blob: 16aae8ef5555b65855f5437a729c13134cf9739b [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
Eric Dumazetbe776282007-03-27 13:53:04 -0700221u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700222EXPORT_SYMBOL(inet_ehash_secret);
223
Eric Dumazetbe776282007-03-27 13:53:04 -0700224/*
225 * inet_ehash_secret must be set exactly once
226 * Instead of using a dedicated spinlock, we (ab)use inetsw_lock
227 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700228void build_ehash_secret(void)
229{
Eric Dumazetbe776282007-03-27 13:53:04 -0700230 u32 rnd;
231 do {
232 get_random_bytes(&rnd, sizeof(rnd));
233 } while (rnd == 0);
234 spin_lock_bh(&inetsw_lock);
235 if (!inet_ehash_secret)
236 inet_ehash_secret = rnd;
237 spin_unlock_bh(&inetsw_lock);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700238}
239EXPORT_SYMBOL(build_ehash_secret);
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241/*
242 * Create an inet socket.
243 */
244
245static int inet_create(struct socket *sock, int protocol)
246{
247 struct sock *sk;
248 struct list_head *p;
249 struct inet_protosw *answer;
250 struct inet_sock *inet;
251 struct proto *answer_prot;
252 unsigned char answer_flags;
253 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700254 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800255 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
David S. Millerb3da2cf2007-03-23 11:40:27 -0700257 if (sock->type != SOCK_RAW &&
258 sock->type != SOCK_DGRAM &&
259 !inet_ehash_secret)
260 build_ehash_secret();
261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 sock->state = SS_UNCONNECTED;
263
264 /* Look for the requested type/protocol pair. */
265 answer = NULL;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700266lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800267 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 rcu_read_lock();
269 list_for_each_rcu(p, &inetsw[sock->type]) {
270 answer = list_entry(p, struct inet_protosw, list);
271
272 /* Check the non-wild match. */
273 if (protocol == answer->protocol) {
274 if (protocol != IPPROTO_IP)
275 break;
276 } else {
277 /* Check for the two wild cases. */
278 if (IPPROTO_IP == protocol) {
279 protocol = answer->protocol;
280 break;
281 }
282 if (IPPROTO_IP == answer->protocol)
283 break;
284 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800285 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 answer = NULL;
287 }
288
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700289 if (unlikely(answer == NULL)) {
290 if (try_loading_module < 2) {
291 rcu_read_unlock();
292 /*
293 * Be more specific, e.g. net-pf-2-proto-132-type-1
294 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
295 */
296 if (++try_loading_module == 1)
297 request_module("net-pf-%d-proto-%d-type-%d",
298 PF_INET, protocol, sock->type);
299 /*
300 * Fall back to generic, e.g. net-pf-2-proto-132
301 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
302 */
303 else
304 request_module("net-pf-%d-proto-%d",
305 PF_INET, protocol);
306 goto lookup_protocol;
307 } else
308 goto out_rcu_unlock;
309 }
310
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 err = -EPERM;
312 if (answer->capability > 0 && !capable(answer->capability))
313 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
315 sock->ops = answer->ops;
316 answer_prot = answer->prot;
317 answer_no_check = answer->no_check;
318 answer_flags = answer->flags;
319 rcu_read_unlock();
320
321 BUG_TRAP(answer_prot->slab != NULL);
322
323 err = -ENOBUFS;
324 sk = sk_alloc(PF_INET, GFP_KERNEL, answer_prot, 1);
325 if (sk == NULL)
326 goto out;
327
328 err = 0;
329 sk->sk_no_check = answer_no_check;
330 if (INET_PROTOSW_REUSE & answer_flags)
331 sk->sk_reuse = 1;
332
333 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800334 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
336 if (SOCK_RAW == sock->type) {
337 inet->num = protocol;
338 if (IPPROTO_RAW == protocol)
339 inet->hdrincl = 1;
340 }
341
342 if (ipv4_config.no_pmtu_disc)
343 inet->pmtudisc = IP_PMTUDISC_DONT;
344 else
345 inet->pmtudisc = IP_PMTUDISC_WANT;
346
347 inet->id = 0;
348
349 sock_init_data(sock, sk);
350
351 sk->sk_destruct = inet_sock_destruct;
352 sk->sk_family = PF_INET;
353 sk->sk_protocol = protocol;
354 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
355
356 inet->uc_ttl = -1;
357 inet->mc_loop = 1;
358 inet->mc_ttl = 1;
359 inet->mc_index = 0;
360 inet->mc_list = NULL;
361
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700362 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
364 if (inet->num) {
365 /* It assumes that any protocol which allows
366 * the user to assign a number at socket
367 * creation time automatically
368 * shares.
