Commit | Line | Data |
---|---|---|
3f421baa 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 | * Support for INET connection oriented protocols. | |
7 | * | |
8 | * Authors: See the TCP sources | |
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 | ||
3f421baa ACM |
16 | #include <linux/module.h> |
17 | #include <linux/jhash.h> | |
18 | ||
19 | #include <net/inet_connection_sock.h> | |
20 | #include <net/inet_hashtables.h> | |
21 | #include <net/inet_timewait_sock.h> | |
22 | #include <net/ip.h> | |
23 | #include <net/route.h> | |
24 | #include <net/tcp_states.h> | |
a019d6fe | 25 | #include <net/xfrm.h> |
3f421baa ACM |
26 | |
27 | #ifdef INET_CSK_DEBUG | |
28 | const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n"; | |
29 | EXPORT_SYMBOL(inet_csk_timer_bug_msg); | |
30 | #endif | |
31 | ||
0bbf87d8 | 32 | void inet_get_local_port_range(struct net *net, int *low, int *high) |
227b60f5 | 33 | { |
95c96174 ED |
34 | unsigned int seq; |
35 | ||
227b60f5 | 36 | do { |
c9d8f1a6 | 37 | seq = read_seqbegin(&net->ipv4.ip_local_ports.lock); |
227b60f5 | 38 | |
c9d8f1a6 CW |
39 | *low = net->ipv4.ip_local_ports.range[0]; |
40 | *high = net->ipv4.ip_local_ports.range[1]; | |
41 | } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq)); | |
227b60f5 SH |
42 | } |
43 | EXPORT_SYMBOL(inet_get_local_port_range); | |
3f421baa | 44 | |
971af18b | 45 | int inet_csk_bind_conflict(const struct sock *sk, |
aacd9289 | 46 | const struct inet_bind_bucket *tb, bool relax) |
3f421baa | 47 | { |
3f421baa | 48 | struct sock *sk2; |
3f421baa | 49 | int reuse = sk->sk_reuse; |
da5e3630 TH |
50 | int reuseport = sk->sk_reuseport; |
51 | kuid_t uid = sock_i_uid((struct sock *)sk); | |
3f421baa | 52 | |
7477fd2e PE |
53 | /* |
54 | * Unlike other sk lookup places we do not check | |
55 | * for sk_net here, since _all_ the socks listed | |
56 | * in tb->owners list belong to the same net - the | |
57 | * one this bucket belongs to. | |
58 | */ | |
59 | ||
b67bfe0d | 60 | sk_for_each_bound(sk2, &tb->owners) { |
3f421baa ACM |
61 | if (sk != sk2 && |
62 | !inet_v6_ipv6only(sk2) && | |
63 | (!sk->sk_bound_dev_if || | |
64 | !sk2->sk_bound_dev_if || | |
65 | sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) { | |
da5e3630 TH |
66 | if ((!reuse || !sk2->sk_reuse || |
67 | sk2->sk_state == TCP_LISTEN) && | |
68 | (!reuseport || !sk2->sk_reuseport || | |
69 | (sk2->sk_state != TCP_TIME_WAIT && | |
70 | !uid_eq(uid, sock_i_uid(sk2))))) { | |
50805466 ED |
71 | |
72 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || | |
73 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
3f421baa | 74 | break; |
8d238b25 | 75 | } |
aacd9289 AC |
76 | if (!relax && reuse && sk2->sk_reuse && |
77 | sk2->sk_state != TCP_LISTEN) { | |
aacd9289 | 78 | |
50805466 ED |
79 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || |
80 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
aacd9289 AC |
81 | break; |
82 | } | |
3f421baa ACM |
83 | } |
84 | } | |
b67bfe0d | 85 | return sk2 != NULL; |
3f421baa | 86 | } |
971af18b ACM |
87 | EXPORT_SYMBOL_GPL(inet_csk_bind_conflict); |
88 | ||
3f421baa ACM |
89 | /* Obtain a reference to a local port for the given sock, |
90 | * if snum is zero it means select any available local port. | |
91 | */ | |
ab1e0a13 | 92 | int inet_csk_get_port(struct sock *sk, unsigned short snum) |
3f421baa | 93 | { |
39d8cda7 | 94 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
3f421baa | 95 | struct inet_bind_hashbucket *head; |
3f421baa | 96 | struct inet_bind_bucket *tb; |
a9d8f911 | 97 | int ret, attempts = 5; |
3b1e0a65 | 98 | struct net *net = sock_net(sk); |
a9d8f911 | 99 | int smallest_size = -1, smallest_rover; |
da5e3630 | 100 | kuid_t uid = sock_i_uid(sk); |
3f421baa ACM |
101 | |
102 | local_bh_disable(); | |
103 | if (!snum) { | |
227b60f5 SH |
104 | int remaining, rover, low, high; |
105 | ||
a9d8f911 | 106 | again: |
0bbf87d8 | 107 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 108 | remaining = (high - low) + 1; |
63862b5b | 109 | smallest_rover = rover = prandom_u32() % remaining + low; |
3f421baa | 110 | |
a9d8f911 | 111 | smallest_size = -1; |
3f421baa | 112 | do { |
122ff243 | 113 | if (inet_is_local_reserved_port(net, rover)) |
e3826f1e | 114 | goto next_nolock; |
7f635ab7 PE |
115 | head = &hashinfo->bhash[inet_bhashfn(net, rover, |
116 | hashinfo->bhash_size)]; | |
3f421baa | 117 | spin_lock(&head->lock); |
b67bfe0d | 118 | inet_bind_bucket_for_each(tb, &head->chain) |
09ad9bc7 | 119 | if (net_eq(ib_net(tb), net) && tb->port == rover) { |
