inet_hashtables.c 14.0 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
		tb->port      = snum;
		tb->fastreuse = 0;
41
		tb->num_owners = 0;
42 43 44 45 46 47 48 49 50
		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
 */
51
void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
52 53 54
{
	if (hlist_empty(&tb->owners)) {
		__hlist_del(&tb->node);
55
		release_net(ib_net(tb));
56 57 58
		kmem_cache_free(cachep, tb);
	}
}
59 60 61 62

void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
		    const unsigned short snum)
{
63 64
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;

65
	atomic_inc(&hashinfo->bsockets);
66

E
Eric Dumazet 已提交
67
	inet_sk(sk)->inet_num = snum;
68
	sk_add_bind_node(sk, &tb->owners);
69
	tb->num_owners++;
70
	inet_csk(sk)->icsk_bind_hash = tb;
71 72 73 74 75
}

/*
 * Get rid of any references to a local port held by the given sock.
 */
76
static void __inet_put_port(struct sock *sk)
77
{
78
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
E
Eric Dumazet 已提交
79
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->inet_num,
80
			hashinfo->bhash_size);
81 82 83
	struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
	struct inet_bind_bucket *tb;

84
	atomic_dec(&hashinfo->bsockets);
85

86
	spin_lock(&head->lock);
87
	tb = inet_csk(sk)->icsk_bind_hash;
88
	__sk_del_bind_node(sk);
89
	tb->num_owners--;
90
	inet_csk(sk)->icsk_bind_hash = NULL;
E
Eric Dumazet 已提交
91
	inet_sk(sk)->inet_num = 0;
92 93 94 95
	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
	spin_unlock(&head->lock);
}

96
void inet_put_port(struct sock *sk)
97 98
{
	local_bh_disable();
99
	__inet_put_port(sk);
100 101 102 103
	local_bh_enable();
}

EXPORT_SYMBOL(inet_put_port);
104

105 106 107
void __inet_inherit_port(struct sock *sk, struct sock *child)
{
	struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
E
Eric Dumazet 已提交
108
	const int bhash = inet_bhashfn(sock_net(sk), inet_sk(child)->inet_num,
109
			table->bhash_size);
110 111 112 113 114 115 116 117 118 119 120 121
	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);

122 123 124 125 126 127 128
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
129
	if (net_eq(sock_net(sk), net) && inet->inet_num == hnum &&
130
			!ipv6_only_sock(sk)) {
E
Eric Dumazet 已提交
131
		__be32 rcv_saddr = inet->inet_rcv_saddr;
132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
		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;
}

147 148 149 150 151 152
/*
 * 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.
 */
153

154

155 156
struct sock *__inet_lookup_listener(struct net *net,
				    struct inet_hashinfo *hashinfo,
157
				    const __be32 daddr, const unsigned short hnum,
158
				    const int dif)
159
{
160 161 162 163 164
	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;
165

166 167 168 169 170 171 172 173 174 175
	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;
		}
176
	}
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
	/*
	 * 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;
		}
192
	}
193 194
	rcu_read_unlock();
	return result;
195
}
196
EXPORT_SYMBOL_GPL(__inet_lookup_listener);
197

198 199
struct sock * __inet_lookup_established(struct net *net,
				  struct inet_hashinfo *hashinfo,
200 201 202 203 204 205 206
				  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;
207
	const struct hlist_nulls_node *node;
208 209 210
	/* Optimize here for direct hit, only listening connections can
	 * have wildcards anyways.
	 */
211
	unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
212
	unsigned int slot = hash & hashinfo->ehash_mask;
213
	struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
214

215 216 217
	rcu_read_lock();
begin:
	sk_nulls_for_each_rcu(sk, node, &head->chain) {
218
		if (INET_MATCH(sk, net, hash, acookie,
219 220 221 222 223 224 225 226 227 228
					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;
		}
229
	}
230 231 232 233 234 235 236
	/*
	 * 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;
237

238
begintw:
239
	/* Must check for a TIME_WAIT'er before going to listener hash. */
240
	sk_nulls_for_each_rcu(sk, node, &head->twchain) {
241
		if (INET_TW_MATCH(sk, net, hash, acookie,
242 243 244 245 246 247 248 249 250 251 252 253
					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;
		}
254
	}
255 256 257 258 259 260 261
	/*
	 * 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;
262 263
	sk = NULL;
out:
264
	rcu_read_unlock();
265 266 267 268
	return sk;
}
EXPORT_SYMBOL_GPL(__inet_lookup_established);

