]>
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> |
c125e80b | 27 | #include <net/sock_reuseport.h> |
3f421baa ACM |
28 | |
29 | #ifdef INET_CSK_DEBUG | |
30 | const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n"; | |
31 | EXPORT_SYMBOL(inet_csk_timer_bug_msg); | |
32 | #endif | |
33 | ||
fe38d2a1 JB |
34 | #if IS_ENABLED(CONFIG_IPV6) |
35 | /* match_wildcard == true: IPV6_ADDR_ANY equals to any IPv6 addresses if IPv6 | |
36 | * only, and any IPv4 addresses if not IPv6 only | |
37 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
38 | * IPV6_ADDR_ANY only equals to IPV6_ADDR_ANY, | |
39 | * and 0.0.0.0 equals to 0.0.0.0 only | |
40 | */ | |
41 | static int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2, | |
42 | bool match_wildcard) | |
43 | { | |
44 | const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2); | |
45 | int sk2_ipv6only = inet_v6_ipv6only(sk2); | |
46 | int addr_type = ipv6_addr_type(&sk->sk_v6_rcv_saddr); | |
47 | int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED; | |
48 | ||
49 | /* if both are mapped, treat as IPv4 */ | |
50 | if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED) { | |
51 | if (!sk2_ipv6only) { | |
52 | if (sk->sk_rcv_saddr == sk2->sk_rcv_saddr) | |
53 | return 1; | |
54 | if (!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr) | |
55 | return match_wildcard; | |
56 | } | |
57 | return 0; | |
58 | } | |
59 | ||
60 | if (addr_type == IPV6_ADDR_ANY && addr_type2 == IPV6_ADDR_ANY) | |
61 | return 1; | |
62 | ||
63 | if (addr_type2 == IPV6_ADDR_ANY && match_wildcard && | |
64 | !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED)) | |
65 | return 1; | |
66 | ||
67 | if (addr_type == IPV6_ADDR_ANY && match_wildcard && | |
68 | !(ipv6_only_sock(sk) && addr_type2 == IPV6_ADDR_MAPPED)) | |
69 | return 1; | |
70 | ||
71 | if (sk2_rcv_saddr6 && | |
72 | ipv6_addr_equal(&sk->sk_v6_rcv_saddr, sk2_rcv_saddr6)) | |
73 | return 1; | |
74 | ||
75 | return 0; | |
76 | } | |
77 | #endif | |
78 | ||
79 | /* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses | |
80 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
81 | * 0.0.0.0 only equals to 0.0.0.0 | |
82 | */ | |
83 | static int ipv4_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2, | |
84 | bool match_wildcard) | |
85 | { | |
86 | if (!ipv6_only_sock(sk2)) { | |
87 | if (sk->sk_rcv_saddr == sk2->sk_rcv_saddr) | |
88 | return 1; | |
89 | if (!sk->sk_rcv_saddr || !sk2->sk_rcv_saddr) | |
90 | return match_wildcard; | |
91 | } | |
92 | return 0; | |
93 | } | |
94 | ||
95 | int inet_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2, | |
96 | bool match_wildcard) | |
97 | { | |
98 | #if IS_ENABLED(CONFIG_IPV6) | |
99 | if (sk->sk_family == AF_INET6) | |
100 | return ipv6_rcv_saddr_equal(sk, sk2, match_wildcard); | |
101 | #endif | |
102 | return ipv4_rcv_saddr_equal(sk, sk2, match_wildcard); | |
103 | } | |
104 | EXPORT_SYMBOL(inet_rcv_saddr_equal); | |
105 | ||
0bbf87d8 | 106 | void inet_get_local_port_range(struct net *net, int *low, int *high) |
227b60f5 | 107 | { |
95c96174 ED |
108 | unsigned int seq; |
109 | ||
227b60f5 | 110 | do { |
c9d8f1a6 | 111 | seq = read_seqbegin(&net->ipv4.ip_local_ports.lock); |
227b60f5 | 112 | |
c9d8f1a6 CW |
113 | *low = net->ipv4.ip_local_ports.range[0]; |
114 | *high = net->ipv4.ip_local_ports.range[1]; | |
115 | } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq)); | |
227b60f5 SH |
116 | } |
117 | EXPORT_SYMBOL(inet_get_local_port_range); | |
3f421baa | 118 | |
971af18b | 119 | int inet_csk_bind_conflict(const struct sock *sk, |
0643ee4f TH |
120 | const struct inet_bind_bucket *tb, bool relax, |
121 | bool reuseport_ok) | |
3f421baa | 122 | { |
3f421baa | 123 | struct sock *sk2; |
0643ee4f TH |
124 | bool reuse = sk->sk_reuse; |
125 | bool reuseport = !!sk->sk_reuseport && reuseport_ok; | |
da5e3630 | 126 | kuid_t uid = sock_i_uid((struct sock *)sk); |
3f421baa | 127 | |
7477fd2e PE |
128 | /* |
129 | * Unlike other sk lookup places we do not check | |
130 | * for sk_net here, since _all_ the socks listed | |
131 | * in tb->owners list belong to the same net - the | |
132 | * one this bucket belongs to. | |
133 | */ | |
134 | ||
b67bfe0d | 135 | sk_for_each_bound(sk2, &tb->owners) { |
3f421baa ACM |
136 | if (sk != sk2 && |
137 | !inet_v6_ipv6only(sk2) && | |
138 | (!sk->sk_bound_dev_if || | |
139 | !sk2->sk_bound_dev_if || | |
140 | sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) { | |
da5e3630 TH |
141 | if ((!reuse || !sk2->sk_reuse || |
142 | sk2->sk_state == TCP_LISTEN) && | |
143 | (!reuseport || !sk2->sk_reuseport || | |
c125e80b CG |
144 | rcu_access_pointer(sk->sk_reuseport_cb) || |
145 | (sk2->sk_state != TCP_TIME_WAIT && | |
da5e3630 | 146 | !uid_eq(uid, sock_i_uid(sk2))))) { |
50805466 ED |
147 | |
148 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || | |
149 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
3f421baa | 150 | break; |
8d238b25 | 151 | } |
aacd9289 AC |
152 | if (!relax && reuse && sk2->sk_reuse && |
153 | sk2->sk_state != TCP_LISTEN) { | |
aacd9289 | 154 | |
50805466 ED |
