]>
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> |
fa76ce73 | 26 | #include <net/tcp.h> |
3f421baa ACM |
27 | |
28 | #ifdef INET_CSK_DEBUG | |
29 | const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n"; | |
30 | EXPORT_SYMBOL(inet_csk_timer_bug_msg); | |
31 | #endif | |
32 | ||
0bbf87d8 | 33 | void inet_get_local_port_range(struct net *net, int *low, int *high) |
227b60f5 | 34 | { |
95c96174 ED |
35 | unsigned int seq; |
36 | ||
227b60f5 | 37 | do { |
c9d8f1a6 | 38 | seq = read_seqbegin(&net->ipv4.ip_local_ports.lock); |
227b60f5 | 39 | |
c9d8f1a6 CW |
40 | *low = net->ipv4.ip_local_ports.range[0]; |
41 | *high = net->ipv4.ip_local_ports.range[1]; | |
42 | } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq)); | |
227b60f5 SH |
43 | } |
44 | EXPORT_SYMBOL(inet_get_local_port_range); | |
3f421baa | 45 | |
971af18b | 46 | int inet_csk_bind_conflict(const struct sock *sk, |
aacd9289 | 47 | const struct inet_bind_bucket *tb, bool relax) |
3f421baa | 48 | { |
3f421baa | 49 | struct sock *sk2; |
3f421baa | 50 | int reuse = sk->sk_reuse; |
da5e3630 TH |
51 | int reuseport = sk->sk_reuseport; |
52 | kuid_t uid = sock_i_uid((struct sock *)sk); | |
3f421baa | 53 | |
7477fd2e PE |
54 | /* |
55 | * Unlike other sk lookup places we do not check | |
56 | * for sk_net here, since _all_ the socks listed | |
57 | * in tb->owners list belong to the same net - the | |
58 | * one this bucket belongs to. | |
59 | */ | |
60 | ||
b67bfe0d | 61 | sk_for_each_bound(sk2, &tb->owners) { |
3f421baa ACM |
62 | if (sk != sk2 && |
63 | !inet_v6_ipv6only(sk2) && | |
64 | (!sk->sk_bound_dev_if || | |
65 | !sk2->sk_bound_dev_if || | |
66 | sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) { | |
da5e3630 TH |
67 | if ((!reuse || !sk2->sk_reuse || |
68 | sk2->sk_state == TCP_LISTEN) && | |
69 | (!reuseport || !sk2->sk_reuseport || | |
70 | (sk2->sk_state != TCP_TIME_WAIT && | |
71 | !uid_eq(uid, sock_i_uid(sk2))))) { | |
50805466 ED |
72 | |
73 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || | |
74 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
3f421baa | 75 | break; |
8d238b25 | 76 | } |
aacd9289 AC |
77 | if (!relax && reuse && sk2->sk_reuse && |
78 | sk2->sk_state != TCP_LISTEN) { | |
aacd9289 | 79 | |
50805466 ED |
80 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || |
81 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
aacd9289 AC |
82 | break; |
83 | } | |
3f421baa ACM |
84 | } |
85 | } | |
b67bfe0d | 86 | return sk2 != NULL; |
3f421baa | 87 | } |
971af18b ACM |
88 | EXPORT_SYMBOL_GPL(inet_csk_bind_conflict); |
89 | ||
3f421baa ACM |
90 | /* Obtain a reference to a local port for the given sock, |
91 | * if snum is zero it means select any available local port. | |
92 | */ | |
ab1e0a13 | 93 | int inet_csk_get_port(struct sock *sk, unsigned short snum) |
3f421baa | 94 | { |
39d8cda7 | 95 | struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo; |
3f421baa | 96 | struct inet_bind_hashbucket *head; |
3f421baa | 97 | struct inet_bind_bucket *tb; |
a9d8f911 | 98 | int ret, attempts = 5; |
3b1e0a65 | 99 | struct net *net = sock_net(sk); |
a9d8f911 | 100 | int smallest_size = -1, smallest_rover; |
da5e3630 | 101 | kuid_t uid = sock_i_uid(sk); |
946f9eb2 | 102 | int attempt_half = (sk->sk_reuse == SK_CAN_REUSE) ? 1 : 0; |
3f421baa ACM |
103 | |
104 | local_bh_disable(); | |
105 | if (!snum) { | |
227b60f5 SH |
106 | int remaining, rover, low, high; |
107 | ||
a9d8f911 | 108 | again: |
0bbf87d8 | 109 | inet_get_local_port_range(net, &low, &high); |
946f9eb2 ED |
110 | if (attempt_half) { |
111 | int half = low + ((high - low) >> 1); | |
112 | ||
113 | if (attempt_half == 1) | |
114 | high = half; | |
115 | else | |
116 | low = half; | |
117 | } | |
a25de534 | 118 | remaining = (high - low) + 1; |
63862b5b | 119 | smallest_rover = rover = prandom_u32() % remaining + low; |
3f421baa | 120 | |
a9d8f911 | 121 | smallest_size = -1; |
3f421baa | 122 | do { |
122ff243 | 123 | if (inet_is_local_reserved_port(net, rover)) |
e3826f1e | 124 | goto next_nolock; |
7f635ab7 PE |
125 | head = &hashinfo->bhash[inet_bhashfn(net, rover, |
126 | hashinfo->bhash_size)]; | |
3f421baa | 127 | spin_lock(&head->lock); |
b67bfe0d | 128 | inet_bind_bucket_for_each(tb, &head->chain) |
09ad9bc7 | 129 | if (net_eq(ib_net(tb), net) && tb->port == rover) { |
da5e3630 TH |
130 | if (((tb->fastreuse > 0 && |
131 | sk->sk_reuse && | |
132 | sk->sk_state != TCP_LISTEN) || | |
133 | (tb->fastreuseport > 0 && | |
134 | sk->sk_reuseport && | |
135 | uid_eq(tb->fastuid, uid))) && | |
a9d8f911 EP |
136 | (tb->num_owners < smallest_size || smallest_size == -1)) { |
137 | smallest_size = tb->num_owners; | |
138 | smallest_rover = rover; | |
a9d8f911 | 139 | } |
aacd9289 | 140 | if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) { |
