blob: 7d12c6a9b19b14d4534e01821e71741e6c5099f0 [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);
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000177 if (!inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 if (sk->sk_prot->get_port(sk, 0)) {
179 release_sock(sk);
180 return -EAGAIN;
181 }
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000182 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 }
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 Paris3f378b62009-11-05 22:18:14 -0800265static int inet_create(struct net *net, struct socket *sock, int protocol,
266 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267{
268 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 struct inet_protosw *answer;
270 struct inet_sock *inet;
271 struct proto *answer_prot;
272 unsigned char answer_flags;
273 char answer_no_check;
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700274 int try_loading_module = 0;
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800275 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
Eric Dumazet04f258c2008-11-23 15:42:23 -0800277 if (unlikely(!inet_ehash_secret))
278 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
279 build_ehash_secret();
David S. Millerb3da2cf2007-03-23 11:40:27 -0700280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 sock->state = SS_UNCONNECTED;
282
283 /* Look for the requested type/protocol pair. */
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700284lookup_protocol:
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800285 err = -ESOCKTNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 rcu_read_lock();
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700287 list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700289 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 /* Check the non-wild match. */
291 if (protocol == answer->protocol) {
292 if (protocol != IPPROTO_IP)
293 break;
294 } else {
295 /* Check for the two wild cases. */
296 if (IPPROTO_IP == protocol) {
297 protocol = answer->protocol;
298 break;
299 }
300 if (IPPROTO_IP == answer->protocol)
301 break;
302 }
Herbert Xu86c8f9d2005-12-02 20:43:26 -0800303 err = -EPROTONOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 }
305
Paul E. McKenney696adfe2008-07-25 01:45:34 -0700306 if (unlikely(err)) {
Arnaldo Carvalho de Melobb97d312005-08-09 20:19:14 -0700307 if (try_loading_module < 2) {
308 rcu_read_unlock();
309 /*
310 * Be more specific, e.g. net-pf-2-proto-132-type-1
311 * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
312 */
313 if (++try_loading_module == 1)
314 request_module("net-pf-%d-proto-%d-type-%d",
315 PF_INET, protocol, sock->type);
316 /*
317 * Fall back to generic, e.g. net-pf-2-proto-132
318 * (net-pf-PF_INET-proto-IPPROTO_SCTP)
319 */
320 else
321 request_module("net-pf-%d-proto-%d",
322 PF_INET, protocol);
323 goto lookup_protocol;
324 } else
325 goto out_rcu_unlock;
326 }
327
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 err = -EPERM;
Eric Parisc84b3262009-11-05 20:45:52 -0800329 if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 goto out_rcu_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Denis V. Lunev2342fd72008-03-24 15:33:42 -0700332 err = -EAFNOSUPPORT;
333 if (!inet_netns_ok(net, protocol))
334 goto out_rcu_unlock;
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 sock->ops = answer->ops;
337 answer_prot = answer->prot;
338 answer_no_check = answer->no_check;
339 answer_flags = answer->flags;
340 rcu_read_unlock();
341
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700342 WARN_ON(answer_prot->slab == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
344 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700345 sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 if (sk == NULL)
347 goto out;
348
349 err = 0;
350 sk->sk_no_check = answer_no_check;
351 if (INET_PROTOSW_REUSE & answer_flags)
352 sk->sk_reuse = 1;
353
354 inet = inet_sk(sk);
Paul Moore469de9b2007-01-09 14:37:06 -0800355 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356
357 if (SOCK_RAW == sock->type) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000358 inet->inet_num = protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 if (IPPROTO_RAW == protocol)
360 inet->hdrincl = 1;
361 }
362
363 if (ipv4_config.no_pmtu_disc)
364 inet->pmtudisc = IP_PMTUDISC_DONT;
365 else
366 inet->pmtudisc = IP_PMTUDISC_WANT;
367
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000368 inet->inet_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
370 sock_init_data(sock, sk);
371
372 sk->sk_destruct = inet_sock_destruct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 sk->sk_protocol = protocol;
374 sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
375
376 inet->uc_ttl = -1;
377 inet->mc_loop = 1;
378 inet->mc_ttl = 1;
Nivedita Singhvif771bef2009-05-28 07:00:46 +0000379 inet->mc_all = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 inet->mc_index = 0;
381 inet->mc_list = NULL;
382
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700383 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000385 if (inet->inet_num) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 /* It assumes that any protocol which allows
387 * the user to assign a number at socket
388 * creation time automatically
389 * shares.
390 */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000391 inet->inet_sport = htons(inet->inet_num);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 /* Add to protocol hash chains. */
393 sk->sk_prot->hash(sk);
394 }
395
396 if (sk->sk_prot->init) {
397 err = sk->sk_prot->init(sk);
398 if (err)
399 sk_common_release(sk);
400 }
401out:
402 return err;
403out_rcu_unlock:
404 rcu_read_unlock();
405 goto out;
406}
407
408
409/*
410 * The peer socket should always be NULL (or else). When we call this
411 * function we are destroying the object and from then on nobody
412 * should refer to it.
413 */
414int inet_release(struct socket *sock)
415{
416 struct sock *sk = sock->sk;
417
418 if (sk) {
419 long timeout;
420
421 /* Applications forget to leave groups before exiting */
422 ip_mc_drop_socket(sk);
423
424 /* If linger is set, we don't return until the close
425 * is complete. Otherwise we return immediately. The
426 * actually closing is done the same either way.
427 *
428 * If the close is due to the process exiting, we never
429 * linger..
