Commit | Line | Data |
---|---|---|
77d8bf9c ACM |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * Generic INET transport hashtables | |
7 | * | |
8 | * Authors: Lotsa people, from code originally in tcp | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public License | |
12 | * as published by the Free Software Foundation; either version | |
13 | * 2 of the License, or (at your option) any later version. | |
14 | */ | |
15 | ||
2d8c4ce5 | 16 | #include <linux/module.h> |
a7f5e7f1 | 17 | #include <linux/random.h> |
f3f05f70 | 18 | #include <linux/sched.h> |
77d8bf9c | 19 | #include <linux/slab.h> |
f3f05f70 | 20 | #include <linux/wait.h> |
77d8bf9c | 21 | |
463c84b9 | 22 | #include <net/inet_connection_sock.h> |
77d8bf9c | 23 | #include <net/inet_hashtables.h> |
6e5714ea | 24 | #include <net/secure_seq.h> |
a7f5e7f1 | 25 | #include <net/ip.h> |
77d8bf9c ACM |
26 | |
27 | /* | |
28 | * Allocate and initialize a new local port bind bucket. | |
29 | * The bindhash mutex for snum's hash chain must be held here. | |
30 | */ | |
e18b890b | 31 | struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep, |
941b1d22 | 32 | struct net *net, |
77d8bf9c ACM |
33 | struct inet_bind_hashbucket *head, |
34 | const unsigned short snum) | |
35 | { | |
54e6ecb2 | 36 | struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC); |
77d8bf9c ACM |
37 | |
38 | if (tb != NULL) { | |
7a9546ee | 39 | write_pnet(&tb->ib_net, hold_net(net)); |
77d8bf9c ACM |
40 | tb->port = snum; |
41 | tb->fastreuse = 0; | |
a9d8f911 | 42 | tb->num_owners = 0; |
77d8bf9c ACM |
43 | INIT_HLIST_HEAD(&tb->owners); |
44 | hlist_add_head(&tb->node, &head->chain); | |
45 | } | |
46 | return tb; | |
47 | } | |
48 | ||
77d8bf9c ACM |
49 | /* |
50 | * Caller must hold hashbucket lock for this tb with local BH disabled | |
51 | */ | |
e18b890b | 52 | void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb) |
77d8bf9c ACM |
53 | { |
54 | if (hlist_empty(&tb->owners)) { | |
55 | __hlist_del(&tb->node); | |
7a9546ee | 56 | release_net(ib_net(tb)); |
77d8bf9c ACM |
57 | kmem_cache_free(cachep, tb); |
58 | } | |
59 | } | |
2d8c4ce5 ACM |
60 | |
61 | void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb, | |
62 | const unsigned short snum) | |
63 | { | |
a9d8f911 EP |
64 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
65 | ||
24dd1fa1 | 66 | atomic_inc(&hashinfo->bsockets); |
a9d8f911 | 67 | |
c720c7e8 | 68 | inet_sk(sk)->inet_num = snum; |
2d8c4ce5 | 69 | sk_add_bind_node(sk, &tb->owners); |
a9d8f911 | 70 | tb->num_owners++; |
463c84b9 | 71 | inet_csk(sk)->icsk_bind_hash = tb; |
2d8c4ce5 ACM |
72 | } |
73 | ||
2d8c4ce5 ACM |
74 | /* |
75 | * Get rid of any references to a local port held by the given sock. | |
76 | */ | |
ab1e0a13 | 77 | static void __inet_put_port(struct sock *sk) |
2d8c4ce5 | 78 | { |
39d8cda7 | 79 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
c720c7e8 | 80 | const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->inet_num, |
7f635ab7 | 81 | hashinfo->bhash_size); |
2d8c4ce5 ACM |
82 | struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash]; |
83 | struct inet_bind_bucket *tb; | |
84 | ||
24dd1fa1 | 85 | atomic_dec(&hashinfo->bsockets); |
