blob: 8a3ac1fa71a976e8c8f4c4f2dfd9fa87af2f9a99 [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 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Florian La Roche, <flla@stud.uni-sb.de>
11 * Alan Cox, <A.Cox@swansea.ac.uk>
12 *
13 * Changes (see also sock.c)
14 *
15 * piggy,
16 * Karl Knutson : Socket protocol table
17 * A.N.Kuznetsov : Socket death error in accept().
18 * John Richardson : Fix non blocking error in connect()
19 * so sockets that fail to connect
20 * don't return -EINPROGRESS.
21 * Alan Cox : Asynchronous I/O support
22 * Alan Cox : Keep correct socket pointer on sock
23 * structures
24 * when accept() ed
25 * Alan Cox : Semantics of SO_LINGER aren't state
26 * moved to close when you look carefully.
27 * With this fixed and the accept bug fixed
28 * some RPC stuff seems happier.
29 * Niibe Yutaka : 4.4BSD style write async I/O
30 * Alan Cox,
31 * Tony Gale : Fixed reuse semantics.
32 * Alan Cox : bind() shouldn't abort existing but dead
33 * sockets. Stops FTP netin:.. I hope.
34 * Alan Cox : bind() works correctly for RAW sockets.
35 * Note that FreeBSD at least was broken
36 * in this respect so be careful with
37 * compatibility tests...
38 * Alan Cox : routing cache support
39 * Alan Cox : memzero the socket structure for
40 * compactness.
41 * Matt Day : nonblock connect error handler
42 * Alan Cox : Allow large numbers of pending sockets
43 * (eg for big web sites), but only if
44 * specifically application requested.
45 * Alan Cox : New buffering throughout IP. Used
46 * dumbly.
47 * Alan Cox : New buffering now used smartly.
48 * Alan Cox : BSD rather than common sense
49 * interpretation of listen.
50 * Germano Caronni : Assorted small races.
51 * Alan Cox : sendmsg/recvmsg basic support.
52 * Alan Cox : Only sendmsg/recvmsg now supported.
53 * Alan Cox : Locked down bind (see security list).
54 * Alan Cox : Loosened bind a little.
55 * Mike McLagan : ADD/DEL DLCI Ioctls
56 * Willy Konynenberg : Transparent proxying support.
57 * David S. Miller : New socket lookup architecture.
58 * Some other random speedups.
59 * Cyrus Durgin : Cleaned up file for kmod hacks.
60 * Andi Kleen : Fix inet_stream_connect TCP race.
61 *
62 * This program is free software; you can redistribute it and/or
63 * modify it under the terms of the GNU General Public License
64 * as published by the Free Software Foundation; either version
65 * 2 of the License, or (at your option) any later version.
66 */
67
Herbert Xuf4c50d92006-06-22 03:02:40 -070068#include <linux/err.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#include <linux/errno.h>
70#include <linux/types.h>
71#include <linux/socket.h>
72#include <linux/in.h>
73#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#include <linux/module.h>
75#include <linux/sched.h>
76#include <linux/timer.h>
77#include <linux/string.h>
78#include <linux/sockios.h>
79#include <linux/net.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080080#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/fcntl.h>
82#include <linux/mm.h>
83#include <linux/interrupt.h>
84#include <linux/stat.h>
85#include <linux/init.h>
86#include <linux/poll.h>
87#include <linux/netfilter_ipv4.h>
David S. Millerb3da2cf2007-03-23 11:40:27 -070088#include <linux/random.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
90#include <asm/uaccess.h>
91#include <asm/system.h>
92
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <linux/inet.h>
94#include <linux/igmp.h>
Arnaldo Carvalho de Melo14c85022005-12-27 02:43:12 -020095#include <linux/inetdevice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070096#include <linux/netdevice.h>
97#include <net/ip.h>
98#include <net/protocol.h>
99#include <net/arp.h>
100#include <net/route.h>
101#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700102#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#include <net/tcp.h>
104#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800105#include <net/udplite.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106#include <linux/skbuff.h>
107#include <net/sock.h>
108#include <net/raw.h>
109#include <net/icmp.h>
110#include <net/ipip.h>
111#include <net/inet_common.h>
112#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700113#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114#ifdef CONFIG_IP_MROUTE
115#include <linux/mroute.h>
116#endif
117
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118extern void ip_mc_drop_socket(struct sock *sk);
119
120/* The inetsw table contains everything that inet_create needs to
121 * build a new socket.
122 */
123static struct list_head inetsw[SOCK_MAX];
124static DEFINE_SPINLOCK(inetsw_lock);
125
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800126struct ipv4_config ipv4_config;
127
128EXPORT_SYMBOL(ipv4_config);
129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/* 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
Hideo Aoki95766ff2007-12-31 00:29:24 -0800139 sk_mem_reclaim(sk);
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
142 printk("Attempt to release TCP socket in state %d %p\n",
143 sk->sk_state, sk);
144 return;
145 }
146 if (!sock_flag(sk, SOCK_DEAD)) {
147 printk("Attempt to release alive inet socket %p\n", sk);
148 return;
149 }
150
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700151 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
152 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
153 WARN_ON(sk->sk_wmem_queued);
154 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
Jesper Juhla51482b2005-11-08 09:41:34 -0800156 kfree(inet->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 dst_release(sk->sk_dst_cache);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700158 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159}
160
161/*
162 * The routines beyond this point handle the behaviour of an AF_INET
163 * socket object. Mostly it punts to the subprotocols of IP to do
164 * the work.
165 */
166
167/*
168 * Automatically bind an unbound socket.