da5e3630 TH |
120 | if (((tb->fastreuse > 0 && |
121 | sk->sk_reuse && | |
122 | sk->sk_state != TCP_LISTEN) || | |
123 | (tb->fastreuseport > 0 && | |
124 | sk->sk_reuseport && | |
125 | uid_eq(tb->fastuid, uid))) && | |
a9d8f911 EP |
126 | (tb->num_owners < smallest_size || smallest_size == -1)) { |
127 | smallest_size = tb->num_owners; | |
128 | smallest_rover = rover; | |
aacd9289 AC |
129 | if (atomic_read(&hashinfo->bsockets) > (high - low) + 1 && |
130 | !inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) { | |
8d238b25 | 131 | snum = smallest_rover; |
fddb7b57 | 132 | goto tb_found; |
a9d8f911 EP |
133 | } |
134 | } | |
aacd9289 | 135 | if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) { |
2b05ad33 | 136 | snum = rover; |
fddb7b57 | 137 | goto tb_found; |
2b05ad33 | 138 | } |
3f421baa | 139 | goto next; |
a9d8f911 | 140 | } |
3f421baa ACM |
141 | break; |
142 | next: | |
143 | spin_unlock(&head->lock); | |
e3826f1e | 144 | next_nolock: |
6df71634 SH |
145 | if (++rover > high) |
146 | rover = low; | |
3f421baa | 147 | } while (--remaining > 0); |
3f421baa ACM |
148 | |
149 | /* Exhausted local port range during search? It is not | |
150 | * possible for us to be holding one of the bind hash | |
151 | * locks if this test triggers, because if 'remaining' | |
152 | * drops to zero, we broke out of the do/while loop at | |
153 | * the top level, not from the 'break;' statement. | |
154 | */ | |
155 | ret = 1; | |
a9d8f911 EP |
156 | if (remaining <= 0) { |
157 | if (smallest_size != -1) { | |
158 | snum = smallest_rover; | |
159 | goto have_snum; | |
160 | } | |
3f421baa | 161 | goto fail; |
a9d8f911 | 162 | } |
3f421baa ACM |
163 | /* OK, here is the one we will use. HEAD is |
164 | * non-NULL and we hold it's mutex. | |
165 | */ | |
166 | snum = rover; | |
167 | } else { | |
a9d8f911 | 168 | have_snum: |
7f635ab7 PE |
169 | head = &hashinfo->bhash[inet_bhashfn(net, snum, |
170 | hashinfo->bhash_size)]; | |
3f421baa | 171 | spin_lock(&head->lock); |
b67bfe0d | 172 | inet_bind_bucket_for_each(tb, &head->chain) |
09ad9bc7 | 173 | if (net_eq(ib_net(tb), net) && tb->port == snum) |
3f421baa ACM |
174 | goto tb_found; |
175 | } | |
176 | tb = NULL; | |
177 | goto tb_not_found; | |
178 | tb_found: | |
179 | if (!hlist_empty(&tb->owners)) { | |
4a17fd52 PE |
180 | if (sk->sk_reuse == SK_FORCE_REUSE) |
181 | goto success; | |
182 | ||
da5e3630 TH |
183 | if (((tb->fastreuse > 0 && |
184 | sk->sk_reuse && sk->sk_state != TCP_LISTEN) || | |
185 | (tb->fastreuseport > 0 && | |
186 | sk->sk_reuseport && uid_eq(tb->fastuid, uid))) && | |
a9d8f911 | 187 | smallest_size == -1) { |
3f421baa ACM |
188 | goto success; |
189 | } else { | |
190 | ret = 1; | |
aacd9289 | 191 | if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true)) { |
da5e3630 | 192 | if (((sk->sk_reuse && sk->sk_state != TCP_LISTEN) || |
9c5e0c0b TH |
193 | (tb->fastreuseport > 0 && |
194 | sk->sk_reuseport && uid_eq(tb->fastuid, uid))) && | |
5add3009 | 195 | smallest_size != -1 && --attempts >= 0) { |
a9d8f911 EP |
196 | spin_unlock(&head->lock); |
197 | goto again; | |
198 | } | |
aacd9289 | 199 | |
3f421baa | 200 | goto fail_unlock; |
a9d8f911 | 201 | } |
3f421baa ACM |
202 | } |
203 | } | |
204 | tb_not_found: | |
205 | ret = 1; | |
941b1d22 PE |
206 | if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep, |
207 | net, head, snum)) == NULL) | |
3f421baa ACM |
208 | goto fail_unlock; |
209 | if (hlist_empty(&tb->owners)) { | |
210 | if (sk->sk_reuse && sk->sk_state != TCP_LISTEN) | |
211 | tb->fastreuse = 1; | |
212 | else | |
213 | tb->fastreuse = 0; | |
da5e3630 TH |
214 | if (sk->sk_reuseport) { |
215 | tb->fastreuseport = 1; | |
216 | tb->fastuid = uid; | |
9c5e0c0b | 217 | } else |
da5e3630 | 218 | tb->fastreuseport = 0; |
da5e3630 TH |
219 | } else { |
220 | if (tb->fastreuse && | |
221 | (!sk->sk_reuse || sk->sk_state == TCP_LISTEN)) | |
222 | tb->fastreuse = 0; | |
223 | if (tb->fastreuseport && | |
9c5e0c0b | 224 | (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid))) |
da5e3630 | 225 | tb->fastreuseport = 0; |
da5e3630 | 226 | } |
3f421baa ACM |
227 | success: |
228 | if (!inet_csk(sk)->icsk_bind_hash) | |
229 | inet_bind_hash(sk, tb, snum); | |
547b792c | 230 | WARN_ON(inet_csk(sk)->icsk_bind_hash != tb); |
e905a9ed | 231 | ret = 0; |
3f421baa ACM |
232 | |
233 | fail_unlock: | |
234 | spin_unlock(&head->lock); | |
235 | fail: | |
236 | local_bh_enable(); | |
237 | return ret; | |
238 | } | |
3f421baa ACM |
239 | EXPORT_SYMBOL_GPL(inet_csk_get_port); |
240 | ||
241 | /* | |