155 | if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr || |
156 | sk2->sk_rcv_saddr == sk->sk_rcv_saddr) | |
aacd9289 AC |
157 | break; |
158 | } | |
3f421baa ACM |
159 | } |
160 | } | |
b67bfe0d | 161 | return sk2 != NULL; |
3f421baa | 162 | } |
971af18b ACM |
163 | EXPORT_SYMBOL_GPL(inet_csk_bind_conflict); |
164 | ||
3f421baa ACM |
165 | /* Obtain a reference to a local port for the given sock, |
166 | * if snum is zero it means select any available local port. | |
ea8add2b | 167 | * We try to allocate an odd port (and leave even ports for connect()) |
3f421baa | 168 | */ |
ab1e0a13 | 169 | int inet_csk_get_port(struct sock *sk, unsigned short snum) |
3f421baa | 170 | { |
ea8add2b ED |
171 | bool reuse = sk->sk_reuse && sk->sk_state != TCP_LISTEN; |
172 | struct inet_hashinfo *hinfo = sk->sk_prot->h.hashinfo; | |
173 | int ret = 1, attempts = 5, port = snum; | |
174 | int smallest_size = -1, smallest_port; | |
3f421baa | 175 | struct inet_bind_hashbucket *head; |
3b1e0a65 | 176 | struct net *net = sock_net(sk); |
ea8add2b ED |
177 | int i, low, high, attempt_half; |
178 | struct inet_bind_bucket *tb; | |
da5e3630 | 179 | kuid_t uid = sock_i_uid(sk); |
ea8add2b | 180 | u32 remaining, offset; |
0643ee4f | 181 | bool reuseport_ok = !!snum; |
3f421baa | 182 | |
ea8add2b ED |
183 | if (port) { |
184 | have_port: | |
185 | head = &hinfo->bhash[inet_bhashfn(net, port, | |
186 | hinfo->bhash_size)]; | |
187 | spin_lock_bh(&head->lock); | |
188 | inet_bind_bucket_for_each(tb, &head->chain) | |
189 | if (net_eq(ib_net(tb), net) && tb->port == port) | |
190 | goto tb_found; | |
227b60f5 | 191 | |
ea8add2b ED |
192 | goto tb_not_found; |
193 | } | |
a9d8f911 | 194 | again: |
ea8add2b ED |
195 | attempt_half = (sk->sk_reuse == SK_CAN_REUSE) ? 1 : 0; |
196 | other_half_scan: | |
197 | inet_get_local_port_range(net, &low, &high); | |
198 | high++; /* [32768, 60999] -> [32768, 61000[ */ | |
199 | if (high - low < 4) | |
200 | attempt_half = 0; | |
201 | if (attempt_half) { | |
202 | int half = low + (((high - low) >> 2) << 1); | |
203 | ||
204 | if (attempt_half == 1) | |
205 | high = half; | |
206 | else | |
207 | low = half; | |
208 | } | |
209 | remaining = high - low; | |
210 | if (likely(remaining > 1)) | |
211 | remaining &= ~1U; | |
3f421baa | 212 | |
ea8add2b ED |
213 | offset = prandom_u32() % remaining; |
214 | /* __inet_hash_connect() favors ports having @low parity | |
215 | * We do the opposite to not pollute connect() users. | |
216 | */ | |
217 | offset |= 1U; | |
218 | smallest_size = -1; | |
219 | smallest_port = low; /* avoid compiler warning */ | |
220 | ||
221 | other_parity_scan: | |
222 | port = low + offset; | |
223 | for (i = 0; i < remaining; i += 2, port += 2) { | |
224 | if (unlikely(port >= high)) | |
225 | port -= remaining; | |
226 | if (inet_is_local_reserved_port(net, port)) | |
227 | continue; | |
228 | head = &hinfo->bhash[inet_bhashfn(net, port, | |
229 | hinfo->bhash_size)]; | |
230 | spin_lock_bh(&head->lock); | |
231 | inet_bind_bucket_for_each(tb, &head->chain) | |
232 | if (net_eq(ib_net(tb), net) && tb->port == port) { | |
233 | if (((tb->fastreuse > 0 && reuse) || | |
234 | (tb->fastreuseport > 0 && | |
235 | sk->sk_reuseport && | |
236 | !rcu_access_pointer(sk->sk_reuseport_cb) && | |
237 | uid_eq(tb->fastuid, uid))) && | |
238 | (tb->num_owners < smallest_size || smallest_size == -1)) { | |
239 | smallest_size = tb->num_owners; | |
240 | smallest_port = port; | |
a9d8f911 | 241 | } |
0643ee4f TH |
242 | if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false, |
243 | reuseport_ok)) | |
ea8add2b ED |
244 | goto tb_found; |
245 | goto next_port; | |
946f9eb2 | 246 | } |
ea8add2b ED |
247 | goto tb_not_found; |
248 | next_port: | |
249 | spin_unlock_bh(&head->lock); | |
250 | cond_resched(); | |
251 | } | |
252 | ||
253 | if (smallest_size != -1) { | |
254 | port = smallest_port; | |
255 | goto have_port; | |
3f421baa | 256 | } |
ea8add2b ED |
257 | offset--; |
258 | if (!(offset & 1)) | |
259 | goto other_parity_scan; | |
260 | ||
261 | if (attempt_half == 1) { | |
262 | /* OK we now try the upper half of the range */ | |
263 | attempt_half = 2; | |
264 | goto other_half_scan; | |
265 | } | |
266 | return ret; | |
267 | ||
268 | tb_not_found: | |
269 | tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep, | |
270 | net, head, port); | |
271 | if (!tb) | |
272 | goto fail_unlock; | |
3f421baa ACM |
273 | tb_found: |
274 | if (!hlist_empty(&tb->owners)) { | |
4a17fd52 PE |
275 | if (sk->sk_reuse == SK_FORCE_REUSE) |
276 | goto success; | |
277 | ||
ea8add2b | 278 | if (((tb->fastreuse > 0 && reuse) || |
da5e3630 | 279 | (tb->fastreuseport > 0 && |
e5fbfc1c | 280 | !rcu_access_pointer(sk->sk_reuseport_cb) && |
ea8add2b ED |
281 | sk->sk_reuseport && uid_eq(tb->fastuid, uid))) && |
282 | smallest_size == -1) | |
3f421baa | 283 | goto success; |
0643ee4f TH |
284 | if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true, |
285 | reuseport_ok)) { | |
ea8add2b ED |
286 | if ((reuse || |
287 | (tb->fastreuseport > 0 && | |
288 | sk->sk_reuseport && | |
