blob: 58c4b0f7c4aae333e5767e331b84bebbe4845bd0 [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>
Herbert Xu73cc19f2008-12-15 23:41:09 -080097#include <net/checksum.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include <net/ip.h>
99#include <net/protocol.h>
100#include <net/arp.h>
101#include <net/route.h>
102#include <net/ip_fib.h>
Arnaldo Carvalho de Melo295f7322005-08-09 20:11:56 -0700103#include <net/inet_connection_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#include <net/tcp.h>
105#include <net/udp.h>
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800106#include <net/udplite.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <linux/skbuff.h>
108#include <net/sock.h>
109#include <net/raw.h>
110#include <net/icmp.h>
111#include <net/ipip.h>
112#include <net/inet_common.h>
113#include <net/xfrm.h>
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -0700114#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115#ifdef CONFIG_IP_MROUTE
116#include <linux/mroute.h>
117#endif
118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
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;
Pavel Emelyanov9ba63972007-12-05 01:38:23 -0800127EXPORT_SYMBOL(ipv4_config);
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129/* New destruction routine */
130
131void inet_sock_destruct(struct sock *sk)
132{
133 struct inet_sock *inet = inet_sk(sk);
134
135 __skb_queue_purge(&sk->sk_receive_queue);
136 __skb_queue_purge(&sk->sk_error_queue);
137
Hideo Aoki95766ff2007-12-31 00:29:24 -0800138 sk_mem_reclaim(sk);
139
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700141 pr_err("Attempt to release TCP socket in state %d %p\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 sk->sk_state, sk);
143 return;
144 }
145 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700146 pr_err("Attempt to release alive inet socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 return;
148 }
149
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700150 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
151 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
152 WARN_ON(sk->sk_wmem_queued);
153 WARN_ON(sk->sk_forward_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154
Jesper Juhla51482b2005-11-08 09:41:34 -0800155 kfree(inet->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 dst_release(sk->sk_dst_cache);
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700157 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700159EXPORT_SYMBOL(inet_sock_destruct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160
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}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700222EXPORT_SYMBOL(inet_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Eric Dumazetbe776282007-03-27 13:53:04 -0700224u32 inet_ehash_secret __read_mostly;
David S. Millerb3da2cf2007-03-23 11:40:27 -0700225EXPORT_SYMBOL(inet_ehash_secret);
226
Eric Dumazetbe776282007-03-27 13:53:04 -0700227/*
228 * inet_ehash_secret must be set exactly once
229 * Instead of using a dedicated spinlock, we (ab)use inetsw_lock
230 */
David S. Millerb3da2cf2007-03-23 11:40:27 -0700231void build_ehash_secret(void)
232{
Eric Dumazetbe776282007-03-27 13:53:04 -0700233 u32 rnd;
234 do {
235 get_random_bytes(&rnd, sizeof(rnd));
236 } while (rnd == 0);
237 spin_lock_bh(&inetsw_lock);
238 if (!inet_ehash_secret)
239 inet_ehash_secret = rnd;
240 spin_unlock_bh(&inetsw_lock);
David S. Millerb3da2cf2007-03-23 11:40:27 -0700241}
242EXPORT_SYMBOL(build_ehash_secret);
243
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700244static inline int inet_netns_ok(struct net *net, int protocol)
245{
246 int hash;
Alexey Dobriyan32613092009-09-14 12:21:47 +0000247 const struct net_protocol *ipprot;
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700248
Eric Dumazet04f258c2008-11-23 15:42:23 -0800249 if (net_eq(net, &init_net))
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700250 return 1;
251
252 hash = protocol & (MAX_INET_PROTOS - 1);
253 ipprot = rcu_dereference(inet_protos[hash]);
254
255 if (ipprot == NULL)
256 /* raw IP is OK */
257 return 1;
258 return ipprot->netns_ok;
259}
260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261/*
262 * Create an inet socket.
263 */
264
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700265static int inet_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266{
267 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 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
Eric Dumazet04f258c2008-11-23 15:42:23 -0800276 if (unlikely(!inet_ehash_secret))
277 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
278 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700279
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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 sk->sk_protocol = protocol;
373 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
374
375 inet->uc_ttl = -1;
376 inet->mc_loop = 1;
377 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000378 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 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}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700439EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
441/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700442int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700443EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
445int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
446{
447 struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
448 struct sock *sk = sock->sk;
449 struct inet_sock *inet = inet_sk(sk);
450 unsigned short snum;
451 int chk_addr_ret;
452 int err;
453
454 /* If the socket has its own bind function then use it. (RAW) */
455 if (sk->sk_prot->bind) {
456 err = sk->sk_prot->bind(sk, uaddr, addr_len);
457 goto out;
458 }
459 err = -EINVAL;
460 if (addr_len < sizeof(struct sockaddr_in))
461 goto out;
462
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900463 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465 /* Not specified by any standard per-se, however it breaks too
466 * many applications when removed. It is unfortunate since
467 * allowing applications to make a non-local bind solves
468 * several problems with systems using dynamic addressing.
