blob: fc0e649cc002beb0631c390e6aed5f1ea3fe7390 [file] [log] [blame]
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -07001/*
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
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070016#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>
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -070025#include <net/xfrm.h>
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070026
27#ifdef INET_CSK_DEBUG
28const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n";
29EXPORT_SYMBOL(inet_csk_timer_bug_msg);
30#endif
31
Amerigo Wange3826f12010-05-05 00:27:06 +000032unsigned long *sysctl_local_reserved_ports;
33EXPORT_SYMBOL(sysctl_local_reserved_ports);
34
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -070035void inet_get_local_port_range(struct net *net, int *low, int *high)
Stephen Hemminger227b60f2007-10-10 17:30:46 -070036{
Eric Dumazet95c96172012-04-15 05:58:06 +000037 unsigned int seq;
38
Stephen Hemminger227b60f2007-10-10 17:30:46 -070039 do {
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -070040 seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
Stephen Hemminger227b60f2007-10-10 17:30:46 -070041
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -070042 *low = net->ipv4.sysctl_local_ports.range[0];
43 *high = net->ipv4.sysctl_local_ports.range[1];
44 } while (read_seqretry(&net->ipv4.sysctl_local_ports.lock, seq));
Stephen Hemminger227b60f2007-10-10 17:30:46 -070045}
46EXPORT_SYMBOL(inet_get_local_port_range);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070047
Arnaldo Carvalho de Melo971af182005-12-13 23:14:47 -080048int inet_csk_bind_conflict(const struct sock *sk,
Alex Copotaacd9282012-04-12 22:21:45 +000049 const struct inet_bind_bucket *tb, bool relax)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070050{
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070051 struct sock *sk2;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070052 int reuse = sk->sk_reuse;
Tom Herbertda5e3632013-01-22 09:50:24 +000053 int reuseport = sk->sk_reuseport;
54 kuid_t uid = sock_i_uid((struct sock *)sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070055
Pavel Emelyanov7477fd2e2008-04-14 02:42:27 -070056 /*
57 * Unlike other sk lookup places we do not check
58 * for sk_net here, since _all_ the socks listed
59 * in tb->owners list belong to the same net - the
60 * one this bucket belongs to.
61 */
62
Sasha Levinb67bfe02013-02-27 17:06:00 -080063 sk_for_each_bound(sk2, &tb->owners) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070064 if (sk != sk2 &&
65 !inet_v6_ipv6only(sk2) &&
66 (!sk->sk_bound_dev_if ||
67 !sk2->sk_bound_dev_if ||
68 sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) {
Tom Herbertda5e3632013-01-22 09:50:24 +000069 if ((!reuse || !sk2->sk_reuse ||
70 sk2->sk_state == TCP_LISTEN) &&
71 (!reuseport || !sk2->sk_reuseport ||
72 (sk2->sk_state != TCP_TIME_WAIT &&
73 !uid_eq(uid, sock_i_uid(sk2))))) {
Eric Dumazet50805462013-10-02 04:29:50 -070074
75 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
76 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070077 break;
David S. Miller8d238b22010-04-28 11:25:59 -070078 }
Alex Copotaacd9282012-04-12 22:21:45 +000079 if (!relax && reuse && sk2->sk_reuse &&
80 sk2->sk_state != TCP_LISTEN) {
Alex Copotaacd9282012-04-12 22:21:45 +000081
Eric Dumazet50805462013-10-02 04:29:50 -070082 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
83 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
Alex Copotaacd9282012-04-12 22:21:45 +000084 break;
85 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070086 }
87 }
Sasha Levinb67bfe02013-02-27 17:06:00 -080088 return sk2 != NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070089}
Arnaldo Carvalho de Melo971af182005-12-13 23:14:47 -080090EXPORT_SYMBOL_GPL(inet_csk_bind_conflict);
91
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070092/* Obtain a reference to a local port for the given sock,
93 * if snum is zero it means select any available local port.
