inet_hashtables.c 15.2 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/secure_seq.h>
25
#include <net/ip.h>
26 27 28 29 30

/*
 * Allocate and initialize a new local port bind bucket.
 * The bindhash mutex for snum's hash chain must be held here.
 */
31
struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep,
32
						 struct net *net,
33 34 35
						 struct inet_bind_hashbucket *head,
						 const unsigned short snum)
{
36
	struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC);
37 38

	if (tb != NULL) {
39
		write_pnet(&tb->ib_net, hold_net(net));
40 41
		tb->port      = snum;
		tb->fastreuse = 0;
42
		tb->num_owners = 0;
43 44 45 46 47 48 49 50 51
		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
 */
52
void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
53 54 55
{
	if (hlist_empty(&tb->owners)) {
		__hlist_del(&tb->node);
56
		release_net(ib_net(tb));
57 58 59
		kmem_cache_free(cachep, tb);
	}
}
60 61 62 63

void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
		    const unsigned short snum)
{
64 65
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;

66
	atomic_inc(&hashinfo->bsockets);
67

68
	inet_sk(sk)->inet_num = snum;
69
	sk_add_bind_node(sk, &tb->owners);
70
	tb->num_owners++;
71
	inet_csk(sk)->icsk_bind_hash = tb;
72 73 74 75 76
}

/*
 * Get rid of any references to a local port held by the given sock.
 */
77
static void __inet_put_port(struct sock *sk)
78
{
79
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
80
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->inet_num,
81
			hashinfo->bhash_size);
82 83 84
	struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
	struct inet_bind_bucket *tb;

85
	atomic_dec(&hashinfo->bsockets);
86

87
	spin_lock(&head->lock);
88
	tb = inet_csk(sk)->icsk_bind_hash;
89
	__sk_del_bind_node(sk);
90
	tb->num_owners--;
91
	inet_csk(sk)->icsk_bind_hash = NULL;
92
	inet_sk(sk)->inet_num = 0;
93 94 95 96
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&head->lock);
}

97
void inet_put_port(struct sock *sk)
98 99
{
	local_bh_disable();
100
	__inet_put_port(sk);
101 102 103
	local_bh_enable();
}
EXPORT_SYMBOL(inet_put_port);
104

105
int __inet_inherit_port(struct sock *sk, struct sock *child)
106 107
{
	struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
108 109
	unsigned short port = inet_sk(child)->inet_num;
	const int bhash = inet_bhashfn(sock_net(sk), port,
110
			table->bhash_size);
111 112 113 114 115
	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
	tb = inet_csk(sk)->icsk_bind_hash;
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
	if (tb->port != port) {
		/* NOTE: using tproxy and redirecting skbs to a proxy
		 * on a different listener port breaks the assumption
		 * that the listener socket's icsk_bind_hash is the same
		 * as that of the child socket. We have to look up or
		 * create a new bind bucket for the child here. */
		struct hlist_node *node;
		inet_bind_bucket_for_each(tb, node, &head->chain) {
			if (net_eq(ib_net(tb), sock_net(sk)) &&
			    tb->port == port)
				break;
		}
		if (!node) {
			tb = inet_bind_bucket_create(table->bind_bucket_cachep,
						     sock_net(sk), head, port);
			if (!tb) {
				spin_unlock(&head->lock);
				return -ENOMEM;
			}
		}
	}
137
	inet_bind_hash(child, tb, port);
138
	spin_unlock(&head->lock);
139 140

	return 0;
141 142 143
}
EXPORT_SYMBOL_GPL(__inet_inherit_port);

144 145 146 147 148 149 150
static inline int compute_score(struct sock *sk, struct net *net,
				const unsigned short hnum, const __be32 daddr,
				const int dif)
{
	int score = -1;
	struct inet_sock *inet = inet_sk(sk);