430 */
431 timeout = 0;
432 if (sock_flag(sk, SOCK_LINGER) &&
433 !(current->flags & PF_EXITING))
434 timeout = sk->sk_lingertime;
435 sock->sk = NULL;
436 sk->sk_prot->close(sk, timeout);
437 }
438 return 0;
439}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700440EXPORT_SYMBOL(inet_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
442/* It is off by default, see below. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700443int sysctl_ip_nonlocal_bind __read_mostly;
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700444EXPORT_SYMBOL(sysctl_ip_nonlocal_bind);
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 &&
Tóth László Attilab9fb1502008-10-01 07:31:24 -0700475 !(inet->freebind || inet->transparent) &&
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;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000498 if (sk->sk_state != TCP_CLOSE || inet->inet_num)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 goto out_release_sock;
500
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000501 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000503 inet->inet_saddr = 0; /* Use device */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
505 /* Make sure we are allowed to bind here. */
506 if (sk->sk_prot->get_port(sk, snum)) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000507 inet->inet_saddr = inet->inet_rcv_saddr = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 err = -EADDRINUSE;
509 goto out_release_sock;
510 }
511
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000512 if (inet->inet_rcv_saddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
514 if (snum)
515 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000516 inet->inet_sport = htons(inet->inet_num);
517 inet->inet_daddr = 0;
518 inet->inet_dport = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 sk_dst_reset(sk);
520 err = 0;
521out_release_sock:
522 release_sock(sk);
523out:
524 return err;
525}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700526EXPORT_SYMBOL(inet_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527
528int inet_dgram_connect(struct socket *sock, struct sockaddr * uaddr,
529 int addr_len, int flags)
530{
531 struct sock *sk = sock->sk;
532
533 if (uaddr->sa_family == AF_UNSPEC)
534 return sk->sk_prot->disconnect(sk, flags);
535
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000536 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 return -EAGAIN;
538 return sk->sk_prot->connect(sk, (struct sockaddr *)uaddr, addr_len);
539}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700540EXPORT_SYMBOL(inet_dgram_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
542static long inet_wait_for_connect(struct sock *sk, long timeo)
543{
544 DEFINE_WAIT(wait);
545
546 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
547
548 /* Basic assumption: if someone sets sk->sk_err, he _must_
549 * change state of the socket from TCP_SYN_*.
550 * Connect() does not allow to get error notifications
551 * without closing the socket.
552 */
553 while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
554 release_sock(sk);
555 timeo = schedule_timeout(timeo);
556 lock_sock(sk);
557 if (signal_pending(current) || !timeo)
558 break;
559 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
560 }
561 finish_wait(sk->sk_sleep, &wait);
562 return timeo;
563}
564
565/*
566 * Connect to a remote host. There is regrettably still a little
567 * TCP 'magic' in here.
568 */
569int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
570 int addr_len, int flags)
571{
572 struct sock *sk = sock->sk;
573 int err;
574 long timeo;
575
576 lock_sock(sk);
577
578 if (uaddr->sa_family == AF_UNSPEC) {
579 err = sk->sk_prot->disconnect(sk, flags);
580 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
581 goto out;
582 }
583
584 switch (sock->state) {
585 default:
586 err = -EINVAL;
587 goto out;
588 case SS_CONNECTED:
589 err = -EISCONN;
590 goto out;
591 case SS_CONNECTING:
592 err = -EALREADY;
593 /* Fall out of switch with err, set for this state */
594 break;
595 case SS_UNCONNECTED:
596 err = -EISCONN;
597 if (sk->sk_state != TCP_CLOSE)
598 goto out;
599
600 err = sk->sk_prot->connect(sk, uaddr, addr_len);
601 if (err < 0)
602 goto out;
603
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900604 sock->state = SS_CONNECTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
606 /* Just entered SS_CONNECTING state; the only
607 * difference is that return value in non-blocking
608 * case is EINPROGRESS, rather than EALREADY.
609 */
610 err = -EINPROGRESS;
611 break;
612 }
613
614 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
615
616 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
617 /* Error code is set above */
618 if (!timeo || !inet_wait_for_connect(sk, timeo))
619 goto out;
620
621 err = sock_intr_errno(timeo);
622 if (signal_pending(current))
623 goto out;
624 }
625
626 /* Connection was closed by RST, timeout, ICMP error
627 * or another process disconnected us.
628 */
629 if (sk->sk_state == TCP_CLOSE)
630 goto sock_error;
631
632 /* sk->sk_err may be not zero now, if RECVERR was ordered by user
633 * and error was received after socket entered established state.
634 * Hence, it is handled normally after connect() return successfully.
635 */
636
637 sock->state = SS_CONNECTED;
638 err = 0;
639out:
640 release_sock(sk);
641 return err;
642
643sock_error:
644 err = sock_error(sk) ? : -ECONNABORTED;
645 sock->state = SS_UNCONNECTED;
646 if (sk->sk_prot->disconnect(sk, flags))
647 sock->state = SS_DISCONNECTING;
648 goto out;
649}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700650EXPORT_SYMBOL(inet_stream_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652/*
653 * Accept a pending connection. The TCP layer now gives BSD semantics.
654 */
655
656int inet_accept(struct socket *sock, struct socket *newsock, int flags)
657{
658 struct sock *sk1 = sock->sk;
659 int err = -EINVAL;
660 struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err);
661
662 if (!sk2)
663 goto do_err;
664
665 lock_sock(sk2);
666
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700667 WARN_ON(!((1 << sk2->sk_state) &
668 (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT | TCPF_CLOSE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670 sock_graft(sk2, newsock);
671
672 newsock->state = SS_CONNECTED;
673 err = 0;
674 release_sock(sk2);
675do_err:
676 return err;
677}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700678EXPORT_SYMBOL(inet_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
680
681/*
682 * This does both peername and sockname.
683 */
684int inet_getname(struct socket *sock, struct sockaddr *uaddr,
685 int *uaddr_len, int peer)
686{
687 struct sock *sk = sock->sk;
688 struct inet_sock *inet = inet_sk(sk);
Cyrill Gorcunov38bfd8f2009-10-29 02:59:18 -0700689 DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
691 sin->sin_family = AF_INET;
692 if (peer) {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000693 if (!inet->inet_dport ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
695 peer == 1))
696 return -ENOTCONN;
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000697 sin->sin_port = inet->inet_dport;
698 sin->sin_addr.s_addr = inet->inet_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 } else {
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000700 __be32 addr = inet->inet_rcv_saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 if (!addr)
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000702 addr = inet->inet_saddr;
703 sin->sin_port = inet->inet_sport;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 sin->sin_addr.s_addr = addr;
705 }
706 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
707 *uaddr_len = sizeof(*sin);
708 return 0;
709}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700710EXPORT_SYMBOL(inet_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
713 size_t size)
714{
715 struct sock *sk = sock->sk;
716
717 /* We may need to bind the socket. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000718 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 return -EAGAIN;
720
721 return sk->sk_prot->sendmsg(iocb, sk, msg, size);
722}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700723EXPORT_SYMBOL(inet_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
725
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700726static ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
727 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728{
729 struct sock *sk = sock->sk;
730
731 /* We may need to bind the socket. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +0000732 if (!inet_sk(sk)->inet_num && inet_autobind(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 return -EAGAIN;
734
735 if (sk->sk_prot->sendpage)
736 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
737 return sock_no_sendpage(sock, page, offset, size, flags);
738}
739
740
741int inet_shutdown(struct socket *sock, int how)
742{
743 struct sock *sk = sock->sk;
744 int err = 0;
745
746 /* This should really check to make sure
747 * the socket is a TCP socket. (WHY AC...)
