blob: 1f418164ebf2720cf96e07846d9abf5fdf2d2135 [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
151 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
152 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
153 BUG_TRAP(!sk->sk_wmem_queued);
154 BUG_TRAP(!sk->sk_forward_alloc);
155
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;
267 struct list_head *p;
268 struct inet_protosw *answer;
269 struct inet_sock *inet;
270 struct proto *answer_prot;
271 unsigned char answer_flags;
272 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700273 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800274 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275
David S. Millerb3da2cf2007-03-23 11:40:27 -0700276 if (sock->type != SOCK_RAW &&
277 sock->type != SOCK_DGRAM &&
278 !inet_ehash_secret)
279 build_ehash_secret();
280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 sock->state = SS_UNCONNECTED;
282
283 /* Look for the requested type/protocol pair. */
284 answer = NULL;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700285lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800286 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 rcu_read_lock();
288 list_for_each_rcu(p, &inetsw[sock->type]) {
289 answer = list_entry(p, struct inet_protosw, list);
290
291 /* Check the non-wild match. */
292 if (protocol == answer->protocol) {
293 if (protocol != IPPROTO_IP)
294 break;
295 } else {
296 /* Check for the two wild cases. */
297 if (IPPROTO_IP == protocol) {
298 protocol = answer->protocol;
299 break;
300 }
301 if (IPPROTO_IP == answer->protocol)
302 break;
303 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800304 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 answer = NULL;
306 }
307
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700308 if (unlikely(answer == NULL)) {
309 if (try_loading_module < 2) {
310 rcu_read_unlock();
311 /*
312 * Be more specific, e.g. net-pf-2-proto-132-type-1
313 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
314 */
315 if (++try_loading_module == 1)
316 request_module("net-pf-%d-proto-%d-type-%d",
317 PF_INET, protocol, sock->type);
318 /*
319 * Fall back to generic, e.g. net-pf-2-proto-132
320 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
321 */
322 else
323 request_module("net-pf-%d-proto-%d",
324 PF_INET, protocol);
325 goto lookup_protocol;
326 } else
327 goto out_rcu_unlock;
328 }
329
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 err = -EPERM;
331 if (answer->capability > 0 && !capable(answer->capability))
332 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700334 err = -EAFNOSUPPORT;
335 if (!inet_netns_ok(net, protocol))
336 goto out_rcu_unlock;
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 sock->ops = answer->ops;
339 answer_prot = answer->prot;
340 answer_no_check = answer->no_check;
341 answer_flags = answer->flags;
342 rcu_read_unlock();
343
344 BUG_TRAP(answer_prot->slab != NULL);
345
346 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700347 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 if (sk == NULL)
349 goto out;
350
351 err = 0;
352 sk->sk_no_check = answer_no_check;
353 if (INET_PROTOSW_REUSE & answer_flags)
354 sk->sk_reuse = 1;
355
356 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800357 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358
359 if (SOCK_RAW == sock->type) {
360 inet->num = protocol;
361 if (IPPROTO_RAW == protocol)
362 inet->hdrincl = 1;
363 }
364
365 if (ipv4_config.no_pmtu_disc)
366 inet->pmtudisc = IP_PMTUDISC_DONT;
367 else
368 inet->pmtudisc = IP_PMTUDISC_WANT;
369
370 inet->id = 0;
371
372 sock_init_data(sock, sk);
373
374 sk->sk_destruct = inet_sock_destruct;
375 sk->sk_family = PF_INET;
376 sk->sk_protocol = protocol;
377 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
378
379 inet->uc_ttl = -1;
380 inet->mc_loop = 1;
381 inet->mc_ttl = 1;
382 inet->mc_index = 0;
383 inet->mc_list = NULL;
384
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700385 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386
387 if (inet->num) {
388 /* It assumes that any protocol which allows
389 * the user to assign a number at socket
390 * creation time automatically
391 * shares.
392 */
393 inet->sport = htons(inet->num);
394 /* Add to protocol hash chains. */
395 sk->sk_prot->hash(sk);
396 }
397
398 if (sk->sk_prot->init) {
399 err = sk->sk_prot->init(sk);
400 if (err)
401 sk_common_release(sk);
402 }
403out:
404 return err;
405out_rcu_unlock:
406 rcu_read_unlock();
407 goto out;
408}
409
410
411/*
412 * The peer socket should always be NULL (or else). When we call this
413 * function we are destroying the object and from then on nobody
414 * should refer to it.
415 */
416int inet_release(struct socket *sock)
417{
418 struct sock *sk = sock->sk;
419
420 if (sk) {
421 long timeout;
422
423 /* Applications forget to leave groups before exiting */
424 ip_mc_drop_socket(sk);
425
426 /* If linger is set, we don't return until the close
427 * is complete. Otherwise we return immediately. The
428 * actually closing is done the same either way.
