1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68#include <linux/err.h>
69#include <linux/errno.h>
70#include <linux/types.h>
71#include <linux/socket.h>
72#include <linux/in.h>
73#include <linux/kernel.h>
74#include <linux/module.h>
75#include <linux/sched.h>
76#include <linux/timer.h>
77#include <linux/string.h>
78#include <linux/sockios.h>
79#include <linux/net.h>
80#include <linux/capability.h>
81#include <linux/fcntl.h>
82#include <linux/mm.h>
83#include <linux/interrupt.h>
84#include <linux/stat.h>
85#include <linux/init.h>
86#include <linux/poll.h>
87#include <linux/netfilter_ipv4.h>
88#include <linux/random.h>
89#include <linux/slab.h>
90
91#include <asm/uaccess.h>
92#include <asm/system.h>
93
94#include <linux/inet.h>
95#include <linux/igmp.h>
96#include <linux/inetdevice.h>
97#include <linux/netdevice.h>
98#include <net/checksum.h>
99#include <net/ip.h>
100#include <net/protocol.h>
101#include <net/arp.h>
102#include <net/route.h>
103#include <net/ip_fib.h>
104#include <net/inet_connection_sock.h>
105#include <net/tcp.h>
106#include <net/udp.h>
107#include <net/udplite.h>
108#include <net/ping.h>
109#include <linux/skbuff.h>
110#include <net/sock.h>
111#include <net/raw.h>
112#include <net/icmp.h>
113#include <net/ipip.h>
114#include <net/inet_common.h>
115#include <net/xfrm.h>
116#include <net/net_namespace.h>
117#ifdef CONFIG_IP_MROUTE
118#include <linux/mroute.h>
119#endif
120
121
122
123
124
125static struct list_head inetsw[SOCK_MAX];
126static DEFINE_SPINLOCK(inetsw_lock);
127
128struct ipv4_config ipv4_config;
129EXPORT_SYMBOL(ipv4_config);
130
131
132
133void inet_sock_destruct(struct sock *sk)
134{
135 struct inet_sock *inet = inet_sk(sk);
136
137 __skb_queue_purge(&sk->sk_receive_queue);
138 __skb_queue_purge(&sk->sk_error_queue);
139
140 sk_mem_reclaim(sk);
141
142 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
143 pr_err("Attempt to release TCP socket in state %d %p\n",
144 sk->sk_state, sk);
145 return;
146 }
147 if (!sock_flag(sk, SOCK_DEAD)) {
148 pr_err("Attempt to release alive inet socket %p\n", sk);
149 return;
150 }
151
152 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
153 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
154 WARN_ON(sk->sk_wmem_queued);
155 WARN_ON(sk->sk_forward_alloc);
156
157 kfree(rcu_dereference_protected(inet->inet_opt, 1));
158 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
159 sk_refcnt_debug_dec(sk);
160}
161EXPORT_SYMBOL(inet_sock_destruct);
162
163
164
165
166
167
168
169
170
171
172
173static int inet_autobind(struct sock *sk)
174{
175 struct inet_sock *inet;
176
177 lock_sock(sk);
178 inet = inet_sk(sk);
179 if (!inet->inet_num) {
180 if (sk->sk_prot->get_port(sk, 0)) {
181 release_sock(sk);
182 return -EAGAIN;
183 }
184 inet->inet_sport = htons(inet->inet_num);
185 }
186 release_sock(sk);
187 return 0;
188}
189
190
191
192
193int inet_listen(struct socket *sock, int backlog)
194{
195 struct sock *sk = sock->sk;
196 unsigned char old_state;
197 int err;
198
199 lock_sock(sk);
200
201 err = -EINVAL;
202 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
203 goto out;
204
205 old_state = sk->sk_state;
206 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
207 goto out;
208
209
210
211
212 if (old_state != TCP_LISTEN) {
213 err = inet_csk_listen_start(sk, backlog);
214 if (err)
215 goto out;
216 }
217 sk->sk_max_ack_backlog = backlog;
218 err = 0;
219
220out:
221 release_sock(sk);
222 return err;
223}
224EXPORT_SYMBOL(inet_listen);
225
226u32 inet_ehash_secret __read_mostly;
227EXPORT_SYMBOL(inet_ehash_secret);
228
229
230
231
232void build_ehash_secret(void)
233{
234 u32 rnd;
235
236 do {
237 get_random_bytes(&rnd, sizeof(rnd));
238 } while (rnd == 0);
239
240 cmpxchg(&inet_ehash_secret, 0, rnd);
241}
242EXPORT_SYMBOL(build_ehash_secret);
243
244static inline int inet_netns_ok(struct net *net, int protocol)
245{
246 int hash;
247 const struct net_protocol *ipprot;
248
249 if (net_eq(net, &init_net))
250 return 1;
251
252 hash = protocol & (MAX_INET_PROTOS - 1);
253 ipprot = rcu_dereference(inet_protos[hash]);
254
255 if (ipprot == NULL)
256
257 return 1;
258 return ipprot->netns_ok;
259}
260
261
262
263
264
265static int inet_create(struct net *net, struct socket *sock, int protocol,
266 int kern)
267{
268 struct sock *sk;
269 struct inet_protosw *answer;
270 struct inet_sock *inet;
271 struct proto *answer_prot;
272 unsigned char answer_flags;
273 char answer_no_check;
274 int try_loading_module = 0;
275 int err;
276
277 if (unlikely(!inet_ehash_secret))
278 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
279 build_ehash_secret();
280
281 sock->state = SS_UNCONNECTED;
282
283
284lookup_protocol:
285 err = -ESOCKTNOSUPPORT;
286 rcu_read_lock();
287 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
288
289 err = 0;
290
291 if (protocol == answer->protocol) {
292 if (protocol != IPPROTO_IP)
293 break;
294 } else {