169 */
170
171static int inet_autobind(struct sock *sk)
172{
173 struct inet_sock *inet;
174 /* We may need to bind the socket. */
175 lock_sock(sk);
176 inet = inet_sk(sk);
177 if (!inet->num) {
178 if (sk->sk_prot->get_port(sk, 0)) {
179 release_sock(sk);
180 return -EAGAIN;
181 }
182 inet->sport = htons(inet->num);
183 }
184 release_sock(sk);
185 return 0;
186}
187
188/*
189 * Move a socket into listening state.
190 */
191int inet_listen(struct socket *sock, int backlog)
192{
193 struct sock *sk = sock->sk;
194 unsigned char old_state;
195 int err;
196
197 lock_sock(sk);
198
199 err = -EINVAL;
200 if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
201 goto out;
202
203 old_state = sk->sk_state;
204 if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
205 goto out;
206
207 /* Really, if the socket is already in listen state
208 * we can only allow the backlog to be adjusted.
209 */
210 if (old_state != TCP_LISTEN) {
Eric Dumazet72a3eff2006-11-16 02:30:37 -0800211 err = inet_csk_listen_start(sk, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 if (err)
213 goto out;
214 }
215 sk->sk_max_ack_backlog = backlog;
216 err = 0;
217
218out:
219 release_sock(sk);
220 return err;
221}
222
Eric Dumazetbe776282007-03-27 13:53:04 -0700223u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700224EXPORT_SYMBOL(inet_ehash_secret);
225
Eric Dumazetbe776282007-03-27 13:53:04 -0700226/*
227 * inet_ehash_secret must be set exactly once
228 * Instead of using a dedicated spinlock, we (ab)use inetsw_lock
229 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700230void build_ehash_secret(void)
231{
Eric Dumazetbe776282007-03-27 13:53:04 -0700232 u32 rnd;
233 do {
234 get_random_bytes(&rnd, sizeof(rnd));
235 } while (rnd == 0);
236 spin_lock_bh(&inetsw_lock);
237 if (!inet_ehash_secret)
238 inet_ehash_secret = rnd;
239 spin_unlock_bh(&inetsw_lock);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700240}
241EXPORT_SYMBOL(build_ehash_secret);
242
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700243static inline int inet_netns_ok(struct net *net, int protocol)
244{
245 int hash;
246 struct net_protocol *ipprot;
247
248 if (net == &init_net)
249 return 1;
250
251 hash = protocol & (MAX_INET_PROTOS - 1);
252 ipprot = rcu_dereference(inet_protos[hash]);
253
254 if (ipprot == NULL)
255 /* raw IP is OK */
256 return 1;
257 return ipprot->netns_ok;
258}
259
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260/*
261 * Create an inet socket.
262 */
263
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700264static int inet_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265{
266 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 struct inet_protosw *answer;
268 struct inet_sock *inet;
269 struct proto *answer_prot;
270 unsigned char answer_flags;
271 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700272 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800273 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
David S. Millerb3da2cf2007-03-23 11:40:27 -0700275 if (sock->type != SOCK_RAW &&
276 sock->type != SOCK_DGRAM &&
277 !inet_ehash_secret)
278 build_ehash_secret();
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 sock->state = SS_UNCONNECTED;
281
282 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700283lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800284 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700286 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700288 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 /* Check the non-wild match. */
290 if (protocol == answer->protocol) {
291 if (protocol != IPPROTO_IP)
292 break;
293 } else {
294 /* Check for the two wild cases. */
295 if (IPPROTO_IP == protocol) {
296 protocol = answer->protocol;
297 break;
298 }
299 if (IPPROTO_IP == answer->protocol)
300 break;
301 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800302 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 }
304
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700305 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700306 if (try_loading_module < 2) {
307 rcu_read_unlock();
308 /*
309 * Be more specific, e.g. net-pf-2-proto-132-type-1
310 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
311 */
312 if (++try_loading_module == 1)
313 request_module("net-pf-%d-proto-%d-type-%d",
314 PF_INET, protocol, sock->type);
315 /*
316 * Fall back to generic, e.g. net-pf-2-proto-132
317 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
318 */
319 else
320 request_module("net-pf-%d-proto-%d",
321 PF_INET, protocol);
322 goto lookup_protocol;
323 } else
324 goto out_rcu_unlock;
325 }
326
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 err = -EPERM;
328 if (answer->capability > 0 && !capable(answer->capability))
329 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700331 err = -EAFNOSUPPORT;
332 if (!inet_netns_ok(net, protocol))
333 goto out_rcu_unlock;
334
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 sock->ops = answer->ops;
336 answer_prot = answer->prot;
337 answer_no_check = answer->no_check;
338 answer_flags = answer->flags;
339 rcu_read_unlock();
340
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700341 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700344 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 if (sk == NULL)
346 goto out;
347
348 err = 0;
349 sk->sk_no_check = answer_no_check;
350 if (INET_PROTOSW_REUSE & answer_flags)
351 sk->sk_reuse = 1;
352
353 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800354 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355
356 if (SOCK_RAW == sock->type) {
357 inet->num = protocol;
358 if (IPPROTO_RAW == protocol)
359 inet->hdrincl = 1;
360 }
361
362 if (ipv4_config.no_pmtu_disc)
363 inet->pmtudisc = IP_PMTUDISC_DONT;
364 else
365 inet->pmtudisc = IP_PMTUDISC_WANT;
366
367 inet->id = 0;
368
369 sock_init_data(sock, sk);
370
371 sk->sk_destruct = inet_sock_destruct;
372 sk->sk_family = PF_INET;
373 sk->sk_protocol = protocol;
374 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
375
376 inet->uc_ttl = -1;
377 inet->mc_loop = 1;
378 inet->mc_ttl = 1;
379 inet->mc_index = 0;
380 inet->mc_list = NULL;
381
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700382 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
384 if (inet->num) {
385 /* It assumes that any protocol which allows
386 * the user to assign a number at socket
387 * creation time automatically
388 * shares.
