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

/*
 * 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->fastreuseport = 0;
43
		tb->num_owners = 0;
44
45
46
47
48
49
50
51
52
		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
 */
53
void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
54
55
56
{
	if (hlist_empty(&tb->owners)) {
		__hlist_del(&tb->node);
57
		release_net(ib_net(tb));
58
59
60
		kmem_cache_free(cachep, tb);
	}
}
61
62
63
64

void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
		    const unsigned short snum)
{
65
66
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;

67
	atomic_inc(&hashinfo->bsockets);
68

69
	inet_sk(sk)->inet_num = snum;
70
	sk_add_bind_node(sk, &tb->owners);
71
	tb->num_owners++;
72
	inet_csk(sk)->icsk_bind_hash = tb;
73
74
75
76
77
}

/*
 * Get rid of any references to a local port held by the given sock.
 */
78
static void __inet_put_port(struct sock *sk)
79
{
80
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
81
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->inet_num,
82
			hashinfo->bhash_size);
83
84
85
	struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
	struct inet_bind_bucket *tb;

86
	atomic_dec(&hashinfo->bsockets);
87

88
	spin_lock(&head->lock);
89
	tb = inet_csk(sk)->icsk_bind_hash;
90
	__sk_del_bind_node(sk);
91
	tb->num_owners--;
92
	inet_csk(sk)->icsk_bind_hash = NULL;
93
	inet_sk(sk)->inet_num = 0;
94
95
96
97
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&head->lock);
}

98
void inet_put_port(struct sock *sk)
99
100
{
	local_bh_disable();
101
	__inet_put_port(sk);
102
103
104
	local_bh_enable();
}
EXPORT_SYMBOL(inet_put_port);
105

106
int __inet_inherit_port(struct sock *sk, struct sock *child)
107
108
{
	struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
109
110
	unsigned short port = inet_sk(child)->inet_num;
	const int bhash = inet_bhashfn(sock_net(sk), port,
111
			table->bhash_size);
112
113
114
115
116
	struct inet_bind_hashbucket *head = &table->bhash[bhash];
	struct inet_bind_bucket *tb;

	spin_lock(&head->lock);
	tb = inet_csk(sk)->icsk_bind_hash;
117
118
119
120
121
122
	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. */
123
		inet_bind_bucket_for_each(tb, &head->chain) {
124
125
126
127
			if (net_eq(ib_net(tb), sock_net(sk)) &&
			    tb->port == port)
				break;
		}
128
		if (!tb) {
129
130
131
132
133
134
135
136
			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
		score = sk->sk_family == PF_INET ? 2 : 1;
155
156
157
		if (rcv_saddr) {
			if (rcv_saddr != daddr)
				return -1;
158
			score += 4;
159
160
161
162
		}
		if (sk->sk_bound_dev_if) {
			if (sk->sk_bound_dev_if != dif)
				return -1;
163
			score += 4;
164
165
166
167
168
		}
	}
	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 saddr, __be16 sport,
180
				    const __be32 daddr, const unsigned short hnum,
181
				    const int dif)
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];
187
188
	int score, hiscore, matches = 0, reuseport = 0;
	u32 phash = 0;
189

190
191
192
	rcu_read_lock();
begin:
	result = NULL;
193
	hiscore = 0;
194
195
196
197
198
	sk_nulls_for_each_rcu(sk, node, &ilb->head) {
		score = compute_score(sk, net, hnum, daddr, dif);
		if (score > hiscore) {
			result = sk;
			hiscore = score;
199
200
201
202
203
204
205
206
207
208
209
			reuseport = sk->sk_reuseport;
			if (reuseport) {
				phash = inet_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);
210
		}
211
	}
212
213
214
215
216
217
218
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) != 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;
		}
227
	}
228
229
	rcu_read_unlock();
	return result;
230
}
231
EXPORT_SYMBOL_GPL(__inet_lookup_listener);
232

Daniel Baluta's avatar
Daniel Baluta committed
233
struct sock *__inet_lookup_established(struct net *net,
234
				  struct inet_hashinfo *hashinfo,
235
236
237
238
239
240
241
				  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;
242
	const struct hlist_nulls_node *node;
243
244
245
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
246
	unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
247
	unsigned int slot = hash & hashinfo->ehash_mask;
248
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
249

250
251
252
	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
253
254
255
256
		if (sk->sk_hash != hash)
			continue;
		if (likely(INET_MATCH(sk, net, acookie,
				      saddr, daddr, ports, dif))) {
257
258
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt)))
				goto begintw;
259
260
			if (unlikely(!INET_MATCH(sk, net, acookie,
						 saddr, daddr, ports, dif))) {
261
262
263
264
265
				sock_put(sk);
				goto begin;
			}
			goto out;
		}
266
	}
267
268
269
270
271
272
273
	/*
	 * 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;
274

