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
90#include <asm/uaccess.h>
91#include <asm/system.h>
92
93#include <linux/inet.h>
94#include <linux/igmp.h>
95#include <linux/inetdevice.h>
96#include <linux/netdevice.h>
97#include <net/checksum.h>
98#include <net/ip.h>
99#include <net/protocol.h>
100#include <net/arp.h>
101#include <net/route.h>
102#include <net/ip_fib.h>
103#include <net/inet_connection_sock.h>
104#include <net/tcp.h>
105#include <net/udp.h>
106#include <net/udplite.h>
107#include <linux/skbuff.h>
108#include <net/sock.h>
109#include <net/raw.h>
110#include <net/icmp.h>
111#include <net/ipip.h>
112#include <net/inet_common.h>
113#include <net/xfrm.h>
114#include <net/net_namespace.h>
115#ifdef CONFIG_IP_MROUTE
116#include <linux/mroute.h>
117#endif
118
119
120
121
122
123static struct list_head inetsw[SOCK_MAX];
124static DEFINE_SPINLOCK(inetsw_lock);
125
126struct ipv4_config ipv4_config;
127EXPORT_SYMBOL(ipv4_config);
128
129
130
131void inet_sock_destruct(struct sock *sk)
132{
133 struct inet_sock *inet = inet_sk(sk);
134
135 __skb_queue_purge(&sk->sk_receive_queue);
136 __skb_queue_purge(&sk->sk_error_queue);
137
138 sk_mem_reclaim(sk);
139
140 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
141 pr_err("Attempt to release TCP socket in state %d %p\n",
142 sk->sk_state, sk);
143 return;
144 }
145 if (!sock_flag(sk, SOCK_DEAD)) {
146 pr_err("Attempt to release alive inet socket %p\n", sk);
147 return;
148 }
149
150 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
151 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
152 WARN_ON(sk->sk_wmem_queued);
153 WARN_ON(sk->sk_forward_alloc);
154
155 kfree(inet->opt);
156 dst_release(sk->sk_dst_cache);
157 sk_refcnt_debug_dec(sk);
158}
159EXPORT_SYMBOL(inet_sock_destruct);
160
161
162
163
164
165
166
167
168
169
170
171static int inet_autobind(struct sock *sk)
172{
173 struct inet_sock *inet;
174
175 lock_sock(sk);
176 inet = inet_sk(sk);
177 if (!inet->num) {
178 if (sk->sk_prot->get_port(sk, 0)) {
179 release_sock(sk);
180 return -EAGAIN;
181 }
182 inet->sport = htons(inet->num);
183 }
184 release_sock(sk);
185 return 0;
186}
187
188
189
190
191int inet_listen(struct socket *sock, int backlog)
192{
193 struct sock *sk = sock->sk;
194 unsigned char old_state;
195 int err;
196
197 lock_sock(sk);
198
199 err = -EINVAL;
200 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
201 goto out;
202
203 old_state = sk->sk_state;
204 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
205 goto out;
206
207
208
209
210 if (old_state != TCP_LISTEN) {
211 err = inet_csk_listen_start(sk, backlog);
212 if (err)
213 goto out;
214 }
215 sk->sk_max_ack_backlog = backlog;
216 err = 0;
217
218out:
219 release_sock(sk);
220 return err;
221}
222EXPORT_SYMBOL(inet_listen);
223
224u32 inet_ehash_secret __read_mostly;
225EXPORT_SYMBOL(inet_ehash_secret);
226
227
228
229
230
231void build_ehash_secret(void)
232{
233 u32 rnd;
234 do {
235 get_random_bytes(&rnd, sizeof(rnd));
236 } while (rnd == 0);
237 spin_lock_bh(&inetsw_lock);
238 if (!inet_ehash_secret)
239 inet_ehash_secret = rnd;
240 spin_unlock_bh(&inetsw_lock);
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{
267 struct sock *sk;
268 struct inet_protosw *answer;
269 struct inet_sock *inet;
270 struct proto *answer_prot;
271 unsigned char answer_flags;
272 char answer_no_check;
273 int try_loading_module = 0;
274 int err;
275
276 if (unlikely(!inet_ehash_secret))
277 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
278 build_ehash_secret();
279
280 sock->state = SS_UNCONNECTED;
281
282
283lookup_protocol:
284 err = -ESOCKTNOSUPPORT;
285 rcu_read_lock();
286 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
287
288 err = 0;
289
290 if (protocol == answer->protocol) {
291 if (protocol != IPPROTO_IP)
292 break;
293 } else {
294
295 if (IPPROTO_IP == protocol) {
296 protocol = answer->protocol;
297 break;
298 }
299 if (IPPROTO_IP == answer->protocol)
300 break;
301 }
302 err = -EPROTONOSUPPORT;
303 }
304
305 if (unlikely(err)) {
306 if (try_loading_module < 2) {
307 rcu_read_unlock();
308
309
310
311
312 if (++try_loading_module == 1)
313 request_module("net-pf-%d-proto-%d-type-%d",
314 PF_INET, protocol, sock->type);
315
316
317
318
319 else
320 request_module("net-pf-%d-proto-%d",
321 PF_INET, protocol);
322 goto lookup_protocol;
323 } else
