[net-next-2.6.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 <acme@mandriva.com>
 *
 *	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/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 int inline 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 <acme@mandriva.com>
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>
d8313f5c	23	#include <net/ip.h>
5324a040	24
9327f705	25	int __inet6_hash(struct sock sk, struct inet_timewait_sock tw)
b1a7ffcb	26	{
39d8cda7	27	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
9327f705	28	int twrefcnt = 0;
b1a7ffcb	29
547b792c	30	WARN_ON(!sk_unhashed(sk));
b1a7ffcb DV	31
b1a7ffcb DV	32	if (sk->sk_state == TCP_LISTEN) {
5caea4ea	33	struct inet_listen_hashbucket *ilb;
3ab5aee7	34
5caea4ea ED	35	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
5caea4ea ED	36	spin_lock(&ilb->lock);
c25eb3bf	37	__sk_nulls_add_node_rcu(sk, &ilb->head);
5caea4ea	38	spin_unlock(&ilb->lock);
b1a7ffcb DV	39	} else {
b1a7ffcb DV	40	unsigned int hash;
3ab5aee7	41	struct hlist_nulls_head *list;
9db66bdc	42	spinlock_t *lock;
3ab5aee7	43
b1a7ffcb	44	sk->sk_hash = hash = inet6_sk_ehashfn(sk);
230140cf ED	45	list = &inet_ehash_bucket(hashinfo, hash)->chain;
230140cf ED	46	lock = inet_ehash_lockp(hashinfo, hash);
9db66bdc	47	spin_lock(lock);
3ab5aee7	48	__sk_nulls_add_node_rcu(sk, list);
9327f705 ED	49	if (tw) {
	50	WARN_ON(sk->sk_hash != tw->tw_hash);
	51	twrefcnt = inet_twsk_unhash(tw);
	52	}
9db66bdc	53	spin_unlock(lock);
b1a7ffcb DV	54	}
b1a7ffcb DV	55
c29a0bc4	56	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
9327f705	57	return twrefcnt;
b1a7ffcb DV	58	}
	59	EXPORT_SYMBOL(__inet6_hash);
	60
	61	/*
	62	* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so
	63	* we need not check it for TCP lookups anymore, thanks Alexey. -DaveM
	64	*
	65	* The sockhash lock must be held as a reader here.
	66	*/
d86e0dac PE	67	struct sock __inet6_lookup_established(struct net net,
d86e0dac PE	68	struct inet_hashinfo *hashinfo,
b1a7ffcb	69	const struct in6_addr *saddr,
d2ecd9cc	70	const __be16 sport,
b1a7ffcb DV	71	const struct in6_addr *daddr,
	72	const u16 hnum,
	73	const int dif)
	74	{
	75	struct sock *sk;
3ab5aee7	76	const struct hlist_nulls_node *node;
4f765d84	77	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
b1a7ffcb DV	78	/* Optimize here for direct hit, only listening connections can
	79	* have wildcards anyways.
