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#define pr_fmt(fmt) "IPv4: " fmt
69
70#include <linux/err.h>
71#include <linux/errno.h>
72#include <linux/types.h>
73#include <linux/socket.h>
74#include <linux/in.h>
75#include <linux/kernel.h>
76#include <linux/module.h>
77#include <linux/sched.h>
78#include <linux/timer.h>
79#include <linux/string.h>
80#include <linux/sockios.h>
81#include <linux/net.h>
82#include <linux/capability.h>
83#include <linux/fcntl.h>
84#include <linux/mm.h>
85#include <linux/interrupt.h>
86#include <linux/stat.h>
87#include <linux/init.h>
88#include <linux/poll.h>
89#include <linux/netfilter_ipv4.h>
90#include <linux/random.h>
91#include <linux/slab.h>
92
93#include <asm/uaccess.h>
94
95#include <linux/inet.h>
96#include <linux/igmp.h>
97#include <linux/inetdevice.h>
98#include <linux/netdevice.h>
99#include <net/checksum.h>
100#include <net/ip.h>
101#include <net/protocol.h>
102#include <net/arp.h>
103#include <net/route.h>
104#include <net/ip_fib.h>
105#include <net/inet_connection_sock.h>
106#include <net/tcp.h>
107#include <net/udp.h>
108#include <net/udplite.h>
109#include <net/ping.h>
110#include <linux/skbuff.h>
111#include <net/sock.h>
112#include <net/raw.h>
113#include <net/icmp.h>
114#include <net/inet_common.h>
115#include <net/xfrm.h>
116#include <net/net_namespace.h>
117#include <net/secure_seq.h>
118#ifdef CONFIG_IP_MROUTE
119#include <linux/mroute.h>
120#endif
121
122
123
124
125
126static struct list_head inetsw[SOCK_MAX];
127static DEFINE_SPINLOCK(inetsw_lock);
128
129struct ipv4_config ipv4_config;
130EXPORT_SYMBOL(ipv4_config);
131
132
133
134void inet_sock_destruct(struct sock *sk)
135{
136 struct inet_sock *inet = inet_sk(sk);
137
138 __skb_queue_purge(&sk->sk_receive_queue);
139 __skb_queue_purge(&sk->sk_error_queue);
140
141 sk_mem_reclaim(sk);
142
143 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
144 pr_err("Attempt to release TCP socket in state %d %p\n",
145 sk->sk_state, sk);
146 return;
147 }
148 if (!sock_flag(sk, SOCK_DEAD)) {
149 pr_err("Attempt to release alive inet socket %p\n", sk);
150 return;
151 }
152
153 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
154 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
155 WARN_ON(sk->sk_wmem_queued);
156 WARN_ON(sk->sk_forward_alloc);
157
158 kfree(rcu_dereference_protected(inet->inet_opt, 1));
159 dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
160 dst_release(sk->sk_rx_dst);
161 sk_refcnt_debug_dec(sk);
162}
163EXPORT_SYMBOL(inet_sock_destruct);
164
165
166
167
168
169
170
171
172
173
174
175static int inet_autobind(struct sock *sk)
176{
177 struct inet_sock *inet;
178
179 lock_sock(sk);
180 inet = inet_sk(sk);
181 if (!inet->inet_num) {
182 if (sk->sk_prot->get_port(sk, 0)) {
183 release_sock(sk);
184 return -EAGAIN;
185 }
186 inet->inet_sport = htons(inet->inet_num);
187 }
188 release_sock(sk);
189 return 0;
190}
191
192
193
194
195int inet_listen(struct socket *sock, int backlog)
196{
197 struct sock *sk = sock->sk;
198 unsigned char old_state;
199 int err;
200
201 lock_sock(sk);
202
203 err = -EINVAL;
204 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
205 goto out;
206
207 old_state = sk->sk_state;
208 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
209 goto out;
210
211
212
213
214 if (old_state != TCP_LISTEN) {
215
216
217
218
219
220
221
222 if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 &&
223 inet_csk(sk)->icsk_accept_queue.fastopenq == NULL) {
224 if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0)
225 err = fastopen_init_queue(sk, backlog);
226 else if ((sysctl_tcp_fastopen &
227 TFO_SERVER_WO_SOCKOPT2) != 0)
228 err = fastopen_init_queue(sk,
229 ((uint)sysctl_tcp_fastopen) >> 16);
230 else
231 err = 0;
232 if (err)
233 goto out;
234 }
235 err = inet_csk_listen_start(sk, backlog);
236 if (err)
237 goto out;
238 }
239 sk->sk_max_ack_backlog = backlog;
240 err = 0;
241
242out:
243 release_sock(sk);
244 return err;
245}
246EXPORT_SYMBOL(inet_listen);
247
248u32 inet_ehash_secret __read_mostly;
249EXPORT_SYMBOL(inet_ehash_secret);
250
251u32 ipv6_hash_secret __read_mostly;
252EXPORT_SYMBOL(ipv6_hash_secret);
253
254
255
256
257
258void build_ehash_secret(void)
259{
260 u32 rnd;
261
262 do {
263 get_random_bytes(&rnd, sizeof(rnd));
264 } while (rnd == 0);
265
266 if (cmpxchg(&inet_ehash_secret, 0, rnd) == 0)
267 get_random_bytes(&ipv6_hash_secret, sizeof(ipv6_hash_secret));
268}
269EXPORT_SYMBOL(build_ehash_secret);
270
271
272
273
274
275static int inet_create(struct net *net, struct socket *sock, int protocol,
276 int kern)
277{
278 struct sock *sk;
279 struct inet_protosw *answer;
280 struct inet_sock *inet;
281 struct proto *answer_prot;
282 unsigned char answer_flags;
283 char answer_no_check;
284 int try_loading_module = 0;
285 int err;
286
287 if (unlikely(!inet_ehash_secret))
288 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
289 build_ehash_secret();
290
291 sock->state = SS_UNCONNECTED;
