inet_hashtables.c 15.6 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
static unsigned int inet_ehashfn(struct net *net, const __be32 laddr,
				 const __u16 lport, const __be32 faddr,
				 const __be16 fport)
{
31 32 33 34
	static u32 inet_ehash_secret __read_mostly;

	net_get_random_once(&inet_ehash_secret, sizeof(inet_ehash_secret));

35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
	return __inet_ehashfn(laddr, lport, faddr, fport,
			      inet_ehash_secret + net_hash_mix(net));
}


static unsigned int inet_sk_ehashfn(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
	const __be32 laddr = inet->inet_rcv_saddr;
	const __u16 lport = inet->inet_num;
	const __be32 faddr = inet->inet_daddr;
	const __be16 fport = inet->inet_dport;
	struct net *net = sock_net(sk);

	return inet_ehashfn(net, laddr, lport, faddr, fport);
}

52 53 54 55
/*
 * Allocate and initialize a new local port bind bucket.
 * The bindhash mutex for snum's hash chain must be held here.
 */
56
struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep,
57
						 struct net *net,
58 59 60
						 struct inet_bind_hashbucket *head,
						 const unsigned short snum)
{
61
	struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC);
62 63

	if (tb != NULL) {
64
		write_pnet(&tb->ib_net, net);
65 66
		tb->port      = snum;
		tb->fastreuse = 0;
67
		tb->fastreuseport = 0;
68
		tb->num_owners = 0;
69 70 71 72 73 74 75 76 77
		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
 */
78
void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
79 80 81 82 83 84
{
	if (hlist_empty(&tb->owners)) {
		__hlist_del(&tb->node);
		kmem_cache_free(cachep, tb);
	}
}
85 86 87 88

void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
		    const unsigned short snum)
{
89 90
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;

91
	atomic_inc(&hashinfo->bsockets);
92

E
Eric Dumazet 已提交
93
	inet_sk(sk)->inet_num = snum;
94
	sk_add_bind_node(sk, &tb->owners);
95
	tb->num_owners++;
96
	inet_csk(sk)->icsk_bind_hash = tb;
97 98 99 100 101
}

/*
 * Get rid of any references to a local port held by the given sock.
 */
102
static void __inet_put_port(struct sock *sk)
103
{
104
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
E
Eric Dumazet 已提交
105
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->inet_num,
106
			hashinfo->bhash_size);
107 108 109
	struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
	struct inet_bind_bucket *tb;

110
	atomic_dec(&hashinfo->bsockets);
111

112
	spin_lock(&head->lock);
113
	tb = inet_csk(sk)->icsk_bind_hash;
114
	__sk_del_bind_node(sk);
115
	tb->num_owners--;
116
	inet_csk(sk)->icsk_bind_hash = NULL;
E
Eric Dumazet 已提交
117
	inet_sk(sk)->inet_num = 0;
118 119 120 121
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&head->lock);
}

122
void inet_put_port(struct sock *sk)
123 124
{
	local_bh_disable();
125
	__inet_put_port(sk);
126 127 128
	local_bh_enable();
}
EXPORT_SYMBOL(inet_put_port);
129

130
int __inet_inherit_port(struct sock *sk, struct sock *child)
131 132
{
	struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
133 134
	unsigned short port = inet_sk(child)->inet_num;
	const int bhash = inet_bhashfn(sock_net(sk), port,
135
			table->bhash_size);
136 137 138 139 140
	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
	tb = inet_csk(sk)->icsk_bind_hash;
141 142 143 144 145 146
	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. */
147
		inet_bind_bucket_for_each(tb, &head->chain) {
148 149 150 151
			if (net_eq(ib_net(tb), sock_net(sk)) &&
			    tb->port == port)
				break;
		}
152
		if (!tb) {
153 154 155 156 157 158 159 160
			tb = inet_bind_bucket_create(table->bind_bucket_cachep,
						     sock_net(sk), head, port);
			if (!tb) {
				spin_unlock(&head->lock);
				return -ENOMEM;
			}
		}
	}
161
	inet_bind_hash(child, tb, port);
162
	spin_unlock(&head->lock);
163 164

	return 0;
165 166 167
}
EXPORT_SYMBOL_GPL(__inet_inherit_port);