429 *
430 * If the close is due to the process exiting, we never
431 * linger..
432 */
433 timeout = 0;
434 if (sock_flag(sk, SOCK_LINGER) &&
435 !(current->flags & PF_EXITING))
436 timeout = sk->sk_lingertime;
437 sock->sk = NULL;
438 sk->sk_prot->close(sk, timeout);
439 }
440 return 0;
441}
442
443/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700444int sysctl_ip_nonlocal_bind __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
447{
448 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
449 struct sock *sk = sock->sk;
450 struct inet_sock *inet = inet_sk(sk);
451 unsigned short snum;
452 int chk_addr_ret;
453 int err;
454
455 /* If the socket has its own bind function then use it. (RAW) */
456 if (sk->sk_prot->bind) {
457 err = sk->sk_prot->bind(sk, uaddr, addr_len);
458 goto out;
459 }
460 err = -EINVAL;
461 if (addr_len < sizeof(struct sockaddr_in))
462 goto out;
463
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900464 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466 /* Not specified by any standard per-se, however it breaks too
467 * many applications when removed. It is unfortunate since
468 * allowing applications to make a non-local bind solves
469 * several problems with systems using dynamic addressing.
470 * (ie. your servers still start up even if your ISDN link
471 * is temporarily down)
472 */
473 err = -EADDRNOTAVAIL;
474 if (!sysctl_ip_nonlocal_bind &&
475 !inet->freebind &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700476 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 chk_addr_ret != RTN_LOCAL &&
478 chk_addr_ret != RTN_MULTICAST &&
479 chk_addr_ret != RTN_BROADCAST)
480 goto out;
481
482 snum = ntohs(addr->sin_port);
483 err = -EACCES;
484 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
485 goto out;
486
487 /* We keep a pair of addresses. rcv_saddr is the one
488 * used by hash lookups, and saddr is used for transmit.
489 *
490 * In the BSD API these are the same except where it
491 * would be illegal to use them (multicast/broadcast) in
492 * which case the sending device address is used.
493 */
494 lock_sock(sk);
495
496 /* Check these errors (active socket, double bind). */
497 err = -EINVAL;
498 if (sk->sk_state != TCP_CLOSE || inet->num)
499 goto out_release_sock;
500
501 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
502 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
503 inet->saddr = 0; /* Use device */
504
505 /* Make sure we are allowed to bind here. */
506 if (sk->sk_prot->get_port(sk, snum)) {
507 inet->saddr = inet->rcv_saddr = 0;
508 err = -EADDRINUSE;
509 goto out_release_sock;
510 }
511
512 if (inet->rcv_saddr)
513 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
514 if (snum)
515 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
516 inet->sport = htons(inet->num);
517 inet->daddr = 0;
518 inet->dport = 0;
519 sk_dst_reset(sk);
520 err = 0;
521out_release_sock:
522 release_sock(sk);
523out:
524 return err;
525}
526
527int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
528 int addr_len, int flags)
529{
530 struct sock *sk = sock->sk;
531
532 if (uaddr->sa_family == AF_UNSPEC)
533 return sk->sk_prot->disconnect(sk, flags);
534
535 if (!inet_sk(sk)->num && inet_autobind(sk))
536 return -EAGAIN;
537 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
538}
539
540static long inet_wait_for_connect(struct sock *sk, long timeo)
541{
542 DEFINE_WAIT(wait);
543
544 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
545
546 /* Basic assumption: if someone sets sk->sk_err, he _must_
547 * change state of the socket from TCP_SYN_*.
548 * Connect() does not allow to get error notifications
549 * without closing the socket.
550 */
551 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
552 release_sock(sk);
553 timeo = schedule_timeout(timeo);
554 lock_sock(sk);
555 if (signal_pending(current) || !timeo)
556 break;
557 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
558 }
559 finish_wait(sk->sk_sleep, &wait);
560 return timeo;
561}
562
563/*
564 * Connect to a remote host. There is regrettably still a little
565 * TCP 'magic' in here.
566 */
567int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
568 int addr_len, int flags)
569{
570 struct sock *sk = sock->sk;
571 int err;
572 long timeo;
573
574 lock_sock(sk);
575
576 if (uaddr->sa_family == AF_UNSPEC) {
577 err = sk->sk_prot->disconnect(sk, flags);
578 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
579 goto out;
580 }
581
582 switch (sock->state) {
583 default:
584 err = -EINVAL;
585 goto out;
586 case SS_CONNECTED:
587 err = -EISCONN;
588 goto out;
589 case SS_CONNECTING:
590 err = -EALREADY;
591 /* Fall out of switch with err, set for this state */
592 break;
593 case SS_UNCONNECTED:
594 err = -EISCONN;
595 if (sk->sk_state != TCP_CLOSE)
596 goto out;
597
598 err = sk->sk_prot->connect(sk, uaddr, addr_len);
599 if (err < 0)
600 goto out;
601
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900602 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
604 /* Just entered SS_CONNECTING state; the only
605 * difference is that return value in non-blocking
606 * case is EINPROGRESS, rather than EALREADY.