292
293
294lookup_protocol:
295 err = -ESOCKTNOSUPPORT;
296 rcu_read_lock();
297 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
298
299 err = 0;
300
301 if (protocol == answer->protocol) {
302 if (protocol != IPPROTO_IP)
303 break;
304 } else {
305
306 if (IPPROTO_IP == protocol) {
307 protocol = answer->protocol;
308 break;
309 }
310 if (IPPROTO_IP == answer->protocol)
311 break;
312 }
313 err = -EPROTONOSUPPORT;
314 }
315
316 if (unlikely(err)) {
317 if (try_loading_module < 2) {
318 rcu_read_unlock();
319
320
321
322
323 if (++try_loading_module == 1)
324 request_module("net-pf-%d-proto-%d-type-%d",
325 PF_INET, protocol, sock->type);
326
327
328
329
330 else
331 request_module("net-pf-%d-proto-%d",
332 PF_INET, protocol);
333 goto lookup_protocol;
334 } else
335 goto out_rcu_unlock;
336 }
337
338 err = -EPERM;
339 if (sock->type == SOCK_RAW && !kern &&
340 !ns_capable(net->user_ns, CAP_NET_RAW))
341 goto out_rcu_unlock;
342
343 sock->ops = answer->ops;
344 answer_prot = answer->prot;
345 answer_no_check = answer->no_check;
346 answer_flags = answer->flags;
347 rcu_read_unlock();
348
349 WARN_ON(answer_prot->slab == NULL);
350
351 err = -ENOBUFS;
352 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
353 if (sk == NULL)
354 goto out;
355
356 err = 0;
357 sk->sk_no_check = answer_no_check;
358 if (INET_PROTOSW_REUSE & answer_flags)
359 sk->sk_reuse = SK_CAN_REUSE;
360
361 inet = inet_sk(sk);
362 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
363
364 inet->nodefrag = 0;
365
366 if (SOCK_RAW == sock->type) {
367 inet->inet_num = protocol;
368 if (IPPROTO_RAW == protocol)
369 inet->hdrincl = 1;
370 }
371
372 if (ipv4_config.no_pmtu_disc)
373 inet->pmtudisc = IP_PMTUDISC_DONT;
374 else
375 inet->pmtudisc = IP_PMTUDISC_WANT;
376
377 inet->inet_id = 0;
378
379 sock_init_data(sock, sk);
380
381 sk->sk_destruct = inet_sock_destruct;
382 sk->sk_protocol = protocol;
383 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
384
385 inet->uc_ttl = -1;
386 inet->mc_loop = 1;
387 inet->mc_ttl = 1;
388 inet->mc_all = 1;
389 inet->mc_index = 0;
390 inet->mc_list = NULL;
391 inet->rcv_tos = 0;
392
393 sk_refcnt_debug_inc(sk);
394
395 if (inet->inet_num) {
396
397
398
399
400
401 inet->inet_sport = htons(inet->inet_num);
402
403 sk->sk_prot->hash(sk);
404 }
405
406 if (sk->sk_prot->init) {
407 err = sk->sk_prot->init(sk);
408 if (err)
409 sk_common_release(sk);
410 }
411out:
412 return err;
413out_rcu_unlock:
414 rcu_read_unlock();
415 goto out;
416}
417
418
419
420
421
422
423
424int inet_release(struct socket *sock)
425{
426 struct sock *sk = sock->sk;
427
428 if (sk) {
429 long timeout;
430
431 sock_rps_reset_flow(sk);
432
433
434 ip_mc_drop_socket(sk);
435
436
437
438
439
440
441
442
443 timeout = 0;
444 if (sock_flag(sk, SOCK_LINGER) &&
445 !(current->flags & PF_EXITING))
446 timeout = sk->sk_lingertime;
447 sock->sk = NULL;
448 sk->sk_prot->close(sk, timeout);
449 }
450 return 0;
451}
452EXPORT_SYMBOL(inet_release);
453
454
455int sysctl_ip_nonlocal_bind __read_mostly;
456EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
457
458int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
459{
460 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
461 struct sock *sk = sock->sk;
462 struct inet_sock *inet = inet_sk(sk);
463 struct net *net = sock_net(sk);
464 unsigned short snum;
465 int chk_addr_ret;
466 int err;
467
468
469 if (sk->sk_prot->bind) {
470 err = sk->sk_prot->bind(sk, uaddr, addr_len);
471 goto out;
472 }
473 err = -EINVAL;
474 if (addr_len < sizeof(struct sockaddr_in))
475 goto out;
476
477 if (addr->sin_family != AF_INET) {
478
479
480
481 err = -EAFNOSUPPORT;
482 if (addr->sin_family != AF_UNSPEC ||
483 addr->sin_addr.s_addr != htonl(INADDR_ANY))
484 goto out;
485 }
486
487 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
488
489
490
491
492
493
494
495
496 err = -EADDRNOTAVAIL;
497 if (!sysctl_ip_nonlocal_bind &&
498 !(inet->freebind || inet->transparent) &&
499 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
500 chk_addr_ret != RTN_LOCAL &&
501 chk_addr_ret != RTN_MULTICAST &&
502 chk_addr_ret != RTN_BROADCAST)
503 goto out;
504
505 snum = ntohs(addr->sin_port);
506 err = -EACCES;
507 if (snum && snum < PROT_SOCK &&
508 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
509 goto out;
510
511
512
513
514
515
516
517
518 lock_sock(sk);
519
520
521 err = -EINVAL;
522 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
523 goto out_release_sock;
524
525 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
526 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
527 inet->inet_saddr = 0;
528
529
530 if (sk->sk_prot->get_port(sk, snum)) {
531 inet->inet_saddr = inet->inet_rcv_saddr = 0;
532 err = -EADDRINUSE;
533 goto out_release_sock;
534 }
535
536 if (inet->inet_rcv_saddr)
537 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
538 if (snum)
539 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
