inet6_hashtables.c 7.8 KB
Newer Older
1 2 3 4 5 6 7
/*
 * 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
 *
8 9
 * Authors:	Lotsa people, from code originally in tcp, generalised here
 * 		by Arnaldo Carvalho de Melo <acme@mandriva.com>
10 11 12 13 14 15 16 17
 *
 *	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>
18
#include <linux/random.h>
19 20 21 22

#include <net/inet_connection_sock.h>
#include <net/inet_hashtables.h>
#include <net/inet6_hashtables.h>
23
#include <net/secure_seq.h>
24
#include <net/ip.h>
25

26
int __inet6_hash(struct sock *sk, struct inet_timewait_sock *tw)
27
{
28
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
29
	int twrefcnt = 0;
30

31
	WARN_ON(!sk_unhashed(sk));
32 33

	if (sk->sk_state == TCP_LISTEN) {
34
		struct inet_listen_hashbucket *ilb;
35

36 37
		ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
		spin_lock(&ilb->lock);
38
		__sk_nulls_add_node_rcu(sk, &ilb->head);
39
		spin_unlock(&ilb->lock);
40 41
	} else {
		unsigned int hash;
42
		struct hlist_nulls_head *list;
43
		spinlock_t *lock;
44

45
		sk->sk_hash = hash = inet6_sk_ehashfn(sk);
46 47
		list = &inet_ehash_bucket(hashinfo, hash)->chain;
		lock = inet_ehash_lockp(hashinfo, hash);
48
		spin_lock(lock);
49
		__sk_nulls_add_node_rcu(sk, list);
50 51 52 53
		if (tw) {
			WARN_ON(sk->sk_hash != tw->tw_hash);
			twrefcnt = inet_twsk_unhash(tw);
		}
54
		spin_unlock(lock);
55 56
	}

57
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
58
	return twrefcnt;
59 60 61 62 63 64 65 66 67
}
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.
 */
68 69
struct sock *__inet6_lookup_established(struct net *net,
					struct inet_hashinfo *hashinfo,
70
					   const struct in6_addr *saddr,
71
					   const __be16 sport,
72 73 74 75 76
					   const struct in6_addr *daddr,
					   const u16 hnum,
					   const int dif)
{
	struct sock *sk;
77
	const struct hlist_nulls_node *node;
78
	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
79 80 81
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
82
	unsigned int hash = inet6_ehashfn(net, daddr, hnum, saddr, sport);
83
	unsigned int slot = hash & hashinfo->ehash_mask;
84
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
85

86 87 88 89

	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
90 91
		if (sk->sk_hash != hash)
			continue;
92 93
		if (!INET6_MATCH(sk, net, saddr, daddr, ports, dif))
			continue;
E
Eric Dumazet 已提交
94 95 96
		if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
			goto out;

97 98 99
		if (unlikely(!INET6_MATCH(sk, net, saddr, daddr, ports, dif))) {
			sock_gen_put(sk);
			goto begin;
100
		}
101
		goto found;
102
	}
103 104 105
	if (get_nulls_value(node) != slot)
		goto begin;
out:
E
Eric Dumazet 已提交
106 107
	sk = NULL;
found:
108
	rcu_read_unlock();
109 110 111 112
	return sk;
}
EXPORT_SYMBOL(__inet6_lookup_established);

113
static inline int compute_score(struct sock *sk, struct net *net,
114 115 116 117 118 119
				const unsigned short hnum,
				const struct in6_addr *daddr,
				const int dif)
{
	int score = -1;

120
	if (net_eq(sock_net(sk), net) && inet_sk(sk)->inet_num == hnum &&
121 122 123
	    sk->sk_family == PF_INET6) {

		score = 1;
124 125
		if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
			if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
126 127 128 129 130 131 132 133 134 135 136 137
				return -1;
			score++;
		}
		if (sk->sk_bound_dev_if) {
			if (sk->sk_bound_dev_if != dif)
				return -1;
			score++;
		}
	}
	return score;
}

