blob: b7b7278d801cf23eaf06b740d87f4e99bd4044d3 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * PF_INET protocol family socket handler.
7 *
8 * Version: $Id: af_inet.c,v 1.137 2002/02/01 22:01:03 davem Exp $
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Changes (see also sock.c)
16 *
17 * piggy,
18 * Karl Knutson : Socket protocol table
19 * A.N.Kuznetsov : Socket death error in accept().
20 * John Richardson : Fix non blocking error in connect()
21 * so sockets that fail to connect
22 * don't return -EINPROGRESS.
23 * Alan Cox : Asynchronous I/O support
24 * Alan Cox : Keep correct socket pointer on sock
25 * structures
26 * when accept() ed
27 * Alan Cox : Semantics of SO_LINGER aren't state
28 * moved to close when you look carefully.
29 * With this fixed and the accept bug fixed
30 * some RPC stuff seems happier.
31 * Niibe Yutaka : 4.4BSD style write async I/O
32 * Alan Cox,
33 * Tony Gale : Fixed reuse semantics.
34 * Alan Cox : bind() shouldn't abort existing but dead
35 * sockets. Stops FTP netin:.. I hope.
36 * Alan Cox : bind() works correctly for RAW sockets.
37 * Note that FreeBSD at least was broken
38 * in this respect so be careful with
39 * compatibility tests...
40 * Alan Cox : routing cache support
41 * Alan Cox : memzero the socket structure for
42 * compactness.
43 * Matt Day : nonblock connect error handler
44 * Alan Cox : Allow large numbers of pending sockets
45 * (eg for big web sites), but only if
46 * specifically application requested.
47 * Alan Cox : New buffering throughout IP. Used
48 * dumbly.
49 * Alan Cox : New buffering now used smartly.
50 * Alan Cox : BSD rather than common sense
51 * interpretation of listen.
52 * Germano Caronni : Assorted small races.
53 * Alan Cox : sendmsg/recvmsg basic support.
54 * Alan Cox : Only sendmsg/recvmsg now supported.
55 * Alan Cox : Locked down bind (see security list).
56 * Alan Cox : Loosened bind a little.
57 * Mike McLagan : ADD/DEL DLCI Ioctls
58 * Willy Konynenberg : Transparent proxying support.
59 * David S. Miller : New socket lookup architecture.
60 * Some other random speedups.
61 * Cyrus Durgin : Cleaned up file for kmod hacks.
62 * Andi Kleen : Fix inet_stream_connect TCP race.
63 *
64 * This program is free software; you can redistribute it and/or
65 * modify it under the terms of the GNU General Public License
66 * as published by the Free Software Foundation; either version
67 * 2 of the License, or (at your option) any later version.
68 */
69
Herbert Xuf4c50d92006-06-22 03:02:40 -070070#include <linux/err.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/errno.h>
72#include <linux/types.h>
73#include <linux/socket.h>
74#include <linux/in.h>
75#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#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>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080082#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#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>
David S. Millerb3da2cf2007-03-23 11:40:27 -070090#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#include <asm/uaccess.h>
93#include <asm/system.h>
94
95#include <linux/smp_lock.h>
96#include <linux/inet.h>
97#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020098#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include <linux/netdevice.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>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700105#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <net/tcp.h>
107#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800108#include <net/udplite.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#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#ifdef CONFIG_IP_MROUTE
117#include <linux/mroute.h>
118#endif
119
Eric Dumazetba899662005-08-26 12:05:31 -0700120DEFINE_SNMP_STAT(struct linux_mib, net_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122extern void ip_mc_drop_socket(struct sock *sk);
123
124/* The inetsw table contains everything that inet_create needs to
125 * build a new socket.
126 */
127static struct list_head inetsw[SOCK_MAX];
128static DEFINE_SPINLOCK(inetsw_lock);
129
130/* New destruction routine */
131
132void inet_sock_destruct(struct sock *sk)
133{
134 struct inet_sock *inet = inet_sk(sk);
135
136 __skb_queue_purge(&sk->sk_receive_queue);
137 __skb_queue_purge(&sk->sk_error_queue);
138
139 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
140 printk("Attempt to release TCP socket in state %d %p\n",
141 sk->sk_state, sk);
142 return;
143 }
144 if (!sock_flag(sk, SOCK_DEAD)) {
145 printk("Attempt to release alive inet socket %p\n", sk);
146 return;
147 }
148
149 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
150 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
151 BUG_TRAP(!sk->sk_wmem_queued);
152 BUG_TRAP(!sk->sk_forward_alloc);
153
Jesper Juhla51482b2005-11-08 09:41:34 -0800154 kfree(inet->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 dst_release(sk->sk_dst_cache);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700156 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157}
158
159/*
160 * The routines beyond this point handle the behaviour of an AF_INET
161 * socket object. Mostly it punts to the subprotocols of IP to do
162 * the work.
163 */
164
165/*
166 * Automatically bind an unbound socket.
167 */
168
169static int inet_autobind(struct sock *sk)
170{
171 struct inet_sock *inet;
172 /* We may need to bind the socket. */
173 lock_sock(sk);
174 inet = inet_sk(sk);
175 if (!inet->num) {
176 if (sk->sk_prot->get_port(sk, 0)) {
177 release_sock(sk);
178 return -EAGAIN;
179 }
180 inet->sport = htons(inet->num);
181 }
182 release_sock(sk);
183 return 0;
184}
185
186/*
187 * Move a socket into listening state.