748 */
749 how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
750 1->2 bit 2 snds.
751 2->3 */
752 if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */
753 return -EINVAL;
754
755 lock_sock(sk);
756 if (sock->state == SS_CONNECTING) {
757 if ((1 << sk->sk_state) &
758 (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
759 sock->state = SS_DISCONNECTING;
760 else
761 sock->state = SS_CONNECTED;
762 }
763
764 switch (sk->sk_state) {
765 case TCP_CLOSE:
766 err = -ENOTCONN;
767 /* Hack to wake up other listeners, who can poll for
768 POLLHUP, even on eg. unconnected UDP sockets -- RR */
769 default:
770 sk->sk_shutdown |= how;
771 if (sk->sk_prot->shutdown)
772 sk->sk_prot->shutdown(sk, how);
773 break;
774
775 /* Remaining two branches are temporary solution for missing
776 * close() in multithreaded environment. It is _not_ a good idea,
777 * but we have no choice until close() is repaired at VFS level.
778 */
779 case TCP_LISTEN:
780 if (!(how & RCV_SHUTDOWN))
781 break;
782 /* Fall through */
783 case TCP_SYN_SENT:
784 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
785 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
786 break;
787 }
788
789 /* Wake up anyone sleeping in poll. */
790 sk->sk_state_change(sk);
791 release_sock(sk);
792 return err;
793}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700794EXPORT_SYMBOL(inet_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
796/*
797 * ioctl() calls you can issue on an INET socket. Most of these are
798 * device configuration and stuff and very rarely used. Some ioctls
799 * pass on to the socket itself.
800 *
801 * NOTE: I like the idea of a module for the config stuff. ie ifconfig
802 * loads the devconfigure module does its configuring and unloads it.
803 * There's a good 20K of config code hanging around the kernel.
804 */
805
806int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
807{
808 struct sock *sk = sock->sk;
809 int err = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900810 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
812 switch (cmd) {
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700813 case SIOCGSTAMP:
814 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
815 break;
816 case SIOCGSTAMPNS:
817 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
818 break;
819 case SIOCADDRT:
820 case SIOCDELRT:
821 case SIOCRTMSG:
822 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
823 break;
824 case SIOCDARP:
825 case SIOCGARP:
826 case SIOCSARP:
827 err = arp_ioctl(net, cmd, (void __user *)arg);
828 break;
829 case SIOCGIFADDR:
830 case SIOCSIFADDR:
831 case SIOCGIFBRDADDR:
832 case SIOCSIFBRDADDR:
833 case SIOCGIFNETMASK:
834 case SIOCSIFNETMASK:
835 case SIOCGIFDSTADDR:
836 case SIOCSIFDSTADDR:
837 case SIOCSIFPFLAGS:
838 case SIOCGIFPFLAGS:
839 case SIOCSIFFLAGS:
840 err = devinet_ioctl(net, cmd, (void __user *)arg);
841 break;
842 default:
843 if (sk->sk_prot->ioctl)
844 err = sk->sk_prot->ioctl(sk, cmd, arg);
845 else
846 err = -ENOIOCTLCMD;
847 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 }
849 return err;
850}
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700851EXPORT_SYMBOL(inet_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800853const struct proto_ops inet_stream_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800854 .family = PF_INET,
855 .owner = THIS_MODULE,
856 .release = inet_release,
857 .bind = inet_bind,
858 .connect = inet_stream_connect,
859 .socketpair = sock_no_socketpair,
860 .accept = inet_accept,
861 .getname = inet_getname,
862 .poll = tcp_poll,
863 .ioctl = inet_ioctl,
864 .listen = inet_listen,
865 .shutdown = inet_shutdown,
866 .setsockopt = sock_common_setsockopt,
867 .getsockopt = sock_common_getsockopt,
David S. Miller3516ffb2007-08-02 19:23:56 -0700868 .sendmsg = tcp_sendmsg,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800869 .recvmsg = sock_common_recvmsg,
870 .mmap = sock_no_mmap,
871 .sendpage = tcp_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800872 .splice_read = tcp_splice_read,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800873#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800874 .compat_setsockopt = compat_sock_common_setsockopt,
875 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800876#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700878EXPORT_SYMBOL(inet_stream_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800880const struct proto_ops inet_dgram_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800881 .family = PF_INET,
882 .owner = THIS_MODULE,
883 .release = inet_release,
884 .bind = inet_bind,
885 .connect = inet_dgram_connect,
886 .socketpair = sock_no_socketpair,
887 .accept = sock_no_accept,
888 .getname = inet_getname,
889 .poll = udp_poll,
890 .ioctl = inet_ioctl,
891 .listen = sock_no_listen,
892 .shutdown = inet_shutdown,
893 .setsockopt = sock_common_setsockopt,
894 .getsockopt = sock_common_getsockopt,
895 .sendmsg = inet_sendmsg,
896 .recvmsg = sock_common_recvmsg,
897 .mmap = sock_no_mmap,
898 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800899#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800900 .compat_setsockopt = compat_sock_common_setsockopt,
901 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800902#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903};
Eric Dumazet3d1427f2009-08-28 23:45:21 -0700904EXPORT_SYMBOL(inet_dgram_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
906/*
907 * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
908 * udp_poll
909 */
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800910static const struct proto_ops inet_sockraw_ops = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800911 .family = PF_INET,
912 .owner = THIS_MODULE,
913 .release = inet_release,
914 .bind = inet_bind,
915 .connect = inet_dgram_connect,
916 .socketpair = sock_no_socketpair,
917 .accept = sock_no_accept,
918 .getname = inet_getname,
919 .poll = datagram_poll,
920 .ioctl = inet_ioctl,
921 .listen = sock_no_listen,
922 .shutdown = inet_shutdown,
923 .setsockopt = sock_common_setsockopt,
924 .getsockopt = sock_common_getsockopt,
925 .sendmsg = inet_sendmsg,
926 .recvmsg = sock_common_recvmsg,
927 .mmap = sock_no_mmap,
928 .sendpage = inet_sendpage,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800929#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800930 .compat_setsockopt = compat_sock_common_setsockopt,
931 .compat_getsockopt = compat_sock_common_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800932#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933};
934
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +0000935static const struct net_proto_family inet_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 .family = PF_INET,
937 .create = inet_create,
938 .owner = THIS_MODULE,
939};
940
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941/* Upon startup we insert all the elements in inetsw_array[] into
942 * the linked list inetsw.
