VIDEOS 》 Linux Kernel TCP Congestion Control Algorithms

Refer:
TCP congestion control :: https://en.wikipedia.org/wiki/TCP_congestion_control
-----
Linux Kernel Source:
Linux Kernel - IPv4 Stack /net/ipv4 :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4
TCP CUBIC: Binary Increase Congestion control for TCP - /net/ipv4/tcp_cubic.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c
TCP-HYBLA Congestion control algorithm - /net/ipv4/tcp_hybla.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_hybla.c
* A TCP Enhancement for Heterogeneous Networks
TCP Illinois congestion control - /net/ipv4/tcp_illinois.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_illinois.c
TCP Westwood+: end-to-end bandwidth estimation for TCP - /net/ipv4/tcp_westwood.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_westwood.c
TCP Vegas congestion control - /net/ipv4/tcp_vegas.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_vegas.c
TCP Veno congestion control - /net/ipv4/tcp_veno.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_veno.c
H-TCP congestion control: TCP for high-speed and long-distance networks - /net/ipv4/tcp_htcp.c :: http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_htcp.c

Also refer the Kernel Architecture difference between a standard Linux System vs Android OS, discussed in this video:
Kernel Architecture - Generic Linux System vs Android - The Linux Channel

Refer:
Wiki:
CUBIC TCP - https://en.wikipedia.org/wiki/CUBIC_TCP
BIC TCP - https://en.wikipedia.org/wiki/BIC_TCP
LFN: long fat networks - https://en.wikipedia.org/wiki/Bandwidth-delay_product
Bandwidth-delay product - https://en.wikipedia.org/wiki/Bandwidth-delay_product
-----
Linux Kernel Source:
TCP CUBIC: Binary Increase Congestion control for TCP v2.3 - http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c
static struct tcp_congestion_ops cubictcp - http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c#L457
static int __init cubictcp_register(void) - http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c#L469
static void __exit cubictcp_unregister(void) - http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c#L504
tcp_register_congestion_control(), tcp_unregister_congestion_control() - http://elixir.free-electrons.com/linux/latest/source/include/net/tcp.h#L1029
Pluggable TCP congestion control support - http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cong.c
struct tcp_congestion_ops data-structure - http://elixir.free-electrons.com/linux/latest/source/include/net/tcp.h#L989

Here is the struct tcp_congestion_ops data-structure (/include/net/tcp.h) from the Kernel-source version 4.14 for quick reference:

struct tcp_congestion_ops {
	struct list_head	list;
	u32 key;
	u32 flags;

	/* initialize private data (optional) */
	void (*init)(struct sock *sk);
	/* cleanup private data  (optional) */
	void (*release)(struct sock *sk);

	/* return slow start threshold (required) */
	u32 (*ssthresh)(struct sock *sk);
	/* do new cwnd calculation (required) */
	void (*cong_avoid)(struct sock *sk, u32 ack, u32 acked);
	/* call before changing ca_state (optional) */
	void (*set_state)(struct sock *sk, u8 new_state);
	/* call when cwnd event occurs (optional) */
	void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
	/* call when ack arrives (optional) */
	void (*in_ack_event)(struct sock *sk, u32 flags);
	/* new value of cwnd after loss (required) */
	u32  (*undo_cwnd)(struct sock *sk);
	/* hook for packet ack accounting (optional) */
	void (*pkts_acked)(struct sock *sk, const struct ack_sample *sample);
	/* suggest number of segments for each skb to transmit (optional) */
	u32 (*tso_segs_goal)(struct sock *sk);
	/* returns the multiplier used in tcp_sndbuf_expand (optional) */
	u32 (*sndbuf_expand)(struct sock *sk);
	/* call when packets are delivered to update cwnd and pacing rate,
	 * after all the ca_state processing. (optional)
	 */
	void (*cong_control)(struct sock *sk, const struct rate_sample *rs);
	/* get info for inet_diag (optional) */
	size_t (*get_info)(struct sock *sk, u32 ext, int *attr,
			   union tcp_cc_info *info);

	char 		name[TCP_CA_NAME_MAX];
	struct module 	*owner;
};

Here is the tcp_register_congestion_control(), tcp_unregister_congestion_control() pluggable congestion control registration APIs (/net/ipv4/tcp_cong.c) from the Kernel-source version 4.14 for quick reference:

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

	ca->key = jhash(ca->name, sizeof(ca->name), strlen(ca->name));

	spin_lock(&tcp_cong_list_lock);
	if (ca->key == TCP_CA_UNSPEC || tcp_ca_find_key(ca->key)) {
		pr_notice("%s already registered or non-unique key\n",
			  ca->name);
		ret = -EEXIST;
	} else {
		list_add_tail_rcu(&ca->list, &tcp_cong_list);
		pr_debug("%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);

	/* Wait for outstanding readers to complete before the
	 * module gets removed entirely.
	 *
	 * A try_module_get() should fail by now as our module is
	 * in "going" state since no refs are held anymore and
	 * module_exit() handler being called.
	 */
	synchronize_rcu();
}
EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);

Here is the struct tcp_congestion_ops cubictcp data-structure instance, cubictcp_register(), cubictcp_unregister() APIs (/net/ipv4/tcp_cubic.c) from the Kernel-source version 4.14 for quick reference:

static struct tcp_congestion_ops cubictcp __read_mostly = {
	.init		= bictcp_init,
	.ssthresh	= bictcp_recalc_ssthresh,
	.cong_avoid	= bictcp_cong_avoid,
	.set_state	= bictcp_state,
	.undo_cwnd	= tcp_reno_undo_cwnd,
	.cwnd_event	= bictcp_cwnd_event,
	.pkts_acked     = bictcp_acked,
	.owner		= THIS_MODULE,
	.name		= "cubic",
};

static int __init cubictcp_register(void)
{
	BUILD_BUG_ON(sizeof(struct bictcp) > ICSK_CA_PRIV_SIZE);

	/* Precompute a bunch of the scaling factors that are used per-packet
	 * based on SRTT of 100ms
	 */

	beta_scale = 8*(BICTCP_BETA_SCALE+beta) / 3
		/ (BICTCP_BETA_SCALE - beta);

	cube_rtt_scale = (bic_scale * 10);	/* 1024*c/rtt */

	/* calculate the "K" for (wmax-cwnd) = c/rtt * K^3
	 *  so K = cubic_root( (wmax-cwnd)*rtt/c )
	 * the unit of K is bictcp_HZ=2^10, not HZ
	 *
	 *  c = bic_scale >> 10
	 *  rtt = 100ms
	 *
	 * the following code has been designed and tested for
	 * cwnd < 1 million packets
	 * RTT < 100 seconds
	 * HZ < 1,000,00  (corresponding to 10 nano-second)
	 */

	/* 1/c * 2^2*bictcp_HZ * srtt */
	cube_factor = 1ull << (10+3*BICTCP_HZ); /* 2^40 */

	/* divide by bic_scale and by constant Srtt (100ms) */
	do_div(cube_factor, bic_scale * 10);

	return tcp_register_congestion_control(&cubictcp);
}

static void __exit cubictcp_unregister(void)
{
	tcp_unregister_congestion_control(&cubictcp);
}

module_init(cubictcp_register);
module_exit(cubictcp_unregister);

MODULE_AUTHOR("Sangtae Ha, Stephen Hemminger");
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("CUBIC TCP");
MODULE_VERSION("2.3");

Here is the LIST_HEAD(tcp_cong_list) linked list, tcp_ca_find() API (/net/ipv4/tcp_cong.c) from the Kernel-source version 4.14 for quick reference:

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;
}

Here is the tcp_init_congestion_control(), tcp_reinit_congestion_control() APIs (/net/ipv4/tcp_cong.c) from the Kernel-source version 4.14 for quick reference:

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

	tcp_sk(sk)->prior_ssthresh = 0;
	if (icsk->icsk_ca_ops->init)
		icsk->icsk_ca_ops->init(sk);
	if (tcp_ca_needs_ecn(sk))
		INET_ECN_xmit(sk);
	else
		INET_ECN_dontxmit(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;
	icsk->icsk_ca_setsockopt = 1;
	memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv));

	if (sk->sk_state != TCP_CLOSE)
		tcp_init_congestion_control(sk);
}

Here is the tcp_set_default_congestion_control(), tcp_get_available_congestion_control(), tcp_set_allowed_congestion_control() APIs (/net/ipv4/tcp_cong.c) from the Kernel-source version 4.14 for quick reference:

/* 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();
}

/* 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;
}

Suggested Topics:

Video Episodes :: Linux Kernel programming