242 | * Wait for an incoming connection, avoid race conditions. This must be called | |
243 | * with the socket locked. | |
244 | */ | |
245 | static int inet_csk_wait_for_connect(struct sock *sk, long timeo) | |
246 | { | |
247 | struct inet_connection_sock *icsk = inet_csk(sk); | |
248 | DEFINE_WAIT(wait); | |
249 | int err; | |
250 | ||
251 | /* | |
252 | * True wake-one mechanism for incoming connections: only | |
253 | * one process gets woken up, not the 'whole herd'. | |
254 | * Since we do not 'race & poll' for established sockets | |
255 | * anymore, the common case will execute the loop only once. | |
256 | * | |
257 | * Subtle issue: "add_wait_queue_exclusive()" will be added | |
258 | * after any current non-exclusive waiters, and we know that | |
259 | * it will always _stay_ after any new non-exclusive waiters | |
260 | * because all non-exclusive waiters are added at the | |
261 | * beginning of the wait-queue. As such, it's ok to "drop" | |
262 | * our exclusiveness temporarily when we get woken up without | |
263 | * having to remove and re-insert us on the wait queue. | |
264 | */ | |
265 | for (;;) { | |
aa395145 | 266 | prepare_to_wait_exclusive(sk_sleep(sk), &wait, |
3f421baa ACM |
267 | TASK_INTERRUPTIBLE); |
268 | release_sock(sk); | |
269 | if (reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
270 | timeo = schedule_timeout(timeo); | |
271 | lock_sock(sk); | |
272 | err = 0; | |
273 | if (!reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
274 | break; | |
275 | err = -EINVAL; | |
276 | if (sk->sk_state != TCP_LISTEN) | |
277 | break; | |
278 | err = sock_intr_errno(timeo); | |
279 | if (signal_pending(current)) | |
280 | break; | |
281 | err = -EAGAIN; | |
282 | if (!timeo) | |
283 | break; | |
284 | } | |
aa395145 | 285 | finish_wait(sk_sleep(sk), &wait); |
3f421baa ACM |
286 | return err; |
287 | } | |
288 | ||
289 | /* | |
290 | * This will accept the next outstanding connection. | |
291 | */ | |
292 | struct sock *inet_csk_accept(struct sock *sk, int flags, int *err) | |
293 | { | |
294 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 295 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
3f421baa | 296 | struct sock *newsk; |
8336886f | 297 | struct request_sock *req; |
3f421baa ACM |
298 | int error; |
299 | ||
300 | lock_sock(sk); | |
301 | ||
302 | /* We need to make sure that this socket is listening, | |
303 | * and that it has something pending. | |
304 | */ | |
305 | error = -EINVAL; | |
306 | if (sk->sk_state != TCP_LISTEN) | |
307 | goto out_err; | |
308 | ||
309 | /* Find already established connection */ | |
8336886f | 310 | if (reqsk_queue_empty(queue)) { |
3f421baa ACM |
311 | long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); |
312 | ||
313 | /* If this is a non blocking socket don't sleep */ | |
314 | error = -EAGAIN; | |
315 | if (!timeo) | |
316 | goto out_err; | |
317 | ||
318 | error = inet_csk_wait_for_connect(sk, timeo); | |
319 | if (error) | |
320 | goto out_err; | |
321 | } | |
8336886f JC |
322 | req = reqsk_queue_remove(queue); |
323 | newsk = req->sk; | |
324 | ||
325 | sk_acceptq_removed(sk); | |
7ab4551f | 326 | if (sk->sk_protocol == IPPROTO_TCP && queue->fastopenq != NULL) { |
8336886f JC |
327 | spin_lock_bh(&queue->fastopenq->lock); |
328 | if (tcp_rsk(req)->listener) { | |
329 | /* We are still waiting for the final ACK from 3WHS | |
330 | * so can't free req now. Instead, we set req->sk to | |
331 | * NULL to signify that the child socket is taken | |
332 | * so reqsk_fastopen_remove() will free the req | |
333 | * when 3WHS finishes (or is aborted). | |
334 | */ | |
335 | req->sk = NULL; | |
336 | req = NULL; | |
337 | } | |
338 | spin_unlock_bh(&queue->fastopenq->lock); | |
339 | } | |
3f421baa ACM |
340 | out: |
341 | release_sock(sk); | |
8336886f JC |
342 | if (req) |
343 | __reqsk_free(req); | |
3f421baa ACM |
344 | return newsk; |
345 | out_err: | |
346 | newsk = NULL; | |
8336886f | 347 | req = NULL; |
3f421baa ACM |
348 | *err = error; |
349 | goto out; | |
350 | } | |
3f421baa ACM |
351 | EXPORT_SYMBOL(inet_csk_accept); |
352 | ||
353 | /* | |
354 | * Using different timers for retransmit, delayed acks and probes | |
e905a9ed | 355 | * We may wish use just one timer maintaining a list of expire jiffies |
3f421baa ACM |
356 | * to optimize. |
357 | */ | |
358 | void inet_csk_init_xmit_timers(struct sock *sk, | |
359 | void (*retransmit_handler)(unsigned long), | |
360 | void (*delack_handler)(unsigned long), | |
361 | void (*keepalive_handler)(unsigned long)) | |
362 | { | |