295
296 if (IPPROTO_IP == protocol) {
297 protocol = answer->protocol;
298 break;
299 }
300 if (IPPROTO_IP == answer->protocol)
301 break;
302 }
303 err = -EPROTONOSUPPORT;
304 }
305
306 if (unlikely(err)) {
307 if (try_loading_module < 2) {
308 rcu_read_unlock();
309
310
311
312
313 if (++try_loading_module == 1)
314 request_module("net-pf-%d-proto-%d-type-%d",
315 PF_INET, protocol, sock->type);
316
317
318
319
320 else
321 request_module("net-pf-%d-proto-%d",
322 PF_INET, protocol);
323 goto lookup_protocol;
324 } else
325 goto out_rcu_unlock;
326 }
327
328 err = -EPERM;
329 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
330 goto out_rcu_unlock;
331
332 err = -EAFNOSUPPORT;
333 if (!inet_netns_ok(net, protocol))
334 goto out_rcu_unlock;
335
336 sock->ops = answer->ops;
337 answer_prot = answer->prot;
338 answer_no_check = answer->no_check;
339 answer_flags = answer->flags;
340 rcu_read_unlock();
341
342 WARN_ON(answer_prot->slab == NULL);
343
344 err = -ENOBUFS;
345 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
346 if (sk == NULL)
347 goto out;
348
349 err = 0;
350 sk->sk_no_check = answer_no_check;
351 if (INET_PROTOSW_REUSE & answer_flags)
352 sk->sk_reuse = 1;
353
354 inet = inet_sk(sk);
355 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
356
357 inet->nodefrag = 0;
358
359 if (SOCK_RAW == sock->type) {
360 inet->inet_num = protocol;
361 if (IPPROTO_RAW == protocol)
362 inet->hdrincl = 1;
363 }
364
365 if (ipv4_config.no_pmtu_disc)
366 inet->pmtudisc = IP_PMTUDISC_DONT;
367 else
368 inet->pmtudisc = IP_PMTUDISC_WANT;
369
370 inet->inet_id = 0;
371
372 sock_init_data(sock, sk);
373
374 sk->sk_destruct = inet_sock_destruct;
375 sk->sk_protocol = protocol;
376 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
377
378 inet->uc_ttl = -1;
379 inet->mc_loop = 1;
380 inet->mc_ttl = 1;
381 inet->mc_all = 1;
382 inet->mc_index = 0;
383 inet->mc_list = NULL;
384
385 sk_refcnt_debug_inc(sk);
386
387 if (inet->inet_num) {
388
389
390
391
392
393 inet->inet_sport = htons(inet->inet_num);
394
395 sk->sk_prot->hash(sk);
396 }
397
398 if (sk->sk_prot->init) {
399 err = sk->sk_prot->init(sk);
400 if (err)
401 sk_common_release(sk);
402 }
403out:
404 return err;
405out_rcu_unlock:
406 rcu_read_unlock();
407 goto out;
408}
409
410
411
412
413
414
415
416int inet_release(struct socket *sock)
417{
418 struct sock *sk = sock->sk;
419
420 if (sk) {
421 long timeout;
422
423 sock_rps_reset_flow(sk);
424
425
426 ip_mc_drop_socket(sk);
427
428
429
430
431
432
433
434
435 timeout = 0;
436 if (sock_flag(sk, SOCK_LINGER) &&
437 !(current->flags & PF_EXITING))
438 timeout = sk->sk_lingertime;
439 sock->sk = NULL;
440 sk->sk_prot->close(sk, timeout);
441 }
442 return 0;
443}
444EXPORT_SYMBOL(inet_release);
445
446
447int sysctl_ip_nonlocal_bind __read_mostly;
448EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
449
450int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
451{
452 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
453 struct sock *sk = sock->sk;
454 struct inet_sock *inet = inet_sk(sk);
455 unsigned short snum;
456 int chk_addr_ret;
457 int err;
458
459
460 if (sk->sk_prot->bind) {
461 err = sk->sk_prot->bind(sk, uaddr, addr_len);
462 goto out;
463 }
464 err = -EINVAL;
465 if (addr_len < sizeof(struct sockaddr_in))
466 goto out;
467
468 if (addr->sin_family != AF_INET) {
469 err = -EAFNOSUPPORT;
470 goto out;
471 }
472
473 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
474
475
476
477
478
479
480
481
482 err = -EADDRNOTAVAIL;
483 if (!sysctl_ip_nonlocal_bind &&
484 !(inet->freebind || inet->transparent) &&
485 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
486 chk_addr_ret != RTN_LOCAL &&
487 chk_addr_ret != RTN_MULTICAST &&
488 chk_addr_ret != RTN_BROADCAST)
489 goto out;
490
491 snum = ntohs(addr->sin_port);
492 err = -EACCES;
493 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
494 goto out;
495
496
497
498
499
500
501
502
503 lock_sock(sk);
504
505
506 err = -EINVAL;
507 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
508 goto out_release_sock;
509
510 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
511 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
512 inet->inet_saddr = 0;
513
514
515 if (sk->sk_prot->get_port(sk, snum)) {
516 inet->inet_saddr = inet->inet_rcv_saddr = 0;
517 err = -EADDRINUSE;
518 goto out_release_sock;
519 }
520
521 if (inet->inet_rcv_saddr)
522 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
523 if (snum)
524 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
525 inet->inet_sport = htons(inet->inet_num);
526 inet->inet_daddr = 0;
527 inet->inet_dport = 0;
528 sk_dst_reset(sk);
529 err = 0;
530out_release_sock:
531 release_sock(sk);
532out:
533 return err;
534}
535EXPORT_SYMBOL(inet_bind);
536
537int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
538 int addr_len, int flags)
539{
540 struct sock *sk = sock->sk;
541