469 * (ie. your servers still start up even if your ISDN link
470 * is temporarily down)
471 */
472 err = -EADDRNOTAVAIL;
473 if (!sysctl_ip_nonlocal_bind &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700474 !(inet->freebind || inet->transparent) &&
Al Viroe6f1ceb2008-03-17 22:44:53 -0700475 addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 chk_addr_ret != RTN_LOCAL &&
477 chk_addr_ret != RTN_MULTICAST &&
478 chk_addr_ret != RTN_BROADCAST)
479 goto out;
480
481 snum = ntohs(addr->sin_port);
482 err = -EACCES;
483 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
484 goto out;
485
486 /* We keep a pair of addresses. rcv_saddr is the one
487 * used by hash lookups, and saddr is used for transmit.
488 *
489 * In the BSD API these are the same except where it
490 * would be illegal to use them (multicast/broadcast) in
491 * which case the sending device address is used.
492 */
493 lock_sock(sk);
494
495 /* Check these errors (active socket, double bind). */
496 err = -EINVAL;
497 if (sk->sk_state != TCP_CLOSE || inet->num)
498 goto out_release_sock;
499
500 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
501 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
502 inet->saddr = 0; /* Use device */
503
504 /* Make sure we are allowed to bind here. */
505 if (sk->sk_prot->get_port(sk, snum)) {
506 inet->saddr = inet->rcv_saddr = 0;
507 err = -EADDRINUSE;
508 goto out_release_sock;
509 }
510
511 if (inet->rcv_saddr)
512 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
513 if (snum)
514 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
515 inet->sport = htons(inet->num);
516 inet->daddr = 0;
517 inet->dport = 0;
518 sk_dst_reset(sk);
519 err = 0;
520out_release_sock:
521 release_sock(sk);
522out:
523 return err;
524}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700525EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
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}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700539EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
541static long inet_wait_for_connect(struct sock *sk, long timeo)
542{
543 DEFINE_WAIT(wait);
544
545 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
546
547 /* Basic assumption: if someone sets sk->sk_err, he _must_
548 * change state of the socket from TCP_SYN_*.
549 * Connect() does not allow to get error notifications
550 * without closing the socket.
551 */
552 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
553 release_sock(sk);
554 timeo = schedule_timeout(timeo);
555 lock_sock(sk);
556 if (signal_pending(current) || !timeo)
557 break;
558 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
559 }
560 finish_wait(sk->sk_sleep, &wait);
561 return timeo;
562}
563
564/*
565 * Connect to a remote host. There is regrettably still a little
566 * TCP 'magic' in here.
567 */
568int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
569 int addr_len, int flags)
570{
571 struct sock *sk = sock->sk;
572 int err;
573 long timeo;
574
575 lock_sock(sk);
576
577 if (uaddr->sa_family == AF_UNSPEC) {
578 err = sk->sk_prot->disconnect(sk, flags);
579 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
580 goto out;
581 }
582
583 switch (sock->state) {
584 default:
585 err = -EINVAL;
586 goto out;
587 case SS_CONNECTED:
588 err = -EISCONN;
589 goto out;
590 case SS_CONNECTING:
591 err = -EALREADY;
592 /* Fall out of switch with err, set for this state */
593 break;
594 case SS_UNCONNECTED:
595 err = -EISCONN;
596 if (sk->sk_state != TCP_CLOSE)
597 goto out;
598
599 err = sk->sk_prot->connect(sk, uaddr, addr_len);
600 if (err < 0)
601 goto out;
602
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900603 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
605 /* Just entered SS_CONNECTING state; the only
606 * difference is that return value in non-blocking
607 * case is EINPROGRESS, rather than EALREADY.
608 */
609 err = -EINPROGRESS;
610 break;
611 }
612
613 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
614
615 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
616 /* Error code is set above */
617 if (!timeo || !inet_wait_for_connect(sk, timeo))
618 goto out;
619
620 err = sock_intr_errno(timeo);
621 if (signal_pending(current))
622 goto out;
623 }
624
625 /* Connection was closed by RST, timeout, ICMP error
626 * or another process disconnected us.
627 */
628 if (sk->sk_state == TCP_CLOSE)
629 goto sock_error;
630
631 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
632 * and error was received after socket entered established state.
633 * Hence, it is handled normally after connect() return successfully.
634 */
635
636 sock->state = SS_CONNECTED;
637 err = 0;
638out:
639 release_sock(sk);
640 return err;
641
642sock_error:
643 err = sock_error(sk) ? : -ECONNABORTED;
644 sock->state = SS_UNCONNECTED;
645 if (sk->sk_prot->disconnect(sk, flags))
646 sock->state = SS_DISCONNECTING;
647 goto out;
648}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700649EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651/*
652 * Accept a pending connection. The TCP layer now gives BSD semantics.
653 */
654
655int inet_accept(struct socket *sock, struct socket *newsock, int flags)
656{
657 struct sock *sk1 = sock->sk;
658 int err = -EINVAL;
659 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
660
661 if (!sk2)
662 goto do_err;
663
664 lock_sock(sk2);
665
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700666 WARN_ON(!((1 << sk2->sk_state) &
667 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
669 sock_graft(sk2, newsock);
670
671 newsock->state = SS_CONNECTED;
672 err = 0;
673 release_sock(sk2);
674do_err:
675 return err;
676}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700677EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
679
680/*
681 * This does both peername and sockname.