289 | !rcu_access_pointer(sk->sk_reuseport_cb) && | |
290 | uid_eq(tb->fastuid, uid))) && | |
9af7e923 | 291 | !snum && smallest_size != -1 && --attempts >= 0) { |
ea8add2b ED |
292 | spin_unlock_bh(&head->lock); |
293 | goto again; | |
a9d8f911 | 294 | } |
ea8add2b | 295 | goto fail_unlock; |
3f421baa | 296 | } |
ea8add2b | 297 | if (!reuse) |
3f421baa | 298 | tb->fastreuse = 0; |
ea8add2b ED |
299 | if (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid)) |
300 | tb->fastreuseport = 0; | |
301 | } else { | |
302 | tb->fastreuse = reuse; | |
da5e3630 TH |
303 | if (sk->sk_reuseport) { |
304 | tb->fastreuseport = 1; | |
305 | tb->fastuid = uid; | |
ea8add2b | 306 | } else { |
da5e3630 | 307 | tb->fastreuseport = 0; |
ea8add2b | 308 | } |
da5e3630 | 309 | } |
3f421baa ACM |
310 | success: |
311 | if (!inet_csk(sk)->icsk_bind_hash) | |
ea8add2b | 312 | inet_bind_hash(sk, tb, port); |
547b792c | 313 | WARN_ON(inet_csk(sk)->icsk_bind_hash != tb); |
e905a9ed | 314 | ret = 0; |
3f421baa ACM |
315 | |
316 | fail_unlock: | |
ea8add2b | 317 | spin_unlock_bh(&head->lock); |
3f421baa ACM |
318 | return ret; |
319 | } | |
3f421baa ACM |
320 | EXPORT_SYMBOL_GPL(inet_csk_get_port); |
321 | ||
322 | /* | |
323 | * Wait for an incoming connection, avoid race conditions. This must be called | |
324 | * with the socket locked. | |
325 | */ | |
326 | static int inet_csk_wait_for_connect(struct sock *sk, long timeo) | |
327 | { | |
328 | struct inet_connection_sock *icsk = inet_csk(sk); | |
329 | DEFINE_WAIT(wait); | |
330 | int err; | |
331 | ||
332 | /* | |
333 | * True wake-one mechanism for incoming connections: only | |
334 | * one process gets woken up, not the 'whole herd'. | |
335 | * Since we do not 'race & poll' for established sockets | |
336 | * anymore, the common case will execute the loop only once. | |
337 | * | |
338 | * Subtle issue: "add_wait_queue_exclusive()" will be added | |
339 | * after any current non-exclusive waiters, and we know that | |
340 | * it will always _stay_ after any new non-exclusive waiters | |
341 | * because all non-exclusive waiters are added at the | |
342 | * beginning of the wait-queue. As such, it's ok to "drop" | |
343 | * our exclusiveness temporarily when we get woken up without | |
344 | * having to remove and re-insert us on the wait queue. | |
345 | */ | |
346 | for (;;) { | |
aa395145 | 347 | prepare_to_wait_exclusive(sk_sleep(sk), &wait, |
3f421baa ACM |
348 | TASK_INTERRUPTIBLE); |
349 | release_sock(sk); | |
350 | if (reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
351 | timeo = schedule_timeout(timeo); | |
cb7cf8a3 | 352 | sched_annotate_sleep(); |
3f421baa ACM |
353 | lock_sock(sk); |
354 | err = 0; | |
355 | if (!reqsk_queue_empty(&icsk->icsk_accept_queue)) | |
356 | break; | |
357 | err = -EINVAL; | |
358 | if (sk->sk_state != TCP_LISTEN) | |
359 | break; | |
360 | err = sock_intr_errno(timeo); | |
361 | if (signal_pending(current)) | |
362 | break; | |
363 | err = -EAGAIN; | |
364 | if (!timeo) | |
365 | break; | |
366 | } | |
aa395145 | 367 | finish_wait(sk_sleep(sk), &wait); |
3f421baa ACM |
368 | return err; |
369 | } | |
370 | ||
371 | /* | |
372 | * This will accept the next outstanding connection. | |
373 | */ | |
374 | struct sock *inet_csk_accept(struct sock *sk, int flags, int *err) | |
375 | { | |
376 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 377 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
8336886f | 378 | struct request_sock *req; |
e3d95ad7 | 379 | struct sock *newsk; |
3f421baa ACM |
380 | int error; |
381 | ||
382 | lock_sock(sk); | |
383 | ||
384 | /* We need to make sure that this socket is listening, | |
385 | * and that it has something pending. | |
386 | */ | |
387 | error = -EINVAL; | |
388 | if (sk->sk_state != TCP_LISTEN) | |
389 | goto out_err; | |
390 | ||
391 | /* Find already established connection */ | |
8336886f | 392 | if (reqsk_queue_empty(queue)) { |
3f421baa ACM |
393 | long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK); |
394 | ||
395 | /* If this is a non blocking socket don't sleep */ | |
396 | error = -EAGAIN; | |
397 | if (!timeo) | |
398 | goto out_err; | |
399 | ||
400 | error = inet_csk_wait_for_connect(sk, timeo); | |
401 | if (error) | |
402 | goto out_err; | |
403 | } | |
fff1f300 | 404 | req = reqsk_queue_remove(queue, sk); |
8336886f JC |
405 | newsk = req->sk; |
406 | ||
e3d95ad7 | 407 | if (sk->sk_protocol == IPPROTO_TCP && |
0536fcc0 ED |
408 | tcp_rsk(req)->tfo_listener) { |
409 | spin_lock_bh(&queue->fastopenq.lock); | |
9439ce00 | 410 | if (tcp_rsk(req)->tfo_listener) { |
8336886f JC |
411 | /* We are still waiting for the final ACK from 3WHS |
412 | * so can't free req now. Instead, we set req->sk to | |
413 | * NULL to signify that the child socket is taken | |
414 | * so reqsk_fastopen_remove() will free the req | |
415 | * when 3WHS finishes (or is aborted). | |
416 | */ | |
417 | req->sk = NULL; | |
418 | req = NULL; | |
419 | } | |
0536fcc0 | 420 | spin_unlock_bh(&queue->fastopenq.lock); |
8336886f | 421 | } |
3f421baa ACM |
422 | out: |