540 inet->inet_sport = htons(inet->inet_num);
541 inet->inet_daddr = 0;
542 inet->inet_dport = 0;
543 sk_dst_reset(sk);
544 err = 0;
545out_release_sock:
546 release_sock(sk);
547out:
548 return err;
549}
550EXPORT_SYMBOL(inet_bind);
551
552int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
553 int addr_len, int flags)
554{
555 struct sock *sk = sock->sk;
556
557 if (addr_len < sizeof(uaddr->sa_family))
558 return -EINVAL;
559 if (uaddr->sa_family == AF_UNSPEC)
560 return sk->sk_prot->disconnect(sk, flags);
561
562 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
563 return -EAGAIN;
564 return sk->sk_prot->connect(sk, uaddr, addr_len);
565}
566EXPORT_SYMBOL(inet_dgram_connect);
567
568static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
569{
570 DEFINE_WAIT(wait);
571
572 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
573 sk->sk_write_pending += writebias;
574
575
576
577
578
579
580 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
581 release_sock(sk);
582 timeo = schedule_timeout(timeo);
583 lock_sock(sk);
584 if (signal_pending(current) || !timeo)
585 break;
586 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
587 }
588 finish_wait(sk_sleep(sk), &wait);
589 sk->sk_write_pending -= writebias;
590 return timeo;
591}
592
593
594
595
596
597int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
598 int addr_len, int flags)
599{
600 struct sock *sk = sock->sk;
601 int err;
602 long timeo;
603
604 if (addr_len < sizeof(uaddr->sa_family))
605 return -EINVAL;
606
607 if (uaddr->sa_family == AF_UNSPEC) {
608 err = sk->sk_prot->disconnect(sk, flags);
609 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
610 goto out;
611 }
612
613 switch (sock->state) {
614 default:
615 err = -EINVAL;
616 goto out;
617 case SS_CONNECTED:
618 err = -EISCONN;
619 goto out;
620 case SS_CONNECTING:
621 err = -EALREADY;
622
623 break;
624 case SS_UNCONNECTED:
625 err = -EISCONN;
626 if (sk->sk_state != TCP_CLOSE)
627 goto out;
628
629 err = sk->sk_prot->connect(sk, uaddr, addr_len);
630 if (err < 0)
631 goto out;
632
633 sock->state = SS_CONNECTING;
634
635
636
637
638
639 err = -EINPROGRESS;
640 break;
641 }
642
643 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
644
645 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
646 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
647 tcp_sk(sk)->fastopen_req &&
648 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
649
650
651 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
652 goto out;
653
654 err = sock_intr_errno(timeo);
655 if (signal_pending(current))
656 goto out;
657 }
658
659
660
661
662 if (sk->sk_state == TCP_CLOSE)
663 goto sock_error;
664
665
666
667
668
669
670 sock->state = SS_CONNECTED;
671 err = 0;
672out:
673 return err;
674
675sock_error:
676 err = sock_error(sk) ? : -ECONNABORTED;
677 sock->state = SS_UNCONNECTED;
678 if (sk->sk_prot->disconnect(sk, flags))
679 sock->state = SS_DISCONNECTING;
680 goto out;
681}
682EXPORT_SYMBOL(__inet_stream_connect);
683
684int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
685 int addr_len, int flags)
686{
687 int err;
688
689 lock_sock(sock->sk);
690 err = __inet_stream_connect(sock, uaddr, addr_len, flags);
691 release_sock(sock->sk);
692 return err;
693}
694EXPORT_SYMBOL(inet_stream_connect);
695
696
697
698
699
700int inet_accept(struct socket *sock, struct socket *newsock, int flags)
701{
702 struct sock *sk1 = sock->sk;
703 int err = -EINVAL;
704 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
705
706 if (!sk2)
707 goto do_err;
708
709 lock_sock(sk2);
710
711 sock_rps_record_flow(sk2);
712 WARN_ON(!((1 << sk2->sk_state) &
713 (TCPF_ESTABLISHED | TCPF_SYN_RECV |
714 TCPF_CLOSE_WAIT | TCPF_CLOSE)));
715
716 sock_graft(sk2, newsock);
717
718 newsock->state = SS_CONNECTED;
719 err = 0;
720 release_sock(sk2);
721do_err:
722 return err;
723}
724EXPORT_SYMBOL(inet_accept);
725
726
727
728
729
730int inet_getname(struct socket *sock, struct sockaddr *uaddr,
731 int *uaddr_len, int peer)
732{
733 struct sock *sk = sock->sk;
734 struct inet_sock *inet = inet_sk(sk);
735 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
736
737 sin->sin_family = AF_INET;
738 if (peer) {
739 if (!inet->inet_dport ||
740 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
741 peer == 1))
742 return -ENOTCONN;
743 sin->sin_port = inet->inet_dport;
744 sin->sin_addr.s_addr = inet->inet_daddr;
745 } else {
746 __be32 addr = inet->inet_rcv_saddr;
747 if (!addr)
748 addr = inet->inet_saddr;
749 sin->sin_port = inet->inet_sport;
750 sin->sin_addr.s_addr = addr;
751 }
752 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
753 *uaddr_len = sizeof(*sin);
754 return 0;
755}
756EXPORT_SYMBOL(inet_getname);
757
758int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
759 size_t size)
760{
761 struct sock *sk = sock->sk;
762
763 sock_rps_record_flow(sk);
764
765
766 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
767 inet_autobind(sk))