542 if (addr_len < sizeof(uaddr->sa_family))
543 return -EINVAL;
544 if (uaddr->sa_family == AF_UNSPEC)
545 return sk->sk_prot->disconnect(sk, flags);
546
547 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
548 return -EAGAIN;
549 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
550}
551EXPORT_SYMBOL(inet_dgram_connect);
552
553static long inet_wait_for_connect(struct sock *sk, long timeo)
554{
555 DEFINE_WAIT(wait);
556
557 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
558
559
560
561
562
563
564 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
565 release_sock(sk);
566 timeo = schedule_timeout(timeo);
567 lock_sock(sk);
568 if (signal_pending(current) || !timeo)
569 break;
570 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
571 }
572 finish_wait(sk_sleep(sk), &wait);
573 return timeo;
574}
575
576
577
578
579
580int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
581 int addr_len, int flags)
582{
583 struct sock *sk = sock->sk;
584 int err;
585 long timeo;
586
587 if (addr_len < sizeof(uaddr->sa_family))
588 return -EINVAL;
589
590 lock_sock(sk);
591
592 if (uaddr->sa_family == AF_UNSPEC) {
593 err = sk->sk_prot->disconnect(sk, flags);
594 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
595 goto out;
596 }
597
598 switch (sock->state) {
599 default:
600 err = -EINVAL;
601 goto out;
602 case SS_CONNECTED:
603 err = -EISCONN;
604 goto out;
605 case SS_CONNECTING:
606 err = -EALREADY;
607
608 break;
609 case SS_UNCONNECTED:
610 err = -EISCONN;
611 if (sk->sk_state != TCP_CLOSE)
612 goto out;
613
614 err = sk->sk_prot->connect(sk, uaddr, addr_len);
615 if (err < 0)
616 goto out;
617
618 sock->state = SS_CONNECTING;
619
620
621
622
623
624 err = -EINPROGRESS;
625 break;
626 }
627
628 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
629
630 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
631
632 if (!timeo || !inet_wait_for_connect(sk, timeo))
633 goto out;
634
635 err = sock_intr_errno(timeo);
636 if (signal_pending(current))
637 goto out;
638 }
639
640
641
642
643 if (sk->sk_state == TCP_CLOSE)
644 goto sock_error;
645
646
647
648
649
650
651 sock->state = SS_CONNECTED;
652 err = 0;
653out:
654 release_sock(sk);
655 return err;
656
657sock_error:
658 err = sock_error(sk) ? : -ECONNABORTED;
659 sock->state = SS_UNCONNECTED;
660 if (sk->sk_prot->disconnect(sk, flags))
661 sock->state = SS_DISCONNECTING;
662 goto out;
663}
664EXPORT_SYMBOL(inet_stream_connect);
665
666
667
668
669
670int inet_accept(struct socket *sock, struct socket *newsock, int flags)
671{
672 struct sock *sk1 = sock->sk;
673 int err = -EINVAL;
674 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
675
676 if (!sk2)
677 goto do_err;
678
679 lock_sock(sk2);
680
681 sock_rps_record_flow(sk2);
682 WARN_ON(!((1 << sk2->sk_state) &
683 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)));
684
685 sock_graft(sk2, newsock);
686
687 newsock->state = SS_CONNECTED;
688 err = 0;
689 release_sock(sk2);
690do_err:
691 return err;
692}
693EXPORT_SYMBOL(inet_accept);
694
695
696
697
698
699int inet_getname(struct socket *sock, struct sockaddr *uaddr,
700 int *uaddr_len, int peer)
701{
702 struct sock *sk = sock->sk;
703 struct inet_sock *inet = inet_sk(sk);
704 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
705
706 sin->sin_family = AF_INET;
707 if (peer) {
708 if (!inet->inet_dport ||
709 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
710 peer == 1))
711 return -ENOTCONN;
712 sin->sin_port = inet->inet_dport;
713 sin->sin_addr.s_addr = inet->inet_daddr;
714 } else {
715 __be32 addr = inet->inet_rcv_saddr;
716 if (!addr)
717 addr = inet->inet_saddr;
718 sin->sin_port = inet->inet_sport;
719 sin->sin_addr.s_addr = addr;
720 }
721 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
722 *uaddr_len = sizeof(*sin);
723 return 0;
724}
725EXPORT_SYMBOL(inet_getname);
726
727int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
728 size_t size)
729{
730 struct sock *sk = sock->sk;
731
732 sock_rps_record_flow(sk);
733
734
735 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
736 inet_autobind(sk))
737 return -EAGAIN;
738
739 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
740}
741EXPORT_SYMBOL(inet_sendmsg);
742
743ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
744 size_t size, int flags)
745{
746 struct sock *sk = sock->sk;
747
748 sock_rps_record_flow(sk);
749
750
751 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
752 inet_autobind(sk))
753 return -EAGAIN;
754
755 if (sk->sk_prot->sendpage)
756 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
757 return sock_no_sendpage(sock, page, offset, size, flags);
758}
759EXPORT_SYMBOL(inet_sendpage);
760
761int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
762 size_t size, int flags)
763{
764 struct sock *sk = sock->sk;
765 int addr_len = 0;
766 int err;
767
768 sock_rps_record_flow(sk);
769
770 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