94 */
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -080095int inet_csk_get_port(struct sock *sk, unsigned short snum)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070096{
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -070097 struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070098 struct inet_bind_hashbucket *head;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070099 struct inet_bind_bucket *tb;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800100 int ret, attempts = 5;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900101 struct net *net = sock_net(sk);
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800102 int smallest_size = -1, smallest_rover;
Tom Herbertda5e3632013-01-22 09:50:24 +0000103 kuid_t uid = sock_i_uid(sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700104
105 local_bh_disable();
106 if (!snum) {
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700107 int remaining, rover, low, high;
108
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800109again:
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700110 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700111 remaining = (high - low) + 1;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800112 smallest_rover = rover = net_random() % remaining + low;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700113
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800114 smallest_size = -1;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700115 do {
Amerigo Wange3826f12010-05-05 00:27:06 +0000116 if (inet_is_reserved_local_port(rover))
117 goto next_nolock;
Pavel Emelyanov7f635ab2008-06-16 17:12:49 -0700118 head = &hashinfo->bhash[inet_bhashfn(net, rover,
119 hashinfo->bhash_size)];
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700120 spin_lock(&head->lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800121 inet_bind_bucket_for_each(tb, &head->chain)
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800122 if (net_eq(ib_net(tb), net) && tb->port == rover) {
Tom Herbertda5e3632013-01-22 09:50:24 +0000123 if (((tb->fastreuse > 0 &&
124 sk->sk_reuse &&
125 sk->sk_state != TCP_LISTEN) ||
126 (tb->fastreuseport > 0 &&
127 sk->sk_reuseport &&
128 uid_eq(tb->fastuid, uid))) &&
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800129 (tb->num_owners < smallest_size || smallest_size == -1)) {
130 smallest_size = tb->num_owners;
131 smallest_rover = rover;
Alex Copotaacd9282012-04-12 22:21:45 +0000132 if (atomic_read(&hashinfo->bsockets) > (high - low) + 1 &&
133 !inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
David S. Miller8d238b22010-04-28 11:25:59 -0700134 snum = smallest_rover;
Flavio Leitnerfddb7b52012-01-25 08:34:52 +0000135 goto tb_found;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800136 }
137 }
Alex Copotaacd9282012-04-12 22:21:45 +0000138 if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
Flavio Leitner2b05ad32012-01-25 08:34:51 +0000139 snum = rover;
Flavio Leitnerfddb7b52012-01-25 08:34:52 +0000140 goto tb_found;
Flavio Leitner2b05ad32012-01-25 08:34:51 +0000141 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700142 goto next;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800143 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700144 break;
145 next:
146 spin_unlock(&head->lock);
Amerigo Wange3826f12010-05-05 00:27:06 +0000147 next_nolock:
Stephen Hemminger6df71632005-11-03 16:33:23 -0800148 if (++rover > high)
149 rover = low;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700150 } while (--remaining > 0);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700151
152 /* Exhausted local port range during search? It is not
153 * possible for us to be holding one of the bind hash
154 * locks if this test triggers, because if 'remaining'
155 * drops to zero, we broke out of the do/while loop at
156 * the top level, not from the 'break;' statement.
157 */
158 ret = 1;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800159 if (remaining <= 0) {
160 if (smallest_size != -1) {
161 snum = smallest_rover;
162 goto have_snum;
163 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700164 goto fail;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800165 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700166 /* OK, here is the one we will use. HEAD is
167 * non-NULL and we hold it's mutex.
168 */
169 snum = rover;
170 } else {
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800171have_snum:
Pavel Emelyanov7f635ab2008-06-16 17:12:49 -0700172 head = &hashinfo->bhash[inet_bhashfn(net, snum,
173 hashinfo->bhash_size)];
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700174 spin_lock(&head->lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800175 inet_bind_bucket_for_each(tb, &head->chain)
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800176 if (net_eq(ib_net(tb), net) && tb->port == snum)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700177 goto tb_found;
178 }
179 tb = NULL;
180 goto tb_not_found;
181tb_found:
182 if (!hlist_empty(&tb->owners)) {
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000183 if (sk->sk_reuse == SK_FORCE_REUSE)
184 goto success;
185
Tom Herbertda5e3632013-01-22 09:50:24 +0000186 if (((tb->fastreuse > 0 &&
187 sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
188 (tb->fastreuseport > 0 &&
189 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800190 smallest_size == -1) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700191 goto success;
192 } else {
193 ret = 1;
Alex Copotaacd9282012-04-12 22:21:45 +0000194 if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true)) {
Tom Herbertda5e3632013-01-22 09:50:24 +0000195 if (((sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000196 (tb->fastreuseport > 0 &&
197 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
Stephen Hemminger5add3002009-02-01 01:40:17 -0800198 smallest_size != -1 && --attempts >= 0) {
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800199 spin_unlock(&head->lock);
200 goto again;
201 }
Alex Copotaacd9282012-04-12 22:21:45 +0000202
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700203 goto fail_unlock;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800204 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700205 }
206 }
207tb_not_found:
208 ret = 1;
Pavel Emelyanov941b1d22008-01-31 05:05:50 -0800209 if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep,
210 net, head, snum)) == NULL)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700211 goto fail_unlock;
212 if (hlist_empty(&tb->owners)) {
213 if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
214 tb->fastreuse = 1;
215 else
216 tb->fastreuse = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000217 if (sk->sk_reuseport) {
218 tb->fastreuseport = 1;
219 tb->fastuid = uid;
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000220 } else
Tom Herbertda5e3632013-01-22 09:50:24 +0000221 tb->fastreuseport = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000222 } else {
223 if (tb->fastreuse &&
224 (!sk->sk_reuse || sk->sk_state == TCP_LISTEN))
225 tb->fastreuse = 0;
226 if (tb->fastreuseport &&
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000227 (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid)))
Tom Herbertda5e3632013-01-22 09:50:24 +0000228 tb->fastreuseport = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000229 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700230success:
231 if (!inet_csk(sk)->icsk_bind_hash)
232 inet_bind_hash(sk, tb, snum);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700233 WARN_ON(inet_csk(sk)->icsk_bind_hash != tb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900234 ret = 0;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700235
236fail_unlock:
237 spin_unlock(&head->lock);
238fail:
239 local_bh_enable();
240 return ret;
241}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700242EXPORT_SYMBOL_GPL(inet_csk_get_port);
243
244/*
245 * Wait for an incoming connection, avoid race conditions. This must be called
246 * with the socket locked.
