blob: 0e4b6eba23408cb0bd4054d6c17939385c56196a [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/inet.h>
96#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020097#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <linux/netdevice.h>
99#include <net/ip.h>
100#include <net/protocol.h>
101#include <net/arp.h>
102#include <net/route.h>
103#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700104#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/tcp.h>
106#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800107#include <net/udplite.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108#include <linux/skbuff.h>
109#include <net/sock.h>
110#include <net/raw.h>
111#include <net/icmp.h>
112#include <net/ipip.h>
113#include <net/inet_common.h>
114#include <net/xfrm.h>
115#ifdef CONFIG_IP_MROUTE
116#include <linux/mroute.h>
117#endif
118
Eric Dumazetba899662005-08-26 12:05:31 -0700119DEFINE_SNMP_STAT(struct linux_mib, net_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121extern void ip_mc_drop_socket(struct sock *sk);
122
123/* The inetsw table contains everything that inet_create needs to
124 * build a new socket.
125 */
126static struct list_head inetsw[SOCK_MAX];
127static DEFINE_SPINLOCK(inetsw_lock);
128
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800129struct ipv4_config ipv4_config;
130
131EXPORT_SYMBOL(ipv4_config);
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133/* New destruction routine */
134
135void inet_sock_destruct(struct sock *sk)
136{
137 struct inet_sock *inet = inet_sk(sk);
138
139 __skb_queue_purge(&sk->sk_receive_queue);
140 __skb_queue_purge(&sk->sk_error_queue);
141
142 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
143 printk("Attempt to release TCP socket in state %d %p\n",
144 sk->sk_state, sk);
145 return;
146 }
147 if (!sock_flag(sk, SOCK_DEAD)) {
148 printk("Attempt to release alive inet socket %p\n", sk);
149 return;
150 }
151
152 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
153 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
154 BUG_TRAP(!sk->sk_wmem_queued);
155 BUG_TRAP(!sk->sk_forward_alloc);
156
Jesper Juhla51482b2005-11-08 09:41:34 -0800157 kfree(inet->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 dst_release(sk->sk_dst_cache);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700159 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160}
161
162/*
163 * The routines beyond this point handle the behaviour of an AF_INET
164 * socket object. Mostly it punts to the subprotocols of IP to do
165 * the work.
166 */
167
168/*
169 * Automatically bind an unbound socket.
170 */
171
172static int inet_autobind(struct sock *sk)
173{
174 struct inet_sock *inet;
175 /* We may need to bind the socket. */
176 lock_sock(sk);
177 inet = inet_sk(sk);
178 if (!inet->num) {
179 if (sk->sk_prot->get_port(sk, 0)) {
180 release_sock(sk);
181 return -EAGAIN;
182 }
183 inet->sport = htons(inet->num);
184 }
185 release_sock(sk);
186 return 0;
187}
188
189/*
190 * Move a socket into listening state.
191 */
192int inet_listen(struct socket *sock, int backlog)
193{
194 struct sock *sk = sock->sk;
195 unsigned char old_state;
196 int err;
197
198 lock_sock(sk);
199
200 err = -EINVAL;
201 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
202 goto out;
203
204 old_state = sk->sk_state;
205 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
206 goto out;
207
208 /* Really, if the socket is already in listen state
209 * we can only allow the backlog to be adjusted.
210 */
211 if (old_state != TCP_LISTEN) {
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800212 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 if (err)
214 goto out;
215 }
216 sk->sk_max_ack_backlog = backlog;
217 err = 0;
218
219out:
220 release_sock(sk);
221 return err;
222}
223
Eric Dumazetbe776282007-03-27 13:53:04 -0700224u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700225EXPORT_SYMBOL(inet_ehash_secret);
226
Eric Dumazetbe776282007-03-27 13:53:04 -0700227/*
228 * inet_ehash_secret must be set exactly once
229 * Instead of using a dedicated spinlock, we (ab)use inetsw_lock
230 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700231void build_ehash_secret(void)
232{
Eric Dumazetbe776282007-03-27 13:53:04 -0700233 u32 rnd;
234 do {
235 get_random_bytes(&rnd, sizeof(rnd));
236 } while (rnd == 0);
237 spin_lock_bh(&inetsw_lock);
238 if (!inet_ehash_secret)
239 inet_ehash_secret = rnd;
240 spin_unlock_bh(&inetsw_lock);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700241}
242EXPORT_SYMBOL(build_ehash_secret);
243
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244/*
245 * Create an inet socket.