168 169 170 171 172 173 174
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);

E
Eric Dumazet 已提交
175
	if (net_eq(sock_net(sk), net) && inet->inet_num == hnum &&
176
			!ipv6_only_sock(sk)) {
E
Eric Dumazet 已提交
177
		__be32 rcv_saddr = inet->inet_rcv_saddr;
178
		score = sk->sk_family == PF_INET ? 2 : 1;
179 180 181
		if (rcv_saddr) {
			if (rcv_saddr != daddr)
				return -1;
182
			score += 4;
183 184 185 186
		}
		if (sk->sk_bound_dev_if) {
			if (sk->sk_bound_dev_if != dif)
				return -1;
187
			score += 4;
188 189 190 191 192
		}
	}
	return score;
}

193 194 195 196 197 198
/*
 * 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.
 */
199

200

201 202
struct sock *__inet_lookup_listener(struct net *net,
				    struct inet_hashinfo *hashinfo,
203
				    const __be32 saddr, __be16 sport,
204
				    const __be32 daddr, const unsigned short hnum,
205
				    const int dif)
206
{
207 208 209 210
	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];
211 212
	int score, hiscore, matches = 0, reuseport = 0;
	u32 phash = 0;
213

214 215 216
	rcu_read_lock();
begin:
	result = NULL;
217
	hiscore = 0;
218 219 220 221 222
	sk_nulls_for_each_rcu(sk, node, &ilb->head) {
		score = compute_score(sk, net, hnum, daddr, dif);
		if (score > hiscore) {
			result = sk;
			hiscore = score;
223 224 225 226 227 228 229 230
			reuseport = sk->sk_reuseport;
			if (reuseport) {
				phash = inet_ehashfn(net, daddr, hnum,
						     saddr, sport);
				matches = 1;
			}
		} else if (score == hiscore && reuseport) {
			matches++;
231
			if (reciprocal_scale(phash, matches) == 0)
232 233
				result = sk;
			phash = next_pseudo_random32(phash);
234
		}
235
	}
236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
	/*
	 * 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;
		}
251
	}
252 253
	rcu_read_unlock();
	return result;
254
}
255
EXPORT_SYMBOL_GPL(__inet_lookup_listener);
256

E
Eric Dumazet 已提交
257 258 259 260 261 262 263 264 265 266 267 268 269
/* All sockets share common refcount, but have different destructors */
void sock_gen_put(struct sock *sk)
{
	if (!atomic_dec_and_test(&sk->sk_refcnt))
		return;

	if (sk->sk_state == TCP_TIME_WAIT)
		inet_twsk_free(inet_twsk(sk));
	else
		sk_free(sk);
}
EXPORT_SYMBOL_GPL(sock_gen_put);

D
Daniel Baluta 已提交
270
struct sock *__inet_lookup_established(struct net *net,
271
				  struct inet_hashinfo *hashinfo,
272 273 274 275
				  const __be32 saddr, const __be16 sport,
				  const __be32 daddr, const u16 hnum,
				  const int dif)
{
276
	INET_ADDR_COOKIE(acookie, saddr, daddr);
277 278
	const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
	struct sock *sk;
279
	const struct hlist_nulls_node *node;
280 281 282
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
283
	unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
284
	unsigned int slot = hash & hashinfo->ehash_mask;
285
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
286

287 288 289
	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
290 291 292 293
		if (sk->sk_hash != hash)
			continue;
		if (likely(INET_MATCH(sk, net, acookie,
				      saddr, daddr, ports, dif))) {
294
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
E
Eric Dumazet 已提交
295
				goto out;
296 297
			if (unlikely(!INET_MATCH(sk, net, acookie,
						 saddr, daddr, ports, dif))) {
E
Eric Dumazet 已提交
298
				sock_gen_put(sk);
299 300
				goto begin;
			}
E
Eric Dumazet 已提交
301
			goto found;
302
		}
303
	}
304 305 306 307 308 309 310
	/*
	 * 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;
311
out:
E
Eric Dumazet 已提交
312 313
	sk = NULL;
found:
314
	rcu_read_unlock();
315 316 317 318
	return sk;
}
EXPORT_SYMBOL_GPL(__inet_lookup_established);