247 */
248static int inet_csk_wait_for_connect(struct sock *sk, long timeo)
249{
250 struct inet_connection_sock *icsk = inet_csk(sk);
251 DEFINE_WAIT(wait);
252 int err;
253
254 /*
255 * True wake-one mechanism for incoming connections: only
256 * one process gets woken up, not the 'whole herd'.
257 * Since we do not 'race & poll' for established sockets
258 * anymore, the common case will execute the loop only once.
259 *
260 * Subtle issue: "add_wait_queue_exclusive()" will be added
261 * after any current non-exclusive waiters, and we know that
262 * it will always _stay_ after any new non-exclusive waiters
263 * because all non-exclusive waiters are added at the
264 * beginning of the wait-queue. As such, it's ok to "drop"
265 * our exclusiveness temporarily when we get woken up without
266 * having to remove and re-insert us on the wait queue.
267 */
268 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +0000269 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700270 TASK_INTERRUPTIBLE);
271 release_sock(sk);
272 if (reqsk_queue_empty(&icsk->icsk_accept_queue))
273 timeo = schedule_timeout(timeo);
274 lock_sock(sk);
275 err = 0;
276 if (!reqsk_queue_empty(&icsk->icsk_accept_queue))
277 break;
278 err = -EINVAL;
279 if (sk->sk_state != TCP_LISTEN)
280 break;
281 err = sock_intr_errno(timeo);
282 if (signal_pending(current))
283 break;
284 err = -EAGAIN;
285 if (!timeo)
286 break;
287 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000288 finish_wait(sk_sleep(sk), &wait);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700289 return err;
290}
291
292/*
293 * This will accept the next outstanding connection.
294 */
295struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
296{
297 struct inet_connection_sock *icsk = inet_csk(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000298 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700299 struct sock *newsk;
Jerry Chu83368862012-08-31 12:29:12 +0000300 struct request_sock *req;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700301 int error;
302
303 lock_sock(sk);
304
305 /* We need to make sure that this socket is listening,
306 * and that it has something pending.
307 */
308 error = -EINVAL;
309 if (sk->sk_state != TCP_LISTEN)
310 goto out_err;
311
312 /* Find already established connection */
Jerry Chu83368862012-08-31 12:29:12 +0000313 if (reqsk_queue_empty(queue)) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700314 long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
315
316 /* If this is a non blocking socket don't sleep */
317 error = -EAGAIN;
318 if (!timeo)
319 goto out_err;
320
321 error = inet_csk_wait_for_connect(sk, timeo);
322 if (error)
323 goto out_err;
324 }
Jerry Chu83368862012-08-31 12:29:12 +0000325 req = reqsk_queue_remove(queue);
326 newsk = req->sk;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700327
Jerry Chu83368862012-08-31 12:29:12 +0000328 sk_acceptq_removed(sk);
Eric Dumazet7ab45512012-09-06 08:07:13 +0000329 if (sk->sk_protocol == IPPROTO_TCP && queue->fastopenq != NULL) {
Jerry Chu83368862012-08-31 12:29:12 +0000330 spin_lock_bh(&queue->fastopenq->lock);
331 if (tcp_rsk(req)->listener) {
332 /* We are still waiting for the final ACK from 3WHS
333 * so can't free req now. Instead, we set req->sk to
334 * NULL to signify that the child socket is taken
335 * so reqsk_fastopen_remove() will free the req
336 * when 3WHS finishes (or is aborted).
337 */
338 req->sk = NULL;
339 req = NULL;
340 }
341 spin_unlock_bh(&queue->fastopenq->lock);
342 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700343out:
344 release_sock(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000345 if (req)
346 __reqsk_free(req);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700347 return newsk;
348out_err:
349 newsk = NULL;
Jerry Chu83368862012-08-31 12:29:12 +0000350 req = NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700351 *err = error;
352 goto out;
353}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700354EXPORT_SYMBOL(inet_csk_accept);
355
356/*
357 * Using different timers for retransmit, delayed acks and probes
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900358 * We may wish use just one timer maintaining a list of expire jiffies
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700359 * to optimize.