a9d8f911 | 86 | |
2d8c4ce5 | 87 | spin_lock(&head->lock); |
463c84b9 | 88 | tb = inet_csk(sk)->icsk_bind_hash; |
2d8c4ce5 | 89 | __sk_del_bind_node(sk); |
a9d8f911 | 90 | tb->num_owners--; |
463c84b9 | 91 | inet_csk(sk)->icsk_bind_hash = NULL; |
c720c7e8 | 92 | inet_sk(sk)->inet_num = 0; |
2d8c4ce5 ACM |
93 | inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb); |
94 | spin_unlock(&head->lock); | |
95 | } | |
96 | ||
ab1e0a13 | 97 | void inet_put_port(struct sock *sk) |
2d8c4ce5 ACM |
98 | { |
99 | local_bh_disable(); | |
ab1e0a13 | 100 | __inet_put_port(sk); |
2d8c4ce5 ACM |
101 | local_bh_enable(); |
102 | } | |
2d8c4ce5 | 103 | EXPORT_SYMBOL(inet_put_port); |
f3f05f70 | 104 | |
093d2823 | 105 | int __inet_inherit_port(struct sock *sk, struct sock *child) |
53083773 PE |
106 | { |
107 | struct inet_hashinfo *table = sk->sk_prot->h.hashinfo; | |
093d2823 BS |
108 | unsigned short port = inet_sk(child)->inet_num; |
109 | const int bhash = inet_bhashfn(sock_net(sk), port, | |
7f635ab7 | 110 | table->bhash_size); |
53083773 PE |
111 | struct inet_bind_hashbucket *head = &table->bhash[bhash]; |
112 | struct inet_bind_bucket *tb; | |
113 | ||
114 | spin_lock(&head->lock); | |
115 | tb = inet_csk(sk)->icsk_bind_hash; | |
093d2823 BS |
116 | if (tb->port != port) { |
117 | /* NOTE: using tproxy and redirecting skbs to a proxy | |
118 | * on a different listener port breaks the assumption | |
119 | * that the listener socket's icsk_bind_hash is the same | |
120 | * as that of the child socket. We have to look up or | |
121 | * create a new bind bucket for the child here. */ | |
122 | struct hlist_node *node; | |
123 | inet_bind_bucket_for_each(tb, node, &head->chain) { | |
124 | if (net_eq(ib_net(tb), sock_net(sk)) && | |
125 | tb->port == port) | |
126 | break; | |
127 | } | |
128 | if (!node) { | |
129 | tb = inet_bind_bucket_create(table->bind_bucket_cachep, | |
130 | sock_net(sk), head, port); | |
131 | if (!tb) { | |
132 | spin_unlock(&head->lock); | |
133 | return -ENOMEM; | |
134 | } | |
135 | } | |
136 | } | |
b4ff3c90 | 137 | inet_bind_hash(child, tb, port); |
53083773 | 138 | spin_unlock(&head->lock); |
093d2823 BS |
139 | |
140 | return 0; | |
53083773 | 141 | } |
53083773 PE |
142 | EXPORT_SYMBOL_GPL(__inet_inherit_port); |
143 | ||
c25eb3bf ED |
144 | static inline int compute_score(struct sock *sk, struct net *net, |
145 | const unsigned short hnum, const __be32 daddr, | |
146 | const int dif) | |
147 | { | |
148 | int score = -1; | |
149 | struct inet_sock *inet = inet_sk(sk); | |
150 | ||
c720c7e8 | 151 | if (net_eq(sock_net(sk), net) && inet->inet_num == hnum && |
c25eb3bf | 152 | !ipv6_only_sock(sk)) { |
c720c7e8 | 153 | __be32 rcv_saddr = inet->inet_rcv_saddr; |
c25eb3bf ED |
154 | score = sk->sk_family == PF_INET ? 1 : 0; |
155 | if (rcv_saddr) { | |
156 | if (rcv_saddr != daddr) | |
157 | return -1; | |
158 | score += 2; | |
159 | } | |
160 | if (sk->sk_bound_dev_if) { | |
161 | if (sk->sk_bound_dev_if != dif) | |
162 | return -1; | |
163 | score += 2; | |
164 | } | |
165 | } | |
166 | return score; | |
167 | } | |
168 | ||
33b62231 ACM |
169 | /* |
170 | * Don't inline this cruft. Here are some nice properties to exploit here. The | |