2b05ad33 | 141 | snum = rover; |
fddb7b57 | 142 | goto tb_found; |
2b05ad33 | 143 | } |
3f421baa | 144 | goto next; |
a9d8f911 | 145 | } |
3f421baa ACM |
146 | break; |
147 | next: | |
148 | spin_unlock(&head->lock); | |
e3826f1e | 149 | next_nolock: |
6df71634 SH |
150 | if (++rover > high) |
151 | rover = low; | |
3f421baa | 152 | } while (--remaining > 0); |
3f421baa ACM |
153 | |
154 | /* Exhausted local port range during search? It is not | |
155 | * possible for us to be holding one of the bind hash | |
156 | * locks if this test triggers, because if 'remaining' | |
157 | * drops to zero, we broke out of the do/while loop at | |
158 | * the top level, not from the 'break;' statement. | |
159 | */ | |
160 | ret = 1; | |
a9d8f911 EP |
161 | if (remaining <= 0) { |
162 | if (smallest_size != -1) { | |
163 | snum = smallest_rover; | |
164 | goto have_snum; | |
165 | } | |
946f9eb2 ED |
166 | if (attempt_half == 1) { |
167 | /* OK we now try the upper half of the range */ | |
168 | attempt_half = 2; | |
169 | goto again; | |
170 | } | |
3f421baa | 171 | goto fail; |
a9d8f911 | 172 | } |
3f421baa ACM |
173 | /* OK, here is the one we will use. HEAD is |
174 | * non-NULL and we hold it's mutex. | |
175 | */ | |
176 | snum = rover; | |
177 | } else { | |
a9d8f911 | 178 | have_snum: |
7f635ab7 PE |
179 | head = &hashinfo->bhash[inet_bhashfn(net, snum, |
180 | hashinfo->bhash_size)]; | |
3f421baa | 181 | spin_lock(&head->lock); |
b67bfe0d | 182 | inet_bind_bucket_for_each(tb, &head->chain) |
09ad9bc7 | 183 | if (net_eq(ib_net(tb), net) && tb->port == snum) |
3f421baa ACM |
184 | goto tb_found; |
185 | } | |
186 | tb = NULL; | |
187 | goto tb_not_found; | |
188 | tb_found: | |
189 | if (!hlist_empty(&tb->owners)) { | |
4a17fd52 PE |
190 | if (sk->sk_reuse == SK_FORCE_REUSE) |
191 | goto success; | |
192 | ||
da5e3630 TH |
193 | if (((tb->fastreuse > 0 && |
194 | sk->sk_reuse && sk->sk_state != TCP_LISTEN) || | |
195 | (tb->fastreuseport > 0 && | |
196 | sk->sk_reuseport && uid_eq(tb->fastuid, uid))) && | |
a9d8f911 | 197 | smallest_size == -1) { |
3f421baa ACM |
198 | goto success; |
199 | } else { | |
200 | ret = 1; | |
aacd9289 | 201 | if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true)) { |
da5e3630 | 202 | if (((sk->sk_reuse && sk->sk_state != TCP_LISTEN) || |
9c5e0c0b TH |
203 | (tb->fastreuseport > 0 && |
204 | sk->sk_reuseport && uid_eq(tb->fastuid, uid))) && | |
5add3009 | 205 | smallest_size != -1 && --attempts >= 0) { |
a9d8f911 EP |
206 | spin_unlock(&head->lock); |
207 | goto again; | |
208 | } | |
aacd9289 | 209 | |
3f421baa | 210 | goto fail_unlock; |
a9d8f911 | 211 | } |
3f421baa ACM |
212 | } |
213 | } | |
214 | tb_not_found: | |
215 | ret = 1; | |
941b1d22 PE |
216 | if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep, |
217 | net, head, snum)) == NULL) | |
3f421baa ACM |
218 | goto fail_unlock; |
219 | if (hlist_empty(&tb->owners)) { | |
220 | if (sk->sk_reuse && sk->sk_state != TCP_LISTEN) | |
221 | tb->fastreuse = 1; | |
222 | else | |
223 | tb->fastreuse = 0; | |
da5e3630 TH |
224 | if (sk->sk_reuseport) { |
225 | tb->fastreuseport = 1; | |
226 | tb->fastuid = uid; | |
9c5e0c0b | 227 | } else |
da5e3630 | 228 | tb->fastreuseport = 0; |
da5e3630 TH |
229 | } else { |
230 | if (tb->fastreuse && | |
231 | (!sk->sk_reuse || sk->sk_state == TCP_LISTEN)) | |
232 | tb->fastreuse = 0; | |
233 | if (tb->fastreuseport && | |
9c5e0c0b | 234 | (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid))) |
da5e3630 | 235 | tb->fastreuseport = 0; |
da5e3630 | 236 | } |
3f421baa ACM |
237 | success: |
238 | if (!inet_csk(sk)->icsk_bind_hash) | |
239 | inet_bind_hash(sk, tb, snum); | |
547b792c | 240 | WARN_ON(inet_csk(sk)->icsk_bind_hash != tb); |
e905a9ed | 241 | ret = 0; |
3f421baa ACM |
242 | |
243 | fail_unlock: | |
244 | spin_unlock(&head->lock); | |
245 | fail: | |
246 | local_bh_enable(); | |
247 | return ret; | |
248 | } | |
3f421baa ACM |
249 | EXPORT_SYMBOL_GPL(inet_csk_get_port); |
250 | ||
251 | /* | |
252 | * Wait for an incoming connection, avoid race conditions. This must be called | |
253 | * with the socket locked. | |
254 | */ | |
255 | static int inet_csk_wait_for_connect(struct sock *sk, long timeo) | |
256 | { | |
257 | struct inet_connection_sock *icsk = inet_csk(sk); | |
258 | DEFINE_WAIT(wait); | |
259 | int err; | |
260 | ||
261 | /* | |
262 | * True wake-one mechanism for incoming connections: only | |
263 | * one process gets woken up, not the 'whole herd'. | |
264 | * Since we do not 'race & poll' for established sockets | |
265 | * anymore, the common case will execute the loop only once. | |
266 | * | |
267 | * Subtle issue: "add_wait_queue_exclusive()" will be added | |
268 | * after any current non-exclusive waiters, and we know that | |
269 | * it will always _stay_ after any new non-exclusive waiters | |