324 goto out_rcu_unlock;
325 }
326
327 err = -EPERM;
328 if (answer->capability > 0 && !capable(answer->capability))
329 goto out_rcu_unlock;
330
331 err = -EAFNOSUPPORT;
332 if (!inet_netns_ok(net, protocol))
333 goto out_rcu_unlock;
334
335 sock->ops = answer->ops;
336 answer_prot = answer->prot;
337 answer_no_check = answer->no_check;
338 answer_flags = answer->flags;
339 rcu_read_unlock();
340
341 WARN_ON(answer_prot->slab == NULL);
342
343 err = -ENOBUFS;
344 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
345 if (sk == NULL)
346 goto out;
347
348 err = 0;
349 sk->sk_no_check = answer_no_check;
350 if (INET_PROTOSW_REUSE & answer_flags)
351 sk->sk_reuse = 1;
352
353 inet = inet_sk(sk);
354 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
355
356 if (SOCK_RAW == sock->type) {
357 inet->num = protocol;
358 if (IPPROTO_RAW == protocol)
359 inet->hdrincl = 1;
360 }
361
362 if (ipv4_config.no_pmtu_disc)
363 inet->pmtudisc = IP_PMTUDISC_DONT;
364 else
365 inet->pmtudisc = IP_PMTUDISC_WANT;
366
367 inet->id = 0;
368
369 sock_init_data(sock, sk);
370
371 sk->sk_destruct = inet_sock_destruct;
372 sk->sk_protocol = protocol;
373 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
374
375 inet->uc_ttl = -1;
376 inet->mc_loop = 1;
377 inet->mc_ttl = 1;
378 inet->mc_all = 1;
379 inet->mc_index = 0;
380 inet->mc_list = NULL;
381
382 sk_refcnt_debug_inc(sk);
383
384 if (inet->num) {
385
386
387
388
389
390 inet->sport = htons(inet->num);
391
392 sk->sk_prot->hash(sk);
393 }
394
395 if (sk->sk_prot->init) {
396 err = sk->sk_prot->init(sk);
397 if (err)
398 sk_common_release(sk);
399 }
400out:
401 return err;
402out_rcu_unlock:
403 rcu_read_unlock();
404 goto out;
405}
406
407
408
409
410
411
412
413int inet_release(struct socket *sock)
414{
415 struct sock *sk = sock->sk;
416
417 if (sk) {
418 long timeout;
419
420
421 ip_mc_drop_socket(sk);
422
423
424
425
426
427
428
429
430 timeout = 0;
431 if (sock_flag(sk, SOCK_LINGER) &&
432 !(current->flags & PF_EXITING))
433 timeout = sk->sk_lingertime;
434 sock->sk = NULL;
435 sk->sk_prot->close(sk, timeout);
436 }
437 return 0;
438}
439EXPORT_SYMBOL(inet_release);
440
441
442int sysctl_ip_nonlocal_bind __read_mostly;
443EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
444
445int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
446{
447 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
448 struct sock *sk = sock->sk;
449 struct inet_sock *inet = inet_sk(sk);
450 unsigned short snum;
451 int chk_addr_ret;
452 int err;
453
454
455 if (sk->sk_prot->bind) {
456 err = sk->sk_prot->bind(sk, uaddr, addr_len);
457 goto out;
458 }
459 err = -EINVAL;
460 if (addr_len < sizeof(struct sockaddr_in))
461 goto out;
462
463 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
464
465
466
467
468
469
470
471
472 err = -EADDRNOTAVAIL;
473 if (!sysctl_ip_nonlocal_bind &&
474 !(inet->freebind || inet->transparent) &&
475 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
476 chk_addr_ret != RTN_LOCAL &&
477 chk_addr_ret != RTN_MULTICAST &&
478 chk_addr_ret != RTN_BROADCAST)
479 goto out;
480
481 snum = ntohs(addr->sin_port);
482 err = -EACCES;
483 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
484 goto out;
485
486
487
488
489
490
491
492
493 lock_sock(sk);
494
495
496 err = -EINVAL;
497 if (sk->sk_state != TCP_CLOSE || inet->num)
498 goto out_release_sock;
499
500 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
501 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
502 inet->saddr = 0;
503
504
505 if (sk->sk_prot->get_port(sk, snum)) {
506 inet->saddr = inet->rcv_saddr = 0;
507 err = -EADDRINUSE;
508 goto out_release_sock;
509 }
510
511 if (inet->rcv_saddr)
512 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
513 if (snum)
514 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
515 inet->sport = htons(inet->num);
516 inet->daddr = 0;
517 inet->dport = 0;
518 sk_dst_reset(sk);
519 err = 0;
520out_release_sock:
521 release_sock(sk);
522out:
523 return err;
524}
525EXPORT_SYMBOL(inet_bind);
526
527int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
528 int addr_len, int flags)
529{
530 struct sock *sk = sock->sk;
531
532 if (uaddr->sa_family == AF_UNSPEC)
533 return sk->sk_prot->disconnect(sk, flags);
534
535 if (!inet_sk(sk)->num && inet_autobind(sk))
536 return -EAGAIN;