768 return -EAGAIN;
769
770 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
771}
772EXPORT_SYMBOL(inet_sendmsg);
773
774ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
775 size_t size, int flags)
776{
777 struct sock *sk = sock->sk;
778
779 sock_rps_record_flow(sk);
780
781
782 if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
783 inet_autobind(sk))
784 return -EAGAIN;
785
786 if (sk->sk_prot->sendpage)
787 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
788 return sock_no_sendpage(sock, page, offset, size, flags);
789}
790EXPORT_SYMBOL(inet_sendpage);
791
792int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
793 size_t size, int flags)
794{
795 struct sock *sk = sock->sk;
796 int addr_len = 0;
797 int err;
798
799 sock_rps_record_flow(sk);
800
801 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
802 flags & ~MSG_DONTWAIT, &addr_len);
803 if (err >= 0)
804 msg->msg_namelen = addr_len;
805 return err;
806}
807EXPORT_SYMBOL(inet_recvmsg);
808
809int inet_shutdown(struct socket *sock, int how)
810{
811 struct sock *sk = sock->sk;
812 int err = 0;
813
814
815
816
817 how++;
818
819
820 if ((how & ~SHUTDOWN_MASK) || !how)
821 return -EINVAL;
822
823 lock_sock(sk);
824 if (sock->state == SS_CONNECTING) {
825 if ((1 << sk->sk_state) &
826 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
827 sock->state = SS_DISCONNECTING;
828 else
829 sock->state = SS_CONNECTED;
830 }
831
832 switch (sk->sk_state) {
833 case TCP_CLOSE:
834 err = -ENOTCONN;
835
836
837 default:
838 sk->sk_shutdown |= how;
839 if (sk->sk_prot->shutdown)
840 sk->sk_prot->shutdown(sk, how);
841 break;
842
843
844
845
846
847 case TCP_LISTEN:
848 if (!(how & RCV_SHUTDOWN))
849 break;
850
851 case TCP_SYN_SENT:
852 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
853 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
854 break;
855 }
856
857
858 sk->sk_state_change(sk);
859 release_sock(sk);
860 return err;
861}
862EXPORT_SYMBOL(inet_shutdown);
863
864
865
866
867
868
869
870
871
872
873
874int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
875{
876 struct sock *sk = sock->sk;
877 int err = 0;
878 struct net *net = sock_net(sk);
879
880 switch (cmd) {
881 case SIOCGSTAMP:
882 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
883 break;
884 case SIOCGSTAMPNS:
885 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
886 break;
887 case SIOCADDRT:
888 case SIOCDELRT:
889 case SIOCRTMSG:
890 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
891 break;
892 case SIOCDARP:
893 case SIOCGARP:
894 case SIOCSARP:
895 err = arp_ioctl(net, cmd, (void __user *)arg);
896 break;
897 case SIOCGIFADDR:
898 case SIOCSIFADDR:
899 case SIOCGIFBRDADDR:
900 case SIOCSIFBRDADDR:
901 case SIOCGIFNETMASK:
902 case SIOCSIFNETMASK:
903 case SIOCGIFDSTADDR:
904 case SIOCSIFDSTADDR:
905 case SIOCSIFPFLAGS:
906 case SIOCGIFPFLAGS:
907 case SIOCSIFFLAGS:
908 err = devinet_ioctl(net, cmd, (void __user *)arg);
909 break;
910 default:
911 if (sk->sk_prot->ioctl)
912 err = sk->sk_prot->ioctl(sk, cmd, arg);
913 else
914 err = -ENOIOCTLCMD;
915 break;
916 }
917 return err;
918}
919EXPORT_SYMBOL(inet_ioctl);
920
921#ifdef CONFIG_COMPAT
922static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
923{
924 struct sock *sk = sock->sk;
925 int err = -ENOIOCTLCMD;
926
927 if (sk->sk_prot->compat_ioctl)
928 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
929
930 return err;
931}
932#endif
933
934const struct proto_ops inet_stream_ops = {
935 .family = PF_INET,
936 .owner = THIS_MODULE,
937 .release = inet_release,
938 .bind = inet_bind,
939 .connect = inet_stream_connect,
940 .socketpair = sock_no_socketpair,
941 .accept = inet_accept,
942 .getname = inet_getname,
943 .poll = tcp_poll,
944 .ioctl = inet_ioctl,
945 .listen = inet_listen,
946 .shutdown = inet_shutdown,
947 .setsockopt = sock_common_setsockopt,
948 .getsockopt = sock_common_getsockopt,
949 .sendmsg = inet_sendmsg,
950 .recvmsg = inet_recvmsg,
951 .mmap = sock_no_mmap,
952 .sendpage = inet_sendpage,
953 .splice_read = tcp_splice_read,
954#ifdef CONFIG_COMPAT
955 .compat_setsockopt = compat_sock_common_setsockopt,
956 .compat_getsockopt = compat_sock_common_getsockopt,
957 .compat_ioctl = inet_compat_ioctl,
958#endif
959};
960EXPORT_SYMBOL(inet_stream_ops);
961
962const struct proto_ops inet_dgram_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 = udp_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};
987EXPORT_SYMBOL(inet_dgram_ops);
988
989
990
991
992
993static const struct proto_ops inet_sockraw_ops = {
994 .family = PF_INET,
995 .owner = THIS_MODULE,
996 .release = inet_release,
997 .bind = inet_bind,
998 .connect = inet_dgram_connect,
999 .socketpair = sock_no_socketpair,
1000 .accept = sock_no_accept,
1001 .getname = inet_getname,
1002 .poll = datagram_poll,
1003 .ioctl = inet_ioctl,
1004 .listen = sock_no_listen,
1005 .shutdown = inet_shutdown,
1006 .setsockopt = sock_common_setsockopt,