246 */
247
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700248static int inet_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249{
250 struct sock *sk;
251 struct list_head *p;
252 struct inet_protosw *answer;
253 struct inet_sock *inet;
254 struct proto *answer_prot;
255 unsigned char answer_flags;
256 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700257 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800258 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700260 if (net != &init_net)
261 return -EAFNOSUPPORT;
262
David S. Millerb3da2cf2007-03-23 11:40:27 -0700263 if (sock->type != SOCK_RAW &&
264 sock->type != SOCK_DGRAM &&
265 !inet_ehash_secret)
266 build_ehash_secret();
267
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 sock->state = SS_UNCONNECTED;
269
270 /* Look for the requested type/protocol pair. */
271 answer = NULL;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700272lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800273 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 rcu_read_lock();
275 list_for_each_rcu(p, &inetsw[sock->type]) {
276 answer = list_entry(p, struct inet_protosw, list);
277
278 /* Check the non-wild match. */
279 if (protocol == answer->protocol) {
280 if (protocol != IPPROTO_IP)
281 break;
282 } else {
283 /* Check for the two wild cases. */
284 if (IPPROTO_IP == protocol) {
285 protocol = answer->protocol;
286 break;
287 }
288 if (IPPROTO_IP == answer->protocol)
289 break;
290 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800291 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 answer = NULL;
293 }
294
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700295 if (unlikely(answer == NULL)) {
296 if (try_loading_module < 2) {
297 rcu_read_unlock();
298 /*
299 * Be more specific, e.g. net-pf-2-proto-132-type-1
300 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
301 */
302 if (++try_loading_module == 1)
303 request_module("net-pf-%d-proto-%d-type-%d",
304 PF_INET, protocol, sock->type);
305 /*
306 * Fall back to generic, e.g. net-pf-2-proto-132
307 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
308 */
309 else
310 request_module("net-pf-%d-proto-%d",
311 PF_INET, protocol);
312 goto lookup_protocol;
313 } else
314 goto out_rcu_unlock;
315 }
316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 err = -EPERM;
318 if (answer->capability > 0 && !capable(answer->capability))
319 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
321 sock->ops = answer->ops;
322 answer_prot = answer->prot;
323 answer_no_check = answer->no_check;
324 answer_flags = answer->flags;
325 rcu_read_unlock();
326
327 BUG_TRAP(answer_prot->slab != NULL);
328
329 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700330 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 if (sk == NULL)
332 goto out;
333
334 err = 0;
335 sk->sk_no_check = answer_no_check;
336 if (INET_PROTOSW_REUSE & answer_flags)
337 sk->sk_reuse = 1;
338
339 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800340 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341
342 if (SOCK_RAW == sock->type) {
343 inet->num = protocol;
344 if (IPPROTO_RAW == protocol)
345 inet->hdrincl = 1;
346 }
347
348 if (ipv4_config.no_pmtu_disc)
349 inet->pmtudisc = IP_PMTUDISC_DONT;
350 else
351 inet->pmtudisc = IP_PMTUDISC_WANT;
352
353 inet->id = 0;
354
355 sock_init_data(sock, sk);
356
357 sk->sk_destruct = inet_sock_destruct;
358 sk->sk_family = PF_INET;
359 sk->sk_protocol = protocol;
360 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
361
362 inet->uc_ttl = -1;
363 inet->mc_loop = 1;
364 inet->mc_ttl = 1;
365 inet->mc_index = 0;
366 inet->mc_list = NULL;
367
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700368 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
370 if (inet->num) {
371 /* It assumes that any protocol which allows
372 * the user to assign a number at socket
373 * creation time automatically
374 * shares.
375 */
376 inet->sport = htons(inet->num);
377 /* Add to protocol hash chains. */
378 sk->sk_prot->hash(sk);
379 }
380
381 if (sk->sk_prot->init) {
382 err = sk->sk_prot->init(sk);
383 if (err)
384 sk_common_release(sk);
385 }
386out:
387 return err;
388out_rcu_unlock:
389 rcu_read_unlock();
390 goto out;
391}
392
393
394/*
395 * The peer socket should always be NULL (or else). When we call this
396 * function we are destroying the object and from then on nobody
397 * should refer to it.
398 */
399int inet_release(struct socket *sock)
400{
401 struct sock *sk = sock->sk;
402
403 if (sk) {
404 long timeout;
405
406 /* Applications forget to leave groups before exiting */
407 ip_mc_drop_socket(sk);
408
409 /* If linger is set, we don't return until the close
410 * is complete. Otherwise we return immediately. The
411 * actually closing is done the same either way.
412 *
413 * If the close is due to the process exiting, we never
414 * linger..
415 */
416 timeout = 0;
417 if (sock_flag(sk, SOCK_LINGER) &&
418 !(current->flags & PF_EXITING))
419 timeout = sk->sk_lingertime;
420 sock->sk = NULL;
421 sk->sk_prot->close(sk, timeout);
422 }
423 return 0;
424}
425
426/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700427int sysctl_ip_nonlocal_bind __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
429int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
430{
431 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
432 struct sock *sk = sock->sk;
433 struct inet_sock *inet = inet_sk(sk);
434 unsigned short snum;
435 int chk_addr_ret;
436 int err;
437
438 /* If the socket has its own bind function then use it. (RAW) */
439 if (sk->sk_prot->bind) {
440 err = sk->sk_prot->bind(sk, uaddr, addr_len);
441 goto out;
442 }
443 err = -EINVAL;
444 if (addr_len < sizeof(struct sockaddr_in))
445 goto out;
446
447 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
448
449 /* Not specified by any standard per-se, however it breaks too
450 * many applications when removed. It is unfortunate since
451 * allowing applications to make a non-local bind solves
452 * several problems with systems using dynamic addressing.