270 | * because all non-exclusive waiters are added at the | |
271 | * beginning of the wait-queue. As such, it's ok to "drop" | |
272 | * our exclusiveness temporarily when we get woken up without | |
273 | * having to remove and re-insert us on the wait queue. | |
274 | */ | |
275 | for (;;) { | |
aa395145 | 276 | prepare_to_wait_exclusive(sk_sleep(sk), &wait, |
3f421baa ACM |
277 | TASK_INTERRUPTIBLE); |
278 | release_sock(sk); | |
279 | if (reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
280 | timeo = schedule_timeout(timeo); | |
cb7cf8a3 | 281 | sched_annotate_sleep(); |
3f421baa ACM |
282 | lock_sock(sk); |
283 | err = 0; | |
284 | if (!reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
285 | break; | |
286 | err = -EINVAL; | |
287 | if (sk->sk_state != TCP_LISTEN) | |
288 | break; | |
289 | err = sock_intr_errno(timeo); | |
290 | if (signal_pending(current)) | |
291 | break; | |
292 | err = -EAGAIN; | |
293 | if (!timeo) | |
294 | break; | |
295 | } | |
aa395145 | 296 | finish_wait(sk_sleep(sk), &wait); |
3f421baa ACM |
297 | return err; |
298 | } | |
299 | ||
300 | /* | |
301 | * This will accept the next outstanding connection. | |
302 | */ | |
303 | struct sock *inet_csk_accept(struct sock *sk, int flags, int *err) | |
304 | { | |
305 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 306 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
8336886f | 307 | struct request_sock *req; |
e3d95ad7 | 308 | struct sock *newsk; |
3f421baa ACM |
309 | int error; |
310 | ||
311 | lock_sock(sk); | |
312 | ||
313 | /* We need to make sure that this socket is listening, | |
314 | * and that it has something pending. | |
315 | */ | |
316 | error = -EINVAL; | |
317 | if (sk->sk_state != TCP_LISTEN) | |
318 | goto out_err; | |
319 | ||
320 | /* Find already established connection */ | |
8336886f | 321 | if (reqsk_queue_empty(queue)) { |
3f421baa ACM |
322 | long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); |
323 | ||
324 | /* If this is a non blocking socket don't sleep */ | |
325 | error = -EAGAIN; | |
326 | if (!timeo) | |
327 | goto out_err; | |
328 | ||
329 | error = inet_csk_wait_for_connect(sk, timeo); | |
330 | if (error) | |
331 | goto out_err; | |
332 | } | |
8336886f JC |
333 | req = reqsk_queue_remove(queue); |
334 | newsk = req->sk; | |
335 | ||
336 | sk_acceptq_removed(sk); | |
e3d95ad7 | 337 | if (sk->sk_protocol == IPPROTO_TCP && |
0536fcc0 ED |
338 | tcp_rsk(req)->tfo_listener) { |
339 | spin_lock_bh(&queue->fastopenq.lock); | |
9439ce00 | 340 | if (tcp_rsk(req)->tfo_listener) { |
8336886f JC |
341 | /* We are still waiting for the final ACK from 3WHS |
342 | * so can't free req now. Instead, we set req->sk to | |
343 | * NULL to signify that the child socket is taken | |
344 | * so reqsk_fastopen_remove() will free the req | |
345 | * when 3WHS finishes (or is aborted). | |
346 | */ | |
347 | req->sk = NULL; | |
348 | req = NULL; | |
349 | } | |
0536fcc0 | 350 | spin_unlock_bh(&queue->fastopenq.lock); |
8336886f | 351 | } |
3f421baa ACM |
352 | out: |
353 | release_sock(sk); | |
8336886f | 354 | if (req) |
13854e5a | 355 | reqsk_put(req); |
3f421baa ACM |
356 | return newsk; |
357 | out_err: | |
358 | newsk = NULL; | |
8336886f | 359 | req = NULL; |
3f421baa ACM |
360 | *err = error; |
361 | goto out; | |
362 | } | |
3f421baa ACM |
363 | EXPORT_SYMBOL(inet_csk_accept); |
364 | ||
365 | /* | |
366 | * Using different timers for retransmit, delayed acks and probes | |
e905a9ed | 367 | * We may wish use just one timer maintaining a list of expire jiffies |
3f421baa ACM |
368 | * to optimize. |
369 | */ | |
370 | void inet_csk_init_xmit_timers(struct sock *sk, | |
371 | void (*retransmit_handler)(unsigned long), | |
372 | void (*delack_handler)(unsigned long), | |
373 | void (*keepalive_handler)(unsigned long)) | |
374 | { | |
375 | struct inet_connection_sock *icsk = inet_csk(sk); | |
376 | ||
b24b8a24 PE |
377 | setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler, |
378 | (unsigned long)sk); | |
379 | setup_timer(&icsk->icsk_delack_timer, delack_handler, | |
380 | (unsigned long)sk); | |
381 | setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk); | |
3f421baa ACM |
382 | icsk->icsk_pending = icsk->icsk_ack.pending = 0; |
383 | } | |
3f421baa ACM |
384 | EXPORT_SYMBOL(inet_csk_init_xmit_timers); |
385 | ||
386 | void inet_csk_clear_xmit_timers(struct sock *sk) | |
387 | { | |
388 | struct inet_connection_sock *icsk = inet_csk(sk); | |
389 | ||
390 | icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0; | |
391 | ||
392 | sk_stop_timer(sk, &icsk->icsk_retransmit_timer); | |
393 | sk_stop_timer(sk, &icsk->icsk_delack_timer); | |
394 | sk_stop_timer(sk, &sk->sk_timer); | |
395 | } | |
3f421baa ACM |
396 | EXPORT_SYMBOL(inet_csk_clear_xmit_timers); |