188 */
189int inet_listen(struct socket *sock, int backlog)
190{
191 struct sock *sk = sock->sk;
192 unsigned char old_state;
193 int err;
194
195 lock_sock(sk);
196
197 err = -EINVAL;
198 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
199 goto out;
200
201 old_state = sk->sk_state;
202 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
203 goto out;
204
205 /* Really, if the socket is already in listen state
206 * we can only allow the backlog to be adjusted.
207 */
208 if (old_state != TCP_LISTEN) {
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800209 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 if (err)
211 goto out;
212 }
213 sk->sk_max_ack_backlog = backlog;
214 err = 0;
215
216out:
217 release_sock(sk);
218 return err;
219}
220
David S. Millerb3da2cf2007-03-23 11:40:27 -0700221u32 inet_ehash_secret;
222EXPORT_SYMBOL(inet_ehash_secret);
223
224void build_ehash_secret(void)
225{
226 while (!inet_ehash_secret)
227 get_random_bytes(&inet_ehash_secret, 4);
228}
229EXPORT_SYMBOL(build_ehash_secret);
230
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231/*
232 * Create an inet socket.
233 */
234
235static int inet_create(struct socket *sock, int protocol)
236{
237 struct sock *sk;
238 struct list_head *p;
239 struct inet_protosw *answer;
240 struct inet_sock *inet;
241 struct proto *answer_prot;
242 unsigned char answer_flags;
243 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700244 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800245 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
David S. Millerb3da2cf2007-03-23 11:40:27 -0700247 if (sock->type != SOCK_RAW &&
248 sock->type != SOCK_DGRAM &&
249 !inet_ehash_secret)
250 build_ehash_secret();
251
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 sock->state = SS_UNCONNECTED;
253
254 /* Look for the requested type/protocol pair. */
255 answer = NULL;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700256lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800257 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 rcu_read_lock();
259 list_for_each_rcu(p, &inetsw[sock->type]) {
260 answer = list_entry(p, struct inet_protosw, list);
261
262 /* Check the non-wild match. */
263 if (protocol == answer->protocol) {
264 if (protocol != IPPROTO_IP)
265 break;
266 } else {
267 /* Check for the two wild cases. */
268 if (IPPROTO_IP == protocol) {
269 protocol = answer->protocol;
270 break;
271 }
272 if (IPPROTO_IP == answer->protocol)
273 break;
274 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800275 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 answer = NULL;
277 }
278
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700279 if (unlikely(answer == NULL)) {
280 if (try_loading_module < 2) {
281 rcu_read_unlock();
282 /*
283 * Be more specific, e.g. net-pf-2-proto-132-type-1
284 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
285 */
286 if (++try_loading_module == 1)
287 request_module("net-pf-%d-proto-%d-type-%d",
288 PF_INET, protocol, sock->type);
289 /*
290 * Fall back to generic, e.g. net-pf-2-proto-132
291 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
292 */
293 else
294 request_module("net-pf-%d-proto-%d",
295 PF_INET, protocol);
296 goto lookup_protocol;
297 } else
298 goto out_rcu_unlock;
299 }
300
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 err = -EPERM;
302 if (answer->capability > 0 && !capable(answer->capability))
303 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
305 sock->ops = answer->ops;
306 answer_prot = answer->prot;
307 answer_no_check = answer->no_check;
308 answer_flags = answer->flags;
309 rcu_read_unlock();
310
311 BUG_TRAP(answer_prot->slab != NULL);
312
313 err = -ENOBUFS;
314 sk = sk_alloc(PF_INET, GFP_KERNEL, answer_prot, 1);
315 if (sk == NULL)
316 goto out;
317
318 err = 0;
319 sk->sk_no_check = answer_no_check;
320 if (INET_PROTOSW_REUSE & answer_flags)
321 sk->sk_reuse = 1;
322
323 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800324 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
326 if (SOCK_RAW == sock->type) {
327 inet->num = protocol;
328 if (IPPROTO_RAW == protocol)
329 inet->hdrincl = 1;
330 }
331
332 if (ipv4_config.no_pmtu_disc)
333 inet->pmtudisc = IP_PMTUDISC_DONT;
334 else
335 inet->pmtudisc = IP_PMTUDISC_WANT;
336
337 inet->id = 0;
338
339 sock_init_data(sock, sk);
340
341 sk->sk_destruct = inet_sock_destruct;
342 sk->sk_family = PF_INET;
343 sk->sk_protocol = protocol;
344 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
345
346 inet->uc_ttl = -1;
347 inet->mc_loop = 1;
348 inet->mc_ttl = 1;
349 inet->mc_index = 0;
350 inet->mc_list = NULL;
351
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700352 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
354 if (inet->num) {
355 /* It assumes that any protocol which allows
356 * the user to assign a number at socket
357 * creation time automatically
358 * shares.