537 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
538}
539EXPORT_SYMBOL(inet_dgram_connect);
540
541static long inet_wait_for_connect(struct sock *sk, long timeo)
542{
543 DEFINE_WAIT(wait);
544
545 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
546
547
548
549
550
551
552 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
553 release_sock(sk);
554 timeo = schedule_timeout(timeo);
555 lock_sock(sk);
556 if (signal_pending(current) || !timeo)
557 break;
558 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
559 }
560 finish_wait(sk->sk_sleep, &wait);
561 return timeo;
562}
563
564
565
566
567
568int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
569 int addr_len, int flags)
570{
571 struct sock *sk = sock->sk;
572 int err;
573 long timeo;
574
575 lock_sock(sk);
576
577 if (uaddr->sa_family == AF_UNSPEC) {
578 err = sk->sk_prot->disconnect(sk, flags);
579 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
580 goto out;
581 }
582
583 switch (sock->state) {
584 default:
585 err = -EINVAL;
586 goto out;
587 case SS_CONNECTED:
588 err = -EISCONN;
589 goto out;
590 case SS_CONNECTING:
591 err = -EALREADY;
592
593 break;
594 case SS_UNCONNECTED:
595 err = -EISCONN;
596 if (sk->sk_state != TCP_CLOSE)
597 goto out;
598
599 err = sk->sk_prot->connect(sk, uaddr, addr_len);
600 if (err < 0)
601 goto out;
602
603 sock->state = SS_CONNECTING;
604
605
606
607
608
609 err = -EINPROGRESS;
610 break;
611 }
612
613 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
614
615 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
616
617 if (!timeo || !inet_wait_for_connect(sk, timeo))
618 goto out;
619
620 err = sock_intr_errno(timeo);
621 if (signal_pending(current))
622 goto out;
623 }
624
625
626
627
628 if (sk->sk_state == TCP_CLOSE)
629 goto sock_error;
630
631
632
633
634
635
636 sock->state = SS_CONNECTED;
637 err = 0;
638out:
639 release_sock(sk);
640 return err;
641
642sock_error:
643 err = sock_error(sk) ? : -ECONNABORTED;
644 sock->state = SS_UNCONNECTED;
645 if (sk->sk_prot->disconnect(sk, flags))
646 sock->state = SS_DISCONNECTING;
647 goto out;
648}
649EXPORT_SYMBOL(inet_stream_connect);
650
651
652
653
654
655int inet_accept(struct socket *sock, struct socket *newsock, int flags)
656{
657 struct sock *sk1 = sock->sk;
658 int err = -EINVAL;
659 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
660
661 if (!sk2)
662 goto do_err;
663
664 lock_sock(sk2);
665
666 WARN_ON(!((1 << sk2->sk_state) &
667 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)));
668
669 sock_graft(sk2, newsock);
670
671 newsock->state = SS_CONNECTED;
672 err = 0;
673 release_sock(sk2);
674do_err:
675 return err;
676}
677EXPORT_SYMBOL(inet_accept);
678
679
680
681
682
683int inet_getname(struct socket *sock, struct sockaddr *uaddr,
684 int *uaddr_len, int peer)
685{
686 struct sock *sk = sock->sk;
687 struct inet_sock *inet = inet_sk(sk);
688 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
689
690 sin->sin_family = AF_INET;
691 if (peer) {
692 if (!inet->dport ||
693 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
694 peer == 1))
695 return -ENOTCONN;
696 sin->sin_port = inet->dport;
697 sin->sin_addr.s_addr = inet->daddr;
698 } else {
699 __be32 addr = inet->rcv_saddr;
700 if (!addr)
701 addr = inet->saddr;
702 sin->sin_port = inet->sport;
703 sin->sin_addr.s_addr = addr;
704 }
705 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
706 *uaddr_len = sizeof(*sin);
707 return 0;
708}
709EXPORT_SYMBOL(inet_getname);
710
711int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
712 size_t size)
713{
714 struct sock *sk = sock->sk;
715
716
717 if (!inet_sk(sk)->num && inet_autobind(sk))
718 return -EAGAIN;
719
720 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
721}
722EXPORT_SYMBOL(inet_sendmsg);
723
724
725static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
726 size_t size, int flags)
727{
728 struct sock *sk = sock->sk;
729
730
731 if (!inet_sk(sk)->num && inet_autobind(sk))
732 return -EAGAIN;
733
734 if (sk->sk_prot->sendpage)
735 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
736 return sock_no_sendpage(sock, page, offset, size, flags);
737}
738
739
740int inet_shutdown(struct socket *sock, int how)
741{
742 struct sock *sk = sock->sk;
743 int err = 0;
744
745
746
747
748 how++;
749
750
751 if ((how & ~SHUTDOWN_MASK) || !how)
752 return -EINVAL;
753
