inet_hashtables.c 13.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/*
 * 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 INET transport hashtables
 *
 * Authors:	Lotsa people, from code originally in tcp
 *
 *	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.
 */

16
#include <linux/module.h>
17
#include <linux/random.h>
18
#include <linux/sched.h>
19
#include <linux/slab.h>
20
#include <linux/wait.h>
21

22
#include <net/inet_connection_sock.h>
23
#include <net/inet_hashtables.h>
24
#include <net/ip.h>
25 26 27 28 29

/*
 * Allocate and initialize a new local port bind bucket.
 * The bindhash mutex for snum's hash chain must be held here.
 */
30
struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep,
31
						 struct net *net,
32 33 34
						 struct inet_bind_hashbucket *head,
						 const unsigned short snum)
{
35
	struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC);
36 37

	if (tb != NULL) {
38
		write_pnet(&tb->ib_net, hold_net(net));
39 40 41 42 43 44 45 46 47 48 49
		tb->port      = snum;
		tb->fastreuse = 0;
		INIT_HLIST_HEAD(&tb->owners);
		hlist_add_head(&tb->node, &head->chain);
	}
	return tb;
}

/*
 * Caller must hold hashbucket lock for this tb with local BH disabled
 */
50
void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
51 52 53
{
	if (hlist_empty(&tb->owners)) {
		__hlist_del(&tb->node);
54
		release_net(ib_net(tb));
55 56 57
		kmem_cache_free(cachep, tb);
	}
}
58 59 60 61

void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
		    const unsigned short snum)
{
62
	inet_sk(sk)->num = snum;
63
	sk_add_bind_node(sk, &tb->owners);
64
	inet_csk(sk)->icsk_bind_hash = tb;
65 66 67 68 69
}

/*
 * Get rid of any references to a local port held by the given sock.
 */
70
static void __inet_put_port(struct sock *sk)
71
{
72
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
73 74
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->num,
			hashinfo->bhash_size);
75 76 77 78
	struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
79
	tb = inet_csk(sk)->icsk_bind_hash;
80
	__sk_del_bind_node(sk);
81 82
	inet_csk(sk)->icsk_bind_hash = NULL;
	inet_sk(sk)->num = 0;
83 84 85 86
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&head->lock);
}

87
void inet_put_port(struct sock *sk)
88 89
{
	local_bh_disable();
90
	__inet_put_port(sk);
91 92 93 94
	local_bh_enable();
}

EXPORT_SYMBOL(inet_put_port);
95

96 97 98
void __inet_inherit_port(struct sock *sk, struct sock *child)
{
	struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
99 100
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(child)->num,
			table->bhash_size);
101 102 103 104 105 106 107 108 109 110 111 112
	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
	tb = inet_csk(sk)->icsk_bind_hash;
	sk_add_bind_node(child, &tb->owners);
	inet_csk(child)->icsk_bind_hash = tb;
	spin_unlock(&head->lock);
}

EXPORT_SYMBOL_GPL(__inet_inherit_port);

113 114 115 116 117 118
/*
 * Don't inline this cruft. Here are some nice properties to exploit here. The
 * BSD API does not allow a listening sock to specify the remote port nor the
 * remote address for the connection. So always assume those are both
 * wildcarded during the search since they can never be otherwise.
 */
119 120
static struct sock *inet_lookup_listener_slow(struct net *net,
					      const struct hlist_head *head,
121
					      const __be32 daddr,
122 123
					      const unsigned short hnum,
					      const int dif)
124 125 126 127 128 129 130 131
{
	struct sock *result = NULL, *sk;
	const struct hlist_node *node;
	int hiscore = -1;

	sk_for_each(sk, node, head) {
		const struct inet_sock *inet = inet_sk(sk);

132
		if (net_eq(sock_net(sk), net) && inet->num == hnum &&
133
				!ipv6_only_sock(sk)) {
134
			const __be32 rcv_saddr = inet->rcv_saddr;
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
			int score = sk->sk_family == PF_INET ? 1 : 0;

			if (rcv_saddr) {
				if (rcv_saddr != daddr)
					continue;
				score += 2;
			}
			if (sk->sk_bound_dev_if) {
				if (sk->sk_bound_dev_if != dif)
					continue;
				score += 2;
			}
			if (score == 5)
				return sk;
			if (score > hiscore) {
				hiscore	= score;
				result	= sk;
			}
		}
	}
	return result;
}
157