943 */
944static struct inet_protosw inetsw_array[] =
945{
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900946 {
947 .type = SOCK_STREAM,
948 .protocol = IPPROTO_TCP,
949 .prot = &tcp_prot,
950 .ops = &inet_stream_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900951 .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,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900961 .no_check = UDP_CSUM_DEFAULT,
962 .flags = INET_PROTOSW_PERMANENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 },
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
966 {
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900967 .type = SOCK_RAW,
968 .protocol = IPPROTO_IP, /* wild card */
969 .prot = &raw_prot,
970 .ops = &inet_sockraw_ops,
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900971 .no_check = UDP_CSUM_DEFAULT,
972 .flags = INET_PROTOSW_REUSE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 }
974};
975
Denis Chengc40f6ff2007-09-16 16:39:25 -0700976#define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
978void inet_register_protosw(struct inet_protosw *p)
979{
980 struct list_head *lh;
981 struct inet_protosw *answer;
982 int protocol = p->protocol;
983 struct list_head *last_perm;
984
985 spin_lock_bh(&inetsw_lock);
986
987 if (p->type >= SOCK_MAX)
988 goto out_illegal;
989
990 /* If we are trying to override a permanent protocol, bail. */
991 answer = NULL;
992 last_perm = &inetsw[p->type];
993 list_for_each(lh, &inetsw[p->type]) {
994 answer = list_entry(lh, struct inet_protosw, list);
995
996 /* Check only the non-wild match. */
997 if (INET_PROTOSW_PERMANENT & answer->flags) {
998 if (protocol == answer->protocol)
999 break;
1000 last_perm = lh;
1001 }
1002
1003 answer = NULL;
1004 }
1005 if (answer)
1006 goto out_permanent;
1007
1008 /* Add the new entry after the last permanent entry if any, so that
1009 * the new entry does not override a permanent entry when matched with
1010 * a wild-card protocol. But it is allowed to override any existing
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001011 * non-permanent entry. This means that when we remove this entry, the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 * system automatically returns to the old behavior.
1013 */
1014 list_add_rcu(&p->list, last_perm);
1015out:
1016 spin_unlock_bh(&inetsw_lock);
1017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 return;
1019
1020out_permanent:
1021 printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
1022 protocol);
1023 goto out;
1024
1025out_illegal:
1026 printk(KERN_ERR
1027 "Ignoring attempt to register invalid socket type %d.\n",
1028 p->type);
1029 goto out;
1030}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001031EXPORT_SYMBOL(inet_register_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
1033void inet_unregister_protosw(struct inet_protosw *p)
1034{
1035 if (INET_PROTOSW_PERMANENT & p->flags) {
1036 printk(KERN_ERR
1037 "Attempt to unregister permanent protocol %d.\n",
1038 p->protocol);
1039 } else {
1040 spin_lock_bh(&inetsw_lock);
1041 list_del_rcu(&p->list);
1042 spin_unlock_bh(&inetsw_lock);
1043
1044 synchronize_net();
1045 }
1046}
Eric Dumazet3d1427f2009-08-28 23:45:21 -07001047EXPORT_SYMBOL(inet_unregister_protosw);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001049/*
1050 * Shall we try to damage output packets if routing dev changes?
1051 */
1052
Brian Haleyab32ea52006-09-22 14:15:41 -07001053int sysctl_ip_dynaddr __read_mostly;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001054
1055static int inet_sk_reselect_saddr(struct sock *sk)
1056{
1057 struct inet_sock *inet = inet_sk(sk);
1058 int err;
1059 struct rtable *rt;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001060 __be32 old_saddr = inet->inet_saddr;
Al Viro714e85b2006-11-14 20:51:49 -08001061 __be32 new_saddr;
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001062 __be32 daddr = inet->inet_daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001063
1064 if (inet->opt && inet->opt->srr)
1065 daddr = inet->opt->faddr;
1066
1067 /* Query new route. */
1068 err = ip_route_connect(&rt, daddr, 0,
1069 RT_CONN_FLAGS(sk),
1070 sk->sk_bound_dev_if,
1071 sk->sk_protocol,
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001072 inet->inet_sport, inet->inet_dport, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001073 if (err)
1074 return err;
1075
1076 sk_setup_caps(sk, &rt->u.dst);
1077
1078 new_saddr = rt->rt_src;
1079
1080 if (new_saddr == old_saddr)
1081 return 0;
1082
1083 if (sysctl_ip_dynaddr > 1) {
Harvey Harrison673d57e2008-10-31 00:53:57 -07001084 printk(KERN_INFO "%s(): shifting inet->saddr from %pI4 to %pI4\n",
1085 __func__, &old_saddr, &new_saddr);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001086 }
1087
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001088 inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001089
1090 /*
1091 * XXX The only one ugly spot where we need to
1092 * XXX really change the sockets identity after
1093 * XXX it has entered the hashes. -DaveM
1094 *
1095 * Besides that, it does not check for connection
1096 * uniqueness. Wait for troubles.
