973ec449 ED	314	u32 snd_cwnd = tp->snd_cwnd;
	315
	316	if (unlikely(!snd_cwnd)) {
	317	pr_err_once("snd_cwnd is nul, please report this bug.\n");
	318	snd_cwnd = 1U;
	319	}
a02ba041	320
a02ba041 SH	321	if (sysctl_tcp_max_ssthresh > 0 && tp->snd_cwnd > sysctl_tcp_max_ssthresh)
a02ba041 SH	322	cnt = sysctl_tcp_max_ssthresh >> 1; /* limited slow start */
886236c1	323	else
973ec449	324	cnt = snd_cwnd; /* exponential increase */
886236c1	325
886236c1	326	tp->snd_cwnd_cnt += cnt;
973ec449 ED	327	while (tp->snd_cwnd_cnt >= snd_cwnd) {
973ec449 ED	328	tp->snd_cwnd_cnt -= snd_cwnd;
9dc27415	329	delta++;
886236c1	330	}
973ec449	331	tp->snd_cwnd = min(snd_cwnd + delta, tp->snd_cwnd_clamp);
40efc6fa SH	332	}
	333	EXPORT_SYMBOL_GPL(tcp_slow_start);
	334
758ce5c8 IJ	335	/* In theory this is tp->snd_cwnd += 1 / tp->snd_cwnd (or alternative w) */
	336	void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w)
	337	{
	338	if (tp->snd_cwnd_cnt >= w) {
	339	if (tp->snd_cwnd < tp->snd_cwnd_clamp)
	340	tp->snd_cwnd++;
	341	tp->snd_cwnd_cnt = 0;
	342	} else {
	343	tp->snd_cwnd_cnt++;
	344	}
	345	}
	346	EXPORT_SYMBOL_GPL(tcp_cong_avoid_ai);
	347
317a76f9 SH	348	/*
	349	* TCP Reno congestion control
	350	* This is special case used for fallback as well.
	351	*/
	352	/* This is Jacobson's slow start and congestion avoidance.
	353	* SIGCOMM '88, p. 328.
	354	*/
c3a05c60	355	void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
317a76f9	356	{
6687e988 ACM	357	struct tcp_sock *tp = tcp_sk(sk);
6687e988 ACM	358
f4805ede	359	if (!tcp_is_cwnd_limited(sk, in_flight))
317a76f9 SH	360	return;
317a76f9 SH	361
7faffa1c	362	/* In "safe" area, increase. */
e905a9ed	363	if (tp->snd_cwnd <= tp->snd_ssthresh)
7faffa1c	364	tcp_slow_start(tp);
e905a9ed	365	/* In dangerous area, increase slowly. */
ca2eb567	366	else
758ce5c8	367	tcp_cong_avoid_ai(tp, tp->snd_cwnd);
317a76f9 SH	368	}
	369	EXPORT_SYMBOL_GPL(tcp_reno_cong_avoid);
	370
	371	/* Slow start threshold is half the congestion window (min 2) */
6687e988	372	u32 tcp_reno_ssthresh(struct sock *sk)
317a76f9	373	{
6687e988	374	const struct tcp_sock *tp = tcp_sk(sk);
317a76f9 SH	375	return max(tp->snd_cwnd >> 1U, 2U);
	376	}
	377	EXPORT_SYMBOL_GPL(tcp_reno_ssthresh);
	378
72dc5b92 SH	379	/* Lower bound on congestion window with halving. */
72dc5b92 SH	380	u32 tcp_reno_min_cwnd(const struct sock *sk)
317a76f9	381	{
6687e988	382	const struct tcp_sock *tp = tcp_sk(sk);
317a76f9 SH	383	return tp->snd_ssthresh/2;
	384	}
	385	EXPORT_SYMBOL_GPL(tcp_reno_min_cwnd);
	386
	387	struct tcp_congestion_ops tcp_reno = {
164891aa	388	.flags = TCP_CONG_NON_RESTRICTED,
317a76f9 SH	389	.name = "reno",
	390	.owner = THIS_MODULE,
	391	.ssthresh = tcp_reno_ssthresh,
	392	.cong_avoid = tcp_reno_cong_avoid,
	393	.min_cwnd = tcp_reno_min_cwnd,
	394	};
	395
5f8ef48d SH	396	/* Initial congestion control used (until SYN)
	397	* really reno under another name so we can tell difference
	398	* during tcp_set_default_congestion_control
	399	*/
	400	struct tcp_congestion_ops tcp_init_congestion_ops = {
	401	.name = "",
	402	.owner = THIS_MODULE,
	403	.ssthresh = tcp_reno_ssthresh,
	404	.cong_avoid = tcp_reno_cong_avoid,
	405	.min_cwnd = tcp_reno_min_cwnd,
	406	};
	407	EXPORT_SYMBOL_GPL(tcp_init_congestion_ops);