754 lock_sock(sk);
755 if (sock->state == SS_CONNECTING) {
756 if ((1 << sk->sk_state) &
757 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
758 sock->state = SS_DISCONNECTING;
759 else
760 sock->state = SS_CONNECTED;
761 }
762
763 switch (sk->sk_state) {
764 case TCP_CLOSE:
765 err = -ENOTCONN;
766
767
768 default:
769 sk->sk_shutdown |= how;
770 if (sk->sk_prot->shutdown)
771 sk->sk_prot->shutdown(sk, how);
772 break;
773
774
775
776
777
778 case TCP_LISTEN:
779 if (!(how & RCV_SHUTDOWN))
780 break;
781
782 case TCP_SYN_SENT:
783 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
784 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
785 break;
786 }
787
788
789 sk->sk_state_change(sk);
790 release_sock(sk);
791 return err;
792}
793EXPORT_SYMBOL(inet_shutdown);
794
795
796
797
798
799
800
801
802
803
804
805int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
806{
807 struct sock *sk = sock->sk;
808 int err = 0;
809 struct net *net = sock_net(sk);
810
811 switch (cmd) {
812 case SIOCGSTAMP:
813 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
814 break;
815 case SIOCGSTAMPNS:
816 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
817 break;
818 case SIOCADDRT:
819 case SIOCDELRT:
820 case SIOCRTMSG:
821 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
822 break;
823 case SIOCDARP:
824 case SIOCGARP:
825 case SIOCSARP:
826 err = arp_ioctl(net, cmd, (void __user *)arg);
827 break;
828 case SIOCGIFADDR:
829 case SIOCSIFADDR:
830 case SIOCGIFBRDADDR:
831 case SIOCSIFBRDADDR:
832 case SIOCGIFNETMASK:
833 case SIOCSIFNETMASK:
834 case SIOCGIFDSTADDR:
835 case SIOCSIFDSTADDR:
836 case SIOCSIFPFLAGS:
837 case SIOCGIFPFLAGS:
838 case SIOCSIFFLAGS:
839 err = devinet_ioctl(net, cmd, (void __user *)arg);
840 break;
841 default:
842 if (sk->sk_prot->ioctl)
843 err = sk->sk_prot->ioctl(sk, cmd, arg);
844 else
845 err = -ENOIOCTLCMD;
846 break;
847 }
848 return err;
849}
850EXPORT_SYMBOL(inet_ioctl);
851
852const struct proto_ops inet_stream_ops = {
853 .family = PF_INET,
854 .owner = THIS_MODULE,
855 .release = inet_release,
856 .bind = inet_bind,
857 .connect = inet_stream_connect,
858 .socketpair = sock_no_socketpair,
859 .accept = inet_accept,
860 .getname = inet_getname,
861 .poll = tcp_poll,
862 .ioctl = inet_ioctl,
863 .listen = inet_listen,
864 .shutdown = inet_shutdown,
865 .setsockopt = sock_common_setsockopt,
866 .getsockopt = sock_common_getsockopt,
867 .sendmsg = tcp_sendmsg,
868 .recvmsg = sock_common_recvmsg,
869 .mmap = sock_no_mmap,
870 .sendpage = tcp_sendpage,
871 .splice_read = tcp_splice_read,
872#ifdef CONFIG_COMPAT
873 .compat_setsockopt = compat_sock_common_setsockopt,
874 .compat_getsockopt = compat_sock_common_getsockopt,
875#endif
876};
877EXPORT_SYMBOL(inet_stream_ops);
878
879const struct proto_ops inet_dgram_ops = {
880 .family = PF_INET,
881 .owner = THIS_MODULE,
882 .release = inet_release,
883 .bind = inet_bind,
884 .connect = inet_dgram_connect,
885 .socketpair = sock_no_socketpair,
886 .accept = sock_no_accept,
887 .getname = inet_getname,
888 .poll = udp_poll,
889 .ioctl = inet_ioctl,
890 .listen = sock_no_listen,
891 .shutdown = inet_shutdown,
892 .setsockopt = sock_common_setsockopt,
893 .getsockopt = sock_common_getsockopt,
894 .sendmsg = inet_sendmsg,
895 .recvmsg = sock_common_recvmsg,
896 .mmap = sock_no_mmap,
897 .sendpage = inet_sendpage,
898#ifdef CONFIG_COMPAT
899 .compat_setsockopt = compat_sock_common_setsockopt,
900 .compat_getsockopt = compat_sock_common_getsockopt,
901#endif
902};
903EXPORT_SYMBOL(inet_dgram_ops);
904
905
906
907
908
909static const struct proto_ops inet_sockraw_ops = {
910 .family = PF_INET,
911 .owner = THIS_MODULE,
912 .release = inet_release,
913 .bind = inet_bind,
914 .connect = inet_dgram_connect,
915 .socketpair = sock_no_socketpair,
916 .accept = sock_no_accept,
917 .getname = inet_getname,
918 .poll = datagram_poll,
919 .ioctl = inet_ioctl,
920 .listen = sock_no_listen,
921 .shutdown = inet_shutdown,
922 .setsockopt = sock_common_setsockopt,
923 .getsockopt = sock_common_getsockopt,
924 .sendmsg = inet_sendmsg,
925 .recvmsg = sock_common_recvmsg,
926 .mmap = sock_no_mmap,
927 .sendpage = inet_sendpage,
928#ifdef CONFIG_COMPAT
929 .compat_setsockopt = compat_sock_common_setsockopt,
930 .compat_getsockopt = compat_sock_common_getsockopt,
931#endif
932};
933