1097 */
1098 __sk_prot_rehash(sk);
1099 return 0;
1100}
1101
1102int inet_sk_rebuild_header(struct sock *sk)
1103{
1104 struct inet_sock *inet = inet_sk(sk);
1105 struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
Al Viro3ca3c682006-09-27 18:28:07 -07001106 __be32 daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001107 int err;
1108
1109 /* Route is OK, nothing to do. */
1110 if (rt)
1111 return 0;
1112
1113 /* Reroute. */
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001114 daddr = inet->inet_daddr;
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001115 if (inet->opt && inet->opt->srr)
1116 daddr = inet->opt->faddr;
1117{
1118 struct flowi fl = {
1119 .oif = sk->sk_bound_dev_if,
Atis Elsts914a9ab2009-10-01 15:16:49 -07001120 .mark = sk->sk_mark,
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001121 .nl_u = {
1122 .ip4_u = {
1123 .daddr = daddr,
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001124 .saddr = inet->inet_saddr,
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001125 .tos = RT_CONN_FLAGS(sk),
1126 },
1127 },
1128 .proto = sk->sk_protocol,
Balazs Scheidlerc8283842008-11-20 01:08:06 -08001129 .flags = inet_sk_flowi_flags(sk),
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001130 .uli_u = {
1131 .ports = {
Eric Dumazetc720c7e2009-10-15 06:30:45 +00001132 .sport = inet->inet_sport,
1133 .dport = inet->inet_dport,
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001134 },
1135 },
1136 };
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001137
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -07001138 security_sk_classify_flow(sk, &fl);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001139 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 0);
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001140}
1141 if (!err)
1142 sk_setup_caps(sk, &rt->u.dst);
1143 else {
1144 /* Routing failed... */
1145 sk->sk_route_caps = 0;
1146 /*
1147 * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1148 * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1149 */
1150 if (!sysctl_ip_dynaddr ||
1151 sk->sk_state != TCP_SYN_SENT ||
1152 (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1153 (err = inet_sk_reselect_saddr(sk)) != 0)
1154 sk->sk_err_soft = -err;
1155 }
1156
1157 return err;
1158}
Arnaldo Carvalho de Melo32519f12005-08-09 19:50:02 -07001159EXPORT_SYMBOL(inet_sk_rebuild_header);
1160
Herbert Xua430a432006-07-08 13:34:56 -07001161static int inet_gso_send_check(struct sk_buff *skb)
1162{
1163 struct iphdr *iph;
Alexey Dobriyan32613092009-09-14 12:21:47 +00001164 const struct net_protocol *ops;
Herbert Xua430a432006-07-08 13:34:56 -07001165 int proto;
1166 int ihl;
1167 int err = -EINVAL;
1168
1169 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1170 goto out;
1171
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001172 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001173 ihl = iph->ihl * 4;
1174 if (ihl < sizeof(*iph))
1175 goto out;
1176
1177 if (unlikely(!pskb_may_pull(skb, ihl)))
1178 goto out;
1179
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001180 __skb_pull(skb, ihl);
1181 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001182 iph = ip_hdr(skb);
Herbert Xua430a432006-07-08 13:34:56 -07001183 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1184 err = -EPROTONOSUPPORT;
1185
1186 rcu_read_lock();
1187 ops = rcu_dereference(inet_protos[proto]);
1188 if (likely(ops && ops->gso_send_check))
1189 err = ops->gso_send_check(skb);
1190 rcu_read_unlock();
1191
1192out:
1193 return err;
1194}
1195
Herbert Xu576a30e2006-06-27 13:22:38 -07001196static struct sk_buff *inet_gso_segment(struct sk_buff *skb, int features)
Herbert Xuf4c50d92006-06-22 03:02:40 -07001197{
1198 struct sk_buff *segs = ERR_PTR(-EINVAL);
1199 struct iphdr *iph;
Alexey Dobriyan32613092009-09-14 12:21:47 +00001200 const struct net_protocol *ops;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001201 int proto;
1202 int ihl;
1203 int id;
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001204 unsigned int offset = 0;
Herbert Xuf4c50d92006-06-22 03:02:40 -07001205
Stephen Hemmingerd212f872007-06-27 00:47:37 -07001206 if (!(features & NETIF_F_V4_CSUM))
1207 features &= ~NETIF_F_SG;
1208
Herbert Xubbcf4672006-07-03 19:38:35 -07001209 if (unlikely(skb_shinfo(skb)->gso_type &
1210 ~(SKB_GSO_TCPV4 |
1211 SKB_GSO_UDP |
1212 SKB_GSO_DODGY |
1213 SKB_GSO_TCP_ECN |
1214 0)))
1215 goto out;
1216
1217 if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001218 goto out;
1219
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001220 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001221 ihl = iph->ihl * 4;
1222 if (ihl < sizeof(*iph))
1223 goto out;
1224
Herbert Xubbcf4672006-07-03 19:38:35 -07001225 if (unlikely(!pskb_may_pull(skb, ihl)))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001226 goto out;
1227
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001228 __skb_pull(skb, ihl);
1229 skb_reset_transport_header(skb);
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001230 iph = ip_hdr(skb);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001231 id = ntohs(iph->id);
1232 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1233 segs = ERR_PTR(-EPROTONOSUPPORT);
1234
1235 rcu_read_lock();
1236 ops = rcu_dereference(inet_protos[proto]);
Herbert Xubbcf4672006-07-03 19:38:35 -07001237 if (likely(ops && ops->gso_segment))