389 */
390 inet->sport = htons(inet->num);
391 /* Add to protocol hash chains. */
392 sk->sk_prot->hash(sk);
393 }
394
395 if (sk->sk_prot->init) {
396 err = sk->sk_prot->init(sk);
397 if (err)
398 sk_common_release(sk);
399 }
400out:
401 return err;
402out_rcu_unlock:
403 rcu_read_unlock();
404 goto out;
405}
406
407
408/*
409 * The peer socket should always be NULL (or else). When we call this
410 * function we are destroying the object and from then on nobody
411 * should refer to it.
412 */
413int inet_release(struct socket *sock)
414{
415 struct sock *sk = sock->sk;
416
417 if (sk) {
418 long timeout;
419
420 /* Applications forget to leave groups before exiting */
421 ip_mc_drop_socket(sk);
422
423 /* If linger is set, we don't return until the close
424 * is complete. Otherwise we return immediately. The
425 * actually closing is done the same either way.
426 *
427 * If the close is due to the process exiting, we never
428 * linger..
429 */
430 timeout = 0;
431 if (sock_flag(sk, SOCK_LINGER) &&
432 !(current->flags & PF_EXITING))
433 timeout = sk->sk_lingertime;
434 sock->sk = NULL;
435 sk->sk_prot->close(sk, timeout);
436 }
437 return 0;
438}
439
440/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700441int sysctl_ip_nonlocal_bind __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
443int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
444{
445 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
446 struct sock *sk = sock->sk;
447 struct inet_sock *inet = inet_sk(sk);
448 unsigned short snum;
449 int chk_addr_ret;
450 int err;
451
452 /* If the socket has its own bind function then use it. (RAW) */
453 if (sk->sk_prot->bind) {
454 err = sk->sk_prot->bind(sk, uaddr, addr_len);
455 goto out;
456 }
457 err = -EINVAL;
458 if (addr_len < sizeof(struct sockaddr_in))
459 goto out;
460
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900461 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462
463 /* Not specified by any standard per-se, however it breaks too
464 * many applications when removed. It is unfortunate since
465 * allowing applications to make a non-local bind solves
466 * several problems with systems using dynamic addressing.
467 * (ie. your servers still start up even if your ISDN link
468 * is temporarily down)
469 */
470 err = -EADDRNOTAVAIL;
471 if (!sysctl_ip_nonlocal_bind &&
472 !inet->freebind &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700473 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 chk_addr_ret != RTN_LOCAL &&
475 chk_addr_ret != RTN_MULTICAST &&
476 chk_addr_ret != RTN_BROADCAST)
477 goto out;
478
479 snum = ntohs(addr->sin_port);
480 err = -EACCES;
481 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
482 goto out;
483
484 /* We keep a pair of addresses. rcv_saddr is the one
485 * used by hash lookups, and saddr is used for transmit.
486 *
487 * In the BSD API these are the same except where it
488 * would be illegal to use them (multicast/broadcast) in
489 * which case the sending device address is used.
490 */
491 lock_sock(sk);
492
493 /* Check these errors (active socket, double bind). */
494 err = -EINVAL;
495 if (sk->sk_state != TCP_CLOSE || inet->num)
496 goto out_release_sock;
497
498 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
499 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
500 inet->saddr = 0; /* Use device */
501
502 /* Make sure we are allowed to bind here. */
503 if (sk->sk_prot->get_port(sk, snum)) {
504 inet->saddr = inet->rcv_saddr = 0;
505 err = -EADDRINUSE;
506 goto out_release_sock;
507 }
508
509 if (inet->rcv_saddr)
510 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
511 if (snum)
512 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
513 inet->sport = htons(inet->num);
514 inet->daddr = 0;
515 inet->dport = 0;
516 sk_dst_reset(sk);
517 err = 0;
518out_release_sock:
519 release_sock(sk);
520out:
521 return err;
522}
523
524int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
525 int addr_len, int flags)
526{
527 struct sock *sk = sock->sk;
528
529 if (uaddr->sa_family == AF_UNSPEC)
530 return sk->sk_prot->disconnect(sk, flags);
531
532 if (!inet_sk(sk)->num && inet_autobind(sk))
533 return -EAGAIN;
534 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
535}
536
537static long inet_wait_for_connect(struct sock *sk, long timeo)
538{
539 DEFINE_WAIT(wait);
540
541 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
542
543 /* Basic assumption: if someone sets sk->sk_err, he _must_
544 * change state of the socket from TCP_SYN_*.
545 * Connect() does not allow to get error notifications
546 * without closing the socket.
547 */
548 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
549 release_sock(sk);
550 timeo = schedule_timeout(timeo);
551 lock_sock(sk);
552 if (signal_pending(current) || !timeo)
553 break;
554 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
555 }
556 finish_wait(sk->sk_sleep, &wait);
557 return timeo;
558}
559
560/*
561 * Connect to a remote host. There is regrettably still a little
562 * TCP 'magic' in here.
563 */
564int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
565 int addr_len, int flags)
566{
567 struct sock *sk = sock->sk;
568 int err;
569 long timeo;
570
571 lock_sock(sk);
572
573 if (uaddr->sa_family == AF_UNSPEC) {
574 err = sk->sk_prot->disconnect(sk, flags);
575 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
576 goto out;
577 }
578
579 switch (sock->state) {
580 default:
581 err = -EINVAL;
582 goto out;
583 case SS_CONNECTED:
584 err = -EISCONN;
585 goto out;
586 case SS_CONNECTING:
587 err = -EALREADY;
588 /* Fall out of switch with err, set for this state */
589 break;
590 case SS_UNCONNECTED:
591 err = -EISCONN;
592 if (sk->sk_state != TCP_CLOSE)
593 goto out;
594
595 err = sk->sk_prot->connect(sk, uaddr, addr_len);
596 if (err < 0)
597 goto out;
598
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900599 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 /* Just entered SS_CONNECTING state; the only
602 * difference is that return value in non-blocking
603 * case is EINPROGRESS, rather than EALREADY.