	80	*/
33de014c	81	unsigned int hash = inet6_ehashfn(net, daddr, hnum, saddr, sport);
f373b53b	82	unsigned int slot = hash & hashinfo->ehash_mask;
3ab5aee7	83	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
b1a7ffcb	84
3ab5aee7 ED	85
	86	rcu_read_lock();
	87	begin:
	88	sk_nulls_for_each_rcu(sk, node, &head->chain) {
b1a7ffcb	89	/* For IPV6 do the cheaper port and family tests first. */
3ab5aee7 ED	90	if (INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	91	if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
	92	goto begintw;
	93	if (!INET6_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	94	sock_put(sk);
	95	goto begin;
	96	}
	97	goto out;
	98	}
b1a7ffcb	99	}
3ab5aee7 ED	100	if (get_nulls_value(node) != slot)
	101	goto begin;
	102
	103	begintw:
b1a7ffcb	104	/* Must check for a TIME_WAIT'er before going to listener hash. */
3ab5aee7 ED	105	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
	106	if (INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	107	if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
	108	sk = NULL;
	109	goto out;
	110	}
	111	if (!INET6_TW_MATCH(sk, net, hash, saddr, daddr, ports, dif)) {
	112	sock_put(sk);
	113	goto begintw;
	114	}
	115	goto out;
	116	}
b1a7ffcb	117	}
3ab5aee7 ED	118	if (get_nulls_value(node) != slot)
	119	goto begintw;
	120	sk = NULL;
	121	out:
	122	rcu_read_unlock();
b1a7ffcb DV	123	return sk;
	124	}
	125	EXPORT_SYMBOL(__inet6_lookup_established);
	126
c25eb3bf ED	127	static int inline compute_score(struct sock sk, struct net net,
	128	const unsigned short hnum,
	129	const struct in6_addr *daddr,
	130	const int dif)
	131	{
	132	int score = -1;
	133
c720c7e8	134	if (net_eq(sock_net(sk), net) && inet_sk(sk)->inet_num == hnum &&
c25eb3bf ED	135	sk->sk_family == PF_INET6) {
	136	const struct ipv6_pinfo *np = inet6_sk(sk);
	137
	138	score = 1;
	139	if (!ipv6_addr_any(&np->rcv_saddr)) {
	140	if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
	141	return -1;
	142	score++;
	143	}
	144	if (sk->sk_bound_dev_if) {
	145	if (sk->sk_bound_dev_if != dif)
	146	return -1;
	147	score++;
	148	}
	149	}
	150	return score;
	151	}
	152
d86e0dac PE	153	struct sock inet6_lookup_listener(struct net net,
	154	struct inet_hashinfo hashinfo, const struct in6_addr daddr,
	155	const unsigned short hnum, const int dif)
5324a040 ACM	156	{
5324a040 ACM	157	struct sock *sk;
c25eb3bf ED	158	const struct hlist_nulls_node *node;
	159	struct sock *result;
	160	int score, hiscore;
	161	unsigned int hash = inet_lhashfn(net, hnum);
	162	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];
	163
	164	rcu_read_lock();
	165	begin:
	166	result = NULL;
	167	hiscore = -1;
	168	sk_nulls_for_each(sk, node, &ilb->head) {
	169	score = compute_score(sk, net, hnum, daddr, dif);
	170	if (score > hiscore) {
	171	hiscore = score;
	172	result = sk;
5324a040 ACM	173	}
5324a040 ACM	174	}
c25eb3bf ED	175	/*
	176	* if the nulls value we got at the end of this lookup is
	177	* not the expected one, we must restart lookup.
	178	* We probably met an item that was moved to another chain.
	179	*/
	180	if (get_nulls_value(node) != hash + LISTENING_NULLS_BASE)
	181	goto begin;
	182	if (result) {
	183	if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt)))
	184	result = NULL;
	185	else if (unlikely(compute_score(result, net, hnum, daddr,
	186	dif) < hiscore)) {
	187	sock_put(result);
	188	goto begin;
	189	}
	190	}
	191	rcu_read_unlock();
5324a040 ACM	192	return result;
	193	}
	194
	195	EXPORT_SYMBOL_GPL(inet6_lookup_listener);
	196
d86e0dac	197	struct sock inet6_lookup(struct net net, struct inet_hashinfo *hashinfo,
d2ecd9cc AV	198	const struct in6_addr *saddr, const __be16 sport,