269 270 271 272 273 274 275
/* 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
276 277
	__be32 daddr = inet->inet_rcv_saddr;
	__be32 saddr = inet->inet_daddr;
278 279
	int dif = sk->sk_bound_dev_if;
	INET_ADDR_COOKIE(acookie, saddr, daddr)
E
Eric Dumazet 已提交
280
	const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport);
281
	struct net *net = sock_net(sk);
E
Eric Dumazet 已提交
282 283
	unsigned int hash = inet_ehashfn(net, daddr, lport,
					 saddr, inet->inet_dport);
284
	struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
285
	spinlock_t *lock = inet_ehash_lockp(hinfo, hash);
286
	struct sock *sk2;
287
	const struct hlist_nulls_node *node;
288
	struct inet_timewait_sock *tw;
289
	int twrefcnt = 0;
290

291
	spin_lock(lock);
292 293

	/* Check TIME-WAIT sockets first. */
294
	sk_nulls_for_each(sk2, node, &head->twchain) {
295 296
		tw = inet_twsk(sk2);

297 298
		if (INET_TW_MATCH(sk2, net, hash, acookie,
					saddr, daddr, ports, dif)) {
299 300 301 302 303 304 305 306 307
			if (twsk_unique(sk, sk2, twp))
				goto unique;
			else
				goto not_unique;
		}
	}
	tw = NULL;

	/* And established part... */
308
	sk_nulls_for_each(sk2, node, &head->chain) {
309 310
		if (INET_MATCH(sk2, net, hash, acookie,
					saddr, daddr, ports, dif))
311 312 313 314 315 316
			goto not_unique;
	}

unique:
	/* Must record num and sport now. Otherwise we will see
	 * in hash table socket with a funny identity. */
E
Eric Dumazet 已提交
317 318
	inet->inet_num = lport;
	inet->inet_sport = htons(lport);
319
	sk->sk_hash = hash;
320
	WARN_ON(!sk_unhashed(sk));
321
	__sk_nulls_add_node_rcu(sk, &head->chain);
322 323 324 325
	if (tw) {
		twrefcnt = inet_twsk_unhash(tw);
		NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
	}
326
	spin_unlock(lock);
327 328
	if (twrefcnt)
		inet_twsk_put(tw);
329
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
330 331 332 333 334 335 336 337 338 339 340 341

	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:
342
	spin_unlock(lock);
343 344 345 346 347 348
	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
349 350 351
	return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr,
					  inet->inet_daddr,
					  inet->inet_dport);
352 353
}

354
void __inet_hash_nolisten(struct sock *sk)
355
{
356
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
357
	struct hlist_nulls_head *list;
358
	spinlock_t *lock;
359 360
	struct inet_ehash_bucket *head;

361
	WARN_ON(!sk_unhashed(sk));
362 363 364 365 366 367

	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);

368
	spin_lock(lock);
369
	__sk_nulls_add_node_rcu(sk, list);
370
	spin_unlock(lock);
371
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
372 373 374
}
EXPORT_SYMBOL_GPL(__inet_hash_nolisten);

375
static void __inet_hash(struct sock *sk)
376
{
377
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
378
	struct inet_listen_hashbucket *ilb;
379 380

	if (sk->sk_state != TCP_LISTEN) {
381
		__inet_hash_nolisten(sk);
382 383 384
		return;
	}

385
	WARN_ON(!sk_unhashed(sk));
386
	ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
387

388
	spin_lock(&ilb->lock);
389
	__sk_nulls_add_node_rcu(sk, &ilb->head);
390
	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
391
	spin_unlock(&ilb->lock);
392
}
393 394 395 396 397 398 399 400 401 402 403 404 405

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)
{
406
	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
407 408
	spinlock_t *lock;
	int done;
409 410

	if (sk_unhashed(sk))
411
		return;
412

413 414 415 416
	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);
417

418 419 420 421
	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);
422
	spin_unlock_bh(lock);
423 424
}
EXPORT_SYMBOL_GPL(inet_unhash);
425

