blob: 14d02ea905b6bea37240f88054f0cd42db73c4c2 [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
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -070032void inet_get_local_port_range(struct net *net, int *low, int *high)
Stephen Hemminger227b60f2007-10-10 17:30:46 -070033{
Eric Dumazet95c96172012-04-15 05:58:06 +000034 unsigned int seq;
35
Stephen Hemminger227b60f2007-10-10 17:30:46 -070036 do {
Cong Wangc9d8f1a2014-05-06 11:02:49 -070037 seq = read_seqbegin(&net->ipv4.ip_local_ports.lock);
Stephen Hemminger227b60f2007-10-10 17:30:46 -070038
Cong Wangc9d8f1a2014-05-06 11:02:49 -070039 *low = net->ipv4.ip_local_ports.range[0];
40 *high = net->ipv4.ip_local_ports.range[1];
41 } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq));
Stephen Hemminger227b60f2007-10-10 17:30:46 -070042}
43EXPORT_SYMBOL(inet_get_local_port_range);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070044
Arnaldo Carvalho de Melo971af182005-12-13 23:14:47 -080045int inet_csk_bind_conflict(const struct sock *sk,
Alex Copotaacd9282012-04-12 22:21:45 +000046 const struct inet_bind_bucket *tb, bool relax)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070047{
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070048 struct sock *sk2;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070049 int reuse = sk->sk_reuse;
Tom Herbertda5e3632013-01-22 09:50:24 +000050 int reuseport = sk->sk_reuseport;
51 kuid_t uid = sock_i_uid((struct sock *)sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070052
Pavel Emelyanov7477fd2e2008-04-14 02:42:27 -070053 /*
54 * Unlike other sk lookup places we do not check
55 * for sk_net here, since _all_ the socks listed
56 * in tb->owners list belong to the same net - the
57 * one this bucket belongs to.
58 */
59
Sasha Levinb67bfe02013-02-27 17:06:00 -080060 sk_for_each_bound(sk2, &tb->owners) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070061 if (sk != sk2 &&
62 !inet_v6_ipv6only(sk2) &&
63 (!sk->sk_bound_dev_if ||
64 !sk2->sk_bound_dev_if ||
65 sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) {
Tom Herbertda5e3632013-01-22 09:50:24 +000066 if ((!reuse || !sk2->sk_reuse ||
67 sk2->sk_state == TCP_LISTEN) &&
68 (!reuseport || !sk2->sk_reuseport ||
69 (sk2->sk_state != TCP_TIME_WAIT &&
70 !uid_eq(uid, sock_i_uid(sk2))))) {
Eric Dumazet50805462013-10-02 04:29:50 -070071
72 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
73 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070074 break;
David S. Miller8d238b22010-04-28 11:25:59 -070075 }
Alex Copotaacd9282012-04-12 22:21:45 +000076 if (!relax && reuse && sk2->sk_reuse &&
77 sk2->sk_state != TCP_LISTEN) {
Alex Copotaacd9282012-04-12 22:21:45 +000078
Eric Dumazet50805462013-10-02 04:29:50 -070079 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
80 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
Alex Copotaacd9282012-04-12 22:21:45 +000081 break;
82 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070083 }
84 }
Sasha Levinb67bfe02013-02-27 17:06:00 -080085 return sk2 != NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070086}
Arnaldo Carvalho de Melo971af182005-12-13 23:14:47 -080087EXPORT_SYMBOL_GPL(inet_csk_bind_conflict);
88
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070089/* Obtain a reference to a local port for the given sock,
90 * if snum is zero it means select any available local port.
91 */
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -080092int inet_csk_get_port(struct sock *sk, unsigned short snum)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070093{
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -070094 struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070095 struct inet_bind_hashbucket *head;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -070096 struct inet_bind_bucket *tb;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -080097 int ret, attempts = 5;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +090098 struct net *net = sock_net(sk);
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -080099 int smallest_size = -1, smallest_rover;
Tom Herbertda5e3632013-01-22 09:50:24 +0000100 kuid_t uid = sock_i_uid(sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700101
102 local_bh_disable();
103 if (!snum) {
Stephen Hemminger227b60f2007-10-10 17:30:46 -0700104 int remaining, rover, low, high;
105
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800106again:
Eric W. Biederman0bbf87d2013-09-28 14:10:59 -0700107 inet_get_local_port_range(net, &low, &high);
Anton Arapova25de532007-10-18 22:00:17 -0700108 remaining = (high - low) + 1;
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -0500109 smallest_rover = rover = prandom_u32() % remaining + low;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700110
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800111 smallest_size = -1;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700112 do {
WANG Cong122ff242014-05-12 16:04:53 -0700113 if (inet_is_local_reserved_port(net, rover))
Amerigo Wange3826f12010-05-05 00:27:06 +0000114 goto next_nolock;
Pavel Emelyanov7f635ab2008-06-16 17:12:49 -0700115 head = &hashinfo->bhash[inet_bhashfn(net, rover,
116 hashinfo->bhash_size)];
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700117 spin_lock(&head->lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800118 inet_bind_bucket_for_each(tb, &head->chain)
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800119 if (net_eq(ib_net(tb), net) && tb->port == rover) {
Tom Herbertda5e3632013-01-22 09:50:24 +0000120 if (((tb->fastreuse > 0 &&
121 sk->sk_reuse &&
122 sk->sk_state != TCP_LISTEN) ||
123 (tb->fastreuseport > 0 &&
124 sk->sk_reuseport &&
125 uid_eq(tb->fastuid, uid))) &&