171 | * BSD API does not allow a listening sock to specify the remote port nor the | |
172 | * remote address for the connection. So always assume those are both | |
173 | * wildcarded during the search since they can never be otherwise. | |
174 | */ | |
e48c414e | 175 | |
c25eb3bf | 176 | |
c67499c0 PE |
177 | struct sock *__inet_lookup_listener(struct net *net, |
178 | struct inet_hashinfo *hashinfo, | |
fb99c848 | 179 | const __be32 daddr, const unsigned short hnum, |
8f491069 | 180 | const int dif) |
99a92ff5 | 181 | { |
c25eb3bf ED |
182 | struct sock *sk, *result; |
183 | struct hlist_nulls_node *node; | |
184 | unsigned int hash = inet_lhashfn(net, hnum); | |
185 | struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash]; | |
186 | int score, hiscore; | |
99a92ff5 | 187 | |
c25eb3bf ED |
188 | rcu_read_lock(); |
189 | begin: | |
190 | result = NULL; | |
191 | hiscore = -1; | |
192 | sk_nulls_for_each_rcu(sk, node, &ilb->head) { | |
193 | score = compute_score(sk, net, hnum, daddr, dif); | |
194 | if (score > hiscore) { | |
195 | result = sk; | |
196 | hiscore = score; | |
197 | } | |
99a92ff5 | 198 | } |
c25eb3bf ED |
199 | /* |
200 | * if the nulls value we got at the end of this lookup is | |
201 | * not the expected one, we must restart lookup. | |
202 | * We probably met an item that was moved to another chain. | |
203 | */ | |
204 | if (get_nulls_value(node) != hash + LISTENING_NULLS_BASE) | |
205 | goto begin; | |
206 | if (result) { | |
207 | if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) | |
208 | result = NULL; | |
209 | else if (unlikely(compute_score(result, net, hnum, daddr, | |
210 | dif) < hiscore)) { | |
211 | sock_put(result); | |
212 | goto begin; | |
213 | } | |
99a92ff5 | 214 | } |
c25eb3bf ED |
215 | rcu_read_unlock(); |
216 | return result; | |
99a92ff5 | 217 | } |
8f491069 | 218 | EXPORT_SYMBOL_GPL(__inet_lookup_listener); |
a7f5e7f1 | 219 | |
5e73ea1a | 220 | struct sock *__inet_lookup_established(struct net *net, |
c67499c0 | 221 | struct inet_hashinfo *hashinfo, |
77a5ba55 PE |
222 | const __be32 saddr, const __be16 sport, |
223 | const __be32 daddr, const u16 hnum, | |
224 | const int dif) | |
225 | { | |
226 | INET_ADDR_COOKIE(acookie, saddr, daddr) | |
227 | const __portpair ports = INET_COMBINED_PORTS(sport, hnum); | |
228 | struct sock *sk; | |
3ab5aee7 | 229 | const struct hlist_nulls_node *node; |
77a5ba55 PE |
230 | /* Optimize here for direct hit, only listening connections can |
231 | * have wildcards anyways. | |
232 | */ | |
9f26b3ad | 233 | unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport); |
f373b53b | 234 | unsigned int slot = hash & hashinfo->ehash_mask; |
3ab5aee7 | 235 | struct inet_ehash_bucket *head = &hashinfo->ehash[slot]; |
77a5ba55 | 236 | |
3ab5aee7 ED |
237 | rcu_read_lock(); |
238 | begin: | |
239 | sk_nulls_for_each_rcu(sk, node, &head->chain) { | |
c67499c0 | 240 | if (INET_MATCH(sk, net, hash, acookie, |
3ab5aee7 ED |
241 | saddr, daddr, ports, dif)) { |
242 | if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) | |
243 | goto begintw; | |
244 | if (unlikely(!INET_MATCH(sk, net, hash, acookie, | |
245 | saddr, daddr, ports, dif))) { | |
246 | sock_put(sk); | |
