[linux.git] / net / ipv6 / inet6_hashtables.c

/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		Generic INET6 transport hashtables
 *
 * Authors:	Lotsa people, from code originally in tcp, generalised here
 * 		by Arnaldo Carvalho de Melo <[email protected]>
 *
 *	This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 */

#include <linux/module.h>
#include <linux/random.h>

#include <net/inet_connection_sock.h>
#include <net/inet_hashtables.h>
#include <net/inet6_hashtables.h>
#include <net/secure_seq.h>
#include <net/ip.h>

int __inet6_hash(struct sock *sk, struct inet_timewait_sock *tw)
{
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
	int twrefcnt = 0;

	WARN_ON(!sk_unhashed(sk));

	if (sk->sk_state == TCP_LISTEN) {
		struct inet_listen_hashbucket *ilb;

		ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
		spin_lock(&ilb->lock);
		__sk_nulls_add_node_rcu(sk, &ilb->head);
		spin_unlock(&ilb->lock);
	} else {
		unsigned int hash;
		struct hlist_nulls_head *list;
		spinlock_t *lock;

		sk->sk_hash = hash = inet6_sk_ehashfn(sk);
		list = &inet_ehash_bucket(hashinfo, hash)->chain;
		lock = inet_ehash_lockp(hashinfo, hash);
		spin_lock(lock);
		__sk_nulls_add_node_rcu(sk, list);
		if (tw) {
			WARN_ON(sk->sk_hash != tw->tw_hash);
			twrefcnt = inet_twsk_unhash(tw);
		}
		spin_unlock(lock);
	}

	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
	return twrefcnt;
}
EXPORT_SYMBOL(__inet6_hash);

/*
 * Sockets in TCP_CLOSE state are _always_ taken out of the hash, so
 * we need not check it for TCP lookups anymore, thanks Alexey. -DaveM
 *
 * The sockhash lock must be held as a reader here.
 */
struct sock *__inet6_lookup_established(struct net *net,
					struct inet_hashinfo *hashinfo,
					   const struct in6_addr *saddr,
					   const __be16 sport,
					   const struct in6_addr *daddr,
					   const u16 hnum,
					   const int dif)
{
	struct sock *sk;
	const struct hlist_nulls_node *node;
	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
	unsigned int hash = inet6_ehashfn(net, daddr, hnum, saddr, sport);
	unsigned int slot = hash & hashinfo->ehash_mask;
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];


	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
		/* For IPV6 do the cheaper port and family tests first. */
		if (INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
				goto begintw;
			if (!INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
				sock_put(sk);
				goto begin;
			}
		goto out;
		}
	}
	if (get_nulls_value(node) != slot)
		goto begin;

begintw:
	/* Must check for a TIME_WAIT'er before going to listener hash. */
	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
		if (INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
				sk = NULL;
				goto out;
			}
			if (!INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
				sock_put(sk);
				goto begintw;
			}
			goto out;
		}
	}
	if (get_nulls_value(node) != slot)
		goto begintw;
	sk = NULL;
out:
	rcu_read_unlock();
	return sk;
}
EXPORT_SYMBOL(__inet6_lookup_established);

static inline int compute_score(struct sock *sk, struct net *net,
				const unsigned short hnum,
				const struct in6_addr *daddr,
				const int dif)
{
	int score = -1;

	if (net_eq(sock_net(sk), net) && inet_sk(sk)->inet_num == hnum &&
	    sk->sk_family == PF_INET6) {
		const struct ipv6_pinfo *np = inet6_sk(sk);

		score = 1;
		if (!ipv6_addr_any(&np->rcv_saddr)) {
			if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
				return -1;
			score++;
		}
		if (sk->sk_bound_dev_if) {
			if (sk->sk_bound_dev_if != dif)
				return -1;
			score++;
		}
	}
	return score;
}

struct sock *inet6_lookup_listener(struct net *net,
		struct inet_hashinfo *hashinfo, const struct in6_addr *daddr,
		const unsigned short hnum, const int dif)
{
	struct sock *sk;
	const struct hlist_nulls_node *node;
	struct sock *result;
	int score, hiscore;
	unsigned int hash = inet_lhashfn(net, hnum);
	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];