369 */
370 inet->sport = htons(inet->num);
371 /* Add to protocol hash chains. */
372 sk->sk_prot->hash(sk);
373 }
374
375 if (sk->sk_prot->init) {
376 err = sk->sk_prot->init(sk);
377 if (err)
378 sk_common_release(sk);
379 }
380out:
381 return err;
382out_rcu_unlock:
383 rcu_read_unlock();
384 goto out;
385}
386
387
388/*
389 * The peer socket should always be NULL (or else). When we call this
390 * function we are destroying the object and from then on nobody
391 * should refer to it.
392 */
393int inet_release(struct socket *sock)
394{
395 struct sock *sk = sock->sk;
396
397 if (sk) {
398 long timeout;
399
400 /* Applications forget to leave groups before exiting */
401 ip_mc_drop_socket(sk);
402
403 /* If linger is set, we don't return until the close
404 * is complete. Otherwise we return immediately. The
405 * actually closing is done the same either way.
406 *
407 * If the close is due to the process exiting, we never
408 * linger..
409 */
410 timeout = 0;
411 if (sock_flag(sk, SOCK_LINGER) &&
412 !(current->flags & PF_EXITING))
413 timeout = sk->sk_lingertime;
414 sock->sk = NULL;
415 sk->sk_prot->close(sk, timeout);
416 }
417 return 0;
418}
419
420/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700421int sysctl_ip_nonlocal_bind __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
423int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
424{
425 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
426 struct sock *sk = sock->sk;
427 struct inet_sock *inet = inet_sk(sk);
428 unsigned short snum;
429 int chk_addr_ret;
430 int err;
431
432 /* If the socket has its own bind function then use it. (RAW) */
433 if (sk->sk_prot->bind) {
434 err = sk->sk_prot->bind(sk, uaddr, addr_len);
435 goto out;
436 }
437 err = -EINVAL;
438 if (addr_len < sizeof(struct sockaddr_in))
439 goto out;
440
441 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
442
443 /* Not specified by any standard per-se, however it breaks too
444 * many applications when removed. It is unfortunate since
445 * allowing applications to make a non-local bind solves
446 * several problems with systems using dynamic addressing.
447 * (ie. your servers still start up even if your ISDN link
448 * is temporarily down)
449 */
450 err = -EADDRNOTAVAIL;
451 if (!sysctl_ip_nonlocal_bind &&
452 !inet->freebind &&
453 addr->sin_addr.s_addr != INADDR_ANY &&
454 chk_addr_ret != RTN_LOCAL &&
455 chk_addr_ret != RTN_MULTICAST &&
456 chk_addr_ret != RTN_BROADCAST)
457 goto out;
458
459 snum = ntohs(addr->sin_port);
460 err = -EACCES;
461 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
462 goto out;
463
464 /* We keep a pair of addresses. rcv_saddr is the one
465 * used by hash lookups, and saddr is used for transmit.
466 *
467 * In the BSD API these are the same except where it
468 * would be illegal to use them (multicast/broadcast) in
469 * which case the sending device address is used.
470 */
471 lock_sock(sk);
472
473 /* Check these errors (active socket, double bind). */
474 err = -EINVAL;
475 if (sk->sk_state != TCP_CLOSE || inet->num)
476 goto out_release_sock;
477
478 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
479 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
480 inet->saddr = 0; /* Use device */
481
482 /* Make sure we are allowed to bind here. */
483 if (sk->sk_prot->get_port(sk, snum)) {
484 inet->saddr = inet->rcv_saddr = 0;
485 err = -EADDRINUSE;
486 goto out_release_sock;
487 }
488
489 if (inet->rcv_saddr)
490 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
491 if (snum)
492 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
493 inet->sport = htons(inet->num);
494 inet->daddr = 0;
495 inet->dport = 0;
496 sk_dst_reset(sk);
497 err = 0;
498out_release_sock:
499 release_sock(sk);
500out:
501 return err;
502}
503
504int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
505 int addr_len, int flags)
506{
507 struct sock *sk = sock->sk;
508
509 if (uaddr->sa_family == AF_UNSPEC)
510 return sk->sk_prot->disconnect(sk, flags);
511
512 if (!inet_sk(sk)->num && inet_autobind(sk))
513 return -EAGAIN;
514 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
515}
516
517static long inet_wait_for_connect(struct sock *sk, long timeo)
518{
519 DEFINE_WAIT(wait);
520
521 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
522
523 /* Basic assumption: if someone sets sk->sk_err, he _must_
524 * change state of the socket from TCP_SYN_*.
525 * Connect() does not allow to get error notifications
526 * without closing the socket.