397 | ||
398 | void inet_csk_delete_keepalive_timer(struct sock *sk) | |
399 | { | |
400 | sk_stop_timer(sk, &sk->sk_timer); | |
401 | } | |
3f421baa ACM |
402 | EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); |
403 | ||
404 | void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) | |
405 | { | |
406 | sk_reset_timer(sk, &sk->sk_timer, jiffies + len); | |
407 | } | |
3f421baa ACM |
408 | EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); |
409 | ||
e5895bc6 | 410 | struct dst_entry *inet_csk_route_req(const struct sock *sk, |
6bd023f3 | 411 | struct flowi4 *fl4, |
ba3f7f04 | 412 | const struct request_sock *req) |
3f421baa | 413 | { |
3f421baa | 414 | const struct inet_request_sock *ireq = inet_rsk(req); |
8b929ab1 ED |
415 | struct net *net = read_pnet(&ireq->ireq_net); |
416 | struct ip_options_rcu *opt = ireq->opt; | |
417 | struct rtable *rt; | |
3f421baa | 418 | |
8b929ab1 | 419 | flowi4_init_output(fl4, ireq->ir_iif, ireq->ir_mark, |
e79d9bc7 | 420 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
8b929ab1 | 421 | sk->sk_protocol, inet_sk_flowi_flags(sk), |
634fb979 | 422 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
8b929ab1 ED |
423 | ireq->ir_loc_addr, ireq->ir_rmt_port, |
424 | htons(ireq->ir_num)); | |
6bd023f3 DM |
425 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
426 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe | 427 | if (IS_ERR(rt)) |
857a6e0a | 428 | goto no_route; |
155e8336 | 429 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
857a6e0a | 430 | goto route_err; |
d8d1f30b | 431 | return &rt->dst; |
857a6e0a IJ |
432 | |
433 | route_err: | |
434 | ip_rt_put(rt); | |
435 | no_route: | |
436 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); | |
437 | return NULL; | |
3f421baa | 438 | } |
3f421baa ACM |
439 | EXPORT_SYMBOL_GPL(inet_csk_route_req); |
440 | ||
a2432c4f | 441 | struct dst_entry *inet_csk_route_child_sock(const struct sock *sk, |
77357a95 DM |
442 | struct sock *newsk, |
443 | const struct request_sock *req) | |
444 | { | |
445 | const struct inet_request_sock *ireq = inet_rsk(req); | |
8b929ab1 | 446 | struct net *net = read_pnet(&ireq->ireq_net); |
77357a95 | 447 | struct inet_sock *newinet = inet_sk(newsk); |
1a7b27c9 | 448 | struct ip_options_rcu *opt; |
77357a95 DM |
449 | struct flowi4 *fl4; |
450 | struct rtable *rt; | |
451 | ||
452 | fl4 = &newinet->cork.fl.u.ip4; | |
1a7b27c9 CP |
453 | |
454 | rcu_read_lock(); | |
455 | opt = rcu_dereference(newinet->inet_opt); | |
8b929ab1 | 456 | flowi4_init_output(fl4, ireq->ir_iif, ireq->ir_mark, |
77357a95 DM |
457 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
458 | sk->sk_protocol, inet_sk_flowi_flags(sk), | |
634fb979 | 459 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
8b929ab1 ED |
460 | ireq->ir_loc_addr, ireq->ir_rmt_port, |
461 | htons(ireq->ir_num)); | |
77357a95 DM |
462 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
463 | rt = ip_route_output_flow(net, fl4, sk); | |
464 | if (IS_ERR(rt)) | |
465 | goto no_route; | |
155e8336 | 466 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
77357a95 | 467 | goto route_err; |
1a7b27c9 | 468 | rcu_read_unlock(); |
77357a95 DM |
469 | return &rt->dst; |
470 | ||
471 | route_err: | |
472 | ip_rt_put(rt); | |
473 | no_route: | |
1a7b27c9 | 474 | rcu_read_unlock(); |
77357a95 DM |
475 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
476 | return NULL; | |
477 | } | |
478 | EXPORT_SYMBOL_GPL(inet_csk_route_child_sock); | |
479 | ||
6b72977b | 480 | static inline u32 inet_synq_hash(const __be32 raddr, const __be16 rport, |
72a3effa | 481 | const u32 rnd, const u32 synq_hsize) |
3f421baa | 482 | { |
6b72977b | 483 | return jhash_2words((__force u32)raddr, (__force u32)rport, rnd) & (synq_hsize - 1); |
3f421baa ACM |
484 | } |
485 | ||
dfd56b8b | 486 | #if IS_ENABLED(CONFIG_IPV6) |
3f421baa ACM |
487 | #define AF_INET_FAMILY(fam) ((fam) == AF_INET) |
488 | #else | |
fa76ce73 | 489 | #define AF_INET_FAMILY(fam) true |
3f421baa ACM |
490 | #endif |
491 | ||
fa76ce73 ED |
492 | /* Note: this is temporary : |
493 | * req sock will no longer be in listener hash table | |
494 | */ | |
495 | struct request_sock *inet_csk_search_req(struct sock *sk, | |
496 | const __be16 rport, | |
497 | const __be32 raddr, | |
7f25afbb | 498 | const __be32 laddr) |
3f421baa | 499 | { |
fa76ce73 | 500 | struct inet_connection_sock *icsk = inet_csk(sk); |
3f421baa | 501 | struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt; |
52452c54 | 502 | struct request_sock *req; |
fa76ce73 ED |
503 | u32 hash = inet_synq_hash(raddr, rport, lopt->hash_rnd, |
504 | lopt->nr_table_entries); | |
3f421baa | 505 | |
b2827053 | 506 | spin_lock(&icsk->icsk_accept_queue.syn_wait_lock); |
fa76ce73 | 507 | for (req = lopt->syn_table[hash]; req != NULL; req = req->dl_next) { |