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800126 (tb->num_owners < smallest_size || smallest_size == -1)) {
127 smallest_size = tb->num_owners;
128 smallest_rover = rover;
Alex Copotaacd9282012-04-12 22:21:45 +0000129 if (atomic_read(&hashinfo->bsockets) > (high - low) + 1 &&
130 !inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
David S. Miller8d238b22010-04-28 11:25:59 -0700131 snum = smallest_rover;
Flavio Leitnerfddb7b52012-01-25 08:34:52 +0000132 goto tb_found;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800133 }
134 }
Alex Copotaacd9282012-04-12 22:21:45 +0000135 if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
Flavio Leitner2b05ad32012-01-25 08:34:51 +0000136 snum = rover;
Flavio Leitnerfddb7b52012-01-25 08:34:52 +0000137 goto tb_found;
Flavio Leitner2b05ad32012-01-25 08:34:51 +0000138 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700139 goto next;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800140 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700141 break;
142 next:
143 spin_unlock(&head->lock);
Amerigo Wange3826f12010-05-05 00:27:06 +0000144 next_nolock:
Stephen Hemminger6df71632005-11-03 16:33:23 -0800145 if (++rover > high)
146 rover = low;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700147 } while (--remaining > 0);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700148
149 /* Exhausted local port range during search? It is not
150 * possible for us to be holding one of the bind hash
151 * locks if this test triggers, because if 'remaining'
152 * drops to zero, we broke out of the do/while loop at
153 * the top level, not from the 'break;' statement.
154 */
155 ret = 1;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800156 if (remaining <= 0) {
157 if (smallest_size != -1) {
158 snum = smallest_rover;
159 goto have_snum;
160 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700161 goto fail;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800162 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700163 /* OK, here is the one we will use. HEAD is
164 * non-NULL and we hold it's mutex.
165 */
166 snum = rover;
167 } else {
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800168have_snum:
Pavel Emelyanov7f635ab2008-06-16 17:12:49 -0700169 head = &hashinfo->bhash[inet_bhashfn(net, snum,
170 hashinfo->bhash_size)];
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700171 spin_lock(&head->lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800172 inet_bind_bucket_for_each(tb, &head->chain)
Octavian Purdila09ad9bc2009-11-25 15:14:13 -0800173 if (net_eq(ib_net(tb), net) && tb->port == snum)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700174 goto tb_found;
175 }
176 tb = NULL;
177 goto tb_not_found;
178tb_found:
179 if (!hlist_empty(&tb->owners)) {
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000180 if (sk->sk_reuse == SK_FORCE_REUSE)
181 goto success;
182
Tom Herbertda5e3632013-01-22 09:50:24 +0000183 if (((tb->fastreuse > 0 &&
184 sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
185 (tb->fastreuseport > 0 &&
186 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800187 smallest_size == -1) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700188 goto success;
189 } else {
190 ret = 1;
Alex Copotaacd9282012-04-12 22:21:45 +0000191 if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true)) {
Tom Herbertda5e3632013-01-22 09:50:24 +0000192 if (((sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000193 (tb->fastreuseport > 0 &&
194 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
Stephen Hemminger5add3002009-02-01 01:40:17 -0800195 smallest_size != -1 && --attempts >= 0) {
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800196 spin_unlock(&head->lock);
197 goto again;
198 }
Alex Copotaacd9282012-04-12 22:21:45 +0000199
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700200 goto fail_unlock;
Evgeniy Polyakova9d8f912009-01-19 16:46:02 -0800201 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700202 }
203 }
204tb_not_found:
205 ret = 1;
Pavel Emelyanov941b1d22008-01-31 05:05:50 -0800206 if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep,
207 net, head, snum)) == NULL)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700208 goto fail_unlock;
209 if (hlist_empty(&tb->owners)) {
210 if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
211 tb->fastreuse = 1;
212 else
213 tb->fastreuse = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000214 if (sk->sk_reuseport) {
215 tb->fastreuseport = 1;
216 tb->fastuid = uid;
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000217 } else
Tom Herbertda5e3632013-01-22 09:50:24 +0000218 tb->fastreuseport = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000219 } else {
220 if (tb->fastreuse &&
221 (!sk->sk_reuse || sk->sk_state == TCP_LISTEN))
222 tb->fastreuse = 0;
223 if (tb->fastreuseport &&
Tom Herbert9c5e0c02013-01-26 07:50:54 +0000224 (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid)))
Tom Herbertda5e3632013-01-22 09:50:24 +0000225 tb->fastreuseport = 0;
Tom Herbertda5e3632013-01-22 09:50:24 +0000226 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700227success:
228 if (!inet_csk(sk)->icsk_bind_hash)
229 inet_bind_hash(sk, tb, snum);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700230 WARN_ON(inet_csk(sk)->icsk_bind_hash != tb);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900231 ret = 0;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700232
233fail_unlock:
234 spin_unlock(&head->lock);
235fail:
236 local_bh_enable();
237 return ret;
238}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700239EXPORT_SYMBOL_GPL(inet_csk_get_port);
240
241/*
242 * Wait for an incoming connection, avoid race conditions. This must be called
243 * with the socket locked.