151
	if (net_eq(sock_net(sk), net) && inet->inet_num == hnum &&
152
			!ipv6_only_sock(sk)) {
153
		__be32 rcv_saddr = inet->inet_rcv_saddr;
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
		score = sk->sk_family == PF_INET ? 1 : 0;
		if (rcv_saddr) {
			if (rcv_saddr != daddr)
				return -1;
			score += 2;
		}
		if (sk->sk_bound_dev_if) {
			if (sk->sk_bound_dev_if != dif)
				return -1;
			score += 2;
		}
	}
	return score;
}

169 170 171 172 173 174
/*
 * 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.
 */
175

176

177 178
struct sock *__inet_lookup_listener(struct net *net,
				    struct inet_hashinfo *hashinfo,
179
				    const __be32 daddr, const unsigned short hnum,
180
				    const int dif)
181
{
182 183 184 185 186
	struct sock *sk, *result;
	struct hlist_nulls_node *node;
	unsigned int hash = inet_lhashfn(net, hnum);
	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];
	int score, hiscore;
187

188 189 190 191 192 193 194 195 196 197
	rcu_read_lock();
begin:
	result = NULL;
	hiscore = -1;
	sk_nulls_for_each_rcu(sk, node, &ilb->head) {
		score = compute_score(sk, net, hnum, daddr, dif);
		if (score > hiscore) {
			result = sk;
			hiscore = score;
		}
198
	}
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
	/*
	 * 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;
		}
214
	}
215 216
	rcu_read_unlock();
	return result;
217
}
218
EXPORT_SYMBOL_GPL(__inet_lookup_listener);
219

Daniel Baluta's avatar
Daniel Baluta committed
220
struct sock *__inet_lookup_established(struct net *net,
221
				  struct inet_hashinfo *hashinfo,
222 223 224 225 226 227 228
				  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;
229
	const struct hlist_nulls_node *node;
230 231 232
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
233
	unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
234
	unsigned int slot = hash & hashinfo->ehash_mask;
235
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
236

237 238 239
	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
240 241 242 243
		if (sk->sk_hash != hash)
			continue;
		if (likely(INET_MATCH(sk, net, acookie,
				      saddr, daddr, ports, dif))) {
244 245
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
				goto begintw;
246 247
			if (unlikely(!INET_MATCH(sk, net, acookie,
						 saddr, daddr, ports, dif))) {
248 249 250 251 252
				sock_put(sk);
				goto begin;
			}
			goto out;
		}
253
	}
254 255 256 257 258 259 260
	/*
	 * 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;
261

262
begintw:
263
	/* Must check for a TIME_WAIT'er before going to listener hash. */
264
	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
265 266 267 268 269
		if (sk->sk_hash != hash)
			continue;
		if (likely(INET_TW_MATCH(sk, net, acookie,
					 saddr, daddr, ports,
					 dif))) {
270 271 272 273
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
				sk = NULL;
				goto out;
			}
274 275 276
			if (unlikely(!INET_TW_MATCH(sk, net, acookie,
						    saddr, daddr, ports,
						    dif))) {
277 278 279 280 281
				sock_put(sk);
				goto begintw;
			}
			goto out;
		}
282
	}
283 284 285 286 287 288 289
	/*
	 * 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;
290 291
	sk = NULL;
out:
292
	rcu_read_unlock();
293 294 295 296
	return sk;
}
EXPORT_SYMBOL_GPL(__inet_lookup_established);

297 298 299 300 301 302 303
/* 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);
304 305
	__be32 daddr = inet->inet_rcv_saddr;
	__be32 saddr = inet->inet_daddr;
306 307
	int dif = sk->sk_bound_dev_if;
	INET_ADDR_COOKIE(acookie, saddr, daddr)
308
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
309
	struct net *net = sock_net(sk);
310 311
	unsigned int hash = inet_ehashfn(net, daddr, lport,
					 saddr, inet->inet_dport);
312
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
313
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
314
	struct sock *sk2;
315
	const struct hlist_nulls_node *node;
316
	struct inet_timewait_sock *tw;
317
	int twrefcnt = 0;
318