527 */
528 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
529 release_sock(sk);
530 timeo = schedule_timeout(timeo);
531 lock_sock(sk);
532 if (signal_pending(current) || !timeo)
533 break;
534 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
535 }
536 finish_wait(sk->sk_sleep, &wait);
537 return timeo;
538}
539
540/*
541 * Connect to a remote host. There is regrettably still a little
542 * TCP 'magic' in here.
543 */
544int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
545 int addr_len, int flags)
546{
547 struct sock *sk = sock->sk;
548 int err;
549 long timeo;
550
551 lock_sock(sk);
552
553 if (uaddr->sa_family == AF_UNSPEC) {
554 err = sk->sk_prot->disconnect(sk, flags);
555 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
556 goto out;
557 }
558
559 switch (sock->state) {
560 default:
561 err = -EINVAL;
562 goto out;
563 case SS_CONNECTED:
564 err = -EISCONN;
565 goto out;
566 case SS_CONNECTING:
567 err = -EALREADY;
568 /* Fall out of switch with err, set for this state */
569 break;
570 case SS_UNCONNECTED:
571 err = -EISCONN;
572 if (sk->sk_state != TCP_CLOSE)
573 goto out;
574
575 err = sk->sk_prot->connect(sk, uaddr, addr_len);
576 if (err < 0)
577 goto out;
578
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900579 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
581 /* Just entered SS_CONNECTING state; the only
582 * difference is that return value in non-blocking
583 * case is EINPROGRESS, rather than EALREADY.
584 */
585 err = -EINPROGRESS;
586 break;
587 }
588
589 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
590
591 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
592 /* Error code is set above */
593 if (!timeo || !inet_wait_for_connect(sk, timeo))
594 goto out;
595
596 err = sock_intr_errno(timeo);
597 if (signal_pending(current))
598 goto out;
599 }
600
601 /* Connection was closed by RST, timeout, ICMP error
602 * or another process disconnected us.
603 */
604 if (sk->sk_state == TCP_CLOSE)
605 goto sock_error;
606
607 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
608 * and error was received after socket entered established state.
609 * Hence, it is handled normally after connect() return successfully.
610 */
611
612 sock->state = SS_CONNECTED;
613 err = 0;
614out:
615 release_sock(sk);
616 return err;
617
618sock_error:
619 err = sock_error(sk) ? : -ECONNABORTED;
620 sock->state = SS_UNCONNECTED;
621 if (sk->sk_prot->disconnect(sk, flags))
622 sock->state = SS_DISCONNECTING;
623 goto out;
624}
625
626/*
627 * Accept a pending connection. The TCP layer now gives BSD semantics.
628 */
629
630int inet_accept(struct socket *sock, struct socket *newsock, int flags)
631{
632 struct sock *sk1 = sock->sk;
633 int err = -EINVAL;
634 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
635
636 if (!sk2)
637 goto do_err;
638
639 lock_sock(sk2);
640
641 BUG_TRAP((1 << sk2->sk_state) &
642 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
643
644 sock_graft(sk2, newsock);
645
646 newsock->state = SS_CONNECTED;
647 err = 0;
648 release_sock(sk2);
649do_err:
650 return err;
651}
652
653
654/*
655 * This does both peername and sockname.
656 */
657int inet_getname(struct socket *sock, struct sockaddr *uaddr,
658 int *uaddr_len, int peer)
659{
660 struct sock *sk = sock->sk;
661 struct inet_sock *inet = inet_sk(sk);
662 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
663
664 sin->sin_family = AF_INET;
665 if (peer) {
666 if (!inet->dport ||
667 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
668 peer == 1))
669 return -ENOTCONN;
670 sin->sin_port = inet->dport;
671 sin->sin_addr.s_addr = inet->daddr;
672 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800673 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 if (!addr)
675 addr = inet->saddr;
676 sin->sin_port = inet->sport;
677 sin->sin_addr.s_addr = addr;
678 }
679 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
680 *uaddr_len = sizeof(*sin);
681 return 0;
682}
683
684int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
685 size_t size)
686{
687 struct sock *sk = sock->sk;
688
689 /* We may need to bind the socket. */
690 if (!inet_sk(sk)->num && inet_autobind(sk))
691 return -EAGAIN;
692
693 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
694}
695
696
697static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
698{
699 struct sock *sk = sock->sk;
700
701 /* We may need to bind the socket. */
702 if (!inet_sk(sk)->num && inet_autobind(sk))
703 return -EAGAIN;
704
705 if (sk->sk_prot->sendpage)
706 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
707 return sock_no_sendpage(sock, page, offset, size, flags);
708}
709
710
711int inet_shutdown(struct socket *sock, int how)
712{
713 struct sock *sk = sock->sk;
714 int err = 0;
715
716 /* This should really check to make sure
717 * the socket is a TCP socket. (WHY AC...)
718 */
719 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
720 1->2 bit 2 snds.