247 | goto begin; | |
248 | } | |
249 | goto out; | |
250 | } | |
77a5ba55 | 251 | } |
3ab5aee7 ED |
252 | /* |
253 | * if the nulls value we got at the end of this lookup is | |
254 | * not the expected one, we must restart lookup. | |
255 | * We probably met an item that was moved to another chain. | |
256 | */ | |
257 | if (get_nulls_value(node) != slot) | |
258 | goto begin; | |
77a5ba55 | 259 | |
3ab5aee7 | 260 | begintw: |
77a5ba55 | 261 | /* Must check for a TIME_WAIT'er before going to listener hash. */ |
3ab5aee7 | 262 | sk_nulls_for_each_rcu(sk, node, &head->twchain) { |
c67499c0 | 263 | if (INET_TW_MATCH(sk, net, hash, acookie, |
3ab5aee7 ED |
264 | saddr, daddr, ports, dif)) { |
265 | if (unlikely(!atomic_inc_not_zero(&sk->sk_refcnt))) { | |
266 | sk = NULL; | |
267 | goto out; | |
268 | } | |
269 | if (unlikely(!INET_TW_MATCH(sk, net, hash, acookie, | |
270 | saddr, daddr, ports, dif))) { | |
271 | sock_put(sk); | |
272 | goto begintw; | |
273 | } | |
274 | goto out; | |
275 | } | |
77a5ba55 | 276 | } |
3ab5aee7 ED |
277 | /* |
278 | * if the nulls value we got at the end of this lookup is | |
279 | * not the expected one, we must restart lookup. | |
280 | * We probably met an item that was moved to another chain. | |
281 | */ | |
282 | if (get_nulls_value(node) != slot) | |
283 | goto begintw; | |
77a5ba55 PE |
284 | sk = NULL; |
285 | out: | |
3ab5aee7 | 286 | rcu_read_unlock(); |
77a5ba55 | 287 | return sk; |
77a5ba55 PE |
288 | } |
289 | EXPORT_SYMBOL_GPL(__inet_lookup_established); | |
290 | ||
a7f5e7f1 ACM |
291 | /* called with local bh disabled */ |
292 | static int __inet_check_established(struct inet_timewait_death_row *death_row, | |
293 | struct sock *sk, __u16 lport, | |
294 | struct inet_timewait_sock **twp) | |
295 | { | |
296 | struct inet_hashinfo *hinfo = death_row->hashinfo; | |
297 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 ED |
298 | __be32 daddr = inet->inet_rcv_saddr; |
299 | __be32 saddr = inet->inet_daddr; | |
a7f5e7f1 ACM |
300 | int dif = sk->sk_bound_dev_if; |
301 | INET_ADDR_COOKIE(acookie, saddr, daddr) | |
c720c7e8 | 302 | const __portpair ports = INET_COMBINED_PORTS(inet->inet_dport, lport); |
9f26b3ad | 303 | struct net *net = sock_net(sk); |
c720c7e8 ED |
304 | unsigned int hash = inet_ehashfn(net, daddr, lport, |
305 | saddr, inet->inet_dport); | |
a7f5e7f1 | 306 | struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash); |
9db66bdc | 307 | spinlock_t *lock = inet_ehash_lockp(hinfo, hash); |
a7f5e7f1 | 308 | struct sock *sk2; |
3ab5aee7 | 309 | const struct hlist_nulls_node *node; |
a7f5e7f1 | 310 | struct inet_timewait_sock *tw; |
13475a30 | 311 | int twrefcnt = 0; |
a7f5e7f1 | 312 | |
9db66bdc | 313 | spin_lock(lock); |
a7f5e7f1 ACM |
314 | |
315 | /* Check TIME-WAIT sockets first. */ | |
3ab5aee7 | 316 | sk_nulls_for_each(sk2, node, &head->twchain) { |
a7f5e7f1 ACM |
317 | tw = inet_twsk(sk2); |
318 | ||
c67499c0 PE |
319 | if (INET_TW_MATCH(sk2, net, hash, acookie, |
320 | saddr, daddr, ports, dif)) { | |
a7f5e7f1 ACM |
321 | if (twsk_unique(sk, sk2, twp)) |
322 | goto unique; | |
323 | else | |
324 | goto not_unique; | |
325 | } | |
326 | } | |
327 | tw = NULL; | |
328 | ||