	rcu_read_lock();
begin:
	result = NULL;
	hiscore = -1;
	sk_nulls_for_each(sk, node, &ilb->head) {
		score = compute_score(sk, net, hnum, daddr, dif);
		if (score > hiscore) {
			hiscore = score;
			result = sk;
		}
	}
	/*
	 * if the nulls value we got at the end of this lookup is
	 * not the expected one, we must restart lookup.
	 * We probably met an item that was moved to another chain.
	 */
	if (get_nulls_value(node) != hash + LISTENING_NULLS_BASE)
		goto begin;
	if (result) {
		if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
			result = NULL;
		else if (unlikely(compute_score(result, net, hnum, daddr,
				  dif) < hiscore)) {
			sock_put(result);
			goto begin;
		}
	}
	rcu_read_unlock();
	return result;
}

EXPORT_SYMBOL_GPL(inet6_lookup_listener);

struct sock *inet6_lookup(struct net *net, struct inet_hashinfo *hashinfo,
			  const struct in6_addr *saddr, const __be16 sport,
			  const struct in6_addr *daddr, const __be16 dport,
			  const int dif)
{
	struct sock *sk;

	local_bh_disable();
	sk = __inet6_lookup(net, hashinfo, saddr, sport, daddr, ntohs(dport), dif);
	local_bh_enable();

	return sk;
}

EXPORT_SYMBOL_GPL(inet6_lookup);

static int __inet6_check_established(struct inet_timewait_death_row *death_row,
				     struct sock *sk, const __u16 lport,
				     struct inet_timewait_sock **twp)
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
	struct inet_sock *inet = inet_sk(sk);
	const struct ipv6_pinfo *np = inet6_sk(sk);
	const struct in6_addr *daddr = &np->rcv_saddr;
	const struct in6_addr *saddr = &np->daddr;
	const int dif = sk->sk_bound_dev_if;
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
	struct net *net = sock_net(sk);
	const unsigned int hash = inet6_ehashfn(net, daddr, lport, saddr,
						inet->inet_dport);
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
	struct sock *sk2;
	const struct hlist_nulls_node *node;
	struct inet_timewait_sock *tw;
	int twrefcnt = 0;

	spin_lock(lock);

	/* Check TIME-WAIT sockets first. */
	sk_nulls_for_each(sk2, node, &head->twchain) {
		tw = inet_twsk(sk2);

		if (INET6_TW_MATCH(sk2, net, hash, saddr, daddr, ports, dif)) {
			if (twsk_unique(sk, sk2, twp))
				goto unique;
			else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
	sk_nulls_for_each(sk2, node, &head->chain) {
		if (INET6_MATCH(sk2, net, hash, saddr, daddr, ports, dif))
			goto not_unique;
	}

unique:
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
	sk->sk_hash = hash;
	WARN_ON(!sk_unhashed(sk));
	__sk_nulls_add_node_rcu(sk, &head->chain);
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
	spin_unlock(lock);
	if (twrefcnt)
		inet_twsk_put(tw);
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);

	if (twp) {
		*twp = tw;
	} else if (tw) {
		/* Silly. Should hash-dance instead... */
		inet_twsk_deschedule(tw, death_row);

		inet_twsk_put(tw);
	}
	return 0;

not_unique:
	spin_unlock(lock);
	return -EADDRNOTAVAIL;
}

static inline u32 inet6_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
	const struct ipv6_pinfo *np = inet6_sk(sk);
	return secure_ipv6_port_ephemeral(np->rcv_saddr.s6_addr32,
					  np->daddr.s6_addr32,
					  inet->inet_dport);
}

int inet6_hash_connect(struct inet_timewait_death_row *death_row,
		       struct sock *sk)
{
	return __inet_hash_connect(death_row, sk, inet6_sk_port_offset(sk),
			__inet6_check_established, __inet6_hash);
}