423 | release_sock(sk); | |
8336886f | 424 | if (req) |
13854e5a | 425 | reqsk_put(req); |
3f421baa ACM |
426 | return newsk; |
427 | out_err: | |
428 | newsk = NULL; | |
8336886f | 429 | req = NULL; |
3f421baa ACM |
430 | *err = error; |
431 | goto out; | |
432 | } | |
3f421baa ACM |
433 | EXPORT_SYMBOL(inet_csk_accept); |
434 | ||
435 | /* | |
436 | * Using different timers for retransmit, delayed acks and probes | |
e905a9ed | 437 | * We may wish use just one timer maintaining a list of expire jiffies |
3f421baa ACM |
438 | * to optimize. |
439 | */ | |
440 | void inet_csk_init_xmit_timers(struct sock *sk, | |
441 | void (*retransmit_handler)(unsigned long), | |
442 | void (*delack_handler)(unsigned long), | |
443 | void (*keepalive_handler)(unsigned long)) | |
444 | { | |
445 | struct inet_connection_sock *icsk = inet_csk(sk); | |
446 | ||
b24b8a24 PE |
447 | setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler, |
448 | (unsigned long)sk); | |
449 | setup_timer(&icsk->icsk_delack_timer, delack_handler, | |
450 | (unsigned long)sk); | |
451 | setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk); | |
3f421baa ACM |
452 | icsk->icsk_pending = icsk->icsk_ack.pending = 0; |
453 | } | |
3f421baa ACM |
454 | EXPORT_SYMBOL(inet_csk_init_xmit_timers); |
455 | ||
456 | void inet_csk_clear_xmit_timers(struct sock *sk) | |
457 | { | |
458 | struct inet_connection_sock *icsk = inet_csk(sk); | |
459 | ||
460 | icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0; | |
461 | ||
462 | sk_stop_timer(sk, &icsk->icsk_retransmit_timer); | |
463 | sk_stop_timer(sk, &icsk->icsk_delack_timer); | |
464 | sk_stop_timer(sk, &sk->sk_timer); | |
465 | } | |
3f421baa ACM |
466 | EXPORT_SYMBOL(inet_csk_clear_xmit_timers); |
467 | ||
468 | void inet_csk_delete_keepalive_timer(struct sock *sk) | |
469 | { | |
470 | sk_stop_timer(sk, &sk->sk_timer); | |
471 | } | |
3f421baa ACM |
472 | EXPORT_SYMBOL(inet_csk_delete_keepalive_timer); |
473 | ||
474 | void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len) | |
475 | { | |
476 | sk_reset_timer(sk, &sk->sk_timer, jiffies + len); | |
477 | } | |
3f421baa ACM |
478 | EXPORT_SYMBOL(inet_csk_reset_keepalive_timer); |
479 | ||
e5895bc6 | 480 | struct dst_entry *inet_csk_route_req(const struct sock *sk, |
6bd023f3 | 481 | struct flowi4 *fl4, |
ba3f7f04 | 482 | const struct request_sock *req) |
3f421baa | 483 | { |
3f421baa | 484 | const struct inet_request_sock *ireq = inet_rsk(req); |
8b929ab1 ED |
485 | struct net *net = read_pnet(&ireq->ireq_net); |
486 | struct ip_options_rcu *opt = ireq->opt; | |
487 | struct rtable *rt; | |
3f421baa | 488 | |
8b929ab1 | 489 | flowi4_init_output(fl4, ireq->ir_iif, ireq->ir_mark, |
e79d9bc7 | 490 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
8b929ab1 | 491 | sk->sk_protocol, inet_sk_flowi_flags(sk), |
634fb979 | 492 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
8b929ab1 | 493 | ireq->ir_loc_addr, ireq->ir_rmt_port, |
e2d118a1 | 494 | htons(ireq->ir_num), sk->sk_uid); |
6bd023f3 DM |
495 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
496 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe | 497 | if (IS_ERR(rt)) |
857a6e0a | 498 | goto no_route; |
155e8336 | 499 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
857a6e0a | 500 | goto route_err; |
d8d1f30b | 501 | return &rt->dst; |
857a6e0a IJ |
502 | |
503 | route_err: | |
504 | ip_rt_put(rt); | |
505 | no_route: | |
b45386ef | 506 | __IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
857a6e0a | 507 | return NULL; |
3f421baa | 508 | } |
3f421baa ACM |
509 | EXPORT_SYMBOL_GPL(inet_csk_route_req); |
510 | ||
a2432c4f | 511 | struct dst_entry *inet_csk_route_child_sock(const struct sock *sk, |
77357a95 DM |
512 | struct sock *newsk, |
513 | const struct request_sock *req) | |
514 | { | |
515 | const struct inet_request_sock *ireq = inet_rsk(req); | |
8b929ab1 | 516 | struct net *net = read_pnet(&ireq->ireq_net); |
77357a95 | 517 | struct inet_sock *newinet = inet_sk(newsk); |
1a7b27c9 | 518 | struct ip_options_rcu *opt; |
77357a95 DM |
519 | struct flowi4 *fl4; |
520 | struct rtable *rt; | |
521 | ||
522 | fl4 = &newinet->cork.fl.u.ip4; | |
1a7b27c9 CP |
523 | |
524 | rcu_read_lock(); | |
525 | opt = rcu_dereference(newinet->inet_opt); | |
8b929ab1 | 526 | flowi4_init_output(fl4, ireq->ir_iif, ireq->ir_mark, |
77357a95 DM |
527 | RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE, |
528 | sk->sk_protocol, inet_sk_flowi_flags(sk), | |
634fb979 | 529 | (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr, |
8b929ab1 | 530 | ireq->ir_loc_addr, ireq->ir_rmt_port, |
e2d118a1 | 531 | htons(ireq->ir_num), sk->sk_uid); |
77357a95 DM |
532 | security_req_classify_flow(req, flowi4_to_flowi(fl4)); |
533 | rt = ip_route_output_flow(net, fl4, sk); | |
534 | if (IS_ERR(rt)) | |
535 | goto no_route; | |
155e8336 | 536 | if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway) |
77357a95 | 537 | goto route_err; |
1a7b27c9 | 538 | rcu_read_unlock(); |
77357a95 DM |
539 | return &rt->dst; |
540 | ||
541 | route_err: | |
542 | ip_rt_put(rt); | |
543 | no_route: | |