158
/* Optimize the common listener case. */
159 160
struct sock *__inet_lookup_listener(struct net *net,
				    struct inet_hashinfo *hashinfo,
161
				    const __be32 daddr, const unsigned short hnum,
162
				    const int dif)
163 164
{
	struct sock *sk = NULL;
165
	struct inet_listen_hashbucket *ilb;
166

167 168 169 170
	ilb = &hashinfo->listening_hash[inet_lhashfn(net, hnum)];
	spin_lock(&ilb->lock);
	if (!hlist_empty(&ilb->head)) {
		const struct inet_sock *inet = inet_sk((sk = __sk_head(&ilb->head)));
171 172 173 174

		if (inet->num == hnum && !sk->sk_node.next &&
		    (!inet->rcv_saddr || inet->rcv_saddr == daddr) &&
		    (sk->sk_family == PF_INET || !ipv6_only_sock(sk)) &&
175
		    !sk->sk_bound_dev_if && net_eq(sock_net(sk), net))
176
			goto sherry_cache;
177
		sk = inet_lookup_listener_slow(net, &ilb->head, daddr, hnum, dif);
178 179 180 181 182
	}
	if (sk) {
sherry_cache:
		sock_hold(sk);
	}
183
	spin_unlock(&ilb->lock);
184 185
	return sk;
}
186
EXPORT_SYMBOL_GPL(__inet_lookup_listener);
187

188 189
struct sock * __inet_lookup_established(struct net *net,
				  struct inet_hashinfo *hashinfo,
190 191 192 193 194 195 196
				  const __be32 saddr, const __be16 sport,
				  const __be32 daddr, const u16 hnum,
				  const int dif)
{
	INET_ADDR_COOKIE(acookie, saddr, daddr)
	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
	struct sock *sk;
197
	const struct hlist_nulls_node *node;
198 199 200
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
201
	unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
202 203
	unsigned int slot = hash & (hashinfo->ehash_size - 1);
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
204

205 206 207
	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
208
		if (INET_MATCH(sk, net, hash, acookie,
209 210 211 212 213 214 215 216 217 218
					saddr, daddr, ports, dif)) {
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
				goto begintw;
			if (unlikely(!INET_MATCH(sk, net, hash, acookie,
				saddr, daddr, ports, dif))) {
				sock_put(sk);
				goto begin;
			}
			goto out;
		}
219
	}
220 221 222 223 224 225 226
	/*
	 * 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) != slot)
		goto begin;
227

228
begintw:
229
	/* Must check for a TIME_WAIT'er before going to listener hash. */
230
	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
231
		if (INET_TW_MATCH(sk, net, hash, acookie,
232 233 234 235 236 237 238 239 240 241 242 243
					saddr, daddr, ports, dif)) {
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
				sk = NULL;
				goto out;
			}
			if (unlikely(!INET_TW_MATCH(sk, net, hash, acookie,
				 saddr, daddr, ports, dif))) {
				sock_put(sk);
				goto begintw;
			}
			goto out;
		}
244
	}
245 246 247 248 249 250 251
	/*
	 * 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) != slot)
		goto begintw;
252 253
	sk = NULL;
out:
254
	rcu_read_unlock();
255 256 257 258
	return sk;
}
EXPORT_SYMBOL_GPL(__inet_lookup_established);

