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.
6 * Support for INET connection oriented protocols.
8 * Authors: See the TCP sources
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.
16 #include <linux/module.h>
17 #include <linux/jhash.h>
19 #include <net/inet_connection_sock.h>
20 #include <net/inet_hashtables.h>
21 #include <net/inet_timewait_sock.h>
23 #include <net/route.h>
24 #include <net/tcp_states.h>
28 const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n";
29 EXPORT_SYMBOL(inet_csk_timer_bug_msg);
32 void inet_get_local_port_range(struct net *net, int *low, int *high)
37 seq = read_seqbegin(&net->ipv4.ip_local_ports.lock);
39 *low = net->ipv4.ip_local_ports.range[0];
40 *high = net->ipv4.ip_local_ports.range[1];
41 } while (read_seqretry(&net->ipv4.ip_local_ports.lock, seq));
43 EXPORT_SYMBOL(inet_get_local_port_range);
45 int inet_csk_bind_conflict(const struct sock *sk,
46 const struct inet_bind_bucket *tb, bool relax)
49 int reuse = sk->sk_reuse;
50 int reuseport = sk->sk_reuseport;
51 kuid_t uid = sock_i_uid((struct sock *)sk);
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.
60 sk_for_each_bound(sk2, &tb->owners) {
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)) {
66 if ((!reuse || !sk2->sk_reuse ||
67 sk2->sk_state == TCP_LISTEN) &&
68 (!reuseport || !sk2->sk_reuseport ||
69 (sk2->sk_state != TCP_TIME_WAIT &&
70 !uid_eq(uid, sock_i_uid(sk2))))) {
72 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
73 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
76 if (!relax && reuse && sk2->sk_reuse &&
77 sk2->sk_state != TCP_LISTEN) {
79 if (!sk2->sk_rcv_saddr || !sk->sk_rcv_saddr ||
80 sk2->sk_rcv_saddr == sk->sk_rcv_saddr)
87 EXPORT_SYMBOL_GPL(inet_csk_bind_conflict);
89 /* Obtain a reference to a local port for the given sock,
90 * if snum is zero it means select any available local port.
92 int inet_csk_get_port(struct sock *sk, unsigned short snum)
94 struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
95 struct inet_bind_hashbucket *head;
96 struct inet_bind_bucket *tb;
97 int ret, attempts = 5;
98 struct net *net = sock_net(sk);
99 int smallest_size = -1, smallest_rover;
100 kuid_t uid = sock_i_uid(sk);
104 int remaining, rover, low, high;
107 inet_get_local_port_range(net, &low, &high);
108 remaining = (high - low) + 1;
109 smallest_rover = rover = prandom_u32() % remaining + low;
113 if (inet_is_local_reserved_port(net, rover))
115 head = &hashinfo->bhash[inet_bhashfn(net, rover,
116 hashinfo->bhash_size)];
117 spin_lock(&head->lock);
118 inet_bind_bucket_for_each(tb, &head->chain)
119 if (net_eq(ib_net(tb), net) && tb->port == rover) {
120 if (((tb->fastreuse > 0 &&
122 sk->sk_state != TCP_LISTEN) ||
123 (tb->fastreuseport > 0 &&
125 uid_eq(tb->fastuid, uid))) &&
126 (tb->num_owners < smallest_size || smallest_size == -1)) {
127 smallest_size = tb->num_owners;
128 smallest_rover = rover;
129 if (atomic_read(&hashinfo->bsockets) > (high - low) + 1 &&
130 !inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
131 snum = smallest_rover;
135 if (!inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, false)) {
143 spin_unlock(&head->lock);
147 } while (--remaining > 0);
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.
156 if (remaining <= 0) {
157 if (smallest_size != -1) {
158 snum = smallest_rover;
163 /* OK, here is the one we will use. HEAD is
164 * non-NULL and we hold it's mutex.