360 */
361void inet_csk_init_xmit_timers(struct sock *sk,
362 void (*retransmit_handler)(unsigned long),
363 void (*delack_handler)(unsigned long),
364 void (*keepalive_handler)(unsigned long))
365{
366 struct inet_connection_sock *icsk = inet_csk(sk);
367
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -0800368 setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler,
369 (unsigned long)sk);
370 setup_timer(&icsk->icsk_delack_timer, delack_handler,
371 (unsigned long)sk);
372 setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700373 icsk->icsk_pending = icsk->icsk_ack.pending = 0;
374}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700375EXPORT_SYMBOL(inet_csk_init_xmit_timers);
376
377void inet_csk_clear_xmit_timers(struct sock *sk)
378{
379 struct inet_connection_sock *icsk = inet_csk(sk);
380
381 icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0;
382
383 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
384 sk_stop_timer(sk, &icsk->icsk_delack_timer);
385 sk_stop_timer(sk, &sk->sk_timer);
386}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700387EXPORT_SYMBOL(inet_csk_clear_xmit_timers);
388
389void inet_csk_delete_keepalive_timer(struct sock *sk)
390{
391 sk_stop_timer(sk, &sk->sk_timer);
392}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700393EXPORT_SYMBOL(inet_csk_delete_keepalive_timer);
394
395void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len)
396{
397 sk_reset_timer(sk, &sk->sk_timer, jiffies + len);
398}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700399EXPORT_SYMBOL(inet_csk_reset_keepalive_timer);
400
Jianjun Kongd93191002008-11-03 00:23:42 -0800401struct dst_entry *inet_csk_route_req(struct sock *sk,
David S. Miller6bd023f2011-05-18 18:32:03 -0400402 struct flowi4 *fl4,
David S. Millerba3f7f02012-07-17 14:02:46 -0700403 const struct request_sock *req)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700404{
405 struct rtable *rt;
406 const struct inet_request_sock *ireq = inet_rsk(req);
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000407 struct ip_options_rcu *opt = inet_rsk(req)->opt;
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700408 struct net *net = sock_net(sk);
David S. Miller3e129392012-07-10 04:01:57 -0700409 int flags = inet_sk_flowi_flags(sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700410
David S. Miller6bd023f2011-05-18 18:32:03 -0400411 flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
David S. Millere79d9bc2011-03-31 04:53:20 -0700412 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
Eric Dumazet74338192012-05-31 21:00:26 +0000413 sk->sk_protocol,
Eric Dumazet7586ece2012-06-20 05:02:19 +0000414 flags,
Eric Dumazet634fb9792013-10-09 15:21:29 -0700415 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
416 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
David S. Miller6bd023f2011-05-18 18:32:03 -0400417 security_req_classify_flow(req, flowi4_to_flowi(fl4));
418 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800419 if (IS_ERR(rt))
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800420 goto no_route;
Julian Anastasov155e8332012-10-08 11:41:18 +0000421 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800422 goto route_err;
Changli Gaod8d1f302010-06-10 23:31:35 -0700423 return &rt->dst;
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800424
425route_err:
426 ip_rt_put(rt);
427no_route:
428 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
429 return NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700430}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700431EXPORT_SYMBOL_GPL(inet_csk_route_req);
432
David S. Miller77357a92011-05-08 14:34:22 -0700433struct dst_entry *inet_csk_route_child_sock(struct sock *sk,
434 struct sock *newsk,
435 const struct request_sock *req)
436{
437 const struct inet_request_sock *ireq = inet_rsk(req);
438 struct inet_sock *newinet = inet_sk(newsk);
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000439 struct ip_options_rcu *opt;
David S. Miller77357a92011-05-08 14:34:22 -0700440 struct net *net = sock_net(sk);
441 struct flowi4 *fl4;
442 struct rtable *rt;
443
444 fl4 = &newinet->cork.fl.u.ip4;
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000445
446 rcu_read_lock();
447 opt = rcu_dereference(newinet->inet_opt);
David S. Miller77357a92011-05-08 14:34:22 -0700448 flowi4_init_output(fl4, sk->sk_bound_dev_if, sk->sk_mark,
449 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
450 sk->sk_protocol, inet_sk_flowi_flags(sk),
Eric Dumazet634fb9792013-10-09 15:21:29 -0700451 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
452 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
David S. Miller77357a92011-05-08 14:34:22 -0700453 security_req_classify_flow(req, flowi4_to_flowi(fl4));
454 rt = ip_route_output_flow(net, fl4, sk);
455 if (IS_ERR(rt))
456 goto no_route;
Julian Anastasov155e8332012-10-08 11:41:18 +0000457 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
David S. Miller77357a92011-05-08 14:34:22 -0700458 goto route_err;
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000459 rcu_read_unlock();
David S. Miller77357a92011-05-08 14:34:22 -0700460 return &rt->dst;
461
462route_err:
463 ip_rt_put(rt);
464no_route:
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000465 rcu_read_unlock();
David S. Miller77357a92011-05-08 14:34:22 -0700466 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