329 | /* And established part... */ | |
3ab5aee7 | 330 | sk_nulls_for_each(sk2, node, &head->chain) { |
c67499c0 PE |
331 | if (INET_MATCH(sk2, net, hash, acookie, |
332 | saddr, daddr, ports, dif)) | |
a7f5e7f1 ACM |
333 | goto not_unique; |
334 | } | |
335 | ||
336 | unique: | |
337 | /* Must record num and sport now. Otherwise we will see | |
338 | * in hash table socket with a funny identity. */ | |
c720c7e8 ED |
339 | inet->inet_num = lport; |
340 | inet->inet_sport = htons(lport); | |
a7f5e7f1 | 341 | sk->sk_hash = hash; |
547b792c | 342 | WARN_ON(!sk_unhashed(sk)); |
3ab5aee7 | 343 | __sk_nulls_add_node_rcu(sk, &head->chain); |
13475a30 ED |
344 | if (tw) { |
345 | twrefcnt = inet_twsk_unhash(tw); | |
346 | NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED); | |
347 | } | |
9db66bdc | 348 | spin_unlock(lock); |
13475a30 ED |
349 | if (twrefcnt) |
350 | inet_twsk_put(tw); | |
c29a0bc4 | 351 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
a7f5e7f1 ACM |
352 | |
353 | if (twp) { | |
354 | *twp = tw; | |
a7f5e7f1 ACM |
355 | } else if (tw) { |
356 | /* Silly. Should hash-dance instead... */ | |
357 | inet_twsk_deschedule(tw, death_row); | |
a7f5e7f1 ACM |
358 | |
359 | inet_twsk_put(tw); | |
360 | } | |
a7f5e7f1 ACM |
361 | return 0; |
362 | ||
363 | not_unique: | |
9db66bdc | 364 | spin_unlock(lock); |
a7f5e7f1 ACM |
365 | return -EADDRNOTAVAIL; |
366 | } | |
367 | ||
368 | static inline u32 inet_sk_port_offset(const struct sock *sk) | |
369 | { | |
370 | const struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 ED |
371 | return secure_ipv4_port_ephemeral(inet->inet_rcv_saddr, |
372 | inet->inet_daddr, | |
373 | inet->inet_dport); | |
a7f5e7f1 ACM |
374 | } |
375 | ||
9327f705 | 376 | int __inet_hash_nolisten(struct sock *sk, struct inet_timewait_sock *tw) |
152da81d | 377 | { |
39d8cda7 | 378 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
3ab5aee7 | 379 | struct hlist_nulls_head *list; |
9db66bdc | 380 | spinlock_t *lock; |
152da81d | 381 | struct inet_ehash_bucket *head; |
9327f705 | 382 | int twrefcnt = 0; |
152da81d | 383 | |
547b792c | 384 | WARN_ON(!sk_unhashed(sk)); |
152da81d PE |
385 | |
386 | sk->sk_hash = inet_sk_ehashfn(sk); | |
387 | head = inet_ehash_bucket(hashinfo, sk->sk_hash); | |
388 | list = &head->chain; | |
389 | lock = inet_ehash_lockp(hashinfo, sk->sk_hash); | |
390 | ||
9db66bdc | 391 | spin_lock(lock); |
3ab5aee7 | 392 | __sk_nulls_add_node_rcu(sk, list); |
9327f705 ED |
393 | if (tw) { |
394 | WARN_ON(sk->sk_hash != tw->tw_hash); | |
395 | twrefcnt = inet_twsk_unhash(tw); | |
396 | } | |
9db66bdc | 397 | spin_unlock(lock); |
c29a0bc4 | 398 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
9327f705 | 399 | return twrefcnt; |
152da81d PE |
400 | } |
401 | EXPORT_SYMBOL_GPL(__inet_hash_nolisten); | |
402 | ||
ab1e0a13 | 403 | static void __inet_hash(struct sock *sk) |
152da81d | 404 | { |
39d8cda7 | 405 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
5caea4ea | 406 | struct inet_listen_hashbucket *ilb; |
152da81d PE |
407 | |
408 | if (sk->sk_state != TCP_LISTEN) { | |
9327f705 | 409 | __inet_hash_nolisten(sk, NULL); |
152da81d PE |
410 | return; |
411 | } | |
412 | ||
547b792c | 413 | WARN_ON(!sk_unhashed(sk)); |