359 */
360 inet->sport = htons(inet->num);
361 /* Add to protocol hash chains. */
362 sk->sk_prot->hash(sk);
363 }
364
365 if (sk->sk_prot->init) {
366 err = sk->sk_prot->init(sk);
367 if (err)
368 sk_common_release(sk);
369 }
370out:
371 return err;
372out_rcu_unlock:
373 rcu_read_unlock();
374 goto out;
375}
376
377
378/*
379 * The peer socket should always be NULL (or else). When we call this
380 * function we are destroying the object and from then on nobody
381 * should refer to it.
382 */
383int inet_release(struct socket *sock)
384{
385 struct sock *sk = sock->sk;
386
387 if (sk) {
388 long timeout;
389
390 /* Applications forget to leave groups before exiting */
391 ip_mc_drop_socket(sk);
392
393 /* If linger is set, we don't return until the close
394 * is complete. Otherwise we return immediately. The
395 * actually closing is done the same either way.
396 *
397 * If the close is due to the process exiting, we never
398 * linger..
399 */
400 timeout = 0;
401 if (sock_flag(sk, SOCK_LINGER) &&
402 !(current->flags & PF_EXITING))
403 timeout = sk->sk_lingertime;
404 sock->sk = NULL;
405 sk->sk_prot->close(sk, timeout);
406 }
407 return 0;
408}
409
410/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700411int sysctl_ip_nonlocal_bind __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
413int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
414{
415 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
416 struct sock *sk = sock->sk;
417 struct inet_sock *inet = inet_sk(sk);
418 unsigned short snum;
419 int chk_addr_ret;
420 int err;
421
422 /* If the socket has its own bind function then use it. (RAW) */
423 if (sk->sk_prot->bind) {
424 err = sk->sk_prot->bind(sk, uaddr, addr_len);
425 goto out;
426 }
427 err = -EINVAL;
428 if (addr_len < sizeof(struct sockaddr_in))
429 goto out;
430
431 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
432
433 /* Not specified by any standard per-se, however it breaks too
434 * many applications when removed. It is unfortunate since
435 * allowing applications to make a non-local bind solves
436 * several problems with systems using dynamic addressing.
437 * (ie. your servers still start up even if your ISDN link
438 * is temporarily down)
439 */
440 err = -EADDRNOTAVAIL;
441 if (!sysctl_ip_nonlocal_bind &&
442 !inet->freebind &&
443 addr->sin_addr.s_addr != INADDR_ANY &&
444 chk_addr_ret != RTN_LOCAL &&
445 chk_addr_ret != RTN_MULTICAST &&
446 chk_addr_ret != RTN_BROADCAST)
447 goto out;
448
449 snum = ntohs(addr->sin_port);
450 err = -EACCES;
451 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
452 goto out;
453
454 /* We keep a pair of addresses. rcv_saddr is the one
455 * used by hash lookups, and saddr is used for transmit.
456 *
457 * In the BSD API these are the same except where it
458 * would be illegal to use them (multicast/broadcast) in
459 * which case the sending device address is used.
460 */
461 lock_sock(sk);
462
463 /* Check these errors (active socket, double bind). */
464 err = -EINVAL;
465 if (sk->sk_state != TCP_CLOSE || inet->num)
466 goto out_release_sock;
467
468 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
469 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
470 inet->saddr = 0; /* Use device */
471
472 /* Make sure we are allowed to bind here. */
473 if (sk->sk_prot->get_port(sk, snum)) {
474 inet->saddr = inet->rcv_saddr = 0;
475 err = -EADDRINUSE;
476 goto out_release_sock;
477 }
478
479 if (inet->rcv_saddr)
480 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
481 if (snum)
482 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
483 inet->sport = htons(inet->num);
484 inet->daddr = 0;
485 inet->dport = 0;
486 sk_dst_reset(sk);
487 err = 0;
488out_release_sock:
489 release_sock(sk);
490out:
491 return err;
492}
493
494int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
495 int addr_len, int flags)
496{
497 struct sock *sk = sock->sk;
498
499 if (uaddr->sa_family == AF_UNSPEC)
500 return sk->sk_prot->disconnect(sk, flags);
501
502 if (!inet_sk(sk)->num && inet_autobind(sk))
503 return -EAGAIN;
504 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
505}
506
507static long inet_wait_for_connect(struct sock *sk, long timeo)
508{
509 DEFINE_WAIT(wait);
510
511 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
512
513 /* Basic assumption: if someone sets sk->sk_err, he _must_
514 * change state of the socket from TCP_SYN_*.
515 * Connect() does not allow to get error notifications
516 * without closing the socket.
517 */
518 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
519 release_sock(sk);
520 timeo = schedule_timeout(timeo);
521 lock_sock(sk);
522 if (signal_pending(current) || !timeo)
523 break;
524 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
525 }
526 finish_wait(sk->sk_sleep, &wait);
527 return timeo;
528}
529
530/*
531 * Connect to a remote host. There is regrettably still a little
532 * TCP 'magic' in here.
