tcp: Namespace-ify sysctl_tcp_default_congestion_control

Make default TCP default congestion control to a per namespace
value. This changes default congestion control to a pointer to congestion ops
(rather than implicit as first element of available lsit).

The congestion control setting of new namespaces is inherited
from the current setting of the root namespace.

Signed-off-by: Stephen Hemminger <[e-mail]>
Reviewed-by: Eric Dumazet <[e-mail]>
Signed-off-by: David S. Miller <[e-mail]>

include/net/netns/ipv4.h

···
 	struct inet_timewait_death_row tcp_death_row;
 	int sysctl_max_syn_backlog;
 	int sysctl_tcp_fastopen;
+	const struct tcp_congestion_ops __rcu  *tcp_congestion_control;
 	struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
 	spinlock_t tcp_fastopen_ctx_lock;
 	unsigned int sysctl_tcp_fastopen_blackhole_timeout;

include/net/tcp.h

···
 void tcp_assign_congestion_control(struct sock *sk);
 void tcp_init_congestion_control(struct sock *sk);
 void tcp_cleanup_congestion_control(struct sock *sk);
-int tcp_set_default_congestion_control(const char *name);
-void tcp_get_default_congestion_control(char *name);
+int tcp_set_default_congestion_control(struct net *net, const char *name);
+void tcp_get_default_congestion_control(struct net *net, char *name);
 void tcp_get_available_congestion_control(char *buf, size_t len);
 void tcp_get_allowed_congestion_control(char *buf, size_t len);
 int tcp_set_allowed_congestion_control(char *allowed);
···
 extern struct tcp_congestion_ops tcp_reno;
 
 struct tcp_congestion_ops *tcp_ca_find_key(u32 key);
-u32 tcp_ca_get_key_by_name(const char *name, bool *ecn_ca);
+u32 tcp_ca_get_key_by_name(struct net *net, const char *name, bool *ecn_ca);
 #ifdef CONFIG_INET
 char *tcp_ca_get_name_by_key(u32 key, char *buffer);
 #else

net/ipv4/fib_semantics.c

···
 			bool ecn_ca = false;
 
 			nla_strlcpy(tmp, nla, sizeof(tmp));
-			val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
+			val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
 		} else {
 			val = nla_get_u32(nla);
 		}
···
 			char tmp[TCP_CA_NAME_MAX];
 
 			nla_strlcpy(tmp, nla, sizeof(tmp));
-			val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
+			val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
 			if (val == TCP_CA_UNSPEC)
 				return -EINVAL;
 		} else {

net/ipv4/sysctl_net_ipv4.c

···
 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 				       void __user *buffer, size_t *lenp, loff_t *ppos)
 {
+	struct net *net = container_of(ctl->data, struct net,
+				       ipv4.tcp_congestion_control);
 	char val[TCP_CA_NAME_MAX];
 	struct ctl_table tbl = {
 		.data = val,
···
 	};
 	int ret;
 
-	tcp_get_default_congestion_control(val);
+	tcp_get_default_congestion_control(net, val);
 
 	ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 	if (write && ret == 0)
-		ret = tcp_set_default_congestion_control(val);
+		ret = tcp_set_default_congestion_control(net, val);
 	return ret;
 }
 
···
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec
-	},
-	{
-		.procname	= "tcp_congestion_control",
-		.mode		= 0644,
-		.maxlen		= TCP_CA_NAME_MAX,
-		.proc_handler	= proc_tcp_congestion_control,
 	},
 #ifdef CONFIG_NETLABEL
 	{
···
 		.extra1		= &one
 	},
 #endif
+	{
+		.procname	= "tcp_congestion_control",
+		.data		= &init_net.ipv4.tcp_congestion_control,
+		.mode		= 0644,
+		.maxlen		= TCP_CA_NAME_MAX,
+		.proc_handler	= proc_tcp_congestion_control,
+	},
 	{
 		.procname	= "tcp_keepalive_time",
 		.data		= &init_net.ipv4.sysctl_tcp_keepalive_time,

net/ipv4/tcp_cong.c

···
 }
 
 /* Must be called with rcu lock held */
-static const struct tcp_congestion_ops *__tcp_ca_find_autoload(const char *name)
+static struct tcp_congestion_ops *tcp_ca_find_autoload(struct net *net,
+						       const char *name)
 {
-	const struct tcp_congestion_ops *ca = tcp_ca_find(name);
+	struct tcp_congestion_ops *ca = tcp_ca_find(name);
+
 #ifdef CONFIG_MODULES
 	if (!ca && capable(CAP_NET_ADMIN)) {
 		rcu_read_unlock();
···
 }
 EXPORT_SYMBOL_GPL(tcp_unregister_congestion_control);
 