682 */
683int inet_getname(struct socket *sock, struct sockaddr *uaddr,
684 int *uaddr_len, int peer)
685{
686 struct sock *sk = sock->sk;
687 struct inet_sock *inet = inet_sk(sk);
688 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
689
690 sin->sin_family = AF_INET;
691 if (peer) {
692 if (!inet->dport ||
693 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
694 peer == 1))
695 return -ENOTCONN;
696 sin->sin_port = inet->dport;
697 sin->sin_addr.s_addr = inet->daddr;
698 } else {
Al Viro714e85b2006-11-14 20:51:49 -0800699 __be32 addr = inet->rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 if (!addr)
701 addr = inet->saddr;
702 sin->sin_port = inet->sport;
703 sin->sin_addr.s_addr = addr;
704 }
705 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
706 *uaddr_len = sizeof(*sin);
707 return 0;
708}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700709EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710
711int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
712 size_t size)
713{
714 struct sock *sk = sock->sk;
715
716 /* We may need to bind the socket. */
717 if (!inet_sk(sk)->num && inet_autobind(sk))
718 return -EAGAIN;
719
720 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
721}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700722EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
724
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700725static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
726 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
728 struct sock *sk = sock->sk;
729
730 /* We may need to bind the socket. */
731 if (!inet_sk(sk)->num && inet_autobind(sk))
732 return -EAGAIN;
733
734 if (sk->sk_prot->sendpage)
735 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
736 return sock_no_sendpage(sock, page, offset, size, flags);
737}
738
739
740int inet_shutdown(struct socket *sock, int how)
741{
742 struct sock *sk = sock->sk;
743 int err = 0;
744
745 /* This should really check to make sure
746 * the socket is a TCP socket. (WHY AC...)
747 */
748 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
749 1->2 bit 2 snds.
750 2->3 */
751 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
752 return -EINVAL;
753
754 lock_sock(sk);
755 if (sock->state == SS_CONNECTING) {
756 if ((1 << sk->sk_state) &
757 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
758 sock->state = SS_DISCONNECTING;
759 else
760 sock->state = SS_CONNECTED;
761 }
762
763 switch (sk->sk_state) {
764 case TCP_CLOSE:
765 err = -ENOTCONN;
766 /* Hack to wake up other listeners, who can poll for
767 POLLHUP, even on eg. unconnected UDP sockets -- RR */
768 default:
769 sk->sk_shutdown |= how;
770 if (sk->sk_prot->shutdown)
771 sk->sk_prot->shutdown(sk, how);
772 break;
773
774 /* Remaining two branches are temporary solution for missing
775 * close() in multithreaded environment. It is _not_ a good idea,
776 * but we have no choice until close() is repaired at VFS level.
777 */
778 case TCP_LISTEN:
779 if (!(how & RCV_SHUTDOWN))
780 break;
781 /* Fall through */
782 case TCP_SYN_SENT:
783 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
784 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
785 break;
786 }
787
788 /* Wake up anyone sleeping in poll. */
789 sk->sk_state_change(sk);
790 release_sock(sk);
791 return err;
792}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700793EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795/*
796 * ioctl() calls you can issue on an INET socket. Most of these are
797 * device configuration and stuff and very rarely used. Some ioctls
798 * pass on to the socket itself.
799 *
800 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
801 * loads the devconfigure module does its configuring and unloads it.
802 * There's a good 20K of config code hanging around the kernel.
803 */
804
805int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
806{
807 struct sock *sk = sock->sk;
808 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900809 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700812 case SIOCGSTAMP:
813 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
814 break;
815 case SIOCGSTAMPNS:
816 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
817 break;
818 case SIOCADDRT:
819 case SIOCDELRT:
820 case SIOCRTMSG:
821 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
822 break;
823 case SIOCDARP:
824 case SIOCGARP:
825 case SIOCSARP:
826 err = arp_ioctl(net, cmd, (void __user *)arg);
827 break;
828 case SIOCGIFADDR:
829 case SIOCSIFADDR:
830 case SIOCGIFBRDADDR:
831 case SIOCSIFBRDADDR:
832 case SIOCGIFNETMASK:
833 case SIOCSIFNETMASK:
834 case SIOCGIFDSTADDR:
835 case SIOCSIFDSTADDR:
836 case SIOCSIFPFLAGS:
837 case SIOCGIFPFLAGS:
838 case SIOCSIFFLAGS:
839 err = devinet_ioctl(net, cmd, (void __user *)arg);
840 break;
841 default:
842 if (sk->sk_prot->ioctl)
843 err = sk->sk_prot->ioctl(sk, cmd, arg);
844 else
845 err = -ENOIOCTLCMD;
846 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 }
848 return err;
849}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700850EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800852const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800853 .family = PF_INET,
854 .owner = THIS_MODULE,
855 .release = inet_release,
856 .bind = inet_bind,
857 .connect = inet_stream_connect,
858 .socketpair = sock_no_socketpair,