771 flags & ~MSG_DONTWAIT, &addr_len);
772 if (err >= 0)
773 msg->msg_namelen = addr_len;
774 return err;
775}
776EXPORT_SYMBOL(inet_recvmsg);
777
778int inet_shutdown(struct socket *sock, int how)
779{
780 struct sock *sk = sock->sk;
781 int err = 0;
782
783
784
785
786 how++;
787
788
789 if ((how & ~SHUTDOWN_MASK) || !how)
790 return -EINVAL;
791
792 lock_sock(sk);
793 if (sock->state == SS_CONNECTING) {
794 if ((1 << sk->sk_state) &
795 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
796 sock->state = SS_DISCONNECTING;
797 else
798 sock->state = SS_CONNECTED;
799 }
800
801 switch (sk->sk_state) {
802 case TCP_CLOSE:
803 err = -ENOTCONN;
804
805
806 default:
807 sk->sk_shutdown |= how;
808 if (sk->sk_prot->shutdown)
809 sk->sk_prot->shutdown(sk, how);
810 break;
811
812
813
814
815
816 case TCP_LISTEN:
817 if (!(how & RCV_SHUTDOWN))
818 break;
819
820 case TCP_SYN_SENT:
821 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
822 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
823 break;
824 }
825
826
827 sk->sk_state_change(sk);
828 release_sock(sk);
829 return err;
830}
831EXPORT_SYMBOL(inet_shutdown);
832
833
834
835
836
837
838
839
840
841
842
843int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
844{
845 struct sock *sk = sock->sk;
846 int err = 0;
847 struct net *net = sock_net(sk);
848
849 switch (cmd) {
850 case SIOCGSTAMP:
851 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
852 break;
853 case SIOCGSTAMPNS:
854 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
855 break;
856 case SIOCADDRT:
857 case SIOCDELRT:
858 case SIOCRTMSG:
859 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
860 break;
861 case SIOCDARP:
862 case SIOCGARP:
863 case SIOCSARP:
864 err = arp_ioctl(net, cmd, (void __user *)arg);
865 break;
866 case SIOCGIFADDR:
867 case SIOCSIFADDR:
868 case SIOCGIFBRDADDR:
869 case SIOCSIFBRDADDR:
870 case SIOCGIFNETMASK:
871 case SIOCSIFNETMASK:
872 case SIOCGIFDSTADDR:
873 case SIOCSIFDSTADDR:
874 case SIOCSIFPFLAGS:
875 case SIOCGIFPFLAGS:
876 case SIOCSIFFLAGS:
877 err = devinet_ioctl(net, cmd, (void __user *)arg);
878 break;
879 default:
880 if (sk->sk_prot->ioctl)
881 err = sk->sk_prot->ioctl(sk, cmd, arg);
882 else
883 err = -ENOIOCTLCMD;
884 break;
885 }
886 return err;
887}
888EXPORT_SYMBOL(inet_ioctl);
889
890#ifdef CONFIG_COMPAT
891int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
892{
893 struct sock *sk = sock->sk;
894 int err = -ENOIOCTLCMD;
895
896 if (sk->sk_prot->compat_ioctl)
897 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
898
899 return err;
900}
901#endif
902
903const struct proto_ops inet_stream_ops = {
904 .family = PF_INET,
905 .owner = THIS_MODULE,
906 .release = inet_release,
907 .bind = inet_bind,
908 .connect = inet_stream_connect,
909 .socketpair = sock_no_socketpair,
910 .accept = inet_accept,
911 .getname = inet_getname,
912 .poll = tcp_poll,
913 .ioctl = inet_ioctl,
914 .listen = inet_listen,
915 .shutdown = inet_shutdown,
916 .setsockopt = sock_common_setsockopt,
917 .getsockopt = sock_common_getsockopt,
918 .sendmsg = inet_sendmsg,
919 .recvmsg = inet_recvmsg,
920 .mmap = sock_no_mmap,
921 .sendpage = inet_sendpage,
922 .splice_read = tcp_splice_read,
923#ifdef CONFIG_COMPAT
924 .compat_setsockopt = compat_sock_common_setsockopt,
925 .compat_getsockopt = compat_sock_common_getsockopt,
926 .compat_ioctl = inet_compat_ioctl,
927#endif
928};
929EXPORT_SYMBOL(inet_stream_ops);
930
931const struct proto_ops inet_dgram_ops = {
932 .family = PF_INET,
933 .owner = THIS_MODULE,
934 .release = inet_release,
935 .bind = inet_bind,
936 .connect = inet_dgram_connect,
937 .socketpair = sock_no_socketpair,
938 .accept = sock_no_accept,
939 .getname = inet_getname,
940 .poll = udp_poll,
941 .ioctl = inet_ioctl,
942 .listen = sock_no_listen,
943 .shutdown = inet_shutdown,
944 .setsockopt = sock_common_setsockopt,
945 .getsockopt = sock_common_getsockopt,
946 .sendmsg = inet_sendmsg,
947 .recvmsg = inet_recvmsg,
948 .mmap = sock_no_mmap,
949 .sendpage = inet_sendpage,
950#ifdef CONFIG_COMPAT
951 .compat_setsockopt = compat_sock_common_setsockopt,
952 .compat_getsockopt = compat_sock_common_getsockopt,
953 .compat_ioctl = inet_compat_ioctl,
954#endif
955};
956EXPORT_SYMBOL(inet_dgram_ops);
957
958
959
960
961
962static const struct proto_ops inet_sockraw_ops = {
963 .family = PF_INET,
964 .owner = THIS_MODULE,
965 .release = inet_release,
966 .bind = inet_bind,
967 .connect = inet_dgram_connect,
968 .socketpair = sock_no_socketpair,
969 .accept = sock_no_accept,
970 .getname = inet_getname,
971 .poll = datagram_poll,
972 .ioctl = inet_ioctl,
973 .listen = sock_no_listen,
974 .shutdown = inet_shutdown,
975 .setsockopt = sock_common_setsockopt,
976 .getsockopt = sock_common_getsockopt,
977 .sendmsg = inet_sendmsg,
978 .recvmsg = inet_recvmsg,
979 .mmap = sock_no_mmap,
980 .sendpage = inet_sendpage,
981#ifdef CONFIG_COMPAT