533 */
534int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
535 int addr_len, int flags)
536{
537 struct sock *sk = sock->sk;
538 int err;
539 long timeo;
540
541 lock_sock(sk);
542
543 if (uaddr->sa_family == AF_UNSPEC) {
544 err = sk->sk_prot->disconnect(sk, flags);
545 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
546 goto out;
547 }
548
549 switch (sock->state) {
550 default:
551 err = -EINVAL;
552 goto out;
553 case SS_CONNECTED:
554 err = -EISCONN;
555 goto out;
556 case SS_CONNECTING:
557 err = -EALREADY;
558 /* Fall out of switch with err, set for this state */
559 break;
560 case SS_UNCONNECTED:
561 err = -EISCONN;
562 if (sk->sk_state != TCP_CLOSE)
563 goto out;
564
565 err = sk->sk_prot->connect(sk, uaddr, addr_len);
566 if (err < 0)
567 goto out;
568
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900569 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
571 /* Just entered SS_CONNECTING state; the only
572 * difference is that return value in non-blocking
573 * case is EINPROGRESS, rather than EALREADY.
574 */
575 err = -EINPROGRESS;
576 break;
577 }
578
579 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
580
581 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
582 /* Error code is set above */
583 if (!timeo || !inet_wait_for_connect(sk, timeo))
584 goto out;
585
586 err = sock_intr_errno(timeo);
587 if (signal_pending(current))
588 goto out;
589 }
590
591 /* Connection was closed by RST, timeout, ICMP error
592 * or another process disconnected us.
593 */
594 if (sk->sk_state == TCP_CLOSE)
595 goto sock_error;
596
597 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
598 * and error was received after socket entered established state.
599 * Hence, it is handled normally after connect() return successfully.
600 */
601
602 sock->state = SS_CONNECTED;
603 err = 0;
604out:
605 release_sock(sk);
606 return err;
607
608sock_error:
609 err = sock_error(sk) ? : -ECONNABORTED;
610 sock->state = SS_UNCONNECTED;
611 if (sk->sk_prot->disconnect(sk, flags))
612 sock->state = SS_DISCONNECTING;
613 goto out;
614}
615
616/*
617 * Accept a pending connection. The TCP layer now gives BSD semantics.
618 */
619
620int inet_accept(struct socket *sock, struct socket *newsock, int flags)
621{
622 struct sock *sk1 = sock->sk;
623 int err = -EINVAL;
624 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
625
626 if (!sk2)
627 goto do_err;
628
629 lock_sock(sk2);
630
631 BUG_TRAP((1 << sk2->sk_state) &
632 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
633
634 sock_graft(sk2, newsock);
635
636 newsock->state = SS_CONNECTED;
637 err = 0;
638 release_sock(sk2);
639do_err:
640 return err;
641}
642
643
644/*
645 * This does both peername and sockname.
646 */
647int inet_getname(struct socket *sock, struct sockaddr *uaddr,
648 int *uaddr_len, int peer)
649{
650 struct sock *sk = sock->sk;
651 struct inet_sock *inet = inet_sk(sk);
652 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
653
654 sin->sin_family = AF_INET;
655 if (peer) {
656 if (!inet->dport ||
657 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
658 peer == 1))
659 return -ENOTCONN;
660 sin->sin_port = inet->dport;
661 sin->sin_addr.s_addr = inet->daddr;
662 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800663 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 if (!addr)
665 addr = inet->saddr;
666 sin->sin_port = inet->sport;
667 sin->sin_addr.s_addr = addr;
668 }
669 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
670 *uaddr_len = sizeof(*sin);
671 return 0;
672}
673
674int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
675 size_t size)
676{
677 struct sock *sk = sock->sk;
678
679 /* We may need to bind the socket. */
680 if (!inet_sk(sk)->num && inet_autobind(sk))
681 return -EAGAIN;
682
683 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
684}
685
686
687static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
688{
689 struct sock *sk = sock->sk;
690
691 /* We may need to bind the socket. */
692 if (!inet_sk(sk)->num && inet_autobind(sk))
693 return -EAGAIN;
694
695 if (sk->sk_prot->sendpage)
696 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
697 return sock_no_sendpage(sock, page, offset, size, flags);
698}
699
700
701int inet_shutdown(struct socket *sock, int how)
702{
703 struct sock *sk = sock->sk;
704 int err = 0;
705
706 /* This should really check to make sure
707 * the socket is a TCP socket. (WHY AC...)
708 */
709 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
710 1->2 bit 2 snds.
711 2->3 */
712 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
713 return -EINVAL;
714
715 lock_sock(sk);
716 if (sock->state == SS_CONNECTING) {
717 if ((1 << sk->sk_state) &
718 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
719 sock->state = SS_DISCONNECTING;
720 else
721 sock->state = SS_CONNECTED;
722 }
723
724 switch (sk->sk_state) {
725 case TCP_CLOSE:
726 err = -ENOTCONN;
727 /* Hack to wake up other listeners, who can poll for
728 POLLHUP, even on eg. unconnected UDP sockets -- RR */
729 default:
730 sk->sk_shutdown |= how;
731 if (sk->sk_prot->shutdown)
732 sk->sk_prot->shutdown(sk, how);
733 break;
734
735 /* Remaining two branches are temporary solution for missing
736 * close() in multithreaded environment. It is _not_ a good idea,
737 * but we have no choice until close() is repaired at VFS level.