363 | struct inet_connection_sock *icsk = inet_csk(sk); | |
364 | ||
b24b8a24 PE |
365 | setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler, |
366 | (unsigned long)sk); | |
367 | setup_timer(&icsk->icsk_delack_timer, delack_handler, | |
368 | (unsigned long)sk); | |
369 | setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk); | |
3f421baa ACM |
370 | icsk->icsk_pending = icsk->icsk_ack.pending = 0; |
371 | } | |
3f421baa ACM |
372 | EXPORT_SYMBOL(inet_csk_init_xmit_timers); |
373 | ||
374 | void inet_csk_clear_xmit_timers(struct sock *sk) | |
375 | { | |
376 | struct inet_connection_sock *icsk = inet_csk(sk); | |
377 | ||
378 | icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0; | |
379 | ||
380 | sk_stop_timer(sk, &icsk->icsk_retransmit_timer); | |
381 | sk_stop_timer(sk, &icsk->icsk_delack_timer); | |
382 | sk_stop_timer(sk, &sk->sk_timer); | |
383 | } | |
3f421baa ACM |
384 | EXPORT_SYMBOL(inet_csk_clear_xmit_timers); |
385 | ||
386 | void inet_csk_delete_keepalive_timer(struct sock *sk) | |
387 | { | |
388 | sk_stop_timer(sk, &sk->sk_timer); | |
389 | } | |
3f421baa ACM |
390 | EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); |
391 | ||
392 | void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) | |
393 | { | |
394 | sk_reset_timer(sk, &sk->sk_timer, jiffies + len); | |
395 | } | |
3f421baa ACM |
396 | EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); |
397 | ||
d9319100 | 398 | struct dst_entry *inet_csk_route_req(struct sock *sk, |
6bd023f3 | 399 | struct flowi4 *fl4, |
ba3f7f04 | 400 | const struct request_sock *req) |
3f421baa ACM |
401 | { |
402 | struct rtable *rt; | |
403 | const struct inet_request_sock *ireq = inet_rsk(req); | |
f6d8bd05 | 404 | struct ip_options_rcu *opt = inet_rsk(req)->opt; |
84a3aa00 | 405 | struct net *net = sock_net(sk); |
3e12939a | 406 | int flags = inet_sk_flowi_flags(sk); |
3f421baa | 407 | |
84f39b08 | 408 | flowi4_init_output(fl4, sk->sk_bound_dev_if, ireq->ir_mark, |
e79d9bc7 | 409 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
7433819a | 410 | sk->sk_protocol, |
7586eceb | 411 | flags, |
634fb979 ED |
412 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
413 | ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport); | |
6bd023f3 DM |
414 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
415 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe | 416 | if (IS_ERR(rt)) |
857a6e0a | 417 | goto no_route; |
155e8336 | 418 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
857a6e0a | 419 | goto route_err; |
d8d1f30b | 420 | return &rt->dst; |
857a6e0a IJ |
421 | |
422 | route_err: | |
423 | ip_rt_put(rt); | |
424 | no_route: | |
425 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); | |
426 | return NULL; | |
3f421baa | 427 | } |
3f421baa ACM |
428 | EXPORT_SYMBOL_GPL(inet_csk_route_req); |
429 | ||
77357a95 DM |
430 | struct dst_entry *inet_csk_route_child_sock(struct sock *sk, |
431 | struct sock *newsk, | |
432 | const struct request_sock *req) | |
433 | { | |
434 | const struct inet_request_sock *ireq = inet_rsk(req); | |
435 | struct inet_sock *newinet = inet_sk(newsk); | |
1a7b27c9 | 436 | struct ip_options_rcu *opt; |
77357a95 DM |
437 | struct net *net = sock_net(sk); |
438 | struct flowi4 *fl4; | |
439 | struct rtable *rt; | |
440 | ||
441 | fl4 = &newinet->cork.fl.u.ip4; | |
1a7b27c9 CP |
442 | |
443 | rcu_read_lock(); | |
444 | opt = rcu_dereference(newinet->inet_opt); | |
84f39b08 | 445 | flowi4_init_output(fl4, sk->sk_bound_dev_if, inet_rsk(req)->ir_mark, |
77357a95 DM |
446 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
447 | sk->sk_protocol, inet_sk_flowi_flags(sk), | |
634fb979 ED |
448 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
449 | ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport); | |
77357a95 DM |
450 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
451 | rt = ip_route_output_flow(net, fl4, sk); | |
452 | if (IS_ERR(rt)) | |
453 | goto no_route; | |
155e8336 | 454 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
77357a95 | 455 | goto route_err; |
1a7b27c9 | 456 | rcu_read_unlock(); |
77357a95 DM |
457 | return &rt->dst; |
458 | ||
459 | route_err: | |
460 | ip_rt_put(rt); | |
461 | no_route: | |
1a7b27c9 | 462 | rcu_read_unlock(); |
77357a95 DM |
463 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
464 | return NULL; | |
465 | } | |
466 | EXPORT_SYMBOL_GPL(inet_csk_route_child_sock); | |
467 | ||
6b72977b | 468 | static inline u32 inet_synq_hash(const __be32 raddr, const __be16 rport, |