982 .compat_setsockopt = compat_sock_common_setsockopt,
983 .compat_getsockopt = compat_sock_common_getsockopt,
984 .compat_ioctl = inet_compat_ioctl,
985#endif
986};
987
988static const struct net_proto_family inet_family_ops = {
989 .family = PF_INET,
990 .create = inet_create,
991 .owner = THIS_MODULE,
992};
993
994
995
996
997static struct inet_protosw inetsw_array[] =
998{
999 {
1000 .type = SOCK_STREAM,
1001 .protocol = IPPROTO_TCP,
1002 .prot = &tcp_prot,
1003 .ops = &inet_stream_ops,
1004 .no_check = 0,
1005 .flags = INET_PROTOSW_PERMANENT |
1006 INET_PROTOSW_ICSK,
1007 },
1008
1009 {
1010 .type = SOCK_DGRAM,
1011 .protocol = IPPROTO_UDP,
1012 .prot = &udp_prot,
1013 .ops = &inet_dgram_ops,
1014 .no_check = UDP_CSUM_DEFAULT,
1015 .flags = INET_PROTOSW_PERMANENT,
1016 },
1017
1018 {
1019 .type = SOCK_DGRAM,
1020 .protocol = IPPROTO_ICMP,
1021 .prot = &ping_prot,
1022 .ops = &inet_dgram_ops,
1023 .no_check = UDP_CSUM_DEFAULT,
1024 .flags = INET_PROTOSW_REUSE,
1025 },
1026
1027 {
1028 .type = SOCK_RAW,
1029 .protocol = IPPROTO_IP,
1030 .prot = &raw_prot,
1031 .ops = &inet_sockraw_ops,
1032 .no_check = UDP_CSUM_DEFAULT,
1033 .flags = INET_PROTOSW_REUSE,
1034 }
1035};
1036
1037#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
1038
1039void inet_register_protosw(struct inet_protosw *p)
1040{
1041 struct list_head *lh;
1042 struct inet_protosw *answer;
1043 int protocol = p->protocol;
1044 struct list_head *last_perm;
1045
1046 spin_lock_bh(&inetsw_lock);
1047
1048 if (p->type >= SOCK_MAX)
1049 goto out_illegal;
1050
1051
1052 answer = NULL;
1053 last_perm = &inetsw[p->type];
1054 list_for_each(lh, &inetsw[p->type]) {
1055 answer = list_entry(lh, struct inet_protosw, list);
1056
1057
1058 if (INET_PROTOSW_PERMANENT & answer->flags) {
1059 if (protocol == answer->protocol)
1060 break;
1061 last_perm = lh;
1062 }
1063
1064 answer = NULL;
1065 }
1066 if (answer)
1067 goto out_permanent;
1068
1069
1070
1071
1072
1073
1074
1075 list_add_rcu(&p->list, last_perm);
1076out:
1077 spin_unlock_bh(&inetsw_lock);
1078
1079 return;
1080
1081out_permanent:
1082 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1083 protocol);
1084 goto out;
1085
1086out_illegal:
1087 printk(KERN_ERR
1088 "Ignoring attempt to register invalid socket type %d.\n",
1089 p->type);
1090 goto out;
1091}
1092EXPORT_SYMBOL(inet_register_protosw);
1093
1094void inet_unregister_protosw(struct inet_protosw *p)
1095{
1096 if (INET_PROTOSW_PERMANENT & p->flags) {
1097 printk(KERN_ERR
1098 "Attempt to unregister permanent protocol %d.\n",
1099 p->protocol);
1100 } else {
1101 spin_lock_bh(&inetsw_lock);
1102 list_del_rcu(&p->list);
1103 spin_unlock_bh(&inetsw_lock);
1104
1105 synchronize_net();
1106 }
1107}
1108EXPORT_SYMBOL(inet_unregister_protosw);
1109
1110
1111
1112
1113
1114int sysctl_ip_dynaddr __read_mostly;
1115
1116static int inet_sk_reselect_saddr(struct sock *sk)
1117{
1118 struct inet_sock *inet = inet_sk(sk);
1119 __be32 old_saddr = inet->inet_saddr;
1120 __be32 daddr = inet->inet_daddr;
1121 struct flowi4 *fl4;
1122 struct rtable *rt;
1123 __be32 new_saddr;
1124 struct ip_options_rcu *inet_opt;
1125
1126 inet_opt = rcu_dereference_protected(inet->inet_opt,
1127 sock_owned_by_user(sk));
1128 if (inet_opt && inet_opt->opt.srr)
1129 daddr = inet_opt->opt.faddr;
1130
1131
1132 fl4 = &inet->cork.fl.u.ip4;
1133 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
1134 sk->sk_bound_dev_if, sk->sk_protocol,
1135 inet->inet_sport, inet->inet_dport, sk, false);
1136 if (IS_ERR(rt))
1137 return PTR_ERR(rt);
1138
1139 sk_setup_caps(sk, &rt->dst);
1140
1141 new_saddr = fl4->saddr;
1142
1143 if (new_saddr == old_saddr)
1144 return 0;
1145
1146 if (sysctl_ip_dynaddr > 1) {
1147 printk(KERN_INFO "%s(): shifting inet->saddr from %pI4 to %pI4\n",
1148 __func__, &old_saddr, &new_saddr);
1149 }
1150
1151 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161 __sk_prot_rehash(sk);
1162 return 0;
1163}
1164
1165int inet_sk_rebuild_header(struct sock *sk)
1166{
1167 struct inet_sock *inet = inet_sk(sk);
1168 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1169 __be32 daddr;
1170 struct ip_options_rcu *inet_opt;
1171 struct flowi4 *fl4;
1172 int err;
1173
1174
1175 if (rt)
1176 return 0;
1177
1178
1179 rcu_read_lock();
1180 inet_opt = rcu_dereference(inet->inet_opt);
1181 daddr = inet->inet_daddr;
1182 if (inet_opt && inet_opt->opt.srr)
1183 daddr = inet_opt->opt.faddr;
1184 rcu_read_unlock();
1185 fl4 = &inet->cork.fl.u.ip4;
1186 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
1187 inet->inet_dport, inet->inet_sport,
1188 sk->sk_protocol, RT_CONN_FLAGS(sk),
1189 sk->sk_bound_dev_if);
1190 if (!IS_ERR(rt)) {
1191 err = 0;
1192 sk_setup_caps(sk, &rt->dst);
1193 } else {
1194 err = PTR_ERR(rt);
1195
1196
1197 sk->sk_route_caps = 0;
1198
1199
1200
1201
1202 if (!sysctl_ip_dynaddr ||
1203 sk->sk_state != TCP_SYN_SENT ||