Herbert Xu576a30e2006-06-27 13:22:38 -07001238 segs = ops->gso_segment(skb, features);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001239 rcu_read_unlock();
1240
Hirofumi Nakagawa801678c2008-04-29 01:03:09 -07001241 if (!segs || IS_ERR(segs))
Herbert Xuf4c50d92006-06-22 03:02:40 -07001242 goto out;
1243
1244 skb = segs;
1245 do {
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -07001246 iph = ip_hdr(skb);
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001247 if (proto == IPPROTO_UDP) {
1248 iph->id = htons(id);
1249 iph->frag_off = htons(offset >> 3);
1250 if (skb->next != NULL)
1251 iph->frag_off |= htons(IP_MF);
1252 offset += (skb->len - skb->mac_len - iph->ihl * 4);
1253 } else
1254 iph->id = htons(id++);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001255 iph->tot_len = htons(skb->len - skb->mac_len);
1256 iph->check = 0;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001257 iph->check = ip_fast_csum(skb_network_header(skb), iph->ihl);
Herbert Xuf4c50d92006-06-22 03:02:40 -07001258 } while ((skb = skb->next));
1259
1260out:
1261 return segs;
1262}
1263
Herbert Xu73cc19f2008-12-15 23:41:09 -08001264static struct sk_buff **inet_gro_receive(struct sk_buff **head,
1265 struct sk_buff *skb)
1266{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001267 const struct net_protocol *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001268 struct sk_buff **pp = NULL;
1269 struct sk_buff *p;
1270 struct iphdr *iph;
Herbert Xua5b1cf22009-05-26 18:50:28 +00001271 unsigned int hlen;
1272 unsigned int off;
Herbert Xu1075f3f2009-05-26 18:50:29 +00001273 unsigned int id;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001274 int flush = 1;
1275 int proto;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001276
Herbert Xua5b1cf22009-05-26 18:50:28 +00001277 off = skb_gro_offset(skb);
1278 hlen = off + sizeof(*iph);
1279 iph = skb_gro_header_fast(skb, off);
1280 if (skb_gro_header_hard(skb, hlen)) {
1281 iph = skb_gro_header_slow(skb, hlen, off);
1282 if (unlikely(!iph))
1283 goto out;
1284 }
Herbert Xu73cc19f2008-12-15 23:41:09 -08001285
Herbert Xu73cc19f2008-12-15 23:41:09 -08001286 proto = iph->protocol & (MAX_INET_PROTOS - 1);
1287
1288 rcu_read_lock();
1289 ops = rcu_dereference(inet_protos[proto]);
1290 if (!ops || !ops->gro_receive)
1291 goto out_unlock;
1292
Herbert Xua5ad24b2009-02-08 18:00:39 +00001293 if (*(u8 *)iph != 0x45)
Herbert Xu73cc19f2008-12-15 23:41:09 -08001294 goto out_unlock;
1295
1296 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
1297 goto out_unlock;
1298
Herbert Xu1075f3f2009-05-26 18:50:29 +00001299 id = ntohl(*(u32 *)&iph->id);
1300 flush = (u16)((ntohl(*(u32 *)iph) ^ skb_gro_len(skb)) | (id ^ IP_DF));
1301 id >>= 16;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001302
1303 for (p = *head; p; p = p->next) {
1304 struct iphdr *iph2;
1305
1306 if (!NAPI_GRO_CB(p)->same_flow)
1307 continue;
1308
1309 iph2 = ip_hdr(p);
1310
Herbert Xua5ad24b2009-02-08 18:00:39 +00001311 if ((iph->protocol ^ iph2->protocol) |
1312 (iph->tos ^ iph2->tos) |
1313 (iph->saddr ^ iph2->saddr) |
1314 (iph->daddr ^ iph2->daddr)) {
Herbert Xu73cc19f2008-12-15 23:41:09 -08001315 NAPI_GRO_CB(p)->same_flow = 0;
1316 continue;
1317 }
1318
1319 /* All fields must match except length and checksum. */
1320 NAPI_GRO_CB(p)->flush |=
Herbert Xua5ad24b2009-02-08 18:00:39 +00001321 (iph->ttl ^ iph2->ttl) |
1322 ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id);
Herbert Xu73cc19f2008-12-15 23:41:09 -08001323
1324 NAPI_GRO_CB(p)->flush |= flush;
1325 }
1326
1327 NAPI_GRO_CB(skb)->flush |= flush;
Herbert Xu86911732009-01-29 14:19:50 +00001328 skb_gro_pull(skb, sizeof(*iph));
1329 skb_set_transport_header(skb, skb_gro_offset(skb));
Herbert Xu73cc19f2008-12-15 23:41:09 -08001330
1331 pp = ops->gro_receive(head, skb);
1332
1333out_unlock:
1334 rcu_read_unlock();
1335
1336out:
1337 NAPI_GRO_CB(skb)->flush |= flush;
1338
1339 return pp;
1340}
1341
1342static int inet_gro_complete(struct sk_buff *skb)
1343{
Alexey Dobriyan32613092009-09-14 12:21:47 +00001344 const struct net_protocol *ops;
Herbert Xu73cc19f2008-12-15 23:41:09 -08001345 struct iphdr *iph = ip_hdr(skb);
1346 int proto = iph->protocol & (MAX_INET_PROTOS - 1);
1347 int err = -ENOSYS;
1348 __be16 newlen = htons(skb->len - skb_network_offset(skb));
1349
1350 csum_replace2(&iph->check, iph->tot_len, newlen);
1351 iph->tot_len = newlen;
1352
1353 rcu_read_lock();
1354 ops = rcu_dereference(inet_protos[proto]);
1355 if (WARN_ON(!ops || !ops->gro_complete))
1356 goto out_unlock;
1357
1358 err = ops->gro_complete(skb);
1359
1360out_unlock:
1361 rcu_read_unlock();
1362
1363 return err;
1364}
1365
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001366int inet_ctl_sock_create(struct sock **sk, unsigned short family,
Denis V. Lunev56772422008-04-03 14:28:30 -07001367 unsigned short type, unsigned char protocol,
1368 struct net *net)
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001369{
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001370 struct socket *sock;
1371 int rc = sock_create_kern(family, type, protocol, &sock);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001372
1373 if (rc == 0) {
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001374 *sk = sock->sk;
1375 (*sk)->sk_allocation = GFP_ATOMIC;
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001376 /*
1377 * Unhash it so that IP input processing does not even see it,
1378 * we do not wish this socket to see incoming packets.