453 * (ie. your servers still start up even if your ISDN link
454 * is temporarily down)
455 */
456 err = -EADDRNOTAVAIL;
457 if (!sysctl_ip_nonlocal_bind &&
458 !inet->freebind &&
459 addr->sin_addr.s_addr != INADDR_ANY &&
460 chk_addr_ret != RTN_LOCAL &&
461 chk_addr_ret != RTN_MULTICAST &&
462 chk_addr_ret != RTN_BROADCAST)
463 goto out;
464
465 snum = ntohs(addr->sin_port);
466 err = -EACCES;
467 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
468 goto out;
469
470 /* We keep a pair of addresses. rcv_saddr is the one
471 * used by hash lookups, and saddr is used for transmit.
472 *
473 * In the BSD API these are the same except where it
474 * would be illegal to use them (multicast/broadcast) in
475 * which case the sending device address is used.
476 */
477 lock_sock(sk);
478
479 /* Check these errors (active socket, double bind). */
480 err = -EINVAL;
481 if (sk->sk_state != TCP_CLOSE || inet->num)
482 goto out_release_sock;
483
484 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
485 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
486 inet->saddr = 0; /* Use device */
487
488 /* Make sure we are allowed to bind here. */
489 if (sk->sk_prot->get_port(sk, snum)) {
490 inet->saddr = inet->rcv_saddr = 0;
491 err = -EADDRINUSE;
492 goto out_release_sock;
493 }
494
495 if (inet->rcv_saddr)
496 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
497 if (snum)
498 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
499 inet->sport = htons(inet->num);
500 inet->daddr = 0;
501 inet->dport = 0;
502 sk_dst_reset(sk);
503 err = 0;
504out_release_sock:
505 release_sock(sk);
506out:
507 return err;
508}
509
510int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
511 int addr_len, int flags)
512{
513 struct sock *sk = sock->sk;
514
515 if (uaddr->sa_family == AF_UNSPEC)
516 return sk->sk_prot->disconnect(sk, flags);
517
518 if (!inet_sk(sk)->num && inet_autobind(sk))
519 return -EAGAIN;
520 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
521}
522
523static long inet_wait_for_connect(struct sock *sk, long timeo)
524{
525 DEFINE_WAIT(wait);
526
527 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
528
529 /* Basic assumption: if someone sets sk->sk_err, he _must_
530 * change state of the socket from TCP_SYN_*.
531 * Connect() does not allow to get error notifications
532 * without closing the socket.
533 */
534 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
535 release_sock(sk);
536 timeo = schedule_timeout(timeo);
537 lock_sock(sk);
538 if (signal_pending(current) || !timeo)
539 break;
540 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
541 }
542 finish_wait(sk->sk_sleep, &wait);
543 return timeo;
544}
545
546/*
547 * Connect to a remote host. There is regrettably still a little
548 * TCP 'magic' in here.
549 */
550int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
551 int addr_len, int flags)
552{
553 struct sock *sk = sock->sk;
554 int err;
555 long timeo;
556
557 lock_sock(sk);
558
559 if (uaddr->sa_family == AF_UNSPEC) {
560 err = sk->sk_prot->disconnect(sk, flags);
561 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
562 goto out;
563 }
564
565 switch (sock->state) {
566 default:
567 err = -EINVAL;
568 goto out;
569 case SS_CONNECTED:
570 err = -EISCONN;
571 goto out;
572 case SS_CONNECTING:
573 err = -EALREADY;
574 /* Fall out of switch with err, set for this state */
575 break;
576 case SS_UNCONNECTED:
577 err = -EISCONN;
578 if (sk->sk_state != TCP_CLOSE)
579 goto out;
580
581 err = sk->sk_prot->connect(sk, uaddr, addr_len);
582 if (err < 0)
583 goto out;
584
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900585 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
587 /* Just entered SS_CONNECTING state; the only
588 * difference is that return value in non-blocking
589 * case is EINPROGRESS, rather than EALREADY.
590 */
591 err = -EINPROGRESS;
592 break;
593 }
594
595 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
596
597 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
598 /* Error code is set above */
599 if (!timeo || !inet_wait_for_connect(sk, timeo))
600 goto out;
601
602 err = sock_intr_errno(timeo);
603 if (signal_pending(current))
604 goto out;
605 }
606
607 /* Connection was closed by RST, timeout, ICMP error
608 * or another process disconnected us.
609 */
610 if (sk->sk_state == TCP_CLOSE)
611 goto sock_error;
612
613 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
614 * and error was received after socket entered established state.
615 * Hence, it is handled normally after connect() return successfully.
616 */
617
618 sock->state = SS_CONNECTED;
619 err = 0;
620out:
621 release_sock(sk);
622 return err;
623
624sock_error:
625 err = sock_error(sk) ? : -ECONNABORTED;
626 sock->state = SS_UNCONNECTED;
627 if (sk->sk_prot->disconnect(sk, flags))
628 sock->state = SS_DISCONNECTING;
629 goto out;
630}
631
632/*
633 * Accept a pending connection. The TCP layer now gives BSD semantics.
634 */
635
636int inet_accept(struct socket *sock, struct socket *newsock, int flags)
637{
638 struct sock *sk1 = sock->sk;
639 int err = -EINVAL;
640 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
641
642 if (!sk2)
643 goto do_err;
644
645 lock_sock(sk2);
646
647 BUG_TRAP((1 << sk2->sk_state) &
648 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
649
650 sock_graft(sk2, newsock);
651
652 newsock->state = SS_CONNECTED;
653 err = 0;
654 release_sock(sk2);
655do_err:
656 return err;
657}
658
659
660/*
661 * This does both peername and sockname.