275
begintw:
276
	/* Must check for a TIME_WAIT'er before going to listener hash. */
277
	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
278
279
280
281
282
		if (sk->sk_hash != hash)
			continue;
		if (likely(INET_TW_MATCH(sk, net, acookie,
					 saddr, daddr, ports,
					 dif))) {
283
284
285
286
			if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) {
				sk = NULL;
				goto out;
			}
287
288
289
			if (unlikely(!INET_TW_MATCH(sk, net, acookie,
						    saddr, daddr, ports,
						    dif))) {
290
291
292
293
294
				sock_put(sk);
				goto begintw;
			}
			goto out;
		}
295
	}
296
297
298
299
300
301
302
	/*
	 * 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;
303
304
	sk = NULL;
out:
305
	rcu_read_unlock();
306
307
308
309
	return sk;
}
EXPORT_SYMBOL_GPL(__inet_lookup_established);

310
311
312
313
314
315
316
/* 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);
317
318
	__be32 daddr = inet->inet_rcv_saddr;
	__be32 saddr = inet->inet_daddr;
319
320
	int dif = sk->sk_bound_dev_if;
	INET_ADDR_COOKIE(acookie, saddr, daddr)
321
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
322
	struct net *net = sock_net(sk);
323
324
	unsigned int hash = inet_ehashfn(net, daddr, lport,
					 saddr, inet->inet_dport);
325
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
326
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
327
	struct sock *sk2;
328
	const struct hlist_nulls_node *node;
329
	struct inet_timewait_sock *tw;
330
	int twrefcnt = 0;
331

332
	spin_lock(lock);
333
334

	/* Check TIME-WAIT sockets first. */
335
	sk_nulls_for_each(sk2, node, &head->twchain) {
336
337
		if (sk2->sk_hash != hash)
			continue;
338

339
340
341
		if (likely(INET_TW_MATCH(sk2, net, acookie,
					 saddr, daddr, ports, dif))) {
			tw = inet_twsk(sk2);
342
343
344
345
346
347
348
349
350
			if (twsk_unique(sk, sk2, twp))
				goto unique;
			else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
351
	sk_nulls_for_each(sk2, node, &head->chain) {
352
353
354
355
		if (sk2->sk_hash != hash)
			continue;
		if (likely(INET_MATCH(sk2, net, acookie,
				      saddr, daddr, ports, dif)))
356
357
358
359
360
361
			goto not_unique;
	}

unique:
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
362
363
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
364
	sk->sk_hash = hash;
365
	WARN_ON(!sk_unhashed(sk));
366
	__sk_nulls_add_node_rcu(sk, &head->chain);
367
368
369
370
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
371
	spin_unlock(lock);
372
373
	if (twrefcnt)
		inet_twsk_put(tw);
374
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
375
376
377
378
379
380
381
382
383
384
385
386

	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:
387
	spin_unlock(lock);
388
389
390
391
392
393
	return -EADDRNOTAVAIL;
}

static inline u32 inet_sk_port_offset(const struct sock *sk)
{
	const struct inet_sock *inet = inet_sk(sk);
394
395
396
	return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr,
					  inet->inet_daddr,
					  inet->inet_dport);
397
398
}

399
int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw)
400
{
401
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
402
	struct hlist_nulls_head *list;
403
	spinlock_t *lock;
404
	struct inet_ehash_bucket *head;
405
	int twrefcnt = 0;
406

407
	WARN_ON(!sk_unhashed(sk));
408
409
410
411
412
413

	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);

414
	spin_lock(lock);
415
	__sk_nulls_add_node_rcu(sk, list);
416
417
418
419
	if (tw) {
		WARN_ON(sk->sk_hash != tw->tw_hash);
		twrefcnt = inet_twsk_unhash(tw);
	}
420
	spin_unlock(lock);
421
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
422
	return twrefcnt;
423
424
425
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);

426
static void __inet_hash(struct sock *sk)
427
{
428
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
429
	struct inet_listen_hashbucket *ilb;
430
431

	if (sk->sk_state != TCP_LISTEN) {
432
		__inet_hash_nolisten(sk, NULL);
433
434
435
		return;
	}

436
	WARN_ON(!sk_unhashed(sk));
437
	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
438

439
	spin_lock(&ilb->lock);
440
	__sk_nulls_add_node_rcu(sk, &ilb->head);
441
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
442
	spin_unlock(&ilb->lock);
443
}
444
445
446
447
448
449
450
451
452
453
454
455
456