244 */
245static int inet_csk_wait_for_connect(struct sock *sk, long timeo)
246{
247 struct inet_connection_sock *icsk = inet_csk(sk);
248 DEFINE_WAIT(wait);
249 int err;
250
251 /*
252 * True wake-one mechanism for incoming connections: only
253 * one process gets woken up, not the 'whole herd'.
254 * Since we do not 'race & poll' for established sockets
255 * anymore, the common case will execute the loop only once.
256 *
257 * Subtle issue: "add_wait_queue_exclusive()" will be added
258 * after any current non-exclusive waiters, and we know that
259 * it will always _stay_ after any new non-exclusive waiters
260 * because all non-exclusive waiters are added at the
261 * beginning of the wait-queue. As such, it's ok to "drop"
262 * our exclusiveness temporarily when we get woken up without
263 * having to remove and re-insert us on the wait queue.
264 */
265 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +0000266 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700267 TASK_INTERRUPTIBLE);
268 release_sock(sk);
269 if (reqsk_queue_empty(&icsk->icsk_accept_queue))
270 timeo = schedule_timeout(timeo);
271 lock_sock(sk);
272 err = 0;
273 if (!reqsk_queue_empty(&icsk->icsk_accept_queue))
274 break;
275 err = -EINVAL;
276 if (sk->sk_state != TCP_LISTEN)
277 break;
278 err = sock_intr_errno(timeo);
279 if (signal_pending(current))
280 break;
281 err = -EAGAIN;
282 if (!timeo)
283 break;
284 }
Eric Dumazetaa395142010-04-20 13:03:51 +0000285 finish_wait(sk_sleep(sk), &wait);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700286 return err;
287}
288
289/*
290 * This will accept the next outstanding connection.
291 */
292struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
293{
294 struct inet_connection_sock *icsk = inet_csk(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000295 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700296 struct sock *newsk;
Jerry Chu83368862012-08-31 12:29:12 +0000297 struct request_sock *req;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700298 int error;
299
300 lock_sock(sk);
301
302 /* We need to make sure that this socket is listening,
303 * and that it has something pending.
304 */
305 error = -EINVAL;
306 if (sk->sk_state != TCP_LISTEN)
307 goto out_err;
308
309 /* Find already established connection */
Jerry Chu83368862012-08-31 12:29:12 +0000310 if (reqsk_queue_empty(queue)) {
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700311 long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
312
313 /* If this is a non blocking socket don't sleep */
314 error = -EAGAIN;
315 if (!timeo)
316 goto out_err;
317
318 error = inet_csk_wait_for_connect(sk, timeo);
319 if (error)
320 goto out_err;
321 }
Jerry Chu83368862012-08-31 12:29:12 +0000322 req = reqsk_queue_remove(queue);
323 newsk = req->sk;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700324
Jerry Chu83368862012-08-31 12:29:12 +0000325 sk_acceptq_removed(sk);
Eric Dumazet7ab45512012-09-06 08:07:13 +0000326 if (sk->sk_protocol == IPPROTO_TCP && queue->fastopenq != NULL) {
Jerry Chu83368862012-08-31 12:29:12 +0000327 spin_lock_bh(&queue->fastopenq->lock);
328 if (tcp_rsk(req)->listener) {
329 /* We are still waiting for the final ACK from 3WHS
330 * so can't free req now. Instead, we set req->sk to
331 * NULL to signify that the child socket is taken
332 * so reqsk_fastopen_remove() will free the req
333 * when 3WHS finishes (or is aborted).
334 */
335 req->sk = NULL;
336 req = NULL;
337 }
338 spin_unlock_bh(&queue->fastopenq->lock);
339 }
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700340out:
341 release_sock(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000342 if (req)
343 __reqsk_free(req);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700344 return newsk;
345out_err:
346 newsk = NULL;
Jerry Chu83368862012-08-31 12:29:12 +0000347 req = NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700348 *err = error;
349 goto out;
350}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700351EXPORT_SYMBOL(inet_csk_accept);
352
353/*
354 * Using different timers for retransmit, delayed acks and probes
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900355 * We may wish use just one timer maintaining a list of expire jiffies
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700356 * to optimize.