EXPORT_SYMBOL_GPL(inet6_hash_connect);
Commit	Line	Data
5324a040 ACM	1	/*
	2	* INET An implementation of the TCP/IP protocol suite for the LINUX
	3	* operating system. INET is implemented using the BSD Socket
	4	* interface as the means of communication with the user level.
	5	*
	6	* Generic INET6 transport hashtables
	7	*
d8313f5c ACM	8	* Authors: Lotsa people, from code originally in tcp, generalised here
d8313f5c ACM	9	* by Arnaldo Carvalho de Melo <[email protected]>
5324a040 ACM	10	*
	11	* This program is free software; you can redistribute it and/or
	12	* modify it under the terms of the GNU General Public License
	13	* as published by the Free Software Foundation; either version
	14	* 2 of the License, or (at your option) any later version.
	15	*/
	16
5324a040	17	#include <linux/module.h>
d8313f5c	18	#include <linux/random.h>
5324a040 ACM	19
	20	#include <net/inet_connection_sock.h>
	21	#include <net/inet_hashtables.h>
	22	#include <net/inet6_hashtables.h>
6e5714ea	23	#include <net/secure_seq.h>
d8313f5c	24	#include <net/ip.h>
5324a040	25
9327f705	26	int __inet6_hash(struct sock sk, struct inet_timewait_sock tw)
b1a7ffcb	27	{
39d8cda7	28	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
9327f705	29	int twrefcnt = 0;
b1a7ffcb	30
547b792c	31	WARN_ON(!sk_unhashed(sk));
b1a7ffcb DV	32
b1a7ffcb DV	33	if (sk->sk_state == TCP_LISTEN) {
5caea4ea	34	struct inet_listen_hashbucket *ilb;
3ab5aee7	35
5caea4ea ED	36	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
5caea4ea ED	37	spin_lock(&ilb->lock);
c25eb3bf	38	__sk_nulls_add_node_rcu(sk, &ilb->head);
5caea4ea	39	spin_unlock(&ilb->lock);
b1a7ffcb DV	40	} else {
b1a7ffcb DV	41	unsigned int hash;
3ab5aee7	42	struct hlist_nulls_head *list;
9db66bdc	43	spinlock_t *lock;
3ab5aee7	44
b1a7ffcb	45	sk->sk_hash = hash = inet6_sk_ehashfn(sk);
230140cf ED	46	list = &inet_ehash_bucket(hashinfo, hash)->chain;
230140cf ED	47	lock = inet_ehash_lockp(hashinfo, hash);
9db66bdc	48	spin_lock(lock);
3ab5aee7	49	__sk_nulls_add_node_rcu(sk, list);
9327f705 ED	50	if (tw) {
	51	WARN_ON(sk->sk_hash != tw->tw_hash);
	52	twrefcnt = inet_twsk_unhash(tw);
	53	}
9db66bdc	54	spin_unlock(lock);
b1a7ffcb DV	55	}
b1a7ffcb DV	56
c29a0bc4	57	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
9327f705	58	return twrefcnt;
b1a7ffcb DV	59	}
	60	EXPORT_SYMBOL(__inet6_hash);
	61
	62	/*
	63	* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so
	64	* we need not check it for TCP lookups anymore, thanks Alexey. -DaveM
	65	*
	66	* The sockhash lock must be held as a reader here.
	67	*/
d86e0dac PE	68	struct sock __inet6_lookup_established(struct net net,
d86e0dac PE	69	struct inet_hashinfo *hashinfo,
b1a7ffcb	70	const struct in6_addr *saddr,
d2ecd9cc	71	const __be16 sport,
b1a7ffcb DV	72	const struct in6_addr *daddr,
	73	const u16 hnum,
	74	const int dif)
	75	{
	76	struct sock *sk;
3ab5aee7	77	const struct hlist_nulls_node *node;
4f765d84	78	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
b1a7ffcb DV	79	/* Optimize here for direct hit, only listening connections can
	80	* have wildcards anyways.
	81	*/
33de014c	82	unsigned int hash = inet6_ehashfn(net, daddr, hnum, saddr, sport);
f373b53b	83	unsigned int slot = hash & hashinfo->ehash_mask;
3ab5aee7	84	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
b1a7ffcb	85
3ab5aee7 ED	86
	87	rcu_read_lock();
	88	begin:
	89	sk_nulls_for_each_rcu(sk, node, &head->chain) {
b1a7ffcb	90	/* For IPV6 do the cheaper port and family tests first. */