607 */
608 err = -EINPROGRESS;
609 break;
610 }
611
612 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
613
614 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
615 /* Error code is set above */
616 if (!timeo || !inet_wait_for_connect(sk, timeo))
617 goto out;
618
619 err = sock_intr_errno(timeo);
620 if (signal_pending(current))
621 goto out;
622 }
623
624 /* Connection was closed by RST, timeout, ICMP error
625 * or another process disconnected us.
626 */
627 if (sk->sk_state == TCP_CLOSE)
628 goto sock_error;
629
630 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
631 * and error was received after socket entered established state.
632 * Hence, it is handled normally after connect() return successfully.
633 */
634
635 sock->state = SS_CONNECTED;
636 err = 0;
637out:
638 release_sock(sk);
639 return err;
640
641sock_error:
642 err = sock_error(sk) ? : -ECONNABORTED;
643 sock->state = SS_UNCONNECTED;
644 if (sk->sk_prot->disconnect(sk, flags))
645 sock->state = SS_DISCONNECTING;
646 goto out;
647}
648
649/*
650 * Accept a pending connection. The TCP layer now gives BSD semantics.
651 */
652
653int inet_accept(struct socket *sock, struct socket *newsock, int flags)
654{
655 struct sock *sk1 = sock->sk;
656 int err = -EINVAL;
657 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
658
659 if (!sk2)
660 goto do_err;
661
662 lock_sock(sk2);
663
664 BUG_TRAP((1 << sk2->sk_state) &
665 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE));
666
667 sock_graft(sk2, newsock);
668
669 newsock->state = SS_CONNECTED;
670 err = 0;
671 release_sock(sk2);
672do_err:
673 return err;
674}
675
676
677/*
678 * This does both peername and sockname.
679 */
680int inet_getname(struct socket *sock, struct sockaddr *uaddr,
681 int *uaddr_len, int peer)
682{
683 struct sock *sk = sock->sk;
684 struct inet_sock *inet = inet_sk(sk);
685 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
686
687 sin->sin_family = AF_INET;
688 if (peer) {
689 if (!inet->dport ||
690 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
691 peer == 1))
692 return -ENOTCONN;
693 sin->sin_port = inet->dport;
694 sin->sin_addr.s_addr = inet->daddr;
695 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800696 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 if (!addr)
698 addr = inet->saddr;
699 sin->sin_port = inet->sport;
700 sin->sin_addr.s_addr = addr;
701 }
702 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
703 *uaddr_len = sizeof(*sin);
704 return 0;
705}
706
707int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
708 size_t size)
709{
710 struct sock *sk = sock->sk;
711
712 /* We may need to bind the socket. */
713 if (!inet_sk(sk)->num && inet_autobind(sk))
714 return -EAGAIN;
715
716 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
717}
718
719
720static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
721{
722 struct sock *sk = sock->sk;
723
724 /* We may need to bind the socket. */
725 if (!inet_sk(sk)->num && inet_autobind(sk))
726 return -EAGAIN;
727
728 if (sk->sk_prot->sendpage)
729 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
730 return sock_no_sendpage(sock, page, offset, size, flags);
731}
732
733
734int inet_shutdown(struct socket *sock, int how)
735{
736 struct sock *sk = sock->sk;
737 int err = 0;
738
739 /* This should really check to make sure
740 * the socket is a TCP socket. (WHY AC...)
741 */
742 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
743 1->2 bit 2 snds.
744 2->3 */
745 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
746 return -EINVAL;
747
748 lock_sock(sk);
749 if (sock->state == SS_CONNECTING) {
750 if ((1 << sk->sk_state) &
751 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
752 sock->state = SS_DISCONNECTING;
753 else
754 sock->state = SS_CONNECTED;
755 }
756
757 switch (sk->sk_state) {
758 case TCP_CLOSE:
759 err = -ENOTCONN;
760 /* Hack to wake up other listeners, who can poll for
761 POLLHUP, even on eg. unconnected UDP sockets -- RR */
762 default:
763 sk->sk_shutdown |= how;
764 if (sk->sk_prot->shutdown)
765 sk->sk_prot->shutdown(sk, how);
766 break;
767
768 /* Remaining two branches are temporary solution for missing
769 * close() in multithreaded environment. It is _not_ a good idea,
770 * but we have no choice until close() is repaired at VFS level.