1a7b27c9 | 544 | rcu_read_unlock(); |
b45386ef | 545 | __IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
77357a95 DM |
546 | return NULL; |
547 | } | |
548 | EXPORT_SYMBOL_GPL(inet_csk_route_child_sock); | |
549 | ||
dfd56b8b | 550 | #if IS_ENABLED(CONFIG_IPV6) |
3f421baa ACM |
551 | #define AF_INET_FAMILY(fam) ((fam) == AF_INET) |
552 | #else | |
fa76ce73 | 553 | #define AF_INET_FAMILY(fam) true |
3f421baa ACM |
554 | #endif |
555 | ||
0c3d79bc JA |
556 | /* Decide when to expire the request and when to resend SYN-ACK */ |
557 | static inline void syn_ack_recalc(struct request_sock *req, const int thresh, | |
558 | const int max_retries, | |
559 | const u8 rskq_defer_accept, | |
560 | int *expire, int *resend) | |
561 | { | |
562 | if (!rskq_defer_accept) { | |
e6c022a4 | 563 | *expire = req->num_timeout >= thresh; |
0c3d79bc JA |
564 | *resend = 1; |
565 | return; | |
566 | } | |
e6c022a4 ED |
567 | *expire = req->num_timeout >= thresh && |
568 | (!inet_rsk(req)->acked || req->num_timeout >= max_retries); | |
0c3d79bc JA |
569 | /* |
570 | * Do not resend while waiting for data after ACK, | |
571 | * start to resend on end of deferring period to give | |
572 | * last chance for data or ACK to create established socket. | |
573 | */ | |
574 | *resend = !inet_rsk(req)->acked || | |
e6c022a4 | 575 | req->num_timeout >= rskq_defer_accept - 1; |
0c3d79bc JA |
576 | } |
577 | ||
1b70e977 | 578 | int inet_rtx_syn_ack(const struct sock *parent, struct request_sock *req) |
e6c022a4 | 579 | { |
1a2c6181 | 580 | int err = req->rsk_ops->rtx_syn_ack(parent, req); |
e6c022a4 ED |
581 | |
582 | if (!err) | |
583 | req->num_retrans++; | |
584 | return err; | |
585 | } | |
586 | EXPORT_SYMBOL(inet_rtx_syn_ack); | |
587 | ||
079096f1 | 588 | /* return true if req was found in the ehash table */ |
b357a364 ED |
589 | static bool reqsk_queue_unlink(struct request_sock_queue *queue, |
590 | struct request_sock *req) | |
591 | { | |
079096f1 | 592 | struct inet_hashinfo *hashinfo = req_to_sk(req)->sk_prot->h.hashinfo; |
5e0724d0 | 593 | bool found = false; |
b357a364 | 594 | |
5e0724d0 ED |
595 | if (sk_hashed(req_to_sk(req))) { |
596 | spinlock_t *lock = inet_ehash_lockp(hashinfo, req->rsk_hash); | |
b357a364 | 597 | |
5e0724d0 ED |
598 | spin_lock(lock); |
599 | found = __sk_nulls_del_node_init_rcu(req_to_sk(req)); | |
600 | spin_unlock(lock); | |
601 | } | |
83fccfc3 | 602 | if (timer_pending(&req->rsk_timer) && del_timer_sync(&req->rsk_timer)) |
b357a364 ED |
603 | reqsk_put(req); |
604 | return found; | |
605 | } | |
606 | ||
607 | void inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req) | |
608 | { | |
609 | if (reqsk_queue_unlink(&inet_csk(sk)->icsk_accept_queue, req)) { | |
610 | reqsk_queue_removed(&inet_csk(sk)->icsk_accept_queue, req); | |
611 | reqsk_put(req); | |
612 | } | |
613 | } | |
614 | EXPORT_SYMBOL(inet_csk_reqsk_queue_drop); | |
615 | ||
f03f2e15 ED |
616 | void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req) |
617 | { | |
618 | inet_csk_reqsk_queue_drop(sk, req); | |
619 | reqsk_put(req); | |
620 | } | |
621 | EXPORT_SYMBOL(inet_csk_reqsk_queue_drop_and_put); | |
622 | ||
fa76ce73 | 623 | static void reqsk_timer_handler(unsigned long data) |
a019d6fe | 624 | { |
fa76ce73 ED |
625 | struct request_sock *req = (struct request_sock *)data; |
626 | struct sock *sk_listener = req->rsk_listener; | |
7c083ecb | 627 | struct net *net = sock_net(sk_listener); |
fa76ce73 | 628 | struct inet_connection_sock *icsk = inet_csk(sk_listener); |
a019d6fe | 629 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
2b41fab7 | 630 | int qlen, expire = 0, resend = 0; |
fa76ce73 | 631 | int max_retries, thresh; |
2b41fab7 | 632 | u8 defer_accept; |
a019d6fe | 633 | |
00fd38d9 | 634 | if (sk_state_load(sk_listener) != TCP_LISTEN) |
079096f1 | 635 | goto drop; |
a019d6fe | 636 | |
7c083ecb | 637 | max_retries = icsk->icsk_syn_retries ? : net->ipv4.sysctl_tcp_synack_retries; |
fa76ce73 | 638 | thresh = max_retries; |
a019d6fe ACM |
639 | /* Normally all the openreqs are young and become mature |
640 | * (i.e. converted to established socket) for first timeout. | |
fd4f2cea | 641 | * If synack was not acknowledged for 1 second, it means |
a019d6fe ACM |
642 | * one of the following things: synack was lost, ack was lost, |
643 | * rtt is high or nobody planned to ack (i.e. synflood). | |
644 | * When server is a bit loaded, queue is populated with old | |
645 | * open requests, reducing effective size of queue. | |
646 | * When server is well loaded, queue size reduces to zero | |
647 | * after several minutes of work. It is not synflood, | |
648 | * it is normal operation. The solution is pruning | |
649 | * too old entries overriding normal timeout, when | |
650 | * situation becomes dangerous. | |
651 | * | |
652 | * Essentially, we reserve half of room for young | |
653 | * embrions; and abort old ones without pity, if old | |
654 | * ones are about to clog our table. | |
655 | */ | |
aac065c5 | 656 | qlen = reqsk_queue_len(queue); |
acb4a6bf | 657 | if ((qlen << 1) > max(8U, sk_listener->sk_max_ack_backlog)) { |