259 260 261 262 263 264 265
/* called with local bh disabled */
static int __inet_check_established(struct inet_timewait_death_row *death_row,
				    struct sock *sk, __u16 lport,
				    struct inet_timewait_sock **twp)
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
	struct inet_sock *inet = inet_sk(sk);
266 267
	__be32 daddr = inet->rcv_saddr;
	__be32 saddr = inet->daddr;
268 269
	int dif = sk->sk_bound_dev_if;
	INET_ADDR_COOKIE(acookie, saddr, daddr)
A
Al Viro 已提交
270
	const __portpair ports = INET_COMBINED_PORTS(inet->dport, lport);
271 272
	struct net *net = sock_net(sk);
	unsigned int hash = inet_ehashfn(net, daddr, lport, saddr, inet->dport);
273
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
274
	rwlock_t *lock = inet_ehash_lockp(hinfo, hash);
275
	struct sock *sk2;
276
	const struct hlist_nulls_node *node;
277 278 279
	struct inet_timewait_sock *tw;

	prefetch(head->chain.first);
280
	write_lock(lock);
281 282

	/* Check TIME-WAIT sockets first. */
283
	sk_nulls_for_each(sk2, node, &head->twchain) {
284 285
		tw = inet_twsk(sk2);

286 287
		if (INET_TW_MATCH(sk2, net, hash, acookie,
					saddr, daddr, ports, dif)) {
288 289 290 291 292 293 294 295 296
			if (twsk_unique(sk, sk2, twp))
				goto unique;
			else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
297
	sk_nulls_for_each(sk2, node, &head->chain) {
298 299
		if (INET_MATCH(sk2, net, hash, acookie,
					saddr, daddr, ports, dif))
300 301 302 303 304 305 306 307 308
			goto not_unique;
	}

unique:
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
	inet->num = lport;
	inet->sport = htons(lport);
	sk->sk_hash = hash;
309
	WARN_ON(!sk_unhashed(sk));
310
	__sk_nulls_add_node_rcu(sk, &head->chain);
311
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
312
	write_unlock(lock);
313 314 315

	if (twp) {
		*twp = tw;
316
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
317 318 319
	} else if (tw) {
		/* Silly. Should hash-dance instead... */
		inet_twsk_deschedule(tw, death_row);
320
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
321 322 323 324 325 326 327

		inet_twsk_put(tw);
	}

	return 0;

not_unique:
328
	write_unlock(lock);
329 330 331 332 333 334
	return -EADDRNOTAVAIL;
}

static inline u32 inet_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
335
	return secure_ipv4_port_ephemeral(inet->rcv_saddr, inet->daddr,
336 337 338
					  inet->dport);
}

339
void __inet_hash_nolisten(struct sock *sk)
340
{
341
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
342
	struct hlist_nulls_head *list;
343 344 345
	rwlock_t *lock;
	struct inet_ehash_bucket *head;

346
	WARN_ON(!sk_unhashed(sk));
347 348 349 350 351 352 353

	sk->sk_hash = inet_sk_ehashfn(sk);
	head = inet_ehash_bucket(hashinfo, sk->sk_hash);
	list = &head->chain;
	lock = inet_ehash_lockp(hashinfo, sk->sk_hash);

	write_lock(lock);
354
	__sk_nulls_add_node_rcu(sk, list);
355
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
356 357 358 359
	write_unlock(lock);
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);

360
static void __inet_hash(struct sock *sk)
361
{
362
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
363
	struct inet_listen_hashbucket *ilb;
364 365

	if (sk->sk_state != TCP_LISTEN) {
366
		__inet_hash_nolisten(sk);
367 368 369
		return;
	}

370
	WARN_ON(!sk_unhashed(sk));
371
	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
372

373 374
	spin_lock(&ilb->lock);
	__sk_add_node(sk, &ilb->head);
375
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
376
	spin_unlock(&ilb->lock);
377
}
378 379 380 381 382 383 384 385 386 387 388 389 390

void inet_hash(struct sock *sk)
{
	if (sk->sk_state != TCP_CLOSE) {
		local_bh_disable();
		__inet_hash(sk);
		local_bh_enable();
	}
}
EXPORT_SYMBOL_GPL(inet_hash);

void inet_unhash(struct sock *sk)
{
391
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
392 393

	if (sk_unhashed(sk))
394
		return;
395 396

	if (sk->sk_state == TCP_LISTEN) {
397 398 399 400
		struct inet_listen_hashbucket *ilb;

		ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
		spin_lock_bh(&ilb->lock);
401 402
		if (__sk_del_node_init(sk))
			sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
403
		spin_unlock_bh(&ilb->lock);
404
	} else {
405 406
		rwlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);