738 */
739 case TCP_LISTEN:
740 if (!(how & RCV_SHUTDOWN))
741 break;
742 /* Fall through */
743 case TCP_SYN_SENT:
744 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
745 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
746 break;
747 }
748
749 /* Wake up anyone sleeping in poll. */
750 sk->sk_state_change(sk);
751 release_sock(sk);
752 return err;
753}
754
755/*
756 * ioctl() calls you can issue on an INET socket. Most of these are
757 * device configuration and stuff and very rarely used. Some ioctls
758 * pass on to the socket itself.
759 *
760 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
761 * loads the devconfigure module does its configuring and unloads it.
762 * There's a good 20K of config code hanging around the kernel.
763 */
764
765int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
766{
767 struct sock *sk = sock->sk;
768 int err = 0;
769
770 switch (cmd) {
771 case SIOCGSTAMP:
772 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
773 break;
Eric Dumazetae40eb12007-03-18 17:33:16 -0700774 case SIOCGSTAMPNS:
775 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
776 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 case SIOCADDRT:
778 case SIOCDELRT:
779 case SIOCRTMSG:
780 err = ip_rt_ioctl(cmd, (void __user *)arg);
781 break;
782 case SIOCDARP:
783 case SIOCGARP:
784 case SIOCSARP:
785 err = arp_ioctl(cmd, (void __user *)arg);
786 break;
787 case SIOCGIFADDR:
788 case SIOCSIFADDR:
789 case SIOCGIFBRDADDR:
790 case SIOCSIFBRDADDR:
791 case SIOCGIFNETMASK:
792 case SIOCSIFNETMASK:
793 case SIOCGIFDSTADDR:
794 case SIOCSIFDSTADDR:
795 case SIOCSIFPFLAGS:
796 case SIOCGIFPFLAGS:
797 case SIOCSIFFLAGS:
798 err = devinet_ioctl(cmd, (void __user *)arg);
799 break;
800 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800801 if (sk->sk_prot->ioctl)
802 err = sk->sk_prot->ioctl(sk, cmd, arg);
803 else
804 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 break;
806 }
807 return err;
808}
809
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800810const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800811 .family = PF_INET,
812 .owner = THIS_MODULE,
813 .release = inet_release,
814 .bind = inet_bind,
815 .connect = inet_stream_connect,
816 .socketpair = sock_no_socketpair,
817 .accept = inet_accept,
818 .getname = inet_getname,
819 .poll = tcp_poll,
820 .ioctl = inet_ioctl,
821 .listen = inet_listen,
822 .shutdown = inet_shutdown,
823 .setsockopt = sock_common_setsockopt,
824 .getsockopt = sock_common_getsockopt,
825 .sendmsg = inet_sendmsg,
826 .recvmsg = sock_common_recvmsg,
827 .mmap = sock_no_mmap,
828 .sendpage = tcp_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800829#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800830 .compat_setsockopt = compat_sock_common_setsockopt,
831 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800832#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833};
834
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800835const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800836 .family = PF_INET,
837 .owner = THIS_MODULE,
838 .release = inet_release,
839 .bind = inet_bind,
840 .connect = inet_dgram_connect,
841 .socketpair = sock_no_socketpair,
842 .accept = sock_no_accept,
843 .getname = inet_getname,
844 .poll = udp_poll,
845 .ioctl = inet_ioctl,
846 .listen = sock_no_listen,
847 .shutdown = inet_shutdown,
848 .setsockopt = sock_common_setsockopt,
849 .getsockopt = sock_common_getsockopt,
850 .sendmsg = inet_sendmsg,
851 .recvmsg = sock_common_recvmsg,
852 .mmap = sock_no_mmap,
853 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800854#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800855 .compat_setsockopt = compat_sock_common_setsockopt,
856 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800857#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858};
859
860/*
861 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
862 * udp_poll
863 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800864static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800865 .family = PF_INET,
866 .owner = THIS_MODULE,
867 .release = inet_release,
868 .bind = inet_bind,
869 .connect = inet_dgram_connect,
870 .socketpair = sock_no_socketpair,
871 .accept = sock_no_accept,
872 .getname = inet_getname,
873 .poll = datagram_poll,
874 .ioctl = inet_ioctl,
875 .listen = sock_no_listen,
876 .shutdown = inet_shutdown,
877 .setsockopt = sock_common_setsockopt,
878 .getsockopt = sock_common_getsockopt,
879 .sendmsg = inet_sendmsg,
880 .recvmsg = sock_common_recvmsg,
881 .mmap = sock_no_mmap,
882 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800883#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800884 .compat_setsockopt = compat_sock_common_setsockopt,
885 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800886#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887};
888
889static struct net_proto_family inet_family_ops = {
890 .family = PF_INET,
891 .create = inet_create,
892 .owner = THIS_MODULE,
893};
894
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895/* Upon startup we insert all the elements in inetsw_array[] into
896 * the linked list inetsw.