467 return NULL;
468}
469EXPORT_SYMBOL_GPL(inet_csk_route_child_sock);
470
Al Viro6b729772006-09-27 18:36:59 -0700471static inline u32 inet_synq_hash(const __be32 raddr, const __be16 rport,
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800472 const u32 rnd, const u32 synq_hsize)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700473{
Al Viro6b729772006-09-27 18:36:59 -0700474 return jhash_2words((__force u32)raddr, (__force u32)rport, rnd) & (synq_hsize - 1);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700475}
476
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000477#if IS_ENABLED(CONFIG_IPV6)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700478#define AF_INET_FAMILY(fam) ((fam) == AF_INET)
479#else
480#define AF_INET_FAMILY(fam) 1
481#endif
482
483struct request_sock *inet_csk_search_req(const struct sock *sk,
484 struct request_sock ***prevp,
Al Viro6b729772006-09-27 18:36:59 -0700485 const __be16 rport, const __be32 raddr,
Al Viro7f25afb2006-09-27 18:27:47 -0700486 const __be32 laddr)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700487{
488 const struct inet_connection_sock *icsk = inet_csk(sk);
489 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
490 struct request_sock *req, **prev;
491
492 for (prev = &lopt->syn_table[inet_synq_hash(raddr, rport, lopt->hash_rnd,
493 lopt->nr_table_entries)];
494 (req = *prev) != NULL;
495 prev = &req->dl_next) {
496 const struct inet_request_sock *ireq = inet_rsk(req);
497
Eric Dumazet634fb9792013-10-09 15:21:29 -0700498 if (ireq->ir_rmt_port == rport &&
499 ireq->ir_rmt_addr == raddr &&
500 ireq->ir_loc_addr == laddr &&
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700501 AF_INET_FAMILY(req->rsk_ops->family)) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700502 WARN_ON(req->sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700503 *prevp = prev;
504 break;
505 }
506 }
507
508 return req;
509}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700510EXPORT_SYMBOL_GPL(inet_csk_search_req);
511
512void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
Arnaldo Carvalho de Meloc2977c22005-12-13 23:15:12 -0800513 unsigned long timeout)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700514{
515 struct inet_connection_sock *icsk = inet_csk(sk);
516 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
Eric Dumazet634fb9792013-10-09 15:21:29 -0700517 const u32 h = inet_synq_hash(inet_rsk(req)->ir_rmt_addr,
518 inet_rsk(req)->ir_rmt_port,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700519 lopt->hash_rnd, lopt->nr_table_entries);
520
521 reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout);
522 inet_csk_reqsk_queue_added(sk, timeout);
523}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000524EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700525
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700526/* Only thing we need from tcp.h */
527extern int sysctl_tcp_synack_retries;
528
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700529
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000530/* Decide when to expire the request and when to resend SYN-ACK */
531static inline void syn_ack_recalc(struct request_sock *req, const int thresh,
532 const int max_retries,
533 const u8 rskq_defer_accept,
534 int *expire, int *resend)
535{
536 if (!rskq_defer_accept) {
Eric Dumazete6c022a2012-10-27 23:16:46 +0000537 *expire = req->num_timeout >= thresh;
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000538 *resend = 1;
539 return;
540 }
Eric Dumazete6c022a2012-10-27 23:16:46 +0000541 *expire = req->num_timeout >= thresh &&
542 (!inet_rsk(req)->acked || req->num_timeout >= max_retries);
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000543 /*
544 * Do not resend while waiting for data after ACK,
545 * start to resend on end of deferring period to give
546 * last chance for data or ACK to create established socket.
547 */
548 *resend = !inet_rsk(req)->acked ||
Eric Dumazete6c022a2012-10-27 23:16:46 +0000549 req->num_timeout >= rskq_defer_accept - 1;
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000550}
551
Eric Dumazete6c022a2012-10-27 23:16:46 +0000552int inet_rtx_syn_ack(struct sock *parent, struct request_sock *req)
553{
Christoph Paasch1a2c6182013-03-17 08:23:34 +0000554 int err = req->rsk_ops->rtx_syn_ack(parent, req);
Eric Dumazete6c022a2012-10-27 23:16:46 +0000555
556 if (!err)
557 req->num_retrans++;
558 return err;
559}
560EXPORT_SYMBOL(inet_rtx_syn_ack);
561
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700562void inet_csk_reqsk_queue_prune(struct sock *parent,
563 const unsigned long interval,
564 const unsigned long timeout,
565 const unsigned long max_rto)
566{
567 struct inet_connection_sock *icsk = inet_csk(parent);
568 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
569 struct listen_sock *lopt = queue->listen_opt;
David S. Millerec0a1962008-06-12 16:31:35 -0700570 int max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries;
571 int thresh = max_retries;
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700572 unsigned long now = jiffies;
573 struct request_sock **reqp, *req;
574 int i, budget;
575
576 if (lopt == NULL || lopt->qlen == 0)
577 return;
578
579 /* Normally all the openreqs are young and become mature
580 * (i.e. converted to established socket) for first timeout.