3f421baa ACM |
508 | const struct inet_request_sock *ireq = inet_rsk(req); |
509 | ||
634fb979 ED |
510 | if (ireq->ir_rmt_port == rport && |
511 | ireq->ir_rmt_addr == raddr && | |
512 | ireq->ir_loc_addr == laddr && | |
3f421baa | 513 | AF_INET_FAMILY(req->rsk_ops->family)) { |
fa76ce73 | 514 | atomic_inc(&req->rsk_refcnt); |
547b792c | 515 | WARN_ON(req->sk); |
3f421baa ACM |
516 | break; |
517 | } | |
518 | } | |
b2827053 | 519 | spin_unlock(&icsk->icsk_accept_queue.syn_wait_lock); |
3f421baa ACM |
520 | |
521 | return req; | |
522 | } | |
3f421baa ACM |
523 | EXPORT_SYMBOL_GPL(inet_csk_search_req); |
524 | ||
525 | void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req, | |
c2977c22 | 526 | unsigned long timeout) |
3f421baa ACM |
527 | { |
528 | struct inet_connection_sock *icsk = inet_csk(sk); | |
529 | struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt; | |
634fb979 ED |
530 | const u32 h = inet_synq_hash(inet_rsk(req)->ir_rmt_addr, |
531 | inet_rsk(req)->ir_rmt_port, | |
3f421baa ACM |
532 | lopt->hash_rnd, lopt->nr_table_entries); |
533 | ||
534 | reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout); | |
535 | inet_csk_reqsk_queue_added(sk, timeout); | |
536 | } | |
4bc2f18b | 537 | EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add); |
3f421baa | 538 | |
a019d6fe ACM |
539 | /* Only thing we need from tcp.h */ |
540 | extern int sysctl_tcp_synack_retries; | |
541 | ||
9f1d2604 | 542 | |
0c3d79bc JA |
543 | /* Decide when to expire the request and when to resend SYN-ACK */ |
544 | static inline void syn_ack_recalc(struct request_sock *req, const int thresh, | |
545 | const int max_retries, | |
546 | const u8 rskq_defer_accept, | |
547 | int *expire, int *resend) | |
548 | { | |
549 | if (!rskq_defer_accept) { | |
e6c022a4 | 550 | *expire = req->num_timeout >= thresh; |
0c3d79bc JA |
551 | *resend = 1; |
552 | return; | |
553 | } | |
e6c022a4 ED |
554 | *expire = req->num_timeout >= thresh && |
555 | (!inet_rsk(req)->acked || req->num_timeout >= max_retries); | |
0c3d79bc JA |
556 | /* |
557 | * Do not resend while waiting for data after ACK, | |
558 | * start to resend on end of deferring period to give | |
559 | * last chance for data or ACK to create established socket. | |
560 | */ | |
561 | *resend = !inet_rsk(req)->acked || | |
e6c022a4 | 562 | req->num_timeout >= rskq_defer_accept - 1; |
0c3d79bc JA |
563 | } |
564 | ||
1b70e977 | 565 | int inet_rtx_syn_ack(const struct sock *parent, struct request_sock *req) |
e6c022a4 | 566 | { |
1a2c6181 | 567 | int err = req->rsk_ops->rtx_syn_ack(parent, req); |
e6c022a4 ED |
568 | |
569 | if (!err) | |
570 | req->num_retrans++; | |
571 | return err; | |
572 | } | |
573 | EXPORT_SYMBOL(inet_rtx_syn_ack); | |
574 | ||
b357a364 ED |
575 | /* return true if req was found in the syn_table[] */ |
576 | static bool reqsk_queue_unlink(struct request_sock_queue *queue, | |
577 | struct request_sock *req) | |
578 | { | |
579 | struct listen_sock *lopt = queue->listen_opt; | |
580 | struct request_sock **prev; | |
581 | bool found = false; | |
582 | ||
583 | spin_lock(&queue->syn_wait_lock); | |
584 | ||
585 | for (prev = &lopt->syn_table[req->rsk_hash]; *prev != NULL; | |
586 | prev = &(*prev)->dl_next) { | |
587 | if (*prev == req) { | |
588 | *prev = req->dl_next; | |
589 | found = true; | |
590 | break; | |
591 | } | |
592 | } | |
593 | ||
594 | spin_unlock(&queue->syn_wait_lock); | |
83fccfc3 | 595 | if (timer_pending(&req->rsk_timer) && del_timer_sync(&req->rsk_timer)) |
b357a364 ED |
596 | reqsk_put(req); |
597 | return found; | |
598 | } | |
599 | ||
600 | void inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req) | |
601 | { | |
602 | if (reqsk_queue_unlink(&inet_csk(sk)->icsk_accept_queue, req)) { | |
603 | reqsk_queue_removed(&inet_csk(sk)->icsk_accept_queue, req); | |
604 | reqsk_put(req); | |
605 | } | |
606 | } | |
607 | EXPORT_SYMBOL(inet_csk_reqsk_queue_drop); | |
608 | ||
fa76ce73 | 609 | static void reqsk_timer_handler(unsigned long data) |
a019d6fe | 610 | { |
fa76ce73 ED |
611 | struct request_sock *req = (struct request_sock *)data; |
612 | struct sock *sk_listener = req->rsk_listener; | |
613 | struct inet_connection_sock *icsk = inet_csk(sk_listener); | |
a019d6fe ACM |
614 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
615 | struct listen_sock *lopt = queue->listen_opt; | |
2b41fab7 | 616 | int qlen, expire = 0, resend = 0; |
fa76ce73 | 617 | int max_retries, thresh; |
2b41fab7 | 618 | u8 defer_accept; |
a019d6fe | 619 | |
fa76ce73 ED |
620 | if (sk_listener->sk_state != TCP_LISTEN || !lopt) { |
621 | reqsk_put(req); | |
a019d6fe | 622 | return; |
fa76ce73 | 623 | } |
a019d6fe | 624 | |
fa76ce73 ED |
625 | max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries; |