426
int __inet_hash_connect(struct inet_timewait_death_row *death_row,
427
		struct sock *sk, u32 port_offset,
428 429
		int (*check_established)(struct inet_timewait_death_row *,
			struct sock *, __u16, struct inet_timewait_sock **),
430
		void (*hash)(struct sock *sk))
431 432
{
	struct inet_hashinfo *hinfo = death_row->hashinfo;
E
Eric Dumazet 已提交
433
	const unsigned short snum = inet_sk(sk)->inet_num;
434 435
	struct inet_bind_hashbucket *head;
	struct inet_bind_bucket *tb;
436
	int ret;
437
	struct net *net = sock_net(sk);
438

439
	if (!snum) {
440
		int i, remaining, low, high, port;
441
		static u32 hint;
442
		u32 offset = hint + port_offset;
443
		struct hlist_node *node;
444
		struct inet_timewait_sock *tw = NULL;
445

446
		inet_get_local_port_range(&low, &high);
447
		remaining = (high - low) + 1;
448

449
		local_bh_disable();
450 451
		for (i = 1; i <= remaining; i++) {
			port = low + (i + offset) % remaining;
452 453
			head = &hinfo->bhash[inet_bhashfn(net, port,
					hinfo->bhash_size)];
454
			spin_lock(&head->lock);
455

456 457 458 459
			/* Does not bother with rcv_saddr checks,
			 * because the established check is already
			 * unique enough.
			 */
460
			inet_bind_bucket_for_each(tb, node, &head->chain) {
O
Octavian Purdila 已提交
461 462
				if (net_eq(ib_net(tb), net) &&
				    tb->port == port) {
463 464
					if (tb->fastreuse >= 0)
						goto next_port;
465
					WARN_ON(hlist_empty(&tb->owners));
466 467
					if (!check_established(death_row, sk,
								port, &tw))
468 469 470 471 472
						goto ok;
					goto next_port;
				}
			}

473 474
			tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
					net, head, port);
475 476 477 478 479 480 481 482 483 484 485 486 487
			if (!tb) {
				spin_unlock(&head->lock);
				break;
			}
			tb->fastreuse = -1;
			goto ok;

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

		return -EADDRNOTAVAIL;
488 489 490 491

ok:
		hint += i;

492 493
		/* Head lock still held and bh's disabled */
		inet_bind_hash(sk, tb, port);
494
		if (sk_unhashed(sk)) {
E
Eric Dumazet 已提交
495
			inet_sk(sk)->inet_sport = htons(port);
496
			hash(sk);
497 498
		}
		spin_unlock(&head->lock);
499

500 501 502 503
		if (tw) {
			inet_twsk_deschedule(tw, death_row);
			inet_twsk_put(tw);
		}
504 505 506

		ret = 0;
		goto out;
507
	}
508

509
	head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
510
	tb  = inet_csk(sk)->icsk_bind_hash;
511 512
	spin_lock_bh(&head->lock);
	if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
513
		hash(sk);
514 515 516 517 518
		spin_unlock_bh(&head->lock);
		return 0;
	} else {
		spin_unlock(&head->lock);
		/* No definite answer... Walk to established hash table */
519
		ret = check_established(death_row, sk, snum, NULL);
520 521 522 523 524
out:
		local_bh_enable();
		return ret;
	}
}
525 526 527 528 529 530 531

/*
 * Bind a port for a connect operation and hash it.
 */
int inet_hash_connect(struct inet_timewait_death_row *death_row,
		      struct sock *sk)
{
532
	return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
533 534
			__inet_check_established, __inet_hash_nolisten);
}
535 536

EXPORT_SYMBOL_GPL(inet_hash_connect);
537 538 539 540 541

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

542
	atomic_set(&h->bsockets, 0);
543
	for (i = 0; i < INET_LHTABLE_SIZE; i++) {
544
		spin_lock_init(&h->listening_hash[i].lock);
545 546 547
		INIT_HLIST_NULLS_HEAD(&h->listening_hash[i].head,
				      i + LISTENING_NULLS_BASE);
		}
548 549 550
}

EXPORT_SYMBOL_GPL(inet_hashinfo_init);