407
		write_lock_bh(lock);
408 409
		if (__sk_nulls_del_node_init_rcu(sk))
			sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
410
		write_unlock_bh(lock);
411 412 413
	}
}
EXPORT_SYMBOL_GPL(inet_unhash);
414

415
int __inet_hash_connect(struct inet_timewait_death_row *death_row,
416
		struct sock *sk, u32 port_offset,
417 418
		int (*check_established)(struct inet_timewait_death_row *,
			struct sock *, __u16, struct inet_timewait_sock **),
419
		void (*hash)(struct sock *sk))
420 421 422
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
	const unsigned short snum = inet_sk(sk)->num;
423 424
	struct inet_bind_hashbucket *head;
	struct inet_bind_bucket *tb;
425
	int ret;
426
	struct net *net = sock_net(sk);
427

428
	if (!snum) {
429
		int i, remaining, low, high, port;
430
		static u32 hint;
431
		u32 offset = hint + port_offset;
432
		struct hlist_node *node;
433
		struct inet_timewait_sock *tw = NULL;
434

435
		inet_get_local_port_range(&low, &high);
436
		remaining = (high - low) + 1;
437

438
		local_bh_disable();
439 440
		for (i = 1; i <= remaining; i++) {
			port = low + (i + offset) % remaining;
441 442
			head = &hinfo->bhash[inet_bhashfn(net, port,
					hinfo->bhash_size)];
443
			spin_lock(&head->lock);
444

445 446 447 448
			/* Does not bother with rcv_saddr checks,
			 * because the established check is already
			 * unique enough.
			 */
449
			inet_bind_bucket_for_each(tb, node, &head->chain) {
450
				if (ib_net(tb) == net && tb->port == port) {
451
					WARN_ON(hlist_empty(&tb->owners));
452 453
					if (tb->fastreuse >= 0)
						goto next_port;
454 455
					if (!check_established(death_row, sk,
								port, &tw))
456 457 458 459 460
						goto ok;
					goto next_port;
				}
			}

461 462
			tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
					net, head, port);
463 464 465 466 467 468 469 470 471 472 473 474 475
			if (!tb) {
				spin_unlock(&head->lock);
				break;
			}
			tb->fastreuse = -1;
			goto ok;

		next_port:
			spin_unlock(&head->lock);
		}
		local_bh_enable();

		return -EADDRNOTAVAIL;
476 477 478 479

ok:
		hint += i;

480 481
		/* Head lock still held and bh's disabled */
		inet_bind_hash(sk, tb, port);
482
		if (sk_unhashed(sk)) {
483
			inet_sk(sk)->sport = htons(port);
484
			hash(sk);
485 486
		}
		spin_unlock(&head->lock);
487

488 489 490 491
		if (tw) {
			inet_twsk_deschedule(tw, death_row);
			inet_twsk_put(tw);
		}
492 493 494

		ret = 0;
		goto out;
495
	}
496

497
	head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
498
	tb  = inet_csk(sk)->icsk_bind_hash;
499 500
	spin_lock_bh(&head->lock);
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
501
		hash(sk);
502 503 504 505 506
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
		spin_unlock(&head->lock);
		/* No definite answer... Walk to established hash table */
507
		ret = check_established(death_row, sk, snum, NULL);
508 509 510 511 512
out:
		local_bh_enable();
		return ret;
	}
}
513 514 515 516 517 518 519

/*
 * Bind a port for a connect operation and hash it.
 */
int inet_hash_connect(struct inet_timewait_death_row *death_row,
		      struct sock *sk)
{
520
	return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
521 522
			__inet_check_established, __inet_hash_nolisten);
}
523 524

EXPORT_SYMBOL_GPL(inet_hash_connect);
525 526 527 528 529 530 531 532 533 534

void inet_hashinfo_init(struct inet_hashinfo *h)
{
	int i;

	for (i = 0; i < INET_LHTABLE_SIZE; i++)
		spin_lock_init(&h->listening_hash[i].lock);
}

EXPORT_SYMBOL_GPL(inet_hashinfo_init);