721 2->3 */
722 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
723 return -EINVAL;
724
725 lock_sock(sk);
726 if (sock->state == SS_CONNECTING) {
727 if ((1 << sk->sk_state) &
728 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
729 sock->state = SS_DISCONNECTING;
730 else
731 sock->state = SS_CONNECTED;
732 }
733
734 switch (sk->sk_state) {
735 case TCP_CLOSE:
736 err = -ENOTCONN;
737 /* Hack to wake up other listeners, who can poll for
738 POLLHUP, even on eg. unconnected UDP sockets -- RR */
739 default:
740 sk->sk_shutdown |= how;
741 if (sk->sk_prot->shutdown)
742 sk->sk_prot->shutdown(sk, how);
743 break;
744
745 /* Remaining two branches are temporary solution for missing
746 * close() in multithreaded environment. It is _not_ a good idea,
747 * but we have no choice until close() is repaired at VFS level.
748 */
749 case TCP_LISTEN:
750 if (!(how & RCV_SHUTDOWN))
751 break;
752 /* Fall through */
753 case TCP_SYN_SENT:
754 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
755 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
756 break;
757 }
758
759 /* Wake up anyone sleeping in poll. */
760 sk->sk_state_change(sk);
761 release_sock(sk);
762 return err;
763}
764
765/*
766 * ioctl() calls you can issue on an INET socket. Most of these are
767 * device configuration and stuff and very rarely used. Some ioctls
768 * pass on to the socket itself.
769 *
770 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
771 * loads the devconfigure module does its configuring and unloads it.
772 * There's a good 20K of config code hanging around the kernel.
773 */
774
775int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
776{
777 struct sock *sk = sock->sk;
778 int err = 0;
779
780 switch (cmd) {
781 case SIOCGSTAMP:
782 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
783 break;
Eric Dumazetae40eb12007-03-18 17:33:16 -0700784 case SIOCGSTAMPNS:
785 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
786 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 case SIOCADDRT:
788 case SIOCDELRT:
789 case SIOCRTMSG:
790 err = ip_rt_ioctl(cmd, (void __user *)arg);
791 break;
792 case SIOCDARP:
793 case SIOCGARP:
794 case SIOCSARP:
795 err = arp_ioctl(cmd, (void __user *)arg);
796 break;
797 case SIOCGIFADDR:
798 case SIOCSIFADDR:
799 case SIOCGIFBRDADDR:
800 case SIOCSIFBRDADDR:
801 case SIOCGIFNETMASK:
802 case SIOCSIFNETMASK:
803 case SIOCGIFDSTADDR:
804 case SIOCSIFDSTADDR:
805 case SIOCSIFPFLAGS:
806 case SIOCGIFPFLAGS:
807 case SIOCSIFFLAGS:
808 err = devinet_ioctl(cmd, (void __user *)arg);
809 break;
810 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800811 if (sk->sk_prot->ioctl)
812 err = sk->sk_prot->ioctl(sk, cmd, arg);
813 else
814 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 break;
816 }
817 return err;
818}
819
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800820const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800821 .family = PF_INET,
822 .owner = THIS_MODULE,
823 .release = inet_release,
824 .bind = inet_bind,
825 .connect = inet_stream_connect,
826 .socketpair = sock_no_socketpair,
827 .accept = inet_accept,
828 .getname = inet_getname,
829 .poll = tcp_poll,
830 .ioctl = inet_ioctl,
831 .listen = inet_listen,
832 .shutdown = inet_shutdown,
833 .setsockopt = sock_common_setsockopt,
834 .getsockopt = sock_common_getsockopt,
835 .sendmsg = inet_sendmsg,
836 .recvmsg = sock_common_recvmsg,
837 .mmap = sock_no_mmap,
838 .sendpage = tcp_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800839#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800840 .compat_setsockopt = compat_sock_common_setsockopt,
841 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800842#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843};
844
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800845const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800846 .family = PF_INET,
847 .owner = THIS_MODULE,
848 .release = inet_release,
849 .bind = inet_bind,
850 .connect = inet_dgram_connect,
851 .socketpair = sock_no_socketpair,
852 .accept = sock_no_accept,
853 .getname = inet_getname,
854 .poll = udp_poll,
855 .ioctl = inet_ioctl,
856 .listen = sock_no_listen,
857 .shutdown = inet_shutdown,
858 .setsockopt = sock_common_setsockopt,
859 .getsockopt = sock_common_getsockopt,
860 .sendmsg = inet_sendmsg,
861 .recvmsg = sock_common_recvmsg,
862 .mmap = sock_no_mmap,
863 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800864#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800865 .compat_setsockopt = compat_sock_common_setsockopt,
866 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800867#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868};
869
870/*
871 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
872 * udp_poll
873 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800874static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800875 .family = PF_INET,
876 .owner = THIS_MODULE,
877 .release = inet_release,
878 .bind = inet_bind,
879 .connect = inet_dgram_connect,
880 .socketpair = sock_no_socketpair,
881 .accept = sock_no_accept,
882 .getname = inet_getname,
883 .poll = datagram_poll,
884 .ioctl = inet_ioctl,
885 .listen = sock_no_listen,
886 .shutdown = inet_shutdown,
887 .setsockopt = sock_common_setsockopt,
888 .getsockopt = sock_common_getsockopt,
889 .sendmsg = inet_sendmsg,
890 .recvmsg = sock_common_recvmsg,
891 .mmap = sock_no_mmap,
892 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800893#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800894 .compat_setsockopt = compat_sock_common_setsockopt,
895 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800896#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897};
898
899static struct net_proto_family inet_family_ops = {
900 .family = PF_INET,
901 .create = inet_create,
902 .owner = THIS_MODULE,
903};
904
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905/* Upon startup we insert all the elements in inetsw_array[] into
906 * the linked list inetsw.