d2ecd9cc AV	199	const struct in6_addr *daddr, const __be16 dport,
5324a040 ACM	200	const int dif)
	201	{
	202	struct sock *sk;
	203
	204	local_bh_disable();
d86e0dac	205	sk = __inet6_lookup(net, hashinfo, saddr, sport, daddr, ntohs(dport), dif);
5324a040 ACM	206	local_bh_enable();
	207
	208	return sk;
	209	}
	210
	211	EXPORT_SYMBOL_GPL(inet6_lookup);
d8313f5c ACM	212
	213	static int __inet6_check_established(struct inet_timewait_death_row *death_row,
	214	struct sock *sk, const __u16 lport,
	215	struct inet_timewait_sock **twp)
	216	{
	217	struct inet_hashinfo *hinfo = death_row->hashinfo;
3759fa9c	218	struct inet_sock *inet = inet_sk(sk);
d8313f5c ACM	219	const struct ipv6_pinfo *np = inet6_sk(sk);
	220	const struct in6_addr *daddr = &np->rcv_saddr;
	221	const struct in6_addr *saddr = &np->daddr;
	222	const int dif = sk->sk_bound_dev_if;
c720c7e8	223	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
33de014c PE	224	struct net *net = sock_net(sk);
33de014c PE	225	const unsigned int hash = inet6_ehashfn(net, daddr, lport, saddr,
c720c7e8	226	inet->inet_dport);
d8313f5c	227	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
9db66bdc	228	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
d8313f5c	229	struct sock *sk2;
3ab5aee7	230	const struct hlist_nulls_node *node;
d8313f5c	231	struct inet_timewait_sock *tw;
13475a30	232	int twrefcnt = 0;
d8313f5c	233
9db66bdc	234	spin_lock(lock);
d8313f5c ACM	235
d8313f5c ACM	236	/* Check TIME-WAIT sockets first. */
3ab5aee7	237	sk_nulls_for_each(sk2, node, &head->twchain) {
d8313f5c ACM	238	tw = inet_twsk(sk2);
d8313f5c ACM	239
d86e0dac	240	if (INET6_TW_MATCH(sk2, net, hash, saddr, daddr, ports, dif)) {
d8313f5c ACM	241	if (twsk_unique(sk, sk2, twp))
	242	goto unique;
	243	else
	244	goto not_unique;
	245	}
	246	}
	247	tw = NULL;
	248
	249	/* And established part... */
3ab5aee7	250	sk_nulls_for_each(sk2, node, &head->chain) {
d86e0dac	251	if (INET6_MATCH(sk2, net, hash, saddr, daddr, ports, dif))
d8313f5c ACM	252	goto not_unique;
	253	}
	254
	255	unique:
3759fa9c HX	256	/* Must record num and sport now. Otherwise we will see
3759fa9c HX	257	* in hash table socket with a funny identity. */
c720c7e8 ED	258	inet->inet_num = lport;
c720c7e8 ED	259	inet->inet_sport = htons(lport);
13475a30	260	sk->sk_hash = hash;
547b792c	261	WARN_ON(!sk_unhashed(sk));
3ab5aee7	262	__sk_nulls_add_node_rcu(sk, &head->chain);
13475a30 ED	263	if (tw) {
	264	twrefcnt = inet_twsk_unhash(tw);
	265	NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	266	}
9db66bdc	267	spin_unlock(lock);
13475a30 ED	268	if (twrefcnt)
13475a30 ED	269	inet_twsk_put(tw);
c29a0bc4	270	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
d8313f5c	271
13475a30	272	if (twp) {
d8313f5c	273	*twp = tw;
13475a30	274	} else if (tw) {
d8313f5c ACM	275	/* Silly. Should hash-dance instead... */
d8313f5c ACM	276	inet_twsk_deschedule(tw, death_row);
d8313f5c ACM	277
	278	inet_twsk_put(tw);
	279	}
	280	return 0;
	281
	282	not_unique:
9db66bdc	283	spin_unlock(lock);
d8313f5c ACM	284	return -EADDRNOTAVAIL;
	285	}
	286
	287	static inline u32 inet6_sk_port_offset(const struct sock *sk)
	288	{
	289	const struct inet_sock *inet = inet_sk(sk);
	290	const struct ipv6_pinfo *np = inet6_sk(sk);
	291	return secure_ipv6_port_ephemeral(np->rcv_saddr.s6_addr32,
	292	np->daddr.s6_addr32,
c720c7e8	293	inet->inet_dport);
d8313f5c ACM	294	}
	295
	296	int inet6_hash_connect(struct inet_timewait_death_row *death_row,
	297	struct sock *sk)
	298	{
5d8c0aa9	299	return __inet_hash_connect(death_row, sk, inet6_sk_port_offset(sk),
5ee31fc1	300	__inet6_check_established, __inet6_hash);
d8313f5c ACM	301	}
	302
	303	EXPORT_SYMBOL_GPL(inet6_hash_connect);