5caea4ea | 414 | ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)]; |
152da81d | 415 | |
5caea4ea | 416 | spin_lock(&ilb->lock); |
c25eb3bf | 417 | __sk_nulls_add_node_rcu(sk, &ilb->head); |
c29a0bc4 | 418 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
5caea4ea | 419 | spin_unlock(&ilb->lock); |
152da81d | 420 | } |
ab1e0a13 ACM |
421 | |
422 | void inet_hash(struct sock *sk) | |
423 | { | |
424 | if (sk->sk_state != TCP_CLOSE) { | |
425 | local_bh_disable(); | |
426 | __inet_hash(sk); | |
427 | local_bh_enable(); | |
428 | } | |
429 | } | |
430 | EXPORT_SYMBOL_GPL(inet_hash); | |
431 | ||
432 | void inet_unhash(struct sock *sk) | |
433 | { | |
39d8cda7 | 434 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
c25eb3bf ED |
435 | spinlock_t *lock; |
436 | int done; | |
ab1e0a13 ACM |
437 | |
438 | if (sk_unhashed(sk)) | |
5caea4ea | 439 | return; |
ab1e0a13 | 440 | |
c25eb3bf ED |
441 | if (sk->sk_state == TCP_LISTEN) |
442 | lock = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)].lock; | |
443 | else | |
444 | lock = inet_ehash_lockp(hashinfo, sk->sk_hash); | |
5caea4ea | 445 | |
c25eb3bf ED |
446 | spin_lock_bh(lock); |
447 | done =__sk_nulls_del_node_init_rcu(sk); | |
c25eb3bf ED |
448 | if (done) |
449 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); | |
920de804 | 450 | spin_unlock_bh(lock); |
ab1e0a13 ACM |
451 | } |
452 | EXPORT_SYMBOL_GPL(inet_unhash); | |
152da81d | 453 | |
5ee31fc1 | 454 | int __inet_hash_connect(struct inet_timewait_death_row *death_row, |
5d8c0aa9 | 455 | struct sock *sk, u32 port_offset, |
5ee31fc1 PE |
456 | int (*check_established)(struct inet_timewait_death_row *, |
457 | struct sock *, __u16, struct inet_timewait_sock **), | |
9327f705 | 458 | int (*hash)(struct sock *sk, struct inet_timewait_sock *twp)) |
a7f5e7f1 ACM |
459 | { |
460 | struct inet_hashinfo *hinfo = death_row->hashinfo; | |
c720c7e8 | 461 | const unsigned short snum = inet_sk(sk)->inet_num; |
e905a9ed YH |
462 | struct inet_bind_hashbucket *head; |
463 | struct inet_bind_bucket *tb; | |
a7f5e7f1 | 464 | int ret; |
3b1e0a65 | 465 | struct net *net = sock_net(sk); |
9327f705 | 466 | int twrefcnt = 1; |
a7f5e7f1 | 467 | |
e905a9ed | 468 | if (!snum) { |
227b60f5 | 469 | int i, remaining, low, high, port; |
a7f5e7f1 | 470 | static u32 hint; |
5d8c0aa9 | 471 | u32 offset = hint + port_offset; |
a7f5e7f1 | 472 | struct hlist_node *node; |
e905a9ed | 473 | struct inet_timewait_sock *tw = NULL; |
a7f5e7f1 | 474 | |
227b60f5 | 475 | inet_get_local_port_range(&low, &high); |
a25de534 | 476 | remaining = (high - low) + 1; |
227b60f5 | 477 | |
e905a9ed | 478 | local_bh_disable(); |
227b60f5 SH |
479 | for (i = 1; i <= remaining; i++) { |
480 | port = low + (i + offset) % remaining; | |
e3826f1e AW |
481 | if (inet_is_reserved_local_port(port)) |
482 | continue; | |
7f635ab7 PE |
483 | head = &hinfo->bhash[inet_bhashfn(net, port, |
484 | hinfo->bhash_size)]; | |
e905a9ed | 485 | spin_lock(&head->lock); |
a7f5e7f1 | 486 | |
e905a9ed YH |
487 | /* Does not bother with rcv_saddr checks, |
488 | * because the established check is already | |
489 | * unique enough. | |
490 | */ | |
a7f5e7f1 | 491 | inet_bind_bucket_for_each(tb, node, &head->chain) { |