169 head = &hashinfo->bhash[inet_bhashfn(net, snum,
170 hashinfo->bhash_size)];
171 spin_lock(&head->lock);
172 inet_bind_bucket_for_each(tb, &head->chain)
173 if (net_eq(ib_net(tb), net) && tb->port == snum)
179 if (!hlist_empty(&tb->owners)) {
180 if (sk->sk_reuse == SK_FORCE_REUSE)
183 if (((tb->fastreuse > 0 &&
184 sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
185 (tb->fastreuseport > 0 &&
186 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
187 smallest_size == -1) {
191 if (inet_csk(sk)->icsk_af_ops->bind_conflict(sk, tb, true)) {
192 if (((sk->sk_reuse && sk->sk_state != TCP_LISTEN) ||
193 (tb->fastreuseport > 0 &&
194 sk->sk_reuseport && uid_eq(tb->fastuid, uid))) &&
195 smallest_size != -1 && --attempts >= 0) {
196 spin_unlock(&head->lock);
206 if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep,
207 net, head, snum)) == NULL)
209 if (hlist_empty(&tb->owners)) {
210 if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
214 if (sk->sk_reuseport) {
215 tb->fastreuseport = 1;
218 tb->fastreuseport = 0;
221 (!sk->sk_reuse || sk->sk_state == TCP_LISTEN))
223 if (tb->fastreuseport &&
224 (!sk->sk_reuseport || !uid_eq(tb->fastuid, uid)))
225 tb->fastreuseport = 0;
228 if (!inet_csk(sk)->icsk_bind_hash)
229 inet_bind_hash(sk, tb, snum);
230 WARN_ON(inet_csk(sk)->icsk_bind_hash != tb);
234 spin_unlock(&head->lock);
239 EXPORT_SYMBOL_GPL(inet_csk_get_port);
242 * Wait for an incoming connection, avoid race conditions. This must be called
243 * with the socket locked.
245 static int inet_csk_wait_for_connect(struct sock *sk, long timeo)
247 struct inet_connection_sock *icsk = inet_csk(sk);
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.
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.
266 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
269 if (reqsk_queue_empty(&icsk->icsk_accept_queue))
270 timeo = schedule_timeout(timeo);
273 if (!reqsk_queue_empty(&icsk->icsk_accept_queue))
276 if (sk->sk_state != TCP_LISTEN)
278 err = sock_intr_errno(timeo);
279 if (signal_pending(current))
285 finish_wait(sk_sleep(sk), &wait);
290 * This will accept the next outstanding connection.
292 struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
294 struct inet_connection_sock *icsk = inet_csk(sk);
295 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
296 struct request_sock *req;
302 /* We need to make sure that this socket is listening,
303 * and that it has something pending.
306 if (sk->sk_state != TCP_LISTEN)
309 /* Find already established connection */
310 if (reqsk_queue_empty(queue)) {
311 long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
313 /* If this is a non blocking socket don't sleep */
318 error = inet_csk_wait_for_connect(sk, timeo);
322 req = reqsk_queue_remove(queue);
325 sk_acceptq_removed(sk);
326 if (sk->sk_protocol == IPPROTO_TCP &&
327 tcp_rsk(req)->tfo_listener &&
329 spin_lock_bh(&queue->fastopenq->lock);
330 if (tcp_rsk(req)->tfo_listener) {
331 /* We are still waiting for the final ACK from 3WHS
332 * so can't free req now. Instead, we set req->sk to
333 * NULL to signify that the child socket is taken
334 * so reqsk_fastopen_remove() will free the req
335 * when 3WHS finishes (or is aborted).