771 */
772 case TCP_LISTEN:
773 if (!(how & RCV_SHUTDOWN))
774 break;
775 /* Fall through */
776 case TCP_SYN_SENT:
777 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
778 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
779 break;
780 }
781
782 /* Wake up anyone sleeping in poll. */
783 sk->sk_state_change(sk);
784 release_sock(sk);
785 return err;
786}
787
788/*
789 * ioctl() calls you can issue on an INET socket. Most of these are
790 * device configuration and stuff and very rarely used. Some ioctls
791 * pass on to the socket itself.
792 *
793 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
794 * loads the devconfigure module does its configuring and unloads it.
795 * There's a good 20K of config code hanging around the kernel.
796 */
797
798int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
799{
800 struct sock *sk = sock->sk;
801 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900802 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803
804 switch (cmd) {
805 case SIOCGSTAMP:
806 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
807 break;
Eric Dumazetae40eb12007-03-18 17:33:16 -0700808 case SIOCGSTAMPNS:
809 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
810 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 case SIOCADDRT:
812 case SIOCDELRT:
813 case SIOCRTMSG:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800814 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 break;
816 case SIOCDARP:
817 case SIOCGARP:
818 case SIOCSARP:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800819 err = arp_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 break;
821 case SIOCGIFADDR:
822 case SIOCSIFADDR:
823 case SIOCGIFBRDADDR:
824 case SIOCSIFBRDADDR:
825 case SIOCGIFNETMASK:
826 case SIOCSIFNETMASK:
827 case SIOCGIFDSTADDR:
828 case SIOCSIFDSTADDR:
829 case SIOCSIFPFLAGS:
830 case SIOCGIFPFLAGS:
831 case SIOCSIFFLAGS:
Denis V. Luneve5b13cb2008-02-28 20:51:43 -0800832 err = devinet_ioctl(net, cmd, (void __user *)arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 break;
834 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800835 if (sk->sk_prot->ioctl)
836 err = sk->sk_prot->ioctl(sk, cmd, arg);
837 else
838 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 break;
840 }
841 return err;
842}
843
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800844const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800845 .family = PF_INET,
846 .owner = THIS_MODULE,
847 .release = inet_release,
848 .bind = inet_bind,
849 .connect = inet_stream_connect,
850 .socketpair = sock_no_socketpair,
851 .accept = inet_accept,
852 .getname = inet_getname,
853 .poll = tcp_poll,
854 .ioctl = inet_ioctl,
855 .listen = inet_listen,
856 .shutdown = inet_shutdown,
857 .setsockopt = sock_common_setsockopt,
858 .getsockopt = sock_common_getsockopt,
David S. Miller3516ffb2007-08-02 19:23:56 -0700859 .sendmsg = tcp_sendmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800860 .recvmsg = sock_common_recvmsg,
861 .mmap = sock_no_mmap,
862 .sendpage = tcp_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800863 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800864#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800865 .compat_setsockopt = compat_sock_common_setsockopt,
866 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800867#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868};
869
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800870const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800871 .family = PF_INET,
872 .owner = THIS_MODULE,
873 .release = inet_release,
874 .bind = inet_bind,
875 .connect = inet_dgram_connect,
876 .socketpair = sock_no_socketpair,
877 .accept = sock_no_accept,
878 .getname = inet_getname,
879 .poll = udp_poll,
880 .ioctl = inet_ioctl,
881 .listen = sock_no_listen,
882 .shutdown = inet_shutdown,
883 .setsockopt = sock_common_setsockopt,
884 .getsockopt = sock_common_getsockopt,
885 .sendmsg = inet_sendmsg,
886 .recvmsg = sock_common_recvmsg,
887 .mmap = sock_no_mmap,
888 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800889#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800890 .compat_setsockopt = compat_sock_common_setsockopt,
891 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800892#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893};
894
895/*
896 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
897 * udp_poll
898 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800899static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800900 .family = PF_INET,
901 .owner = THIS_MODULE,
902 .release = inet_release,
903 .bind = inet_bind,
904 .connect = inet_dgram_connect,
905 .socketpair = sock_no_socketpair,
906 .accept = sock_no_accept,
907 .getname = inet_getname,
908 .poll = datagram_poll,
909 .ioctl = inet_ioctl,
910 .listen = sock_no_listen,
911 .shutdown = inet_shutdown,
912 .setsockopt = sock_common_setsockopt,
913 .getsockopt = sock_common_getsockopt,
914 .sendmsg = inet_sendmsg,
915 .recvmsg = sock_common_recvmsg,
916 .mmap = sock_no_mmap,
917 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800918#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800919 .compat_setsockopt = compat_sock_common_setsockopt,
920 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800921#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922};
923
924static struct net_proto_family inet_family_ops = {
925 .family = PF_INET,
926 .create = inet_create,
927 .owner = THIS_MODULE,
928};
929
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930/* Upon startup we insert all the elements in inetsw_array[] into
931 * the linked list inetsw.