1007 .getsockopt = sock_common_getsockopt,
1008 .sendmsg = inet_sendmsg,
1009 .recvmsg = inet_recvmsg,
1010 .mmap = sock_no_mmap,
1011 .sendpage = inet_sendpage,
1012#ifdef CONFIG_COMPAT
1013 .compat_setsockopt = compat_sock_common_setsockopt,
1014 .compat_getsockopt = compat_sock_common_getsockopt,
1015 .compat_ioctl = inet_compat_ioctl,
1016#endif
1017};
1018
1019static const struct net_proto_family inet_family_ops = {
1020 .family = PF_INET,
1021 .create = inet_create,
1022 .owner = THIS_MODULE,
1023};
1024
1025
1026
1027
1028static struct inet_protosw inetsw_array[] =
1029{
1030 {
1031 .type = SOCK_STREAM,
1032 .protocol = IPPROTO_TCP,
1033 .prot = &tcp_prot,
1034 .ops = &inet_stream_ops,
1035 .no_check = 0,
1036 .flags = INET_PROTOSW_PERMANENT |
1037 INET_PROTOSW_ICSK,
1038 },
1039
1040 {
1041 .type = SOCK_DGRAM,
1042 .protocol = IPPROTO_UDP,
1043 .prot = &udp_prot,
1044 .ops = &inet_dgram_ops,
1045 .no_check = UDP_CSUM_DEFAULT,
1046 .flags = INET_PROTOSW_PERMANENT,
1047 },
1048
1049 {
1050 .type = SOCK_DGRAM,
1051 .protocol = IPPROTO_ICMP,
1052 .prot = &ping_prot,
1053 .ops = &inet_dgram_ops,
1054 .no_check = UDP_CSUM_DEFAULT,
1055 .flags = INET_PROTOSW_REUSE,
1056 },
1057
1058 {
1059 .type = SOCK_RAW,
1060 .protocol = IPPROTO_IP,
1061 .prot = &raw_prot,
1062 .ops = &inet_sockraw_ops,
1063 .no_check = UDP_CSUM_DEFAULT,
1064 .flags = INET_PROTOSW_REUSE,
1065 }
1066};
1067
1068#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
1069
1070void inet_register_protosw(struct inet_protosw *p)
1071{
1072 struct list_head *lh;
1073 struct inet_protosw *answer;
1074 int protocol = p->protocol;
1075 struct list_head *last_perm;
1076
1077 spin_lock_bh(&inetsw_lock);
1078
1079 if (p->type >= SOCK_MAX)
1080 goto out_illegal;
1081
1082
1083 answer = NULL;
1084 last_perm = &inetsw[p->type];
1085 list_for_each(lh, &inetsw[p->type]) {
1086 answer = list_entry(lh, struct inet_protosw, list);
1087
1088
1089 if (INET_PROTOSW_PERMANENT & answer->flags) {
1090 if (protocol == answer->protocol)
1091 break;
1092 last_perm = lh;
1093 }
1094
1095 answer = NULL;
1096 }
1097 if (answer)
1098 goto out_permanent;
1099
1100
1101
1102
1103
1104
1105
1106 list_add_rcu(&p->list, last_perm);
1107out:
1108 spin_unlock_bh(&inetsw_lock);
1109
1110 return;
1111
1112out_permanent:
1113 pr_err("Attempt to override permanent protocol %d\n", protocol);
1114 goto out;
1115
1116out_illegal:
1117 pr_err("Ignoring attempt to register invalid socket type %d\n",
1118 p->type);
1119 goto out;
1120}
1121EXPORT_SYMBOL(inet_register_protosw);
1122
1123void inet_unregister_protosw(struct inet_protosw *p)
1124{
1125 if (INET_PROTOSW_PERMANENT & p->flags) {
1126 pr_err("Attempt to unregister permanent protocol %d\n",
1127 p->protocol);
1128 } else {
1129 spin_lock_bh(&inetsw_lock);
1130 list_del_rcu(&p->list);
1131 spin_unlock_bh(&inetsw_lock);
1132
1133 synchronize_net();
1134 }
1135}
1136EXPORT_SYMBOL(inet_unregister_protosw);
1137
1138
1139
1140
1141
1142int sysctl_ip_dynaddr __read_mostly;
1143
1144static int inet_sk_reselect_saddr(struct sock *sk)
1145{
1146 struct inet_sock *inet = inet_sk(sk);
1147 __be32 old_saddr = inet->inet_saddr;
1148 __be32 daddr = inet->inet_daddr;
1149 struct flowi4 *fl4;
1150 struct rtable *rt;
1151 __be32 new_saddr;
1152 struct ip_options_rcu *inet_opt;
1153
1154 inet_opt = rcu_dereference_protected(inet->inet_opt,
1155 sock_owned_by_user(sk));
1156 if (inet_opt && inet_opt->opt.srr)
1157 daddr = inet_opt->opt.faddr;
1158
1159
1160 fl4 = &inet->cork.fl.u.ip4;
1161 rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
1162 sk->sk_bound_dev_if, sk->sk_protocol,
1163 inet->inet_sport, inet->inet_dport, sk, false);
1164 if (IS_ERR(rt))
1165 return PTR_ERR(rt);
1166
1167 sk_setup_caps(sk, &rt->dst);
1168
1169 new_saddr = fl4->saddr;
1170
1171 if (new_saddr == old_saddr)
1172 return 0;
1173
1174 if (sysctl_ip_dynaddr > 1) {
1175 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1176 __func__, &old_saddr, &new_saddr);
1177 }
1178
1179 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189 __sk_prot_rehash(sk);
1190 return 0;
1191}
1192
1193int inet_sk_rebuild_header(struct sock *sk)
1194{
1195 struct inet_sock *inet = inet_sk(sk);
1196 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1197 __be32 daddr;
1198 struct ip_options_rcu *inet_opt;
1199 struct flowi4 *fl4;
1200 int err;
1201
1202
1203 if (rt)
1204 return 0;
1205
1206
1207 rcu_read_lock();
1208 inet_opt = rcu_dereference(inet->inet_opt);
1209 daddr = inet->inet_daddr;
1210 if (inet_opt && inet_opt->opt.srr)
1211 daddr = inet_opt->opt.faddr;
1212 rcu_read_unlock();
1213 fl4 = &inet->cork.fl.u.ip4;
1214 rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
1215 inet->inet_dport, inet->inet_sport,
1216 sk->sk_protocol, RT_CONN_FLAGS(sk),
1217 sk->sk_bound_dev_if);
1218 if (!IS_ERR(rt)) {
1219 err = 0;
1220 sk_setup_caps(sk, &rt->dst);
1221 } else {
1222 err = PTR_ERR(rt);
1223
1224
1225 sk->sk_route_caps = 0;
1226
1227
1228
1229