340 spin_unlock_bh(&queue->fastopenq->lock);
353 EXPORT_SYMBOL(inet_csk_accept);
356 * Using different timers for retransmit, delayed acks and probes
357 * We may wish use just one timer maintaining a list of expire jiffies
360 void inet_csk_init_xmit_timers(struct sock *sk,
361 void (*retransmit_handler)(unsigned long),
362 void (*delack_handler)(unsigned long),
363 void (*keepalive_handler)(unsigned long))
365 struct inet_connection_sock *icsk = inet_csk(sk);
367 setup_timer(&icsk->icsk_retransmit_timer, retransmit_handler,
369 setup_timer(&icsk->icsk_delack_timer, delack_handler,
371 setup_timer(&sk->sk_timer, keepalive_handler, (unsigned long)sk);
372 icsk->icsk_pending = icsk->icsk_ack.pending = 0;
374 EXPORT_SYMBOL(inet_csk_init_xmit_timers);
376 void inet_csk_clear_xmit_timers(struct sock *sk)
378 struct inet_connection_sock *icsk = inet_csk(sk);
380 icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0;
382 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
383 sk_stop_timer(sk, &icsk->icsk_delack_timer);
384 sk_stop_timer(sk, &sk->sk_timer);
386 EXPORT_SYMBOL(inet_csk_clear_xmit_timers);
388 void inet_csk_delete_keepalive_timer(struct sock *sk)
390 sk_stop_timer(sk, &sk->sk_timer);
392 EXPORT_SYMBOL(inet_csk_delete_keepalive_timer);
394 void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len)
396 sk_reset_timer(sk, &sk->sk_timer, jiffies + len);
398 EXPORT_SYMBOL(inet_csk_reset_keepalive_timer);
400 struct dst_entry *inet_csk_route_req(struct sock *sk,
402 const struct request_sock *req)
405 const struct inet_request_sock *ireq = inet_rsk(req);
406 struct ip_options_rcu *opt = inet_rsk(req)->opt;
407 struct net *net = sock_net(sk);
408 int flags = inet_sk_flowi_flags(sk);
410 flowi4_init_output(fl4, sk->sk_bound_dev_if, ireq->ir_mark,
411 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
414 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
415 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
416 security_req_classify_flow(req, flowi4_to_flowi(fl4));
417 rt = ip_route_output_flow(net, fl4, sk);
420 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
427 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
430 EXPORT_SYMBOL_GPL(inet_csk_route_req);
432 struct dst_entry *inet_csk_route_child_sock(struct sock *sk,
434 const struct request_sock *req)
436 const struct inet_request_sock *ireq = inet_rsk(req);
437 struct inet_sock *newinet = inet_sk(newsk);
438 struct ip_options_rcu *opt;
439 struct net *net = sock_net(sk);
443 fl4 = &newinet->cork.fl.u.ip4;
446 opt = rcu_dereference(newinet->inet_opt);
447 flowi4_init_output(fl4, sk->sk_bound_dev_if, inet_rsk(req)->ir_mark,
448 RT_CONN_FLAGS(sk), RT_SCOPE_UNIVERSE,
449 sk->sk_protocol, inet_sk_flowi_flags(sk),
450 (opt && opt->opt.srr) ? opt->opt.faddr : ireq->ir_rmt_addr,
451 ireq->ir_loc_addr, ireq->ir_rmt_port, inet_sk(sk)->inet_sport);
452 security_req_classify_flow(req, flowi4_to_flowi(fl4));
453 rt = ip_route_output_flow(net, fl4, sk);
456 if (opt && opt->opt.is_strictroute && rt->rt_uses_gateway)
465 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
468 EXPORT_SYMBOL_GPL(inet_csk_route_child_sock);
470 static inline u32 inet_synq_hash(const __be32 raddr, const __be16 rport,
471 const u32 rnd, const u32 synq_hsize)
473 return jhash_2words((__force u32)raddr, (__force u32)rport, rnd) & (synq_hsize - 1);
476 #if IS_ENABLED(CONFIG_IPV6)
477 #define AF_INET_FAMILY(fam) ((fam) == AF_INET)
479 #define AF_INET_FAMILY(fam) 1
482 struct request_sock *inet_csk_search_req(const struct sock *sk,
483 const __be16 rport, const __be32 raddr,
486 const struct inet_connection_sock *icsk = inet_csk(sk);
487 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
488 struct request_sock *req;
490 for (req = lopt->syn_table[inet_synq_hash(raddr, rport, lopt->hash_rnd,
491 lopt->nr_table_entries)];
493 req = req->dl_next) {
494 const struct inet_request_sock *ireq = inet_rsk(req);
496 if (ireq->ir_rmt_port == rport &&
497 ireq->ir_rmt_addr == raddr &&
498 ireq->ir_loc_addr == laddr &&
499 AF_INET_FAMILY(req->rsk_ops->family)) {
507 EXPORT_SYMBOL_GPL(inet_csk_search_req);
509 void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
510 unsigned long timeout)
512 struct inet_connection_sock *icsk = inet_csk(sk);
513 struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
514 const u32 h = inet_synq_hash(inet_rsk(req)->ir_rmt_addr,
515 inet_rsk(req)->ir_rmt_port,
516 lopt->hash_rnd, lopt->nr_table_entries);
518 reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout);
519 inet_csk_reqsk_queue_added(sk, timeout);
521 EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add);
523 /* Only thing we need from tcp.h */
524 extern int sysctl_tcp_synack_retries;
527 /* Decide when to expire the request and when to resend SYN-ACK */
528 static inline void syn_ack_recalc(struct request_sock *req, const int thresh,
529 const int max_retries,
530 const u8 rskq_defer_accept,
531 int *expire, int *resend)
533 if (!rskq_defer_accept) {
534 *expire = req->num_timeout >= thresh;
538 *expire = req->num_timeout >= thresh &&
539 (!inet_rsk(req)->acked || req->num_timeout >= max_retries);
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.