319 320 321 322 323 324 325
/* 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);
E
Eric Dumazet 已提交
326 327
	__be32 daddr = inet->inet_rcv_saddr;
	__be32 saddr = inet->inet_daddr;
328
	int dif = sk->sk_bound_dev_if;
329
	INET_ADDR_COOKIE(acookie, saddr, daddr);
E
Eric Dumazet 已提交
330
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
331
	struct net *net = sock_net(sk);
E
Eric Dumazet 已提交
332 333
	unsigned int hash = inet_ehashfn(net, daddr, lport,
					 saddr, inet->inet_dport);
334
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
335
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
336
	struct sock *sk2;
337
	const struct hlist_nulls_node *node;
E
Eric Dumazet 已提交
338
	struct inet_timewait_sock *tw = NULL;
339
	int twrefcnt = 0;
340

341
	spin_lock(lock);
342

343
	sk_nulls_for_each(sk2, node, &head->chain) {
344 345
		if (sk2->sk_hash != hash)
			continue;
E
Eric Dumazet 已提交
346

347
		if (likely(INET_MATCH(sk2, net, acookie,
E
Eric Dumazet 已提交
348 349 350 351 352 353
					 saddr, daddr, ports, dif))) {
			if (sk2->sk_state == TCP_TIME_WAIT) {
				tw = inet_twsk(sk2);
				if (twsk_unique(sk, sk2, twp))
					break;
			}
354
			goto not_unique;
E
Eric Dumazet 已提交
355
		}
356 357 358
	}

	/* Must record num and sport now. Otherwise we will see
E
Eric Dumazet 已提交
359 360
	 * in hash table socket with a funny identity.
	 */
E
Eric Dumazet 已提交
361 362
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
363
	sk->sk_hash = hash;
364
	WARN_ON(!sk_unhashed(sk));
365
	__sk_nulls_add_node_rcu(sk, &head->chain);
366 367 368 369
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
370
	spin_unlock(lock);
371 372
	if (twrefcnt)
		inet_twsk_put(tw);
373
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
374 375 376 377 378 379 380 381 382 383 384 385

	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:
386
	spin_unlock(lock);
387 388 389 390 391 392
	return -EADDRNOTAVAIL;
}

static inline u32 inet_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
E
Eric Dumazet 已提交
393 394 395
	return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr,
					  inet->inet_daddr,
					  inet->inet_dport);
396 397
}

398
int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw)
399
{
400
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
401
	struct hlist_nulls_head *list;
402
	spinlock_t *lock;
403
	struct inet_ehash_bucket *head;
404
	int twrefcnt = 0;
405

406
	WARN_ON(!sk_unhashed(sk));
407 408 409 410 411 412

	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);

413
	spin_lock(lock);
414
	__sk_nulls_add_node_rcu(sk, list);
415 416 417 418
	if (tw) {
		WARN_ON(sk->sk_hash != tw->tw_hash);
		twrefcnt = inet_twsk_unhash(tw);
	}
419
	spin_unlock(lock);
420
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
421
	return twrefcnt;
422 423 424
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);

425
static void __inet_hash(struct sock *sk)
426
{
427
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
428
	struct inet_listen_hashbucket *ilb;
429 430

	if (sk->sk_state != TCP_LISTEN) {
431
		__inet_hash_nolisten(sk, NULL);
432 433 434
		return;
	}

435
	WARN_ON(!sk_unhashed(sk));
436
	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
437

438
	spin_lock(&ilb->lock);
439
	__sk_nulls_add_node_rcu(sk, &ilb->head);
440
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
441
	spin_unlock(&ilb->lock);
442
}
443 444 445 446 447 448 449 450 451 452 453 454 455

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)
{
456
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
457 458
	spinlock_t *lock;
	int done;
459 460

	if (sk_unhashed(sk))
461
		return;
462

463 464 465 466
	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);
467

468
	spin_lock_bh(lock);
C
Camelia Groza 已提交
469
	done = __sk_nulls_del_node_init_rcu(sk);
470 471
	if (done)
		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
472
	spin_unlock_bh(lock);
473 474
}
EXPORT_SYMBOL_GPL(inet_unhash);
475