604 */
605 err = -EINPROGRESS;
606 break;
607 }
608
609 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
610
611 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
612 /* Error code is set above */
613 if (!timeo || !inet_wait_for_connect(sk, timeo))
614 goto out;
615
616 err = sock_intr_errno(timeo);
617 if (signal_pending(current))
618 goto out;
619 }
620
621 /* Connection was closed by RST, timeout, ICMP error
622 * or another process disconnected us.
623 */
624 if (sk->sk_state == TCP_CLOSE)
625 goto sock_error;
626
627 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
628 * and error was received after socket entered established state.
629 * Hence, it is handled normally after connect() return successfully.
630 */
631
632 sock->state = SS_CONNECTED;
633 err = 0;
634out:
635 release_sock(sk);
636 return err;
637
638sock_error:
639 err = sock_error(sk) ? : -ECONNABORTED;
640 sock->state = SS_UNCONNECTED;
641 if (sk->sk_prot->disconnect(sk, flags))
642 sock->state = SS_DISCONNECTING;
643 goto out;
644}
645
646/*
647 * Accept a pending connection. The TCP layer now gives BSD semantics.
648 */
649
650int inet_accept(struct socket *sock, struct socket *newsock, int flags)
651{
652 struct sock *sk1 = sock->sk;
653 int err = -EINVAL;
654 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
655
656 if (!sk2)
657 goto do_err;
658
659 lock_sock(sk2);
660
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700661 WARN_ON(!((1 << sk2->sk_state) &
662 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
664 sock_graft(sk2, newsock);
665
666 newsock->state = SS_CONNECTED;
667 err = 0;
668 release_sock(sk2);
669do_err:
670 return err;
671}
672
673
674/*
675 * This does both peername and sockname.
676 */
677int inet_getname(struct socket *sock, struct sockaddr *uaddr,
678 int *uaddr_len, int peer)
679{
680 struct sock *sk = sock->sk;
681 struct inet_sock *inet = inet_sk(sk);
682 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
683
684 sin->sin_family = AF_INET;
685 if (peer) {
686 if (!inet->dport ||
687 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
688 peer == 1))
689 return -ENOTCONN;
690 sin->sin_port = inet->dport;
691 sin->sin_addr.s_addr = inet->daddr;
692 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800693 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 if (!addr)
695 addr = inet->saddr;
696 sin->sin_port = inet->sport;
697 sin->sin_addr.s_addr = addr;
698 }
699 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
700 *uaddr_len = sizeof(*sin);
701 return 0;
702}
703
704int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
705 size_t size)
706{
707 struct sock *sk = sock->sk;
708
709 /* We may need to bind the socket. */
710 if (!inet_sk(sk)->num && inet_autobind(sk))
711 return -EAGAIN;
712
713 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
714}
715
716
717static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
718{
719 struct sock *sk = sock->sk;
720
721 /* We may need to bind the socket. */
722 if (!inet_sk(sk)->num && inet_autobind(sk))
723 return -EAGAIN;
724
725 if (sk->sk_prot->sendpage)
726 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
727 return sock_no_sendpage(sock, page, offset, size, flags);
728}
729
730
731int inet_shutdown(struct socket *sock, int how)
732{
733 struct sock *sk = sock->sk;
734 int err = 0;
735
736 /* This should really check to make sure
737 * the socket is a TCP socket. (WHY AC...)
738 */
739 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
740 1->2 bit 2 snds.
741 2->3 */
742 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
743 return -EINVAL;
744
745 lock_sock(sk);
746 if (sock->state == SS_CONNECTING) {
747 if ((1 << sk->sk_state) &
748 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
749 sock->state = SS_DISCONNECTING;
750 else
751 sock->state = SS_CONNECTED;
752 }
753
754 switch (sk->sk_state) {
755 case TCP_CLOSE:
756 err = -ENOTCONN;
757 /* Hack to wake up other listeners, who can poll for
758 POLLHUP, even on eg. unconnected UDP sockets -- RR */
759 default:
760 sk->sk_shutdown |= how;
761 if (sk->sk_prot->shutdown)
762 sk->sk_prot->shutdown(sk, how);
763 break;
764
765 /* Remaining two branches are temporary solution for missing
766 * close() in multithreaded environment. It is _not_ a good idea,
767 * but we have no choice until close() is repaired at VFS level.
768 */
769 case TCP_LISTEN:
770 if (!(how & RCV_SHUTDOWN))
771 break;
772 /* Fall through */
773 case TCP_SYN_SENT:
774 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
775 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
776 break;
777 }
778
779 /* Wake up anyone sleeping in poll. */
780 sk->sk_state_change(sk);
781 release_sock(sk);
782 return err;
783}
784
785/*
786 * ioctl() calls you can issue on an INET socket. Most of these are
787 * device configuration and stuff and very rarely used. Some ioctls
788 * pass on to the socket itself.
789 *
790 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
791 * loads the devconfigure module does its configuring and unloads it.
792 * There's a good 20K of config code hanging around the kernel.