545 *resend = !inet_rsk(req)->acked ||
546 req->num_timeout >= rskq_defer_accept - 1;
549 int inet_rtx_syn_ack(struct sock *parent, struct request_sock *req)
551 int err = req->rsk_ops->rtx_syn_ack(parent, req);
557 EXPORT_SYMBOL(inet_rtx_syn_ack);
559 void inet_csk_reqsk_queue_prune(struct sock *parent,
560 const unsigned long interval,
561 const unsigned long timeout,
562 const unsigned long max_rto)
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;
567 int max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries;
568 int thresh = max_retries;
569 unsigned long now = jiffies;
570 struct request_sock **reqp, *req;
573 if (lopt == NULL || lopt->qlen == 0)
576 /* Normally all the openreqs are young and become mature
577 * (i.e. converted to established socket) for first timeout.
578 * If synack was not acknowledged for 1 second, it means
579 * one of the following things: synack was lost, ack was lost,
580 * rtt is high or nobody planned to ack (i.e. synflood).
581 * When server is a bit loaded, queue is populated with old
582 * open requests, reducing effective size of queue.
583 * When server is well loaded, queue size reduces to zero
584 * after several minutes of work. It is not synflood,
585 * it is normal operation. The solution is pruning
586 * too old entries overriding normal timeout, when
587 * situation becomes dangerous.
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.
593 if (lopt->qlen>>(lopt->max_qlen_log-1)) {
594 int young = (lopt->qlen_young<<1);
597 if (lopt->qlen < young)
604 if (queue->rskq_defer_accept)
605 max_retries = queue->rskq_defer_accept;
607 budget = 2 * (lopt->nr_table_entries / (timeout / interval));
608 i = lopt->clock_hand;
611 reqp = &lopt->syn_table[i];
614 write_lock(&queue->syn_wait_lock);
615 while ((req = *reqp) != NULL) {
616 if (time_after_eq(now, req->expires)) {
617 int expire = 0, resend = 0;
619 syn_ack_recalc(req, thresh, max_retries,
620 queue->rskq_defer_accept,
622 req->rsk_ops->syn_ack_timeout(parent, req);
625 !inet_rtx_syn_ack(parent, req) ||
626 inet_rsk(req)->acked)) {
629 if (req->num_timeout++ == 0)
631 timeo = min(timeout << req->num_timeout,
633 req->expires = now + timeo;
634 reqp = &req->dl_next;
638 /* Drop this request */
639 *reqp = req->dl_next;
640 reqsk_queue_removed(queue, req);
644 reqp = &req->dl_next;
646 write_unlock(&queue->syn_wait_lock);
648 i = (i + 1) & (lopt->nr_table_entries - 1);
650 } while (--budget > 0);
652 lopt->clock_hand = i;
655 inet_csk_reset_keepalive_timer(parent, interval);
657 EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_prune);
660 * inet_csk_clone_lock - clone an inet socket, and lock its clone
661 * @sk: the socket to clone
663 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
665 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
667 struct sock *inet_csk_clone_lock(const struct sock *sk,
668 const struct request_sock *req,
669 const gfp_t priority)
671 struct sock *newsk = sk_clone_lock(sk, priority);