476
int __inet_hash_connect(struct inet_timewait_death_row *death_row,
477
		struct sock *sk, u32 port_offset,
478 479
		int (*check_established)(struct inet_timewait_death_row *,
			struct sock *, __u16, struct inet_timewait_sock **),
480
		int (*hash)(struct sock *sk, struct inet_timewait_sock *twp))
481 482
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
E
Eric Dumazet 已提交
483
	const unsigned short snum = inet_sk(sk)->inet_num;
484 485
	struct inet_bind_hashbucket *head;
	struct inet_bind_bucket *tb;
486
	int ret;
487
	struct net *net = sock_net(sk);
488
	int twrefcnt = 1;
489

490
	if (!snum) {
491
		int i, remaining, low, high, port;
492
		static u32 hint;
493
		u32 offset = hint + port_offset;
494
		struct inet_timewait_sock *tw = NULL;
495

496
		inet_get_local_port_range(net, &low, &high);
497
		remaining = (high - low) + 1;
498

499
		local_bh_disable();
500 501
		for (i = 1; i <= remaining; i++) {
			port = low + (i + offset) % remaining;
502
			if (inet_is_local_reserved_port(net, port))
503
				continue;
504 505
			head = &hinfo->bhash[inet_bhashfn(net, port,
					hinfo->bhash_size)];
506
			spin_lock(&head->lock);
507

508 509 510 511
			/* Does not bother with rcv_saddr checks,
			 * because the established check is already
			 * unique enough.
			 */
512
			inet_bind_bucket_for_each(tb, &head->chain) {
O
Octavian Purdila 已提交
513 514
				if (net_eq(ib_net(tb), net) &&
				    tb->port == port) {
515 516
					if (tb->fastreuse >= 0 ||
					    tb->fastreuseport >= 0)
517
						goto next_port;
518
					WARN_ON(hlist_empty(&tb->owners));
519 520
					if (!check_established(death_row, sk,
								port, &tw))
521 522 523 524 525
						goto ok;
					goto next_port;
				}
			}

526 527
			tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
					net, head, port);
528 529 530 531 532
			if (!tb) {
				spin_unlock(&head->lock);
				break;
			}
			tb->fastreuse = -1;
533
			tb->fastreuseport = -1;
534 535 536 537 538 539 540 541
			goto ok;

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

		return -EADDRNOTAVAIL;
542 543 544 545

ok:
		hint += i;

546 547
		/* Head lock still held and bh's disabled */
		inet_bind_hash(sk, tb, port);
548
		if (sk_unhashed(sk)) {
E
Eric Dumazet 已提交
549
			inet_sk(sk)->inet_sport = htons(port);
550
			twrefcnt += hash(sk, tw);
551
		}
552 553
		if (tw)
			twrefcnt += inet_twsk_bind_unhash(tw, hinfo);
554
		spin_unlock(&head->lock);
555

556 557
		if (tw) {
			inet_twsk_deschedule(tw, death_row);
558 559 560 561
			while (twrefcnt) {
				twrefcnt--;
				inet_twsk_put(tw);
			}
562
		}
563 564 565

		ret = 0;
		goto out;
566
	}
567

568
	head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
569
	tb  = inet_csk(sk)->icsk_bind_hash;
570 571
	spin_lock_bh(&head->lock);
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
572
		hash(sk, NULL);
573 574 575 576 577
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
		spin_unlock(&head->lock);
		/* No definite answer... Walk to established hash table */
578
		ret = check_established(death_row, sk, snum, NULL);
579 580 581 582 583
out:
		local_bh_enable();
		return ret;
	}
}
584 585 586 587 588 589 590

/*
 * Bind a port for a connect operation and hash it.
 */
int inet_hash_connect(struct inet_timewait_death_row *death_row,
		      struct sock *sk)
{
591
	return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
592 593
			__inet_check_established, __inet_hash_nolisten);
}
594
EXPORT_SYMBOL_GPL(inet_hash_connect);
595 596 597 598 599

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

600
	atomic_set(&h->bsockets, 0);
601
	for (i = 0; i < INET_LHTABLE_SIZE; i++) {
602
		spin_lock_init(&h->listening_hash[i].lock);
603 604 605
		INIT_HLIST_NULLS_HEAD(&h->listening_hash[i].head,
				      i + LISTENING_NULLS_BASE);
		}
606 607
}
EXPORT_SYMBOL_GPL(inet_hashinfo_init);