Eric Dumazetfd4f2ce2012-04-12 19:48:40 +0000581 * If synack was not acknowledged for 1 second, it means
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700582 * one of the following things: synack was lost, ack was lost,
583 * rtt is high or nobody planned to ack (i.e. synflood).
584 * When server is a bit loaded, queue is populated with old
585 * open requests, reducing effective size of queue.
586 * When server is well loaded, queue size reduces to zero
587 * after several minutes of work. It is not synflood,
588 * it is normal operation. The solution is pruning
589 * too old entries overriding normal timeout, when
590 * situation becomes dangerous.
591 *
592 * Essentially, we reserve half of room for young
593 * embrions; and abort old ones without pity, if old
594 * ones are about to clog our table.
595 */
596 if (lopt->qlen>>(lopt->max_qlen_log-1)) {
597 int young = (lopt->qlen_young<<1);
598
599 while (thresh > 2) {
600 if (lopt->qlen < young)
601 break;
602 thresh--;
603 young <<= 1;
604 }
605 }
606
David S. Millerec0a1962008-06-12 16:31:35 -0700607 if (queue->rskq_defer_accept)
608 max_retries = queue->rskq_defer_accept;
609
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700610 budget = 2 * (lopt->nr_table_entries / (timeout / interval));
611 i = lopt->clock_hand;
612
613 do {
614 reqp=&lopt->syn_table[i];
615 while ((req = *reqp) != NULL) {
616 if (time_after_eq(now, req->expires)) {
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000617 int expire = 0, resend = 0;
618
619 syn_ack_recalc(req, thresh, max_retries,
620 queue->rskq_defer_accept,
621 &expire, &resend);
Eric Dumazetc72e1182012-04-12 22:16:05 +0000622 req->rsk_ops->syn_ack_timeout(parent, req);
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000623 if (!expire &&
624 (!resend ||
Eric Dumazete6c022a2012-10-27 23:16:46 +0000625 !inet_rtx_syn_ack(parent, req) ||
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000626 inet_rsk(req)->acked)) {
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700627 unsigned long timeo;
628
Eric Dumazete6c022a2012-10-27 23:16:46 +0000629 if (req->num_timeout++ == 0)
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700630 lopt->qlen_young--;
Eric Dumazete6c022a2012-10-27 23:16:46 +0000631 timeo = min(timeout << req->num_timeout,
632 max_rto);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700633 req->expires = now + timeo;
634 reqp = &req->dl_next;
635 continue;
636 }
637
638 /* Drop this request */
639 inet_csk_reqsk_queue_unlink(parent, req, reqp);
640 reqsk_queue_removed(queue, req);
641 reqsk_free(req);
642 continue;
643 }
644 reqp = &req->dl_next;
645 }
646
647 i = (i + 1) & (lopt->nr_table_entries - 1);
648
649 } while (--budget > 0);
650
651 lopt->clock_hand = i;
652
653 if (lopt->qlen)
654 inet_csk_reset_keepalive_timer(parent, interval);
655}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700656EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_prune);
657
Eric Dumazete56c57d2011-11-08 17:07:07 -0500658/**
659 * inet_csk_clone_lock - clone an inet socket, and lock its clone
660 * @sk: the socket to clone
661 * @req: request_sock
662 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
663 *
664 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
665 */
666struct sock *inet_csk_clone_lock(const struct sock *sk,
667 const struct request_sock *req,
668 const gfp_t priority)
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700669{
Eric Dumazete56c57d2011-11-08 17:07:07 -0500670 struct sock *newsk = sk_clone_lock(sk, priority);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700671
672 if (newsk != NULL) {
673 struct inet_connection_sock *newicsk = inet_csk(newsk);
674
675 newsk->sk_state = TCP_SYN_RECV;
676 newicsk->icsk_bind_hash = NULL;
677
Eric Dumazet634fb9792013-10-09 15:21:29 -0700678 inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port;
Eric Dumazetb44084c2013-10-10 00:04:37 -0700679 inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num;
680 inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700681 newsk->sk_write_space = sk_stream_write_space;
682
683 newicsk->icsk_retransmits = 0;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300684 newicsk->icsk_backoff = 0;
685 newicsk->icsk_probes_out = 0;
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700686
687 /* Deinitialize accept_queue to trap illegal accesses. */
688 memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue));
Venkat Yekkirala4237c752006-07-24 23:32:50 -0700689
690 security_inet_csk_clone(newsk, req);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700691 }
692 return newsk;
693}
Eric Dumazete56c57d2011-11-08 17:07:07 -0500694EXPORT_SYMBOL_GPL(inet_csk_clone_lock);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700695
696/*
697 * At this point, there should be no process reference to this
698 * socket, and thus no user references at all. Therefore we
699 * can assume the socket waitqueue is inactive and nobody will
700 * try to jump onto it.