662 */
663int inet_getname(struct socket *sock, struct sockaddr *uaddr,
664 int *uaddr_len, int peer)
665{
666 struct sock *sk = sock->sk;
667 struct inet_sock *inet = inet_sk(sk);
668 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
669
670 sin->sin_family = AF_INET;
671 if (peer) {
672 if (!inet->dport ||
673 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
674 peer == 1))
675 return -ENOTCONN;
676 sin->sin_port = inet->dport;
677 sin->sin_addr.s_addr = inet->daddr;
678 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800679 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (!addr)
681 addr = inet->saddr;
682 sin->sin_port = inet->sport;
683 sin->sin_addr.s_addr = addr;
684 }
685 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
686 *uaddr_len = sizeof(*sin);
687 return 0;
688}
689
690int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
691 size_t size)
692{
693 struct sock *sk = sock->sk;
694
695 /* We may need to bind the socket. */
696 if (!inet_sk(sk)->num && inet_autobind(sk))
697 return -EAGAIN;
698
699 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
700}
701
702
703static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
704{
705 struct sock *sk = sock->sk;
706
707 /* We may need to bind the socket. */
708 if (!inet_sk(sk)->num && inet_autobind(sk))
709 return -EAGAIN;
710
711 if (sk->sk_prot->sendpage)
712 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
713 return sock_no_sendpage(sock, page, offset, size, flags);
714}
715
716
717int inet_shutdown(struct socket *sock, int how)
718{
719 struct sock *sk = sock->sk;
720 int err = 0;
721
722 /* This should really check to make sure
723 * the socket is a TCP socket. (WHY AC...)
724 */
725 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
726 1->2 bit 2 snds.
727 2->3 */
728 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
729 return -EINVAL;
730
731 lock_sock(sk);
732 if (sock->state == SS_CONNECTING) {
733 if ((1 << sk->sk_state) &
734 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
735 sock->state = SS_DISCONNECTING;
736 else
737 sock->state = SS_CONNECTED;
738 }
739
740 switch (sk->sk_state) {
741 case TCP_CLOSE:
742 err = -ENOTCONN;
743 /* Hack to wake up other listeners, who can poll for
744 POLLHUP, even on eg. unconnected UDP sockets -- RR */
745 default:
746 sk->sk_shutdown |= how;
747 if (sk->sk_prot->shutdown)
748 sk->sk_prot->shutdown(sk, how);
749 break;
750
751 /* Remaining two branches are temporary solution for missing
752 * close() in multithreaded environment. It is _not_ a good idea,
753 * but we have no choice until close() is repaired at VFS level.
754 */
755 case TCP_LISTEN:
756 if (!(how & RCV_SHUTDOWN))
757 break;
758 /* Fall through */
759 case TCP_SYN_SENT:
760 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
761 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
762 break;
763 }
764
765 /* Wake up anyone sleeping in poll. */
766 sk->sk_state_change(sk);
767 release_sock(sk);
768 return err;
769}
770
771/*
772 * ioctl() calls you can issue on an INET socket. Most of these are
773 * device configuration and stuff and very rarely used. Some ioctls
774 * pass on to the socket itself.
775 *
776 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
777 * loads the devconfigure module does its configuring and unloads it.
778 * There's a good 20K of config code hanging around the kernel.
779 */
780
781int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
782{
783 struct sock *sk = sock->sk;
784 int err = 0;
785
786 switch (cmd) {
787 case SIOCGSTAMP:
788 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
789 break;
Eric Dumazetae40eb12007-03-18 17:33:16 -0700790 case SIOCGSTAMPNS:
791 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
792 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 case SIOCADDRT:
794 case SIOCDELRT:
795 case SIOCRTMSG:
796 err = ip_rt_ioctl(cmd, (void __user *)arg);
797 break;
798 case SIOCDARP:
799 case SIOCGARP:
800 case SIOCSARP:
801 err = arp_ioctl(cmd, (void __user *)arg);
802 break;
803 case SIOCGIFADDR:
804 case SIOCSIFADDR:
805 case SIOCGIFBRDADDR:
806 case SIOCSIFBRDADDR:
807 case SIOCGIFNETMASK:
808 case SIOCSIFNETMASK:
809 case SIOCGIFDSTADDR:
810 case SIOCSIFDSTADDR:
811 case SIOCSIFPFLAGS:
812 case SIOCGIFPFLAGS:
813 case SIOCSIFFLAGS:
814 err = devinet_ioctl(cmd, (void __user *)arg);
815 break;
816 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800817 if (sk->sk_prot->ioctl)
818 err = sk->sk_prot->ioctl(sk, cmd, arg);
819 else
820 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 break;
822 }
823 return err;
824}
825
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800826const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800827 .family = PF_INET,
828 .owner = THIS_MODULE,
829 .release = inet_release,
830 .bind = inet_bind,
831 .connect = inet_stream_connect,
832 .socketpair = sock_no_socketpair,
833 .accept = inet_accept,
834 .getname = inet_getname,
835 .poll = tcp_poll,
836 .ioctl = inet_ioctl,
837 .listen = inet_listen,
838 .shutdown = inet_shutdown,
839 .setsockopt = sock_common_setsockopt,
840 .getsockopt = sock_common_getsockopt,
David S. Miller3516ffb2007-08-02 19:23:56 -0700841 .sendmsg = tcp_sendmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800842 .recvmsg = sock_common_recvmsg,
843 .mmap = sock_no_mmap,
844 .sendpage = tcp_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800845 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800846#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800847 .compat_setsockopt = compat_sock_common_setsockopt,