09ad9bc7 OP |
492 | if (net_eq(ib_net(tb), net) && |
493 | tb->port == port) { | |
e905a9ed YH |
494 | if (tb->fastreuse >= 0) |
495 | goto next_port; | |
a9d8f911 | 496 | WARN_ON(hlist_empty(&tb->owners)); |
5ee31fc1 PE |
497 | if (!check_established(death_row, sk, |
498 | port, &tw)) | |
e905a9ed YH |
499 | goto ok; |
500 | goto next_port; | |
501 | } | |
502 | } | |
503 | ||
941b1d22 PE |
504 | tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep, |
505 | net, head, port); | |
e905a9ed YH |
506 | if (!tb) { |
507 | spin_unlock(&head->lock); | |
508 | break; | |
509 | } | |
510 | tb->fastreuse = -1; | |
511 | goto ok; | |
512 | ||
513 | next_port: | |
514 | spin_unlock(&head->lock); | |
515 | } | |
516 | local_bh_enable(); | |
517 | ||
518 | return -EADDRNOTAVAIL; | |
a7f5e7f1 ACM |
519 | |
520 | ok: | |
521 | hint += i; | |
522 | ||
e905a9ed YH |
523 | /* Head lock still held and bh's disabled */ |
524 | inet_bind_hash(sk, tb, port); | |
a7f5e7f1 | 525 | if (sk_unhashed(sk)) { |
c720c7e8 | 526 | inet_sk(sk)->inet_sport = htons(port); |
9327f705 | 527 | twrefcnt += hash(sk, tw); |
e905a9ed | 528 | } |
3cdaedae ED |
529 | if (tw) |
530 | twrefcnt += inet_twsk_bind_unhash(tw, hinfo); | |
e905a9ed | 531 | spin_unlock(&head->lock); |
a7f5e7f1 | 532 | |
e905a9ed YH |
533 | if (tw) { |
534 | inet_twsk_deschedule(tw, death_row); | |
9327f705 ED |
535 | while (twrefcnt) { |
536 | twrefcnt--; | |
537 | inet_twsk_put(tw); | |
538 | } | |
e905a9ed | 539 | } |
a7f5e7f1 ACM |
540 | |
541 | ret = 0; | |
542 | goto out; | |
e905a9ed | 543 | } |
a7f5e7f1 | 544 | |
7f635ab7 | 545 | head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)]; |
e905a9ed | 546 | tb = inet_csk(sk)->icsk_bind_hash; |
a7f5e7f1 ACM |
547 | spin_lock_bh(&head->lock); |
548 | if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) { | |
9327f705 | 549 | hash(sk, NULL); |
a7f5e7f1 ACM |
550 | spin_unlock_bh(&head->lock); |
551 | return 0; | |
552 | } else { | |
553 | spin_unlock(&head->lock); | |
554 | /* No definite answer... Walk to established hash table */ | |
5ee31fc1 | 555 | ret = check_established(death_row, sk, snum, NULL); |
a7f5e7f1 ACM |
556 | out: |
557 | local_bh_enable(); | |
558 | return ret; | |
559 | } | |
560 | } | |
5ee31fc1 PE |
561 | |
562 | /* | |
563 | * Bind a port for a connect operation and hash it. | |
564 | */ | |
565 | int inet_hash_connect(struct inet_timewait_death_row *death_row, | |
566 | struct sock *sk) | |
567 | { | |
5d8c0aa9 | 568 | return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk), |
5ee31fc1 PE |
569 | __inet_check_established, __inet_hash_nolisten); |
570 | } | |
a7f5e7f1 | 571 | EXPORT_SYMBOL_GPL(inet_hash_connect); |
5caea4ea ED |
572 | |
573 | void inet_hashinfo_init(struct inet_hashinfo *h) | |
574 | { | |
575 | int i; | |
576 | ||
24dd1fa1 | 577 | atomic_set(&h->bsockets, 0); |
c25eb3bf | 578 | for (i = 0; i < INET_LHTABLE_SIZE; i++) { |
5caea4ea | 579 | spin_lock_init(&h->listening_hash[i].lock); |
c25eb3bf ED |
580 | INIT_HLIST_NULLS_HEAD(&h->listening_hash[i].head, |
581 | i + LISTENING_NULLS_BASE); | |
582 | } | |
5caea4ea | 583 | } |
5caea4ea | 584 | EXPORT_SYMBOL_GPL(inet_hashinfo_init); |