934static struct net_proto_family inet_family_ops = {
935 .family = PF_INET,
936 .create = inet_create,
937 .owner = THIS_MODULE,
938};
939
940
941
942
943static struct inet_protosw inetsw_array[] =
944{
945 {
946 .type = SOCK_STREAM,
947 .protocol = IPPROTO_TCP,
948 .prot = &tcp_prot,
949 .ops = &inet_stream_ops,
950 .capability = -1,
951 .no_check = 0,
952 .flags = INET_PROTOSW_PERMANENT |
953 INET_PROTOSW_ICSK,
954 },
955
956 {
957 .type = SOCK_DGRAM,
958 .protocol = IPPROTO_UDP,
959 .prot = &udp_prot,
960 .ops = &inet_dgram_ops,
961 .capability = -1,
962 .no_check = UDP_CSUM_DEFAULT,
963 .flags = INET_PROTOSW_PERMANENT,
964 },
965
966
967 {
968 .type = SOCK_RAW,
969 .protocol = IPPROTO_IP,
970 .prot = &raw_prot,
971 .ops = &inet_sockraw_ops,
972 .capability = CAP_NET_RAW,
973 .no_check = UDP_CSUM_DEFAULT,
974 .flags = INET_PROTOSW_REUSE,
975 }
976};
977
978#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
979
980void inet_register_protosw(struct inet_protosw *p)
981{
982 struct list_head *lh;
983 struct inet_protosw *answer;
984 int protocol = p->protocol;
985 struct list_head *last_perm;
986
987 spin_lock_bh(&inetsw_lock);
988
989 if (p->type >= SOCK_MAX)
990 goto out_illegal;
991
992
993 answer = NULL;
994 last_perm = &inetsw[p->type];
995 list_for_each(lh, &inetsw[p->type]) {
996 answer = list_entry(lh, struct inet_protosw, list);
997
998
999 if (INET_PROTOSW_PERMANENT & answer->flags) {
1000 if (protocol == answer->protocol)
1001 break;
1002 last_perm = lh;
1003 }
1004
1005 answer = NULL;
1006 }
1007 if (answer)
1008 goto out_permanent;
1009
1010
1011
1012
1013
1014
1015
1016 list_add_rcu(&p->list, last_perm);
1017out:
1018 spin_unlock_bh(&inetsw_lock);
1019
1020 return;
1021
1022out_permanent:
1023 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1024 protocol);
1025 goto out;
1026
1027out_illegal:
1028 printk(KERN_ERR
1029 "Ignoring attempt to register invalid socket type %d.\n",
1030 p->type);
1031 goto out;
1032}
1033EXPORT_SYMBOL(inet_register_protosw);
1034
1035void inet_unregister_protosw(struct inet_protosw *p)
1036{
1037 if (INET_PROTOSW_PERMANENT & p->flags) {
1038 printk(KERN_ERR
1039 "Attempt to unregister permanent protocol %d.\n",
1040 p->protocol);
1041 } else {
1042 spin_lock_bh(&inetsw_lock);
1043 list_del_rcu(&p->list);
1044 spin_unlock_bh(&inetsw_lock);
1045
1046 synchronize_net();
1047 }
1048}
1049EXPORT_SYMBOL(inet_unregister_protosw);
1050
1051
1052
1053
1054
1055int sysctl_ip_dynaddr __read_mostly;
1056
1057static int inet_sk_reselect_saddr(struct sock *sk)
1058{
1059 struct inet_sock *inet = inet_sk(sk);
1060 int err;
1061 struct rtable *rt;
1062 __be32 old_saddr = inet->saddr;
1063 __be32 new_saddr;
1064 __be32 daddr = inet->daddr;
1065
1066 if (inet->opt && inet->opt->srr)
1067 daddr = inet->opt->faddr;
1068
1069
1070 err = ip_route_connect(&rt, daddr, 0,
1071 RT_CONN_FLAGS(sk),
1072 sk->sk_bound_dev_if,
1073 sk->sk_protocol,
1074 inet->sport, inet->dport, sk, 0);
1075 if (err)
1076 return err;
1077
1078 sk_setup_caps(sk, &rt->u.dst);
1079
1080 new_saddr = rt->rt_src;
1081
1082 if (new_saddr == old_saddr)
1083 return 0;
1084
1085 if (sysctl_ip_dynaddr > 1) {
1086 printk(KERN_INFO "%s(): shifting inet->saddr from %pI4 to %pI4\n",
1087 __func__, &old_saddr, &new_saddr);
1088 }
1089
1090 inet->saddr = inet->rcv_saddr = new_saddr;
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100 __sk_prot_rehash(sk);
1101 return 0;
1102}
1103
1104int inet_sk_rebuild_header(struct sock *sk)
1105{
1106 struct inet_sock *inet = inet_sk(sk);
1107 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1108 __be32 daddr;
1109 int err;
1110
1111
1112 if (rt)
1113 return 0;
1114
1115
1116 daddr = inet->daddr;
1117 if (inet->opt && inet->opt->srr)
1118 daddr = inet->opt->faddr;
1119{
1120 struct flowi fl = {
1121 .oif = sk->sk_bound_dev_if,
1122 .mark = sk->sk_mark,
1123 .nl_u = {
1124 .ip4_u = {
1125 .daddr = daddr,
1126 .saddr = inet->saddr,
1127 .tos = RT_CONN_FLAGS(sk),
1128 },
1129 },
1130 .proto = sk->sk_protocol,
1131 .flags = inet_sk_flowi_flags(sk),
1132 .uli_u = {
1133 .ports = {
1134 .sport = inet->sport,
1135 .dport = inet->dport,
1136 },
1137 },
1138 };
1139
1140 security_sk_classify_flow(sk, &fl);
1141 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 0);
1142}
1143 if (!err)
1144 sk_setup_caps(sk, &rt->u.dst);
1145 else {
1146
1147 sk->sk_route_caps = 0;