1230 if (!sysctl_ip_dynaddr ||
1231 sk->sk_state != TCP_SYN_SENT ||
1232 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1233 (err = inet_sk_reselect_saddr(sk)) != 0)
1234 sk->sk_err_soft = -err;
1235 }
1236
1237 return err;
1238}
1239EXPORT_SYMBOL(inet_sk_rebuild_header);
1240
1241static int inet_gso_send_check(struct sk_buff *skb)
1242{
1243 const struct net_offload *ops;
1244 const struct iphdr *iph;
1245 int proto;
1246 int ihl;
1247 int err = -EINVAL;
1248
1249 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1250 goto out;
1251
1252 iph = ip_hdr(skb);
1253 ihl = iph->ihl * 4;
1254 if (ihl < sizeof(*iph))
1255 goto out;
1256
1257 if (unlikely(!pskb_may_pull(skb, ihl)))
1258 goto out;
1259
1260 __skb_pull(skb, ihl);
1261 skb_reset_transport_header(skb);
1262 iph = ip_hdr(skb);
1263 proto = iph->protocol;
1264 err = -EPROTONOSUPPORT;
1265
1266 rcu_read_lock();
1267 ops = rcu_dereference(inet_offloads[proto]);
1268 if (likely(ops && ops->callbacks.gso_send_check))
1269 err = ops->callbacks.gso_send_check(skb);
1270 rcu_read_unlock();
1271
1272out:
1273 return err;
1274}
1275
1276static struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1277 netdev_features_t features)
1278{
1279 struct sk_buff *segs = ERR_PTR(-EINVAL);
1280 const struct net_offload *ops;
1281 struct iphdr *iph;
1282 int proto;
1283 int ihl;
1284 int id;
1285 unsigned int offset = 0;
1286 bool tunnel;
1287
1288 if (unlikely(skb_shinfo(skb)->gso_type &
1289 ~(SKB_GSO_TCPV4 |
1290 SKB_GSO_UDP |
1291 SKB_GSO_DODGY |
1292 SKB_GSO_TCP_ECN |
1293 SKB_GSO_GRE |
1294 SKB_GSO_TCPV6 |
1295 SKB_GSO_UDP_TUNNEL |
1296 SKB_GSO_MPLS |
1297 0)))
1298 goto out;
1299
1300 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1301 goto out;
1302
1303 iph = ip_hdr(skb);
1304 ihl = iph->ihl * 4;
1305 if (ihl < sizeof(*iph))
1306 goto out;
1307
1308 if (unlikely(!pskb_may_pull(skb, ihl)))
1309 goto out;
1310
1311 tunnel = !!skb->encapsulation;
1312
1313 __skb_pull(skb, ihl);
1314 skb_reset_transport_header(skb);
1315 iph = ip_hdr(skb);
1316 id = ntohs(iph->id);
1317 proto = iph->protocol;
1318 segs = ERR_PTR(-EPROTONOSUPPORT);
1319
1320 rcu_read_lock();
1321 ops = rcu_dereference(inet_offloads[proto]);
1322 if (likely(ops && ops->callbacks.gso_segment))
1323 segs = ops->callbacks.gso_segment(skb, features);
1324 rcu_read_unlock();
1325
1326 if (IS_ERR_OR_NULL(segs))
1327 goto out;
1328
1329 skb = segs;
1330 do {
1331 iph = ip_hdr(skb);
1332 if (!tunnel && proto == IPPROTO_UDP) {
1333 iph->id = htons(id);
1334 iph->frag_off = htons(offset >> 3);
1335 if (skb->next != NULL)
1336 iph->frag_off |= htons(IP_MF);
1337 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1338 } else {
1339 iph->id = htons(id++);
1340 }
1341 iph->tot_len = htons(skb->len - skb->mac_len);
1342 iph->check = 0;
1343 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
1344 } while ((skb = skb->next));
1345
1346out:
1347 return segs;
1348}
1349
1350static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1351 struct sk_buff *skb)
1352{
1353 const struct net_offload *ops;
1354 struct sk_buff **pp = NULL;
1355 struct sk_buff *p;
1356 const struct iphdr *iph;
1357 unsigned int hlen;
1358 unsigned int off;
1359 unsigned int id;
1360 int flush = 1;
1361 int proto;
1362
1363 off = skb_gro_offset(skb);
1364 hlen = off + sizeof(*iph);
1365 iph = skb_gro_header_fast(skb, off);
1366 if (skb_gro_header_hard(skb, hlen)) {
1367 iph = skb_gro_header_slow(skb, hlen, off);
1368 if (unlikely(!iph))
1369 goto out;
1370 }
1371
1372 proto = iph->protocol;
1373
1374 rcu_read_lock();
1375 ops = rcu_dereference(inet_offloads[proto]);
1376 if (!ops || !ops->callbacks.gro_receive)
1377 goto out_unlock;
1378
1379 if (*(u8 *)iph != 0x45)
1380 goto out_unlock;
1381
1382 if (unlikely(ip_fast_csum((u8 *)iph, 5)))
1383 goto out_unlock;
1384
1385 id = ntohl(*(__be32 *)&iph->id);
1386 flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
1387 id >>= 16;
1388
1389 for (p = *head; p; p = p->next) {
1390 struct iphdr *iph2;
1391
1392 if (!NAPI_GRO_CB(p)->same_flow)
1393 continue;
1394
1395 iph2 = ip_hdr(p);
1396
1397 if ((iph->protocol ^ iph2->protocol) |
1398 ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1399 ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
1400 NAPI_GRO_CB(p)->same_flow = 0;
1401 continue;
1402 }
1403
1404
1405 NAPI_GRO_CB(p)->flush |=
1406 (iph->ttl ^ iph2->ttl) |
1407 (iph->tos ^ iph2->tos) |
1408 (__force int)((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)) |
1409 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
1410
1411 NAPI_GRO_CB(p)->flush |= flush;
1412 }
1413
1414 NAPI_GRO_CB(skb)->flush |= flush;
1415 skb_gro_pull(skb, sizeof(*iph));
1416 skb_set_transport_header(skb, skb_gro_offset(skb));
1417
1418 pp = ops->callbacks.gro_receive(head, skb);
1419
1420out_unlock:
1421 rcu_read_unlock();
1422
1423out:
1424 NAPI_GRO_CB(skb)->flush |= flush;
1425
1426 return pp;
1427}
1428
1429static int inet_gro_complete(struct sk_buff *skb)