932 */
933static struct inet_protosw inetsw_array[] =
934{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900935 {
936 .type = SOCK_STREAM,
937 .protocol = IPPROTO_TCP,
938 .prot = &tcp_prot,
939 .ops = &inet_stream_ops,
940 .capability = -1,
941 .no_check = 0,
942 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800943 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900944 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900946 {
947 .type = SOCK_DGRAM,
948 .protocol = IPPROTO_UDP,
949 .prot = &udp_prot,
950 .ops = &inet_dgram_ops,
951 .capability = -1,
952 .no_check = UDP_CSUM_DEFAULT,
953 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900955
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956
957 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900958 .type = SOCK_RAW,
959 .protocol = IPPROTO_IP, /* wild card */
960 .prot = &raw_prot,
961 .ops = &inet_sockraw_ops,
962 .capability = CAP_NET_RAW,
963 .no_check = UDP_CSUM_DEFAULT,
964 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 }
966};
967
Denis Chengc40f6ff2007-09-16 16:39:25 -0700968#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
970void inet_register_protosw(struct inet_protosw *p)
971{
972 struct list_head *lh;
973 struct inet_protosw *answer;
974 int protocol = p->protocol;
975 struct list_head *last_perm;
976
977 spin_lock_bh(&inetsw_lock);
978
979 if (p->type >= SOCK_MAX)
980 goto out_illegal;
981
982 /* If we are trying to override a permanent protocol, bail. */
983 answer = NULL;
984 last_perm = &inetsw[p->type];
985 list_for_each(lh, &inetsw[p->type]) {
986 answer = list_entry(lh, struct inet_protosw, list);
987
988 /* Check only the non-wild match. */
989 if (INET_PROTOSW_PERMANENT & answer->flags) {
990 if (protocol == answer->protocol)
991 break;
992 last_perm = lh;
993 }
994
995 answer = NULL;
996 }
997 if (answer)
998 goto out_permanent;
999
1000 /* Add the new entry after the last permanent entry if any, so that
1001 * the new entry does not override a permanent entry when matched with
1002 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001003 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 * system automatically returns to the old behavior.
1005 */
1006 list_add_rcu(&p->list, last_perm);
1007out:
1008 spin_unlock_bh(&inetsw_lock);
1009
1010 synchronize_net();
1011
1012 return;
1013
1014out_permanent:
1015 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1016 protocol);
1017 goto out;
1018
1019out_illegal:
1020 printk(KERN_ERR
1021 "Ignoring attempt to register invalid socket type %d.\n",
1022 p->type);
1023 goto out;
1024}
1025
1026void inet_unregister_protosw(struct inet_protosw *p)
1027{
1028 if (INET_PROTOSW_PERMANENT & p->flags) {
1029 printk(KERN_ERR
1030 "Attempt to unregister permanent protocol %d.\n",
1031 p->protocol);
1032 } else {
1033 spin_lock_bh(&inetsw_lock);
1034 list_del_rcu(&p->list);
1035 spin_unlock_bh(&inetsw_lock);
1036
1037 synchronize_net();
1038 }
1039}
1040
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001041/*
1042 * Shall we try to damage output packets if routing dev changes?
1043 */
1044
Brian Haleyab32ea52006-09-22 14:15:41 -07001045int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001046
1047static int inet_sk_reselect_saddr(struct sock *sk)
1048{
1049 struct inet_sock *inet = inet_sk(sk);
1050 int err;
1051 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001052 __be32 old_saddr = inet->saddr;
1053 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001054 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001055
1056 if (inet->opt && inet->opt->srr)
1057 daddr = inet->opt->faddr;
1058
1059 /* Query new route. */
1060 err = ip_route_connect(&rt, daddr, 0,
1061 RT_CONN_FLAGS(sk),
1062 sk->sk_bound_dev_if,
1063 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001064 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001065 if (err)
1066 return err;
1067
1068 sk_setup_caps(sk, &rt->u.dst);
1069
1070 new_saddr = rt->rt_src;
1071
1072 if (new_saddr == old_saddr)
1073 return 0;
1074
1075 if (sysctl_ip_dynaddr > 1) {
1076 printk(KERN_INFO "%s(): shifting inet->"
YOSHIFUJI Hideakia7d632b2008-04-14 04:09:00 -07001077 "saddr from " NIPQUAD_FMT " to " NIPQUAD_FMT "\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001078 __func__,
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001079 NIPQUAD(old_saddr),
1080 NIPQUAD(new_saddr));
1081 }
1082
1083 inet->saddr = inet->rcv_saddr = new_saddr;
1084
1085 /*
1086 * XXX The only one ugly spot where we need to
1087 * XXX really change the sockets identity after
1088 * XXX it has entered the hashes. -DaveM
1089 *
1090 * Besides that, it does not check for connection
1091 * uniqueness. Wait for troubles.