357 */
358void inet_csk_init_xmit_timers(struct sock *sk,
359 void (*retransmit_handler)(unsigned long),
360 void (*delack_handler)(unsigned long),
361 void (*keepalive_handler)(unsigned long))
362{
363 struct inet_connection_sock *icsk = inet_csk(sk);
364
Pavel Emelyanovb24b8a22008-01-23 21:20:07 -0800365 setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler,
366 (unsigned long)sk);
367 setup_timer(&icsk->icsk_delack_timer, delack_handler,
368 (unsigned long)sk);
369 setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700370 icsk->icsk_pending = icsk->icsk_ack.pending = 0;
371}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700372EXPORT_SYMBOL(inet_csk_init_xmit_timers);
373
374void inet_csk_clear_xmit_timers(struct sock *sk)
375{
376 struct inet_connection_sock *icsk = inet_csk(sk);
377
378 icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0;
379
380 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
381 sk_stop_timer(sk, &icsk->icsk_delack_timer);
382 sk_stop_timer(sk, &sk->sk_timer);
383}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700384EXPORT_SYMBOL(inet_csk_clear_xmit_timers);
385
386void inet_csk_delete_keepalive_timer(struct sock *sk)
387{
388 sk_stop_timer(sk, &sk->sk_timer);
389}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700390EXPORT_SYMBOL(inet_csk_delete_keepalive_timer);
391
392void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len)
393{
394 sk_reset_timer(sk, &sk->sk_timer, jiffies + len);
395}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700396EXPORT_SYMBOL(inet_csk_reset_keepalive_timer);
397
Jianjun Kongd93191002008-11-03 00:23:42 -0800398struct dst_entry *inet_csk_route_req(struct sock *sk,
David S. Miller6bd023f2011-05-18 18:32:03 -0400399 struct flowi4 *fl4,
David S. Millerba3f7f02012-07-17 14:02:46 -0700400 const struct request_sock *req)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700401{
402 struct rtable *rt;
403 const struct inet_request_sock *ireq = inet_rsk(req);
Eric Dumazetf6d8bd02011-04-21 09:45:37 +0000404 struct ip_options_rcu *opt = inet_rsk(req)->opt;
Pavel Emelyanov84a3aa02008-07-16 20:19:08 -0700405 struct net *net = sock_net(sk);
David S. Miller3e129392012-07-10 04:01:57 -0700406 int flags = inet_sk_flowi_flags(sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700407
Lorenzo Colitti84f39b02014-05-13 10:17:35 -0700408 flowi4_init_output(fl4, sk->sk_bound_dev_if, ireq->ir_mark,
David S. Millere79d9bc2011-03-31 04:53:20 -0700409 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
Eric Dumazet74338192012-05-31 21:00:26 +0000410 sk->sk_protocol,
Eric Dumazet7586ece2012-06-20 05:02:19 +0000411 flags,
Eric Dumazet634fb9792013-10-09 15:21:29 -0700412 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
413 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
David S. Miller6bd023f2011-05-18 18:32:03 -0400414 security_req_classify_flow(req, flowi4_to_flowi(fl4));
415 rt = ip_route_output_flow(net, fl4, sk);
David S. Millerb23dd4f2011-03-02 14:31:35 -0800416 if (IS_ERR(rt))
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800417 goto no_route;
Julian Anastasov155e8332012-10-08 11:41:18 +0000418 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800419 goto route_err;
Changli Gaod8d1f302010-06-10 23:31:35 -0700420 return &rt->dst;
Ilpo Järvinen857a6e02008-12-14 23:13:08 -0800421
422route_err:
423 ip_rt_put(rt);
424no_route:
425 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
426 return NULL;
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700427}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700428EXPORT_SYMBOL_GPL(inet_csk_route_req);
429
David S. Miller77357a92011-05-08 14:34:22 -0700430struct dst_entry *inet_csk_route_child_sock(struct sock *sk,
431 struct sock *newsk,
432 const struct request_sock *req)
433{
434 const struct inet_request_sock *ireq = inet_rsk(req);
435 struct inet_sock *newinet = inet_sk(newsk);
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000436 struct ip_options_rcu *opt;
David S. Miller77357a92011-05-08 14:34:22 -0700437 struct net *net = sock_net(sk);
438 struct flowi4 *fl4;
439 struct rtable *rt;
440
441 fl4 = &newinet->cork.fl.u.ip4;
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000442
443 rcu_read_lock();
444 opt = rcu_dereference(newinet->inet_opt);
Lorenzo Colitti84f39b02014-05-13 10:17:35 -0700445 flowi4_init_output(fl4, sk->sk_bound_dev_if, inet_rsk(req)->ir_mark,
David S. Miller77357a92011-05-08 14:34:22 -0700446 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
447 sk->sk_protocol, inet_sk_flowi_flags(sk),
Eric Dumazet634fb9792013-10-09 15:21:29 -0700448 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
449 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
David S. Miller77357a92011-05-08 14:34:22 -0700450 security_req_classify_flow(req, flowi4_to_flowi(fl4));
451 rt = ip_route_output_flow(net, fl4, sk);
452 if (IS_ERR(rt))
453 goto no_route;
Julian Anastasov155e8332012-10-08 11:41:18 +0000454 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
David S. Miller77357a92011-05-08 14:34:22 -0700455 goto route_err;
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000456 rcu_read_unlock();
David S. Miller77357a92011-05-08 14:34:22 -0700457 return &rt->dst;
458
459route_err:
460 ip_rt_put(rt);
461no_route:
Christoph Paasch1a7b27c2012-08-20 02:52:09 +0000462 rcu_read_unlock();
David S. Miller77357a92011-05-08 14:34:22 -0700463 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