1430{
1431 __be16 newlen = htons(skb->len - skb_network_offset(skb));
1432 struct iphdr *iph = ip_hdr(skb);
1433 const struct net_offload *ops;
1434 int proto = iph->protocol;
1435 int err = -ENOSYS;
1436
1437 csum_replace2(&iph->check, iph->tot_len, newlen);
1438 iph->tot_len = newlen;
1439
1440 rcu_read_lock();
1441 ops = rcu_dereference(inet_offloads[proto]);
1442 if (WARN_ON(!ops || !ops->callbacks.gro_complete))
1443 goto out_unlock;
1444
1445 err = ops->callbacks.gro_complete(skb);
1446
1447out_unlock:
1448 rcu_read_unlock();
1449
1450 return err;
1451}
1452
1453int inet_ctl_sock_create(struct sock **sk, unsigned short family,
1454 unsigned short type, unsigned char protocol,
1455 struct net *net)
1456{
1457 struct socket *sock;
1458 int rc = sock_create_kern(family, type, protocol, &sock);
1459
1460 if (rc == 0) {
1461 *sk = sock->sk;
1462 (*sk)->sk_allocation = GFP_ATOMIC;
1463
1464
1465
1466
1467 (*sk)->sk_prot->unhash(*sk);
1468
1469 sk_change_net(*sk, net);
1470 }
1471 return rc;
1472}
1473EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1474
1475unsigned long snmp_fold_field(void __percpu *mib[], int offt)
1476{
1477 unsigned long res = 0;
1478 int i, j;
1479
1480 for_each_possible_cpu(i) {
1481 for (j = 0; j < SNMP_ARRAY_SZ; j++)
1482 res += *(((unsigned long *) per_cpu_ptr(mib[j], i)) + offt);
1483 }
1484 return res;
1485}
1486EXPORT_SYMBOL_GPL(snmp_fold_field);
1487
1488#if BITS_PER_LONG==32
1489
1490u64 snmp_fold_field64(void __percpu *mib[], int offt, size_t syncp_offset)
1491{
1492 u64 res = 0;
1493 int cpu;
1494
1495 for_each_possible_cpu(cpu) {
1496 void *bhptr;
1497 struct u64_stats_sync *syncp;
1498 u64 v;
1499 unsigned int start;
1500
1501 bhptr = per_cpu_ptr(mib[0], cpu);
1502 syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1503 do {
1504 start = u64_stats_fetch_begin_bh(syncp);
1505 v = *(((u64 *) bhptr) + offt);
1506 } while (u64_stats_fetch_retry_bh(syncp, start));
1507
1508 res += v;
1509 }
1510 return res;
1511}
1512EXPORT_SYMBOL_GPL(snmp_fold_field64);
1513#endif
1514
1515int snmp_mib_init(void __percpu *ptr[2], size_t mibsize, size_t align)
1516{
1517 BUG_ON(ptr == NULL);
1518 ptr[0] = __alloc_percpu(mibsize, align);
1519 if (!ptr[0])
1520 return -ENOMEM;
1521#if SNMP_ARRAY_SZ == 2
1522 ptr[1] = __alloc_percpu(mibsize, align);
1523 if (!ptr[1]) {
1524 free_percpu(ptr[0]);
1525 ptr[0] = NULL;
1526 return -ENOMEM;
1527 }
1528#endif
1529 return 0;
1530}
1531EXPORT_SYMBOL_GPL(snmp_mib_init);
1532
1533#ifdef CONFIG_IP_MULTICAST
1534static const struct net_protocol igmp_protocol = {
1535 .handler = igmp_rcv,
1536 .netns_ok = 1,
1537};
1538#endif
1539
1540static const struct net_protocol tcp_protocol = {
1541 .early_demux = tcp_v4_early_demux,
1542 .handler = tcp_v4_rcv,
1543 .err_handler = tcp_v4_err,
1544 .no_policy = 1,
1545 .netns_ok = 1,
1546};
1547
1548static const struct net_protocol udp_protocol = {
1549 .handler = udp_rcv,
1550 .err_handler = udp_err,
1551 .no_policy = 1,
1552 .netns_ok = 1,
1553};
1554
1555static const struct net_protocol icmp_protocol = {
1556 .handler = icmp_rcv,
1557 .err_handler = icmp_err,
1558 .no_policy = 1,
1559 .netns_ok = 1,
1560};
1561
1562static __net_init int ipv4_mib_init_net(struct net *net)
1563{
1564 if (snmp_mib_init((void __percpu **)net->mib.tcp_statistics,
1565 sizeof(struct tcp_mib),
1566 __alignof__(struct tcp_mib)) < 0)
1567 goto err_tcp_mib;
1568 if (snmp_mib_init((void __percpu **)net->mib.ip_statistics,
1569 sizeof(struct ipstats_mib),
1570 __alignof__(struct ipstats_mib)) < 0)
1571 goto err_ip_mib;
1572 if (snmp_mib_init((void __percpu **)net->mib.net_statistics,
1573 sizeof(struct linux_mib),
1574 __alignof__(struct linux_mib)) < 0)
1575 goto err_net_mib;
1576 if (snmp_mib_init((void __percpu **)net->mib.udp_statistics,
1577 sizeof(struct udp_mib),
1578 __alignof__(struct udp_mib)) < 0)
1579 goto err_udp_mib;
1580 if (snmp_mib_init((void __percpu **)net->mib.udplite_statistics,
1581 sizeof(struct udp_mib),
1582 __alignof__(struct udp_mib)) < 0)
1583 goto err_udplite_mib;
1584 if (snmp_mib_init((void __percpu **)net->mib.icmp_statistics,
1585 sizeof(struct icmp_mib),
1586 __alignof__(struct icmp_mib)) < 0)
1587 goto err_icmp_mib;
1588 net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1589 GFP_KERNEL);
1590 if (!net->mib.icmpmsg_statistics)
1591 goto err_icmpmsg_mib;
1592
1593 tcp_mib_init(net);
1594 return 0;
1595
1596err_icmpmsg_mib:
1597 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1598err_icmp_mib:
1599 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1600err_udplite_mib:
1601 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1602err_udp_mib:
1603 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1604err_net_mib:
1605 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1606err_ip_mib:
1607 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
1608err_tcp_mib:
1609 return -ENOMEM;
1610}
1611