859 .accept = inet_accept,
860 .getname = inet_getname,
861 .poll = tcp_poll,
862 .ioctl = inet_ioctl,
863 .listen = inet_listen,
864 .shutdown = inet_shutdown,
865 .setsockopt = sock_common_setsockopt,
866 .getsockopt = sock_common_getsockopt,
David S. Miller3516ffb2007-08-02 19:23:56 -0700867 .sendmsg = tcp_sendmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800868 .recvmsg = sock_common_recvmsg,
869 .mmap = sock_no_mmap,
870 .sendpage = tcp_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800871 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800872#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800873 .compat_setsockopt = compat_sock_common_setsockopt,
874 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800875#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700877EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800879const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800880 .family = PF_INET,
881 .owner = THIS_MODULE,
882 .release = inet_release,
883 .bind = inet_bind,
884 .connect = inet_dgram_connect,
885 .socketpair = sock_no_socketpair,
886 .accept = sock_no_accept,
887 .getname = inet_getname,
888 .poll = udp_poll,
889 .ioctl = inet_ioctl,
890 .listen = sock_no_listen,
891 .shutdown = inet_shutdown,
892 .setsockopt = sock_common_setsockopt,
893 .getsockopt = sock_common_getsockopt,
894 .sendmsg = inet_sendmsg,
895 .recvmsg = sock_common_recvmsg,
896 .mmap = sock_no_mmap,
897 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800898#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800899 .compat_setsockopt = compat_sock_common_setsockopt,
900 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800901#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700903EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
905/*
906 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
907 * udp_poll
908 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800909static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800910 .family = PF_INET,
911 .owner = THIS_MODULE,
912 .release = inet_release,
913 .bind = inet_bind,
914 .connect = inet_dgram_connect,
915 .socketpair = sock_no_socketpair,
916 .accept = sock_no_accept,
917 .getname = inet_getname,
918 .poll = datagram_poll,
919 .ioctl = inet_ioctl,
920 .listen = sock_no_listen,
921 .shutdown = inet_shutdown,
922 .setsockopt = sock_common_setsockopt,
923 .getsockopt = sock_common_getsockopt,
924 .sendmsg = inet_sendmsg,
925 .recvmsg = sock_common_recvmsg,
926 .mmap = sock_no_mmap,
927 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800928#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800929 .compat_setsockopt = compat_sock_common_setsockopt,
930 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800931#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932};
933
934static struct net_proto_family inet_family_ops = {
935 .family = PF_INET,
936 .create = inet_create,
937 .owner = THIS_MODULE,
938};
939
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940/* Upon startup we insert all the elements in inetsw_array[] into
941 * the linked list inetsw.
942 */
943static struct inet_protosw inetsw_array[] =
944{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900945 {
946 .type = SOCK_STREAM,
947 .protocol = IPPROTO_TCP,
948 .prot = &tcp_prot,
949 .ops = &inet_stream_ops,
950 .capability = -1,
951 .no_check = 0,
952 .flags = INET_PROTOSW_PERMANENT |
Arnaldo Carvalho de Melod83d8462005-12-13 23:26:10 -0800953 INET_PROTOSW_ICSK,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900954 },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900956 {
957 .type = SOCK_DGRAM,
958 .protocol = IPPROTO_UDP,
959 .prot = &udp_prot,
960 .ops = &inet_dgram_ops,
961 .capability = -1,
962 .no_check = UDP_CSUM_DEFAULT,
963 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900965
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
967 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900968 .type = SOCK_RAW,
969 .protocol = IPPROTO_IP, /* wild card */
970 .prot = &raw_prot,
971 .ops = &inet_sockraw_ops,
972 .capability = CAP_NET_RAW,
973 .no_check = UDP_CSUM_DEFAULT,
974 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 }
976};
977
Denis Chengc40f6ff2007-09-16 16:39:25 -0700978#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
980void inet_register_protosw(struct inet_protosw *p)
981{
982 struct list_head *lh;
983 struct inet_protosw *answer;
984 int protocol = p->protocol;
985 struct list_head *last_perm;
986
987 spin_lock_bh(&inetsw_lock);
988
989 if (p->type >= SOCK_MAX)
990 goto out_illegal;
991
992 /* If we are trying to override a permanent protocol, bail. */
993 answer = NULL;
994 last_perm = &inetsw[p->type];
995 list_for_each(lh, &inetsw[p->type]) {
996 answer = list_entry(lh, struct inet_protosw, list);
997
998 /* Check only the non-wild match. */
999 if (INET_PROTOSW_PERMANENT & answer->flags) {
1000 if (protocol == answer->protocol)
1001 break;
1002 last_perm = lh;
1003 }
1004
1005 answer = NULL;
1006 }
1007 if (answer)
1008 goto out_permanent;
1009
1010 /* Add the new entry after the last permanent entry if any, so that
1011 * the new entry does not override a permanent entry when matched with
1012 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001013 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 * system automatically returns to the old behavior.