1148
1149
1150
1151
1152 if (!sysctl_ip_dynaddr ||
1153 sk->sk_state != TCP_SYN_SENT ||
1154 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1155 (err = inet_sk_reselect_saddr(sk)) != 0)
1156 sk->sk_err_soft = -err;
1157 }
1158
1159 return err;
1160}
1161EXPORT_SYMBOL(inet_sk_rebuild_header);
1162
1163static int inet_gso_send_check(struct sk_buff *skb)
1164{
1165 struct iphdr *iph;
1166 const struct net_protocol *ops;
1167 int proto;
1168 int ihl;
1169 int err = -EINVAL;
1170
1171 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1172 goto out;
1173
1174 iph = ip_hdr(skb);
1175 ihl = iph->ihl * 4;
1176 if (ihl < sizeof(*iph))
1177 goto out;
1178
1179 if (unlikely(!pskb_may_pull(skb, ihl)))
1180 goto out;
1181
1182 __skb_pull(skb, ihl);
1183 skb_reset_transport_header(skb);
1184 iph = ip_hdr(skb);
1185 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1186 err = -EPROTONOSUPPORT;
1187
1188 rcu_read_lock();
1189 ops = rcu_dereference(inet_protos[proto]);
1190 if (likely(ops && ops->gso_send_check))
1191 err = ops->gso_send_check(skb);
1192 rcu_read_unlock();
1193
1194out:
1195 return err;
1196}
1197
1198static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
1199{
1200 struct sk_buff *segs = ERR_PTR(-EINVAL);
1201 struct iphdr *iph;
1202 const struct net_protocol *ops;
1203 int proto;
1204 int ihl;
1205 int id;
1206 unsigned int offset = 0;
1207
1208 if (!(features & NETIF_F_V4_CSUM))
1209 features &= ~NETIF_F_SG;
1210
1211 if (unlikely(skb_shinfo(skb)->gso_type &
1212 ~(SKB_GSO_TCPV4 |
1213 SKB_GSO_UDP |
1214 SKB_GSO_DODGY |
1215 SKB_GSO_TCP_ECN |
1216 0)))
1217 goto out;
1218
1219 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1220 goto out;
1221
1222 iph = ip_hdr(skb);
1223 ihl = iph->ihl * 4;
1224 if (ihl < sizeof(*iph))
1225 goto out;
1226
1227 if (unlikely(!pskb_may_pull(skb, ihl)))
1228 goto out;
1229
1230 __skb_pull(skb, ihl);
1231 skb_reset_transport_header(skb);
1232 iph = ip_hdr(skb);
1233 id = ntohs(iph->id);
1234 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1235 segs = ERR_PTR(-EPROTONOSUPPORT);
1236
1237 rcu_read_lock();
1238 ops = rcu_dereference(inet_protos[proto]);
1239 if (likely(ops && ops->gso_segment))
1240 segs = ops->gso_segment(skb, features);
1241 rcu_read_unlock();
1242
1243 if (!segs || IS_ERR(segs))
1244 goto out;
1245
1246 skb = segs;
1247 do {
1248 iph = ip_hdr(skb);
1249 if (proto == IPPROTO_UDP) {
1250 iph->id = htons(id);
1251 iph->frag_off = htons(offset >> 3);
1252 if (skb->next != NULL)
1253 iph->frag_off |= htons(IP_MF);
1254 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1255 } else
1256 iph->id = htons(id++);
1257 iph->tot_len = htons(skb->len - skb->mac_len);
1258 iph->check = 0;
1259 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
1260 } while ((skb = skb->next));
1261
1262out:
1263 return segs;
1264}
1265
1266static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1267 struct sk_buff *skb)
1268{
1269 const struct net_protocol *ops;
1270 struct sk_buff **pp = NULL;
1271 struct sk_buff *p;
1272 struct iphdr *iph;
1273 unsigned int hlen;
1274 unsigned int off;
1275 unsigned int id;
1276 int flush = 1;
1277 int proto;
1278
1279 off = skb_gro_offset(skb);
1280 hlen = off + sizeof(*iph);
1281 iph = skb_gro_header_fast(skb, off);
1282 if (skb_gro_header_hard(skb, hlen)) {
1283 iph = skb_gro_header_slow(skb, hlen, off);
1284 if (unlikely(!iph))
1285 goto out;
1286 }
1287
1288 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1289
1290 rcu_read_lock();
1291 ops = rcu_dereference(inet_protos[proto]);
1292 if (!ops || !ops->gro_receive)
1293 goto out_unlock;
1294
1295 if (*(u8 *)iph != 0x45)
1296 goto out_unlock;
1297
1298 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1299 goto out_unlock;
1300
1301 id = ntohl(*(u32 *)&iph->id);
1302 flush = (u16)((ntohl(*(u32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
1303 id >>= 16;
1304
1305 for (p = *head; p; p = p->next) {
1306 struct iphdr *iph2;
1307
1308 if (!NAPI_GRO_CB(p)->same_flow)
1309 continue;
1310
1311 iph2 = ip_hdr(p);
1312
1313 if ((iph->protocol ^ iph2->protocol) |
1314 (iph->tos ^ iph2->tos) |
1315 (iph->saddr ^ iph2->saddr) |
1316 (iph->daddr ^ iph2->daddr)) {
1317 NAPI_GRO_CB(p)->same_flow = 0;
1318 continue;
1319 }
1320
1321
1322 NAPI_GRO_CB(p)->flush |=
1323 (iph->ttl ^ iph2->ttl) |
1324 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