907 */
908static struct inet_protosw inetsw_array[] =
909{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900910 {
911 .type = SOCK_STREAM,
912 .protocol = IPPROTO_TCP,
913 .prot = &tcp_prot,
914 .ops = &inet_stream_ops,
915 .capability = -1,
916 .no_check = 0,
917 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800918 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900919 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900921 {
922 .type = SOCK_DGRAM,
923 .protocol = IPPROTO_UDP,
924 .prot = &udp_prot,
925 .ops = &inet_dgram_ops,
926 .capability = -1,
927 .no_check = UDP_CSUM_DEFAULT,
928 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900930
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
932 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900933 .type = SOCK_RAW,
934 .protocol = IPPROTO_IP, /* wild card */
935 .prot = &raw_prot,
936 .ops = &inet_sockraw_ops,
937 .capability = CAP_NET_RAW,
938 .no_check = UDP_CSUM_DEFAULT,
939 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 }
941};
942
943#define INETSW_ARRAY_LEN (sizeof(inetsw_array) / sizeof(struct inet_protosw))
944
945void inet_register_protosw(struct inet_protosw *p)
946{
947 struct list_head *lh;
948 struct inet_protosw *answer;
949 int protocol = p->protocol;
950 struct list_head *last_perm;
951
952 spin_lock_bh(&inetsw_lock);
953
954 if (p->type >= SOCK_MAX)
955 goto out_illegal;
956
957 /* If we are trying to override a permanent protocol, bail. */
958 answer = NULL;
959 last_perm = &inetsw[p->type];
960 list_for_each(lh, &inetsw[p->type]) {
961 answer = list_entry(lh, struct inet_protosw, list);
962
963 /* Check only the non-wild match. */
964 if (INET_PROTOSW_PERMANENT & answer->flags) {
965 if (protocol == answer->protocol)
966 break;
967 last_perm = lh;
968 }
969
970 answer = NULL;
971 }
972 if (answer)
973 goto out_permanent;
974
975 /* Add the new entry after the last permanent entry if any, so that
976 * the new entry does not override a permanent entry when matched with
977 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900978 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 * system automatically returns to the old behavior.
980 */
981 list_add_rcu(&p->list, last_perm);
982out:
983 spin_unlock_bh(&inetsw_lock);
984
985 synchronize_net();
986
987 return;
988
989out_permanent:
990 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
991 protocol);
992 goto out;
993
994out_illegal:
995 printk(KERN_ERR
996 "Ignoring attempt to register invalid socket type %d.\n",
997 p->type);
998 goto out;
999}
1000
1001void inet_unregister_protosw(struct inet_protosw *p)
1002{
1003 if (INET_PROTOSW_PERMANENT & p->flags) {
1004 printk(KERN_ERR
1005 "Attempt to unregister permanent protocol %d.\n",
1006 p->protocol);
1007 } else {
1008 spin_lock_bh(&inetsw_lock);
1009 list_del_rcu(&p->list);
1010 spin_unlock_bh(&inetsw_lock);
1011
1012 synchronize_net();
1013 }
1014}
1015
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001016/*
1017 * Shall we try to damage output packets if routing dev changes?