464 return NULL;
465}
466EXPORT_SYMBOL_GPL(inet_csk_route_child_sock);
467
Al Viro6b729772006-09-27 18:36:59 -0700468static inline u32 inet_synq_hash(const __be32 raddr, const __be16 rport,
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800469 const u32 rnd, const u32 synq_hsize)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700470{
Al Viro6b729772006-09-27 18:36:59 -0700471 return jhash_2words((__force u32)raddr, (__force u32)rport, rnd) & (synq_hsize - 1);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700472}
473
Eric Dumazetdfd56b82011-12-10 09:48:31 +0000474#if IS_ENABLED(CONFIG_IPV6)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700475#define AF_INET_FAMILY(fam) ((fam) == AF_INET)
476#else
477#define AF_INET_FAMILY(fam) 1
478#endif
479
480struct request_sock *inet_csk_search_req(const struct sock *sk,
481 struct request_sock ***prevp,
Al Viro6b729772006-09-27 18:36:59 -0700482 const __be16 rport, const __be32 raddr,
Al Viro7f25afb2006-09-27 18:27:47 -0700483 const __be32 laddr)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700484{
485 const struct inet_connection_sock *icsk = inet_csk(sk);
486 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
487 struct request_sock *req, **prev;
488
489 for (prev = &lopt->syn_table[inet_synq_hash(raddr, rport, lopt->hash_rnd,
490 lopt->nr_table_entries)];
491 (req = *prev) != NULL;
492 prev = &req->dl_next) {
493 const struct inet_request_sock *ireq = inet_rsk(req);
494
Eric Dumazet634fb9792013-10-09 15:21:29 -0700495 if (ireq->ir_rmt_port == rport &&
496 ireq->ir_rmt_addr == raddr &&
497 ireq->ir_loc_addr == laddr &&
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700498 AF_INET_FAMILY(req->rsk_ops->family)) {
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700499 WARN_ON(req->sk);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700500 *prevp = prev;
501 break;
502 }
503 }
504
505 return req;
506}
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700507EXPORT_SYMBOL_GPL(inet_csk_search_req);
508
509void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
Arnaldo Carvalho de Meloc2977c22005-12-13 23:15:12 -0800510 unsigned long timeout)
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700511{
512 struct inet_connection_sock *icsk = inet_csk(sk);
513 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
Eric Dumazet634fb9792013-10-09 15:21:29 -0700514 const u32 h = inet_synq_hash(inet_rsk(req)->ir_rmt_addr,
515 inet_rsk(req)->ir_rmt_port,
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700516 lopt->hash_rnd, lopt->nr_table_entries);
517
518 reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout);
519 inet_csk_reqsk_queue_added(sk, timeout);
520}
Eric Dumazet4bc2f182010-07-09 21:22:10 +0000521EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add);
Arnaldo Carvalho de Melo3f421ba2005-08-09 20:11:08 -0700522
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700523/* Only thing we need from tcp.h */
524extern int sysctl_tcp_synack_retries;
525
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700526
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000527/* Decide when to expire the request and when to resend SYN-ACK */
528static inline void syn_ack_recalc(struct request_sock *req, const int thresh,
529 const int max_retries,
530 const u8 rskq_defer_accept,
531 int *expire, int *resend)
532{
533 if (!rskq_defer_accept) {
Eric Dumazete6c022a2012-10-27 23:16:46 +0000534 *expire = req->num_timeout >= thresh;
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000535 *resend = 1;
536 return;
537 }
Eric Dumazete6c022a2012-10-27 23:16:46 +0000538 *expire = req->num_timeout >= thresh &&
539 (!inet_rsk(req)->acked || req->num_timeout >= max_retries);
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000540 /*
541 * Do not resend while waiting for data after ACK,
542 * start to resend on end of deferring period to give
543 * last chance for data or ACK to create established socket.
544 */
545 *resend = !inet_rsk(req)->acked ||
Eric Dumazete6c022a2012-10-27 23:16:46 +0000546 req->num_timeout >= rskq_defer_accept - 1;
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000547}
548
Eric Dumazete6c022a2012-10-27 23:16:46 +0000549int inet_rtx_syn_ack(struct sock *parent, struct request_sock *req)
550{
Christoph Paasch1a2c6182013-03-17 08:23:34 +0000551 int err = req->rsk_ops->rtx_syn_ack(parent, req);
Eric Dumazete6c022a2012-10-27 23:16:46 +0000552
553 if (!err)
554 req->num_retrans++;
555 return err;
556}
557EXPORT_SYMBOL(inet_rtx_syn_ack);
558
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700559void inet_csk_reqsk_queue_prune(struct sock *parent,
560 const unsigned long interval,
561 const unsigned long timeout,
562 const unsigned long max_rto)
563{
564 struct inet_connection_sock *icsk = inet_csk(parent);
565 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
566 struct listen_sock *lopt = queue->listen_opt;
David S. Millerec0a1962008-06-12 16:31:35 -0700567 int max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries;
568 int thresh = max_retries;
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700569 unsigned long now = jiffies;
570 struct request_sock **reqp, *req;
571 int i, budget;
572
573 if (lopt == NULL || lopt->qlen == 0)
574 return;
575
576 /* Normally all the openreqs are young and become mature
577 * (i.e. converted to established socket) for first timeout.