793 */
794
795int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
796{
797 struct sock *sk = sock->sk;
798 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900799 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
801 switch (cmd) {
802 case SIOCGSTAMP:
803 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
804 break;
Eric Dumazetae40eb12007-03-18 17:33:16 -0700805 case SIOCGSTAMPNS:
806 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
807 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 case SIOCADDRT:
809 case SIOCDELRT:
810 case SIOCRTMSG:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800811 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 break;
813 case SIOCDARP:
814 case SIOCGARP:
815 case SIOCSARP:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800816 err = arp_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 break;
818 case SIOCGIFADDR:
819 case SIOCSIFADDR:
820 case SIOCGIFBRDADDR:
821 case SIOCSIFBRDADDR:
822 case SIOCGIFNETMASK:
823 case SIOCSIFNETMASK:
824 case SIOCGIFDSTADDR:
825 case SIOCSIFDSTADDR:
826 case SIOCSIFPFLAGS:
827 case SIOCGIFPFLAGS:
828 case SIOCSIFFLAGS:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800829 err = devinet_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 break;
831 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800832 if (sk->sk_prot->ioctl)
833 err = sk->sk_prot->ioctl(sk, cmd, arg);
834 else
835 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 break;
837 }
838 return err;
839}
840
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800841const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800842 .family = PF_INET,
843 .owner = THIS_MODULE,
844 .release = inet_release,
845 .bind = inet_bind,
846 .connect = inet_stream_connect,
847 .socketpair = sock_no_socketpair,
848 .accept = inet_accept,
849 .getname = inet_getname,
850 .poll = tcp_poll,
851 .ioctl = inet_ioctl,
852 .listen = inet_listen,
853 .shutdown = inet_shutdown,
854 .setsockopt = sock_common_setsockopt,
855 .getsockopt = sock_common_getsockopt,
David S. Miller3516ffb2007-08-02 19:23:56 -0700856 .sendmsg = tcp_sendmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800857 .recvmsg = sock_common_recvmsg,
858 .mmap = sock_no_mmap,
859 .sendpage = tcp_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800860 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800861#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800862 .compat_setsockopt = compat_sock_common_setsockopt,
863 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800864#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865};
866
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800867const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800868 .family = PF_INET,
869 .owner = THIS_MODULE,
870 .release = inet_release,
871 .bind = inet_bind,
872 .connect = inet_dgram_connect,
873 .socketpair = sock_no_socketpair,
874 .accept = sock_no_accept,
875 .getname = inet_getname,
876 .poll = udp_poll,
877 .ioctl = inet_ioctl,
878 .listen = sock_no_listen,
879 .shutdown = inet_shutdown,
880 .setsockopt = sock_common_setsockopt,
881 .getsockopt = sock_common_getsockopt,
882 .sendmsg = inet_sendmsg,
883 .recvmsg = sock_common_recvmsg,
884 .mmap = sock_no_mmap,
885 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800886#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800887 .compat_setsockopt = compat_sock_common_setsockopt,
888 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800889#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890};
891
892/*
893 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
894 * udp_poll
895 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800896static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800897 .family = PF_INET,
898 .owner = THIS_MODULE,
899 .release = inet_release,
900 .bind = inet_bind,
901 .connect = inet_dgram_connect,
902 .socketpair = sock_no_socketpair,
903 .accept = sock_no_accept,
904 .getname = inet_getname,
905 .poll = datagram_poll,
906 .ioctl = inet_ioctl,
907 .listen = sock_no_listen,
908 .shutdown = inet_shutdown,
909 .setsockopt = sock_common_setsockopt,
910 .getsockopt = sock_common_getsockopt,
911 .sendmsg = inet_sendmsg,
912 .recvmsg = sock_common_recvmsg,
913 .mmap = sock_no_mmap,
914 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800915#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800916 .compat_setsockopt = compat_sock_common_setsockopt,
917 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800918#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919};
920
921static struct net_proto_family inet_family_ops = {
922 .family = PF_INET,
923 .create = inet_create,
924 .owner = THIS_MODULE,
925};
926
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927/* Upon startup we insert all the elements in inetsw_array[] into
928 * the linked list inetsw.
929 */
930static struct inet_protosw inetsw_array[] =
931{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900932 {
933 .type = SOCK_STREAM,
934 .protocol = IPPROTO_TCP,
935 .prot = &tcp_prot,
936 .ops = &inet_stream_ops,
937 .capability = -1,
938 .no_check = 0,
939 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800940 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900941 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900943 {
944 .type = SOCK_DGRAM,
945 .protocol = IPPROTO_UDP,
946 .prot = &udp_prot,
947 .ops = &inet_dgram_ops,
948 .capability = -1,
949 .no_check = UDP_CSUM_DEFAULT,
950 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
954 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900955 .type = SOCK_RAW,
956 .protocol = IPPROTO_IP, /* wild card */
957 .prot = &raw_prot,
958 .ops = &inet_sockraw_ops,
959 .capability = CAP_NET_RAW,
960 .no_check = UDP_CSUM_DEFAULT,
961 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 }
963};
964
Denis Chengc40f6ff2007-09-16 16:39:25 -0700965#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
967void inet_register_protosw(struct inet_protosw *p)
968{
969 struct list_head *lh;
970 struct inet_protosw *answer;
971 int protocol = p->protocol;
972 struct list_head *last_perm;
973
974 spin_lock_bh(&inetsw_lock);
975
976 if (p->type >= SOCK_MAX)
977 goto out_illegal;
978
979 /* If we are trying to override a permanent protocol, bail. */
980 answer = NULL;
981 last_perm = &inetsw[p->type];
982 list_for_each(lh, &inetsw[p->type]) {
983 answer = list_entry(lh, struct inet_protosw, list);
984
985 /* Check only the non-wild match. */
986 if (INET_PROTOSW_PERMANENT & answer->flags) {
987 if (protocol == answer->protocol)
988 break;
989 last_perm = lh;
990 }
991
992 answer = NULL;
993 }
994 if (answer)
995 goto out_permanent;
996
997 /* Add the new entry after the last permanent entry if any, so that
998 * the new entry does not override a permanent entry when matched with
999 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001000 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 * system automatically returns to the old behavior.