626 | thresh = max_retries; | |
a019d6fe ACM |
627 | /* Normally all the openreqs are young and become mature |
628 | * (i.e. converted to established socket) for first timeout. | |
fd4f2cea | 629 | * If synack was not acknowledged for 1 second, it means |
a019d6fe ACM |
630 | * one of the following things: synack was lost, ack was lost, |
631 | * rtt is high or nobody planned to ack (i.e. synflood). | |
632 | * When server is a bit loaded, queue is populated with old | |
633 | * open requests, reducing effective size of queue. | |
634 | * When server is well loaded, queue size reduces to zero | |
635 | * after several minutes of work. It is not synflood, | |
636 | * it is normal operation. The solution is pruning | |
637 | * too old entries overriding normal timeout, when | |
638 | * situation becomes dangerous. | |
639 | * | |
640 | * Essentially, we reserve half of room for young | |
641 | * embrions; and abort old ones without pity, if old | |
642 | * ones are about to clog our table. | |
643 | */ | |
2b41fab7 ED |
644 | qlen = listen_sock_qlen(lopt); |
645 | if (qlen >> (lopt->max_qlen_log - 1)) { | |
fa76ce73 | 646 | int young = listen_sock_young(lopt) << 1; |
a019d6fe ACM |
647 | |
648 | while (thresh > 2) { | |
2b41fab7 | 649 | if (qlen < young) |
a019d6fe ACM |
650 | break; |
651 | thresh--; | |
652 | young <<= 1; | |
653 | } | |
654 | } | |
2b41fab7 ED |
655 | defer_accept = READ_ONCE(queue->rskq_defer_accept); |
656 | if (defer_accept) | |
657 | max_retries = defer_accept; | |
658 | syn_ack_recalc(req, thresh, max_retries, defer_accept, | |
fa76ce73 | 659 | &expire, &resend); |
42cb80a2 | 660 | req->rsk_ops->syn_ack_timeout(req); |
fa76ce73 ED |
661 | if (!expire && |
662 | (!resend || | |
663 | !inet_rtx_syn_ack(sk_listener, req) || | |
664 | inet_rsk(req)->acked)) { | |
665 | unsigned long timeo; | |
666 | ||
667 | if (req->num_timeout++ == 0) | |
668 | atomic_inc(&lopt->young_dec); | |
669 | timeo = min(TCP_TIMEOUT_INIT << req->num_timeout, TCP_RTO_MAX); | |
670 | mod_timer_pinned(&req->rsk_timer, jiffies + timeo); | |
671 | return; | |
672 | } | |
673 | inet_csk_reqsk_queue_drop(sk_listener, req); | |
674 | reqsk_put(req); | |
675 | } | |
ec0a1966 | 676 | |
fa76ce73 ED |
677 | void reqsk_queue_hash_req(struct request_sock_queue *queue, |
678 | u32 hash, struct request_sock *req, | |
679 | unsigned long timeout) | |
680 | { | |
681 | struct listen_sock *lopt = queue->listen_opt; | |
a019d6fe | 682 | |
fa76ce73 ED |
683 | req->num_retrans = 0; |
684 | req->num_timeout = 0; | |
685 | req->sk = NULL; | |
a019d6fe | 686 | |
29c68526 ED |
687 | setup_timer(&req->rsk_timer, reqsk_timer_handler, (unsigned long)req); |
688 | mod_timer_pinned(&req->rsk_timer, jiffies + timeout); | |
689 | req->rsk_hash = hash; | |
690 | ||
fa76ce73 ED |
691 | /* before letting lookups find us, make sure all req fields |
692 | * are committed to memory and refcnt initialized. | |
693 | */ | |
694 | smp_wmb(); | |
695 | atomic_set(&req->rsk_refcnt, 2); | |
a019d6fe | 696 | |
b2827053 | 697 | spin_lock(&queue->syn_wait_lock); |
fa76ce73 ED |
698 | req->dl_next = lopt->syn_table[hash]; |
699 | lopt->syn_table[hash] = req; | |
b2827053 | 700 | spin_unlock(&queue->syn_wait_lock); |
a019d6fe | 701 | } |
fa76ce73 | 702 | EXPORT_SYMBOL(reqsk_queue_hash_req); |
a019d6fe | 703 | |
e56c57d0 ED |
704 | /** |
705 | * inet_csk_clone_lock - clone an inet socket, and lock its clone | |
706 | * @sk: the socket to clone | |
707 | * @req: request_sock | |
708 | * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) | |
709 | * | |
710 | * Caller must unlock socket even in error path (bh_unlock_sock(newsk)) | |
711 | */ | |
712 | struct sock *inet_csk_clone_lock(const struct sock *sk, | |
713 | const struct request_sock *req, | |
714 | const gfp_t priority) | |
9f1d2604 | 715 | { |
e56c57d0 | 716 | struct sock *newsk = sk_clone_lock(sk, priority); |
9f1d2604 | 717 | |
00db4124 | 718 | if (newsk) { |
9f1d2604 ACM |
719 | struct inet_connection_sock *newicsk = inet_csk(newsk); |
720 | ||
721 | newsk->sk_state = TCP_SYN_RECV; | |
722 | newicsk->icsk_bind_hash = NULL; | |
723 | ||
634fb979 | 724 | inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port; |
b44084c2 ED |
725 | inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num; |
726 | inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num); | |
9f1d2604 ACM |
727 | newsk->sk_write_space = sk_stream_write_space; |
728 | ||
84f39b08 | 729 | newsk->sk_mark = inet_rsk(req)->ir_mark; |
33cf7c90 ED |
730 | atomic64_set(&newsk->sk_cookie, |
731 | atomic64_read(&inet_rsk(req)->ir_cookie)); | |
84f39b08 | 732 | |
9f1d2604 | 733 | newicsk->icsk_retransmits = 0; |
6687e988 ACM |
734 | newicsk->icsk_backoff = 0; |
735 | newicsk->icsk_probes_out = 0; | |
9f1d2604 ACM |
736 | |