Eric Dumazetfd4f2ce2012-04-12 19:48:40 +0000578 * If synack was not acknowledged for 1 second, it means
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700579 * one of the following things: synack was lost, ack was lost,
580 * rtt is high or nobody planned to ack (i.e. synflood).
581 * When server is a bit loaded, queue is populated with old
582 * open requests, reducing effective size of queue.
583 * When server is well loaded, queue size reduces to zero
584 * after several minutes of work. It is not synflood,
585 * it is normal operation. The solution is pruning
586 * too old entries overriding normal timeout, when
587 * situation becomes dangerous.
588 *
589 * Essentially, we reserve half of room for young
590 * embrions; and abort old ones without pity, if old
591 * ones are about to clog our table.
592 */
593 if (lopt->qlen>>(lopt->max_qlen_log-1)) {
594 int young = (lopt->qlen_young<<1);
595
596 while (thresh > 2) {
597 if (lopt->qlen < young)
598 break;
599 thresh--;
600 young <<= 1;
601 }
602 }
603
David S. Millerec0a1962008-06-12 16:31:35 -0700604 if (queue->rskq_defer_accept)
605 max_retries = queue->rskq_defer_accept;
606
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700607 budget = 2 * (lopt->nr_table_entries / (timeout / interval));
608 i = lopt->clock_hand;
609
610 do {
611 reqp=&lopt->syn_table[i];
612 while ((req = *reqp) != NULL) {
613 if (time_after_eq(now, req->expires)) {
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000614 int expire = 0, resend = 0;
615
616 syn_ack_recalc(req, thresh, max_retries,
617 queue->rskq_defer_accept,
618 &expire, &resend);
Eric Dumazetc72e1182012-04-12 22:16:05 +0000619 req->rsk_ops->syn_ack_timeout(parent, req);
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000620 if (!expire &&
621 (!resend ||
Eric Dumazete6c022a2012-10-27 23:16:46 +0000622 !inet_rtx_syn_ack(parent, req) ||
Julian Anastasov0c3d79b2009-10-19 10:03:58 +0000623 inet_rsk(req)->acked)) {
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700624 unsigned long timeo;
625
Eric Dumazete6c022a2012-10-27 23:16:46 +0000626 if (req->num_timeout++ == 0)
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700627 lopt->qlen_young--;
Eric Dumazete6c022a2012-10-27 23:16:46 +0000628 timeo = min(timeout << req->num_timeout,
629 max_rto);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700630 req->expires = now + timeo;
631 reqp = &req->dl_next;
632 continue;
633 }
634
635 /* Drop this request */
636 inet_csk_reqsk_queue_unlink(parent, req, reqp);
637 reqsk_queue_removed(queue, req);
638 reqsk_free(req);
639 continue;
640 }
641 reqp = &req->dl_next;
642 }
643
644 i = (i + 1) & (lopt->nr_table_entries - 1);
645
646 } while (--budget > 0);
647
648 lopt->clock_hand = i;
649
650 if (lopt->qlen)
651 inet_csk_reset_keepalive_timer(parent, interval);
652}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700653EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_prune);
654
Eric Dumazete56c57d2011-11-08 17:07:07 -0500655/**
656 * inet_csk_clone_lock - clone an inet socket, and lock its clone
657 * @sk: the socket to clone
658 * @req: request_sock
659 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
660 *
661 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
662 */
663struct sock *inet_csk_clone_lock(const struct sock *sk,
664 const struct request_sock *req,
665 const gfp_t priority)
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700666{
Eric Dumazete56c57d2011-11-08 17:07:07 -0500667 struct sock *newsk = sk_clone_lock(sk, priority);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700668
669 if (newsk != NULL) {
670 struct inet_connection_sock *newicsk = inet_csk(newsk);
671
672 newsk->sk_state = TCP_SYN_RECV;
673 newicsk->icsk_bind_hash = NULL;
674
Eric Dumazet634fb9792013-10-09 15:21:29 -0700675 inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port;
Eric Dumazetb44084c2013-10-10 00:04:37 -0700676 inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num;
677 inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700678 newsk->sk_write_space = sk_stream_write_space;
679
Lorenzo Colitti84f39b02014-05-13 10:17:35 -0700680 newsk->sk_mark = inet_rsk(req)->ir_mark;
681
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700682 newicsk->icsk_retransmits = 0;
Arnaldo Carvalho de Melo6687e982005-08-10 04:03:31 -0300683 newicsk->icsk_backoff = 0;
684 newicsk->icsk_probes_out = 0;
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700685
686 /* Deinitialize accept_queue to trap illegal accesses. */
687 memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue));
Venkat Yekkirala4237c752006-07-24 23:32:50 -0700688
689 security_inet_csk_clone(newsk, req);
Arnaldo Carvalho de Melo9f1d2602005-08-09 20:11:24 -0700690 }
691 return newsk;
692}
Eric Dumazete56c57d2011-11-08 17:07:07 -0500693EXPORT_SYMBOL_GPL(inet_csk_clone_lock);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700694
695/*
696 * At this point, there should be no process reference to this
697 * socket, and thus no user references at all. Therefore we
698 * can assume the socket waitqueue is inactive and nobody will
699 * try to jump onto it.