848 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800849#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850};
851
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800852const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800853 .family = PF_INET,
854 .owner = THIS_MODULE,
855 .release = inet_release,
856 .bind = inet_bind,
857 .connect = inet_dgram_connect,
858 .socketpair = sock_no_socketpair,
859 .accept = sock_no_accept,
860 .getname = inet_getname,
861 .poll = udp_poll,
862 .ioctl = inet_ioctl,
863 .listen = sock_no_listen,
864 .shutdown = inet_shutdown,
865 .setsockopt = sock_common_setsockopt,
866 .getsockopt = sock_common_getsockopt,
867 .sendmsg = inet_sendmsg,
868 .recvmsg = sock_common_recvmsg,
869 .mmap = sock_no_mmap,
870 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800871#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800872 .compat_setsockopt = compat_sock_common_setsockopt,
873 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800874#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875};
876
877/*
878 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
879 * udp_poll
880 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800881static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800882 .family = PF_INET,
883 .owner = THIS_MODULE,
884 .release = inet_release,
885 .bind = inet_bind,
886 .connect = inet_dgram_connect,
887 .socketpair = sock_no_socketpair,
888 .accept = sock_no_accept,
889 .getname = inet_getname,
890 .poll = datagram_poll,
891 .ioctl = inet_ioctl,
892 .listen = sock_no_listen,
893 .shutdown = inet_shutdown,
894 .setsockopt = sock_common_setsockopt,
895 .getsockopt = sock_common_getsockopt,
896 .sendmsg = inet_sendmsg,
897 .recvmsg = sock_common_recvmsg,
898 .mmap = sock_no_mmap,
899 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800900#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800901 .compat_setsockopt = compat_sock_common_setsockopt,
902 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800903#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904};
905
906static struct net_proto_family inet_family_ops = {
907 .family = PF_INET,
908 .create = inet_create,
909 .owner = THIS_MODULE,
910};
911
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912/* Upon startup we insert all the elements in inetsw_array[] into
913 * the linked list inetsw.
914 */
915static struct inet_protosw inetsw_array[] =
916{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900917 {
918 .type = SOCK_STREAM,
919 .protocol = IPPROTO_TCP,
920 .prot = &tcp_prot,
921 .ops = &inet_stream_ops,
922 .capability = -1,
923 .no_check = 0,
924 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800925 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900926 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900928 {
929 .type = SOCK_DGRAM,
930 .protocol = IPPROTO_UDP,
931 .prot = &udp_prot,
932 .ops = &inet_dgram_ops,
933 .capability = -1,
934 .no_check = UDP_CSUM_DEFAULT,
935 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900937
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900940 .type = SOCK_RAW,
941 .protocol = IPPROTO_IP, /* wild card */
942 .prot = &raw_prot,
943 .ops = &inet_sockraw_ops,
944 .capability = CAP_NET_RAW,
945 .no_check = UDP_CSUM_DEFAULT,
946 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 }
948};
949
Denis Chengc40f6ff2007-09-16 16:39:25 -0700950#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
952void inet_register_protosw(struct inet_protosw *p)
953{
954 struct list_head *lh;
955 struct inet_protosw *answer;
956 int protocol = p->protocol;
957 struct list_head *last_perm;
958
959 spin_lock_bh(&inetsw_lock);
960
961 if (p->type >= SOCK_MAX)
962 goto out_illegal;
963
964 /* If we are trying to override a permanent protocol, bail. */
965 answer = NULL;
966 last_perm = &inetsw[p->type];
967 list_for_each(lh, &inetsw[p->type]) {
968 answer = list_entry(lh, struct inet_protosw, list);
969
970 /* Check only the non-wild match. */
971 if (INET_PROTOSW_PERMANENT & answer->flags) {
972 if (protocol == answer->protocol)
973 break;
974 last_perm = lh;
975 }
976
977 answer = NULL;
978 }
979 if (answer)
980 goto out_permanent;
981
982 /* Add the new entry after the last permanent entry if any, so that
983 * the new entry does not override a permanent entry when matched with
984 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900985 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 * system automatically returns to the old behavior.
987 */
988 list_add_rcu(&p->list, last_perm);
989out:
990 spin_unlock_bh(&inetsw_lock);
991
992 synchronize_net();
993
994 return;
995
996out_permanent:
997 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
998 protocol);
999 goto out;
1000
1001out_illegal:
1002 printk(KERN_ERR
1003 "Ignoring attempt to register invalid socket type %d.\n",
1004 p->type);
1005 goto out;
1006}
1007
1008void inet_unregister_protosw(struct inet_protosw *p)
1009{
1010 if (INET_PROTOSW_PERMANENT & p->flags) {
1011 printk(KERN_ERR
1012 "Attempt to unregister permanent protocol %d.\n",
1013 p->protocol);
1014 } else {
1015 spin_lock_bh(&inetsw_lock);
1016 list_del_rcu(&p->list);
1017 spin_unlock_bh(&inetsw_lock);
1018
1019 synchronize_net();
1020 }
1021}
1022
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001023/*
1024 * Shall we try to damage output packets if routing dev changes?