72a3effa | 469 | const u32 rnd, const u32 synq_hsize) |
3f421baa | 470 | { |
6b72977b | 471 | return jhash_2words((__force u32)raddr, (__force u32)rport, rnd) & (synq_hsize - 1); |
3f421baa ACM |
472 | } |
473 | ||
dfd56b8b | 474 | #if IS_ENABLED(CONFIG_IPV6) |
3f421baa ACM |
475 | #define AF_INET_FAMILY(fam) ((fam) == AF_INET) |
476 | #else | |
477 | #define AF_INET_FAMILY(fam) 1 | |
478 | #endif | |
479 | ||
480 | struct request_sock *inet_csk_search_req(const struct sock *sk, | |
481 | struct request_sock ***prevp, | |
6b72977b | 482 | const __be16 rport, const __be32 raddr, |
7f25afbb | 483 | const __be32 laddr) |
3f421baa ACM |
484 | { |
485 | const struct inet_connection_sock *icsk = inet_csk(sk); | |
486 | struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt; | |
487 | struct request_sock *req, **prev; | |
488 | ||
489 | for (prev = &lopt->syn_table[inet_synq_hash(raddr, rport, lopt->hash_rnd, | |
490 | lopt->nr_table_entries)]; | |
491 | (req = *prev) != NULL; | |
492 | prev = &req->dl_next) { | |
493 | const struct inet_request_sock *ireq = inet_rsk(req); | |
494 | ||
634fb979 ED |
495 | if (ireq->ir_rmt_port == rport && |
496 | ireq->ir_rmt_addr == raddr && | |
497 | ireq->ir_loc_addr == laddr && | |
3f421baa | 498 | AF_INET_FAMILY(req->rsk_ops->family)) { |
547b792c | 499 | WARN_ON(req->sk); |
3f421baa ACM |
500 | *prevp = prev; |
501 | break; | |
502 | } | |
503 | } | |
504 | ||
505 | return req; | |
506 | } | |
3f421baa ACM |
507 | EXPORT_SYMBOL_GPL(inet_csk_search_req); |
508 | ||
509 | void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req, | |
c2977c22 | 510 | unsigned long timeout) |
3f421baa ACM |
511 | { |
512 | struct inet_connection_sock *icsk = inet_csk(sk); | |
513 | struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt; | |
634fb979 ED |
514 | const u32 h = inet_synq_hash(inet_rsk(req)->ir_rmt_addr, |
515 | inet_rsk(req)->ir_rmt_port, | |
3f421baa ACM |
516 | lopt->hash_rnd, lopt->nr_table_entries); |
517 | ||
518 | reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout); | |
519 | inet_csk_reqsk_queue_added(sk, timeout); | |
520 | } | |
4bc2f18b | 521 | EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add); |
3f421baa | 522 | |
a019d6fe ACM |
523 | /* Only thing we need from tcp.h */ |
524 | extern int sysctl_tcp_synack_retries; | |
525 | ||
9f1d2604 | 526 | |
0c3d79bc JA |
527 | /* Decide when to expire the request and when to resend SYN-ACK */ |
528 | static inline void syn_ack_recalc(struct request_sock *req, const int thresh, | |
529 | const int max_retries, | |
530 | const u8 rskq_defer_accept, | |
531 | int *expire, int *resend) | |
532 | { | |
533 | if (!rskq_defer_accept) { | |
e6c022a4 | 534 | *expire = req->num_timeout >= thresh; |
0c3d79bc JA |
535 | *resend = 1; |
536 | return; | |
537 | } | |
e6c022a4 ED |
538 | *expire = req->num_timeout >= thresh && |
539 | (!inet_rsk(req)->acked || req->num_timeout >= max_retries); | |
0c3d79bc JA |
540 | /* |
541 | * Do not resend while waiting for data after ACK, | |
542 | * start to resend on end of deferring period to give | |
543 | * last chance for data or ACK to create established socket. | |
544 | */ | |
545 | *resend = !inet_rsk(req)->acked || | |
e6c022a4 | 546 | req->num_timeout >= rskq_defer_accept - 1; |
0c3d79bc JA |
547 | } |
548 | ||
e6c022a4 ED |
549 | int inet_rtx_syn_ack(struct sock *parent, struct request_sock *req) |
550 | { | |
1a2c6181 | 551 | int err = req->rsk_ops->rtx_syn_ack(parent, req); |
e6c022a4 ED |
552 | |
553 | if (!err) | |
554 | req->num_retrans++; | |
555 | return err; | |
556 | } | |
557 | EXPORT_SYMBOL(inet_rtx_syn_ack); | |
558 | ||
a019d6fe ACM |
559 | void inet_csk_reqsk_queue_prune(struct sock *parent, |
560 | const unsigned long interval, | |
561 | const unsigned long timeout, | |
562 | const unsigned long max_rto) | |
563 | { | |
564 | struct inet_connection_sock *icsk = inet_csk(parent); | |
565 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; | |
566 | struct listen_sock *lopt = queue->listen_opt; | |
ec0a1966 DM |
567 | int max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries; |
568 | int thresh = max_retries; | |
a019d6fe ACM |
569 | unsigned long now = jiffies; |
570 | struct request_sock **reqp, *req; | |
571 | int i, budget; | |
572 | ||
573 | if (lopt == NULL || lopt->qlen == 0) | |
574 | return; | |
575 | ||
576 | /* Normally all the openreqs are young and become mature | |
577 | * (i.e. converted to established socket) for first timeout. | |
fd4f2cea | 578 | * If synack was not acknowledged for 1 second, it means |
a019d6fe ACM |