700 */
701void inet_csk_destroy_sock(struct sock *sk)
702{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700703 WARN_ON(sk->sk_state != TCP_CLOSE);
704 WARN_ON(!sock_flag(sk, SOCK_DEAD));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700705
706 /* It cannot be in hash table! */
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700707 WARN_ON(!sk_unhashed(sk));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700708
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000709 /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */
710 WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700711
712 sk->sk_prot->destroy(sk);
713
714 sk_stream_kill_queues(sk);
715
716 xfrm_sk_free_policy(sk);
717
718 sk_refcnt_debug_release(sk);
719
Eric Dumazetdd24c002008-11-25 21:17:14 -0800720 percpu_counter_dec(sk->sk_prot->orphan_count);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700721 sock_put(sk);
722}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700723EXPORT_SYMBOL(inet_csk_destroy_sock);
724
Christoph Paasche337e242012-12-14 04:07:58 +0000725/* This function allows to force a closure of a socket after the call to
726 * tcp/dccp_create_openreq_child().
727 */
728void inet_csk_prepare_forced_close(struct sock *sk)
Christoph Paaschc10cb5f2013-03-07 02:34:33 +0000729 __releases(&sk->sk_lock.slock)
Christoph Paasche337e242012-12-14 04:07:58 +0000730{
731 /* sk_clone_lock locked the socket and set refcnt to 2 */
732 bh_unlock_sock(sk);
733 sock_put(sk);
734
735 /* The below has to be done to allow calling inet_csk_destroy_sock */
736 sock_set_flag(sk, SOCK_DEAD);
737 percpu_counter_inc(sk->sk_prot->orphan_count);
738 inet_sk(sk)->inet_num = 0;
739}
740EXPORT_SYMBOL(inet_csk_prepare_forced_close);
741
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700742int inet_csk_listen_start(struct sock *sk, const int nr_table_entries)
743{
744 struct inet_sock *inet = inet_sk(sk);
745 struct inet_connection_sock *icsk = inet_csk(sk);
746 int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries);
747
748 if (rc != 0)
749 return rc;
750
751 sk->sk_max_ack_backlog = 0;
752 sk->sk_ack_backlog = 0;
753 inet_csk_delack_init(sk);
754
755 /* There is race window here: we announce ourselves listening,
756 * but this transition is still not validated by get_port().
757 * It is OK, because this socket enters to hash table only
758 * after validation is complete.
759 */
760 sk->sk_state = TCP_LISTEN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000761 if (!sk->sk_prot->get_port(sk, inet->inet_num)) {
762 inet->inet_sport = htons(inet->inet_num);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700763
764 sk_dst_reset(sk);
765 sk->sk_prot->hash(sk);
766
767 return 0;
768 }
769
770 sk->sk_state = TCP_CLOSE;
771 __reqsk_queue_destroy(&icsk->icsk_accept_queue);
772 return -EADDRINUSE;
773}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700774EXPORT_SYMBOL_GPL(inet_csk_listen_start);
775
776/*
777 * This routine closes sockets which have been at least partially
778 * opened, but not yet accepted.
779 */
780void inet_csk_listen_stop(struct sock *sk)
781{
782 struct inet_connection_sock *icsk = inet_csk(sk);
Jerry Chu83368862012-08-31 12:29:12 +0000783 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700784 struct request_sock *acc_req;
785 struct request_sock *req;
786
787 inet_csk_delete_keepalive_timer(sk);
788
789 /* make all the listen_opt local to us */
Jerry Chu83368862012-08-31 12:29:12 +0000790 acc_req = reqsk_queue_yank_acceptq(queue);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700791
792 /* Following specs, it would be better either to send FIN
793 * (and enter FIN-WAIT-1, it is normal close)
794 * or to send active reset (abort).
795 * Certainly, it is pretty dangerous while synflood, but it is
796 * bad justification for our negligence 8)
797 * To be honest, we are not able to make either
798 * of the variants now. --ANK
799 */
Jerry Chu83368862012-08-31 12:29:12 +0000800 reqsk_queue_destroy(queue);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700801
802 while ((req = acc_req) != NULL) {
803 struct sock *child = req->sk;
804
805 acc_req = req->dl_next;
806
807 local_bh_disable();
808 bh_lock_sock(child);
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700809 WARN_ON(sock_owned_by_user(child));
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700810 sock_hold(child);
811
812 sk->sk_prot->disconnect(child, O_NONBLOCK);
813
814 sock_orphan(child);
815
Herbert Xueb4dea52008-12-29 23:04:08 -0800816 percpu_counter_inc(sk->sk_prot->orphan_count);
817
Eric Dumazet7ab45512012-09-06 08:07:13 +0000818 if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->listener) {
Jerry Chu83368862012-08-31 12:29:12 +0000819 BUG_ON(tcp_sk(child)->fastopen_rsk != req);
820 BUG_ON(sk != tcp_rsk(req)->listener);
821
822 /* Paranoid, to prevent race condition if
823 * an inbound pkt destined for child is
824 * blocked by sock lock in tcp_v4_rcv().