1018 */
1019
Brian Haleyab32ea52006-09-22 14:15:41 -07001020int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001021
1022static int inet_sk_reselect_saddr(struct sock *sk)
1023{
1024 struct inet_sock *inet = inet_sk(sk);
1025 int err;
1026 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001027 __be32 old_saddr = inet->saddr;
1028 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001029 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001030
1031 if (inet->opt && inet->opt->srr)
1032 daddr = inet->opt->faddr;
1033
1034 /* Query new route. */
1035 err = ip_route_connect(&rt, daddr, 0,
1036 RT_CONN_FLAGS(sk),
1037 sk->sk_bound_dev_if,
1038 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001039 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001040 if (err)
1041 return err;
1042
1043 sk_setup_caps(sk, &rt->u.dst);
1044
1045 new_saddr = rt->rt_src;
1046
1047 if (new_saddr == old_saddr)
1048 return 0;
1049
1050 if (sysctl_ip_dynaddr > 1) {
1051 printk(KERN_INFO "%s(): shifting inet->"
1052 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1053 __FUNCTION__,
1054 NIPQUAD(old_saddr),
1055 NIPQUAD(new_saddr));
1056 }
1057
1058 inet->saddr = inet->rcv_saddr = new_saddr;
1059
1060 /*
1061 * XXX The only one ugly spot where we need to
1062 * XXX really change the sockets identity after
1063 * XXX it has entered the hashes. -DaveM
1064 *
1065 * Besides that, it does not check for connection
1066 * uniqueness. Wait for troubles.
1067 */
1068 __sk_prot_rehash(sk);
1069 return 0;
1070}
1071
1072int inet_sk_rebuild_header(struct sock *sk)
1073{
1074 struct inet_sock *inet = inet_sk(sk);
1075 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001076 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001077 int err;
1078
1079 /* Route is OK, nothing to do. */
1080 if (rt)
1081 return 0;
1082
1083 /* Reroute. */
1084 daddr = inet->daddr;
1085 if (inet->opt && inet->opt->srr)
1086 daddr = inet->opt->faddr;
1087{
1088 struct flowi fl = {
1089 .oif = sk->sk_bound_dev_if,
1090 .nl_u = {
1091 .ip4_u = {
1092 .daddr = daddr,
1093 .saddr = inet->saddr,
1094 .tos = RT_CONN_FLAGS(sk),
1095 },
1096 },
1097 .proto = sk->sk_protocol,
1098 .uli_u = {
1099 .ports = {
1100 .sport = inet->sport,
1101 .dport = inet->dport,
1102 },
1103 },
1104 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001105
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001106 security_sk_classify_flow(sk, &fl);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001107 err = ip_route_output_flow(&rt, &fl, sk, 0);
1108}
1109 if (!err)
1110 sk_setup_caps(sk, &rt->u.dst);
1111 else {
1112 /* Routing failed... */
1113 sk->sk_route_caps = 0;
1114 /*
1115 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1116 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1117 */
1118 if (!sysctl_ip_dynaddr ||
1119 sk->sk_state != TCP_SYN_SENT ||
1120 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1121 (err = inet_sk_reselect_saddr(sk)) != 0)
1122 sk->sk_err_soft = -err;
1123 }
1124
1125 return err;
1126}
1127
1128EXPORT_SYMBOL(inet_sk_rebuild_header);
1129
Herbert Xua430a432006-07-08 13:34:56 -07001130static int inet_gso_send_check(struct sk_buff *skb)
1131{
1132 struct iphdr *iph;
1133 struct net_protocol *ops;
1134 int proto;
1135 int ihl;
1136 int err = -EINVAL;
1137
1138 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1139 goto out;
1140
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001141 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001142 ihl = iph->ihl * 4;
1143 if (ihl < sizeof(*iph))
1144 goto out;
1145
1146 if (unlikely(!pskb_may_pull(skb, ihl)))
1147 goto out;
1148
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001149 __skb_pull(skb, ihl);
1150 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001151 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001152 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1153 err = -EPROTONOSUPPORT;
1154
1155 rcu_read_lock();
1156 ops = rcu_dereference(inet_protos[proto]);
1157 if (likely(ops && ops->gso_send_check))
1158 err = ops->gso_send_check(skb);
1159 rcu_read_unlock();
1160
1161out:
1162 return err;
1163}
1164
Herbert Xu576a30e2006-06-27 13:22:38 -07001165static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001166{
1167 struct sk_buff *segs = ERR_PTR(-EINVAL);
1168 struct iphdr *iph;
1169 struct net_protocol *ops;
1170 int proto;
1171 int ihl;
1172 int id;
1173
Herbert Xubbcf4672006-07-03 19:38:35 -07001174 if (unlikely(skb_shinfo(skb)->gso_type &
1175 ~(SKB_GSO_TCPV4 |
1176 SKB_GSO_UDP |
1177 SKB_GSO_DODGY |
1178 SKB_GSO_TCP_ECN |
1179 0)))
1180 goto out;
1181
1182 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001183 goto out;
1184
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001185 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001186 ihl = iph->ihl * 4;
1187 if (ihl < sizeof(*iph))
1188 goto out;
1189
Herbert Xubbcf4672006-07-03 19:38:35 -07001190 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001191 goto out;
1192
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001193 __skb_pull(skb, ihl);
1194 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001195 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001196 id = ntohs(iph->id);
1197 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1198 segs = ERR_PTR(-EPROTONOSUPPORT);
1199
1200 rcu_read_lock();