319
	spin_lock(lock);
320 321

	/* Check TIME-WAIT sockets first. */
322
	sk_nulls_for_each(sk2, node, &head->twchain) {
323 324
		if (sk2->sk_hash != hash)
			continue;
325

326 327 328
		if (likely(INET_TW_MATCH(sk2, net, acookie,
					 saddr, daddr, ports, dif))) {
			tw = inet_twsk(sk2);
329 330 331 332 333 334 335 336 337
			if (twsk_unique(sk, sk2, twp))
				goto unique;
			else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
338
	sk_nulls_for_each(sk2, node, &head->chain) {
339 340 341 342
		if (sk2->sk_hash != hash)
			continue;
		if (likely(INET_MATCH(sk2, net, acookie,
				      saddr, daddr, ports, dif)))
343 344 345 346 347 348
			goto not_unique;
	}

unique:
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
349 350
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
351
	sk->sk_hash = hash;
352
	WARN_ON(!sk_unhashed(sk));
353
	__sk_nulls_add_node_rcu(sk, &head->chain);
354 355 356 357
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
358
	spin_unlock(lock);
359 360
	if (twrefcnt)
		inet_twsk_put(tw);
361
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
362 363 364 365 366 367 368 369 370 371 372 373

	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:
374
	spin_unlock(lock);
375 376 377 378 379 380
	return -EADDRNOTAVAIL;
}

static inline u32 inet_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
381 382 383
	return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr,
					  inet->inet_daddr,
					  inet->inet_dport);
384 385
}

386
int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw)
387
{
388
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
389
	struct hlist_nulls_head *list;
390
	spinlock_t *lock;
391
	struct inet_ehash_bucket *head;
392
	int twrefcnt = 0;
393

394
	WARN_ON(!sk_unhashed(sk));
395 396 397 398 399 400

	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);

401
	spin_lock(lock);
402
	__sk_nulls_add_node_rcu(sk, list);
403 404 405 406
	if (tw) {
		WARN_ON(sk->sk_hash != tw->tw_hash);
		twrefcnt = inet_twsk_unhash(tw);
	}
407
	spin_unlock(lock);
408
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
409
	return twrefcnt;
410 411 412
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);

413
static void __inet_hash(struct sock *sk)
414
{
415
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
416
	struct inet_listen_hashbucket *ilb;
417 418

	if (sk->sk_state != TCP_LISTEN) {
419
		__inet_hash_nolisten(sk, NULL);
420 421 422
		return;
	}

423
	WARN_ON(!sk_unhashed(sk));
424
	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
425

426
	spin_lock(&ilb->lock);
427
	__sk_nulls_add_node_rcu(sk, &ilb->head);
428
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
429
	spin_unlock(&ilb->lock);
430
}
431 432 433 434 435 436 437 438 439 440 441 442 443

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)
{
444
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
445 446
	spinlock_t *lock;
	int done;
447 448

	if (sk_unhashed(sk))
449
		return;
450

451 452 453 454
	if (sk->sk_state == TCP_LISTEN)
		lock = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)].lock;
	else
		lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
455

456 457 458 459
	spin_lock_bh(lock);
	done =__sk_nulls_del_node_init_rcu(sk);
	if (done)
		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
460
	spin_unlock_bh(lock);
461 462
}
EXPORT_SYMBOL_GPL(inet_unhash);
463

464
int __inet_hash_connect(struct inet_timewait_death_row *death_row,
465
		struct sock *sk, u32 port_offset,
466 467
		int (*check_established)(struct inet_timewait_death_row *,
			struct sock *, __u16, struct inet_timewait_sock **),
468
		int (*hash)(struct sock *sk, struct inet_timewait_sock *twp))
469 470
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
471
	const unsigned short snum = inet_sk(sk)->inet_num;
472 473
	struct inet_bind_hashbucket *head;
	struct inet_bind_bucket *tb;
474
	int ret;
475
	struct net *net = sock_net(sk);
476
	int twrefcnt = 1;
477