825 * Also to satisfy an assertion in
826 * tcp_v4_destroy_sock().
827 */
828 tcp_sk(child)->fastopen_rsk = NULL;
829 sock_put(sk);
830 }
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700831 inet_csk_destroy_sock(child);
832
833 bh_unlock_sock(child);
834 local_bh_enable();
835 sock_put(child);
836
837 sk_acceptq_removed(sk);
838 __reqsk_free(req);
839 }
Jerry Chu83368862012-08-31 12:29:12 +0000840 if (queue->fastopenq != NULL) {
841 /* Free all the reqs queued in rskq_rst_head. */
842 spin_lock_bh(&queue->fastopenq->lock);
843 acc_req = queue->fastopenq->rskq_rst_head;
844 queue->fastopenq->rskq_rst_head = NULL;
845 spin_unlock_bh(&queue->fastopenq->lock);
846 while ((req = acc_req) != NULL) {
847 acc_req = req->dl_next;
848 __reqsk_free(req);
849 }
850 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700851 WARN_ON(sk->sk_ack_backlog);
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700852}
Arnaldo Carvalho de Meloa019d6f2005-08-09 20:15:09 -0700853EXPORT_SYMBOL_GPL(inet_csk_listen_stop);
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800854
855void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr)
856{
857 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
858 const struct inet_sock *inet = inet_sk(sk);
859
860 sin->sin_family = AF_INET;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000861 sin->sin_addr.s_addr = inet->inet_daddr;
862 sin->sin_port = inet->inet_dport;
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800863}
Arnaldo Carvalho de Meloaf05dc92005-12-13 23:16:04 -0800864EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr);
Arnaldo Carvalho de Meloc4d93902006-03-20 22:01:03 -0800865
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800866#ifdef CONFIG_COMPAT
867int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname,
868 char __user *optval, int __user *optlen)
869{
David S. Millerdbeff122006-03-20 22:52:32 -0800870 const struct inet_connection_sock *icsk = inet_csk(sk);
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800871
872 if (icsk->icsk_af_ops->compat_getsockopt != NULL)
873 return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname,
874 optval, optlen);
875 return icsk->icsk_af_ops->getsockopt(sk, level, optname,
876 optval, optlen);
877}
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800878EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt);
879
880int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -0700881 char __user *optval, unsigned int optlen)
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800882{
David S. Millerdbeff122006-03-20 22:52:32 -0800883 const struct inet_connection_sock *icsk = inet_csk(sk);
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800884
885 if (icsk->icsk_af_ops->compat_setsockopt != NULL)
886 return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname,
887 optval, optlen);
888 return icsk->icsk_af_ops->setsockopt(sk, level, optname,
889 optval, optlen);
890}
Arnaldo Carvalho de Melodec73ff2006-03-20 22:46:16 -0800891EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt);
892#endif
David S. Miller80d0a692012-07-16 03:28:06 -0700893
894static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl)
895{
Eric Dumazet5abf7f72012-07-17 22:42:13 +0200896 const struct inet_sock *inet = inet_sk(sk);
897 const struct ip_options_rcu *inet_opt;
David S. Miller80d0a692012-07-16 03:28:06 -0700898 __be32 daddr = inet->inet_daddr;
899 struct flowi4 *fl4;
900 struct rtable *rt;
901
902 rcu_read_lock();
903 inet_opt = rcu_dereference(inet->inet_opt);
904 if (inet_opt && inet_opt->opt.srr)
905 daddr = inet_opt->opt.faddr;
906 fl4 = &fl->u.ip4;
907 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr,
908 inet->inet_saddr, inet->inet_dport,
909 inet->inet_sport, sk->sk_protocol,
910 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if);
911 if (IS_ERR(rt))
912 rt = NULL;
913 if (rt)
914 sk_setup_caps(sk, &rt->dst);
915 rcu_read_unlock();
916
917 return &rt->dst;
918}
919
920struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu)
921{
922 struct dst_entry *dst = __sk_dst_check(sk, 0);
923 struct inet_sock *inet = inet_sk(sk);
924
925 if (!dst) {
926 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
927 if (!dst)
928 goto out;
929 }
David S. Miller6700c272012-07-17 03:29:28 -0700930 dst->ops->update_pmtu(dst, sk, NULL, mtu);
David S. Miller80d0a692012-07-16 03:28:06 -0700931
932 dst = __sk_dst_check(sk, 0);
933 if (!dst)
934 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
935out:
936 return dst;
937}
938EXPORT_SYMBOL_GPL(inet_csk_update_pmtu);