1002 */
1003 list_add_rcu(&p->list, last_perm);
1004out:
1005 spin_unlock_bh(&inetsw_lock);
1006
1007 synchronize_net();
1008
1009 return;
1010
1011out_permanent:
1012 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1013 protocol);
1014 goto out;
1015
1016out_illegal:
1017 printk(KERN_ERR
1018 "Ignoring attempt to register invalid socket type %d.\n",
1019 p->type);
1020 goto out;
1021}
1022
1023void inet_unregister_protosw(struct inet_protosw *p)
1024{
1025 if (INET_PROTOSW_PERMANENT & p->flags) {
1026 printk(KERN_ERR
1027 "Attempt to unregister permanent protocol %d.\n",
1028 p->protocol);
1029 } else {
1030 spin_lock_bh(&inetsw_lock);
1031 list_del_rcu(&p->list);
1032 spin_unlock_bh(&inetsw_lock);
1033
1034 synchronize_net();
1035 }
1036}
1037
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001038/*
1039 * Shall we try to damage output packets if routing dev changes?
1040 */
1041
Brian Haleyab32ea52006-09-22 14:15:41 -07001042int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001043
1044static int inet_sk_reselect_saddr(struct sock *sk)
1045{
1046 struct inet_sock *inet = inet_sk(sk);
1047 int err;
1048 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001049 __be32 old_saddr = inet->saddr;
1050 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001051 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001052
1053 if (inet->opt && inet->opt->srr)
1054 daddr = inet->opt->faddr;
1055
1056 /* Query new route. */
1057 err = ip_route_connect(&rt, daddr, 0,
1058 RT_CONN_FLAGS(sk),
1059 sk->sk_bound_dev_if,
1060 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001061 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001062 if (err)
1063 return err;
1064
1065 sk_setup_caps(sk, &rt->u.dst);
1066
1067 new_saddr = rt->rt_src;
1068
1069 if (new_saddr == old_saddr)
1070 return 0;
1071
1072 if (sysctl_ip_dynaddr > 1) {
1073 printk(KERN_INFO "%s(): shifting inet->"
YOSHIFUJI Hideakia7d632b2008-04-14 04:09:00 -07001074 "saddr from " NIPQUAD_FMT " to " NIPQUAD_FMT "\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001075 __func__,
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001076 NIPQUAD(old_saddr),
1077 NIPQUAD(new_saddr));
1078 }
1079
1080 inet->saddr = inet->rcv_saddr = new_saddr;
1081
1082 /*
1083 * XXX The only one ugly spot where we need to
1084 * XXX really change the sockets identity after
1085 * XXX it has entered the hashes. -DaveM
1086 *
1087 * Besides that, it does not check for connection
1088 * uniqueness. Wait for troubles.
1089 */
1090 __sk_prot_rehash(sk);
1091 return 0;
1092}
1093
1094int inet_sk_rebuild_header(struct sock *sk)
1095{
1096 struct inet_sock *inet = inet_sk(sk);
1097 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001098 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001099 int err;
1100
1101 /* Route is OK, nothing to do. */
1102 if (rt)
1103 return 0;
1104
1105 /* Reroute. */
1106 daddr = inet->daddr;
1107 if (inet->opt && inet->opt->srr)
1108 daddr = inet->opt->faddr;
1109{
1110 struct flowi fl = {
1111 .oif = sk->sk_bound_dev_if,
1112 .nl_u = {
1113 .ip4_u = {
1114 .daddr = daddr,
1115 .saddr = inet->saddr,
1116 .tos = RT_CONN_FLAGS(sk),
1117 },
1118 },
1119 .proto = sk->sk_protocol,
1120 .uli_u = {
1121 .ports = {
1122 .sport = inet->sport,
1123 .dport = inet->dport,
1124 },
1125 },
1126 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001127
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001128 security_sk_classify_flow(sk, &fl);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001129 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001130}
1131 if (!err)
1132 sk_setup_caps(sk, &rt->u.dst);
1133 else {
1134 /* Routing failed... */
1135 sk->sk_route_caps = 0;
1136 /*
1137 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1138 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1139 */
1140 if (!sysctl_ip_dynaddr ||
1141 sk->sk_state != TCP_SYN_SENT ||
1142 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1143 (err = inet_sk_reselect_saddr(sk)) != 0)
1144 sk->sk_err_soft = -err;
1145 }
1146
1147 return err;
1148}
1149
1150EXPORT_SYMBOL(inet_sk_rebuild_header);
1151
Herbert Xua430a432006-07-08 13:34:56 -07001152static int inet_gso_send_check(struct sk_buff *skb)
1153{
1154 struct iphdr *iph;
1155 struct net_protocol *ops;
1156 int proto;
1157 int ihl;
1158 int err = -EINVAL;
1159
1160 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1161 goto out;
1162
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001163 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001164 ihl = iph->ihl * 4;
1165 if (ihl < sizeof(*iph))
1166 goto out;
1167
1168 if (unlikely(!pskb_may_pull(skb, ihl)))
1169 goto out;
1170
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001171 __skb_pull(skb, ihl);
1172 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001173 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001174 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1175 err = -EPROTONOSUPPORT;
1176
1177 rcu_read_lock();
1178 ops = rcu_dereference(inet_protos[proto]);
1179 if (likely(ops && ops->gso_send_check))
1180 err = ops->gso_send_check(skb);
1181 rcu_read_unlock();
1182
1183out:
1184 return err;
1185}
1186
Herbert Xu576a30e2006-06-27 13:22:38 -07001187static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001188{
1189 struct sk_buff *segs = ERR_PTR(-EINVAL);
1190 struct iphdr *iph;
1191 struct net_protocol *ops;
1192 int proto;
1193 int ihl;
1194 int id;
1195
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001196 if (!(features & NETIF_F_V4_CSUM))
1197 features &= ~NETIF_F_SG;
1198
Herbert Xubbcf4672006-07-03 19:38:35 -07001199 if (unlikely(skb_shinfo(skb)->gso_type &
1200 ~(SKB_GSO_TCPV4 |
1201 SKB_GSO_UDP |
1202 SKB_GSO_DODGY |
1203 SKB_GSO_TCP_ECN |
1204 0)))
1205 goto out;
1206
1207 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001208 goto out;
1209
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001210 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001211 ihl = iph->ihl * 4;
1212 if (ihl < sizeof(*iph))
1213 goto out;
1214
Herbert Xubbcf4672006-07-03 19:38:35 -07001215 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001216 goto out;
1217
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001218 __skb_pull(skb, ihl);
1219 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001220 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001221 id = ntohs(iph->id);