674 struct inet_connection_sock *newicsk = inet_csk(newsk);
676 newsk->sk_state = TCP_SYN_RECV;
677 newicsk->icsk_bind_hash = NULL;
679 inet_sk(newsk)->inet_dport = inet_rsk(req)->ir_rmt_port;
680 inet_sk(newsk)->inet_num = inet_rsk(req)->ir_num;
681 inet_sk(newsk)->inet_sport = htons(inet_rsk(req)->ir_num);
682 newsk->sk_write_space = sk_stream_write_space;
684 newsk->sk_mark = inet_rsk(req)->ir_mark;
685 atomic64_set(&newsk->sk_cookie,
686 atomic64_read(&inet_rsk(req)->ir_cookie));
688 newicsk->icsk_retransmits = 0;
689 newicsk->icsk_backoff = 0;
690 newicsk->icsk_probes_out = 0;
692 /* Deinitialize accept_queue to trap illegal accesses. */
693 memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue));
695 security_inet_csk_clone(newsk, req);
699 EXPORT_SYMBOL_GPL(inet_csk_clone_lock);
702 * At this point, there should be no process reference to this
703 * socket, and thus no user references at all. Therefore we
704 * can assume the socket waitqueue is inactive and nobody will
705 * try to jump onto it.
707 void inet_csk_destroy_sock(struct sock *sk)
709 WARN_ON(sk->sk_state != TCP_CLOSE);
710 WARN_ON(!sock_flag(sk, SOCK_DEAD));
712 /* It cannot be in hash table! */
713 WARN_ON(!sk_unhashed(sk));
715 /* If it has not 0 inet_sk(sk)->inet_num, it must be bound */
716 WARN_ON(inet_sk(sk)->inet_num && !inet_csk(sk)->icsk_bind_hash);
718 sk->sk_prot->destroy(sk);
720 sk_stream_kill_queues(sk);
722 xfrm_sk_free_policy(sk);
724 sk_refcnt_debug_release(sk);
726 percpu_counter_dec(sk->sk_prot->orphan_count);
729 EXPORT_SYMBOL(inet_csk_destroy_sock);
731 /* This function allows to force a closure of a socket after the call to
732 * tcp/dccp_create_openreq_child().
734 void inet_csk_prepare_forced_close(struct sock *sk)
735 __releases(&sk->sk_lock.slock)
737 /* sk_clone_lock locked the socket and set refcnt to 2 */
741 /* The below has to be done to allow calling inet_csk_destroy_sock */
742 sock_set_flag(sk, SOCK_DEAD);
743 percpu_counter_inc(sk->sk_prot->orphan_count);
744 inet_sk(sk)->inet_num = 0;
746 EXPORT_SYMBOL(inet_csk_prepare_forced_close);
748 int inet_csk_listen_start(struct sock *sk, const int nr_table_entries)
750 struct inet_sock *inet = inet_sk(sk);
751 struct inet_connection_sock *icsk = inet_csk(sk);
752 int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries);
757 sk->sk_max_ack_backlog = 0;
758 sk->sk_ack_backlog = 0;
759 inet_csk_delack_init(sk);
761 /* There is race window here: we announce ourselves listening,
762 * but this transition is still not validated by get_port().
763 * It is OK, because this socket enters to hash table only
764 * after validation is complete.
766 sk->sk_state = TCP_LISTEN;
767 if (!sk->sk_prot->get_port(sk, inet->inet_num)) {
768 inet->inet_sport = htons(inet->inet_num);
771 sk->sk_prot->hash(sk);
776 sk->sk_state = TCP_CLOSE;
777 __reqsk_queue_destroy(&icsk->icsk_accept_queue);
780 EXPORT_SYMBOL_GPL(inet_csk_listen_start);
783 * This routine closes sockets which have been at least partially
784 * opened, but not yet accepted.