579 | * one of the following things: synack was lost, ack was lost, |
580 | * rtt is high or nobody planned to ack (i.e. synflood). | |
581 | * When server is a bit loaded, queue is populated with old | |
582 | * open requests, reducing effective size of queue. | |
583 | * When server is well loaded, queue size reduces to zero | |
584 | * after several minutes of work. It is not synflood, | |
585 | * it is normal operation. The solution is pruning | |
586 | * too old entries overriding normal timeout, when | |
587 | * situation becomes dangerous. | |
588 | * | |
589 | * Essentially, we reserve half of room for young | |
590 | * embrions; and abort old ones without pity, if old | |
591 | * ones are about to clog our table. | |
592 | */ | |
593 | if (lopt->qlen>>(lopt->max_qlen_log-1)) { | |
594 | int young = (lopt->qlen_young<<1); | |
595 | ||
596 | while (thresh > 2) { | |
597 | if (lopt->qlen < young) | |
598 | break; | |
599 | thresh--; | |
600 | young <<= 1; | |
601 | } | |
602 | } | |
603 | ||
ec0a1966 DM |
604 | if (queue->rskq_defer_accept) |
605 | max_retries = queue->rskq_defer_accept; | |
606 | ||
a019d6fe ACM |
607 | budget = 2 * (lopt->nr_table_entries / (timeout / interval)); |
608 | i = lopt->clock_hand; | |
609 | ||
610 | do { | |
611 | reqp=&lopt->syn_table[i]; | |
612 | while ((req = *reqp) != NULL) { | |
613 | if (time_after_eq(now, req->expires)) { | |
0c3d79bc JA |
614 | int expire = 0, resend = 0; |
615 | ||
616 | syn_ack_recalc(req, thresh, max_retries, | |
617 | queue->rskq_defer_accept, | |
618 | &expire, &resend); | |
c72e1183 | 619 | req->rsk_ops->syn_ack_timeout(parent, req); |
0c3d79bc JA |
620 | if (!expire && |
621 | (!resend || | |
e6c022a4 | 622 | !inet_rtx_syn_ack(parent, req) || |
0c3d79bc | 623 | inet_rsk(req)->acked)) { |
a019d6fe ACM |
624 | unsigned long timeo; |
625 | ||
e6c022a4 | 626 | if (req->num_timeout++ == 0) |
a019d6fe | 627 | lopt->qlen_young--; |
e6c022a4 ED |
628 | timeo = min(timeout << req->num_timeout, |
629 | max_rto); | |
a019d6fe ACM |
630 | req->expires = now + timeo; |
631 | reqp = &req->dl_next; | |
632 | continue; | |
633 | } | |
634 | ||
635 | /* Drop this request */ | |
636 | inet_csk_reqsk_queue_unlink(parent, req, reqp); | |
637 | reqsk_queue_removed(queue, req); | |
638 | reqsk_free(req); | |
639 | continue; | |
640 | } | |
641 | reqp = &req->dl_next; | |
642 | } | |
643 | ||
644 | i = (i + 1) & (lopt->nr_table_entries - 1); | |
645 | ||
646 | } while (--budget > 0); | |
647 | ||
648 | lopt->clock_hand = i; | |
649 | ||
650 | if (lopt->qlen) | |
651 | inet_csk_reset_keepalive_timer(parent, interval); | |
652 | } | |
a019d6fe ACM |
653 | EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_prune); |
654 | ||
e56c57d0 ED |
655 | /** |
656 | * inet_csk_clone_lock - clone an inet socket, and lock its clone | |
657 | * @sk: the socket to clone | |
658 | * @req: request_sock | |
659 | * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) | |
660 | * | |
661 | * Caller must unlock socket even in error path (bh_unlock_sock(newsk)) | |
662 | */ | |
663 | struct sock *inet_csk_clone_lock(const struct sock *sk, | |
664 | const struct request_sock *req, | |
665 | const gfp_t priority) | |
9f1d2604 | 666 | { |
e56c57d0 | 667 | struct sock *newsk = sk_clone_lock(sk, priority); |
9f1d2604 ACM |
668 | |
669 | if (newsk != NULL) { | |
670 | struct inet_connection_sock *newicsk = inet_csk(newsk); | |
671 | ||
672 | newsk->sk_state = TCP_SYN_RECV; | |
673 | newicsk->icsk_bind_hash = NULL; | |
674 | ||
634fb979 | 675 | inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port; |
b44084c2 ED |
676 | inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num; |
677 | inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num); | |
9f1d2604 ACM |
678 | newsk->sk_write_space = sk_stream_write_space; |
679 | ||
84f39b08 LC |
680 | newsk->sk_mark = inet_rsk(req)->ir_mark; |
681 | ||
9f1d2604 | 682 | newicsk->icsk_retransmits = 0; |
6687e988 ACM |
683 | newicsk->icsk_backoff = 0; |
684 | newicsk->icsk_probes_out = 0; | |
9f1d2604 ACM |
685 | |
686 | /* Deinitialize accept_queue to trap illegal accesses. */ | |
687 | memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue)); | |
4237c75c VY |
688 | |
689 | security_inet_csk_clone(newsk, req); | |
9f1d2604 ACM |
690 | } |
691 | return newsk; | |
692 | } | |
e56c57d0 | 693 | EXPORT_SYMBOL_GPL(inet_csk_clone_lock); |
a019d6fe ACM |
694 | |
695 | /* | |
696 | * At this point, there should be no process reference to this | |
697 | * socket, and thus no user references at all. Therefore we | |