1222 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1223 segs = ERR_PTR(-EPROTONOSUPPORT);
1224
1225 rcu_read_lock();
1226 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001227 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001228 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001229 rcu_read_unlock();
1230
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001231 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001232 goto out;
1233
1234 skb = segs;
1235 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001236 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001237 iph->id = htons(id++);
1238 iph->tot_len = htons(skb->len - skb->mac_len);
1239 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001240 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001241 } while ((skb = skb->next));
1242
1243out:
1244 return segs;
1245}
1246
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001247int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001248 unsigned short type, unsigned char protocol,
1249 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001250{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001251 struct socket *sock;
1252 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001253
1254 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001255 *sk = sock->sk;
1256 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001257 /*
1258 * Unhash it so that IP input processing does not even see it,
1259 * we do not wish this socket to see incoming packets.
1260 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001261 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001262
1263 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001264 }
1265 return rc;
1266}
1267
1268EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1269
Herbert Xu5e0f0432007-04-24 21:53:35 -07001270unsigned long snmp_fold_field(void *mib[], int offt)
1271{
1272 unsigned long res = 0;
1273 int i;
1274
1275 for_each_possible_cpu(i) {
1276 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1277 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1278 }
1279 return res;
1280}
1281EXPORT_SYMBOL_GPL(snmp_fold_field);
1282
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001283int snmp_mib_init(void *ptr[2], size_t mibsize)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001284{
1285 BUG_ON(ptr == NULL);
1286 ptr[0] = __alloc_percpu(mibsize);
1287 if (!ptr[0])
1288 goto err0;
1289 ptr[1] = __alloc_percpu(mibsize);
1290 if (!ptr[1])
1291 goto err1;
1292 return 0;
1293err1:
1294 free_percpu(ptr[0]);
1295 ptr[0] = NULL;
1296err0:
1297 return -ENOMEM;
1298}
1299EXPORT_SYMBOL_GPL(snmp_mib_init);
1300
1301void snmp_mib_free(void *ptr[2])
1302{
1303 BUG_ON(ptr == NULL);
1304 free_percpu(ptr[0]);
1305 free_percpu(ptr[1]);
1306 ptr[0] = ptr[1] = NULL;
1307}
1308EXPORT_SYMBOL_GPL(snmp_mib_free);
1309
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310#ifdef CONFIG_IP_MULTICAST
1311static struct net_protocol igmp_protocol = {
1312 .handler = igmp_rcv,
1313};
1314#endif
1315
1316static struct net_protocol tcp_protocol = {
1317 .handler = tcp_v4_rcv,
1318 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001319 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001320 .gso_segment = tcp_tso_segment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001322 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323};
1324
1325static struct net_protocol udp_protocol = {
1326 .handler = udp_rcv,
1327 .err_handler = udp_err,
1328 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001329 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330};
1331
1332static struct net_protocol icmp_protocol = {
1333 .handler = icmp_rcv,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001334 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001335 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336};
1337
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001338static __net_init int ipv4_mib_init_net(struct net *net)
1339{
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001340 if (snmp_mib_init((void **)net->mib.tcp_statistics,
1341 sizeof(struct tcp_mib)) < 0)
1342 goto err_tcp_mib;
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001343 if (snmp_mib_init((void **)net->mib.ip_statistics,
1344 sizeof(struct ipstats_mib)) < 0)
1345 goto err_ip_mib;
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001346 if (snmp_mib_init((void **)net->mib.net_statistics,
1347 sizeof(struct linux_mib)) < 0)
1348 goto err_net_mib;
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001349 if (snmp_mib_init((void **)net->mib.udp_statistics,
1350 sizeof(struct udp_mib)) < 0)
1351 goto err_udp_mib;
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001352 if (snmp_mib_init((void **)net->mib.udplite_statistics,
1353 sizeof(struct udp_mib)) < 0)
1354 goto err_udplite_mib;
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001355 if (snmp_mib_init((void **)net->mib.icmp_statistics,
1356 sizeof(struct icmp_mib)) < 0)
1357 goto err_icmp_mib;
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001358 if (snmp_mib_init((void **)net->mib.icmpmsg_statistics,
1359 sizeof(struct icmpmsg_mib)) < 0)
1360 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001361
1362 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001363 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001364
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001365err_icmpmsg_mib:
1366 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001367err_icmp_mib:
1368 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001369err_udplite_mib:
1370 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001371err_udp_mib:
1372 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001373err_net_mib:
1374 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001375err_ip_mib:
1376 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001377err_tcp_mib:
1378 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001379}
1380
1381static __net_exit void ipv4_mib_exit_net(struct net *net)
1382{
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001383 snmp_mib_free((void **)net->mib.icmpmsg_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001384 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001385 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001386 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001387 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001388 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001389 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001390}