1025 */
1026
Brian Haleyab32ea52006-09-22 14:15:41 -07001027int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001028
1029static int inet_sk_reselect_saddr(struct sock *sk)
1030{
1031 struct inet_sock *inet = inet_sk(sk);
1032 int err;
1033 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001034 __be32 old_saddr = inet->saddr;
1035 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001036 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001037
1038 if (inet->opt && inet->opt->srr)
1039 daddr = inet->opt->faddr;
1040
1041 /* Query new route. */
1042 err = ip_route_connect(&rt, daddr, 0,
1043 RT_CONN_FLAGS(sk),
1044 sk->sk_bound_dev_if,
1045 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001046 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001047 if (err)
1048 return err;
1049
1050 sk_setup_caps(sk, &rt->u.dst);
1051
1052 new_saddr = rt->rt_src;
1053
1054 if (new_saddr == old_saddr)
1055 return 0;
1056
1057 if (sysctl_ip_dynaddr > 1) {
1058 printk(KERN_INFO "%s(): shifting inet->"
1059 "saddr from %d.%d.%d.%d to %d.%d.%d.%d\n",
1060 __FUNCTION__,
1061 NIPQUAD(old_saddr),
1062 NIPQUAD(new_saddr));
1063 }
1064
1065 inet->saddr = inet->rcv_saddr = new_saddr;
1066
1067 /*
1068 * XXX The only one ugly spot where we need to
1069 * XXX really change the sockets identity after
1070 * XXX it has entered the hashes. -DaveM
1071 *
1072 * Besides that, it does not check for connection
1073 * uniqueness. Wait for troubles.
1074 */
1075 __sk_prot_rehash(sk);
1076 return 0;
1077}
1078
1079int inet_sk_rebuild_header(struct sock *sk)
1080{
1081 struct inet_sock *inet = inet_sk(sk);
1082 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001083 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001084 int err;
1085
1086 /* Route is OK, nothing to do. */
1087 if (rt)
1088 return 0;
1089
1090 /* Reroute. */
1091 daddr = inet->daddr;
1092 if (inet->opt && inet->opt->srr)
1093 daddr = inet->opt->faddr;
1094{
1095 struct flowi fl = {
1096 .oif = sk->sk_bound_dev_if,
1097 .nl_u = {
1098 .ip4_u = {
1099 .daddr = daddr,
1100 .saddr = inet->saddr,
1101 .tos = RT_CONN_FLAGS(sk),
1102 },
1103 },
1104 .proto = sk->sk_protocol,
1105 .uli_u = {
1106 .ports = {
1107 .sport = inet->sport,
1108 .dport = inet->dport,
1109 },
1110 },
1111 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001112
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001113 security_sk_classify_flow(sk, &fl);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001114 err = ip_route_output_flow(&rt, &fl, sk, 0);
1115}
1116 if (!err)
1117 sk_setup_caps(sk, &rt->u.dst);
1118 else {
1119 /* Routing failed... */
1120 sk->sk_route_caps = 0;
1121 /*
1122 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1123 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1124 */
1125 if (!sysctl_ip_dynaddr ||
1126 sk->sk_state != TCP_SYN_SENT ||
1127 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1128 (err = inet_sk_reselect_saddr(sk)) != 0)
1129 sk->sk_err_soft = -err;
1130 }
1131
1132 return err;
1133}
1134
1135EXPORT_SYMBOL(inet_sk_rebuild_header);
1136
Herbert Xua430a432006-07-08 13:34:56 -07001137static int inet_gso_send_check(struct sk_buff *skb)
1138{
1139 struct iphdr *iph;
1140 struct net_protocol *ops;
1141 int proto;
1142 int ihl;
1143 int err = -EINVAL;
1144
1145 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1146 goto out;
1147
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001148 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001149 ihl = iph->ihl * 4;
1150 if (ihl < sizeof(*iph))
1151 goto out;
1152
1153 if (unlikely(!pskb_may_pull(skb, ihl)))
1154 goto out;
1155
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001156 __skb_pull(skb, ihl);
1157 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001158 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001159 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1160 err = -EPROTONOSUPPORT;
1161
1162 rcu_read_lock();
1163 ops = rcu_dereference(inet_protos[proto]);
1164 if (likely(ops && ops->gso_send_check))
1165 err = ops->gso_send_check(skb);
1166 rcu_read_unlock();
1167
1168out:
1169 return err;
1170}
1171
Herbert Xu576a30e2006-06-27 13:22:38 -07001172static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001173{
1174 struct sk_buff *segs = ERR_PTR(-EINVAL);
1175 struct iphdr *iph;
1176 struct net_protocol *ops;
1177 int proto;
1178 int ihl;
1179 int id;
1180
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001181 if (!(features & NETIF_F_V4_CSUM))
1182 features &= ~NETIF_F_SG;
1183
Herbert Xubbcf4672006-07-03 19:38:35 -07001184 if (unlikely(skb_shinfo(skb)->gso_type &
1185 ~(SKB_GSO_TCPV4 |
1186 SKB_GSO_UDP |
1187 SKB_GSO_DODGY |
1188 SKB_GSO_TCP_ECN |
1189 0)))
1190 goto out;
1191
1192 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001193 goto out;
1194
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001195 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001196 ihl = iph->ihl * 4;
1197 if (ihl < sizeof(*iph))
1198 goto out;
1199
Herbert Xubbcf4672006-07-03 19:38:35 -07001200 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001201 goto out;
1202
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001203 __skb_pull(skb, ihl);
1204 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001205 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001206 id = ntohs(iph->id);
1207 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1208 segs = ERR_PTR(-EPROTONOSUPPORT);
1209
1210 rcu_read_lock();
1211 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001212 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001213 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001214 rcu_read_unlock();