1325
1326 NAPI_GRO_CB(p)->flush |= flush;
1327 }
1328
1329 NAPI_GRO_CB(skb)->flush |= flush;
1330 skb_gro_pull(skb, sizeof(*iph));
1331 skb_set_transport_header(skb, skb_gro_offset(skb));
1332
1333 pp = ops->gro_receive(head, skb);
1334
1335out_unlock:
1336 rcu_read_unlock();
1337
1338out:
1339 NAPI_GRO_CB(skb)->flush |= flush;
1340
1341 return pp;
1342}
1343
1344static int inet_gro_complete(struct sk_buff *skb)
1345{
1346 const struct net_protocol *ops;
1347 struct iphdr *iph = ip_hdr(skb);
1348 int proto = iph->protocol & (MAX_INET_PROTOS - 1);
1349 int err = -ENOSYS;
1350 __be16 newlen = htons(skb->len - skb_network_offset(skb));
1351
1352 csum_replace2(&iph->check, iph->tot_len, newlen);
1353 iph->tot_len = newlen;
1354
1355 rcu_read_lock();
1356 ops = rcu_dereference(inet_protos[proto]);
1357 if (WARN_ON(!ops || !ops->gro_complete))
1358 goto out_unlock;
1359
1360 err = ops->gro_complete(skb);
1361
1362out_unlock:
1363 rcu_read_unlock();
1364
1365 return err;
1366}
1367
1368int inet_ctl_sock_create(struct sock **sk, unsigned short family,
1369 unsigned short type, unsigned char protocol,
1370 struct net *net)
1371{
1372 struct socket *sock;
1373 int rc = sock_create_kern(family, type, protocol, &sock);
1374
1375 if (rc == 0) {
1376 *sk = sock->sk;
1377 (*sk)->sk_allocation = GFP_ATOMIC;
1378
1379
1380
1381
1382 (*sk)->sk_prot->unhash(*sk);
1383
1384 sk_change_net(*sk, net);
1385 }
1386 return rc;
1387}
1388EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1389
1390unsigned long snmp_fold_field(void *mib[], int offt)
1391{
1392 unsigned long res = 0;
1393 int i;
1394
1395 for_each_possible_cpu(i) {
1396 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1397 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1398 }
1399 return res;
1400}
1401EXPORT_SYMBOL_GPL(snmp_fold_field);
1402
1403int snmp_mib_init(void *ptr[2], size_t mibsize)
1404{
1405 BUG_ON(ptr == NULL);
1406 ptr[0] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
1407 if (!ptr[0])
1408 goto err0;
1409 ptr[1] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
1410 if (!ptr[1])
1411 goto err1;
1412 return 0;
1413err1:
1414 free_percpu(ptr[0]);
1415 ptr[0] = NULL;
1416err0:
1417 return -ENOMEM;
1418}
1419EXPORT_SYMBOL_GPL(snmp_mib_init);
1420
1421void snmp_mib_free(void *ptr[2])
1422{
1423 BUG_ON(ptr == NULL);
1424 free_percpu(ptr[0]);
1425 free_percpu(ptr[1]);
1426 ptr[0] = ptr[1] = NULL;
1427}
1428EXPORT_SYMBOL_GPL(snmp_mib_free);
1429
1430#ifdef CONFIG_IP_MULTICAST
1431static const struct net_protocol igmp_protocol = {
1432 .handler = igmp_rcv,
1433 .netns_ok = 1,
1434};
1435#endif
1436
1437static const struct net_protocol tcp_protocol = {
1438 .handler = tcp_v4_rcv,
1439 .err_handler = tcp_v4_err,
1440 .gso_send_check = tcp_v4_gso_send_check,
1441 .gso_segment = tcp_tso_segment,
1442 .gro_receive = tcp4_gro_receive,
1443 .gro_complete = tcp4_gro_complete,
1444 .no_policy = 1,
1445 .netns_ok = 1,
1446};
1447
1448static const struct net_protocol udp_protocol = {
1449 .handler = udp_rcv,
1450 .err_handler = udp_err,
1451 .gso_send_check = udp4_ufo_send_check,
1452 .gso_segment = udp4_ufo_fragment,
1453 .no_policy = 1,
1454 .netns_ok = 1,
1455};
1456
1457static const struct net_protocol icmp_protocol = {
1458 .handler = icmp_rcv,
1459 .no_policy = 1,
1460 .netns_ok = 1,
1461};
1462
1463static __net_init int ipv4_mib_init_net(struct net *net)
1464{
1465 if (snmp_mib_init((void **)net->mib.tcp_statistics,
1466 sizeof(struct tcp_mib)) < 0)
1467 goto err_tcp_mib;
1468 if (snmp_mib_init((void **)net->mib.ip_statistics,
1469 sizeof(struct ipstats_mib)) < 0)
1470 goto err_ip_mib;
1471 if (snmp_mib_init((void **)net->mib.net_statistics,
1472 sizeof(struct linux_mib)) < 0)
1473 goto err_net_mib;
1474 if (snmp_mib_init((void **)net->mib.udp_statistics,
1475 sizeof(struct udp_mib)) < 0)
1476 goto err_udp_mib;
1477 if (snmp_mib_init((void **)net->mib.udplite_statistics,
1478 sizeof(struct udp_mib)) < 0)
1479 goto err_udplite_mib;
1480 if (snmp_mib_init((void **)net->mib.icmp_statistics,
1481 sizeof(struct icmp_mib)) < 0)
1482 goto err_icmp_mib;
1483 if (snmp_mib_init((void **)net->mib.icmpmsg_statistics,
1484 sizeof(struct icmpmsg_mib)) < 0)
1485 goto err_icmpmsg_mib;
1486
1487 tcp_mib_init(net);
1488 return 0;
1489
1490err_icmpmsg_mib:
1491 snmp_mib_free((void **)net->mib.icmp_statistics);