aac065c5 | 658 | int young = reqsk_queue_len_young(queue) << 1; |
a019d6fe ACM |
659 | |
660 | while (thresh > 2) { | |
2b41fab7 | 661 | if (qlen < young) |
a019d6fe ACM |
662 | break; |
663 | thresh--; | |
664 | young <<= 1; | |
665 | } | |
666 | } | |
2b41fab7 ED |
667 | defer_accept = READ_ONCE(queue->rskq_defer_accept); |
668 | if (defer_accept) | |
669 | max_retries = defer_accept; | |
670 | syn_ack_recalc(req, thresh, max_retries, defer_accept, | |
fa76ce73 | 671 | &expire, &resend); |
42cb80a2 | 672 | req->rsk_ops->syn_ack_timeout(req); |
fa76ce73 ED |
673 | if (!expire && |
674 | (!resend || | |
675 | !inet_rtx_syn_ack(sk_listener, req) || | |
676 | inet_rsk(req)->acked)) { | |
677 | unsigned long timeo; | |
678 | ||
679 | if (req->num_timeout++ == 0) | |
aac065c5 | 680 | atomic_dec(&queue->young); |
fa76ce73 | 681 | timeo = min(TCP_TIMEOUT_INIT << req->num_timeout, TCP_RTO_MAX); |
f3438bc7 | 682 | mod_timer(&req->rsk_timer, jiffies + timeo); |
fa76ce73 ED |
683 | return; |
684 | } | |
079096f1 | 685 | drop: |
f03f2e15 | 686 | inet_csk_reqsk_queue_drop_and_put(sk_listener, req); |
fa76ce73 | 687 | } |
ec0a1966 | 688 | |
079096f1 ED |
689 | static void reqsk_queue_hash_req(struct request_sock *req, |
690 | unsigned long timeout) | |
fa76ce73 | 691 | { |
fa76ce73 ED |
692 | req->num_retrans = 0; |
693 | req->num_timeout = 0; | |
694 | req->sk = NULL; | |
a019d6fe | 695 | |
f3438bc7 TG |
696 | setup_pinned_timer(&req->rsk_timer, reqsk_timer_handler, |
697 | (unsigned long)req); | |
698 | mod_timer(&req->rsk_timer, jiffies + timeout); | |
29c68526 | 699 | |
079096f1 | 700 | inet_ehash_insert(req_to_sk(req), NULL); |
fa76ce73 ED |
701 | /* before letting lookups find us, make sure all req fields |
702 | * are committed to memory and refcnt initialized. | |
703 | */ | |
704 | smp_wmb(); | |
ca6fb065 | 705 | atomic_set(&req->rsk_refcnt, 2 + 1); |
079096f1 | 706 | } |
a019d6fe | 707 | |
079096f1 ED |
708 | void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req, |
709 | unsigned long timeout) | |
710 | { | |
711 | reqsk_queue_hash_req(req, timeout); | |
712 | inet_csk_reqsk_queue_added(sk); | |
a019d6fe | 713 | } |
079096f1 | 714 | EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add); |
a019d6fe | 715 | |
e56c57d0 ED |
716 | /** |
717 | * inet_csk_clone_lock - clone an inet socket, and lock its clone | |
718 | * @sk: the socket to clone | |
719 | * @req: request_sock | |
720 | * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc) | |
721 | * | |
722 | * Caller must unlock socket even in error path (bh_unlock_sock(newsk)) | |
723 | */ | |
724 | struct sock *inet_csk_clone_lock(const struct sock *sk, | |
725 | const struct request_sock *req, | |
726 | const gfp_t priority) | |
9f1d2604 | 727 | { |
e56c57d0 | 728 | struct sock *newsk = sk_clone_lock(sk, priority); |
9f1d2604 | 729 | |
00db4124 | 730 | if (newsk) { |
9f1d2604 ACM |
731 | struct inet_connection_sock *newicsk = inet_csk(newsk); |
732 | ||
733 | newsk->sk_state = TCP_SYN_RECV; | |
734 | newicsk->icsk_bind_hash = NULL; | |
735 | ||
634fb979 | 736 | inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port; |
b44084c2 ED |
737 | inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num; |
738 | inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num); | |
9f1d2604 ACM |
739 | newsk->sk_write_space = sk_stream_write_space; |
740 | ||
85017869 ED |
741 | /* listeners have SOCK_RCU_FREE, not the children */ |
742 | sock_reset_flag(newsk, SOCK_RCU_FREE); | |
743 | ||
84f39b08 | 744 | newsk->sk_mark = inet_rsk(req)->ir_mark; |
33cf7c90 ED |
745 | atomic64_set(&newsk->sk_cookie, |
746 | atomic64_read(&inet_rsk(req)->ir_cookie)); | |
84f39b08 | 747 | |
9f1d2604 | 748 | newicsk->icsk_retransmits = 0; |
6687e988 ACM |
749 | newicsk->icsk_backoff = 0; |
750 | newicsk->icsk_probes_out = 0; | |
9f1d2604 ACM |
751 | |
752 | /* Deinitialize accept_queue to trap illegal accesses. */ | |
753 | memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue)); | |
4237c75c VY |
754 | |
755 | security_inet_csk_clone(newsk, req); | |
9f1d2604 ACM |
756 | } |
757 | return newsk; | |
758 | } | |
e56c57d0 | 759 | EXPORT_SYMBOL_GPL(inet_csk_clone_lock); |
a019d6fe ACM |
760 | |
761 | /* | |
762 | * At this point, there should be no process reference to this | |
763 | * socket, and thus no user references at all. Therefore we | |
764 | * can assume the socket waitqueue is inactive and nobody will | |
765 | * try to jump onto it. | |
766 | */ | |
767 | void inet_csk_destroy_sock(struct sock *sk) | |
768 | { | |
547b792c IJ |
769 | WARN_ON(sk->sk_state != TCP_CLOSE); |
770 | WARN_ON(!sock_flag(sk, SOCK_DEAD)); | |
a019d6fe ACM |
771 | |
772 | /* It cannot be in hash table! */ | |
547b792c | 773 | WARN_ON(!sk_unhashed(sk)); |
a019d6fe | 774 | |
c720c7e8 ED |
775 | /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */ |
776 | WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash); | |
a019d6fe ACM |
777 | |
778 | sk->sk_prot->destroy(sk); | |
779 | ||
780 | sk_stream_kill_queues(sk); | |
781 | ||