897 */
898static struct inet_protosw inetsw_array[] =
899{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900900 {
901 .type = SOCK_STREAM,
902 .protocol = IPPROTO_TCP,
903 .prot = &tcp_prot,
904 .ops = &inet_stream_ops,
905 .capability = -1,
906 .no_check = 0,
907 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800908 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900909 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900911 {
912 .type = SOCK_DGRAM,
913 .protocol = IPPROTO_UDP,
914 .prot = &udp_prot,
915 .ops = &inet_dgram_ops,
916 .capability = -1,
917 .no_check = UDP_CSUM_DEFAULT,
918 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900920
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921
922 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900923 .type = SOCK_RAW,
924 .protocol = IPPROTO_IP, /* wild card */
925 .prot = &raw_prot,
926 .ops = &inet_sockraw_ops,
927 .capability = CAP_NET_RAW,
928 .no_check = UDP_CSUM_DEFAULT,
929 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 }
931};
932
933#define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
934
935void inet_register_protosw(struct inet_protosw *p)
936{
937 struct list_head *lh;
938 struct inet_protosw *answer;
939 int protocol = p->protocol;
940 struct list_head *last_perm;
941
942 spin_lock_bh(&inetsw_lock);
943
944 if (p->type >= SOCK_MAX)
945 goto out_illegal;
946
947 /* If we are trying to override a permanent protocol, bail. */
948 answer = NULL;
949 last_perm = &inetsw[p->type];
950 list_for_each(lh, &inetsw[p->type]) {
951 answer = list_entry(lh, struct inet_protosw, list);
952
953 /* Check only the non-wild match. */
954 if (INET_PROTOSW_PERMANENT & answer->flags) {
955 if (protocol == answer->protocol)
956 break;
957 last_perm = lh;
958 }
959
960 answer = NULL;
961 }
962 if (answer)
963 goto out_permanent;
964
965 /* Add the new entry after the last permanent entry if any, so that
966 * the new entry does not override a permanent entry when matched with
967 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900968 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 * system automatically returns to the old behavior.
970 */
971 list_add_rcu(&p->list, last_perm);
972out:
973 spin_unlock_bh(&inetsw_lock);
974
975 synchronize_net();
976
977 return;
978
979out_permanent:
980 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
981 protocol);
982 goto out;
983
984out_illegal:
985 printk(KERN_ERR
986 "Ignoring attempt to register invalid socket type %d.\n",
987 p->type);
988 goto out;
989}
990
991void inet_unregister_protosw(struct inet_protosw *p)
992{
993 if (INET_PROTOSW_PERMANENT & p->flags) {
994 printk(KERN_ERR
995 "Attempt to unregister permanent protocol %d.\n",
996 p->protocol);
997 } else {
998 spin_lock_bh(&inetsw_lock);
999 list_del_rcu(&p->list);
1000 spin_unlock_bh(&inetsw_lock);
1001
1002 synchronize_net();
1003 }
1004}
1005
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001006/*
1007 * Shall we try to damage output packets if routing dev changes?
1008 */
1009
Brian Haleyab32ea52006-09-22 14:15:41 -07001010int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001011
1012static int inet_sk_reselect_saddr(struct sock *sk)
1013{
1014 struct inet_sock *inet = inet_sk(sk);
1015 int err;
1016 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001017 __be32 old_saddr = inet->saddr;
1018 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001019 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001020
1021 if (inet->opt && inet->opt->srr)
1022 daddr = inet->opt->faddr;
1023
1024 /* Query new route. */
1025 err = ip_route_connect(&rt, daddr, 0,
1026 RT_CONN_FLAGS(sk),
1027 sk->sk_bound_dev_if,
1028 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001029 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001030 if (err)
1031 return err;
1032
1033 sk_setup_caps(sk, &rt->u.dst);
1034
1035 new_saddr = rt->rt_src;
1036
1037 if (new_saddr == old_saddr)
1038 return 0;
1039
1040 if (sysctl_ip_dynaddr > 1) {
1041 printk(KERN_INFO "%s(): shifting inet->"
1042 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1043 __FUNCTION__,
1044 NIPQUAD(old_saddr),
1045 NIPQUAD(new_saddr));
1046 }
1047
1048 inet->saddr = inet->rcv_saddr = new_saddr;
1049
1050 /*
1051 * XXX The only one ugly spot where we need to
1052 * XXX really change the sockets identity after
1053 * XXX it has entered the hashes. -DaveM
1054 *
1055 * Besides that, it does not check for connection
1056 * uniqueness. Wait for troubles.