701 */
702void inet_csk_destroy_sock(struct sock *sk)
703{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700704 WARN_ON(sk->sk_state != TCP_CLOSE);
705 WARN_ON(!sock_flag(sk, SOCK_DEAD));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700706
707 /* It cannot be in hash table! */
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700708 WARN_ON(!sk_unhashed(sk));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700709
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000710 /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */
711 WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700712
713 sk->sk_prot->destroy(sk);
714
715 sk_stream_kill_queues(sk);
716
717 xfrm_sk_free_policy(sk);
718
719 sk_refcnt_debug_release(sk);
720
Eric Dumazetdd24c002008-11-25 21:17:14 -0800721 percpu_counter_dec(sk->sk_prot->orphan_count);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700722 sock_put(sk);
723}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700724EXPORT_SYMBOL(inet_csk_destroy_sock);
725
Christoph Paasche337e242012-12-14 04:07:58 +0000726/* This function allows to force a closure of a socket after the call to
727 * tcp/dccp_create_openreq_child().
728 */
729void inet_csk_prepare_forced_close(struct sock *sk)
Christoph Paaschc10cb5f2013-03-07 02:34:33 +0000730 __releases(&sk->sk_lock.slock)
Christoph Paasche337e242012-12-14 04:07:58 +0000731{
732 /* sk_clone_lock locked the socket and set refcnt to 2 */
733 bh_unlock_sock(sk);
734 sock_put(sk);
735
736 /* The below has to be done to allow calling inet_csk_destroy_sock */
737 sock_set_flag(sk, SOCK_DEAD);
738 percpu_counter_inc(sk->sk_prot->orphan_count);
739 inet_sk(sk)->inet_num = 0;
740}
741EXPORT_SYMBOL(inet_csk_prepare_forced_close);
742
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700743int inet_csk_listen_start(struct sock *sk, const int nr_table_entries)
744{
745 struct inet_sock *inet = inet_sk(sk);
746 struct inet_connection_sock *icsk = inet_csk(sk);
747 int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries);
748
749 if (rc != 0)
750 return rc;
751
752 sk->sk_max_ack_backlog = 0;
753 sk->sk_ack_backlog = 0;
754 inet_csk_delack_init(sk);
755
756 /* There is race window here: we announce ourselves listening,
757 * but this transition is still not validated by get_port().
758 * It is OK, because this socket enters to hash table only
759 * after validation is complete.
760 */
761 sk->sk_state = TCP_LISTEN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000762 if (!sk->sk_prot->get_port(sk, inet->inet_num)) {
763 inet->inet_sport = htons(inet->inet_num);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700764
765 sk_dst_reset(sk);
766 sk->sk_prot->hash(sk);
767
768 return 0;
769 }
770
771 sk->sk_state = TCP_CLOSE;
772 __reqsk_queue_destroy(&icsk->icsk_accept_queue);
773 return -EADDRINUSE;
774}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700775EXPORT_SYMBOL_GPL(inet_csk_listen_start);
776
777/*
778 * This routine closes sockets which have been at least partially
779 * opened, but not yet accepted.
780 */
781void inet_csk_listen_stop(struct sock *sk)
782{
783 struct inet_connection_sock *icsk = inet_csk(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000784 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700785 struct request_sock *acc_req;
786 struct request_sock *req;
787
788 inet_csk_delete_keepalive_timer(sk);
789
790 /* make all the listen_opt local to us */
Jerry Chu83368862012-08-31 12:29:12 +0000791 acc_req = reqsk_queue_yank_acceptq(queue);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700792
793 /* Following specs, it would be better either to send FIN
794 * (and enter FIN-WAIT-1, it is normal close)
795 * or to send active reset (abort).
796 * Certainly, it is pretty dangerous while synflood, but it is
797 * bad justification for our negligence 8)
798 * To be honest, we are not able to make either
799 * of the variants now. --ANK
800 */
Jerry Chu83368862012-08-31 12:29:12 +0000801 reqsk_queue_destroy(queue);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700802
803 while ((req = acc_req) != NULL) {
804 struct sock *child = req->sk;
805
806 acc_req = req->dl_next;
807
808 local_bh_disable();
809 bh_lock_sock(child);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700810 WARN_ON(sock_owned_by_user(child));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700811 sock_hold(child);
812
813 sk->sk_prot->disconnect(child, O_NONBLOCK);
814
815 sock_orphan(child);
816
Herbert Xueb4dea52008-12-29 23:04:08 -0800817 percpu_counter_inc(sk->sk_prot->orphan_count);
818
Eric Dumazet7ab45512012-09-06 08:07:13 +0000819 if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->listener) {
Jerry Chu83368862012-08-31 12:29:12 +0000820 BUG_ON(tcp_sk(child)->fastopen_rsk != req);
821 BUG_ON(sk != tcp_rsk(req)->listener);
822
823 /* Paranoid, to prevent race condition if
824 * an inbound pkt destined for child is
825 * blocked by sock lock in tcp_v4_rcv().