782 | xfrm_sk_free_policy(sk); | |
783 | ||
784 | sk_refcnt_debug_release(sk); | |
785 | ||
777c6ae5 | 786 | local_bh_disable(); |
dd24c001 | 787 | percpu_counter_dec(sk->sk_prot->orphan_count); |
777c6ae5 | 788 | local_bh_enable(); |
a019d6fe ACM |
789 | sock_put(sk); |
790 | } | |
a019d6fe ACM |
791 | EXPORT_SYMBOL(inet_csk_destroy_sock); |
792 | ||
e337e24d CP |
793 | /* This function allows to force a closure of a socket after the call to |
794 | * tcp/dccp_create_openreq_child(). | |
795 | */ | |
796 | void inet_csk_prepare_forced_close(struct sock *sk) | |
c10cb5fc | 797 | __releases(&sk->sk_lock.slock) |
e337e24d CP |
798 | { |
799 | /* sk_clone_lock locked the socket and set refcnt to 2 */ | |
800 | bh_unlock_sock(sk); | |
801 | sock_put(sk); | |
802 | ||
803 | /* The below has to be done to allow calling inet_csk_destroy_sock */ | |
804 | sock_set_flag(sk, SOCK_DEAD); | |
805 | percpu_counter_inc(sk->sk_prot->orphan_count); | |
806 | inet_sk(sk)->inet_num = 0; | |
807 | } | |
808 | EXPORT_SYMBOL(inet_csk_prepare_forced_close); | |
809 | ||
f985c65c | 810 | int inet_csk_listen_start(struct sock *sk, int backlog) |
a019d6fe | 811 | { |
a019d6fe | 812 | struct inet_connection_sock *icsk = inet_csk(sk); |
10cbc8f1 | 813 | struct inet_sock *inet = inet_sk(sk); |
086c653f | 814 | int err = -EADDRINUSE; |
a019d6fe | 815 | |
ef547f2a | 816 | reqsk_queue_alloc(&icsk->icsk_accept_queue); |
a019d6fe | 817 | |
f985c65c | 818 | sk->sk_max_ack_backlog = backlog; |
a019d6fe ACM |
819 | sk->sk_ack_backlog = 0; |
820 | inet_csk_delack_init(sk); | |
821 | ||
822 | /* There is race window here: we announce ourselves listening, | |
823 | * but this transition is still not validated by get_port(). | |
824 | * It is OK, because this socket enters to hash table only | |
825 | * after validation is complete. | |
826 | */ | |
00fd38d9 | 827 | sk_state_store(sk, TCP_LISTEN); |
c720c7e8 ED |
828 | if (!sk->sk_prot->get_port(sk, inet->inet_num)) { |
829 | inet->inet_sport = htons(inet->inet_num); | |
a019d6fe ACM |
830 | |
831 | sk_dst_reset(sk); | |
086c653f | 832 | err = sk->sk_prot->hash(sk); |
a019d6fe | 833 | |
086c653f CG |
834 | if (likely(!err)) |
835 | return 0; | |
a019d6fe ACM |
836 | } |
837 | ||
838 | sk->sk_state = TCP_CLOSE; | |
086c653f | 839 | return err; |
a019d6fe | 840 | } |
a019d6fe ACM |
841 | EXPORT_SYMBOL_GPL(inet_csk_listen_start); |
842 | ||
ebb516af ED |
843 | static void inet_child_forget(struct sock *sk, struct request_sock *req, |
844 | struct sock *child) | |
845 | { | |
846 | sk->sk_prot->disconnect(child, O_NONBLOCK); | |
847 | ||
848 | sock_orphan(child); | |
849 | ||
850 | percpu_counter_inc(sk->sk_prot->orphan_count); | |
851 | ||
852 | if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->tfo_listener) { | |
853 | BUG_ON(tcp_sk(child)->fastopen_rsk != req); | |
854 | BUG_ON(sk != req->rsk_listener); | |
855 | ||
856 | /* Paranoid, to prevent race condition if | |
857 | * an inbound pkt destined for child is | |
858 | * blocked by sock lock in tcp_v4_rcv(). | |
859 | * Also to satisfy an assertion in | |
860 | * tcp_v4_destroy_sock(). | |
861 | */ | |
862 | tcp_sk(child)->fastopen_rsk = NULL; | |
863 | } | |
864 | inet_csk_destroy_sock(child); | |
865 | reqsk_put(req); | |
866 | } | |
867 | ||
7716682c ED |
868 | struct sock *inet_csk_reqsk_queue_add(struct sock *sk, |
869 | struct request_sock *req, | |
870 | struct sock *child) | |
ebb516af ED |
871 | { |
872 | struct request_sock_queue *queue = &inet_csk(sk)->icsk_accept_queue; | |
873 | ||
874 | spin_lock(&queue->rskq_lock); | |
875 | if (unlikely(sk->sk_state != TCP_LISTEN)) { | |
876 | inet_child_forget(sk, req, child); | |
7716682c | 877 | child = NULL; |
ebb516af ED |
878 | } else { |
879 | req->sk = child; | |
880 | req->dl_next = NULL; | |
881 | if (queue->rskq_accept_head == NULL) | |
882 | queue->rskq_accept_head = req; | |
883 | else | |
884 | queue->rskq_accept_tail->dl_next = req; | |
885 | queue->rskq_accept_tail = req; | |
886 | sk_acceptq_added(sk); | |
887 | } | |
888 | spin_unlock(&queue->rskq_lock); | |
7716682c | 889 | return child; |
ebb516af ED |
890 | } |
891 | EXPORT_SYMBOL(inet_csk_reqsk_queue_add); | |
892 | ||
5e0724d0 ED |
893 | struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child, |
894 | struct request_sock *req, bool own_req) | |
895 | { | |
896 | if (own_req) { | |
897 | inet_csk_reqsk_queue_drop(sk, req); | |
898 | reqsk_queue_removed(&inet_csk(sk)->icsk_accept_queue, req); | |
7716682c ED |
899 | if (inet_csk_reqsk_queue_add(sk, req, child)) |
900 | return child; | |
5e0724d0 ED |
901 | } |
902 | /* Too bad, another child took ownership of the request, undo. */ | |
903 | bh_unlock_sock(child); | |
904 | sock_put(child); | |
905 | return NULL; | |
906 | } | |
907 | EXPORT_SYMBOL(inet_csk_complete_hashdance); | |
908 | ||
a019d6fe ACM |
909 | /* |
910 | * This routine closes sockets which have been at least partially | |
911 | * opened, but not yet accepted. | |
912 | */ | |
913 | void inet_csk_listen_stop(struct sock *sk) | |
914 | { | |
915 | struct inet_connection_sock *icsk = inet_csk(sk); | |