737 | /* Deinitialize accept_queue to trap illegal accesses. */ | |
738 | memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue)); | |
4237c75c VY |
739 | |
740 | security_inet_csk_clone(newsk, req); | |
9f1d2604 ACM |
741 | } |
742 | return newsk; | |
743 | } | |
e56c57d0 | 744 | EXPORT_SYMBOL_GPL(inet_csk_clone_lock); |
a019d6fe ACM |
745 | |
746 | /* | |
747 | * At this point, there should be no process reference to this | |
748 | * socket, and thus no user references at all. Therefore we | |
749 | * can assume the socket waitqueue is inactive and nobody will | |
750 | * try to jump onto it. | |
751 | */ | |
752 | void inet_csk_destroy_sock(struct sock *sk) | |
753 | { | |
547b792c IJ |
754 | WARN_ON(sk->sk_state != TCP_CLOSE); |
755 | WARN_ON(!sock_flag(sk, SOCK_DEAD)); | |
a019d6fe ACM |
756 | |
757 | /* It cannot be in hash table! */ | |
547b792c | 758 | WARN_ON(!sk_unhashed(sk)); |
a019d6fe | 759 | |
c720c7e8 ED |
760 | /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */ |
761 | WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash); | |
a019d6fe ACM |
762 | |
763 | sk->sk_prot->destroy(sk); | |
764 | ||
765 | sk_stream_kill_queues(sk); | |
766 | ||
767 | xfrm_sk_free_policy(sk); | |
768 | ||
769 | sk_refcnt_debug_release(sk); | |
770 | ||
dd24c001 | 771 | percpu_counter_dec(sk->sk_prot->orphan_count); |
a019d6fe ACM |
772 | sock_put(sk); |
773 | } | |
a019d6fe ACM |
774 | EXPORT_SYMBOL(inet_csk_destroy_sock); |
775 | ||
e337e24d CP |
776 | /* This function allows to force a closure of a socket after the call to |
777 | * tcp/dccp_create_openreq_child(). | |
778 | */ | |
779 | void inet_csk_prepare_forced_close(struct sock *sk) | |
c10cb5fc | 780 | __releases(&sk->sk_lock.slock) |
e337e24d CP |
781 | { |
782 | /* sk_clone_lock locked the socket and set refcnt to 2 */ | |
783 | bh_unlock_sock(sk); | |
784 | sock_put(sk); | |
785 | ||
786 | /* The below has to be done to allow calling inet_csk_destroy_sock */ | |
787 | sock_set_flag(sk, SOCK_DEAD); | |
788 | percpu_counter_inc(sk->sk_prot->orphan_count); | |
789 | inet_sk(sk)->inet_num = 0; | |
790 | } | |
791 | EXPORT_SYMBOL(inet_csk_prepare_forced_close); | |
792 | ||
a019d6fe ACM |
793 | int inet_csk_listen_start(struct sock *sk, const int nr_table_entries) |
794 | { | |
795 | struct inet_sock *inet = inet_sk(sk); | |
796 | struct inet_connection_sock *icsk = inet_csk(sk); | |
797 | int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries); | |
798 | ||
799 | if (rc != 0) | |
800 | return rc; | |
801 | ||
802 | sk->sk_max_ack_backlog = 0; | |
803 | sk->sk_ack_backlog = 0; | |
804 | inet_csk_delack_init(sk); | |
805 | ||
806 | /* There is race window here: we announce ourselves listening, | |
807 | * but this transition is still not validated by get_port(). | |
808 | * It is OK, because this socket enters to hash table only | |
809 | * after validation is complete. | |
810 | */ | |
811 | sk->sk_state = TCP_LISTEN; | |
c720c7e8 ED |
812 | if (!sk->sk_prot->get_port(sk, inet->inet_num)) { |
813 | inet->inet_sport = htons(inet->inet_num); | |
a019d6fe ACM |
814 | |
815 | sk_dst_reset(sk); | |
816 | sk->sk_prot->hash(sk); | |
817 | ||
818 | return 0; | |
819 | } | |
820 | ||
821 | sk->sk_state = TCP_CLOSE; | |
822 | __reqsk_queue_destroy(&icsk->icsk_accept_queue); | |
823 | return -EADDRINUSE; | |
824 | } | |
a019d6fe ACM |
825 | EXPORT_SYMBOL_GPL(inet_csk_listen_start); |
826 | ||
827 | /* | |
828 | * This routine closes sockets which have been at least partially | |
829 | * opened, but not yet accepted. | |
830 | */ | |
831 | void inet_csk_listen_stop(struct sock *sk) | |
832 | { | |
833 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 834 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
a019d6fe ACM |
835 | struct request_sock *acc_req; |
836 | struct request_sock *req; | |
837 | ||
a019d6fe | 838 | /* make all the listen_opt local to us */ |
8336886f | 839 | acc_req = reqsk_queue_yank_acceptq(queue); |
a019d6fe ACM |
840 | |
841 | /* Following specs, it would be better either to send FIN | |
842 | * (and enter FIN-WAIT-1, it is normal close) | |
843 | * or to send active reset (abort). | |
844 | * Certainly, it is pretty dangerous while synflood, but it is | |
845 | * bad justification for our negligence 8) | |
846 | * To be honest, we are not able to make either | |
847 | * of the variants now. --ANK | |
848 | */ | |
8336886f | 849 | reqsk_queue_destroy(queue); |
a019d6fe ACM |
850 | |
851 | while ((req = acc_req) != NULL) { | |
852 | struct sock *child = req->sk; | |
853 | ||
854 | acc_req = req->dl_next; | |
855 | ||
856 | local_bh_disable(); | |
857 | bh_lock_sock(child); | |
547b792c | 858 | WARN_ON(sock_owned_by_user(child)); |
a019d6fe ACM |
859 | sock_hold(child); |
860 | ||
861 | sk->sk_prot->disconnect(child, O_NONBLOCK); | |
862 | ||
863 | sock_orphan(child); | |