1057 */
1058 __sk_prot_rehash(sk);
1059 return 0;
1060}
1061
1062int inet_sk_rebuild_header(struct sock *sk)
1063{
1064 struct inet_sock *inet = inet_sk(sk);
1065 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001066 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001067 int err;
1068
1069 /* Route is OK, nothing to do. */
1070 if (rt)
1071 return 0;
1072
1073 /* Reroute. */
1074 daddr = inet->daddr;
1075 if (inet->opt && inet->opt->srr)
1076 daddr = inet->opt->faddr;
1077{
1078 struct flowi fl = {
1079 .oif = sk->sk_bound_dev_if,
1080 .nl_u = {
1081 .ip4_u = {
1082 .daddr = daddr,
1083 .saddr = inet->saddr,
1084 .tos = RT_CONN_FLAGS(sk),
1085 },
1086 },
1087 .proto = sk->sk_protocol,
1088 .uli_u = {
1089 .ports = {
1090 .sport = inet->sport,
1091 .dport = inet->dport,
1092 },
1093 },
1094 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001095
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001096 security_sk_classify_flow(sk, &fl);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001097 err = ip_route_output_flow(&rt, &fl, sk, 0);
1098}
1099 if (!err)
1100 sk_setup_caps(sk, &rt->u.dst);
1101 else {
1102 /* Routing failed... */
1103 sk->sk_route_caps = 0;
1104 /*
1105 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1106 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1107 */
1108 if (!sysctl_ip_dynaddr ||
1109 sk->sk_state != TCP_SYN_SENT ||
1110 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1111 (err = inet_sk_reselect_saddr(sk)) != 0)
1112 sk->sk_err_soft = -err;
1113 }
1114
1115 return err;
1116}
1117
1118EXPORT_SYMBOL(inet_sk_rebuild_header);
1119
Herbert Xua430a432006-07-08 13:34:56 -07001120static int inet_gso_send_check(struct sk_buff *skb)
1121{
1122 struct iphdr *iph;
1123 struct net_protocol *ops;
1124 int proto;
1125 int ihl;
1126 int err = -EINVAL;
1127
1128 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1129 goto out;
1130
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001131 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001132 ihl = iph->ihl * 4;
1133 if (ihl < sizeof(*iph))
1134 goto out;
1135
1136 if (unlikely(!pskb_may_pull(skb, ihl)))
1137 goto out;
1138
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001139 __skb_pull(skb, ihl);
1140 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001141 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001142 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1143 err = -EPROTONOSUPPORT;
1144
1145 rcu_read_lock();
1146 ops = rcu_dereference(inet_protos[proto]);
1147 if (likely(ops && ops->gso_send_check))
1148 err = ops->gso_send_check(skb);
1149 rcu_read_unlock();
1150
1151out:
1152 return err;
1153}
1154
Herbert Xu576a30e2006-06-27 13:22:38 -07001155static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001156{
1157 struct sk_buff *segs = ERR_PTR(-EINVAL);
1158 struct iphdr *iph;
1159 struct net_protocol *ops;
1160 int proto;
1161 int ihl;
1162 int id;
1163
Herbert Xubbcf4672006-07-03 19:38:35 -07001164 if (unlikely(skb_shinfo(skb)->gso_type &
1165 ~(SKB_GSO_TCPV4 |
1166 SKB_GSO_UDP |
1167 SKB_GSO_DODGY |
1168 SKB_GSO_TCP_ECN |
1169 0)))
1170 goto out;
1171
1172 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001173 goto out;
1174
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001175 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001176 ihl = iph->ihl * 4;
1177 if (ihl < sizeof(*iph))
1178 goto out;
1179
Herbert Xubbcf4672006-07-03 19:38:35 -07001180 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001181 goto out;
1182
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001183 __skb_pull(skb, ihl);
1184 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001185 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001186 id = ntohs(iph->id);
1187 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1188 segs = ERR_PTR(-EPROTONOSUPPORT);
1189
1190 rcu_read_lock();
1191 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001192 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001193 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001194 rcu_read_unlock();
1195
Herbert Xu576a30e2006-06-27 13:22:38 -07001196 if (!segs || unlikely(IS_ERR(segs)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001197 goto out;
1198
1199 skb = segs;
1200 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001201 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001202 iph->id = htons(id++);
1203 iph->tot_len = htons(skb->len - skb->mac_len);
1204 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001205 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001206 } while ((skb = skb->next));
1207
1208out:
1209 return segs;
1210}
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212#ifdef CONFIG_IP_MULTICAST
1213static struct net_protocol igmp_protocol = {
1214 .handler = igmp_rcv,
1215};
1216#endif
1217
1218static struct net_protocol tcp_protocol = {
1219 .handler = tcp_v4_rcv,
1220 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001221 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001222 .gso_segment = tcp_tso_segment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 .no_policy = 1,
1224};
1225
1226static struct net_protocol udp_protocol = {
1227 .handler = udp_rcv,
1228 .err_handler = udp_err,
1229 .no_policy = 1,
1230};
1231
1232static struct net_protocol icmp_protocol = {
1233 .handler = icmp_rcv,
1234};
1235
1236static int __init init_ipv4_mibs(void)
1237{
1238 net_statistics[0] = alloc_percpu(struct linux_mib);
1239 net_statistics[1] = alloc_percpu(struct linux_mib);
1240 ip_statistics[0] = alloc_percpu(struct ipstats_mib);
1241 ip_statistics[1] = alloc_percpu(struct ipstats_mib);
1242 icmp_statistics[0] = alloc_percpu(struct icmp_mib);
1243 icmp_statistics[1] = alloc_percpu(struct icmp_mib);
1244 tcp_statistics[0] = alloc_percpu(struct tcp_mib);
1245 tcp_statistics[1] = alloc_percpu(struct tcp_mib);
1246 udp_statistics[0] = alloc_percpu(struct udp_mib);
1247 udp_statistics[1] = alloc_percpu(struct udp_mib);
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001248 udplite_statistics[0] = alloc_percpu(struct udp_mib);
1249 udplite_statistics[1] = alloc_percpu(struct udp_mib);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 if (!