698 | * can assume the socket waitqueue is inactive and nobody will | |
699 | * try to jump onto it. | |
700 | */ | |
701 | void inet_csk_destroy_sock(struct sock *sk) | |
702 | { | |
547b792c IJ |
703 | WARN_ON(sk->sk_state != TCP_CLOSE); |
704 | WARN_ON(!sock_flag(sk, SOCK_DEAD)); | |
a019d6fe ACM |
705 | |
706 | /* It cannot be in hash table! */ | |
547b792c | 707 | WARN_ON(!sk_unhashed(sk)); |
a019d6fe | 708 | |
c720c7e8 ED |
709 | /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */ |
710 | WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash); | |
a019d6fe ACM |
711 | |
712 | sk->sk_prot->destroy(sk); | |
713 | ||
714 | sk_stream_kill_queues(sk); | |
715 | ||
716 | xfrm_sk_free_policy(sk); | |
717 | ||
718 | sk_refcnt_debug_release(sk); | |
719 | ||
dd24c001 | 720 | percpu_counter_dec(sk->sk_prot->orphan_count); |
a019d6fe ACM |
721 | sock_put(sk); |
722 | } | |
a019d6fe ACM |
723 | EXPORT_SYMBOL(inet_csk_destroy_sock); |
724 | ||
e337e24d CP |
725 | /* This function allows to force a closure of a socket after the call to |
726 | * tcp/dccp_create_openreq_child(). | |
727 | */ | |
728 | void inet_csk_prepare_forced_close(struct sock *sk) | |
c10cb5fc | 729 | __releases(&sk->sk_lock.slock) |
e337e24d CP |
730 | { |
731 | /* sk_clone_lock locked the socket and set refcnt to 2 */ | |
732 | bh_unlock_sock(sk); | |
733 | sock_put(sk); | |
734 | ||
735 | /* The below has to be done to allow calling inet_csk_destroy_sock */ | |
736 | sock_set_flag(sk, SOCK_DEAD); | |
737 | percpu_counter_inc(sk->sk_prot->orphan_count); | |
738 | inet_sk(sk)->inet_num = 0; | |
739 | } | |
740 | EXPORT_SYMBOL(inet_csk_prepare_forced_close); | |
741 | ||
a019d6fe ACM |
742 | int inet_csk_listen_start(struct sock *sk, const int nr_table_entries) |
743 | { | |
744 | struct inet_sock *inet = inet_sk(sk); | |
745 | struct inet_connection_sock *icsk = inet_csk(sk); | |
746 | int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries); | |
747 | ||
748 | if (rc != 0) | |
749 | return rc; | |
750 | ||
751 | sk->sk_max_ack_backlog = 0; | |
752 | sk->sk_ack_backlog = 0; | |
753 | inet_csk_delack_init(sk); | |
754 | ||
755 | /* There is race window here: we announce ourselves listening, | |
756 | * but this transition is still not validated by get_port(). | |
757 | * It is OK, because this socket enters to hash table only | |
758 | * after validation is complete. | |
759 | */ | |
760 | sk->sk_state = TCP_LISTEN; | |
c720c7e8 ED |
761 | if (!sk->sk_prot->get_port(sk, inet->inet_num)) { |
762 | inet->inet_sport = htons(inet->inet_num); | |
a019d6fe ACM |
763 | |
764 | sk_dst_reset(sk); | |
765 | sk->sk_prot->hash(sk); | |
766 | ||
767 | return 0; | |
768 | } | |
769 | ||
770 | sk->sk_state = TCP_CLOSE; | |
771 | __reqsk_queue_destroy(&icsk->icsk_accept_queue); | |
772 | return -EADDRINUSE; | |
773 | } | |
a019d6fe ACM |
774 | EXPORT_SYMBOL_GPL(inet_csk_listen_start); |
775 | ||
776 | /* | |
777 | * This routine closes sockets which have been at least partially | |
778 | * opened, but not yet accepted. | |
779 | */ | |
780 | void inet_csk_listen_stop(struct sock *sk) | |
781 | { | |
782 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 783 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
a019d6fe ACM |
784 | struct request_sock *acc_req; |
785 | struct request_sock *req; | |
786 | ||
787 | inet_csk_delete_keepalive_timer(sk); | |
788 | ||
789 | /* make all the listen_opt local to us */ | |
8336886f | 790 | acc_req = reqsk_queue_yank_acceptq(queue); |
a019d6fe ACM |
791 | |
792 | /* Following specs, it would be better either to send FIN | |
793 | * (and enter FIN-WAIT-1, it is normal close) | |
794 | * or to send active reset (abort). | |
795 | * Certainly, it is pretty dangerous while synflood, but it is | |
796 | * bad justification for our negligence 8) | |
797 | * To be honest, we are not able to make either | |
798 | * of the variants now. --ANK | |
799 | */ | |
8336886f | 800 | reqsk_queue_destroy(queue); |
a019d6fe ACM |
801 | |
802 | while ((req = acc_req) != NULL) { | |
803 | struct sock *child = req->sk; | |
804 | ||
805 | acc_req = req->dl_next; | |
806 | ||
807 | local_bh_disable(); | |
808 | bh_lock_sock(child); | |
547b792c | 809 | WARN_ON(sock_owned_by_user(child)); |
a019d6fe ACM |
810 | sock_hold(child); |
811 | ||
812 | sk->sk_prot->disconnect(child, O_NONBLOCK); | |
813 | ||
814 | sock_orphan(child); | |
815 | ||
eb4dea58 HX |
816 | percpu_counter_inc(sk->sk_prot->orphan_count); |
817 | ||
7ab4551f | 818 | if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->listener) { |