1379 */
Denis V. Luneveee4fe42008-04-03 14:27:58 -07001380 (*sk)->sk_prot->unhash(*sk);
Denis V. Lunev56772422008-04-03 14:28:30 -07001381
1382 sk_change_net(*sk, net);
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001383 }
1384 return rc;
1385}
Denis V. Lunev3d58b5f2008-04-03 14:22:32 -07001386EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1387
Herbert Xu5e0f0432007-04-24 21:53:35 -07001388unsigned long snmp_fold_field(void *mib[], int offt)
1389{
1390 unsigned long res = 0;
1391 int i;
1392
1393 for_each_possible_cpu(i) {
1394 res += *(((unsigned long *) per_cpu_ptr(mib[0], i)) + offt);
1395 res += *(((unsigned long *) per_cpu_ptr(mib[1], i)) + offt);
1396 }
1397 return res;
1398}
1399EXPORT_SYMBOL_GPL(snmp_fold_field);
1400
YOSHIFUJI Hideakic69bce22008-01-23 22:31:45 -08001401int snmp_mib_init(void *ptr[2], size_t mibsize)
Herbert Xu5e0f0432007-04-24 21:53:35 -07001402{
1403 BUG_ON(ptr == NULL);
Rusty Russell313e4582009-02-20 16:29:08 +09001404 ptr[0] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
Herbert Xu5e0f0432007-04-24 21:53:35 -07001405 if (!ptr[0])
1406 goto err0;
Rusty Russell313e4582009-02-20 16:29:08 +09001407 ptr[1] = __alloc_percpu(mibsize, __alignof__(unsigned long long));
Herbert Xu5e0f0432007-04-24 21:53:35 -07001408 if (!ptr[1])
1409 goto err1;
1410 return 0;
1411err1:
1412 free_percpu(ptr[0]);
1413 ptr[0] = NULL;
1414err0:
1415 return -ENOMEM;
1416}
1417EXPORT_SYMBOL_GPL(snmp_mib_init);
1418
1419void snmp_mib_free(void *ptr[2])
1420{
1421 BUG_ON(ptr == NULL);
1422 free_percpu(ptr[0]);
1423 free_percpu(ptr[1]);
1424 ptr[0] = ptr[1] = NULL;
1425}
1426EXPORT_SYMBOL_GPL(snmp_mib_free);
1427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428#ifdef CONFIG_IP_MULTICAST
Alexey Dobriyan32613092009-09-14 12:21:47 +00001429static const struct net_protocol igmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 .handler = igmp_rcv,
Alexey Dobriyanb4ee07d2008-12-25 16:42:23 -08001431 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432};
1433#endif
1434
Alexey Dobriyan32613092009-09-14 12:21:47 +00001435static const struct net_protocol tcp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 .handler = tcp_v4_rcv,
1437 .err_handler = tcp_v4_err,
Herbert Xua430a432006-07-08 13:34:56 -07001438 .gso_send_check = tcp_v4_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001439 .gso_segment = tcp_tso_segment,
Herbert Xubf296b12008-12-15 23:43:36 -08001440 .gro_receive = tcp4_gro_receive,
1441 .gro_complete = tcp4_gro_complete,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001443 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444};
1445
Alexey Dobriyan32613092009-09-14 12:21:47 +00001446static const struct net_protocol udp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 .handler = udp_rcv,
1448 .err_handler = udp_err,
Sridhar Samudralad7ca4cc2009-07-09 08:09:47 +00001449 .gso_send_check = udp4_ufo_send_check,
1450 .gso_segment = udp4_ufo_fragment,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001452 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453};
1454
Alexey Dobriyan32613092009-09-14 12:21:47 +00001455static const struct net_protocol icmp_protocol = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 .handler = icmp_rcv,
Herbert Xu8b7817f2007-12-12 10:44:43 -08001457 .no_policy = 1,
Denis V. Lunev92f1fec2008-03-24 15:34:06 -07001458 .netns_ok = 1,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459};
1460
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001461static __net_init int ipv4_mib_init_net(struct net *net)
1462{
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001463 if (snmp_mib_init((void **)net->mib.tcp_statistics,
1464 sizeof(struct tcp_mib)) < 0)
1465 goto err_tcp_mib;
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001466 if (snmp_mib_init((void **)net->mib.ip_statistics,
1467 sizeof(struct ipstats_mib)) < 0)
1468 goto err_ip_mib;
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001469 if (snmp_mib_init((void **)net->mib.net_statistics,
1470 sizeof(struct linux_mib)) < 0)
1471 goto err_net_mib;
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001472 if (snmp_mib_init((void **)net->mib.udp_statistics,
1473 sizeof(struct udp_mib)) < 0)
1474 goto err_udp_mib;
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001475 if (snmp_mib_init((void **)net->mib.udplite_statistics,
1476 sizeof(struct udp_mib)) < 0)
1477 goto err_udplite_mib;
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001478 if (snmp_mib_init((void **)net->mib.icmp_statistics,
1479 sizeof(struct icmp_mib)) < 0)
1480 goto err_icmp_mib;
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001481 if (snmp_mib_init((void **)net->mib.icmpmsg_statistics,
1482 sizeof(struct icmpmsg_mib)) < 0)
1483 goto err_icmpmsg_mib;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001484
1485 tcp_mib_init(net);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001486 return 0;
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001487
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001488err_icmpmsg_mib:
1489 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001490err_icmp_mib:
1491 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001492err_udplite_mib:
1493 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001494err_udp_mib:
1495 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001496err_net_mib:
1497 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001498err_ip_mib:
1499 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001500err_tcp_mib:
1501 return -ENOMEM;
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001502}
1503
1504static __net_exit void ipv4_mib_exit_net(struct net *net)
1505{
Pavel Emelyanov923c6582008-07-18 04:04:22 -07001506 snmp_mib_free((void **)net->mib.icmpmsg_statistics);
Pavel Emelyanovb60538a2008-07-18 04:04:02 -07001507 snmp_mib_free((void **)net->mib.icmp_statistics);
Pavel Emelyanov386019d2008-07-18 04:03:45 -07001508 snmp_mib_free((void **)net->mib.udplite_statistics);
Pavel Emelyanov2f275f92008-07-18 04:03:27 -07001509 snmp_mib_free((void **)net->mib.udp_statistics);
Pavel Emelyanov61a7e262008-07-18 04:03:08 -07001510 snmp_mib_free((void **)net->mib.net_statistics);
Pavel Emelyanova20f5792008-07-18 04:02:42 -07001511 snmp_mib_free((void **)net->mib.ip_statistics);
Pavel Emelyanov57ef42d2008-07-18 04:02:08 -07001512 snmp_mib_free((void **)net->mib.tcp_statistics);