826 * Also to satisfy an assertion in
827 * tcp_v4_destroy_sock().
828 */
829 tcp_sk(child)->fastopen_rsk = NULL;
830 sock_put(sk);
831 }
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700832 inet_csk_destroy_sock(child);
833
834 bh_unlock_sock(child);
835 local_bh_enable();
836 sock_put(child);
837
838 sk_acceptq_removed(sk);
839 __reqsk_free(req);
840 }
Jerry Chu83368862012-08-31 12:29:12 +0000841 if (queue->fastopenq != NULL) {
842 /* Free all the reqs queued in rskq_rst_head. */
843 spin_lock_bh(&queue->fastopenq->lock);
844 acc_req = queue->fastopenq->rskq_rst_head;
845 queue->fastopenq->rskq_rst_head = NULL;
846 spin_unlock_bh(&queue->fastopenq->lock);
847 while ((req = acc_req) != NULL) {
848 acc_req = req->dl_next;
849 __reqsk_free(req);
850 }
851 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700852 WARN_ON(sk->sk_ack_backlog);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700853}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700854EXPORT_SYMBOL_GPL(inet_csk_listen_stop);
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800855
856void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr)
857{
858 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
859 const struct inet_sock *inet = inet_sk(sk);
860
861 sin->sin_family = AF_INET;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000862 sin->sin_addr.s_addr = inet->inet_daddr;
863 sin->sin_port = inet->inet_dport;
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800864}
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800865EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr);
Arnaldo Carvalho de Meloc4d93902006-03-20 22:01:03 -0800866
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800867#ifdef CONFIG_COMPAT
868int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname,
869 char __user *optval, int __user *optlen)
870{
David S. Millerdbeff122006-03-20 22:52:32 -0800871 const struct inet_connection_sock *icsk = inet_csk(sk);
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800872
873 if (icsk->icsk_af_ops->compat_getsockopt != NULL)
874 return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname,
875 optval, optlen);
876 return icsk->icsk_af_ops->getsockopt(sk, level, optname,
877 optval, optlen);
878}
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800879EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt);
880
881int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -0700882 char __user *optval, unsigned int optlen)
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800883{
David S. Millerdbeff122006-03-20 22:52:32 -0800884 const struct inet_connection_sock *icsk = inet_csk(sk);
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800885
886 if (icsk->icsk_af_ops->compat_setsockopt != NULL)
887 return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname,
888 optval, optlen);
889 return icsk->icsk_af_ops->setsockopt(sk, level, optname,
890 optval, optlen);
891}
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800892EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt);
893#endif
David S. Miller80d0a692012-07-16 03:28:06 -0700894
895static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl)
896{
Eric Dumazet5abf7f72012-07-17 22:42:13 +0200897 const struct inet_sock *inet = inet_sk(sk);
898 const struct ip_options_rcu *inet_opt;
David S. Miller80d0a692012-07-16 03:28:06 -0700899 __be32 daddr = inet->inet_daddr;
900 struct flowi4 *fl4;
901 struct rtable *rt;
902
903 rcu_read_lock();
904 inet_opt = rcu_dereference(inet->inet_opt);
905 if (inet_opt && inet_opt->opt.srr)
906 daddr = inet_opt->opt.faddr;
907 fl4 = &fl->u.ip4;
908 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr,
909 inet->inet_saddr, inet->inet_dport,
910 inet->inet_sport, sk->sk_protocol,
911 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if);
912 if (IS_ERR(rt))
913 rt = NULL;
914 if (rt)
915 sk_setup_caps(sk, &rt->dst);
916 rcu_read_unlock();
917
918 return &rt->dst;
919}
920
921struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu)
922{
923 struct dst_entry *dst = __sk_dst_check(sk, 0);
924 struct inet_sock *inet = inet_sk(sk);
925
926 if (!dst) {
927 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
928 if (!dst)
929 goto out;
930 }
David S. Miller6700c272012-07-17 03:29:28 -0700931 dst->ops->update_pmtu(dst, sk, NULL, mtu);
David S. Miller80d0a692012-07-16 03:28:06 -0700932
933 dst = __sk_dst_check(sk, 0);
934 if (!dst)
935 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
936out:
937 return dst;
938}
939EXPORT_SYMBOL_GPL(inet_csk_update_pmtu);