1015 */
1016 list_add_rcu(&p->list, last_perm);
1017out:
1018 spin_unlock_bh(&inetsw_lock);
1019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 return;
1021
1022out_permanent:
1023 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1024 protocol);
1025 goto out;
1026
1027out_illegal:
1028 printk(KERN_ERR
1029 "Ignoring attempt to register invalid socket type %d.\n",
1030 p->type);
1031 goto out;
1032}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001033EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034
1035void inet_unregister_protosw(struct inet_protosw *p)
1036{
1037 if (INET_PROTOSW_PERMANENT & p->flags) {
1038 printk(KERN_ERR
1039 "Attempt to unregister permanent protocol %d.\n",
1040 p->protocol);
1041 } else {
1042 spin_lock_bh(&inetsw_lock);
1043 list_del_rcu(&p->list);
1044 spin_unlock_bh(&inetsw_lock);
1045
1046 synchronize_net();
1047 }
1048}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001049EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001051/*
1052 * Shall we try to damage output packets if routing dev changes?
1053 */
1054
Brian Haleyab32ea52006-09-22 14:15:41 -07001055int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001056
1057static int inet_sk_reselect_saddr(struct sock *sk)
1058{
1059 struct inet_sock *inet = inet_sk(sk);
1060 int err;
1061 struct rtable *rt;
Al Viro714e85b2006-11-14 20:51:49 -08001062 __be32 old_saddr = inet->saddr;
1063 __be32 new_saddr;
Al Virobada8ad2006-09-26 21:27:15 -07001064 __be32 daddr = inet->daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001065
1066 if (inet->opt && inet->opt->srr)
1067 daddr = inet->opt->faddr;
1068
1069 /* Query new route. */
1070 err = ip_route_connect(&rt, daddr, 0,
1071 RT_CONN_FLAGS(sk),
1072 sk->sk_bound_dev_if,
1073 sk->sk_protocol,
David S. Miller8eb90862007-02-08 02:09:21 -08001074 inet->sport, inet->dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001075 if (err)
1076 return err;
1077
1078 sk_setup_caps(sk, &rt->u.dst);
1079
1080 new_saddr = rt->rt_src;
1081
1082 if (new_saddr == old_saddr)
1083 return 0;
1084
1085 if (sysctl_ip_dynaddr > 1) {
Harvey Harrison673d57e2008-10-31 00:53:57 -07001086 printk(KERN_INFO "%s(): shifting inet->saddr from %pI4 to %pI4\n",
1087 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001088 }
1089
1090 inet->saddr = inet->rcv_saddr = new_saddr;
1091
1092 /*
1093 * XXX The only one ugly spot where we need to
1094 * XXX really change the sockets identity after
1095 * XXX it has entered the hashes. -DaveM
1096 *
1097 * Besides that, it does not check for connection
1098 * uniqueness. Wait for troubles.
1099 */
1100 __sk_prot_rehash(sk);
1101 return 0;
1102}
1103
1104int inet_sk_rebuild_header(struct sock *sk)
1105{
1106 struct inet_sock *inet = inet_sk(sk);
1107 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001108 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001109 int err;
1110
1111 /* Route is OK, nothing to do. */
1112 if (rt)
1113 return 0;
1114
1115 /* Reroute. */
1116 daddr = inet->daddr;
1117 if (inet->opt && inet->opt->srr)
1118 daddr = inet->opt->faddr;
1119{
1120 struct flowi fl = {
1121 .oif = sk->sk_bound_dev_if,
1122 .nl_u = {
1123 .ip4_u = {
1124 .daddr = daddr,
1125 .saddr = inet->saddr,
1126 .tos = RT_CONN_FLAGS(sk),
1127 },
1128 },
1129 .proto = sk->sk_protocol,
Balazs Scheidlerc8283842008-11-20 01:08:06 -08001130 .flags = inet_sk_flowi_flags(sk),
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001131 .uli_u = {
1132 .ports = {
1133 .sport = inet->sport,
1134 .dport = inet->dport,
1135 },
1136 },
1137 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001138
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001139 security_sk_classify_flow(sk, &fl);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001140 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001141}
1142 if (!err)
1143 sk_setup_caps(sk, &rt->u.dst);
1144 else {
1145 /* Routing failed... */
1146 sk->sk_route_caps = 0;
1147 /*
1148 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1149 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1150 */
1151 if (!sysctl_ip_dynaddr ||
1152 sk->sk_state != TCP_SYN_SENT ||
1153 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1154 (err = inet_sk_reselect_saddr(sk)) != 0)
1155 sk->sk_err_soft = -err;
1156 }
1157
1158 return err;
1159}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001160EXPORT_SYMBOL(inet_sk_rebuild_header);
1161
Herbert Xua430a432006-07-08 13:34:56 -07001162static int inet_gso_send_check(struct sk_buff *skb)
1163{
1164 struct iphdr *iph;
Alexey Dobriyan32613092009-09-14 12:21:47 +00001165 const struct net_protocol *ops;
Herbert Xua430a432006-07-08 13:34:56 -07001166 int proto;
1167 int ihl;
1168 int err = -EINVAL;
1169
1170 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1171 goto out;
1172
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001173 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001174 ihl = iph->ihl * 4;
1175 if (ihl < sizeof(*iph))
1176 goto out;
1177
1178 if (unlikely(!pskb_may_pull(skb, ihl)))
1179 goto out;
1180
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001181 __skb_pull(skb, ihl);
1182 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001183 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001184 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1185 err = -EPROTONOSUPPORT;
1186
1187 rcu_read_lock();
1188 ops = rcu_dereference(inet_protos[proto]);
1189 if (likely(ops && ops->gso_send_check))
1190 err = ops->gso_send_check(skb);
1191 rcu_read_unlock();
1192
1193out:
1194 return err;
1195}
1196
Herbert Xu576a30e2006-06-27 13:22:38 -07001197static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001198{
1199 struct sk_buff *segs = ERR_PTR(-EINVAL);
1200 struct iphdr *iph;
Alexey Dobriyan32613092009-09-14 12:21:47 +00001201 const struct net_protocol *ops;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001202 int proto;
1203 int ihl;
1204 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001205 unsigned int offset = 0;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001206
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001207 if (!(features & NETIF_F_V4_CSUM))