138
struct sock *inet6_lookup_listener(struct net *net,
139 140
		struct inet_hashinfo *hashinfo, const struct in6_addr *saddr,
		const __be16 sport, const struct in6_addr *daddr,
141
		const unsigned short hnum, const int dif)
142 143
{
	struct sock *sk;
144 145
	const struct hlist_nulls_node *node;
	struct sock *result;
146 147
	int score, hiscore, matches = 0, reuseport = 0;
	u32 phash = 0;
148 149 150 151 152 153
	unsigned int hash = inet_lhashfn(net, hnum);
	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];

	rcu_read_lock();
begin:
	result = NULL;
154
	hiscore = 0;
155 156 157 158 159
	sk_nulls_for_each(sk, node, &ilb->head) {
		score = compute_score(sk, net, hnum, daddr, dif);
		if (score > hiscore) {
			hiscore = score;
			result = sk;
160 161 162 163 164 165 166 167 168 169 170
			reuseport = sk->sk_reuseport;
			if (reuseport) {
				phash = inet6_ehashfn(net, daddr, hnum,
						      saddr, sport);
				matches = 1;
			}
		} else if (score == hiscore && reuseport) {
			matches++;
			if (((u64)phash * matches) >> 32 == 0)
				result = sk;
			phash = next_pseudo_random32(phash);
171 172
		}
	}
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
	/*
	 * 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();
190 191 192 193 194
	return result;
}

EXPORT_SYMBOL_GPL(inet6_lookup_listener);

195
struct sock *inet6_lookup(struct net *net, struct inet_hashinfo *hashinfo,
196 197
			  const struct in6_addr *saddr, const __be16 sport,
			  const struct in6_addr *daddr, const __be16 dport,
198 199 200 201 202
			  const int dif)
{
	struct sock *sk;

	local_bh_disable();
203
	sk = __inet6_lookup(net, hashinfo, saddr, sport, daddr, ntohs(dport), dif);
204 205 206 207 208 209
	local_bh_enable();

	return sk;
}

EXPORT_SYMBOL_GPL(inet6_lookup);
210 211 212 213 214 215

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;
216
	struct inet_sock *inet = inet_sk(sk);
217 218
	const struct in6_addr *daddr = &sk->sk_v6_rcv_saddr;
	const struct in6_addr *saddr = &sk->sk_v6_daddr;
219
	const int dif = sk->sk_bound_dev_if;
220
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
221 222
	struct net *net = sock_net(sk);
	const unsigned int hash = inet6_ehashfn(net, daddr, lport, saddr,
223
						inet->inet_dport);
224
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
225
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
226
	struct sock *sk2;
227
	const struct hlist_nulls_node *node;
E
Eric Dumazet 已提交
228
	struct inet_timewait_sock *tw = NULL;
229
	int twrefcnt = 0;
230

231
	spin_lock(lock);
232

E
Eric Dumazet 已提交
233
	sk_nulls_for_each(sk2, node, &head->chain) {
234 235
		if (sk2->sk_hash != hash)
			continue;
236

237 238
		if (likely(INET6_MATCH(sk2, net, saddr, daddr, ports, dif))) {
			if (sk2->sk_state == TCP_TIME_WAIT) {
E
Eric Dumazet 已提交
239 240
				tw = inet_twsk(sk2);
				if (twsk_unique(sk, sk2, twp))
241
					break;
E
Eric Dumazet 已提交
242
			}
243
			goto not_unique;
244
		}
245 246
	}

247
	/* Must record num and sport now. Otherwise we will see
248 249
	 * in hash table socket with a funny identity.
	 */
250 251
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
252
	sk->sk_hash = hash;
253
	WARN_ON(!sk_unhashed(sk));
254
	__sk_nulls_add_node_rcu(sk, &head->chain);
255 256 257 258
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
259
	spin_unlock(lock);
260 261
	if (twrefcnt)
		inet_twsk_put(tw);
262
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
263

264
	if (twp) {
265
		*twp = tw;
266
	} else if (tw) {
267 268 269 270 271 272 273 274
		/* Silly. Should hash-dance instead... */
		inet_twsk_deschedule(tw, death_row);

		inet_twsk_put(tw);
	}
	return 0;

not_unique:
275
	spin_unlock(lock);
276 277 278 279 280 281
	return -EADDRNOTAVAIL;
}

static inline u32 inet6_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
282 283 284

	return secure_ipv6_port_ephemeral(sk->sk_v6_rcv_saddr.s6_addr32,
					  sk->sk_v6_daddr.s6_addr32,
285
					  inet->inet_dport);
286 287 288 289 290
}

int inet6_hash_connect(struct inet_timewait_death_row *death_row,
		       struct sock *sk)
{
291
	return __inet_hash_connect(death_row, sk, inet6_sk_port_offset(sk),
292
			__inet6_check_established, __inet6_hash);
293 294 295
}

EXPORT_SYMBOL_GPL(inet6_hash_connect);
新手
引导
客服 返回
顶部