1204 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1205 (err = inet_sk_reselect_saddr(sk)) != 0)
1206 sk->sk_err_soft = -err;
1207 }
1208
1209 return err;
1210}
1211EXPORT_SYMBOL(inet_sk_rebuild_header);
1212
1213static int inet_gso_send_check(struct sk_buff *skb)
1214{
1215 const struct iphdr *iph;
1216 const struct net_protocol *ops;
1217 int proto;
1218 int ihl;
1219 int err = -EINVAL;
1220
1221 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1222 goto out;
1223
1224 iph = ip_hdr(skb);
1225 ihl = iph->ihl * 4;
1226 if (ihl < sizeof(*iph))
1227 goto out;
1228
1229 if (unlikely(!pskb_may_pull(skb, ihl)))
1230 goto out;
1231
1232 __skb_pull(skb, ihl);
1233 skb_reset_transport_header(skb);
1234 iph = ip_hdr(skb);
1235 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1236 err = -EPROTONOSUPPORT;
1237
1238 rcu_read_lock();
1239 ops = rcu_dereference(inet_protos[proto]);
1240 if (likely(ops && ops->gso_send_check))
1241 err = ops->gso_send_check(skb);
1242 rcu_read_unlock();
1243
1244out:
1245 return err;
1246}
1247
1248static struct sk_buff *inet_gso_segment(struct sk_buff *skb, u32 features)
1249{
1250 struct sk_buff *segs = ERR_PTR(-EINVAL);
1251 struct iphdr *iph;
1252 const struct net_protocol *ops;
1253 int proto;
1254 int ihl;
1255 int id;
1256 unsigned int offset = 0;
1257
1258 if (!(features & NETIF_F_V4_CSUM))
1259 features &= ~NETIF_F_SG;
1260
1261 if (unlikely(skb_shinfo(skb)->gso_type &
1262 ~(SKB_GSO_TCPV4 |
1263 SKB_GSO_UDP |
1264 SKB_GSO_DODGY |
1265 SKB_GSO_TCP_ECN |
1266 0)))
1267 goto out;
1268
1269 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1270 goto out;
1271
1272 iph = ip_hdr(skb);
1273 ihl = iph->ihl * 4;
1274 if (ihl < sizeof(*iph))
1275 goto out;
1276
1277 if (unlikely(!pskb_may_pull(skb, ihl)))
1278 goto out;
1279
1280 __skb_pull(skb, ihl);
1281 skb_reset_transport_header(skb);
1282 iph = ip_hdr(skb);
1283 id = ntohs(iph->id);
1284 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1285 segs = ERR_PTR(-EPROTONOSUPPORT);
1286
1287 rcu_read_lock();
1288 ops = rcu_dereference(inet_protos[proto]);
1289 if (likely(ops && ops->gso_segment))
1290 segs = ops->gso_segment(skb, features);
1291 rcu_read_unlock();
1292
1293 if (!segs || IS_ERR(segs))
1294 goto out;
1295
1296 skb = segs;
1297 do {
1298 iph = ip_hdr(skb);
1299 if (proto == IPPROTO_UDP) {
1300 iph->id = htons(id);
1301 iph->frag_off = htons(offset >> 3);
1302 if (skb->next != NULL)
1303 iph->frag_off |= htons(IP_MF);
1304 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1305 } else
1306 iph->id = htons(id++);
1307 iph->tot_len = htons(skb->len - skb->mac_len);
1308 iph->check = 0;
1309 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
1310 } while ((skb = skb->next));
1311
1312out:
1313 return segs;
1314}
1315
1316static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1317 struct sk_buff *skb)
1318{
1319 const struct net_protocol *ops;
1320 struct sk_buff **pp = NULL;
1321 struct sk_buff *p;
1322 const struct iphdr *iph;
1323 unsigned int hlen;
1324 unsigned int off;
1325 unsigned int id;
1326 int flush = 1;
1327 int proto;
1328
1329 off = skb_gro_offset(skb);
1330 hlen = off + sizeof(*iph);
1331 iph = skb_gro_header_fast(skb, off);
1332 if (skb_gro_header_hard(skb, hlen)) {
1333 iph = skb_gro_header_slow(skb, hlen, off);
1334 if (unlikely(!iph))
1335 goto out;
1336 }
1337
1338 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1339
1340 rcu_read_lock();
1341 ops = rcu_dereference(inet_protos[proto]);
1342 if (!ops || !ops->gro_receive)
1343 goto out_unlock;
1344
1345 if (*(u8 *)iph != 0x45)
1346 goto out_unlock;
1347
1348 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1349 goto out_unlock;
1350
1351 id = ntohl(*(__be32 *)&iph->id);
1352 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
1353 id >>= 16;
1354
1355 for (p = *head; p; p = p->next) {
1356 struct iphdr *iph2;
1357
1358 if (!NAPI_GRO_CB(p)->same_flow)
1359 continue;
1360
1361 iph2 = ip_hdr(p);
1362
1363 if ((iph->protocol ^ iph2->protocol) |
1364 (iph->tos ^ iph2->tos) |
1365 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1366 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
1367 NAPI_GRO_CB(p)->same_flow = 0;
1368 continue;
1369 }
1370
1371
1372 NAPI_GRO_CB(p)->flush |=
1373 (iph->ttl ^ iph2->ttl) |
1374 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
1375
1376 NAPI_GRO_CB(p)->flush |= flush;
1377 }
1378
1379 NAPI_GRO_CB(skb)->flush |= flush;
1380 skb_gro_pull(skb, sizeof(*iph));
1381 skb_set_transport_header(skb, skb_gro_offset(skb));
1382
1383 pp = ops->gro_receive(head, skb);
1384
1385out_unlock:
1386 rcu_read_unlock();
1387
1388out:
1389 NAPI_GRO_CB(skb)->flush |= flush;
1390
1391 return pp;
1392}
1393
1394static int inet_gro_complete(struct sk_buff *skb)
1395{
1396 const struct net_protocol *ops;
1397 struct iphdr *iph = ip_hdr(skb);
1398 int proto = iph->protocol & (MAX_INET_PROTOS - 1);
1399 int err = -ENOSYS;