1201 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001202 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001203 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001204 rcu_read_unlock();
1205
Herbert Xu576a30e2006-06-27 13:22:38 -07001206 if (!segs || unlikely(IS_ERR(segs)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001207 goto out;
1208
1209 skb = segs;
1210 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001211 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001212 iph->id = htons(id++);
1213 iph->tot_len = htons(skb->len - skb->mac_len);
1214 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001215 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001216 } while ((skb = skb->next));
1217
1218out:
1219 return segs;
1220}
1221
Herbert Xu5e0f0432007-04-24 21:53:35 -07001222unsigned long snmp_fold_field(void *mib[], int offt)
1223{
1224 unsigned long res = 0;
1225 int i;
1226
1227 for_each_possible_cpu(i) {
1228 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1229 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1230 }
1231 return res;
1232}
1233EXPORT_SYMBOL_GPL(snmp_fold_field);
1234
1235int snmp_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
1236{
1237 BUG_ON(ptr == NULL);
1238 ptr[0] = __alloc_percpu(mibsize);
1239 if (!ptr[0])
1240 goto err0;
1241 ptr[1] = __alloc_percpu(mibsize);
1242 if (!ptr[1])
1243 goto err1;
1244 return 0;
1245err1:
1246 free_percpu(ptr[0]);
1247 ptr[0] = NULL;
1248err0:
1249 return -ENOMEM;
1250}
1251EXPORT_SYMBOL_GPL(snmp_mib_init);
1252
1253void snmp_mib_free(void *ptr[2])
1254{
1255 BUG_ON(ptr == NULL);
1256 free_percpu(ptr[0]);
1257 free_percpu(ptr[1]);
1258 ptr[0] = ptr[1] = NULL;
1259}
1260EXPORT_SYMBOL_GPL(snmp_mib_free);
1261
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262#ifdef CONFIG_IP_MULTICAST
1263static struct net_protocol igmp_protocol = {
1264 .handler = igmp_rcv,
1265};
1266#endif
1267
1268static struct net_protocol tcp_protocol = {
1269 .handler = tcp_v4_rcv,
1270 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001271 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001272 .gso_segment = tcp_tso_segment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 .no_policy = 1,
1274};
1275
1276static struct net_protocol udp_protocol = {
1277 .handler = udp_rcv,
1278 .err_handler = udp_err,
1279 .no_policy = 1,
1280};
1281
1282static struct net_protocol icmp_protocol = {
1283 .handler = icmp_rcv,
1284};
1285
1286static int __init init_ipv4_mibs(void)
1287{
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001288 if (snmp_mib_init((void **)net_statistics,
1289 sizeof(struct linux_mib),
1290 __alignof__(struct linux_mib)) < 0)
1291 goto err_net_mib;
1292 if (snmp_mib_init((void **)ip_statistics,
1293 sizeof(struct ipstats_mib),
1294 __alignof__(struct ipstats_mib)) < 0)
1295 goto err_ip_mib;
1296 if (snmp_mib_init((void **)icmp_statistics,
1297 sizeof(struct icmp_mib),
1298 __alignof__(struct icmp_mib)) < 0)
1299 goto err_icmp_mib;
1300 if (snmp_mib_init((void **)tcp_statistics,
1301 sizeof(struct tcp_mib),
1302 __alignof__(struct tcp_mib)) < 0)
1303 goto err_tcp_mib;
1304 if (snmp_mib_init((void **)udp_statistics,
1305 sizeof(struct udp_mib),
1306 __alignof__(struct udp_mib)) < 0)
1307 goto err_udp_mib;
1308 if (snmp_mib_init((void **)udplite_statistics,
1309 sizeof(struct udp_mib),
1310 __alignof__(struct udp_mib)) < 0)
1311 goto err_udplite_mib;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001313 tcp_mib_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
1315 return 0;
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001316
1317err_udplite_mib:
1318 snmp_mib_free((void **)udp_statistics);
1319err_udp_mib:
1320 snmp_mib_free((void **)tcp_statistics);
1321err_tcp_mib:
1322 snmp_mib_free((void **)icmp_statistics);
1323err_icmp_mib:
1324 snmp_mib_free((void **)ip_statistics);
1325err_ip_mib:
1326 snmp_mib_free((void **)net_statistics);
1327err_net_mib:
1328 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329}
1330
1331static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Herbert Xu30e224d2005-07-05 14:40:10 -07001333/*
1334 * IP protocol layer initialiser
1335 */
1336
1337static struct packet_type ip_packet_type = {
1338 .type = __constant_htons(ETH_P_IP),
1339 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001340 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001341 .gso_segment = inet_gso_segment,
Herbert Xu30e224d2005-07-05 14:40:10 -07001342};
1343
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344static int __init inet_init(void)
1345{
1346 struct sk_buff *dummy_skb;
1347 struct inet_protosw *q;
1348 struct list_head *r;
1349 int rc = -EINVAL;
1350
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001351 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
1353 rc = proto_register(&tcp_prot, 1);
1354 if (rc)
1355 goto out;
1356
1357 rc = proto_register(&udp_prot, 1);
1358 if (rc)
1359 goto out_unregister_tcp_proto;
1360
1361 rc = proto_register(&raw_prot, 1);
1362 if (rc)
1363 goto out_unregister_udp_proto;
1364
1365 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001366 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 */
1368
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001369 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
1371 /*
1372 * Add all the base protocols.