1092 */
1093 __sk_prot_rehash(sk);
1094 return 0;
1095}
1096
1097int inet_sk_rebuild_header(struct sock *sk)
1098{
1099 struct inet_sock *inet = inet_sk(sk);
1100 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001101 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001102 int err;
1103
1104 /* Route is OK, nothing to do. */
1105 if (rt)
1106 return 0;
1107
1108 /* Reroute. */
1109 daddr = inet->daddr;
1110 if (inet->opt && inet->opt->srr)
1111 daddr = inet->opt->faddr;
1112{
1113 struct flowi fl = {
1114 .oif = sk->sk_bound_dev_if,
1115 .nl_u = {
1116 .ip4_u = {
1117 .daddr = daddr,
1118 .saddr = inet->saddr,
1119 .tos = RT_CONN_FLAGS(sk),
1120 },
1121 },
1122 .proto = sk->sk_protocol,
1123 .uli_u = {
1124 .ports = {
1125 .sport = inet->sport,
1126 .dport = inet->dport,
1127 },
1128 },
1129 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001130
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001131 security_sk_classify_flow(sk, &fl);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001132 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001133}
1134 if (!err)
1135 sk_setup_caps(sk, &rt->u.dst);
1136 else {
1137 /* Routing failed... */
1138 sk->sk_route_caps = 0;
1139 /*
1140 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1141 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1142 */
1143 if (!sysctl_ip_dynaddr ||
1144 sk->sk_state != TCP_SYN_SENT ||
1145 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1146 (err = inet_sk_reselect_saddr(sk)) != 0)
1147 sk->sk_err_soft = -err;
1148 }
1149
1150 return err;
1151}
1152
1153EXPORT_SYMBOL(inet_sk_rebuild_header);
1154
Herbert Xua430a432006-07-08 13:34:56 -07001155static int inet_gso_send_check(struct sk_buff *skb)
1156{
1157 struct iphdr *iph;
1158 struct net_protocol *ops;
1159 int proto;
1160 int ihl;
1161 int err = -EINVAL;
1162
1163 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1164 goto out;
1165
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001166 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001167 ihl = iph->ihl * 4;
1168 if (ihl < sizeof(*iph))
1169 goto out;
1170
1171 if (unlikely(!pskb_may_pull(skb, ihl)))
1172 goto out;
1173
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001174 __skb_pull(skb, ihl);
1175 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001176 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001177 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1178 err = -EPROTONOSUPPORT;
1179
1180 rcu_read_lock();
1181 ops = rcu_dereference(inet_protos[proto]);
1182 if (likely(ops && ops->gso_send_check))
1183 err = ops->gso_send_check(skb);
1184 rcu_read_unlock();
1185
1186out:
1187 return err;
1188}
1189
Herbert Xu576a30e2006-06-27 13:22:38 -07001190static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001191{
1192 struct sk_buff *segs = ERR_PTR(-EINVAL);
1193 struct iphdr *iph;
1194 struct net_protocol *ops;
1195 int proto;
1196 int ihl;
1197 int id;
1198
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001199 if (!(features & NETIF_F_V4_CSUM))
1200 features &= ~NETIF_F_SG;
1201
Herbert Xubbcf4672006-07-03 19:38:35 -07001202 if (unlikely(skb_shinfo(skb)->gso_type &
1203 ~(SKB_GSO_TCPV4 |
1204 SKB_GSO_UDP |
1205 SKB_GSO_DODGY |
1206 SKB_GSO_TCP_ECN |
1207 0)))
1208 goto out;
1209
1210 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001211 goto out;
1212
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001213 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001214 ihl = iph->ihl * 4;
1215 if (ihl < sizeof(*iph))
1216 goto out;
1217
Herbert Xubbcf4672006-07-03 19:38:35 -07001218 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001219 goto out;
1220
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001221 __skb_pull(skb, ihl);
1222 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001223 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001224 id = ntohs(iph->id);
1225 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1226 segs = ERR_PTR(-EPROTONOSUPPORT);
1227
1228 rcu_read_lock();
1229 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001230 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001231 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001232 rcu_read_unlock();
1233
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001234 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001235 goto out;
1236
1237 skb = segs;
1238 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001239 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001240 iph->id = htons(id++);
1241 iph->tot_len = htons(skb->len - skb->mac_len);
1242 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001243 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001244 } while ((skb = skb->next));
1245
1246out:
1247 return segs;
1248}
1249
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001250int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001251 unsigned short type, unsigned char protocol,
1252 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001253{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001254 struct socket *sock;
1255 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001256
1257 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001258 *sk = sock->sk;
1259 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001260 /*
1261 * Unhash it so that IP input processing does not even see it,
1262 * we do not wish this socket to see incoming packets.