3ab5aee7 ED	91	if (INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	92	if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
	93	goto begintw;
	94	if (!INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	95	sock_put(sk);
	96	goto begin;
	97	}
	98	goto out;
	99	}
b1a7ffcb	100	}
3ab5aee7 ED	101	if (get_nulls_value(node) != slot)
	102	goto begin;
	103
	104	begintw:
b1a7ffcb	105	/* Must check for a TIME_WAIT'er before going to listener hash. */
3ab5aee7 ED	106	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
	107	if (INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	108	if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
	109	sk = NULL;
	110	goto out;
	111	}
	112	if (!INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	113	sock_put(sk);
	114	goto begintw;
	115	}
	116	goto out;
	117	}
b1a7ffcb	118	}
3ab5aee7 ED	119	if (get_nulls_value(node) != slot)
	120	goto begintw;
	121	sk = NULL;
	122	out:
	123	rcu_read_unlock();
b1a7ffcb DV	124	return sk;
	125	}
	126	EXPORT_SYMBOL(__inet6_lookup_established);
	127
42b16b3f	128	static inline int compute_score(struct sock sk, struct net net,
c25eb3bf ED	129	const unsigned short hnum,
	130	const struct in6_addr *daddr,
	131	const int dif)
	132	{
	133	int score = -1;
	134
c720c7e8	135	if (net_eq(sock_net(sk), net) && inet_sk(sk)->inet_num == hnum &&
c25eb3bf ED	136	sk->sk_family == PF_INET6) {
	137	const struct ipv6_pinfo *np = inet6_sk(sk);
	138
	139	score = 1;
	140	if (!ipv6_addr_any(&np->rcv_saddr)) {
	141	if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
	142	return -1;
	143	score++;
	144	}
	145	if (sk->sk_bound_dev_if) {
	146	if (sk->sk_bound_dev_if != dif)
	147	return -1;
	148	score++;
	149	}
	150	}
	151	return score;
	152	}
	153
d86e0dac PE	154	struct sock inet6_lookup_listener(struct net net,
	155	struct inet_hashinfo hashinfo, const struct in6_addr daddr,
	156	const unsigned short hnum, const int dif)
5324a040 ACM	157	{
5324a040 ACM	158	struct sock *sk;
c25eb3bf ED	159	const struct hlist_nulls_node *node;
	160	struct sock *result;
	161	int score, hiscore;
	162	unsigned int hash = inet_lhashfn(net, hnum);
	163	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];
	164
	165	rcu_read_lock();
	166	begin:
	167	result = NULL;
	168	hiscore = -1;
	169	sk_nulls_for_each(sk, node, &ilb->head) {
	170	score = compute_score(sk, net, hnum, daddr, dif);
	171	if (score > hiscore) {
	172	hiscore = score;
	173	result = sk;
5324a040 ACM	174	}
5324a040 ACM	175	}
c25eb3bf ED	176	/*
	177	* if the nulls value we got at the end of this lookup is
	178	* not the expected one, we must restart lookup.
	179	* We probably met an item that was moved to another chain.
	180	*/
	181	if (get_nulls_value(node) != hash + LISTENING_NULLS_BASE)
	182	goto begin;
	183	if (result) {
	184	if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
	185	result = NULL;
	186	else if (unlikely(compute_score(result, net, hnum, daddr,
	187	dif) < hiscore)) {
	188	sock_put(result);
	189	goto begin;
	190	}
	191	}
	192	rcu_read_unlock();
5324a040 ACM	193	return result;
	194	}
	195
	196	EXPORT_SYMBOL_GPL(inet6_lookup_listener);
	197
d86e0dac	198	struct sock inet6_lookup(struct net net, struct inet_hashinfo *hashinfo,
d2ecd9cc AV	199	const struct in6_addr *saddr, const __be16 sport,
d2ecd9cc AV	200	const struct in6_addr *daddr, const __be16 dport,
5324a040 ACM	201	const int dif)
	202	{
	203	struct sock *sk;
	204
	205	local_bh_disable();
d86e0dac	206	sk = __inet6_lookup(net, hashinfo, saddr, sport, daddr, ntohs(dport), dif);