1612static __net_exit void ipv4_mib_exit_net(struct net *net)
1613{
1614 kfree(net->mib.icmpmsg_statistics);
1615 snmp_mib_free((void __percpu **)net->mib.icmp_statistics);
1616 snmp_mib_free((void __percpu **)net->mib.udplite_statistics);
1617 snmp_mib_free((void __percpu **)net->mib.udp_statistics);
1618 snmp_mib_free((void __percpu **)net->mib.net_statistics);
1619 snmp_mib_free((void __percpu **)net->mib.ip_statistics);
1620 snmp_mib_free((void __percpu **)net->mib.tcp_statistics);
1621}
1622
1623static __net_initdata struct pernet_operations ipv4_mib_ops = {
1624 .init = ipv4_mib_init_net,
1625 .exit = ipv4_mib_exit_net,
1626};
1627
1628static int __init init_ipv4_mibs(void)
1629{
1630 return register_pernet_subsys(&ipv4_mib_ops);
1631}
1632
1633static int ipv4_proc_init(void);
1634
1635
1636
1637
1638
1639static struct packet_offload ip_packet_offload __read_mostly = {
1640 .type = cpu_to_be16(ETH_P_IP),
1641 .callbacks = {
1642 .gso_send_check = inet_gso_send_check,
1643 .gso_segment = inet_gso_segment,
1644 .gro_receive = inet_gro_receive,
1645 .gro_complete = inet_gro_complete,
1646 },
1647};
1648
1649static int __init ipv4_offload_init(void)
1650{
1651
1652
1653
1654 if (udpv4_offload_init() < 0)
1655 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
1656 if (tcpv4_offload_init() < 0)
1657 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
1658
1659 dev_add_offload(&ip_packet_offload);
1660 return 0;
1661}
1662
1663fs_initcall(ipv4_offload_init);
1664
1665static struct packet_type ip_packet_type __read_mostly = {
1666 .type = cpu_to_be16(ETH_P_IP),
1667 .func = ip_rcv,
1668};
1669
1670static int __init inet_init(void)
1671{
1672 struct inet_protosw *q;
1673 struct list_head *r;
1674 int rc = -EINVAL;
1675
1676 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
1677
1678 sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1679 if (!sysctl_local_reserved_ports)
1680 goto out;
1681
1682 rc = proto_register(&tcp_prot, 1);
1683 if (rc)
1684 goto out_free_reserved_ports;
1685
1686 rc = proto_register(&udp_prot, 1);
1687 if (rc)
1688 goto out_unregister_tcp_proto;
1689
1690 rc = proto_register(&raw_prot, 1);
1691 if (rc)
1692 goto out_unregister_udp_proto;
1693
1694 rc = proto_register(&ping_prot, 1);
1695 if (rc)
1696 goto out_unregister_raw_proto;
1697
1698
1699
1700
1701
1702 (void)sock_register(&inet_family_ops);
1703
1704#ifdef CONFIG_SYSCTL
1705 ip_static_sysctl_init();
1706#endif
1707
1708 tcp_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1709
1710
1711
1712
1713
1714 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1715 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
1716 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1717 pr_crit("%s: Cannot add UDP protocol\n", __func__);
1718 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1719 pr_crit("%s: Cannot add TCP protocol\n", __func__);
1720#ifdef CONFIG_IP_MULTICAST
1721 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1722 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
1723#endif
1724
1725
1726 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1727 INIT_LIST_HEAD(r);
1728
1729 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1730 inet_register_protosw(q);
1731
1732
1733
1734
1735
1736 arp_init();
1737
1738
1739
1740
1741
1742 ip_init();
1743
1744 tcp_v4_init();
1745
1746
1747 tcp_init();
1748
1749
1750 udp_init();
1751
1752
1753 udplite4_register();
1754
1755 ping_init();
1756
1757
1758
1759
1760
1761 if (icmp_init() < 0)
1762 panic("Failed to create the ICMP control socket.\n");
1763
1764
1765
1766
1767#if defined(CONFIG_IP_MROUTE)
1768 if (ip_mr_init())
1769 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
1770#endif
1771
1772
1773
1774
1775 if (init_ipv4_mibs())
1776 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
1777
1778 ipv4_proc_init();
1779
1780 ipfrag_init();
1781
1782 dev_add_pack(&ip_packet_type);
1783
1784 rc = 0;
1785out:
1786 return rc;
1787out_unregister_raw_proto:
1788 proto_unregister(&raw_prot);
1789out_unregister_udp_proto:
1790 proto_unregister(&udp_prot);
1791out_unregister_tcp_proto:
1792 proto_unregister(&tcp_prot);
1793out_free_reserved_ports:
1794 kfree(sysctl_local_reserved_ports);
1795 goto out;
1796}
1797
1798fs_initcall(inet_init);
1799
1800
1801
1802#ifdef CONFIG_PROC_FS
1803static int __init ipv4_proc_init(void)
1804{
1805 int rc = 0;
1806
1807 if (raw_proc_init())
1808 goto out_raw;
1809 if (tcp4_proc_init())
1810 goto out_tcp;
1811 if (udp4_proc_init())
1812 goto out_udp;
1813 if (ping_proc_init())
1814 goto out_ping;
1815 if (ip_misc_proc_init())
1816 goto out_misc;
1817out:
1818 return rc;
1819out_misc:
1820 ping_proc_exit();
1821out_ping:
1822 udp4_proc_exit();
1823out_udp:
1824 tcp4_proc_exit();
1825out_tcp:
1826 raw_proc_exit();
1827out_raw:
1828 rc = -ENOMEM;
1829 goto out;
1830}
1831
1832#else
1833static int __init ipv4_proc_init(void)
1834{
1835 return 0;
1836}
1837#endif
1838
1839MODULE_ALIAS_NETPROTO(PF_INET);
1840
1841