1215
Herbert Xu576a30e2006-06-27 13:22:38 -07001216 if (!segs || unlikely(IS_ERR(segs)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001217 goto out;
1218
1219 skb = segs;
1220 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001221 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001222 iph->id = htons(id++);
1223 iph->tot_len = htons(skb->len - skb->mac_len);
1224 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001225 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001226 } while ((skb = skb->next));
1227
1228out:
1229 return segs;
1230}
1231
Herbert Xu5e0f0432007-04-24 21:53:35 -07001232unsigned long snmp_fold_field(void *mib[], int offt)
1233{
1234 unsigned long res = 0;
1235 int i;
1236
1237 for_each_possible_cpu(i) {
1238 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1239 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1240 }
1241 return res;
1242}
1243EXPORT_SYMBOL_GPL(snmp_fold_field);
1244
1245int snmp_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
1246{
1247 BUG_ON(ptr == NULL);
1248 ptr[0] = __alloc_percpu(mibsize);
1249 if (!ptr[0])
1250 goto err0;
1251 ptr[1] = __alloc_percpu(mibsize);
1252 if (!ptr[1])
1253 goto err1;
1254 return 0;
1255err1:
1256 free_percpu(ptr[0]);
1257 ptr[0] = NULL;
1258err0:
1259 return -ENOMEM;
1260}
1261EXPORT_SYMBOL_GPL(snmp_mib_init);
1262
1263void snmp_mib_free(void *ptr[2])
1264{
1265 BUG_ON(ptr == NULL);
1266 free_percpu(ptr[0]);
1267 free_percpu(ptr[1]);
1268 ptr[0] = ptr[1] = NULL;
1269}
1270EXPORT_SYMBOL_GPL(snmp_mib_free);
1271
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272#ifdef CONFIG_IP_MULTICAST
1273static struct net_protocol igmp_protocol = {
1274 .handler = igmp_rcv,
1275};
1276#endif
1277
1278static struct net_protocol tcp_protocol = {
1279 .handler = tcp_v4_rcv,
1280 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001281 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001282 .gso_segment = tcp_tso_segment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 .no_policy = 1,
1284};
1285
1286static struct net_protocol udp_protocol = {
1287 .handler = udp_rcv,
1288 .err_handler = udp_err,
1289 .no_policy = 1,
1290};
1291
1292static struct net_protocol icmp_protocol = {
1293 .handler = icmp_rcv,
1294};
1295
1296static int __init init_ipv4_mibs(void)
1297{
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001298 if (snmp_mib_init((void **)net_statistics,
1299 sizeof(struct linux_mib),
1300 __alignof__(struct linux_mib)) < 0)
1301 goto err_net_mib;
1302 if (snmp_mib_init((void **)ip_statistics,
1303 sizeof(struct ipstats_mib),
1304 __alignof__(struct ipstats_mib)) < 0)
1305 goto err_ip_mib;
1306 if (snmp_mib_init((void **)icmp_statistics,
1307 sizeof(struct icmp_mib),
1308 __alignof__(struct icmp_mib)) < 0)
1309 goto err_icmp_mib;
David L Stevens96793b42007-09-17 09:57:33 -07001310 if (snmp_mib_init((void **)icmpmsg_statistics,
1311 sizeof(struct icmpmsg_mib),
1312 __alignof__(struct icmpmsg_mib)) < 0)
1313 goto err_icmpmsg_mib;
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001314 if (snmp_mib_init((void **)tcp_statistics,
1315 sizeof(struct tcp_mib),
1316 __alignof__(struct tcp_mib)) < 0)
1317 goto err_tcp_mib;
1318 if (snmp_mib_init((void **)udp_statistics,
1319 sizeof(struct udp_mib),
1320 __alignof__(struct udp_mib)) < 0)
1321 goto err_udp_mib;
1322 if (snmp_mib_init((void **)udplite_statistics,
1323 sizeof(struct udp_mib),
1324 __alignof__(struct udp_mib)) < 0)
1325 goto err_udplite_mib;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001327 tcp_mib_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
1329 return 0;
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001330
1331err_udplite_mib:
1332 snmp_mib_free((void **)udp_statistics);
1333err_udp_mib:
1334 snmp_mib_free((void **)tcp_statistics);
1335err_tcp_mib:
David L Stevens96793b42007-09-17 09:57:33 -07001336 snmp_mib_free((void **)icmpmsg_statistics);
1337err_icmpmsg_mib:
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001338 snmp_mib_free((void **)icmp_statistics);
1339err_icmp_mib:
1340 snmp_mib_free((void **)ip_statistics);
1341err_ip_mib:
1342 snmp_mib_free((void **)net_statistics);
1343err_net_mib:
1344 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345}
1346
1347static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
Herbert Xu30e224d2005-07-05 14:40:10 -07001349/*
1350 * IP protocol layer initialiser
1351 */
1352
1353static struct packet_type ip_packet_type = {
1354 .type = __constant_htons(ETH_P_IP),
1355 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001356 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001357 .gso_segment = inet_gso_segment,
Herbert Xu30e224d2005-07-05 14:40:10 -07001358};
1359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360static int __init inet_init(void)
1361{
1362 struct sk_buff *dummy_skb;
1363 struct inet_protosw *q;
1364 struct list_head *r;
1365 int rc = -EINVAL;
1366
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001367 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
1369 rc = proto_register(&tcp_prot, 1);
1370 if (rc)
1371 goto out;
1372
1373 rc = proto_register(&udp_prot, 1);
1374 if (rc)
1375 goto out_unregister_tcp_proto;
1376
1377 rc = proto_register(&raw_prot, 1);
1378 if (rc)
1379 goto out_unregister_udp_proto;
1380
1381 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001382 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 */
1384
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001385 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386
1387 /*
1388 * Add all the base protocols.