478
	if (!snum) {
479
		int i, remaining, low, high, port;
480
		static u32 hint;
481
		u32 offset = hint + port_offset;
482
		struct hlist_node *node;
483
		struct inet_timewait_sock *tw = NULL;
484

485
		inet_get_local_port_range(&low, &high);
486
		remaining = (high - low) + 1;
487

488
		local_bh_disable();
489 490
		for (i = 1; i <= remaining; i++) {
			port = low + (i + offset) % remaining;
491 492
			if (inet_is_reserved_local_port(port))
				continue;
493 494
			head = &hinfo->bhash[inet_bhashfn(net, port,
					hinfo->bhash_size)];
495
			spin_lock(&head->lock);
496

497 498 499 500
			/* Does not bother with rcv_saddr checks,
			 * because the established check is already
			 * unique enough.
			 */
501
			inet_bind_bucket_for_each(tb, node, &head->chain) {
502 503
				if (net_eq(ib_net(tb), net) &&
				    tb->port == port) {
504 505
					if (tb->fastreuse >= 0)
						goto next_port;
506
					WARN_ON(hlist_empty(&tb->owners));
507 508
					if (!check_established(death_row, sk,
								port, &tw))
509 510 511 512 513
						goto ok;
					goto next_port;
				}
			}

514 515
			tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
					net, head, port);
516 517 518 519 520 521 522 523 524 525 526 527 528
			if (!tb) {
				spin_unlock(&head->lock);
				break;
			}
			tb->fastreuse = -1;
			goto ok;

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

		return -EADDRNOTAVAIL;
529 530 531 532

ok:
		hint += i;

533 534
		/* Head lock still held and bh's disabled */
		inet_bind_hash(sk, tb, port);
535
		if (sk_unhashed(sk)) {
536
			inet_sk(sk)->inet_sport = htons(port);
537
			twrefcnt += hash(sk, tw);
538
		}
539 540
		if (tw)
			twrefcnt += inet_twsk_bind_unhash(tw, hinfo);
541
		spin_unlock(&head->lock);
542

543 544
		if (tw) {
			inet_twsk_deschedule(tw, death_row);
545 546 547 548
			while (twrefcnt) {
				twrefcnt--;
				inet_twsk_put(tw);
			}
549
		}
550 551 552

		ret = 0;
		goto out;
553
	}
554

555
	head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
556
	tb  = inet_csk(sk)->icsk_bind_hash;
557 558
	spin_lock_bh(&head->lock);
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
559
		hash(sk, NULL);
560 561 562 563 564
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
		spin_unlock(&head->lock);
		/* No definite answer... Walk to established hash table */
565
		ret = check_established(death_row, sk, snum, NULL);
566 567 568 569 570
out:
		local_bh_enable();
		return ret;
	}
}
571 572 573 574 575 576 577

/*
 * Bind a port for a connect operation and hash it.
 */
int inet_hash_connect(struct inet_timewait_death_row *death_row,
		      struct sock *sk)
{
578
	return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
579 580
			__inet_check_established, __inet_hash_nolisten);
}
581
EXPORT_SYMBOL_GPL(inet_hash_connect);
582 583 584 585 586

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

587
	atomic_set(&h->bsockets, 0);
588
	for (i = 0; i < INET_LHTABLE_SIZE; i++) {
589
		spin_lock_init(&h->listening_hash[i].lock);
590 591 592
		INIT_HLIST_NULLS_HEAD(&h->listening_hash[i].head,
				      i + LISTENING_NULLS_BASE);
		}
593 594
}
EXPORT_SYMBOL_GPL(inet_hashinfo_init);