1208 features &= ~NETIF_F_SG;
1209
Herbert Xubbcf4672006-07-03 19:38:35 -07001210 if (unlikely(skb_shinfo(skb)->gso_type &
1211 ~(SKB_GSO_TCPV4 |
1212 SKB_GSO_UDP |
1213 SKB_GSO_DODGY |
1214 SKB_GSO_TCP_ECN |
1215 0)))
1216 goto out;
1217
1218 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001219 goto out;
1220
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001221 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001222 ihl = iph->ihl * 4;
1223 if (ihl < sizeof(*iph))
1224 goto out;
1225
Herbert Xubbcf4672006-07-03 19:38:35 -07001226 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001227 goto out;
1228
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001229 __skb_pull(skb, ihl);
1230 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001231 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001232 id = ntohs(iph->id);
1233 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1234 segs = ERR_PTR(-EPROTONOSUPPORT);
1235
1236 rcu_read_lock();
1237 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001238 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001239 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001240 rcu_read_unlock();
1241
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001242 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001243 goto out;
1244
1245 skb = segs;
1246 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001247 iph = ip_hdr(skb);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001248 if (proto == IPPROTO_UDP) {
1249 iph->id = htons(id);
1250 iph->frag_off = htons(offset >> 3);
1251 if (skb->next != NULL)
1252 iph->frag_off |= htons(IP_MF);
1253 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1254 } else
1255 iph->id = htons(id++);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001256 iph->tot_len = htons(skb->len - skb->mac_len);
1257 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001258 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001259 } while ((skb = skb->next));
1260
1261out:
1262 return segs;
1263}
1264
Herbert Xu73cc19f2008-12-15 23:41:09 -08001265static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1266 struct sk_buff *skb)
1267{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001268 const struct net_protocol *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001269 struct sk_buff **pp = NULL;
1270 struct sk_buff *p;
1271 struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001272 unsigned int hlen;
1273 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001274 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001275 int flush = 1;
1276 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001277
Herbert Xua5b1cf22009-05-26 18:50:28 +00001278 off = skb_gro_offset(skb);
1279 hlen = off + sizeof(*iph);
1280 iph = skb_gro_header_fast(skb, off);
1281 if (skb_gro_header_hard(skb, hlen)) {
1282 iph = skb_gro_header_slow(skb, hlen, off);
1283 if (unlikely(!iph))
1284 goto out;
1285 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001286
Herbert Xu73cc19f2008-12-15 23:41:09 -08001287 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1288
1289 rcu_read_lock();
1290 ops = rcu_dereference(inet_protos[proto]);
1291 if (!ops || !ops->gro_receive)
1292 goto out_unlock;
1293
Herbert Xua5ad24b2009-02-08 18:00:39 +00001294 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001295 goto out_unlock;
1296
1297 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1298 goto out_unlock;
1299
Herbert Xu1075f3f2009-05-26 18:50:29 +00001300 id = ntohl(*(u32 *)&iph->id);
1301 flush = (u16)((ntohl(*(u32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
1302 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001303
1304 for (p = *head; p; p = p->next) {
1305 struct iphdr *iph2;
1306
1307 if (!NAPI_GRO_CB(p)->same_flow)
1308 continue;
1309
1310 iph2 = ip_hdr(p);
1311
Herbert Xua5ad24b2009-02-08 18:00:39 +00001312 if ((iph->protocol ^ iph2->protocol) |
1313 (iph->tos ^ iph2->tos) |
1314 (iph->saddr ^ iph2->saddr) |
1315 (iph->daddr ^ iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001316 NAPI_GRO_CB(p)->same_flow = 0;
1317 continue;
1318 }
1319
1320 /* All fields must match except length and checksum. */
1321 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001322 (iph->ttl ^ iph2->ttl) |
1323 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001324
1325 NAPI_GRO_CB(p)->flush |= flush;
1326 }
1327
1328 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001329 skb_gro_pull(skb, sizeof(*iph));
1330 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001331
1332 pp = ops->gro_receive(head, skb);
1333
1334out_unlock:
1335 rcu_read_unlock();
1336
1337out:
1338 NAPI_GRO_CB(skb)->flush |= flush;
1339
1340 return pp;
1341}
1342
1343static int inet_gro_complete(struct sk_buff *skb)
1344{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001345 const struct net_protocol *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001346 struct iphdr *iph = ip_hdr(skb);
1347 int proto = iph->protocol & (MAX_INET_PROTOS - 1);
1348 int err = -ENOSYS;
1349 __be16 newlen = htons(skb->len - skb_network_offset(skb));
1350
1351 csum_replace2(&iph->check, iph->tot_len, newlen);
1352 iph->tot_len = newlen;
1353
1354 rcu_read_lock();
1355 ops = rcu_dereference(inet_protos[proto]);
1356 if (WARN_ON(!ops || !ops->gro_complete))
1357 goto out_unlock;
1358
1359 err = ops->gro_complete(skb);
1360
1361out_unlock:
1362 rcu_read_unlock();
1363
1364 return err;
1365}
1366
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001367int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001368 unsigned short type, unsigned char protocol,
1369 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001370{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001371 struct socket *sock;
1372 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001373
1374 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001375 *sk = sock->sk;
1376 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001377 /*
1378 * Unhash it so that IP input processing does not even see it,
1379 * we do not wish this socket to see incoming packets.