1373 */
1374
1375 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1376 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1377 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1378 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1379 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1380 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1381#ifdef CONFIG_IP_MULTICAST
1382 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1383 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1384#endif
1385
1386 /* Register the socket-side information for inet_create. */
1387 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1388 INIT_LIST_HEAD(r);
1389
1390 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1391 inet_register_protosw(q);
1392
1393 /*
1394 * Set the ARP module up
1395 */
1396
1397 arp_init();
1398
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001399 /*
1400 * Set the IP module up
1401 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
1403 ip_init();
1404
1405 tcp_v4_init(&inet_family_ops);
1406
1407 /* Setup TCP slab cache for open requests. */
1408 tcp_init();
1409
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001410 /* Add UDP-Lite (RFC 3828) */
1411 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
1413 /*
1414 * Set the ICMP layer up
1415 */
1416
1417 icmp_init(&inet_family_ops);
1418
1419 /*
1420 * Initialise the multicast router
1421 */
1422#if defined(CONFIG_IP_MROUTE)
1423 ip_mr_init();
1424#endif
1425 /*
1426 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001427 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
Stephen Hemminger132adf52007-03-08 20:44:43 -08001429 if (init_ipv4_mibs())
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001431
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 ipv4_proc_init();
1433
1434 ipfrag_init();
1435
Herbert Xu30e224d2005-07-05 14:40:10 -07001436 dev_add_pack(&ip_packet_type);
1437
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 rc = 0;
1439out:
1440 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441out_unregister_udp_proto:
1442 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001443out_unregister_tcp_proto:
1444 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 goto out;
1446}
1447
Heiko Carstensa536e072006-04-28 15:19:17 -07001448fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449
1450/* ------------------------------------------------------------------------ */
1451
1452#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453static int __init ipv4_proc_init(void)
1454{
1455 int rc = 0;
1456
1457 if (raw_proc_init())
1458 goto out_raw;
1459 if (tcp4_proc_init())
1460 goto out_tcp;
1461 if (udp4_proc_init())
1462 goto out_udp;
1463 if (fib_proc_init())
1464 goto out_fib;
1465 if (ip_misc_proc_init())
1466 goto out_misc;
1467out:
1468 return rc;
1469out_misc:
1470 fib_proc_exit();
1471out_fib:
1472 udp4_proc_exit();
1473out_udp:
1474 tcp4_proc_exit();
1475out_tcp:
1476 raw_proc_exit();
1477out_raw:
1478 rc = -ENOMEM;
1479 goto out;
1480}
1481
1482#else /* CONFIG_PROC_FS */
1483static int __init ipv4_proc_init(void)
1484{
1485 return 0;
1486}
1487#endif /* CONFIG_PROC_FS */
1488
1489MODULE_ALIAS_NETPROTO(PF_INET);
1490
1491EXPORT_SYMBOL(inet_accept);
1492EXPORT_SYMBOL(inet_bind);
1493EXPORT_SYMBOL(inet_dgram_connect);
1494EXPORT_SYMBOL(inet_dgram_ops);
1495EXPORT_SYMBOL(inet_getname);
1496EXPORT_SYMBOL(inet_ioctl);
1497EXPORT_SYMBOL(inet_listen);
1498EXPORT_SYMBOL(inet_register_protosw);
1499EXPORT_SYMBOL(inet_release);
1500EXPORT_SYMBOL(inet_sendmsg);
1501EXPORT_SYMBOL(inet_shutdown);
1502EXPORT_SYMBOL(inet_sock_destruct);
1503EXPORT_SYMBOL(inet_stream_connect);
1504EXPORT_SYMBOL(inet_stream_ops);
1505EXPORT_SYMBOL(inet_unregister_protosw);
1506EXPORT_SYMBOL(net_statistics);
Neil Hormancdac4e02005-06-13 15:12:33 -07001507EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);