Pavel Emelyanov9b4661b2008-07-18 04:01:44 -07001513}
1514
1515static __net_initdata struct pernet_operations ipv4_mib_ops = {
1516 .init = ipv4_mib_init_net,
1517 .exit = ipv4_mib_exit_net,
1518};
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520static int __init init_ipv4_mibs(void)
1521{
Pavel Emelyanovd89cbbb2008-07-18 04:04:51 -07001522 return register_pernet_subsys(&ipv4_mib_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523}
1524
1525static int ipv4_proc_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
Herbert Xu30e224d2005-07-05 14:40:10 -07001527/*
1528 * IP protocol layer initialiser
1529 */
1530
Stephen Hemminger7546dd92009-03-09 08:18:29 +00001531static struct packet_type ip_packet_type __read_mostly = {
Harvey Harrison09640e62009-02-01 00:45:17 -08001532 .type = cpu_to_be16(ETH_P_IP),
Herbert Xu30e224d2005-07-05 14:40:10 -07001533 .func = ip_rcv,
Herbert Xua430a432006-07-08 13:34:56 -07001534 .gso_send_check = inet_gso_send_check,
Herbert Xuf4c50d92006-06-22 03:02:40 -07001535 .gso_segment = inet_gso_segment,
Herbert Xu73cc19f2008-12-15 23:41:09 -08001536 .gro_receive = inet_gro_receive,
1537 .gro_complete = inet_gro_complete,
Herbert Xu30e224d2005-07-05 14:40:10 -07001538};
1539
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540static int __init inet_init(void)
1541{
1542 struct sk_buff *dummy_skb;
1543 struct inet_protosw *q;
1544 struct list_head *r;
1545 int rc = -EINVAL;
1546
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001547 BUILD_BUG_ON(sizeof(struct inet_skb_parm) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
1549 rc = proto_register(&tcp_prot, 1);
1550 if (rc)
1551 goto out;
1552
1553 rc = proto_register(&udp_prot, 1);
1554 if (rc)
1555 goto out_unregister_tcp_proto;
1556
1557 rc = proto_register(&raw_prot, 1);
1558 if (rc)
1559 goto out_unregister_udp_proto;
1560
1561 /*
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001562 * Tell SOCKET that we are alive...
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 */
1564
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001565 (void)sock_register(&inet_family_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566
Al Virobd7b1532008-07-15 16:00:59 -04001567#ifdef CONFIG_SYSCTL
1568 ip_static_sysctl_init();
1569#endif
1570
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 /*
1572 * Add all the base protocols.
1573 */
1574
1575 if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1576 printk(KERN_CRIT "inet_init: Cannot add ICMP protocol\n");
1577 if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1578 printk(KERN_CRIT "inet_init: Cannot add UDP protocol\n");
1579 if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1580 printk(KERN_CRIT "inet_init: Cannot add TCP protocol\n");
1581#ifdef CONFIG_IP_MULTICAST
1582 if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1583 printk(KERN_CRIT "inet_init: Cannot add IGMP protocol\n");
1584#endif
1585
1586 /* Register the socket-side information for inet_create. */
1587 for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1588 INIT_LIST_HEAD(r);
1589
1590 for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1591 inet_register_protosw(q);
1592
1593 /*
1594 * Set the ARP module up
1595 */
1596
1597 arp_init();
1598
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001599 /*
1600 * Set the IP module up
1601 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602
1603 ip_init();
1604
Denis V. Lunev9b0f9762008-02-29 11:13:15 -08001605 tcp_v4_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607 /* Setup TCP slab cache for open requests. */
1608 tcp_init();
1609
Hideo Aoki95766ff2007-12-31 00:29:24 -08001610 /* Setup UDP memory threshold */
1611 udp_init();
1612
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001613 /* Add UDP-Lite (RFC 3828) */
1614 udplite4_register();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615
1616 /*
1617 * Set the ICMP layer up
1618 */
1619
Denis V. Luneva5710d62008-02-29 11:14:50 -08001620 if (icmp_init() < 0)
1621 panic("Failed to create the ICMP control socket.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623 /*
1624 * Initialise the multicast router
1625 */
1626#if defined(CONFIG_IP_MROUTE)
Wang Chen03d2f892008-07-03 12:13:36 +08001627 if (ip_mr_init())
1628 printk(KERN_CRIT "inet_init: Cannot init ipv4 mroute\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629#endif
1630 /*
1631 * Initialise per-cpu ipv4 mibs
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001632 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633
Stephen Hemminger132adf52007-03-08 20:44:43 -08001634 if (init_ipv4_mibs())
Wang Chen03d2f892008-07-03 12:13:36 +08001635 printk(KERN_CRIT "inet_init: Cannot init ipv4 mibs\n");
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +09001636
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 ipv4_proc_init();
1638
1639 ipfrag_init();
1640
Herbert Xu30e224d2005-07-05 14:40:10 -07001641 dev_add_pack(&ip_packet_type);
1642
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 rc = 0;
1644out:
1645 return rc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646out_unregister_udp_proto:
1647 proto_unregister(&udp_prot);
Olaf Kirch321efff2006-09-27 16:33:45 -07001648out_unregister_tcp_proto:
1649 proto_unregister(&tcp_prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 goto out;
1651}
1652
Heiko Carstensa536e072006-04-28 15:19:17 -07001653fs_initcall(inet_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654
1655/* ------------------------------------------------------------------------ */
1656
1657#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658static int __init ipv4_proc_init(void)
1659{
1660 int rc = 0;
1661
1662 if (raw_proc_init())
1663 goto out_raw;
1664 if (tcp4_proc_init())
1665 goto out_tcp;
1666 if (udp4_proc_init())
1667 goto out_udp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 if (ip_misc_proc_init())
1669 goto out_misc;
1670out:
1671 return rc;
1672out_misc:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 udp4_proc_exit();
1674out_udp:
1675 tcp4_proc_exit();
1676out_tcp:
1677 raw_proc_exit();
1678out_raw:
1679 rc = -ENOMEM;
1680 goto out;
1681}
1682
1683#else /* CONFIG_PROC_FS */
1684static int __init ipv4_proc_init(void)
1685{
1686 return 0;
1687}
1688#endif /* CONFIG_PROC_FS */
1689
1690MODULE_ALIAS_NETPROTO(PF_INET);
1691