8336886f | 916 | struct request_sock_queue *queue = &icsk->icsk_accept_queue; |
fff1f300 | 917 | struct request_sock *next, *req; |
a019d6fe ACM |
918 | |
919 | /* Following specs, it would be better either to send FIN | |
920 | * (and enter FIN-WAIT-1, it is normal close) | |
921 | * or to send active reset (abort). | |
922 | * Certainly, it is pretty dangerous while synflood, but it is | |
923 | * bad justification for our negligence 8) | |
924 | * To be honest, we are not able to make either | |
925 | * of the variants now. --ANK | |
926 | */ | |
fff1f300 | 927 | while ((req = reqsk_queue_remove(queue, sk)) != NULL) { |
a019d6fe ACM |
928 | struct sock *child = req->sk; |
929 | ||
a019d6fe ACM |
930 | local_bh_disable(); |
931 | bh_lock_sock(child); | |
547b792c | 932 | WARN_ON(sock_owned_by_user(child)); |
a019d6fe ACM |
933 | sock_hold(child); |
934 | ||
ebb516af | 935 | inet_child_forget(sk, req, child); |
a019d6fe ACM |
936 | bh_unlock_sock(child); |
937 | local_bh_enable(); | |
938 | sock_put(child); | |
939 | ||
92d6f176 | 940 | cond_resched(); |
a019d6fe | 941 | } |
0536fcc0 | 942 | if (queue->fastopenq.rskq_rst_head) { |
8336886f | 943 | /* Free all the reqs queued in rskq_rst_head. */ |
0536fcc0 | 944 | spin_lock_bh(&queue->fastopenq.lock); |
fff1f300 | 945 | req = queue->fastopenq.rskq_rst_head; |
0536fcc0 ED |
946 | queue->fastopenq.rskq_rst_head = NULL; |
947 | spin_unlock_bh(&queue->fastopenq.lock); | |
fff1f300 ED |
948 | while (req != NULL) { |
949 | next = req->dl_next; | |
13854e5a | 950 | reqsk_put(req); |
fff1f300 | 951 | req = next; |
8336886f JC |
952 | } |
953 | } | |
ebb516af | 954 | WARN_ON_ONCE(sk->sk_ack_backlog); |
a019d6fe | 955 | } |
a019d6fe | 956 | EXPORT_SYMBOL_GPL(inet_csk_listen_stop); |
af05dc93 ACM |
957 | |
958 | void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr) | |
959 | { | |
960 | struct sockaddr_in *sin = (struct sockaddr_in *)uaddr; | |
961 | const struct inet_sock *inet = inet_sk(sk); | |
962 | ||
963 | sin->sin_family = AF_INET; | |
c720c7e8 ED |
964 | sin->sin_addr.s_addr = inet->inet_daddr; |
965 | sin->sin_port = inet->inet_dport; | |
af05dc93 | 966 | } |
af05dc93 | 967 | EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr); |
c4d93909 | 968 | |
dec73ff0 ACM |
969 | #ifdef CONFIG_COMPAT |
970 | int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname, | |
971 | char __user *optval, int __user *optlen) | |
972 | { | |
dbeff12b | 973 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 | 974 | |
00db4124 | 975 | if (icsk->icsk_af_ops->compat_getsockopt) |
dec73ff0 ACM |
976 | return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname, |
977 | optval, optlen); | |
978 | return icsk->icsk_af_ops->getsockopt(sk, level, optname, | |
979 | optval, optlen); | |
980 | } | |
dec73ff0 ACM |
981 | EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt); |
982 | ||
983 | int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 984 | char __user *optval, unsigned int optlen) |
dec73ff0 | 985 | { |
dbeff12b | 986 | const struct inet_connection_sock *icsk = inet_csk(sk); |
dec73ff0 | 987 | |
00db4124 | 988 | if (icsk->icsk_af_ops->compat_setsockopt) |
dec73ff0 ACM |
989 | return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname, |
990 | optval, optlen); | |
991 | return icsk->icsk_af_ops->setsockopt(sk, level, optname, | |
992 | optval, optlen); | |
993 | } | |
dec73ff0 ACM |
994 | EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt); |
995 | #endif | |
80d0a69f DM |
996 | |
997 | static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl) | |
998 | { | |
5abf7f7e ED |
999 | const struct inet_sock *inet = inet_sk(sk); |
1000 | const struct ip_options_rcu *inet_opt; | |
80d0a69f DM |
1001 | __be32 daddr = inet->inet_daddr; |
1002 | struct flowi4 *fl4; | |
1003 | struct rtable *rt; | |
1004 | ||
1005 | rcu_read_lock(); | |
1006 | inet_opt = rcu_dereference(inet->inet_opt); | |
1007 | if (inet_opt && inet_opt->opt.srr) | |
1008 | daddr = inet_opt->opt.faddr; | |
1009 | fl4 = &fl->u.ip4; | |
1010 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, | |
1011 | inet->inet_saddr, inet->inet_dport, | |
1012 | inet->inet_sport, sk->sk_protocol, | |
1013 | RT_CONN_FLAGS(sk), sk->sk_bound_dev_if); | |
1014 | if (IS_ERR(rt)) | |
1015 | rt = NULL; | |
1016 | if (rt) | |
1017 | sk_setup_caps(sk, &rt->dst); | |
1018 | rcu_read_unlock(); | |
1019 | ||
1020 | return &rt->dst; | |
1021 | } | |
1022 | ||
1023 | struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu) | |
1024 | { | |
1025 | struct dst_entry *dst = __sk_dst_check(sk, 0); | |
1026 | struct inet_sock *inet = inet_sk(sk); | |
1027 | ||
1028 | if (!dst) { | |
1029 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
1030 | if (!dst) | |
1031 | goto out; | |
1032 | } | |
6700c270 | 1033 | dst->ops->update_pmtu(dst, sk, NULL, mtu); |
80d0a69f DM |
1034 | |
1035 | dst = __sk_dst_check(sk, 0); | |
1036 | if (!dst) | |
1037 | dst = inet_csk_rebuild_route(sk, &inet->cork.fl); | |
1038 | out: | |
1039 | return dst; | |
1040 | } | |
1041 | EXPORT_SYMBOL_GPL(inet_csk_update_pmtu); |