786 void inet_csk_listen_stop(struct sock *sk)
788 struct inet_connection_sock *icsk = inet_csk(sk);
789 struct request_sock_queue *queue = &icsk->icsk_accept_queue;
790 struct request_sock *acc_req;
791 struct request_sock *req;
793 inet_csk_delete_keepalive_timer(sk);
795 /* make all the listen_opt local to us */
796 acc_req = reqsk_queue_yank_acceptq(queue);
798 /* Following specs, it would be better either to send FIN
799 * (and enter FIN-WAIT-1, it is normal close)
800 * or to send active reset (abort).
801 * Certainly, it is pretty dangerous while synflood, but it is
802 * bad justification for our negligence 8)
803 * To be honest, we are not able to make either
804 * of the variants now. --ANK
806 reqsk_queue_destroy(queue);
808 while ((req = acc_req) != NULL) {
809 struct sock *child = req->sk;
811 acc_req = req->dl_next;
815 WARN_ON(sock_owned_by_user(child));
818 sk->sk_prot->disconnect(child, O_NONBLOCK);
822 percpu_counter_inc(sk->sk_prot->orphan_count);
824 if (sk->sk_protocol == IPPROTO_TCP && tcp_rsk(req)->tfo_listener) {
825 BUG_ON(tcp_sk(child)->fastopen_rsk != req);
826 BUG_ON(sk != req->rsk_listener);
828 /* Paranoid, to prevent race condition if
829 * an inbound pkt destined for child is
830 * blocked by sock lock in tcp_v4_rcv().
831 * Also to satisfy an assertion in
832 * tcp_v4_destroy_sock().
834 tcp_sk(child)->fastopen_rsk = NULL;
836 inet_csk_destroy_sock(child);
838 bh_unlock_sock(child);
842 sk_acceptq_removed(sk);
845 if (queue->fastopenq != NULL) {
846 /* Free all the reqs queued in rskq_rst_head. */
847 spin_lock_bh(&queue->fastopenq->lock);
848 acc_req = queue->fastopenq->rskq_rst_head;
849 queue->fastopenq->rskq_rst_head = NULL;
850 spin_unlock_bh(&queue->fastopenq->lock);
851 while ((req = acc_req) != NULL) {
852 acc_req = req->dl_next;
856 WARN_ON(sk->sk_ack_backlog);
858 EXPORT_SYMBOL_GPL(inet_csk_listen_stop);
860 void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr)
862 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
863 const struct inet_sock *inet = inet_sk(sk);
865 sin->sin_family = AF_INET;
866 sin->sin_addr.s_addr = inet->inet_daddr;
867 sin->sin_port = inet->inet_dport;
869 EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr);
872 int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname,
873 char __user *optval, int __user *optlen)
875 const struct inet_connection_sock *icsk = inet_csk(sk);
877 if (icsk->icsk_af_ops->compat_getsockopt != NULL)
878 return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname,
880 return icsk->icsk_af_ops->getsockopt(sk, level, optname,
883 EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt);
885 int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname,
886 char __user *optval, unsigned int optlen)
888 const struct inet_connection_sock *icsk = inet_csk(sk);
890 if (icsk->icsk_af_ops->compat_setsockopt != NULL)
891 return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname,
893 return icsk->icsk_af_ops->setsockopt(sk, level, optname,
896 EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt);
899 static struct dst_entry *inet_csk_rebuild_route(struct sock *sk, struct flowi *fl)
901 const struct inet_sock *inet = inet_sk(sk);
902 const struct ip_options_rcu *inet_opt;
903 __be32 daddr = inet->inet_daddr;
908 inet_opt = rcu_dereference(inet->inet_opt);
909 if (inet_opt && inet_opt->opt.srr)
910 daddr = inet_opt->opt.faddr;
912 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr,
913 inet->inet_saddr, inet->inet_dport,
914 inet->inet_sport, sk->sk_protocol,
915 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if);
919 sk_setup_caps(sk, &rt->dst);
925 struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu)
927 struct dst_entry *dst = __sk_dst_check(sk, 0);
928 struct inet_sock *inet = inet_sk(sk);
931 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
935 dst->ops->update_pmtu(dst, sk, NULL, mtu);
937 dst = __sk_dst_check(sk, 0);
939 dst = inet_csk_rebuild_route(sk, &inet->cork.fl);
943 EXPORT_SYMBOL_GPL(inet_csk_update_pmtu);