1400 __be16 newlen = htons(skb->len - skb_network_offset(skb));
1401
1402 csum_replace2(&iph->check, iph->tot_len, newlen);
1403 iph->tot_len = newlen;
1404
1405 rcu_read_lock();
1406 ops = rcu_dereference(inet_protos[proto]);
1407 if (WARN_ON(!ops || !ops->gro_complete))
1408 goto out_unlock;
1409
1410 err = ops->gro_complete(skb);
1411
1412out_unlock:
1413 rcu_read_unlock();
1414
1415 return err;
1416}
1417
1418int inet_ctl_sock_create(struct sock **sk, unsigned short family,
1419 unsigned short type, unsigned char protocol,
1420 struct net *net)
1421{
1422 struct socket *sock;
1423 int rc = sock_create_kern(family, type, protocol, &sock);
1424
1425 if (rc == 0) {
1426 *sk = sock->sk;
1427 (*sk)->sk_allocation = GFP_ATOMIC;
1428
1429
1430
1431
1432 (*sk)->sk_prot->unhash(*sk);
1433
1434 sk_change_net(*sk, net);
1435 }
1436 return rc;
1437}
1438EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1439
1440unsigned long snmp_fold_field(void __percpu *mib[], int offt)
1441{
1442 unsigned long res = 0;
1443 int i;
1444
1445 for_each_possible_cpu(i) {
1446 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1447 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1448 }
1449 return res;
1450}
1451EXPORT_SYMBOL_GPL(snmp_fold_field);
1452
1453#if BITS_PER_LONG==32
1454
1455u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1456{
1457 u64 res = 0;
1458 int cpu;
1459
1460 for_each_possible_cpu(cpu) {
1461 void *bhptr, *userptr;
1462 struct u64_stats_sync *syncp;
1463 u64 v_bh, v_user;
1464 unsigned int start;
1465
1466
1467 bhptr = per_cpu_ptr(SNMP_STAT_BHPTR(mib), cpu);
1468 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1469 do {
1470 start = u64_stats_fetch_begin_bh(syncp);
1471 v_bh = *(((u64 *) bhptr) + offt);
1472 } while (u64_stats_fetch_retry_bh(syncp, start));
1473
1474
1475 userptr = per_cpu_ptr(SNMP_STAT_USRPTR(mib), cpu);
1476 syncp = (struct u64_stats_sync *)(userptr + syncp_offset);
1477 do {
1478 start = u64_stats_fetch_begin(syncp);
1479 v_user = *(((u64 *) userptr) + offt);
1480 } while (u64_stats_fetch_retry(syncp, start));
1481
1482 res += v_bh + v_user;
1483 }
1484 return res;
1485}
1486EXPORT_SYMBOL_GPL(snmp_fold_field64);
1487#endif
1488
1489int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
1490{
1491 BUG_ON(ptr == NULL);
1492 ptr[0] = __alloc_percpu(mibsize, align);
1493 if (!ptr[0])
1494 goto err0;
1495 ptr[1] = __alloc_percpu(mibsize, align);
1496 if (!ptr[1])
1497 goto err1;
1498 return 0;
1499err1:
1500 free_percpu(ptr[0]);
1501 ptr[0] = NULL;
1502err0:
1503 return -ENOMEM;
1504}
1505EXPORT_SYMBOL_GPL(snmp_mib_init);
1506
1507void snmp_mib_free(void __percpu *ptr[2])
1508{
1509 BUG_ON(ptr == NULL);
1510 free_percpu(ptr[0]);
1511 free_percpu(ptr[1]);
1512 ptr[0] = ptr[1] = NULL;
1513}
1514EXPORT_SYMBOL_GPL(snmp_mib_free);
1515
1516#ifdef CONFIG_IP_MULTICAST
1517static const struct net_protocol igmp_protocol = {
1518 .handler = igmp_rcv,
1519 .netns_ok = 1,
1520};
1521#endif
1522
1523static const struct net_protocol tcp_protocol = {
1524 .handler = tcp_v4_rcv,
1525 .err_handler = tcp_v4_err,
1526 .gso_send_check = tcp_v4_gso_send_check,
1527 .gso_segment = tcp_tso_segment,
1528 .gro_receive = tcp4_gro_receive,
1529 .gro_complete = tcp4_gro_complete,
1530 .no_policy = 1,
1531 .netns_ok = 1,
1532};
1533
1534static const struct net_protocol udp_protocol = {
1535 .handler = udp_rcv,
1536 .err_handler = udp_err,
1537 .gso_send_check = udp4_ufo_send_check,
1538 .gso_segment = udp4_ufo_fragment,
1539 .no_policy = 1,
1540 .netns_ok = 1,
1541};
1542
1543static const struct net_protocol icmp_protocol = {
1544 .handler = icmp_rcv,
1545 .err_handler = ping_err,
1546 .no_policy = 1,
1547 .netns_ok = 1,
1548};
1549
1550static __net_init int ipv4_mib_init_net(struct net *net)
1551{
1552 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
1553 sizeof(struct tcp_mib),
1554 __alignof__(struct tcp_mib)) < 0)
1555 goto err_tcp_mib;
1556 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
1557 sizeof(struct ipstats_mib),
1558 __alignof__(struct ipstats_mib)) < 0)
1559 goto err_ip_mib;
1560 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
1561 sizeof(struct linux_mib),
1562 __alignof__(struct linux_mib)) < 0)
1563 goto err_net_mib;
1564 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
1565 sizeof(struct udp_mib),
1566 __alignof__(struct udp_mib)) < 0)
1567 goto err_udp_mib;
1568 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
1569 sizeof(struct udp_mib),
1570 __alignof__(struct udp_mib)) < 0)
1571 goto err_udplite_mib;
1572 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
1573 sizeof(struct icmp_mib),
1574 __alignof__(struct icmp_mib)) < 0)
1575 goto err_icmp_mib;
1576 if (snmp_mib_init((void __percpu **)net->mib.icmpmsg_statistics,
1577 sizeof(struct icmpmsg_mib),
1578 __alignof__(struct icmpmsg_mib)) < 0)
1579 goto err_icmpmsg_mib;
1580
1581 tcp_mib_init(net);
1582 return 0;
1583
1584err_icmpmsg_mib:
1585 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1586err_icmp_mib:
1587 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1588err_udplite_mib:
1589 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1590err_udp_mib:
1591 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1592err_net_mib:
1593 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1594err_ip_mib:
1595 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
1596err_tcp_mib:
1597 return -ENOMEM;
1598}
1599
1600static __net_exit void ipv4_mib_exit_net(struct net *net)
1601{
1602 snmp_mib_free((void __percpu **)net->mib.icmpmsg_statistics);
1603 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1604 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1605 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1606 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1607 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1608 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
1609}
1610
1611static __net_initdata struct pernet_operations ipv4_mib_ops = {
1612 .init = ipv4_mib_init_net,
1613 .exit = ipv4_mib_exit_net,
1614};
1615
1616static int __init init_ipv4_mibs(void)
1617{
1618 return register_pernet_subsys(&ipv4_mib_ops);
1619}
1620
1621static int ipv4_proc_init(void);
1622
1623
1624
1625
1626
1627static struct packet_type ip_packet_type __read_mostly = {
1628 .type = cpu_to_be16(ETH_P_IP),
1629 .func = ip_rcv,
1630 .gso_send_check = inet_gso_send_check,
1631 .gso_segment = inet_gso_segment,
1632 .gro_receive = inet_gro_receive,
1633 .gro_complete = inet_gro_complete,
1634};
1635
1636static int __init inet_init(void)
1637{
1638 struct sk_buff *dummy_skb;
1639 struct inet_protosw *q;
1640 struct list_head *r;
1641 int rc = -EINVAL;
1642
1643 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
1644
1645 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1646 if (!sysctl_local_reserved_ports)
1647 goto out;
1648
1649 rc = proto_register(&tcp_prot, 1);
1650 if (rc)
1651 goto out_free_reserved_ports;
1652
1653 rc = proto_register(&udp_prot, 1);
1654 if (rc)
1655 goto out_unregister_tcp_proto;
1656
1657 rc = proto_register(&raw_prot, 1);
1658 if (rc)
1659 goto out_unregister_udp_proto;
1660
1661 rc = proto_register(&ping_prot, 1);
1662 if (rc)
1663 goto out_unregister_raw_proto;
1664
1665
1666
1667
1668
1669 (void)sock_register(&inet_family_ops);
1670
1671#ifdef CONFIG_SYSCTL
1672 ip_static_sysctl_init();
1673#endif
1674
1675
1676
1677
1678
1679 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1680 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1681 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1682 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1683 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1684 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1685#ifdef CONFIG_IP_MULTICAST
1686 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1687 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1688#endif
1689
1690
1691 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1692 INIT_LIST_HEAD(r);
1693
1694 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1695 inet_register_protosw(q);
1696
1697
1698
1699
1700
1701 arp_init();
1702
1703
1704
1705
1706
1707 ip_init();
1708
1709 tcp_v4_init();
1710
1711
1712 tcp_init();
1713
1714
1715 udp_init();
1716
1717
1718 udplite4_register();
1719
1720 ping_init();
1721
1722
1723
1724
1725
1726 if (icmp_init() < 0)
1727 panic("Failed to create the ICMP control socket.\n");
1728
1729
1730
1731
1732#if defined(CONFIG_IP_MROUTE)
1733 if (ip_mr_init())
1734 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
1735#endif
1736
1737
1738
1739
1740 if (init_ipv4_mibs())
1741 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
1742
1743 ipv4_proc_init();
1744
1745 ipfrag_init();
1746
1747 dev_add_pack(&ip_packet_type);
1748
1749 rc = 0;
1750out:
1751 return rc;
1752out_unregister_raw_proto:
1753 proto_unregister(&raw_prot);
1754out_unregister_udp_proto:
1755 proto_unregister(&udp_prot);
1756out_unregister_tcp_proto:
1757 proto_unregister(&tcp_prot);
1758out_free_reserved_ports:
1759 kfree(sysctl_local_reserved_ports);
1760 goto out;
1761}
1762
1763fs_initcall(inet_init);
1764
1765
1766
1767#ifdef CONFIG_PROC_FS
1768static int __init ipv4_proc_init(void)
1769{
1770 int rc = 0;
1771
1772 if (raw_proc_init())
1773 goto out_raw;
1774 if (tcp4_proc_init())
1775 goto out_tcp;
1776 if (udp4_proc_init())
1777 goto out_udp;
1778 if (ping_proc_init())
1779 goto out_ping;
1780 if (ip_misc_proc_init())
1781 goto out_misc;
1782out:
1783 return rc;
1784out_misc:
1785 ping_proc_exit();
1786out_ping:
1787 udp4_proc_exit();
1788out_udp:
1789 tcp4_proc_exit();
1790out_tcp:
1791 raw_proc_exit();
1792out_raw:
1793 rc = -ENOMEM;
1794 goto out;
1795}
1796
1797#else
1798static int __init ipv4_proc_init(void)
1799{
1800 return 0;
1801}
1802#endif
1803
1804MODULE_ALIAS_NETPROTO(PF_INET);
1805
1806