1263 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001264 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001265
1266 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001267 }
1268 return rc;
1269}
1270
1271EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1272
Herbert Xu5e0f0432007-04-24 21:53:35 -07001273unsigned long snmp_fold_field(void *mib[], int offt)
1274{
1275 unsigned long res = 0;
1276 int i;
1277
1278 for_each_possible_cpu(i) {
1279 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1280 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1281 }
1282 return res;
1283}
1284EXPORT_SYMBOL_GPL(snmp_fold_field);
1285
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001286int snmp_mib_init(void *ptr[2], size_t mibsize)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001287{
1288 BUG_ON(ptr == NULL);
1289 ptr[0] = __alloc_percpu(mibsize);
1290 if (!ptr[0])
1291 goto err0;
1292 ptr[1] = __alloc_percpu(mibsize);
1293 if (!ptr[1])
1294 goto err1;
1295 return 0;
1296err1:
1297 free_percpu(ptr[0]);
1298 ptr[0] = NULL;
1299err0:
1300 return -ENOMEM;
1301}
1302EXPORT_SYMBOL_GPL(snmp_mib_init);
1303
1304void snmp_mib_free(void *ptr[2])
1305{
1306 BUG_ON(ptr == NULL);
1307 free_percpu(ptr[0]);
1308 free_percpu(ptr[1]);
1309 ptr[0] = ptr[1] = NULL;
1310}
1311EXPORT_SYMBOL_GPL(snmp_mib_free);
1312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313#ifdef CONFIG_IP_MULTICAST
1314static struct net_protocol igmp_protocol = {
1315 .handler = igmp_rcv,
1316};
1317#endif
1318
1319static struct net_protocol tcp_protocol = {
1320 .handler = tcp_v4_rcv,
1321 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001322 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001323 .gso_segment = tcp_tso_segment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001325 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326};
1327
1328static struct net_protocol udp_protocol = {
1329 .handler = udp_rcv,
1330 .err_handler = udp_err,
1331 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001332 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333};
1334
1335static struct net_protocol icmp_protocol = {
1336 .handler = icmp_rcv,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001337 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001338 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339};
1340
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001341static __net_init int ipv4_mib_init_net(struct net *net)
1342{
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001343 if (snmp_mib_init((void **)net->mib.tcp_statistics,
1344 sizeof(struct tcp_mib)) < 0)
1345 goto err_tcp_mib;
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001346 if (snmp_mib_init((void **)net->mib.ip_statistics,
1347 sizeof(struct ipstats_mib)) < 0)
1348 goto err_ip_mib;
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001349 if (snmp_mib_init((void **)net->mib.net_statistics,
1350 sizeof(struct linux_mib)) < 0)
1351 goto err_net_mib;
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001352 if (snmp_mib_init((void **)net->mib.udp_statistics,
1353 sizeof(struct udp_mib)) < 0)
1354 goto err_udp_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001355
1356 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001357 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001358
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001359err_udp_mib:
1360 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001361err_net_mib:
1362 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001363err_ip_mib:
1364 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001365err_tcp_mib:
1366 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001367}
1368
1369static __net_exit void ipv4_mib_exit_net(struct net *net)
1370{
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001371 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001372 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001373 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001374 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001375}
1376
1377static __net_initdata struct pernet_operations ipv4_mib_ops = {
1378 .init = ipv4_mib_init_net,
1379 .exit = ipv4_mib_exit_net,
1380};
1381
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382static int __init init_ipv4_mibs(void)
1383{
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001384 if (snmp_mib_init((void **)icmp_statistics,
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001385 sizeof(struct icmp_mib)) < 0)
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001386 goto err_icmp_mib;
David L Stevens96793b42007-09-17 09:57:33 -07001387 if (snmp_mib_init((void **)icmpmsg_statistics,
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001388 sizeof(struct icmpmsg_mib)) < 0)
David L Stevens96793b42007-09-17 09:57:33 -07001389 goto err_icmpmsg_mib;
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001390 if (snmp_mib_init((void **)udplite_statistics,
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001391 sizeof(struct udp_mib)) < 0)
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001392 goto err_udplite_mib;
David S. Millerdb8dac22008-03-06 16:22:02 -08001393
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001394 if (register_pernet_subsys(&ipv4_mib_ops))
1395 goto err_net;
1396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 return 0;
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001398
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001399err_net:
1400 snmp_mib_free((void **)udplite_statistics);
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001401err_udplite_mib:
David L Stevens96793b42007-09-17 09:57:33 -07001402 snmp_mib_free((void **)icmpmsg_statistics);
1403err_icmpmsg_mib:
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001404 snmp_mib_free((void **)icmp_statistics);
1405err_icmp_mib:
YOSHIFUJI Hideaki33490172007-04-20 15:57:15 -07001406 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407}
1408
1409static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410
Herbert Xu30e224d2005-07-05 14:40:10 -07001411/*
1412 * IP protocol layer initialiser
1413 */
1414
1415static struct packet_type ip_packet_type = {
1416 .type = __constant_htons(ETH_P_IP),
1417 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001418 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001419 .gso_segment = inet_gso_segment,
Herbert Xu30e224d2005-07-05 14:40:10 -07001420};
1421
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422static int __init inet_init(void)
1423{
1424 struct sk_buff *dummy_skb;
1425 struct inet_protosw *q;
1426 struct list_head *r;
1427 int rc = -EINVAL;
1428
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001429 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
1431 rc = proto_register(&tcp_prot, 1);
1432 if (rc)
1433 goto out;
1434
1435 rc = proto_register(&udp_prot, 1);
1436 if (rc)
1437 goto out_unregister_tcp_proto;
1438
1439 rc = proto_register(&raw_prot, 1);
1440 if (rc)
1441 goto out_unregister_udp_proto;
1442
1443 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001444 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 */
1446
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001447 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
1449 /*
1450 * Add all the base protocols.