1251 (net_statistics[0] && net_statistics[1] && ip_statistics[0]
1252 && ip_statistics[1] && tcp_statistics[0] && tcp_statistics[1]
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001253 && udp_statistics[0] && udp_statistics[1]
1254 && udplite_statistics[0] && udplite_statistics[1] ) )
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 return -ENOMEM;
1256
1257 (void) tcp_mib_init();
1258
1259 return 0;
1260}
1261
1262static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
Herbert Xu30e224d2005-07-05 14:40:10 -07001264/*
1265 * IP protocol layer initialiser
1266 */
1267
1268static struct packet_type ip_packet_type = {
1269 .type = __constant_htons(ETH_P_IP),
1270 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001271 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001272 .gso_segment = inet_gso_segment,
Herbert Xu30e224d2005-07-05 14:40:10 -07001273};
1274
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275static int __init inet_init(void)
1276{
1277 struct sk_buff *dummy_skb;
1278 struct inet_protosw *q;
1279 struct list_head *r;
1280 int rc = -EINVAL;
1281
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001282 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283
1284 rc = proto_register(&tcp_prot, 1);
1285 if (rc)
1286 goto out;
1287
1288 rc = proto_register(&udp_prot, 1);
1289 if (rc)
1290 goto out_unregister_tcp_proto;
1291
1292 rc = proto_register(&raw_prot, 1);
1293 if (rc)
1294 goto out_unregister_udp_proto;
1295
1296 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001297 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 */
1299
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001300 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301
1302 /*
1303 * Add all the base protocols.
1304 */
1305
1306 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1307 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1308 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1309 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1310 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1311 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1312#ifdef CONFIG_IP_MULTICAST
1313 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1314 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1315#endif
1316
1317 /* Register the socket-side information for inet_create. */
1318 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1319 INIT_LIST_HEAD(r);
1320
1321 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1322 inet_register_protosw(q);
1323
1324 /*
1325 * Set the ARP module up
1326 */
1327
1328 arp_init();
1329
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001330 /*
1331 * Set the IP module up
1332 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
1334 ip_init();
1335
1336 tcp_v4_init(&inet_family_ops);
1337
1338 /* Setup TCP slab cache for open requests. */
1339 tcp_init();
1340
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001341 /* Add UDP-Lite (RFC 3828) */
1342 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
1344 /*
1345 * Set the ICMP layer up
1346 */
1347
1348 icmp_init(&inet_family_ops);
1349
1350 /*
1351 * Initialise the multicast router
1352 */
1353#if defined(CONFIG_IP_MROUTE)
1354 ip_mr_init();
1355#endif
1356 /*
1357 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001358 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
Stephen Hemminger132adf52007-03-08 20:44:43 -08001360 if (init_ipv4_mibs())
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001362
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 ipv4_proc_init();
1364
1365 ipfrag_init();
1366
Herbert Xu30e224d2005-07-05 14:40:10 -07001367 dev_add_pack(&ip_packet_type);
1368
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 rc = 0;
1370out:
1371 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372out_unregister_udp_proto:
1373 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001374out_unregister_tcp_proto:
1375 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 goto out;
1377}
1378
Heiko Carstensa536e072006-04-28 15:19:17 -07001379fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
1381/* ------------------------------------------------------------------------ */
1382
1383#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384static int __init ipv4_proc_init(void)
1385{
1386 int rc = 0;
1387
1388 if (raw_proc_init())
1389 goto out_raw;
1390 if (tcp4_proc_init())
1391 goto out_tcp;
1392 if (udp4_proc_init())
1393 goto out_udp;
1394 if (fib_proc_init())
1395 goto out_fib;
1396 if (ip_misc_proc_init())
1397 goto out_misc;
1398out:
1399 return rc;
1400out_misc:
1401 fib_proc_exit();
1402out_fib:
1403 udp4_proc_exit();
1404out_udp:
1405 tcp4_proc_exit();
1406out_tcp:
1407 raw_proc_exit();
1408out_raw:
1409 rc = -ENOMEM;
1410 goto out;
1411}
1412
1413#else /* CONFIG_PROC_FS */
1414static int __init ipv4_proc_init(void)
1415{
1416 return 0;
1417}
1418#endif /* CONFIG_PROC_FS */
1419
1420MODULE_ALIAS_NETPROTO(PF_INET);
1421
1422EXPORT_SYMBOL(inet_accept);
1423EXPORT_SYMBOL(inet_bind);
1424EXPORT_SYMBOL(inet_dgram_connect);
1425EXPORT_SYMBOL(inet_dgram_ops);
1426EXPORT_SYMBOL(inet_getname);
1427EXPORT_SYMBOL(inet_ioctl);
1428EXPORT_SYMBOL(inet_listen);
1429EXPORT_SYMBOL(inet_register_protosw);
1430EXPORT_SYMBOL(inet_release);
1431EXPORT_SYMBOL(inet_sendmsg);
1432EXPORT_SYMBOL(inet_shutdown);
1433EXPORT_SYMBOL(inet_sock_destruct);
1434EXPORT_SYMBOL(inet_stream_connect);
1435EXPORT_SYMBOL(inet_stream_ops);
1436EXPORT_SYMBOL(inet_unregister_protosw);
1437EXPORT_SYMBOL(net_statistics);
Neil Hormancdac4e02005-06-13 15:12:33 -07001438EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);