-u32 tcp_ca_get_key_by_name(const char *name, bool *ecn_ca)
+u32 tcp_ca_get_key_by_name(struct net *net, const char *name, bool *ecn_ca)
 {
 	const struct tcp_congestion_ops *ca;
 	u32 key = TCP_CA_UNSPEC;
···
 	might_sleep();
 
 	rcu_read_lock();
-	ca = __tcp_ca_find_autoload(name);
+	ca = tcp_ca_find_autoload(net, name);
 	if (ca) {
 		key = ca->key;
 		*ecn_ca = ca->flags & TCP_CONG_NEEDS_ECN;
···
 /* Assign choice of congestion control. */
 void tcp_assign_congestion_control(struct sock *sk)
 {
+	struct net *net = sock_net(sk);
 	struct inet_connection_sock *icsk = inet_csk(sk);
-	struct tcp_congestion_ops *ca;
+	const 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:
+	ca = rcu_dereference(net->ipv4.tcp_congestion_control);
+	if (unlikely(!try_module_get(ca->owner)))
+		ca = &tcp_reno;
+	icsk->icsk_ca_ops = ca;
 	rcu_read_unlock();
-	memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv));
 
+	memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv));
 	if (ca->flags & TCP_CONG_NEEDS_ECN)
 		INET_ECN_xmit(sk);
 	else
···
 }
 
 /* Used by sysctl to change default congestion control */
-int tcp_set_default_congestion_control(const char *name)
+int tcp_set_default_congestion_control(struct net *net, const char *name)
 {
 	struct tcp_congestion_ops *ca;
-	int ret = -ENOENT;
+	const struct tcp_congestion_ops *prev;
+	int ret;
 
-	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);
+	rcu_read_lock();
+	ca = tcp_ca_find_autoload(net, name);
+	if (!ca) {
+		ret = -ENOENT;
+	} else if (!try_module_get(ca->owner)) {
+		ret = -EBUSY;
+	} else {
+		prev = xchg(&net->ipv4.tcp_congestion_control, ca);
+		if (prev)
+			module_put(prev->owner);
 
-		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);
+		ca->flags |= TCP_CONG_NON_RESTRICTED;
 		ret = 0;
 	}
-	spin_unlock(&tcp_cong_list_lock);
+	rcu_read_unlock();
 
 	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);
+	return tcp_set_default_congestion_control(&init_net,
+						  CONFIG_DEFAULT_TCP_CONG);
 }
 late_initcall(tcp_congestion_default);
 
···
 }
 
 /* Get current default congestion control */
-void tcp_get_default_congestion_control(char *name)
+void tcp_get_default_congestion_control(struct net *net, char *name)
 {
-	struct tcp_congestion_ops *ca;
-	/* We will always have reno... */
-	BUG_ON(list_empty(&tcp_cong_list));
+	const struct tcp_congestion_ops *ca;
 
 	rcu_read_lock();
-	ca = list_entry(tcp_cong_list.next, struct tcp_congestion_ops, list);
+	ca = rcu_dereference(net->ipv4.tcp_congestion_control);
 	strncpy(name, ca->name, TCP_CA_NAME_MAX);
 	rcu_read_unlock();
 }
···
 	if (!load)
 		ca = tcp_ca_find(name);
 	else
-		ca = __tcp_ca_find_autoload(name);
+		ca = tcp_ca_find_autoload(sock_net(sk), name);
+
 	/* No change asking for existing value */
 	if (ca == icsk->icsk_ca_ops) {
 		icsk->icsk_ca_setsockopt = 1;
 		goto out;
 	}
+
 	if (!ca) {
 		err = -ENOENT;
 	} else if (!load) {

net/ipv4/tcp_ipv4.c

···
 {
 	int cpu;
 
+	module_put(net->ipv4.tcp_congestion_control->owner);
+
 	for_each_possible_cpu(cpu)
 		inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv4.tcp_sk, cpu));
 	free_percpu(net->ipv4.tcp_sk);
···
 	spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock);
 	net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60;
 	atomic_set(&net->ipv4.tfo_active_disable_times, 0);
+
+	/* Reno is always built in */
+	if (!net_eq(net, &init_net) &&
+	    try_module_get(init_net.ipv4.tcp_congestion_control->owner))
+		net->ipv4.tcp_congestion_control = init_net.ipv4.tcp_congestion_control;
+	else
+		net->ipv4.tcp_congestion_control = &tcp_reno;
 
 	return 0;
 fail:

net/ipv6/route.c

···
 static int ip6_convert_metrics(struct mx6_config *mxc,
 			       const struct fib6_config *cfg)
 {
+	struct net *net = cfg->fc_nlinfo.nl_net;
 	bool ecn_ca = false;
 	struct nlattr *nla;
 	int remaining;
···
 			char tmp[TCP_CA_NAME_MAX];
 
 			nla_strlcpy(tmp, nla, sizeof(tmp));
-			val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
+			val = tcp_ca_get_key_by_name(net, tmp, &ecn_ca);
 			if (val == TCP_CA_UNSPEC)
 				goto err;
 		} else {