5324a040 ACM	207	local_bh_enable();
	208
	209	return sk;
	210	}
	211
	212	EXPORT_SYMBOL_GPL(inet6_lookup);
d8313f5c ACM	213
	214	static int __inet6_check_established(struct inet_timewait_death_row *death_row,
	215	struct sock *sk, const __u16 lport,
	216	struct inet_timewait_sock **twp)
	217	{
	218	struct inet_hashinfo *hinfo = death_row->hashinfo;
3759fa9c	219	struct inet_sock *inet = inet_sk(sk);
d8313f5c ACM	220	const struct ipv6_pinfo *np = inet6_sk(sk);
	221	const struct in6_addr *daddr = &np->rcv_saddr;
	222	const struct in6_addr *saddr = &np->daddr;
	223	const int dif = sk->sk_bound_dev_if;
c720c7e8	224	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
33de014c PE	225	struct net *net = sock_net(sk);
33de014c PE	226	const unsigned int hash = inet6_ehashfn(net, daddr, lport, saddr,
c720c7e8	227	inet->inet_dport);
d8313f5c	228	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
9db66bdc	229	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
d8313f5c	230	struct sock *sk2;
3ab5aee7	231	const struct hlist_nulls_node *node;
d8313f5c	232	struct inet_timewait_sock *tw;
13475a30	233	int twrefcnt = 0;
d8313f5c	234
9db66bdc	235	spin_lock(lock);
d8313f5c ACM	236
d8313f5c ACM	237	/* Check TIME-WAIT sockets first. */
3ab5aee7	238	sk_nulls_for_each(sk2, node, &head->twchain) {
d8313f5c ACM	239	tw = inet_twsk(sk2);
d8313f5c ACM	240
d86e0dac	241	if (INET6_TW_MATCH(sk2, net, hash, saddr, daddr, ports, dif)) {
d8313f5c ACM	242	if (twsk_unique(sk, sk2, twp))
	243	goto unique;
	244	else
	245	goto not_unique;
	246	}
	247	}
	248	tw = NULL;
	249
	250	/* And established part... */
3ab5aee7	251	sk_nulls_for_each(sk2, node, &head->chain) {
d86e0dac	252	if (INET6_MATCH(sk2, net, hash, saddr, daddr, ports, dif))
d8313f5c ACM	253	goto not_unique;
	254	}
	255
	256	unique:
3759fa9c HX	257	/* Must record num and sport now. Otherwise we will see
3759fa9c HX	258	* in hash table socket with a funny identity. */
c720c7e8 ED	259	inet->inet_num = lport;
c720c7e8 ED	260	inet->inet_sport = htons(lport);
13475a30	261	sk->sk_hash = hash;
547b792c	262	WARN_ON(!sk_unhashed(sk));
3ab5aee7	263	__sk_nulls_add_node_rcu(sk, &head->chain);
13475a30 ED	264	if (tw) {
	265	twrefcnt = inet_twsk_unhash(tw);
	266	NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	267	}
9db66bdc	268	spin_unlock(lock);
13475a30 ED	269	if (twrefcnt)
13475a30 ED	270	inet_twsk_put(tw);
c29a0bc4	271	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
d8313f5c	272
13475a30	273	if (twp) {
d8313f5c	274	*twp = tw;
13475a30	275	} else if (tw) {
d8313f5c ACM	276	/* Silly. Should hash-dance instead... */
d8313f5c ACM	277	inet_twsk_deschedule(tw, death_row);
d8313f5c ACM	278
	279	inet_twsk_put(tw);
	280	}
	281	return 0;
	282
	283	not_unique:
9db66bdc	284	spin_unlock(lock);
d8313f5c ACM	285	return -EADDRNOTAVAIL;
	286	}
	287
	288	static inline u32 inet6_sk_port_offset(const struct sock *sk)
	289	{
	290	const struct inet_sock *inet = inet_sk(sk);
	291	const struct ipv6_pinfo *np = inet6_sk(sk);
	292	return secure_ipv6_port_ephemeral(np->rcv_saddr.s6_addr32,
	293	np->daddr.s6_addr32,
c720c7e8	294	inet->inet_dport);
d8313f5c ACM	295	}
	296
	297	int inet6_hash_connect(struct inet_timewait_death_row *death_row,
	298	struct sock *sk)
	299	{
5d8c0aa9	300	return __inet_hash_connect(death_row, sk, inet6_sk_port_offset(sk),
5ee31fc1	301	__inet6_check_established, __inet6_hash);
d8313f5c ACM	302	}
	303
	304	EXPORT_SYMBOL_GPL(inet6_hash_connect);