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)
{
457
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
458
459
	spinlock_t *lock;
	int done;
460
461

	if (sk_unhashed(sk))
462
		return;
463

464
465
466
467
	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);
468

469
	spin_lock_bh(lock);
470
	done = __sk_nulls_del_node_init_rcu(sk);
471
472
	if (done)
		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
473
	spin_unlock_bh(lock);
474
475
}
EXPORT_SYMBOL_GPL(inet_unhash);
476

477
int __inet_hash_connect(struct inet_timewait_death_row *death_row,
478
		struct sock *sk, u32 port_offset,
479
480
		int (*check_established)(struct inet_timewait_death_row *,
			struct sock *, __u16, struct inet_timewait_sock **),
481
		int (*hash)(struct sock *sk, struct inet_timewait_sock *twp))
482
483
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
484
	const unsigned short snum = inet_sk(sk)->inet_num;
485
486
	struct inet_bind_hashbucket *head;
	struct inet_bind_bucket *tb;
487
	int ret;
488
	struct net *net = sock_net(sk);
489
	int twrefcnt = 1;
490

491
	if (!snum) {
492
		int i, remaining, low, high, port;
493
		static u32 hint;
494
		u32 offset = hint + port_offset;
495
		struct inet_timewait_sock *tw = NULL;
496

497
		inet_get_local_port_range(&low, &high);
498
		remaining = (high - low) + 1;
499

500
		local_bh_disable();
501
502
		for (i = 1; i <= remaining; i++) {
			port = low + (i + offset) % remaining;
503
504
			if (inet_is_reserved_local_port(port))
				continue;
505
506
			head = &hinfo->bhash[inet_bhashfn(net, port,
					hinfo->bhash_size)];
507
			spin_lock(&head->lock);
508

509
510
511
512
			/* Does not bother with rcv_saddr checks,
			 * because the established check is already
			 * unique enough.
			 */
513
			inet_bind_bucket_for_each(tb, &head->chain) {
514
515
				if (net_eq(ib_net(tb), net) &&
				    tb->port == port) {
516
517
					if (tb->fastreuse >= 0 ||
					    tb->fastreuseport >= 0)
518
						goto next_port;
519
					WARN_ON(hlist_empty(&tb->owners));
520
521
					if (!check_established(death_row, sk,
								port, &tw))
522
523
524
525
526
						goto ok;
					goto next_port;
				}
			}

527
528
			tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
					net, head, port);
529
530
531
532
533
			if (!tb) {
				spin_unlock(&head->lock);
				break;
			}
			tb->fastreuse = -1;
534
			tb->fastreuseport = -1;
535
536
537
538
539
540
541
542
			goto ok;

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

		return -EADDRNOTAVAIL;
543
544
545
546

ok:
		hint += i;

547
548
		/* Head lock still held and bh's disabled */
		inet_bind_hash(sk, tb, port);
549
		if (sk_unhashed(sk)) {
550
			inet_sk(sk)->inet_sport = htons(port);
551
			twrefcnt += hash(sk, tw);
552
		}
553
554
		if (tw)
			twrefcnt += inet_twsk_bind_unhash(tw, hinfo);
555
		spin_unlock(&head->lock);
556

557
558
		if (tw) {
			inet_twsk_deschedule(tw, death_row);
559
560
561
562
			while (twrefcnt) {
				twrefcnt--;
				inet_twsk_put(tw);
			}
563
		}
564
565
566

		ret = 0;
		goto out;
567
	}
568

569
	head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
570
	tb  = inet_csk(sk)->icsk_bind_hash;
571
572
	spin_lock_bh(&head->lock);
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
573
		hash(sk, NULL);
574
575
576
577
578
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
		spin_unlock(&head->lock);
		/* No definite answer... Walk to established hash table */
579
		ret = check_established(death_row, sk, snum, NULL);
580
581
582
583
584
out:
		local_bh_enable();
		return ret;
	}
}
585
586
587
588
589
590
591

/*
 * Bind a port for a connect operation and hash it.
 */
int inet_hash_connect(struct inet_timewait_death_row *death_row,
		      struct sock *sk)
{
592
	return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
593
594
			__inet_check_established, __inet_hash_nolisten);
}
595
EXPORT_SYMBOL_GPL(inet_hash_connect);
596
597
598
599
600

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

601
	atomic_set(&h->bsockets, 0);
602
	for (i = 0; i < INET_LHTABLE_SIZE; i++) {
603
		spin_lock_init(&h->listening_hash[i].lock);
604
605
606
		INIT_HLIST_NULLS_HEAD(&h->listening_hash[i].head,
				      i + LISTENING_NULLS_BASE);
		}
607
608
}
EXPORT_SYMBOL_GPL(inet_hashinfo_init);