1389 */
1390
1391 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1392 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1393 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1394 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1395 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1396 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1397#ifdef CONFIG_IP_MULTICAST
1398 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1399 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1400#endif
1401
1402 /* Register the socket-side information for inet_create. */
1403 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1404 INIT_LIST_HEAD(r);
1405
1406 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1407 inet_register_protosw(q);
1408
1409 /*
1410 * Set the ARP module up
1411 */
1412
1413 arp_init();
1414
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001415 /*
1416 * Set the IP module up
1417 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 ip_init();
1420
1421 tcp_v4_init(&inet_family_ops);
1422
1423 /* Setup TCP slab cache for open requests. */
1424 tcp_init();
1425
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001426 /* Add UDP-Lite (RFC 3828) */
1427 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
1429 /*
1430 * Set the ICMP layer up
1431 */
1432
1433 icmp_init(&inet_family_ops);
1434
1435 /*
1436 * Initialise the multicast router
1437 */
1438#if defined(CONFIG_IP_MROUTE)
1439 ip_mr_init();
1440#endif
1441 /*
1442 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001443 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
Stephen Hemminger132adf52007-03-08 20:44:43 -08001445 if (init_ipv4_mibs())
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n"); ;
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001447
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 ipv4_proc_init();
1449
1450 ipfrag_init();
1451
Herbert Xu30e224d2005-07-05 14:40:10 -07001452 dev_add_pack(&ip_packet_type);
1453
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 rc = 0;
1455out:
1456 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457out_unregister_udp_proto:
1458 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001459out_unregister_tcp_proto:
1460 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 goto out;
1462}
1463
Heiko Carstensa536e072006-04-28 15:19:17 -07001464fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466/* ------------------------------------------------------------------------ */
1467
1468#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469static int __init ipv4_proc_init(void)
1470{
1471 int rc = 0;
1472
1473 if (raw_proc_init())
1474 goto out_raw;
1475 if (tcp4_proc_init())
1476 goto out_tcp;
1477 if (udp4_proc_init())
1478 goto out_udp;
1479 if (fib_proc_init())
1480 goto out_fib;
1481 if (ip_misc_proc_init())
1482 goto out_misc;
1483out:
1484 return rc;
1485out_misc:
1486 fib_proc_exit();
1487out_fib:
1488 udp4_proc_exit();
1489out_udp:
1490 tcp4_proc_exit();
1491out_tcp:
1492 raw_proc_exit();
1493out_raw:
1494 rc = -ENOMEM;
1495 goto out;
1496}
1497
1498#else /* CONFIG_PROC_FS */
1499static int __init ipv4_proc_init(void)
1500{
1501 return 0;
1502}
1503#endif /* CONFIG_PROC_FS */
1504
1505MODULE_ALIAS_NETPROTO(PF_INET);
1506
1507EXPORT_SYMBOL(inet_accept);
1508EXPORT_SYMBOL(inet_bind);
1509EXPORT_SYMBOL(inet_dgram_connect);
1510EXPORT_SYMBOL(inet_dgram_ops);
1511EXPORT_SYMBOL(inet_getname);
1512EXPORT_SYMBOL(inet_ioctl);
1513EXPORT_SYMBOL(inet_listen);
1514EXPORT_SYMBOL(inet_register_protosw);
1515EXPORT_SYMBOL(inet_release);
1516EXPORT_SYMBOL(inet_sendmsg);
1517EXPORT_SYMBOL(inet_shutdown);
1518EXPORT_SYMBOL(inet_sock_destruct);
1519EXPORT_SYMBOL(inet_stream_connect);
1520EXPORT_SYMBOL(inet_stream_ops);
1521EXPORT_SYMBOL(inet_unregister_protosw);
1522EXPORT_SYMBOL(net_statistics);
Neil Hormancdac4e02005-06-13 15:12:33 -07001523EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);