1451 */
1452
1453 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1454 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1455 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1456 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1457 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1458 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1459#ifdef CONFIG_IP_MULTICAST
1460 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1461 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1462#endif
1463
1464 /* Register the socket-side information for inet_create. */
1465 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1466 INIT_LIST_HEAD(r);
1467
1468 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1469 inet_register_protosw(q);
1470
1471 /*
1472 * Set the ARP module up
1473 */
1474
1475 arp_init();
1476
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001477 /*
1478 * Set the IP module up
1479 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
1481 ip_init();
1482
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001483 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484
1485 /* Setup TCP slab cache for open requests. */
1486 tcp_init();
1487
Hideo Aoki95766ff2007-12-31 00:29:24 -08001488 /* Setup UDP memory threshold */
1489 udp_init();
1490
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001491 /* Add UDP-Lite (RFC 3828) */
1492 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
1494 /*
1495 * Set the ICMP layer up
1496 */
1497
Denis V. Luneva5710d62008-02-29 11:14:50 -08001498 if (icmp_init() < 0)
1499 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500
1501 /*
1502 * Initialise the multicast router
1503 */
1504#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001505 if (ip_mr_init())
1506 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507#endif
1508 /*
1509 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001510 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
Stephen Hemminger132adf52007-03-08 20:44:43 -08001512 if (init_ipv4_mibs())
Wang Chen03d2f892008-07-03 12:13:36 +08001513 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001514
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 ipv4_proc_init();
1516
1517 ipfrag_init();
1518
Herbert Xu30e224d2005-07-05 14:40:10 -07001519 dev_add_pack(&ip_packet_type);
1520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 rc = 0;
1522out:
1523 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524out_unregister_udp_proto:
1525 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001526out_unregister_tcp_proto:
1527 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 goto out;
1529}
1530
Heiko Carstensa536e072006-04-28 15:19:17 -07001531fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
1533/* ------------------------------------------------------------------------ */
1534
1535#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536static int __init ipv4_proc_init(void)
1537{
1538 int rc = 0;
1539
1540 if (raw_proc_init())
1541 goto out_raw;
1542 if (tcp4_proc_init())
1543 goto out_tcp;
1544 if (udp4_proc_init())
1545 goto out_udp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 if (ip_misc_proc_init())
1547 goto out_misc;
1548out:
1549 return rc;
1550out_misc:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 udp4_proc_exit();
1552out_udp:
1553 tcp4_proc_exit();
1554out_tcp:
1555 raw_proc_exit();
1556out_raw:
1557 rc = -ENOMEM;
1558 goto out;
1559}
1560
1561#else /* CONFIG_PROC_FS */
1562static int __init ipv4_proc_init(void)
1563{
1564 return 0;
1565}
1566#endif /* CONFIG_PROC_FS */
1567
1568MODULE_ALIAS_NETPROTO(PF_INET);
1569
1570EXPORT_SYMBOL(inet_accept);
1571EXPORT_SYMBOL(inet_bind);
1572EXPORT_SYMBOL(inet_dgram_connect);
1573EXPORT_SYMBOL(inet_dgram_ops);
1574EXPORT_SYMBOL(inet_getname);
1575EXPORT_SYMBOL(inet_ioctl);
1576EXPORT_SYMBOL(inet_listen);
1577EXPORT_SYMBOL(inet_register_protosw);
1578EXPORT_SYMBOL(inet_release);
1579EXPORT_SYMBOL(inet_sendmsg);
1580EXPORT_SYMBOL(inet_shutdown);
1581EXPORT_SYMBOL(inet_sock_destruct);
1582EXPORT_SYMBOL(inet_stream_connect);
1583EXPORT_SYMBOL(inet_stream_ops);
1584EXPORT_SYMBOL(inet_unregister_protosw);
Neil Hormancdac4e02005-06-13 15:12:33 -07001585EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);