1492err_icmp_mib:
1493 snmp_mib_free((void **)net->mib.udplite_statistics);
1494err_udplite_mib:
1495 snmp_mib_free((void **)net->mib.udp_statistics);
1496err_udp_mib:
1497 snmp_mib_free((void **)net->mib.net_statistics);
1498err_net_mib:
1499 snmp_mib_free((void **)net->mib.ip_statistics);
1500err_ip_mib:
1501 snmp_mib_free((void **)net->mib.tcp_statistics);
1502err_tcp_mib:
1503 return -ENOMEM;
1504}
1505
1506static __net_exit void ipv4_mib_exit_net(struct net *net)
1507{
1508 snmp_mib_free((void **)net->mib.icmpmsg_statistics);
1509 snmp_mib_free((void **)net->mib.icmp_statistics);
1510 snmp_mib_free((void **)net->mib.udplite_statistics);
1511 snmp_mib_free((void **)net->mib.udp_statistics);
1512 snmp_mib_free((void **)net->mib.net_statistics);
1513 snmp_mib_free((void **)net->mib.ip_statistics);
1514 snmp_mib_free((void **)net->mib.tcp_statistics);
1515}
1516
1517static __net_initdata struct pernet_operations ipv4_mib_ops = {
1518 .init = ipv4_mib_init_net,
1519 .exit = ipv4_mib_exit_net,
1520};
1521
1522static int __init init_ipv4_mibs(void)
1523{
1524 return register_pernet_subsys(&ipv4_mib_ops);
1525}
1526
1527static int ipv4_proc_init(void);
1528
1529
1530
1531
1532
1533static struct packet_type ip_packet_type __read_mostly = {
1534 .type = cpu_to_be16(ETH_P_IP),
1535 .func = ip_rcv,
1536 .gso_send_check = inet_gso_send_check,
1537 .gso_segment = inet_gso_segment,
1538 .gro_receive = inet_gro_receive,
1539 .gro_complete = inet_gro_complete,
1540};
1541
1542static int __init inet_init(void)
1543{
1544 struct sk_buff *dummy_skb;
1545 struct inet_protosw *q;
1546 struct list_head *r;
1547 int rc = -EINVAL;
1548
1549 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
1550
1551 rc = proto_register(&tcp_prot, 1);
1552 if (rc)
1553 goto out;
1554
1555 rc = proto_register(&udp_prot, 1);
1556 if (rc)
1557 goto out_unregister_tcp_proto;
1558
1559 rc = proto_register(&raw_prot, 1);
1560 if (rc)
1561 goto out_unregister_udp_proto;
1562
1563
1564
1565
1566
1567 (void)sock_register(&inet_family_ops);
1568
1569#ifdef CONFIG_SYSCTL
1570 ip_static_sysctl_init();
1571#endif
1572
1573
1574
1575
1576
1577 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1578 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1579 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1580 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1581 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1582 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1583#ifdef CONFIG_IP_MULTICAST
1584 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1585 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1586#endif
1587
1588
1589 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1590 INIT_LIST_HEAD(r);
1591
1592 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1593 inet_register_protosw(q);
1594
1595
1596
1597
1598
1599 arp_init();
1600
1601
1602
1603
1604
1605 ip_init();
1606
1607 tcp_v4_init();
1608
1609
1610 tcp_init();
1611
1612
1613 udp_init();
1614
1615
1616 udplite4_register();
1617
1618
1619
1620
1621
1622 if (icmp_init() < 0)
1623 panic("Failed to create the ICMP control socket.\n");
1624
1625
1626
1627
1628#if defined(CONFIG_IP_MROUTE)
1629 if (ip_mr_init())
1630 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
1631#endif
1632
1633
1634
1635
1636 if (init_ipv4_mibs())
1637 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
1638
1639 ipv4_proc_init();
1640
1641 ipfrag_init();
1642
1643 dev_add_pack(&ip_packet_type);
1644
1645 rc = 0;
1646out:
1647 return rc;
1648out_unregister_udp_proto:
1649 proto_unregister(&udp_prot);
1650out_unregister_tcp_proto:
1651 proto_unregister(&tcp_prot);
1652 goto out;
1653}
1654
1655fs_initcall(inet_init);
1656
1657
1658
1659#ifdef CONFIG_PROC_FS
1660static int __init ipv4_proc_init(void)
1661{
1662 int rc = 0;
1663
1664 if (raw_proc_init())
1665 goto out_raw;
1666 if (tcp4_proc_init())
1667 goto out_tcp;
1668 if (udp4_proc_init())
1669 goto out_udp;
1670 if (ip_misc_proc_init())
1671 goto out_misc;
1672out:
1673 return rc;
1674out_misc:
1675 udp4_proc_exit();
1676out_udp:
1677 tcp4_proc_exit();
1678out_tcp:
1679 raw_proc_exit();
1680out_raw:
1681 rc = -ENOMEM;
1682 goto out;
1683}
1684
1685#else
1686static int __init ipv4_proc_init(void)
1687{
1688 return 0;
1689}
1690#endif
1691
1692MODULE_ALIAS_NETPROTO(PF_INET);
1693
1694