1380 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001381 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001382
1383 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001384 }
1385 return rc;
1386}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001387EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1388
Herbert Xu5e0f0432007-04-24 21:53:35 -07001389unsigned long snmp_fold_field(void *mib[], int offt)
1390{
1391 unsigned long res = 0;
1392 int i;
1393
1394 for_each_possible_cpu(i) {
1395 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1396 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1397 }
1398 return res;
1399}
1400EXPORT_SYMBOL_GPL(snmp_fold_field);
1401
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001402int snmp_mib_init(void *ptr[2], size_t mibsize)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001403{
1404 BUG_ON(ptr == NULL);
Rusty Russell313e4582009-02-20 16:29:08 +09001405 ptr[0] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
Herbert Xu5e0f0432007-04-24 21:53:35 -07001406 if (!ptr[0])
1407 goto err0;
Rusty Russell313e4582009-02-20 16:29:08 +09001408 ptr[1] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
Herbert Xu5e0f0432007-04-24 21:53:35 -07001409 if (!ptr[1])
1410 goto err1;
1411 return 0;
1412err1:
1413 free_percpu(ptr[0]);
1414 ptr[0] = NULL;
1415err0:
1416 return -ENOMEM;
1417}
1418EXPORT_SYMBOL_GPL(snmp_mib_init);
1419
1420void snmp_mib_free(void *ptr[2])
1421{
1422 BUG_ON(ptr == NULL);
1423 free_percpu(ptr[0]);
1424 free_percpu(ptr[1]);
1425 ptr[0] = ptr[1] = NULL;
1426}
1427EXPORT_SYMBOL_GPL(snmp_mib_free);
1428
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001430static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001432 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433};
1434#endif
1435
Alexey Dobriyan32613092009-09-14 12:21:47 +00001436static const struct net_protocol tcp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 .handler = tcp_v4_rcv,
1438 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001439 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001440 .gso_segment = tcp_tso_segment,
Herbert Xubf296b12008-12-15 23:43:36 -08001441 .gro_receive = tcp4_gro_receive,
1442 .gro_complete = tcp4_gro_complete,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001444 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445};
1446
Alexey Dobriyan32613092009-09-14 12:21:47 +00001447static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 .handler = udp_rcv,
1449 .err_handler = udp_err,
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001450 .gso_send_check = udp4_ufo_send_check,
1451 .gso_segment = udp4_ufo_fragment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001453 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454};
1455
Alexey Dobriyan32613092009-09-14 12:21:47 +00001456static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 .handler = icmp_rcv,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001458 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001459 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460};
1461
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001462static __net_init int ipv4_mib_init_net(struct net *net)
1463{
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001464 if (snmp_mib_init((void **)net->mib.tcp_statistics,
1465 sizeof(struct tcp_mib)) < 0)
1466 goto err_tcp_mib;
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001467 if (snmp_mib_init((void **)net->mib.ip_statistics,
1468 sizeof(struct ipstats_mib)) < 0)
1469 goto err_ip_mib;
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001470 if (snmp_mib_init((void **)net->mib.net_statistics,
1471 sizeof(struct linux_mib)) < 0)
1472 goto err_net_mib;
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001473 if (snmp_mib_init((void **)net->mib.udp_statistics,
1474 sizeof(struct udp_mib)) < 0)
1475 goto err_udp_mib;
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001476 if (snmp_mib_init((void **)net->mib.udplite_statistics,
1477 sizeof(struct udp_mib)) < 0)
1478 goto err_udplite_mib;
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001479 if (snmp_mib_init((void **)net->mib.icmp_statistics,
1480 sizeof(struct icmp_mib)) < 0)
1481 goto err_icmp_mib;
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001482 if (snmp_mib_init((void **)net->mib.icmpmsg_statistics,
1483 sizeof(struct icmpmsg_mib)) < 0)
1484 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001485
1486 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001487 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001488
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001489err_icmpmsg_mib:
1490 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001491err_icmp_mib:
1492 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001493err_udplite_mib:
1494 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001495err_udp_mib:
1496 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001497err_net_mib:
1498 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001499err_ip_mib:
1500 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001501err_tcp_mib:
1502 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001503}
1504
1505static __net_exit void ipv4_mib_exit_net(struct net *net)
1506{
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001507 snmp_mib_free((void **)net->mib.icmpmsg_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001508 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001509 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001510 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001511 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001512 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001513 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001514}