8336886f JC |
819 | BUG_ON(tcp_sk(child)->fastopen_rsk != req); |
820 | BUG_ON(sk != tcp_rsk(req)->listener); | |
821 | ||
822 | /* Paranoid, to prevent race condition if | |
823 | * an inbound pkt destined for child is | |
824 | * blocked by sock lock in tcp_v4_rcv(). | |
825 | * Also to satisfy an assertion in | |
826 | * tcp_v4_destroy_sock(). | |
827 | */ | |
828 | tcp_sk(child)->fastopen_rsk = NULL; | |
829 | sock_put(sk); | |
830 | } | |
a019d6fe ACM |
831 | inet_csk_destroy_sock(child); |
832 | ||
833 | bh_unlock_sock(child); | |
834 | local_bh_enable(); | |
835 | sock_put(child); | |
836 | ||
837 | sk_acceptq_removed(sk); | |
838 | __reqsk_free(req); | |
839 | } | |
8336886f JC |
840 | if (queue->fastopenq != NULL) { |
841 | /* Free all the reqs queued in rskq_rst_head. */ | |
842 | spin_lock_bh(&queue->fastopenq->lock); | |
843 | acc_req = queue->fastopenq->rskq_rst_head; | |
844 | queue->fastopenq->rskq_rst_head = NULL; | |
845 | spin_unlock_bh(&queue->fastopenq->lock); | |
846 | while ((req = acc_req) != NULL) { | |
847 | acc_req = req->dl_next; | |
848 | __reqsk_free(req); | |
849 | } | |
850 | } | |
547b792c | 851 | WARN_ON(sk->sk_ack_backlog); |
a019d6fe | 852 | } |
a019d6fe | 853 | EXPORT_SYMBOL_GPL(inet_csk_listen_stop); |
af05dc93 ACM |
854 | |
855 | void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr) | |
856 | { | |
857 | struct sockaddr_in *sin = (struct sockaddr_in *)uaddr; | |
858 | const struct inet_sock *inet = inet_sk(sk); | |
859 | ||
860 | sin->sin_family = AF_INET; | |
c720c7e8 ED |
861 | sin->sin_addr.s_addr = inet->inet_daddr; |
862 | sin->sin_port = inet->inet_dport; | |
af05dc93 | 863 | } |
af05dc93 | 864 | EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr); |
c4d93909 | 865 | |
dec73ff0 ACM |
866 | #ifdef CONFIG_COMPAT |
867 | int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname, | |
868 | char __user *optval, int __user *optlen) | |
869 | { | |
dbeff12b | 870 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 ACM |
871 | |
872 | if (icsk->icsk_af_ops->compat_getsockopt != NULL) | |
873 | return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname, | |
874 | optval, optlen); | |
875 | return icsk->icsk_af_ops->getsockopt(sk, level, optname, | |
876 | optval, optlen); | |
877 | } | |
dec73ff0 ACM |
878 | EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt); |
879 | ||
880 | int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 881 | char __user *optval, unsigned int optlen) |
dec73ff0 | 882 | { |
dbeff12b | 883 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 ACM |
884 | |
885 | if (icsk->icsk_af_ops->compat_setsockopt != NULL) | |
886 | return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname, | |
887 | optval, optlen); | |
888 | return icsk->icsk_af_ops->setsockopt(sk, level, optname, | |
889 | optval, optlen); | |
890 | } | |
dec73ff0 ACM |
891 | EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt); |
892 | #endif | |
80d0a69f DM |
893 | |
894 | static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl) | |
895 | { | |
5abf7f7e ED |
896 | const struct inet_sock *inet = inet_sk(sk); |
897 | const struct ip_options_rcu *inet_opt; | |
80d0a69f DM |
898 | __be32 daddr = inet->inet_daddr; |
899 | struct flowi4 *fl4; | |
900 | struct rtable *rt; | |
901 | ||
902 | rcu_read_lock(); | |
903 | inet_opt = rcu_dereference(inet->inet_opt); | |
904 | if (inet_opt && inet_opt->opt.srr) | |
905 | daddr = inet_opt->opt.faddr; | |
906 | fl4 = &fl->u.ip4; | |
907 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, | |
908 | inet->inet_saddr, inet->inet_dport, | |
909 | inet->inet_sport, sk->sk_protocol, | |
910 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if); | |
911 | if (IS_ERR(rt)) | |
912 | rt = NULL; | |
913 | if (rt) | |
914 | sk_setup_caps(sk, &rt->dst); | |
915 | rcu_read_unlock(); | |
916 | ||
917 | return &rt->dst; | |
918 | } | |
919 | ||
920 | struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu) | |
921 | { | |
922 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
923 | struct inet_sock *inet = inet_sk(sk); | |
924 | ||
925 | if (!dst) { | |
926 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
927 | if (!dst) | |
928 | goto out; | |
929 | } | |
6700c270 | 930 | dst->ops->update_pmtu(dst, sk, NULL, mtu); |
80d0a69f DM |
931 | |
932 | dst = __sk_dst_check(sk, 0); | |
933 | if (!dst) | |
934 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
935 | out: | |
936 | return dst; | |
937 | } | |
938 | EXPORT_SYMBOL_GPL(inet_csk_update_pmtu); |