864 | ||
eb4dea58 HX |
865 | percpu_counter_inc(sk->sk_prot->orphan_count); |
866 | ||
9439ce00 | 867 | if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->tfo_listener) { |
8336886f | 868 | BUG_ON(tcp_sk(child)->fastopen_rsk != req); |
9439ce00 | 869 | BUG_ON(sk != req->rsk_listener); |
8336886f JC |
870 | |
871 | /* Paranoid, to prevent race condition if | |
872 | * an inbound pkt destined for child is | |
873 | * blocked by sock lock in tcp_v4_rcv(). | |
874 | * Also to satisfy an assertion in | |
875 | * tcp_v4_destroy_sock(). | |
876 | */ | |
877 | tcp_sk(child)->fastopen_rsk = NULL; | |
8336886f | 878 | } |
a019d6fe ACM |
879 | inet_csk_destroy_sock(child); |
880 | ||
881 | bh_unlock_sock(child); | |
882 | local_bh_enable(); | |
883 | sock_put(child); | |
884 | ||
885 | sk_acceptq_removed(sk); | |
13854e5a | 886 | reqsk_put(req); |
a019d6fe | 887 | } |
0536fcc0 | 888 | if (queue->fastopenq.rskq_rst_head) { |
8336886f | 889 | /* Free all the reqs queued in rskq_rst_head. */ |
0536fcc0 ED |
890 | spin_lock_bh(&queue->fastopenq.lock); |
891 | acc_req = queue->fastopenq.rskq_rst_head; | |
892 | queue->fastopenq.rskq_rst_head = NULL; | |
893 | spin_unlock_bh(&queue->fastopenq.lock); | |
8336886f JC |
894 | while ((req = acc_req) != NULL) { |
895 | acc_req = req->dl_next; | |
13854e5a | 896 | reqsk_put(req); |
8336886f JC |
897 | } |
898 | } | |
547b792c | 899 | WARN_ON(sk->sk_ack_backlog); |
a019d6fe | 900 | } |
a019d6fe | 901 | EXPORT_SYMBOL_GPL(inet_csk_listen_stop); |
af05dc93 ACM |
902 | |
903 | void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr) | |
904 | { | |
905 | struct sockaddr_in *sin = (struct sockaddr_in *)uaddr; | |
906 | const struct inet_sock *inet = inet_sk(sk); | |
907 | ||
908 | sin->sin_family = AF_INET; | |
c720c7e8 ED |
909 | sin->sin_addr.s_addr = inet->inet_daddr; |
910 | sin->sin_port = inet->inet_dport; | |
af05dc93 | 911 | } |
af05dc93 | 912 | EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr); |
c4d93909 | 913 | |
dec73ff0 ACM |
914 | #ifdef CONFIG_COMPAT |
915 | int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname, | |
916 | char __user *optval, int __user *optlen) | |
917 | { | |
dbeff12b | 918 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 | 919 | |
00db4124 | 920 | if (icsk->icsk_af_ops->compat_getsockopt) |
dec73ff0 ACM |
921 | return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname, |
922 | optval, optlen); | |
923 | return icsk->icsk_af_ops->getsockopt(sk, level, optname, | |
924 | optval, optlen); | |
925 | } | |
dec73ff0 ACM |
926 | EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt); |
927 | ||
928 | int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 929 | char __user *optval, unsigned int optlen) |
dec73ff0 | 930 | { |
dbeff12b | 931 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 | 932 | |
00db4124 | 933 | if (icsk->icsk_af_ops->compat_setsockopt) |
dec73ff0 ACM |
934 | return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname, |
935 | optval, optlen); | |
936 | return icsk->icsk_af_ops->setsockopt(sk, level, optname, | |
937 | optval, optlen); | |
938 | } | |
dec73ff0 ACM |
939 | EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt); |
940 | #endif | |
80d0a69f DM |
941 | |
942 | static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl) | |
943 | { | |
5abf7f7e ED |
944 | const struct inet_sock *inet = inet_sk(sk); |
945 | const struct ip_options_rcu *inet_opt; | |
80d0a69f DM |
946 | __be32 daddr = inet->inet_daddr; |
947 | struct flowi4 *fl4; | |
948 | struct rtable *rt; | |
949 | ||
950 | rcu_read_lock(); | |
951 | inet_opt = rcu_dereference(inet->inet_opt); | |
952 | if (inet_opt && inet_opt->opt.srr) | |
953 | daddr = inet_opt->opt.faddr; | |
954 | fl4 = &fl->u.ip4; | |
955 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, | |
956 | inet->inet_saddr, inet->inet_dport, | |
957 | inet->inet_sport, sk->sk_protocol, | |
958 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if); | |
959 | if (IS_ERR(rt)) | |
960 | rt = NULL; | |
961 | if (rt) | |
962 | sk_setup_caps(sk, &rt->dst); | |
963 | rcu_read_unlock(); | |
964 | ||
965 | return &rt->dst; | |
966 | } | |
967 | ||
968 | struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu) | |
969 | { | |
970 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
971 | struct inet_sock *inet = inet_sk(sk); | |
972 | ||
973 | if (!dst) { | |
974 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
975 | if (!dst) | |
976 | goto out; | |
977 | } | |
6700c270 | 978 | dst->ops->update_pmtu(dst, sk, NULL, mtu); |
80d0a69f DM |
979 | |
980 | dst = __sk_dst_check(sk, 0); | |
981 | if (!dst) | |
982 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
983 | out: | |
984 | return dst; | |
985 | } | |
986 | EXPORT_SYMBOL_GPL(inet_csk_update_pmtu); |