1515
1516static __net_initdata struct pernet_operations ipv4_mib_ops = {
1517 .init = ipv4_mib_init_net,
1518 .exit = ipv4_mib_exit_net,
1519};
1520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521static int __init init_ipv4_mibs(void)
1522{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001523 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524}
1525
1526static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Herbert Xu30e224d2005-07-05 14:40:10 -07001528/*
1529 * IP protocol layer initialiser
1530 */
1531
Stephen Hemminger7546dd92009-03-09 08:18:29 +00001532static struct packet_type ip_packet_type __read_mostly = {
Harvey Harrison09640e62009-02-01 00:45:17 -08001533 .type = cpu_to_be16(ETH_P_IP),
Herbert Xu30e224d2005-07-05 14:40:10 -07001534 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001535 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001536 .gso_segment = inet_gso_segment,
Herbert Xu73cc19f2008-12-15 23:41:09 -08001537 .gro_receive = inet_gro_receive,
1538 .gro_complete = inet_gro_complete,
Herbert Xu30e224d2005-07-05 14:40:10 -07001539};
1540
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541static int __init inet_init(void)
1542{
1543 struct sk_buff *dummy_skb;
1544 struct inet_protosw *q;
1545 struct list_head *r;
1546 int rc = -EINVAL;
1547
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001548 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549
1550 rc = proto_register(&tcp_prot, 1);
1551 if (rc)
1552 goto out;
1553
1554 rc = proto_register(&udp_prot, 1);
1555 if (rc)
1556 goto out_unregister_tcp_proto;
1557
1558 rc = proto_register(&raw_prot, 1);
1559 if (rc)
1560 goto out_unregister_udp_proto;
1561
1562 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001563 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 */
1565
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001566 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
Al Virobd7b1532008-07-15 16:00:59 -04001568#ifdef CONFIG_SYSCTL
1569 ip_static_sysctl_init();
1570#endif
1571
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 /*
1573 * Add all the base protocols.
1574 */
1575
1576 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1577 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1578 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1579 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1580 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1581 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1582#ifdef CONFIG_IP_MULTICAST
1583 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1584 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1585#endif
1586
1587 /* Register the socket-side information for inet_create. */
1588 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1589 INIT_LIST_HEAD(r);
1590
1591 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1592 inet_register_protosw(q);
1593
1594 /*
1595 * Set the ARP module up
1596 */
1597
1598 arp_init();
1599
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001600 /*
1601 * Set the IP module up
1602 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603
1604 ip_init();
1605
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001606 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607
1608 /* Setup TCP slab cache for open requests. */
1609 tcp_init();
1610
Hideo Aoki95766ff2007-12-31 00:29:24 -08001611 /* Setup UDP memory threshold */
1612 udp_init();
1613
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001614 /* Add UDP-Lite (RFC 3828) */
1615 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
1617 /*
1618 * Set the ICMP layer up
1619 */
1620
Denis V. Luneva5710d62008-02-29 11:14:50 -08001621 if (icmp_init() < 0)
1622 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623
1624 /*
1625 * Initialise the multicast router
1626 */
1627#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001628 if (ip_mr_init())
1629 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630#endif
1631 /*
1632 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001633 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
Stephen Hemminger132adf52007-03-08 20:44:43 -08001635 if (init_ipv4_mibs())
Wang Chen03d2f892008-07-03 12:13:36 +08001636 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001637
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 ipv4_proc_init();
1639
1640 ipfrag_init();
1641
Herbert Xu30e224d2005-07-05 14:40:10 -07001642 dev_add_pack(&ip_packet_type);
1643
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 rc = 0;
1645out:
1646 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647out_unregister_udp_proto:
1648 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001649out_unregister_tcp_proto:
1650 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 goto out;
1652}
1653
Heiko Carstensa536e072006-04-28 15:19:17 -07001654fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655
1656/* ------------------------------------------------------------------------ */
1657
1658#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659static int __init ipv4_proc_init(void)
1660{
1661 int rc = 0;
1662
1663 if (raw_proc_init())
1664 goto out_raw;
1665 if (tcp4_proc_init())
1666 goto out_tcp;
1667 if (udp4_proc_init())
1668 goto out_udp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 if (ip_misc_proc_init())
1670 goto out_misc;
1671out:
1672 return rc;
1673out_misc:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 udp4_proc_exit();
1675out_udp:
1676 tcp4_proc_exit();
1677out_tcp:
1678 raw_proc_exit();
1679out_raw:
1680 rc = -ENOMEM;
1681 goto out;
1682}
1683
1684#else /* CONFIG_PROC_FS */
1685static int __init ipv4_proc_init(void)
1686{
1687 return 0;
1688}
1689#endif /* CONFIG_PROC_FS */
1690
1691MODULE_ALIAS_NETPROTO(PF_INET);
1692