1391
1392static __net_initdata struct pernet_operations ipv4_mib_ops = {
1393 .init = ipv4_mib_init_net,
1394 .exit = ipv4_mib_exit_net,
1395};
1396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397static int __init init_ipv4_mibs(void)
1398{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001399 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400}
1401
1402static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Herbert Xu30e224d2005-07-05 14:40:10 -07001404/*
1405 * IP protocol layer initialiser
1406 */
1407
1408static struct packet_type ip_packet_type = {
1409 .type = __constant_htons(ETH_P_IP),
1410 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001411 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001412 .gso_segment = inet_gso_segment,
Herbert Xu30e224d2005-07-05 14:40:10 -07001413};
1414
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415static int __init inet_init(void)
1416{
1417 struct sk_buff *dummy_skb;
1418 struct inet_protosw *q;
1419 struct list_head *r;
1420 int rc = -EINVAL;
1421
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001422 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
1424 rc = proto_register(&tcp_prot, 1);
1425 if (rc)
1426 goto out;
1427
1428 rc = proto_register(&udp_prot, 1);
1429 if (rc)
1430 goto out_unregister_tcp_proto;
1431
1432 rc = proto_register(&raw_prot, 1);
1433 if (rc)
1434 goto out_unregister_udp_proto;
1435
1436 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001437 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 */
1439
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001440 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
Al Virobd7b1532008-07-15 16:00:59 -04001442#ifdef CONFIG_SYSCTL
1443 ip_static_sysctl_init();
1444#endif
1445
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 /*
1447 * Add all the base protocols.
1448 */
1449
1450 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1451 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1452 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1453 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1454 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1455 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1456#ifdef CONFIG_IP_MULTICAST
1457 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1458 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1459#endif
1460
1461 /* Register the socket-side information for inet_create. */
1462 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1463 INIT_LIST_HEAD(r);
1464
1465 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1466 inet_register_protosw(q);
1467
1468 /*
1469 * Set the ARP module up
1470 */
1471
1472 arp_init();
1473
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001474 /*
1475 * Set the IP module up
1476 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
1478 ip_init();
1479
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001480 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
1482 /* Setup TCP slab cache for open requests. */
1483 tcp_init();
1484
Hideo Aoki95766ff2007-12-31 00:29:24 -08001485 /* Setup UDP memory threshold */
1486 udp_init();
1487
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001488 /* Add UDP-Lite (RFC 3828) */
1489 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490
1491 /*
1492 * Set the ICMP layer up
1493 */
1494
Denis V. Luneva5710d62008-02-29 11:14:50 -08001495 if (icmp_init() < 0)
1496 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497
1498 /*
1499 * Initialise the multicast router
1500 */
1501#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001502 if (ip_mr_init())
1503 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504#endif
1505 /*
1506 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001507 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
Stephen Hemminger132adf52007-03-08 20:44:43 -08001509 if (init_ipv4_mibs())
Wang Chen03d2f892008-07-03 12:13:36 +08001510 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001511
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 ipv4_proc_init();
1513
1514 ipfrag_init();
1515
Herbert Xu30e224d2005-07-05 14:40:10 -07001516 dev_add_pack(&ip_packet_type);
1517
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 rc = 0;
1519out:
1520 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521out_unregister_udp_proto:
1522 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001523out_unregister_tcp_proto:
1524 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 goto out;
1526}
1527
Heiko Carstensa536e072006-04-28 15:19:17 -07001528fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
1530/* ------------------------------------------------------------------------ */
1531
1532#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533static int __init ipv4_proc_init(void)
1534{
1535 int rc = 0;
1536
1537 if (raw_proc_init())
1538 goto out_raw;
1539 if (tcp4_proc_init())
1540 goto out_tcp;
1541 if (udp4_proc_init())
1542 goto out_udp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 if (ip_misc_proc_init())
1544 goto out_misc;
1545out:
1546 return rc;
1547out_misc:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 udp4_proc_exit();
1549out_udp:
1550 tcp4_proc_exit();
1551out_tcp:
1552 raw_proc_exit();
1553out_raw:
1554 rc = -ENOMEM;
1555 goto out;
1556}
1557
1558#else /* CONFIG_PROC_FS */
1559static int __init ipv4_proc_init(void)
1560{
1561 return 0;
1562}
1563#endif /* CONFIG_PROC_FS */
1564
1565MODULE_ALIAS_NETPROTO(PF_INET);
1566
1567EXPORT_SYMBOL(inet_accept);
1568EXPORT_SYMBOL(inet_bind);
1569EXPORT_SYMBOL(inet_dgram_connect);
1570EXPORT_SYMBOL(inet_dgram_ops);
1571EXPORT_SYMBOL(inet_getname);
1572EXPORT_SYMBOL(inet_ioctl);
1573EXPORT_SYMBOL(inet_listen);
1574EXPORT_SYMBOL(inet_register_protosw);
1575EXPORT_SYMBOL(inet_release);
1576EXPORT_SYMBOL(inet_sendmsg);
1577EXPORT_SYMBOL(inet_shutdown);
1578EXPORT_SYMBOL(inet_sock_destruct);
1579EXPORT_SYMBOL(inet_stream_connect);
1580EXPORT_SYMBOL(inet_stream_ops);
1581EXPORT_SYMBOL(inet_unregister_protosw);
Neil Hormancdac4e02005-06-13 15:12:33 -07001582EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);